0% found this document useful (0 votes)

111 views

How Dictatorships Work

The document introduces the topic of studying dictatorships. It notes that while democracies have been extensively studied, dictatorships are less understood due to opaque decision-making and heterogeneity among regimes. The author aims to develop realistic theories of authoritarian rule by analyzing characteristics of groups that seize power and influence policymaking and citizen welfare after taking control. Traits of the "seizure group" shape governance and outcomes within a dictatorship.

Uploaded by

alan

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

111 views

How Dictatorships Work

Uploaded by

alan

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 228

1

Introduction

Since humans began to live in settled communities, most have lived under
autocracy.1 Dictatorships still rule roughly 40 percent of the world’s nations.
All international wars since the end of World War I have involved dictator-
ships. Two-thirds of civil wars and ethnic conflicts since World War II have
erupted in countries under authoritarian rule.2 Since the fall of the Berlin Wall,
dictatorships have perpetrated nearly 85 percent of mass killings by govern-
ments.3 In other words, dictatorships affect millions of people’s lives (and
deaths) and initiate most of the urgent international challenges that policy
makers face. We cannot avoid dealing with them. And yet a limited understand-
ing of how dictatorships work undermines our ability to influence and negotiate
with them.
Most academic analyses of how governments work have focused on democ-
racies. We therefore know much less about dictatorial decision-making than
about democratic. As a further complication, dictatorships differ not only from
democracies but also from each other, and these differences have consequences
for citizen welfare and international conflict. Although some dictatorships
initiate more than their share of wars and political violence, many other

1
The absence of fair, reasonably competitive elections through which citizens choose those who
make policies on their behalf defines autocracy or dictatorship. The coding rules that operational-
ize this definition of dictatorship can be found at http://sites.psu.edu/dictators.
2
These figures were calculated using data on regime type from Geddes, Wright, and Frantz (2014),
data on civil wars from Themner and Wallensteen (2014), data on ethnic conflict from Wimmer,
Cederman, and Min (2009), and data on mass killings from Ulfelder and Valentino (2008). Civil
war statistics are calculated from all civil war years, including internationalized civil conflicts.
Ethnic war and mass-killing statistics are calculated from the onset years of conflicts. The years
included are 1946–2010.
3
Calculated from data on one-sided mass killings from 1989 to 2010 from Eck and Hultman
(2007). The figure excludes the genocide in 1994 Rwanda, which is hard to classify.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:33:41, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.001
2 Introduction

dictatorships live in peace with their neighbors and refrain from oppressing
citizens. The fastest-growing countries in the world are dictatorships, but the
most economically mismanaged are dictatorships as well. Some dictatorships
have followed policies to equalize incomes, but others have raised inequality to
astonishing levels. Abstract theories that treat all dictatorships as the same
cannot make sense of these differences. We need more realistic theories.
A great deal has been written about specific autocracies by individuals with
impressive local expertise, but only a few comparative studies grounded in
evidence exist. We also have some interesting theories of dictatorship, but fewer
theories firmly anchored in the real world. We know little about why some
dictatorships establish stable government while others suffer continuous
upheaval, why some create democratic-looking political institutions to engage
citizens and others do not, why some distribute benefits broadly while others
concentrate wealth within a small group of regime supporters, or why some last
many decades but many collapse within a year or two. In short, we understand
little about how dictatorships work and why they sometimes fail to work.
Some of the reasons why analysts have made less progress in the study of
dictatorship than democracy are obvious. Dictatorial decision-making often
occurs in secret, while policy-making and leadership choice in democracies
are relatively transparent (Lewis 1978, 622). Decision-making opacity inter-
feres with understanding why dictatorships do things. Small dictatorial elite
groups usually make decisions in informal settings: “formal institutions are not
necessarily the place to look when you want to understand everyday operating
procedures” in dictatorships (Fitzpatrick 2015, 278). Legislative debates and
votes often ratify policy choices made elsewhere, and cabinet ministers may be
the implementers of decisions, not the decision makers. Democracies publish
great quantities of data about themselves, making it easier for scholars to
investigate them. Not only do dictatorships publish less, but what they do
publish may be purposely inaccurate (Magee and Doces 2014). Election results
may reflect the resource advantage enjoyed by the ruling party rather than
voters’ preferences about who should rule, and published results may not match
votes cast.
For all these reasons, in order to make progress in understanding dictator-
ships, we need much more systematic information about them than has been
available. More challenging, the information needs to reflect informal aspects
of real dictatorial decision-making, not just the formal characteristics of rule
included in many existing data sets. More detailed information has begun to be
collected in recent years, facilitating the development of new theories about
dictatorship (Goemans, Gleditsch, and Chiozza 2009; Cheibub, Gandhi, and
Vreeland 2010; Svolik 2012). In this book we use additional newly collected
data to build on these efforts, which enables us to take another step toward
explaining political choices in dictatorships.
A second, perhaps less obvious, reason for the difficulty in developing a
systematic understanding of authoritarian politics is the great heterogeneity

across autocracies in the way decisions are made and leaders chosen, which
groups influence these decisions and who is excluded, who supports the dicta-
torial elite, and who benefits from their decisions. The policy-making process in
Saudi Arabia is quite different from that in China, and both differ from
decision-making procedures during military rule in Argentina between
1976 and 1983. Thus, while many theories about how democratic governments
function fit all democracies, useful theories about authoritarian politics have to
explain and take into account the differences among dictatorships.
This means that in order to develop theories about authoritarian politics, we
may first need to explain their differences. Our approach to doing that begins
with identifying characteristics of the different groups that seize dictatorial
power, which we can observe before the dictatorship begins. Groups potentially
able to initiate dictatorship have varying capacities, resources, organizational
structures, ways of making decisions, and distributions of within-group power.
After the seizure of power, these characteristics shape the way decisions are
made in the ensuing dictatorship and who can influence them. Preexisting traits
of the seizure group, as we show below, affect which citizens outside the group
can influence decisions, which part of the citizenry is likely to support the
regime, how the dictatorship responds to citizens who oppose it, which domes-
tic and international policies the dictatorship chooses, and what kinds of formal
political institutions it establishes to solve its cooperation problems, monitor
potential opponents, and incorporate citizens into unchallenging forms of
participation.
We use the term “seizure group” to refer to the small group that literally
ousts the incumbent and takes over in order to initiate dictatorship, as well as
their organized support base. For example, when military officers seize power
in a coup, the seizure group includes both the individual coup plotters and the
part of the military (possibly all of it) that provides less active support for them.
Seizure groups are thus similar to Haber’s (2006) “launching groups” except
that we make no assumptions about their ability to solve collective action
problems or oust the dictator they install. Nor do we assume that the group
that helps the dictator seize power takes over the bureaucracy, courts, police,
and military because, in the real world, they often do not. Frequently, in fact,
these institutions remain staffed or partly staffed by individuals hired by those
who ruled before the seizure of power.
As an example, consider a coup such as the one in Argentina in 1976: the
commanding officer of the armed forces, supported by a consensus among
other high-ranking officers, seizes control of government. In this kind of situ-
ation, before the coup most members of the seizure group have direct control of
weapons and expertise in using them but less expertise in bargaining. They have
a hierarchical organization structure that can usually ensure disciplined imple-
mentation of orders by soldiers (though not civilians), as well as a technocratic
decision-making style, with decisions concentrated at the top of the military
hierarchy. These characteristics can be expected to influence the way decisions

are made after the seizure of power. Because of strong norms about the chain of
command within the military, for example, military seizure groups usually
choose their highest-ranked officer as dictator.
Contrast this portrait of military rule with rule by a party that won an earlier
free and fair election, but then used its control of the legislature to pass laws
that severely disadvantaged the opposition, thus “authoritarianizing” the
government.4 For example, between 2002 and 2005, the elected Chávez
government in Venezuela used harassment and intimidation of the opposition,
arrests of opposition leaders, and interference with the media to authoritaria-
nize the political system. Other ways that democratically elected governments
have authoritarianized include banning opposition parties and closing
legislatures.
A party leadership that initially grew out of competitive election campaigns
has developed very different capacities, organization, and resources than a
military seizure group. Such party-based seizure groups typically lack weapons
and expertise in using them, but have a great deal of experience in bargaining,
cooptation, and electoral mobilization. The party may have a hierarchical
organization in the sense that decision-making is concentrated in the party
leader, but its activists and employees tend to be undisciplined because of the
use of jobs and other benefits to coopt opponents and buy support from people
with diverse interests. As a consequence, decisions made at the top may be
distorted during implementation to benefit local officials or simply not imple-
mented. Like the central traits of military seizure groups, these characteristics
also tend to carry over into post-seizure dictatorships established by parties that
were once fairly elected.
Several intuitions lie behind the claim that preexisting characteristics of the
group that establishes the dictatorship persist and shape political processes that
follow. First, we expect the inner circle of the dictatorship to be chosen from the
seizure group. Second, we expect groups represented in the inner circles of
dictatorships to dominate early decision-making and to have more influence on
decisions than excluded groups throughout the life of the dictatorship. We also
expect organized included groups to wield more power than unorganized ones.
Parties and militaries are large, often well-organized groups frequently repre-
sented in seizure groups and thus initially in the dictator’s inner circle. The
dictator’s inner circle may also represent the interests of particular class, ethnic,
religious, or regional groups, but since such groups tend to be loosely organ-
ized, we expect them to have less capacity to influence decisions and implemen-
tation than more effectively organized groups. Third, we expect groups that
have developed skills and routinized ways of interacting and making decisions
to gravitate toward these same ways of doing things immediately after seizures
of power. Our theories build on these intuitions.

4
This is a frequent means of establishing dictatorship, as shown in Chapter 2.

We expect preexisting characteristics of the seizure group to inﬂuence the

kind of autocratic regime that emerges after a seizure of power. By “regime” we
mean the set of very basic formal and informal rules for choosing leaders and
policies. We “measure” regimes as the continuous country-years in which
the same group – though not necessarily the same individuals – controls the
government and uses the same basic rules. Regimes can and often do include the
tenures of more than one dictator, as in China under Communist Party rule or
Saudi Arabia under the Al Saud family dynasty. “Very basic” rules include such
unspoken requirements as the necessity for paramount leaders to come from
particular ethnic groups or from highest-ranking officers, not necessarily
electoral rules or constitutional provisions, which the dictatorial inner circle
can change and/or abrogate. Because dictatorships lack third-party enforce-
ment of formal political rules (Svolik 2012), the kinds of formal political insti-
tutions that shape politics in democracies have less influence on the behavior of
elites in dictatorships. The basic rules that define dictatorial regimes are those,
whether formal or informal, that really shape the choice of top leaders and
important policies.

implementing our definition of regime

For gathering the data on which this study depends, we relied on a set of
detailed rules to identify the beginnings and ends of autocratic regimes. In
keeping with much of the literature, countries are coded as democratic if
government leaders achieve power through direct, reasonably fair competitive
election; indirect election by democratically elected assemblies; or constitu-
tional succession to democratically elected executives.
Events that deﬁne the beginning of dictatorship include the following:
Government leaders achieve power through some means other than a direct,
reasonably fair competitive election; or indirect election by a body at least
60 percent of which was elected in direct, reasonably fair, competitive
elections; or constitutional succession to a democratically elected executive.
The government achieved power through competitive elections, as described
above, but later changed the formal or informal rules such that competition
in subsequent elections was limited.
If competitive elections are held to choose the government, but the military
either prevents one or more parties for which substantial numbers of citizens
would be expected to vote from competing, or dictates policy choice in
important policy areas (e.g., foreign policy in the Middle East). We label
such regimes “indirect military rule.”
These rules are mostly uncontroversial, but they lead to a few coding deci-
sions with which others may disagree. For example, we code the government of
Indonesia’s popular ﬁrst leader, Suharto, as authoritarian because he was not
elected before taking power. Some other independence leaders whose

governments were coded as democratic at independence (because of fair com-

petitive pre-independence elections) were later classified as authoritarian after
they banned an opposition party, arrested opposition leaders, or used violence
and intimidation against opposition voters.
Once a country-year is coded as authoritarian, successive years in the same
country are coded as part of the same regime until one of the following events
occurs:
A competitive election (as defined above) for the executive, or for the body
that chooses the executive, occurs and is won by a person other than the
incumbent or someone closely allied with the incumbent; and the individual
or party elected is allowed to take office. The end date for the regime is the
election, but the regime is coded as ending only if the candidate or party
elected is allowed to take power.
Or the ruling group markedly changes the rules for choosing leaders and
policies such that the identity of the group from which leaders can be chosen
or the group that can choose major policies changes.
Or the government is ousted by a coup, popular uprising, rebellion, civil
war, invasion, or other violent means, and replaced by a different regime
(defined as above: a government that follows different rules for choosing
leaders and policies).
We code competitive elections as ending dictatorships only if the incumbent
is defeated because so many dictatorships hold competitive elections. Many of
the ways that dictatorships manipulate electoral outcomes do not occur on
election day or during vote counting, so foreign observers may not see rigging.
This makes it difficult to judge whether elections are free and fair. We use
incumbent turnover because it is a clear indicator that the dictatorship did not
control the election outcome. This is a conservative rule in the sense that we
code dictatorships as continuing unless we are sure they have ended. This rule
leads to a few controversial classifications; for example, we code Ghana’s last
democratization as of the 2000 election, which led to the first incumbent
turnover since Flight Lieutenant Jerry Rawlings’s seizure of power, but many
analysts code it as of 1992, the first multiparty election.
The most difficult coding decisions involve coups (defined as the overthrow
of the incumbent leader by members of the military of the regime being
ousted),5 which sometimes replace the whole ruling group and sometimes
replace only the dictator in an ongoing regime. We classify coups as regime
changes if they replace the incumbent government with one supported by
regions, religions, ethnicities, or tribes different from those that supported the
ousted incumbent or if they eliminate civilian collaborators from the inner

5
Our deﬁnition is thus consistent with everyday usage, but differs from that of Svolik (2012) and
Roessler (2016), who label any replacement of the dictator by regime insiders as a coup.

circle. Such coups are coded as regime changes because they change the com-
position of the group that can influence policy and leadership choice. For
example, we treat the coup that replaced the military dictatorship led by
Colonel Saye Zerbo in Burkina Faso with one led by Captain Thomas Sankara
as a regime change because Sankara’s military faction was rooted in different
ethnic groups than Zerbo’s (Englebert 1998, 51–65). If a coup simply replaces a
ruling general with another general from the military command council, with-
out changing the underlying group from which leaders are selected, we code it
as a leader change, but not a regime change.
In a very small number of instances, we also classified leader changes in
party-led dictatorships as regime changes because of dramatic changes in the
ethnic, religious, regional, or tribal base of the ruling group initiated by the new
dictator. In nearly all situations, a peaceful transition from one leader to
another in a dominant-party regime would be considered a leader change in
an ongoing dictatorship. However, Paul Biya’s succession in Cameroon is one
of the handful that we coded as a new regime. The coalition that had supported
Ahmadou Ahidjo, Cameroon’s first leader, was multiregional and multiethnic
though northern Muslims were favored; Ahidjo was a northern Muslim and
Biya, his prime minister, a southern Christian. Soon after becoming president,
Biya began narrowing the group with political influence and concentrating
power in his own small (southern) ethnic group at the expense of the coalition
he had inherited. The post-1983 government is treated as a different regime
because the regional and ethnic bases of policy influence changed along with
the group from which officials were selected (Harkness 2014, 598–99).6

the groups that initiate dictatorships

Two kinds of groups establish most contemporary dictatorships: groups of
ofﬁcers and soldiers, and groups of civilians organized into parties. Historically,
tribal or clan leaders supported by armed followers began most monarchies,
but this means of establishing dictatorship may have disappeared. Outgoing
colonial rulers (re)established a number of post–World War II monarchies.

Seizures Led by a Group of Military Ofﬁcers

Military officers usually achieve power via coups, though they are sometimes
handed political control during popular uprisings. After the seizure, officers
initially decide who will rule the country and make basic policy decisions.
Policy choice in some areas may be delegated to civilians, especially for deci-
sions that require technical expertise, but officers choose which civilians and

6
For the full coding rules for identifying regime beginnings and ends, see the Appendix to this
chapter. The data on regimes and all coding rules for deﬁning them are available at http://sites
.psu.edu/dictators.

can dismiss them. Other officers may have the capacity to constrain the military
dictator and force him to consult about major decisions, because the wide
dispersion among officers of arms and men under orders ensures that they
can make credible threats to oust the dictator if he fails to consult or to heed
their advice.7 Such credible threats to depose give the dictator incentives to
consult with other officers, but only if they can in turn make credible promises
to support him if he consults. The dictator has less reason to consult with
officers from a recently created, undisciplined, or factionalized officer corps,
which cannot make credible promises of support because those bargaining with
the dictator cannot count on being obeyed by junior officers and thus cannot
prevent rogue coups.
An example of a seizure of power by a unified military is the 1980 coup in
Turkey led by General Kenan Evren and the rest of the military high command.
The Turkish army has a long history of disciplined professionalism. The coup
was planned by the high command and voted on by the generals at the War
Academy.8 They ruled through the National Security Council, a consultative
body composed of the service chiefs and the commander of the gendarmerie.
During the dictatorship, officers made many key policy decisions and chose
civilian technocrats to handle the economy. Officers also planned and oversaw
the orderly return to civilian rule that ended the dictatorship.

Seizures Led by Parties

Parties achieve dictatorial power in three main ways: via “authoritarianization”
after winning competitive elections, by armed insurgency, and through impos-
ition by a foreign occupier. The different means of establishing dictatorship are
associated with different internal party-governing structures and different ways
of interacting with ordinary people. But in all, party leaders and procedures
control personnel appointments and hence the political careers of those who
wish to share power and influence or work for the government after the seizure
of power. High party officials, who can include officers as well as civilians, can
constrain the dictator if his hold on office depends ultimately on their support.
Dictatorial ruling parties range from highly organized, disciplined networks
with tentacles reaching into every neighborhood and village to cliques of the
dictator’s friends who can mobilize public employees to turn out votes for the
dictator in sham elections but perform few other functions. Parties that led a

7
Historically, the vast majority of dictators and members of their inner circles have been male. The
data set used as the basis for most empirical statements in this book includes one female dictator,
who served as regent for a year during the minority of the prince designated as successor to the
deceased king of Swaziland.
8
This experience contrasts of course with the failed Turkish coup of 2016, which was organized by
one faction of the military but defeated by a combination of courageous civilian mobilization and
loyal troops from opposing factions.

long struggle to mobilize popular support prior to the seizure of power tend to
be more organized and disciplined than those that were cobbled together during
the last competitive election before authoritarianization or those created after
the seizure of power in order to reward the dictator’s supporters with public
employment and other benefits.
As an example of party seizure of power, the United National Independence
Party (UNIP) of Zambia used a typical authoritarianization strategy. UNIP and
its leader Kenneth Kaunda had led the independence movement and won a fair
pre-independence election in 1964. The party enjoyed widespread popularity
at independence and developed an effective grassroots political network by
fighting the competitive elections held before authoritarianization. UNIP
transformed itself into a dominant-party dictatorship by using intimidation
and violence against opponents to ensure its victory in the 1967 by-elections.
It then ruled Zambia until 1991, maintaining its grip on power by banning rival
political parties and repressing opponents. During UNIP rule, civilian party
members dominated policy-making and governance. The party elite had little
ability to constrain its leader, however. Until 1967, Kaunda unilaterally chose
the party’s executive committee members. Then, after the first internal party
elections, the executive committee factionalized along ethnic/regional lines,
leading Kaunda to reassert personal control over it to reduce ethnic conflict
(Molteno 1974, 67–68; Tordoff and Molteno 1974, 9–11, 29, 35). If the
dictator chooses the members of the leadership group, they cannot limit his
decision-making autonomy.

Seizures Led by What Becomes a Ruling Family

Seizures of power that result in family-controlled dictatorships have occurred
via the conquest of territories by a family-led group and their armed supporters,
usurpation of the throne of an established monarchy by the armed followers of
a different family or clan, and imposition by an outgoing colonial power. After
such seizures, a ruling family chooses future leaders and plays an important role
in the decision-making inner circle, though economic and some other aspects of
policy are often delegated to commoners with expertise. The monarch’s broth-
ers, uncles, and/or sons often control the most important ministries and lead the
military and security services (Herb 1999). In this way, power is dispersed
within the ruling lineage, which both limits the discretion of the monarch and
protects the ruling family from external challenges. Powerful members of the
family can inﬂuence policy decisions because, in extreme circumstances, they
can dismiss the monarch.
What became the ruling family of Oman established the Al Said dynasty via
traditional military prowess. In 1741 Ahmed bin Said al Busaidi, governor of
Sohar on the coast of what is now Oman, led the city’s defense against a Persian
invasion during a very chaotic time. As a result, he was formally chosen as
imam in about 1744. The Al Said have remained in power as traditional sultans

ever since. In the current era, Said bin Taimur inherited the throne when his
father abdicated in 1932. His son, Qabus bin Said Al Said, ousted him in
1970 at the behest of the rest of the family (and with support from the British)
for obstructing investment and development. Qabus continues to rule today
(Mohamedi 1994; Smyth 1994; Plekhanov 2004, 94–99).
In contrast to Oman, Jordan’s Hashemite dynasty is a British creation. The
ﬁrst Jordanian monarch, King Abdullah I, was the son of the Ottoman emir of
Mecca, who claimed a hereditary right to rule in the Hijaz (now part of Saudi
Arabia). He was a leader of the Arab nationalist movement against Ottoman
rule and sided with the British during World War I. In 1921, the British
appointed him emir of Jordan, a state constructed by the British, and he
remained in power when Jordan gained independence in 1946. When Abdullah
was assassinated in 1951, his oldest son succeeded to the throne but abdicated
in favor of his underage son the following year due to mental illness. A Regency
Council controlled the country until Crown Prince Hussein came of age.
Hussein died in 1999 and was succeeded by his oldest son, who continues to
rule (Haddad 1971, 484–91; Lewis 1989; Wilson 1990; “Background Note:
Jordan” 2011).

conflict and bargaining within the seizure group

The different groups that initiate dictatorship have different arrangements for
organizing themselves, making decisions, and taking action. The procedures
and norms established in the seizure group before the dictatorship exists then
influence who rules after the seizure of power, the kind of institutions they
create to organize their rule, and their first policy choices. The group’s preex-
isting rules and procedures for making decisions provide a focal point in the
chaos that characterizes the early weeks of many dictatorships. Many seizure
groups apparently give little thought to the practical details of what they will do
after ousting the old regime. Following the Iraqi coup in 1968, for example,
“Like all other post-coup governments in Iraq, al-Bakr and his colleagues had
no very clear idea about politics or administration on a day to day basis”
(Farouk-Sluglett and Sluglett 1987, 116). After the Sudanese coup in 1958,
Woodward says, “it was clear that few plans for the post-coup situation had
been made” (1990, 102). Some plotters have not even chosen who will rule the
nation. Describing the 2008 coup in Mauritania, Pazzanita notes, “Although
the 6 August coup was swiftly and efficiently executed . . . the composition
of the HCE (Haute Conseil d’Etat, the new ruling council) evidently was not
well thought-out beforehand” (2008, 160). Sometimes more than one member
of the seizure group assumes he will become regime leader.
Under pressure to make many decisions quickly, the new rulers tend to rely
on their existing leadership and familiar decision-making structures. After the
seizure of power, members of the group must quickly decide who will lead, who
will fill other top offices, and how much power the new leader will have relative

to other members of the inner circle. These are difficult, hugely consequential,
and potentially dangerous decisions that can provoke conflict within the group.
Preexisting traits of seizure groups shape dictatorial choices immediately
after they gain power, but power struggles, policy failures, and other events
sometimes change the power of the dictator relative to other members of the
dictatorial elite in later months and years. Disagreements within the ruling
group can trigger splits among previously close allies. Power struggles are
common during the early years of dictatorships. Winners may jail or kill losers,
torture losers’ supporters, and impoverish their families. Power struggles are
frequently entangled with institutional innovations and reversals. These con-
flicts and institutional experiments can change the distribution of power within
the ruling group, either strengthening procedures for consultation or increasing
the concentration of power and resources in the new dictator’s hands. Post-
seizure power struggles can result in much greater concentration of power in
the dictator’s hands than his comrades intended or foresaw.
We label dictators who have concentrated powers in their own hands
personalist. Personalism tends to develop after the seizure of power, as we
show below, when seizure groups are factionalized and lack discipline.
A disunited group cannot prevent the new dictator from playing off first one
faction and then another against the others, in the process ridding himself of the
supporters most capable of challenging his decisions and power grabs. If
dictators can choose the members of the regime’s top decision-making inner
circle, they can change its composition without taking into account party
procedures, the military chain of command, or, in monarchies, the opinions
of ruling-family members. The dictator’s control over appointments to the inner
circle means that he can threaten his lieutenants with exclusion from power and
benefits, but the lieutenants cannot credibly threaten him with ouster.
A unified seizure group, in contrast, can enforce its standard procedures for
promoting officers or choosing members of the party executive committee
because it can overthrow a dictator who violates group norms. It can thus limit
the dictator’s discretion over the composition of the dictatorial elite. It can also
block the dictator’s efforts to take personal control of internal security services.
A dictator who can spy on, intimidate, or kill other members of the dictatorial
elite cannot be constrained by them. Other members of the dictatorial elite who
understand this will try to retain control of the security forces within the party
or regular military chain of command rather than permitting the dictator to
take personal control of them.
The leader of a new dictatorship rarely controls recruitment to the inner
circle or the internal security services the morning after the seizure of power.
These are weapons he may acquire, however, through the jockeying over power
among members of the seizure group during the first months and years after the
group has gained control. Personalist rule can arise in any kind of seizure
group. Like Idi Amin of Uganda, personalist dictators often originate in the
military, leading many observers to refer to the dictatorships they lead as

military regimes despite the marginalization of most ofﬁcers from decision-

making. And like Hugo Chávez of Venezuela, personalist dictators often rely
on parties to organize supporters. Nevertheless, they achieve so much control
over the lives and prospects of party officials and military officers that they are
largely unconstrained by these institutions. Though personalization develops
after the seizure of power, the internal characteristics of the seizure group that
give the dictator advantages in bargaining with other members of the inner
circle predate the seizure. Discipline and unity take time to develop in organiza-
tions and cannot be produced overnight when the challenge of controlling a
new dictator arises.
Bargaining and conflict between the dictator and members of the inner circle
are central features of authoritarian politics and among the things we would
most like to understand, since they affect international behavior and many
other policy choices. Previous research has shown, for example, that dictators
who have concentrated great power in their own hands start more wars than
more constrained dictators and pursue more erratic economic strategies (Frantz
and Ezrow 2011; Weeks 2012). The bargaining and conflict within the dicta-
torial inner circle can also motivate the initiation or change of formal political
institutions, which can help routinize decision-making, end destructive power
struggles, or consolidate the power of one man. Institutional choice is another
aspect of authoritarian politics we would like to understand since some insti-
tutions seem to stabilize authoritarian rule, while others contribute to
political chaos.
This book addresses these subjects. It is about authoritarian politics. That is,
the book explains why some dictators make certain policy choices while others
make different ones, and then how these decisions affect the dictatorship’s
vulnerability to threats to its survival. It thus takes a step toward answering
some of the questions policy makers and other observers would like to answer.
The first stage of our explanation begins with preexisting characteristics of
seizure groups, not only because they help explain dictatorial decision pro-
cesses, but also because these characteristics are exogenous to the dictatorship;
that is, they predate it. Other observable features of dictatorships, such as how
many parties they allow, whether they have a legislature, and the extent to
which they rig their elections, are strategic choices made by the dictatorial elite
after the seizure of power and thus are part of what needs to be explained.
Our starting point for thinking about autocratic differences is not the only
one possible. Other analysts, such as Hadenius and Teorell (2007), Gandhi
(2008), and Boix and Svolik (2013), have proposed alternatives based on
whether the dictatorship has a legislature and the number of parties allowed
in it. In a related vein, Levitsky and Way (2010) classify dictatorships based on
how much competition they allow in elections. These classifications are useful
for some purposes, but the characteristics they use as starting points reflect
attempts by ruling groups to retain power. These features should not therefore
be used to explain authoritarian durability, for example, because we cannot

rule out the possibility that something that contributed to the dictatorial deci-
sion to establish the institution being investigated also caused the outcome
(Pepinksy 2014). We therefore opt for a theoretical approach based on exogen-
ous characteristics of the seizure group, measured before the existence of the
dictatorship.
For theory testing purposes, the resources and capacities of the seizure group
before it takes control (or the observable traits that reflect unobservable
resources and capacities) are exogenous. Capturing these characteristics, how-
ever, requires detailed and subtle data. The data need to allow the assessment of
characteristics of pre-seizure party organizations and of how factionalized the
military that seizes power was before the seizure. Military factionalization and
weak party organization, as we show in what follows, predispose seizure
groups toward the personalization of dictatorial control.
To test other arguments, we also need to assess post-seizure features of
dictatorial rule, and the real world of dictatorships is complicated. Officers
sometimes lead regimes in which a ruling party makes central decisions and
most officers simply obey orders. In other military-led dictatorships, one officer
may manage to concentrate a great deal of power in his own hands, in the
process excluding most other officers from decision-making. Since we want to
distinguish these kinds of dictatorship from those in which a group of high-
ranking officers rule in a somewhat collegial fashion, we need information
beyond knowing whether the dictator wears a uniform. Ruling parties exist in
most dictatorships, including many regimes controlled by military officers and a
couple of monarchies, as well as regimes actually led and controlled by parties.
Thus, we cannot use the simple existence of a regime-support party to infer that
party institutions really control dictatorial decision-making. Ideally, we would
like to be able to distinguish dictatorships led by groups of officers representing
the military institution from other dictatorships led by individual officers, and
we would like to distinguish dictatorships in which party officials have some
ability to constrain the dictator from those with toothless parties. For these
reasons, we collected and make use of the Authoritarian Regimes Data Set,
which was designed to capture the distinctions we consider theoretically
important, including, among other things, specific features of the seizure group
both before and after the seizure of power.9

plan of the book

The order of the book follows the common sequence of challenges faced
by dictatorial elites: (1) initiation, the seizure of power; (2) elite consolidation;
(3) the extension of rule to society – policy implementation and information

9
The dataset, along with a codebook that explains the rules for coding, is available at our data
website: http://sites.psu.edu/dictators/how-dictatorships-work/.

gathering; and (4) breakdown.10 To correct common misperceptions about

dictatorships, Chapter 2 provides some concrete facts about how they come
into existence and what happens immediately after seizures of power. It shows
the frequency of different methods of seizing control, the kinds of regimes that
dictatorial seizures of power replace, and which kinds of groups use which
methods to gain political power. Most dictatorships replace an earlier auto-
cratic regime rather than a democracy – as, for example, the replacement of
monarchies by military officers in Afghanistan, Burundi, Cambodia, Egypt,
Iraq, Libya, and Yemen. Knowing something about how dictatorships begin
helps us to understand their later political choices.
Coups are the most common means of initiating new dictatorships, simply
because they are easier to organize than insurgencies or popular uprisings.
Scholars have disagreed about the reasons for military seizures of power. In
Chapter 3, we investigate the reasons for coups that initiate dictatorships. We
find no relationship between mass popular mobilization, rebellion, or income
inequality and coups that replace either democracies or incumbent dictator-
ships. Thus, we find no evidence that officers represent the interests of economic
elites. Instead, we find support for Nordlinger’s (1977) claim that coups reflect
the interests of officers.
Whether they secure power through a coup or in some other way, once in
control of the capital, members of the seizure group immediately confront all
the problems that helped make the overthrow of the old regime possible. They
also have to establish or reorganize internal security police to prevent the
mobilization of opposition. While juggling these problems and tasks, they must
also assign spheres of authority to different members of the new ruling group
and devise methods for settling disputes among themselves. The starting point
for assigning tasks and making decisions is the preexisting organization of the
seizure group and its norms and procedures for choosing leaders, making
choices, and maintaining internal unity.
Chapters 4 and 5 focus on the elite consolidation stage, which is dominated
by sometimes violent struggles over power and distribution within the inner
circle of the dictatorship, beginning soon after the seizure of power (Jowitt
1975). Chapter 4 develops our central theory of politics in dictatorships. The
choice of one member of the seizure group as regime leader (dictator) creates a
serious control problem for the rest of the dictatorial inner circle. Much of the
conflict during the first years of dictatorships arises from the inner-circle
members’ efforts to control the dictator, and the dictator’s attempts to escape

10
We revise and extend Kenneth Jowitt’s (1975) implicitly evolutionary argument that successful
Leninist regimes exhibit different traits during different stages of their existence. Building on
Crane Brinton (1938), Morris Janowitz (1977, 8) also describes a “natural history of revolu-
tion,” in which a chaotic revolutionary period is followed by the consolidation of power by one
man, itself followed by some relaxation. We suggest a general explanation for these patterns,
which we think apply to nonrevolutionary dictatorships as well as post-revolutionary ones.

control. This struggle leads either to the concentration of power and resources
in the dictator’s hands or to the reinforcement of somewhat more collegial rule.
In order to limit the dictator’s personal discretion, his closest allies must be able
to credibly threaten him with ouster. Threats of ouster are more credible when
the ruling group is unified. This chapter shows how factionalism within the
seizure group enables the dictator to consolidate personal power. Using detailed
historical data on observable features of dictatorial rule, we offer the first
empirical evidence linking pre-seizure characteristics of the seizure group to
the consolidation of personal power in the hands of dictators.
Chapter 5 introduces the complication of wide dispersal of control over arms
across members of the dictatorial elite. In armed seizure groups, typically most
members of the inner circle have direct access to weapons and command over
men trained to use them. This makes it possible for many different individuals
to threaten the dictator with ouster, thus increasing the bargaining power of
members of the inner circle relative to the dictator. All else equal, the increased
bargaining power of members of the inner circle should lead to more constraint
on the dictator, as the ease of ouster makes the dictator’s promises to share
more credible. In factionalized armed seizure groups, however, other members
of the inner circle cannot make the credible commitments to support the
dictator that are needed to solidify their side of the exchange. Their promises
to support the dictator if he shares power are not credible because factionalism
or indiscipline undermines their control of their subordinates, who also have
guns. Consequently, they cannot sustain power-sharing bargains.
A dictator facing this situation has little reason to share power with other
officers. Instead, he has strong incentives both to invest in new security forces
loyal to himself to counterbalance his unruly and potentially disloyal military
supporters (discussed in Chapter 7) and to create civilian support organizations
to diversify his support, thus reducing his dependence on armed members of the
ruling group. Military dictators who cannot count on the rest of the military for
support because of factionalism or indiscipline in the army try to marginalize
most of the military from decision-making, and to shift the support base of the
regime to civilians, who are less threatening because unarmed. Dictators often
do this by having themselves popularly elected, creating a civilian support
party, appointing a civilian cabinet, and dissolving the military ruling council.
Observers label this series of events civilianization, and sometimes even inter-
pret it as democratization, but it is a dictatorial strategy to survive and consoli-
date personal power in the face of a factionalized and unreliable military
support group.
Once decision-making within the elite has become somewhat routinized, the
next challenge facing dictatorships is extending their rule over society. They
must either create new agencies for pursuing the radical changes in policy that
motivated the seizure of power or trust implementation to public employees
likely to have been the patronage appointments of the ousted government. In
order to rule, they have to have information about what is happening in

different parts of the country and how their policies are working. They also
have to be able to monitor the behavior of lower-level officials to make sure
their policies are not being sabotaged. These are difficult problems for dictators
and their allies. Officials tasked with reporting on local conditions often have
incentives to misreport conditions on the ground.
Dictators use a number of strategies to try to ensure the loyalty of officials
and routinize the collection of accurate information about the grassroots. In
Chapter 6 we focus on their use of mass organizations that engage citizens,
especially ruling parties, elections, and legislatures, to incentivize information
gathering and good behavior in officials. Dictatorial elites often make ruling-
party membership a condition for public employment to try to ensure the
loyalty of officials. Ruling parties link central elites to vast numbers of officials
and state employees. Elites hope to exchange jobs and other benefits for loyalty,
effort, and honesty from officials. Without monitoring, however, they cannot
ensure that officials live up to their side of the bargain.
Elections for legislative and local offices can partially substitute for regular
monitoring. Although the ruling party is unlikely to lose elections, individual
officials can lose if citizens are fed up with them. Bad local election outcomes
notify central authorities about especially corrupt, incompetent, or abusive offi-
cials and unworkable policies. Local and legislative offices are highly valuable
because of the benefits that accompany office, so the possibility of losing elections
incentivizes officials to extend their distributive networks down to the grassroots,
limit theft, and lobby central officials for benefits for their areas. Future elections
also motivate officials to report local problems and policy failures to the center
and to compete for access to local public goods. In these ways, elections and the
knowledge that they will face future elections incentivize the transmission of
information about local conditions to the center. Even choice-free elections can
serve this purpose because competition for ruling-party nominations is intense
regardless of whether there is partisan competition.
In Chapter 7 we consider the other side of monitoring and information
gathering: spying. We describe the coercive institutions that protect dictator-
ships. We stress the differences between the interests of army officers and those
of internal security police. We describe the sources of military autonomy from
the dictatorships they are supposed to protect and how dictators try to over-
come their autonomy. In contrast to the army, internal security services are
usually created anew to serve dictatorships and thus have less autonomy. Their
main tasks are to spy on and intimidate anyone who might oppose the dictator
or dictatorship, which includes members of the dictatorial elite, the ruling
party, and military officers. Internal security agencies can be controlled by,
and thus serve as agents for, the ruling party, the military high command, or the
dictator himself. In this chapter, we analyze how control of internal security
changes the distribution of power within the dictatorial inner circle.
Chapter 8 analyzes authoritarian breakdown. Events beyond the dictator-
ship’s control can reduce the costs of various behaviors – such as participating

in demonstrations, campaigning for the opposition, and plotting – that in turn

reduce regime survival chances. In this chapter, we show how characteristics of
the dictatorship shape its ability to withstand such challenges. We describe how
the relationship between the dictator and his inner circle interacts with insti-
tutions originally created for other purposes – such as parties formed to win
democratic elections – to reduce vulnerability to different kinds of challenge.
This interaction explains why economic crisis dooms some dictatorships but
not others and why some have more difficulty surviving succession struggles
after a dictator dies. The chapter highlights the relationship between past insti-
tutional choices and regime survival. We also show how different expected
post-exit fates affect the responses of both dictators and their closest allies to
popular opposition. We bring these various strands together to explain, first,
why some kinds of dictatorship are more resilient than others in the face of
challenges and, second, why some kinds of dictatorship tend to permit fair,
contested elections when faced with widespread opposition and thus to exit
peacefully, while others clutch desperately at power until bloodily removed.
In Chapter 9, the concluding chapter, we summarize the facts established
and arguments made in other chapters. Several of our findings lead directly to
policy recommendations. For example, the evidence we offer implies that policy
makers should hesitate to intervene in personalist dictatorships (such as
Moammar Qaddafi’s in Libya or Saddam Hussein’s in Iraq). Such dictators
may be hated and incompetent, and they may have committed horrific human
rights abuses, but deposing them may nevertheless make the average citizen of
the countries they have ruled worse off. The ouster of personalist dictatorships
is less likely to result in democratization than the overthrow of other kinds of
dictatorship (Geddes, Wright, and Frantz 2014). A violent ouster will probably
result in a new autocracy, but could end in civil war or a failed state.
These outcomes occur because the destruction of both political and civil
society institutions under personalist rule leaves nations that have endured it
with very little human infrastructure with which subsequent political leaders
can build stable government. The decimation of institutions under personalist
rule often includes the military and police, leaving these forces subsequently
incapable of maintaining order or defending the new government from violent
attacks by supporters of the ousted dictator. In addition, the personalist dicta-
tor’s systematic elimination of politically talented potential rivals reduces the
quality of the pool from which new leaders can come. These several kinds of
damage mean that post-intervention governments struggle to carry out ordin-
ary functions such as keeping streets safe and delivering water.
The foreign intervener is then likely to be blamed for the violence and decline in
living standards that follow intervention. For these reasons, intervention to over-
throw a personalist dictator is unlikely to result in a friendly country with a stable
government willing to protect the intervener’s economic and security interests.
We conclude with a series of policy recommendations that, like this one, are
implied by our findings.

Coding Rules for Authoritarian Regimes

Dictatorship begins when any one of the following has occurred:

The government leader achieved power through some means other than a
direct, reasonably fair competitive election; or indirect election by an assem-
bly at least 60 percent of which was elected in direct, reasonably fair,
competitive elections; or constitutional succession to a democratically elected
executive.
○ Elections are not considered reasonably competitive if one or more large

party is not allowed to participate; and/or if there are widespread reports

of violence, jailing, and/or intimidation of opposition leaders or support-
ers; and/or if there are credible reports of vote fraud widespread enough to
change election outcomes (especially if reported by international obser-
vers); and/or if the incumbent so dominates political resources and the
media that observers do not consider elections fair.
○ Elections are not considered reasonably fair if less than 10 percent of the

population (equivalent to about 40 percent of the adult male population)

was eligible to vote.
○ Regimes are not coded authoritarian if an elected executive is ousted by

the military, nonconstitutional legislative action, or popular pressure, but

is succeeded by a constitutionally mandated successor and the successor
behaves in accordance with the constitution. (Such governments may be
unconstitutional, but they are not authoritarian regimes because they
continue to follow democratic rules concerning succession, length of term,
means of choosing the next executive, and legislative-executive
relationship.)
Or the government achieved power through competitive elections, as
described above, but subsequently changed the formal or informal rules such
that competition in subsequent elections was limited.

○ Events and rule changes that should be coded as causing a transition from
democracy to autocracy in electoral regimes:
Opposition parties representing more than 20 percent of voters are banned
or not allowed to run candidates in elections.
Most opposition parties are forced to merge with the ruling party.
Legislature is closed unconstitutionally.
There are widespread reports of violence and/or intimidation of oppos-
ition leaders or supporters; exclusion of opposition deputies from the
legislature; the jailing of one or more opposition leaders.
There are credible reports of vote fraud widespread enough to change
election outcomes (especially if reported by international observers).
Election results are annulled.
○ Start of autocracy dates from change in rules if formal rules are changed,

from date the legislature closed, from date of campaign in which violence
was ﬁrst reported, from election in which fraud is reported; from date
when annulment occurred; or from date when deputies were excluded or
when opposition leaders were jailed.
○ The following irregularities are not coded as authoritarian:

Reports of vote buying (because it is very common in democracies)

Scattered reports of fraud
Fraud complaints by the opposition without supporting evidence or
corroboration by neutral observers
Opposition boycott of election in the absence of other evidence of
unfairness
Or if competitive elections are held to choose the government, but the
military either prevents one or more parties that substantial numbers of
citizens would be expected to vote for from competing, or dictates policy
choice in important policy areas (e.g., foreign policy in the Middle East). We
label such regimes “indirect military rule.”
The start date for monarchical regimes is the first year of the dynasty if the
country was independent in 1946; or the first year of independence.
Once a country-year is coded as authoritarian, successive years in the same
country are coded as part of the same regime until one of the events identified
below as ending a regime occurs.
Authoritarian regimes end when any of the following occurs:
A competitive election (as defined above) for the executive, or for the body that
chooses the executive, occurs and is won by a person other than the incumbent
or someone closely allied with the incumbent; and the individual or party
elected is allowed to take office. The end date is the election, but the regime is
coded as ending only if the candidate or party elected is allowed to take power.
○ Remember that in cases of indirect military rule, the incumbent leader is

the top military ofﬁcer. If leaders of an indirect military regime change the
rules such that all major parties and population groups are permitted to

compete in fair elections, and the civilian winner is allowed to take ofﬁce
and to make policy in areas previously reserved for the military, this
change is coded as democratization. The authoritarian end date is the date
of the election, but cases are not included unless the person elected is
allowed to take ofﬁce.
○ If a country has both a popularly elected president and a prime minister

chosen by the elected legislature, and it is not clear which has the most
political power, loss of either ofﬁce by the incumbent party indicates the
end of the authoritarian regime.
Or the government is ousted by a coup, popular uprising, rebellion, civil
war, invasion, or other violent means, and replaced by a different regime
(deﬁned as above, as a government that follows different rules for choosing
leaders and policies). Regimes should be coded as ending if:
○ Civil war, invasion, popular uprising, or rebellion brings to power indi-

viduals from regions, religions, ethnicities, or tribes different from those

who ruled before (i.e., the group from which leaders can be chosen has
changed).
○ A coup (deﬁned as the overthrow of the incumbent leader by members

of the military of the regime being ousted) replaces the government with
one supported by different regions, religions, ethnicities, or tribes. If a
coup simply replaces an incumbent general from one military faction
with a general from another without changing the group from which
leaders are selected, this is coded as a leader change, not a regime
change.
○ Assassinations are treated like coups; i.e., if the assassinated incumbent is

replaced by someone else from within the same ruling group, it is not
coded as a regime end. If the assassinated incumbent is replaced by
someone from a different group, as described above, the assassination is
counted as a regime end.
Or the ruling group markedly changes the rules for choosing leaders and
policies such that the identity of the group from which leaders can be chosen
or the group that can choose major policies changes. Examples of regime
changes implemented by leaders of the incumbent regime include:
○ The new regime leader after a regular authoritarian succession (e.g., the

dictator dies and is succeeded by his constitutional successor) replaces the

most important members of the ruling group with individuals drawn from
a different region, religion, tribe, or ethnicity and changes the basic rules of
how the regime functions.
○ Transitions from military rule to indirect military rule, which occur when

military regime leaders allow the election of a civilian government that has
many of the powers of a democratic government, but military leaders
maintain substantial control over leader and policy choice, either by
preventing parties for which large numbers of citizens would be expected
to vote from competing or by directly controlling the selection of

important cabinet posts and policies. Indirect military regimes are coded as
distinct from the prior military-led regime because many in the leadership
are chosen through fair elections, and these elected officials control
important aspects of policy; they are not simply puppets. Transitions from
indirect military rule to democracy are coded from the date of the fair,
competitive election.
Transitions from indirect military rule to other forms of autocracy occur
when the elected civilian partner of an indirect military regime is removed
from office by the military partner or some other group. These changes
usually occur via coup.
Country-years are excluded from the authoritarian regimes data set if:
Country is democratic (defined as above)
Country has a provisional government charged with conducting elections as
part of a transition to democracy, if the elections actually take place and the
candidate and party elected are allowed to take office; or if a provisional
government that is following the rules set up as part of a transition to
democracy exists on January 1, but is ousted before the transition is com-
plete by a group different from the one that held power before the provi-
sional government was established; or if the provisional government
remained in power on January 1, 2010, the last date coded
○ To be considered transitional, the majority of top leaders cannot have been

top members of the prior authoritarian regime during the months preced-
ing the change in leadership.
○ If instead of holding elections, the provisional government converts itself

into the “permanent” government, it is coded as authoritarian.

○ If elections are held but elected leaders are not permitted to take ofﬁce, it is

coded depending on who prevents them from taking office and who
governs instead.
If actors from the old regime (from before the provisional government)
prevent those who won elections from taking office and return to power
themselves, the provisional government and the one that succeeded it are
coded as a continuation of the authoritarian regime that preceded the
provisional government.
If actors from the old regime prevent those who won the elections from
taking office but replace them with a government drawn from a different
group than the one that ruled before (e.g., the military that used to rule
replaces elected civilians with a civilian whose base of support lies partly
outside the military), the new government is coded as a new authoritarian
regime.
If actors from the old regime prevent those who won the elections from taking
office, but they are then ousted by a group that forms a government based on
a different support group, using different rules for choosing leaders and
policies, the new government is coded as a new authoritarian regime.

Country is not independent

Cases are excluded if foreign troops occupy the country and the occupier
governs it, or if occupation is the only thing that prevents the country from
being coded as democratic. Cases are not excluded if a foreign power inﬂu-
ences the (authoritarian) government but allows it to make many decisions.
Country has no government or has multiple governments, no one of which
controls most of the resources of the state.
○ Cases are not excluded because of civil war if one government still controls

signiﬁcant territory and the capital.

Autocratic Seizures of Power

Most thinking about dictatorship pictures the starting point as the violent
overthrow of democracy – something like the bloody Chilean coup of Septem-
ber 11, 1973, in which military forces led by General Augusto Pinochet
overthrew elected President Salvador Allende. But in reality, less than a third
(30 percent) of post–World War II autocratic regimes began by replacing
democracies. Among these democratic breakdowns, more than a quarter
(28 percent) were “self-coups,” meaning that a democratically elected govern-
ment “authoritarianized” itself, usually by banning the opposition, arresting
its leaders, or closing the legislature. The Chilean scenario, the violent over-
throw of a democratic government, initiated only 20 percent of post-1946
dictatorships.1
In this chapter, we provide this kind of basic information about how dicta-
torships form, and we describe the conditions facing seizure groups the day
after the seizure of power. The ﬁrst sections show how dictatorships begin and
who begins them. We start with some facts about the initiation of dictatorship
because differences among seizure groups have long-term consequences, as we
show in later chapters. Next, we analyze the distribution of power between the
leader and other plotters before they have seized power. We then describe the
situation that faces seizure groups the morning after they oust the old regime
and the kinds of ruling groups that different kinds of seizure group typically
create. Our aim is to provide some background for the analysis in the chapters
to come.

1
“Violent” is deﬁned here as overthrow by a coup or armed rebellion.

Downloaded from https://www.cambridge.org/core. HarrassowitzN. American Lib Sub Svc, on 12 Aug 2018 at 07:37:03, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.002
26 Initiation

who do dictatorial seizure groups oust?

The largest proportion of post-war dictatorships began by ousting an earlier
dictator and his supporters as, for example, when a popular uprising ousted the
Shah of Iran in 1979 and ushered in the theocracy led by Ayatollah Khomeini,
or when the insurgency led by Laurent Kabila toppled Mobutu Sese Seku in
what is now the Democratic Republic of Congo. The regimes introduced by
these interventions differed from the dictatorships they replaced in that the
identity of the group controlling government changed. The new rulers altered
the pool from which leaders and elites could be drawn, as well as other basic
rules for making leadership and policy choices. In short, following the defin-
ition of regime used in this book, long authoritarian spells in these and many
other countries include two or more different authoritarian regimes.2 Forty-two
percent of autocratic regimes begin with the ouster of a different dictatorship.
Many of the remaining dictatorships, 26 percent of the total from 1946 to
2010, began with the ouster of foreign rulers or a handover from foreigners to
an undemocratic government. Most of these foreign initiations involved transi-
tions from colonial control at independence.3 In the remaining few cases,
autocratic rule has succeeded a period of warlordism.
This variation in the status quo ante may explain some of the difficulty
scholars have had in discovering systematic reasons for the initiation of dicta-
torship. The conditions that motivate the creation of an insurgency to end
colonial rule may be quite different from the conditions that motivate the ouster
of a populist democracy or the overthrow of an arbitrary traditional monarchy.
These figures also suggest that studying only those dictatorships that follow
democracies, as occurs inadvertently when analysts rely on data sets that code
only transitions between democracy and autocracy, could lead to a biased
understanding of both the causes of dictatorship and how they operate in
practice.

how dictatorships begin

We sometimes speak of dictators as though they rule alone, but individuals
lack the resources to overthrow or transform existing governments (Haber 2006).

2
Of course, not all leadership changes are regime changes, only those that accompany basic
changes in the identity of the group in power and the formal and informal rules for choosing
leaders and policies. See Chapter 1 for our deﬁnition of regime and Geddes, Wright, and Frantz
(2014) for an elaboration of how regime changes differ from leadership changes and undifferen-
tiated spells of consecutive years under authoritarian rule.
3
As colonialism has become uncommon and democracy has spread, these ﬁgures may change just
because there are fewer foreign occupiers and autocrats to oust. Since democracy is less suscep-
tible to overthrow than autocracy, however, the number of dictatorships may simply fall rather
than the proportions changing.

Only groups can do it. Groups plotting to attain dictatorial power choose
methods of achieving their goal that make the most of their own resources
and capacities relative to those of the group to be ousted or defeated. For
example, military ofﬁcers, who have an advantage in deploying force and the
threat of force, tend to take power through coups, which rely on credible
threats of armed violence. Democratically elected incumbent groups usually
assume dictatorial control through “authoritarianization,” a strategy
unavailable to nonincumbent seizure groups. They choose this strategy
because they already control their countries’ legal and judicial systems and
can thus most easily initiate dictatorship via legal changes like banning
opposition.
The empirical record indicates that modern dictatorships begin in six main
ways.

Coups (deﬁned as ousters carried out by members of the military of the

government being overthrown) replace the incumbent government with one
preferred by military ofﬁcers. Coups can replace either democratic or auto-
cratic regimes.
Insurgents defeat the incumbent militarily and replace it with their own
leaders. Insurgency is usually used to replace foreigners or incumbent auto-
crats with a new dictatorship supported by different groups. This strategy is
rarely successful against democracy.
Popular uprisings persuade incumbents to hand power to opposition or
seemingly neutral leaders such as military ofﬁcers. Popular uprisings can
result in dictatorship when the interim leader reneges on promises to democ-
ratize. Popular uprisings have occurred against both democratic and auto-
cratic incumbents.
Foreign conquest or imposition leads to the eventual control of the state by
a group preferred by the invaders. Foreign impositions have ended both
democratic and autocratic regimes, as well as nonstate forms of rule.
Autocratic elites initiate rule changes that alter the kinds of groups permitted
in the regime’s inner circle.
Competitively elected elites initiate rule changes that prohibit opposition
groups from competing effectively, a process we refer to as
“authoritarianization.” Such rule changes replace democracy with autocracy
under the same ruling group.

The last two kinds of seizure differ from the others in that incumbents
remain in the ruling group; such “seizures” are motivated not by the desire to
redistribute away from incumbents, as most other dictatorial seizures are, but
to prevent redistribution away from the group currently in power, which might
follow fully competitive elections.
Table 2.1 shows the distribution of these methods of seizure for post–World
War II dictatorships.

table 2.1 Proportion of dictatorships begun by different kinds of

seizures of power

Kind of Seizure Percentage of All Seizures

Coup 45
Foreign imposition* 16
Authoritarianization 15
Insurgency 14
Popular uprising 5
Elite rule change 5
* Includes colonial handovers to ruling groups not chosen in democratic elections.

As Table 2.1 shows, coups initiate the largest proportion of dictatorships.4

By definition, military officers carry out coups. Sometimes the officers have
civilian supporters, but coups nevertheless usually lead to government con-
trolled by an officer or group of officers. Examples include the 1973 Chilean
coup mentioned above and the 1963 ouster by a coalition of officers with
Ba’thist sympathies of an earlier Syrian dictatorship led by the military high
command in collaboration with an elected civilian president. The 1963 coup
was coded as a regime initiation because it replaced a government led by a
military high command consciously selected to unify the military despite its
partisan factions with one led by a single faction, the pro-Ba’thists (Be’eri 1970,
150–53; Haddad 1971, 294). The 1963 coup eventuated in the Assad
dictatorship that still rules much of Syria as this is written.
Foreigners imposed about 16 percent of post-1946 dictatorships, either as
outgoing colonial powers or as occupying forces. Outgoing colonial powers
created quite a few post–World War II dictatorships by handing power either to
a civilian party-led regime with formally democratic institutions or to a
monarchy that was constitutional on paper but not in practice. Soviet occupy-
ing forces created many of the rest. The first wave was the party-led regimes
with communist institutions imposed by Soviet occupation forces during and
after World War II. The second occurred when a number of ex-Soviet nations
achieved independence under governments chosen before independence by the
Soviets. Where we have labeled the initiation of a dictatorship foreign-imposed,
occupiers usually influenced the choice of individual leaders as well as the rules
and institutions for making future leadership and policy choices. Some foreign
impositions involved a finger on the scale rather than all-out coercion, as when
foreigners outlawed popular leftist parties during the period leading to pre-
independence elections in order to help a favored party or when they chose to

4
Coups have declined in frequency since the end of the Cold War (Kendall-Taylor and Frantz
2014; Marinov and Goemans 2014), but still accounted for more than one-third of all authori-
tarian initiations between 1990 and 2010.

negotiate independence with monarchists rather than republicans. Military

occupation and all-out repression occurred in other cases, as in Eastern Europe.
Since few colonies remain, foreign impositions have become less common.
Democratically elected incumbents initiate a surprising number of dictator-
ships by changing political institutions or practices to authoritarianize the
political system. Examples include the incremental transformation of Vene-
zuelan politics from democratic to authoritarian during the presidency of Hugo
Chávez and President Alberto Fujimori’s unconstitutional closure of Congress
and dismissal of judges (autogolpe) in Peru. Elected presidents supported by
ruling parties carry out most authoritarianizations.
Fourteen percent of autocracies begin when armed rebels defeat incumbents.
Parties lead most insurgencies. Until the end of colonization, many insurgencies
ousted foreigners. Examples include the Algerian, Vietnamese, and Mozambi-
can wars of independence. Insurgencies that have replaced domestic incum-
bents include the Sandinistas’ defeat of the Somoza family dictatorship in
Nicaragua in 1979 and the defeat of the Habré dictatorship in Chad by Idriss
Déby’s forces in 1990.
Popular uprisings and elite rule changes are the least common ways new
autocracies start.5 Popular uprisings generally demand democracy, and the
individuals to whom power is entrusted usually announce transitions to dem-
ocracy, but they sometimes consolidate authoritarian rule instead. Examples
include the uprisings that ousted Robert Guéï in Cote d’Ivoire in 2000 and the
Duvalier family dictatorship in Haiti in 1986. Guéï’s overthrow brought to
power Laurent Gbagbo, who had probably won the election that protesters
accused Guéï of stealing, but Gbagbo then governed as an autocrat (Englebert
2003, 332). A military-civilian transitional ruling council tasked with holding
elections replaced the Duvaliers, but ofﬁcers in the council prevented democra-
tization (Payne and Sutton 1993, 80–89).
Elite rule changes initiate new regimes in the sense that they change the
identity of groups that control leadership and policy choice, even though the
current incumbent remains a part of the inner circle of the new dictatorship.
An example is the transition from direct to indirect military rule in Guate-
mala in 1985, when military rulers agreed to allow competitive elections for
the presidency and legislature, but prevented popular leftist parties from
participating (which is the reason the case is coded as authoritarian, despite
competitive elections). The military also retained control over important
aspects of policy as a condition for allowing real civilian control in other
policy areas. Elected civilian President Cerezo said he “held no more than
30 percent of power with no control over the armed forces” (Schirmer 1998,
176). Under our deﬁnition of regime, Cerezo’s election began a new regime
because the pre-1985 political system limited leaders to high-ranking military

5
Since 1990, popular uprisings have become more common.

ofﬁcers, but the post-1985 regime involved a power-sharing arrangement

between officers and competitively elected civilians.
The initiation of the rump Yugoslavian (Serbian and Montenegrin) govern-
ment under the same leadership as before the breakup of pre-1990 Yugoslavia
is another example of an elite-initiated rule change. Slobodan Milošević, who
had led Serbia before the breakup but ruled pre-1990 Yugoslavia in coalition
with the leaders of five other nationalities, remained Serbia’s leader but within a
ruling group that included only Serbians and Montenegrins after the breakup.
Elite rule changes, like authoritarianizations, do not fit the standard image of
autocratic seizures of power, but, as Table 2.1 shows, they initiate a number of
dictatorships.
Current (or former) military officers and groups organized as parties carry
out a large majority of transitions to autocracy. Organized groups have a big
advantage over the unorganized when it comes to difficult and dangerous
endeavors like overthrowing governments. Soldiers carry out coups, but they
also sometimes lead other kinds of seizures. Ex-officers lead 8 percent of the
insurgencies that initiate dictatorships; officers implement more than half of the
elite rule changes that result in new regimes; and they are handed power during
almost one-third of the popular uprisings that end up failing to democratize.
Parties lead three-quarters of authoritarianizations, most insurgencies, and
some popular uprisings. Half of autocracies imposed by foreign powers are
led by dominant parties, as in Eastern Europe after World War II and a number
of postcolonial regimes in Africa and Asia.
Military and party leadership are not mutually exclusive. Some seizures
involve both officers or ex-officers and parties. The exceptions to leadership
by a party or the military include a few instances in which popular uprisings
have catapulted party-less opposition leaders to power; some insurgencies that
were not organized as parties before the seizure, though insurgent organizations
always perform some of the functions usually taken care of by parties; and
some cases in which foreign rulers left power in the hands of individuals not
supported by parties (usually monarchs, but Hamid Karzai in Afghanistan is
another). Table 2.2 shows the frequency of seizures led by current (or former)
military officers and/or parties.
Ousters of incumbent governments, whether democratic or autocratic, usu-
ally occur when things are going badly, that is, when the economy is in trouble,
when disorder and violence make people feel unsafe, and/or when scandal or
arbitrary brutality has discredited incumbents. These are the reasons coups are
often “welcomed throughout society because, initially, the military coup means
all things to all men. The army is popular not because of what it stands for (which
nobody knows, at first), but because of what, quite patently, it has fought
against” (Finer 1976, 104). After the overthrow of the Iraqi monarchy, the first
army broadcast invited citizens to “go out and watch the edifices of tyranny
crumble. Within the hour a mob of hundreds of thousands was milling through
the streets screaming its joy and its thirst for vengeance” (Dann 1969, 33).

table 2.2 Proportion of dictatorships begun by different kinds of

seizure groups

Seizure Leadership Percentage of All Seizures

Military 50
Party-based 40
Neither 18
Both 8
Note: Leadership of the seizure group is coded as military if the regime seized
power via coup or the military put the ﬁrst regime leader into ofﬁce. It is coded as
party if what became the regime support party existed prior to the seizure. The
sample consists of 280 regimes that appear in the post-1946 data, including those
that seized power before 1946.

Citizens welcomed the 1980 coup in Burkina Faso and also the 1982 coup that
ousted those who had been welcomed in 1980 (Otayek 1986). Surveys carried
out after the 1966 coup in Argentina and the 1992 autogolpe in Peru show
popular support for the takeover in each instance at above 60 percent (O’Don-
nell 1973, 39; Stokes 2001, 142).
When things are going badly for the incumbent and ordinary people are
thoroughly fed up, it is not uncommon for more than one group to plot during
the months before the successful ouster and for rumors of coups to circulate.
Sometimes newspaper editorials publicly urge the military to end an unsuccess-
ful government.6 Multiple different actors often reach the conclusion that
drastic action is needed to “save” the country or that bold action to advance
their own interests has become feasible. Several separate groups of officers may
be plotting coups at the same time, and sometimes senior officers stage coups to
forestall junior officers whom they know to be plotting. In Libya during
the months before Moammar Qaddafi’s coup, for example, “at least three
and possibly four groups were jostling to unseat the king” (Blundy and Lycett
1987, 53). Because multiple groups were known to be plotting, “Several officers
took part in [Qaddafi’s] coup without knowing who was leading it” (First
1974, 108–9). In such circumstances, there is quite a bit of luck involved in
which plotters complete their preparations first to become the actual seizure
group. One of the reasons no one opposed the Qaddafi coup is that the seizure
group did not announce their identity, and many other officers assumed that a
more senior group known to be plotting was in charge.
The choice of the method by which to replace an incumbent regime and
install a new dictatorship is largely opportunistic. Coups are the most frequent
method for ousting fellow citizens (as opposed to foreigners) simply because
officers’ access to weapons and command of soldiers make them much easier to

6
Stepan (1971), for example, reports the frequency of editorials calling on the military to oust the
elected government before the 1964 coup in Brazil.

organize than other overthrows. Minority military factions carry out many
coups. Qaddafi was a twenty-seven-year-old captain supported by a small
number of other low-ranking officers carrying small arms and forty-eight
rounds of ammunition (Singh 2014, 29). Small numbers of conspirators can
succeed in part because of military discipline; lower-ranked officers and soldiers
usually follow orders so few have to know about the plot. Describing the first
Nigerian coup, Robin Luckham writes, “the other officers and men thought
they were going out only on a night exercise” (1971, 31). After a coup attempt
in Guatemala, Jennifer Schirmer reports that “few lower-ranking officers and
none of the soldiers were told that the purpose of the mobilization was to
overthrow the government” (1998, 218). Rebellions and popular uprisings, in
contrast, require substantial voluntary sacrifices by participants. Coups oust
64 percent of autocracies that are replaced by new autocracies and 60 percent
of democracies that suffer breakdowns.7
A foreign occupier or colonial power, however, has never been ousted by
coup. The reason is obvious, but we mention the point because foreign occupa-
tion, whether after invasion or colonial incursion, is almost the only situation in
which the armed forces do not have an advantage when it comes to toppling
incumbents. Otherwise, the armed forces have a clear edge, which begins soon
after a military is created, as the many coups carried out by small, newly created
armies shortly after independence demonstrate.
This advantage means that military officers dissatisfied with an incumbent
government face fewer impediments to acting than do similarly dissatisfied
civilians. As a consequence, civilian groups that rate their chances of getting
into power through legal means as low, whether because they cannot attract
wide popular support or because they face an incumbent who prevents fair
elections, often ally with military factions. They hope to achieve power by
outmaneuvering officers after a coup ends the old regime. For example, the
communists in Afghanistan and the Ba’th Party in Syria and Iraq used this
strategy. It worked, in the sense that the party achieved power rather than being
marginalized by its erstwhile military ally, on the first try in Syria, on the second
in Afghanistan, and on the third in Iraq. After the first effort in Afghanistan and
the first two in Iraq, the military ally who assumed leadership right after the
coup quickly eliminated the civilian party ally from the government.
While force is the obvious choice for military officers contemplating a seizure
of power, the comparative advantage of party-based groups lies elsewhere, so
they choose nonviolent means of seizure when possible. They most often
initiate dictatorship via authoritarianization because it is relatively easy once
a party controls a democratic government. Authoritarianization also has a high

7
These ﬁgures for coups are larger than those in Table 2.1 because the universe of autocratic
initiations used for the table includes 26 percent that arose in territory that was either foreign-
occupied or not independent prior to the seizure of power that started an autocratic regime. The
ﬁgures on this page are for coups that end regimes in independent countries.

success rate, though we know of a few cases in which a democratically elected

president’s proposal to merge all parties into one (and thus authoritarianize)
was successfully resisted and a couple in which such a proposal led to the
elected government’s overthrow. The authoritarianization strategy is obviously
available only to groups that already control the government, however. Some
of the authoritarianizations included in the data we use happened during the
run-up to independence or during foreign occupation. The foreign occupier
may have promoted it because it ensured control by a group friendly to the
occupier’s interests, or foreigners may have cared more about extricating
themselves than about who ruled after their departure.
Insurgency, in contrast, has very high costs for participants compared with
either coups or authoritarianization. Insurgent leaders must sustain armed
forces for long periods compared with those engaged in coups. Insurgencies
last about a decade on average (Fearon 2007), whereas coups are typically over
within a day or two. Insurgent leaders must ﬁnd some way to fund their forces
since the government does not pay their salaries or buy their weapons – as it
does for soldiers involved in coups. The chances of death, disability, privation,
and loss of livelihood are much higher for participants in insurgencies than for
participants in coups or popular uprisings. Furthermore, insurgency is unlikely
to succeed. Insurgent groups secure outright victories against incumbents only
about a third of the time (Fearon 2007). By contrast, a bit more than half of all
coup attempts succeed (Powell and Thyne 2011). Probably for these reasons,
more insurgencies have ousted colonial powers and foreign occupiers than
domestic governments. In other words, insurgency is rarely the strategy of
choice; it is used when no other strategy is available.
Popular uprisings also sometimes initiate authoritarian regimes, but this is
not usually the intention of those who joined the uprising. When massive
demonstrations persuade an incumbent to resign, power is handed to interim
leaders to manage the transition. These leaders are supposed to oversee com-
petitive elections to choose a new government, but they may opt to consolidate
power in their own hands instead, initiating a dictatorship.
To summarize this section, the selection of means to overthrow governments
is largely instrumental. The method of seizure reﬂects the resources and capaci-
ties that comprise the comparative advantage of different kinds of seizure
groups. Most autocracies begin with coups because the command of men and
weapons gives coup plotters a great advantage relative to other would-be
seizure groups. Coups are the easiest and cheapest means available, unless the
group bent on establishing a dictatorship already controls government as the
elected incumbent.

before the seizure of power

Many are called to the vocation of dictatorship, but few are chosen. The
majority of seizure groups fail. Apart from groups that already control

government at the time of authoritarianization, members of would-be seizure

groups lead difficult lives. Some are jailed or killed by the governments they
seek to overthrow. They also risk their careers, income, and liberty. These
difficulties and dangers put strong evolutionary pressures on would-be seizure
groups. Those we observe after the initiation of dictatorship are highly selected
for traits that were useful in getting them to that point. Because this selection
pressure weeds out much of the natural diversity among those who aspire to
establish dictatorships, we can make some generalizations about those who
succeed.
Centralization of authority within the group of plotters contributes to dis-
cipline and secrecy, both of which help seizure groups survive in a hostile
environment. Nevertheless, successful seizure groups must usually also main-
tain a somewhat collegial internal organization in order to retain sufficient
membership.8 Seizure-group leaders cannot usually impose heavy costs on
those who abandon the cause. Out-of-power leaders have little ability beyond
moral authority and force of personality to enforce discipline. They usually
have no police apparatus at their disposal and few means to intimidate. Indi-
viduals who disagree with a leader can simply leave the party or coup conspir-
acy. It is estimated, for example, that 90 percent of those who joined the
Communist Party in the United States had left by the end of the first year.
Leaving the party or coup conspiracy entails few costs for the one who leaves.
In contrast, the loss of comrades can be quite damaging to remaining conspir-
ators. They may be left with too few allies for effective action, and their former
colleagues may turn them in to authorities. In short, if leaders demand too
much obedience and subordination from adherents, they may simply desert,
leaving the group too small to accomplish its goals – arguably the fate of all the
western communist parties rigidly controlled by the Comintern.
To maintain the minimum support needed for effective action, plot leaders
must therefore consult and be somewhat open to the ideas and interests of
members of the group. As a result, successful seizure groups usually enforce
norms of discipline and secrecy with respect to outsiders, but relationships
within the group tend to be relatively collegial before seizures of power.
Seizure groups need some support from the populace as well and thus must
be somewhat responsive to popular aspirations. A group bent on seizing power
does not need majority support, but it needs some. Regardless of the mode of
seizing power, groups need the “contingent consent” of the ruled (Levi 1997).
To attract the support they need, military seizure groups typically promise
things nearly all citizens want, such as public order, growth, and an end to

8
We build on Kenneth Jowitt’s (1975) argument about traits that help Leninist parties succeed at
different stages. During the chaotic period of power seizure, according to Jowitt, collegial
relationships within the party give cadres the autonomy and discretion to respond to changing
conditions, thereby contributing to survival. A different set of traits contributes to survival during
the post-seizure period.

corruption in their first broadcast informing the people of a coup. Other kinds
of seizure groups make equally attractive promises.
Seizure groups are more likely to succeed when they (1) articulate goals that
are attractive and intelligible to many ordinary people, such as peace, prosper-
ity, and the end of corruption (Jowitt 1975) and (2) choose a moment to seize
power when disgust with incumbents has spread through much of the popu-
lace. When both these conditions hold, seizure groups can attract broad,
though often temporary, public support. Elite groups as well as ordinary
citizens often support a dictatorial seizure of power simply because they want
to oust the old order.
During insurgencies, the evolutionary pressures on seizure groups are even
stronger. If the struggle is long and violent, the seizure group needs to draw
manpower and other resources from the population, and it cannot do this
entirely through coercion. Where the struggle to oust the old regime has been
prolonged and the seizure group has needed to maintain popular support for
some years and also to recruit fighters and extract resources from the popula-
tion, successful groups have generally provided more than promises. As Samuel
Popkin (1979) argues, change-oriented movements are more likely to succeed in
attracting mass support if their leaders can provide real benefits to those they
seek to mobilize. Since they have few material resources to hand out, they often
provide goods such as land redistribution and organization for self-defense in
the areas they control, along with individual benefits that require cadre labor
but little money.9 Literacy campaigns are a common strategy. These benefits are
valued by recipients and create loyalty that ideology and promises alone could
not elicit. Leninist parties in China and Vietnam relied on these strategies
during the years of insurgency (Johnson 1962; Popkin 1979). Before the seizure
of power, communist parties redistributed land to tillers. Only after they
controlled the government, military, and security services did they carry out
unpopular collectivizations of agriculture.
The point of this discussion of collegial relationships within potential seizure
groups and groups’ efforts to respond to popular demands is to highlight the
contrast with the change in power relationships that occurs after seizures of
power. The bargaining power of followers relative to leaders within seizure
groups and of ordinary citizens relative to seizure group members is much
stronger before the seizure of power than afterward.
The dependence of leaders on the led carries over into the immediate after-
math of the seizure, when the regime is still uncertain of its grip on power and
reliant on much of the ousted incumbent’s bureaucratic and military apparatus.
The seizure group’s responsiveness to citizens declines, however, as the new
leadership gains control over the implementing arms of government, such as the

9
See Mampilly (2011) for a study of how rebel groups provide public goods to citizens who live in
territory they control.

police and tax authorities. The longer-term fate of collegial leadership in

dictatorships is discussed in Chapters 4 and 5.

the morning after a seizure of power

Nonincumbent groups seize control of government in order to change public
policies and redistribute the fruits of control of the state.10 Even if personal
ambition motivates participants, groups that lead seizures also expect to be able
to advance public welfare by improving economic performance, restoring
order, and achieving other widely shared goals, though these expectations arise
more from inexperience and ideological commitment than realistic assessment.
The fruits to be redistributed include status, power, and opportunities as well as
material goods. Seizure groups expect to be able to redistribute away from the
individuals and groups they believe benefited from the status quo and toward
groups seen as having a legitimate claim to more, including themselves. The
anticipated redistribution sometimes includes private property but invariably
includes the direct benefits of controlling government. In short, many of the
same goals motivate authoritarian seizure groups as motivate out-of-power
parties competing in democratic elections.
Nonincumbent seizure groups, however, tend to lack experience in govern-
ment, and they also often lack a detailed plan for what to do if they succeed in
capturing power. Before the overthrow, seizure groups typically have
developed a thoroughgoing critique of the government they hope to oust, which
is needed to attract support, and they usually have a practical plan for how to
force it from office.
For military officers, this usually means careful, detailed plans for which
officers will order troops they command to surround or seize key government
facilities.11 Plotters have to include officers who actually command troops
stationed near facilities to be taken. Depending on the kind of defenses the
incumbent controls, plotters may need the cooperation of officers who com-
mand tank regiments or heavy artillery. They may have to coordinate with the
police and air force. If plotters include highest-ranking officers, they can min-
imize opposition to their plans by retiring officers loyal to the incumbent before
they attempt the coup. Much of the planning effort for coups goes into securing
the support of other officers so that the ruler being ousted will not have an
equally threatening armed force on his side (Potash 1969, 1980, 1996; Stepan
1971; Decalo 1976; Nordlinger 1977; Fontana 1987). One of the goals of
careful coup planning is to minimize the chance of bloodshed or, worse,

10
Dictatorships begun by the authoritarianization of a democratic government or elite rule changes
are exceptions.
11
There are of course exceptions. Sometimes barely planned coups succeed by luck. The 1981 coup
in Ghana led by J. J. Rawlings and the 1980 coup led by Samuel Doe in Liberia are examples of
barely planned coups.

protracted armed conflict (Potash 1969, 1980, 1996). Because of the obvious
and credible threat posed by tanks and armies drawn up around the presiden-
tial palace, coups have been accomplished without bloodshed about two-thirds
of the time.12
Party-based nonincumbent seizure groups have a more difficult task because,
unlike officers, they cannot order others to participate. They must usually
attract large numbers of other civilians to demonstrate, strike, or fight along
with them. The leadership’s main task beforehand is the persuasion and organ-
ization of supporters drawn from the initially uncommitted population. Great
effort has to be put into gathering the resources and attracting and training
enough militants to mobilize large numbers into the streets and keep them there
beyond the first few days, or to recruit and supply an insurgent army and keep
it alive and fighting for months or years.
In other words, for nonincumbent seizure groups – regardless of whether
they are led by a party or the military – the plotting, planning, and organizing
required for a successful overthrow usually take a lot of effort, attention to
detail, and strategic thinking. The urgent need to get this part of the plan right
limits the amount of attention that goes into planning what to do after the
overthrow, and the need for secrecy can limit the seizure group’s ability to
consult experts and interest groups. Moreover, sometimes the details about
who will rule post-seizure and what policies will be followed need to be kept
vague in order to attract key support. After the 1958 Iraqi coup, for example,
“Those in power lacked both experience and a shared ideology, with the result
that fundamental issues of principle, such as who was in command, and what
form of government and political system should be adopted, remained unre-
solved” (Farouk-Sluglett and Sluglett 1987, 52). Describing another Iraqi coup
ten years later, the same authors say that “the coup had been carried out by a
group of officers, each of whom felt equally entitled to play a, or the, key role in
government” (119).
For all these reasons, nonincumbent seizure groups usually achieve office
without a detailed policy plan. Those who carry out seizures of power typically
have idealized goals they expect to accomplish, and they may have decided to
initiate a few grand policy changes, such as nationalization of natural resources,
land redistribution, or economic liberalization. Their lack of practical experi-
ence in government limits their ability to make detailed plans for how to
implement their policies, however. Members of the seizure group may agree
on the general direction of policy changes, but not on specific strategies for how
to carry them out, elicit the cooperation of affected citizens, or impose the
policies on those who will inevitably be damaged by them. They usually have
not made decisions about priorities among desired changes, speed of

12
This estimate was tabulated using the Center for Systemic Peace’s “Coups d’Etat, 1946–2013”
data set (available at www.systemicpeace.org), which provides data on the number of deaths
associated with successful coups worldwide.

implementation, or the mix of coercion and cooptation that should characterize

their interaction with different groups of citizens. Differences of opinion about
these things cause conflict within the group.
After their accession, they face immediately all the economic and other
problems that helped make their seizure of power possible, as well as the urgent
need to make a large number of appointments and policy decisions quickly.
They have to adapt the group’s previous decision-making procedures to policy-
making on the fly. They have to either create human structures for implement-
ing policies and protecting themselves or make use of the perhaps disloyal
public employees hired during the ousted regime. And they also have to create
or adapt police and other agencies that interact with the public to prevent the
mobilization of opposition.
While engaged in a frenetic effort to deal with problems, they must also
devise methods for reaching policy decisions and find ways to resolve disagree-
ments among themselves. Thus, the early days of dictatorship tend to be
anxious, chaotic, and full of uncertainty.13

post-seizure organization
Regardless of how seizure groups achieve power, the starting point for
assigning tasks and making decisions is the group’s preexisting organization
and its norms and procedures for choosing leaders, making choices, and
maintaining internal cohesion. Preexisting organization and norms vary by
kind of group. In this section we discuss characteristic differences among
seizure groups.14
Coups tend to lead to the occupation of the top offices in the new dictator-
ship by officers, regardless of what civilian allies might have hoped. Because of
strong hierarchical norms within the military, plotters often choose the top-
ranked officer who cooperated with plotters as regime leader because other
officers would balk at being led by a lower-ranked officer. Higher-ranked
officers who opposed the coup tend to be retired (or jailed, exiled, or killed)
in order to prevent potential destabilizing challenges to the chain of command.
After the coup led by Qaddafi (a captain), 430 of Libya’s 600 officers were
retired, jailed, or posted abroad (Blundy and Lycett 1987, 64).

13
Seizures led by incumbents face fewer challenges of course. Pushing rule changes through
legislatures is less dangerous than insurgency or staging coups. Incumbent seizure groups may
have ﬁlled the army and police with supporters before authoritarianization, eliminating one of
the most urgent tasks facing other seizure groups. Seizure groups that authoritarianize an elected
government also have more government experience than other seizure groups, and they can
typically count on the loyalty of existing bureaucrats (many of whom they appointed). In
addition, the gradual and subtle nature of some authoritarianizations can leave citizens without
a focal point around which opposition might coalesce.
14
We do not discuss monarchies in this section because foreigners handed control over to nearly all
authoritarian monarchies that have ruled since 1946.

The clear formal hierarchy of command in military seizure groups tends to

centralize decision-making within an inner circle of highly ranked officers. The
wide dispersion of arms and command of troops across the officer corps,
however, means that before making key decisions, military dictators must
consult with other officers or risk a coup led by their colleagues. To deal with
that basic feature of military rule, regime leaders include as many officers as
they believe are needed to ensure the cooperation of all parts of the armed
forces in the ruling inner circle. For highly professionalized militaries, this
means a junta made up of the commanders of the branches of the armed forces
and often also the commander of the national police. In more factionalized
militaries, however, the ruling military council may include many more officers
in order to ensure ethnic representation or to include all officers who command
troops. In some, even representatives of noncommissioned officers and soldiers
are included.
Uniformed members of the inner circle serve as representatives of their
subordinates and thus facilitate consultation between the leader and the forces
that serve as his primary support base. Officers in the inner circle are expected
both to advocate for policies that serve their subordinates’ interests or reflect
their opinions and to ensure subordinates’ loyalty to the leadership. The pol-
icies of dictators established by military seizure groups thus tend to reflect the
views of their officer corps, leading to increased spending on the military (Bove
and Brauner 2014), pay raises, immediate promotions of those involved in the
coup and sometimes all officers, violent responses to minority demands for
greater rights and autonomy (Nordlinger 1977), and the adoption of policies
favored by other officers regardless of whether the dictator himself shares
their ideas. For example, Mathieu Kérékou in Benin and Didier Ratsiraka in
Madagascar introduced socialist rhetoric and policies, though neither had
displayed much earlier leftism (Decalo 1979, 1986).15
The need to retain the support of other officers also constrains the dictator’s
discretion over personnel appointments. All dictators try to appoint loyalists to
key posts, but the need to conform to military promotion norms in order to
avoid alienating other officers limits the discretion of military dictators. Even
after the murder of senior commanders during coups, “coup participants still
[keep] up the normal command relationships between each other” (Luckham
1971, 31). As a result, military dictators often lack the full control of member-
ship in the dictatorial elite enjoyed by some other dictators because they can
only choose appointees from among candidates with the right rank and senior-
ity.16 Policy implementation can also be undermined because appointments to

15
Decalo (1986) describes Ratsiraka as a “bourgeois technocrat . . . with no previous ideological
convictions” (129).
16
For multiple examples, see Potash’s (1969, 1980, 1996) careful descriptions of decision-making
in several Argentine military dictatorships.

ministries and state agencies depend on the promotion and retirement norms of
the armed forces rather than substantive competence (North 1983, 273).
At the time of the seizure, the military usually lacks an organized civilian
network capable of mobilizing, or even reaching and gathering information
from, ordinary citizens. Officers’ training to obey and to give orders exacer-
bates this deficiency, as does their lack of training or experience in persuasion,
bargaining, and compromise.
Civilian seizure groups are more heterogeneous than military seizure groups.
Some are armed while others are not. We first describe those that seize power
via armed insurgency. Successful insurgency leads to control of the dictator-
ship’s inner circle by the leadership of the rebellion. Military and civilian roles
tend to be fused in insurgencies, and that often persists in the post-seizure inner
circle.
Insurgent seizure groups share some (but not all) of the characteristics of
military seizure groups. Arms and command over troops are widely dispersed
within insurgent groups, but military promotion norms do not constrain the
leader’s discretion over appointments, increasing his power relative to others in
the inner circle. Successful insurgent groups develop strong discipline, but
where regional commanders have led autonomous insurgent forces in different
parts of the country, intense factionalism can develop. That is, discipline may
operate within region-based factions, but not necessarily across them. The
various factions will then need to be represented in the post-seizure dictatorial
elite for the same reason that ruling military councils include many officers
when the military is factionalized. Examples of factionalized insurgencies
include the multiple armed forces that cooperated with difficulty during the
Mexican revolution and the several separate regional Khmer Rouge armies in
Cambodia, each led by its own party/military commander before the fall of
Phnom Penh.
Since successful insurgency requires the maintenance of organized networks
for recruiting fighters, extracting resources, training soldiers, and disciplining
dissidents, insurgent groups often have much more of the organizational struc-
ture needed to reach most citizens and to incorporate them into the seizure
group’s project than do other seizure groups. Internal party institutions in
dictatorships brought to power by insurgency thus tend to be stronger relative
to society but weaker relative to the dictator than those of parties that achieved
power via authoritarianization or popular uprising. Because they have often
developed internal security services before the seizure, they can very quickly
consolidate control over the population once the ruling group has control of the
resources of the state.
Insurgents who have achieved power usually purge and reorganize, or
replace, the armed forces and security services of the ousted regime. They
may have sufficient trained soldiers and cadres to replace the incumbent army
and part of the incumbent administration rather than having to retrain
and monitor (or compromise with) them, as most other seizure groups do.

Former insurgents thus have less need to coopt opposition leaders and more
ability to respond to opposition with coercion than do dominant parties that
achieved power by nonviolent means. In short, their ability to penetrate both
state and society, and thus to implement dramatic policy shifts, initially tends to
be much greater than that of military or unarmed party-led seizure groups.
Because of the great costs and risks associated with joining insurgencies,
those who become their leaders and most dedicated militants tend to have
commitments to extreme ideologies or policies (DeNardo 1985). Only individ-
uals who want radical changes would shoulder such costs.17 Consequently,
insurgent seizure groups often want to impose unpopular policy changes – for
example, collectivization of agriculture or the exclusion of distrusted classes or
ethnicities from government jobs. Because of the prior development of extensive
cadre networks and security agencies, they may also have greater capacity to
carry out such changes than do military or unarmed party-led seizure groups.
Unarmed civilian seizure groups, in contrast, tend to be more loosely organ-
ized. Intensive organization is not needed to win a democratic election, the
precondition for authoritarianization, which is the most common way for
civilian seizure groups to achieve power. Nor is much organization needed to
mobilize people into the streets for a few days or weeks. There are a few well-
known cases in which well-organized groups seized power by mobilizing ordin-
ary people for a popular uprising (e.g., the Iranian revolution), but most of the
time popular uprising is a tactic chosen by groups that lack the organization
and resources to maintain insurgency or a multiyear electoral campaign against
an authoritarian incumbent. In short, unarmed civilian seizure groups are
usually loosely organized, partly because they did not need high internal unity
in order to achieve power, but also because they so often include many different
parties or groups that joined forces only to win an election or oust a previous
incumbent.
Authoritarianization usually results in civilian leadership of the new dicta-
torship, with power centered in the hands of the incumbent. It involves a
change in the relationship between the ruling party and the opposition, but
often not in the composition of the government’s inner circle. Parties that
achieve control via authoritarianization very often coopt the leaders of other
parties, thus reducing societal opposition at the expense of their own discipline
and ideological coherence. They may also coopt military ofﬁcers, but they
rarely bring them into the regime’s inner circle.
Paradoxically, loose party organization gives leaders more autonomy rela-
tive to the rest of the inner circle than military dictators initially enjoy. Both
electoral and revolutionary parties have rules for choosing leaders, making
strategy choices, and delegating tasks. Leaders of out-of-power parties can

17
A seminal article by Kalt and Zupan (1984) developed the logic linking political activism with
extremism in democracies. The logic is even more compelling when activism is physically
dangerous.

usually be removed by followers. Relative bargaining power often changes after

the seizure of power, however, and the members of the inner circles of dictator-
ships that achieved power via popular uprising or authoritarianization may
have limited ability to constrain the dictator's policy choices because they lack
unity and arms.
At the same time, the weak organization of such parties can also mean that
the dictator cannot restrain or discipline party subordinates, which undermines
policy implementation and efforts to limit corruption. At the beginning, such
groups usually have very broad citizen support, but linkages with ordinary
citizens tend to be superficial and ultimately depend on delivering benefits. In
contrast to the military, such civilian seizure groups usually have little coercive
capacity but a great deal of experience in persuasion and compromise. They are
good at coopting opposition elites, but bad at delivering on their promises to
ordinary citizens, which requires a disciplined and honest cadre of officials to
implement reforms and distributive decisions.
These characteristics of the seizure group set in motion political processes
that define the dictatorship initially. These features in turn affect who can
become the leaders of the new dictatorship, how centralized power is within
the dictatorial inner circle on the morning after the seizure, and how the new
elite responds to citizen demands and opposition.

conclusion
In this chapter, we have provided some basic facts about how dictatorships
begin and who leads them. We also described the immediate aftermath of
different kinds of seizures of power. We noted that most dictatorships replace
earlier autocracies and that the regimes they oust are often incompetent, dis-
honest, or both, as well as repressive. Since World War II, military coups have
established more dictatorships than other means of seizing power. Authoritar-
ianization by a democratically elected incumbent, insurgency, and foreign
imposition are the other common ways of establishing dictatorship.
Groups of military officers or civilian-led parties install most dictatorships.
We see decisions about how to seize power as strategic and opportunistic. That
is, military officers have a comparative advantage in using violence and threats
of violence to take over governments, so that is what they usually use. Incum-
bent elected parties that decide to prevent future electoral defeats can most
easily sustain their rule by using their control over the legislature and courts to
change formal or informal rules in ways that prevent fair competition. Insur-
gency is a very costly and difficult way to gain political control, so it is only used
when other means are unavailable, as, for example, against foreign occupiers.
The initiation of new dictatorships by insurgents has become less frequent as
foreign occupation has become less common and as more incumbent dictator-
ships have allowed semi-competitive elections, giving opponents a less costly
way to challenge them.

Because of the difﬁculty and risk of trying to force out an incumbent

government, groups that seek to do so focus their efforts on organizing the
details of the physical overthrow, often at the expense of planning what to do
post-seizure. In order to attract sufficient support, they may need to promise
different things to different people or keep their policy goals vague. The need
for secrecy during plotting means they cannot usually choose important cabinet
officials, develop detailed implementation plans, or consult experts for policy
advice before the seizure of power. Sometimes they have not even decided who
will lead the new government. The lack of detailed planning about what to do
post-seizure tends to make the first months after forcible impositions of dicta-
torship chaotic. Power struggles occur between supporters of different policies
and different individual officeholders. Strategic skills not previously displayed
in public and unforeseen events affect who wins post-seizure power struggles,
which add to the difficulty of predicting what kind of policies will emerge after
the establishment of a new dictatorship. It may take time for observers and even
participants to figure out what kind of regime is being created.

What Do We Know about Coups?

In 1952, the arbitrary and corrupt Egyptian monarchy ended. Troops com-
manded by a group of young officers who had named themselves the Free
Officers occupied the army’s general headquarters and arrested twenty of the
highest-ranking officers – the king’s strongest military loyalists. Tanks simul-
taneously surrounded the broadcasting building, the telephone exchange, the
railroad station, and the airports. This classic coup took only a few hours. Two
guards were killed. No one else resisted. The Free Officers did not arrest the
king or his ministers. Three days later, they asked the king to abdicate, and he
left the country, ending the monarchy (Haddad 1973, 21–23).
King Farouk had ruled “comfortably for sixteen years and would have
probably ruled longer if revolutionary action had depended on the people’s
initiative,” according to George Haddad (1973, 10). The Free Officers’ leader,
Colonel Gamal Abdel Nasser, stated that the army was “‘the force to do the
job’ because it had more cohesion, its men trusted each other . . . and it had
‘enough material strength at its disposal to guarantee a swift and decisive
action’” (Haddad 1973, 15). The Free Officers had been plotting since at least
1949. Soldiers blamed Egypt’s poor performance in the Palestine War of
1948 on senior officers, the king, and his cronies, which alienated much of
the officer corps from the monarchy. A scandal involving the corrupt procure-
ment of defective arms had become public and implicated the king. In short,
much of the officer corps felt betrayed by the king and the officers he had
promoted to the highest ranks, but most ordinary Egyptians appeared content
with the monarchy. When violent conflict with the British over control of the
Suez Canal led to a destructive riot in Cairo and a cabinet crisis, the Free
Officers decided the time to end the old regime had come (Haddad 1973,
7–11). The plotters struck during a high point of cabinet disorganization and
popular anger with the government’s Suez policy in order to reduce the chances
of bloodshed, effective opposition, and failure.

Coups, as noted in the previous chapter, are the most frequent means of
initiating new dictatorships and also the most common way of ending them,
as we show in later chapters. Yet we lack a good general understanding of why
they happen. When insurgents seize control of a government or an elected
ruling party authoritarianizes a democracy, it seems self-evident that providing
benefits for themselves and their supporters motivates their actions. Analysts
debate the motivations behind military seizures of power, however.
Some scholars have viewed officers as defending the interests of economic
elites, and thus as more likely to seize power if popular mobilization becomes
threatening to elites or if politicians responsive to popular interests propose
redistribution (e.g., O’Donnell 1973; Acemoglu and Robinson 2001). In this
view, officers are allied with economic elites and serve as their agents rather
than pursuing their own interests, not only when plotting coups but also while
they control governments. If this view is correct, officers’ comparative advan-
tage in overthrowing governments gives economic elites a powerful weapon to
use in bargaining with incumbent governments whether democratic or
autocratic.
Scholars such as Nordlinger have challenged this idea, however, noting that
“the personal interests of officers – their desire for promotions, political ambi-
tions, and fear of dismissal” (1977, 66) have motivated many coups. Nordlin-
ger also observes that although some military-led governments choose policies
that benefit existing economic elites, others expropriate private property and
pursue redistributive policies. Svolik (2012) has proposed an interesting syn-
thesis of these two views: he suggests that incumbent political leaders increase
the funding of their military agents when organized mass opposition threatens
their political control, but that then the military uses the additional resources to
seize power for itself, rather than serving as a loyal agent.
In this chapter, we examine what currently available evidence can tell us
about soldiers’ seizures of power. We first discuss the various ways that coups
are used to change the political status quo and why we might expect these
different uses to be associated with different officer motivations, political
contexts, or social forces. Next, we compare the conditions associated with
coups that end democracies and those that replace one dictatorship with
another. At issue is whether popular mobilization against economic elites or
incumbent political leaders increases the likelihood of military intervention.
Claims that the military intervenes to prevent redistribution usually assume
implicitly that coups oust democratic governments sympathetic to popular
demands, which leads to the expectation that the circumstances associated with
coups differ for democracies and dictatorships. In the final empirical section, we
retest Svolik’s argument about the association between popular mobilization
and coups. We find no relationship between various measures of mobilized
popular opposition and regime-change coups. This finding indicates that coups
have little to do with elite fears of redistribution or with the buildup of military

resources in response to popular mobilization. Instead, we ﬁnd that ofﬁcers’

grievances are associated with coups.

coups for various purposes

Investigating the conditions associated with coups requires some preliminary
discussion because of the various kinds of change accomplished via coup.
Coups are military officers’ instrument of choice in several different situations,
but here we are interested only in those coups that bring new dictatorships to
power. Regime-initiating coups include the highly visible events that replace an
elected civilian leader with someone who wears a uniform and carries a
weapon, but coups also oust incumbent dictatorships. Such coups replace one
dictatorship with another. The coups that ousted monarchies in Egypt (1952),
Iraq (1958), and Libya (1969) are examples. Coups against incumbent
dictatorships can oust any type of preexisting autocratic ruling group, even
those controlled by rival ethnicity-based military factions. These coups result in
new autocratic regimes.
Officers also use coups as a means of changing leaders in ongoing dictator-
ships, however, especially those led by the military. These leader-shuffling
events are not military seizures of power because they replace one military
dictator with another from the same ruling group; the basic formal and infor-
mal rules remain unchanged. The coups in Argentina during the early 1970s
that replaced one general with another are examples. They did not begin
(or end) the dictatorship, which lasted from 1966 to 1973 and ended in an
election. Each coup replaced the junta leader but did not replace the group – in
this case top military officers – that could select leaders and make key policy
decisions. Other examples include the 1980 and 1984 coups in Mauritania,
each of which sacked one member of the Military Committee for National
Salvation and replaced him with another allied officer. We do not treat these
leader replacements as military seizures of power because the same group of
elites within the military continues to rule before and after the leader change.
We emphasize the difference between regime-initiating coups and leader-
shuffling coups because available coup data sets do not distinguish the two, and
most analysts have used these data without disaggregating.1 In what follows,
we show that lumping together leader-shuffling coups and those that initiate
new dictatorships can blur our understanding of the causes of regime-initiating
coups. To more accurately model the process by which a military faction ousts
an incumbent government that did not include members of the plotters’ faction,
we exclude reshuffling coups from the sample used in the first tests below and
focus only on regime-initiating military seizures of power. We explore the
implications of doing so for previous research as well as for our own analysis.

1
Aksoy, Carter, and Wright (2015) are an exception.

preconditions associated with regime-change coups

We begin by examining two kinds of regime-initiating coups: those that replace
democracy with dictatorship and those that replace one autocracy with
another. We examine these separately to see if they have different causes. The
first sample includes only incumbent dictatorships. The second sample includes
only incumbent democracies. The second analysis is similar to research on the
causes of democratic breakdown. We treat authoritarianizations by the incum-
bent democratic leadership – or what others call power grabs or autogolpes by
the democratic incumbent – as right-censored events.2
We have good measures of popular mobilization (protest, violent protest,
and civil conflict), which are needed to test arguments about the effects of
widespread popular opposition on the likelihood of military seizures of power.
A finding that popular participation in protests or civil conflict increases the
likelihood of coups would support the view of officers as agents of civilian elites
since widespread popular mobilization might be expected to lead to elite fears
of redistribution or loss of political control.
We lack the kinds of measures of soldiers’ specific grievances needed for
careful tests of Nordlinger’s argument about soldiers’ reasons for seizing
power. As an approximation of the latter, we use information about officers’
past experiences and measures of ethnic representation that we believe might
affect soldiers’ grievances.
We first examine the factors that prior studies have most often linked to
military intervention: level of development, oil wealth, growth, international
and civil conflict, protest, and ethnic diversity.3 We also include indicator
variables to investigate the possible effects of earlier political experience. The
first identifies countries that were democratic before the current dictatorship (in
the sample of incumbent dictatorships); the second identifies countries that
had previous experience with military rule. Earlier studies have found that a
history of past military intervention increases the likelihood of coups. To
account for unmeasured factors that vary by country, we estimate a
model with country-level random effects and standard errors clustered by
country.4 All specifications include decade fixed effects and a logged measure
of duration dependence.

2
Authoritarianization is the other way democracies are replaced by dictatorship. We treat them as
right-censored in order to avoid inadvertently treating them as surviving democracies.
3
Level of development is measured as the log of GDP per capita, lagged one year. Oil wealth is
measured as the log of oil income, lagged one year. Growth is lagged one year. These variables come
from the Ethnic Power Relations data set (Wimmer, Cederman, and Min 2009). Data on inter-
national and civil conﬂict are lagged dummy variables for conﬂict exceeding 1,000 battle deaths in a
calendar year, from Themnér and Wallensteen (2014). Data on violent and nonviolent protest
movements are lagged one year, from the NAVCO2 project (Chenoweth and Lewis 2013).
4
Variables for the manner in which the autocratic regime seized power, used in models 2 and 3, are
from our data, http://sites.psu.edu/dictators/how-dictatorships-work/. They do not vary

For the sample of incumbent dictatorships, we also examine two character-

istics of the incumbent regime that could influence the likelihood that officers
develop grievances against regime elites: how the dictatorship came to power in
the first place and the ethnic composition of the officer corps and government.
We include the way the incumbent dictatorship achieved power as a proxy
indicator of the congruence of interests between military officers and ruling
elites. We expect more shared interests after insurgent seizures of power
because new rebel governments often replace the existing officer corps with
rebel officers.
We also investigate the effects of ethnicity because of literature linking ethnic
exclusion with different forms of political instability, such as coups and civil
wars. For example, Roessler (2011, 2016) argues that leaders who use ethnic
criteria to exclude rivals within their initial support coalition may decrease the
risk of coups but increase the risk of ethnic insurgency. To investigate the
relationship between ethnic inclusion and coups, we include variables that
measure ethnic fractionalization in the society and the share of the politically
relevant ethnic population that is excluded from participation in the incumbent
government.5 This measure of ethnic exclusion focuses on the groups that share
executive political power, not the military. To measure ethnic inclusiveness in
the military, we include an indicator variable for whether the officer corps is
ethnically heterogeneous, that is, whether it includes members of all larger
ethnic groups. This variable, from our own data collection, exists only for the
sample of incumbent dictatorships.
Figure 3.1 shows the results for models of regime-initiating coups against
incumbent dictatorships. The shapes (diamond, square, and triangle) indicate
the point estimates produced by different models, while the lines represent
90 (thick) and 95 (thin) percent confidence intervals. The first specification,
shown as diamonds, excludes the seizure-of-power and ethnicity variables. The
second, with square point estimates, adds the seizure-of-power variables, while
the third specification (triangles) adds the ethnicity variables. Because the
estimator is nonlinear, the estimates are not directly comparable across
variables.
In these models, the only structural variables consistently associated with
military seizures of power are the level of development and population size: the
military is less likely to initiate a (new) dictatorship in wealthier countries, as
also found in previous research (Londregan and Poole 1990). We find no
consistent association between military seizures of power and oil income,

substantially within most countries, so we use a country random-effects estimator. Of the 118
countries, 65 have no variation in coup seizure, 100 have no variation in foreign seizure, and
97 have no variation in rebel seizure.
5
Ethnic exclusion and fractionalization data, both lagged one year, are from version 3 of the Ethnic
Power Relations data (Wimmer, Cederman, and Min 2009).

Incumbent dictatorships

Duration (log)
GDP pc (log)
Oil pc (log)
Population (log)
Civil conflict
Int'l conflict
Violent protest
Non-viol. protest
Seizure

Prior democracy
Prior military
Growth
Coup
Rebellion
Foreign
Ethnic frac.
Excluded pop.
Hetero military

–2 –1 0 1 2
Coefficient estimate

Base model Add seizure Add ethnicity

figure 3.1 Regime-initiating coups against incumbent dictatorships.

economic growth, civil or international conflict, and protest, all of which have
been suggested as causes of coups by other analysts. The lack of relationship
between protest and coups means that our findings fail to support arguments
that officers seize power in order to defend the interests of economic elites
threatened by popular mobilization. Prior regime type, however, does appear to
be correlated with coups: dictatorships in countries that had earlier experienced
military rule or democracy are more likely to be ended by coups than dictator-
ships in countries previously ruled by civilian dictatorships.
Next, we consider some circumstances that might contribute to the develop-
ment of officers’ grievances against incumbent rulers. We expect the makeup of
the dictatorship’s inner circle and its relationship with military leadership to
affect the congruence of interests between rulers and officers, and thus the
likelihood of regime-change coups motivated by officers’ grievances. We use
the way the dictatorship first achieved power (coup, rebellion, foreign

imposition, and nonviolent) as an indicator of the military’s relationship to the

dictatorial elite. The excluded category of seizure includes the less violent means
of achieving dictatorial power (authoritarianization, elite rule changes, and
popular uprising).
The variables for how the dictatorship initially seized power do not vary
over time within individual dictatorships. Instead, they identify a characteristic
of the seizure group measured prior to the initiation of dictatorship. This last
point is important because the dictator cannot change this characteristic in
order to forestall coups. That is, these characteristics of the seizure group are
exogenous with respect to later attempts to prevent coups.
The coefficient labeled “Coup” estimates the chances of a regime-change
coup in year N against dictatorships first established by coup relative to the left-
out group, dictatorships that achieved power in less violent ways. We expect
military dictators to consult more with other officers than do other dictators
and to share many of their interests. Consequently, we might expect fewer
grievances to develop in the officer corps during periods of military rule, leading
to fewer regime-change coups. On the other hand, consultation does not
eliminate factionalism or personal ambition, other causes of coups. The officer
corps of newly independent countries were especially prone to factionalism for
the reasons described later in the discussion of ethnic heterogeneity. As a result
of these cross-cutting influences, we may see no difference in coup susceptibility
between regimes initiated by coup and those initiated by less forceful means.
In contrast, we expect no cross-cutting influences in regimes initiated by
rebellion. Successful rebellion leads to political control by the insurgents and
often the replacement of the preexisting army by the rebel army. Consequently,
the dictatorial elite and the new army share interests, ideas, experiences, and
often friendship, which should reduce the seriousness of differences between
officers and rulers and hence officer grievances.
Foreign imposition of dictatorship also often goes along with replacing,
retraining, and reorganizing both the military and the internal security forces
of the occupied country. These strategies aim at coup-proofing. The occupier
may also station troops in the country for some years after the formal occupa-
tion has ended. For example, Vietnam kept troops in Cambodia for ten years
after its replacement of the Khmer Rouge government in 1979. We expect
military reorganization, improved internal security, and the stationing of for-
eign troops to deter coups against foreign-imposed dictatorships.
We test these expectations in the second model in Figure 3.1. We find that
coups are unlikely to unseat dictatorships that come to power via rebellion or
foreign imposition. The likelihood of regime-change coups in military-led dic-
tatorships (initiated by coup) is not statistically different from their likelihood
in regimes that came to power by more peaceful means. These results suggest an
association between officers’ grievances and regime-change coups.
Once we account for how a dictatorship seized power, the estimates for
earlier democratic experience or military rule lose statistical significance, in part

because military regimes differ systematically from other dictatorships in how

they end (Geddes 1999). These earlier regime endings are the same events used
to code the current regime’s method of power seizure. Our estimates thus
suggest that some of the explanatory power attributed by other research to
earlier experience with military rule may simply reflect the fact that researchers
have yet to examine systematically how characteristics of different autocratic
seizure groups affect the likelihood of later coups.
The final specification in Figure 3.1 adds ethnicity variables to the model:
ethnic fractionalization of the population (ethnic fractionalization), the share of
the ethnically relevant population excluded from executive political power
(excluded), and a measure of whether the officer corps includes members of
the larger ethnic groups in the country (heterogeneous military). Ethnic fractio-
nalization and the exclusion of representatives of significant ethnic groups from
political influence do not correlate with military seizures of power against
dictatorships. An ethnically heterogeneous officer corps, however, is associated
with a higher likelihood of regime-change coups.
A heterogeneous officer corps means that men from several ethnicities have
enough control over armed men to oust a government. Most colonial govern-
ments tended to recruit soldiers from relatively underdeveloped regions, with
the result that the first officers at independence often came from one region and
one or a few ethnic groups. Political leaders at independence usually came from
more developed parts of the country, and thus from different regions and
ethnicities (Horowitz 1985). When post-independence leaders took over
recruitment of officers, they usually either recruited from their own regions
(to ensure officer loyalty) or from all regions – to try to create a truly national
army. This recruitment sequence led to a post-independence officer corps that
was heterogeneous but nevertheless stratified by ethnicity. Top officers often
came from one area and lower-ranking officers from others. As a result, we see
quite a few coups carried out by junior officers from one ethnic cluster against
ruling senior officers from another. These coups reflect the grievances of junior
officers over slow promotions and their regions’ share of federal spending.
We view the pattern of officer recruitment before and after independence as
the underlying reason for the association between regime-change coups and an
ethnically mixed officer corps. We note that coups carried out by officers from
heterogeneous armed forces were much more likely during the years when the
forces of newly independent countries still reflected colonial recruitment
(1957–1974) than afterward. Indeed, we show (in the replication files) that
regime-change coups in countries with heterogeneous militaries were most
common in the first five years of a dictatorship’s existence and quite rare
afterward. This finding adds to the evidence that officers’ grievances can
motivate coups.
These empirical tests improve on the standard model of coups by adding
variables that reflect characteristics of the autocratic leadership group that
bargains with the military, as well as characteristics of the officer corps itself.

table 3.1 Area under the ROC for models of regime-change

coups in dictatorships

Model Area under the ROC

Structural 0.748
Structural + Seizure Type 0.771
Structural + Seizure Type + ethnicity 0.776

As shown in Table 3.1, the model with standard variables has an area under the
ROC – a measure of in-sample predictive power – of 0.748. Adding measures
of the seizure type – a proxy for the relationship between the incumbent ruling
group and the military – improves model accuracy, with an area under the
ROC of nearly 0.771. The final model that adds ethnic variables again increases
model accuracy, with an area under the ROC of 0.776. The difference in
predictive accuracy between the first and third models is statistically significant
at the p<0.05 level, indicating a substantial improvement in model accuracy.
Equally important, the variables that capture the relationship between the
incumbent leadership group and the military are exogenous to dictatorial
attempts to retain power because they measure a characteristic that cannot be
altered after the seizure of power, namely, how the incumbent ruling group
seized power in the first place.
Figure 3.2 shows the results from a similar series of tests that investigate the
factors associated with military coups against democratic leaders. In the base-
line model, shown as diamonds, we find that the only significant correlates of
such coups are the level of development, the Cold War (1947–1989) period,
and whether the country had experienced an earlier military intervention.6 As
with coups that end dictatorships, there is no consistent association between
coups that end democracies and oil income, conflict, or protest. Instead, the
findings are consistent with those who argue that poor countries are likely to
experience coups (Londregan and Poole 1990) and that democracies preceded
by military rule are more vulnerable to coups (Cheibub 2006).
Next, we add to the model a set of variables that measures how the demo-
cratic regime came into existence. A large number of post-1946 democracies
(31 percent) began with a competitive election (election) held by an outgoing
dictatorship. An additional 22 percent of democracies arose in the aftermath of
popular uprisings (uprising), while roughly 20 percent were formed by armed
force, including those initiated after coups (coups), rebellions, or foreign inva-
sions (rebel/foreign). The remainder were initiated by elite rule changes such as
suffrage expansions, which is the reference category.

6
We also ﬁnd that longer-lasting democracies are less likely to end in coups. Without modeling
potential nonproportional hazards, we cannot interpret democratic duration as “consolidation.”
See Svolik (2008) on modeling consolidation of democracy using a split-population estimator.

Incumbent democracies
Duration (log)
GDP pc (log)
Oil pc (log)
Population (log)
Civil conflict
Int'l conflict
Violent protest
Non-viol. protest
Seizure

Prior military
Cold War
Growth
Coup
Rebel/Foreign
Election
Uprising
Ethnic frac.
Excluded pop.

–2 0 2 4
Coefficient estimate

Base model Add seizure Add ethnicity

figure 3.2 Coups against incumbent democracies.

We add these variables to keep the tests as similar as possible, but we do not
interpret them the same way as in dictatorships. The events that usher in
democracy often do not determine the identity of subsequent leaders and thus
do not have the same implications for the later relationship between political
leaders and military ofﬁcers. That is, when coups or popular uprisings lead to
democracy, the transitional leadership that takes power initially oversees a fair,
competitive election that chooses who rules. Elite rule changes that initiate
democracy also lead to fair, competitive elections. Since leadership during the
democratic regime is chosen in elections regardless of how the democracy was
initiated, we would not expect particular modes of initiation to be associated
with later military grievances. Foreign or rebel initiation might, however, deter
coups by reducing the likelihood that they would succeed.
Adding these variables does not change the ﬁndings for level of development
and earlier experience with military rule. Democracies initiated after armed

seizures of power by rebel groups or foreigners are less likely to be ousted in a

coup than those initiated by elite rule changes (the reference category), though
this relationship may be spurious. One-third of the foreign-initiated democra-
cies arose after US liberation of Western Europe during World War II. None of
the foreign-initiated democracies later fell to a coup. This probably has more to
do with the past history and economic performance of these countries than with
how democracy began. Initiation by coup, election, or popular uprising has no
effect on the likelihood of succumbing to a later coup, as expected.
In the final model, we add ethnicity variables: ethnic fractionalization of the
population and the share of the ethnically relevant population excluded from
executive political power. We do not have data from democracies on whether
the military is ethnically heterogeneous, so we cannot test this idea for the
sample of democracies. While ethnic fractionalization per se is not associated
with coups, ethnic political exclusion increases the likelihood of coups against
democracies.7 This suggests that the military is more likely to intervene against
incumbent democrats when executive power is not shared widely across ethnic
groups. This finding is consistent with the earlier argument that coups are more
likely when dissatisfaction with the incumbent is widespread. It also seems
likely that when a large part of the citizenry lacks representation in the execu-
tive, many officers also lack representation, leading to grievances.

inequality and coups

In this section, we test Svolik’s (2012) argument that threats from organized
societal opposition against incumbent dictators increase their allocations to the
military, which in turn heightens the military’s ability to seize power for itself.
Officers, in this view, are most likely to replace the incumbent regime when
political elites face the threat of “mass, organized, and potentially violent
opposition” (2012, 125). Figures 3.1 and 3.3 show the effects of violent and
nonviolent protest as well as civil conflict on regime-change coups. These
coefficients are not close to standard levels of statistical significance. These
results thus fail to support the idea that widespread popular opposition
increases the likelihood of military intervention, either directly or through the
mechanism Svolik suggests.
Svolik’s own tests, however, use income inequality as a proxy for mass,
potentially violent opposition rather than direct measures of protest. So, our
next tests assess the effect of inequality on coups. Svolik expects popular
challenges to be most threatening at middle levels of income inequality. He
then shows that middling inequality predicts coups in a data set that combines

7
Adding seizure and ethnic variables to the model of coups against incumbent democracies does
not substantially improve the predictive accuracy of the model, in part because the baseline model
has high in-sample predictive power, with an area under the ROC of 0.88.

Entry Exit
Gini
2
Gini
GDP pc
Growth
Fuel exp
Trade open
Cold war
Dem. nbr.
Ethnic frac
Int'l conflict
Civil conflict
Prior mil. leader
Log time
Mil. leader

–4 –2 0 2 –4 –2 0 2
Coefficient estimate

All coups Seizure Reshuffle

figure 3.3 Causes of coups in dictatorships, by coup type.

leader-shufﬂing and regime-change coups. It is possible, however, that coups

that initiate new dictatorships might be caused by factors different from those
associated with leader-shuffling coups, particularly if the latter serve as a
mechanism to maintain leader accountability to other elites. Some military
regimes use reshuffling coups to replace leaders much the way parliamentary
regimes rely on votes of no confidence. We investigate the possibility that the
causes of regime-change coups differ from those of reshuffling coups in the later
analysis.
To do this, we first verify the main finding from Svolik’s analysis. The data
he used separately identify coups that result in the initiation of an individual
officer’s tenure as dictator and coups that remove a dictator from power
(meaning that reshuffling coups are included in both samples, along with
regime-change coups). We then further disaggregate his data to distinguish
regime-change coups from leader-reshuffling coups. We separate his leader-
entry coup data into two groups: (1) those that correspond to military seizures
of power, i.e. regime changes, and (2) reshuffling leadership changes in ongoing
authoritarian regimes. Roughly 47 percent of Svolik’s entry coups are military
seizures of power, while 53 percent replace leaders in ongoing dictatorships.
For coups that involve the exit of dictators, 39 percent initiate regime transi-
tions. The other 61 percent are leadership reshuffles among members of

the same ruling group. Altogether, the majority of coups in these data are
reshuffling coups.
Figure 3.3 shows the results from this analysis. The left panel presents the
estimates for leader-entry coups, while the right panel depicts the estimates for
leader-exit coups.8 The top coefficient estimate (diamond shape) for each
variable comes from the specification verifying Svolik’s results, and thus reflects
data that include both regime-change and leader-shuffling coups.9 The middle
estimate (triangle shape) for each examines only regime-change coups, that is,
coups that establish new autocratic regimes. The bottom estimate (square
shape) for each comes from a model that examines only reshuffling coups.
The main variables of interest are the Gini coefficient (i.e., inequality) and
the squared term that captures the nonlinearity in the hypothesized effect of
inequality. For both entry and exit coups, the coefficients of interest are statis-
tically different from zero in the verification model (top estimates, using a
sample that combines regime-change and reshuffling coups) and the model that
examines reshuffling coups only (bottom estimates). For regime-change coups –
those in which an out-of-power military faction seizes power from the incum-
bent dictatorship and establishes a new autocratic regime – the estimates for the
inequality variables are small and not close to statistical significance. Civil
conflict, arguably a direct measure of “mass, organized, and potentially violent
opposition,” may reduce the likelihood of regime-change coups rather than
raising it, as claimed in Svolik’s argument, though it is not statistically signifi-
cant for coups that establish new dictatorships (entry coups). It does not
increase the likelihood of either kind of coup. Thus, we find no support for
Svolik’s argument with regard to regime-change coups.
These findings suggest that Svolik’s empirical results depend primarily on
reshuffling coups within ongoing dictatorships. Eighty-five percent of leader-
change coups occur in military-led dictatorships, however. In other words,
most of these coups replace one military dictator with another rather than
replacing a civilian dictator with a military one, as would be expected if they
reflect a transition from rulers with a comparative advantage in cooptation to
rulers with a comparative advantage in repression, as Svolik argues.

conclusion
Our ﬁndings on the causes of coups fail to support the claim that popular
mobilization in protest or rebellion increases the likelihood of coups against

8
We drop international conflict from the specification because there are no observations in which
international conflict occurs in the same year as a regime-change exit.
9
These reported results are not exactly the same as those reported in Svolik (2012) because the
verification file does not include code setting the random seed for estimating a random coeffi-
cients model using the gllamm package. For substantive purposes, however, the top estimates we
report are identical to those reported by Svolik.

either democratic or authoritarian governments. Instead, our results suggest

that officers are motivated by their own interests. Coups that end democracies
are less likely when the incumbent government includes representatives of all
the country’s ethnic groups, implying that all ethnic groups represented in the
officer corps are represented as well.
Regime-change coups that oust dictatorships are less likely when the dicta-
torship originally achieved power by means of insurgency or foreign impos-
ition. We interpret this result as indicating that officers who share experiences,
values, ideas, and friendship (developed during the insurgency that created the
regime or imposed by foreign occupiers) with regime leaders are less likely to
develop grievances against them. We also find that ethnic heterogeneity in the
army is associated with more regime-change coups. Ethnic heterogeneity pre-
disposes the military to factionalism. The grievances of one ethnic faction can
motivate coups against dictatorships led by other ethnic factions. This result
suggests one of the mechanisms underlying the inability of factionalized armies
to provide a stable basis for rule, which we analyze in Chapter 5.
In the retest of Svolik’s (2012) argument about the effect of income inequality
on the likelihood of coups, we show that his results depend on combining leader-
shuffling coups with regime-change coups. When the two kinds of coups are
looked at separately, we find that middling levels of inequality are associated
with leader-shuffling coups in ongoing dictatorships, but not with coups that
initiate new dictatorships. Leader-shuffling coups usually replace one military
dictator with another, rather than replacing civilian dictators with military
rulers, as Svolik theorized. Regime-change coups are unrelated to inequality.
The direct indicators of threatening mass opposition (protest and civil conflict)
also have no effect on the likelihood of regime-change coups. In short, we find no
support for the idea that regime-change coups are motivated, either directly or
indirectly, by elite fears of popular opposition or redistribution.
Instead, the variables associated with military seizures of power, especially
those that end earlier dictatorships, reflect characteristics of the officer corps
and the group of political leaders with which the military bargains. Although
theoretical models of coups have focused on this bargaining relationship, to
date there have been few empirical tests of the effects of characteristics of the
actors doing the bargaining. We show that when dictatorial leaders share many
interests and experiences with military officers – which we argue happens more
often after seizures of power by armed rebellion or foreign invasion – their
military forces are less likely to oust them.
These results justify treating the military as an interest group and organized
political actor in its own right. In this book, we build on that result. We pay
attention to the special capacities the military can deploy and the consequences
of military organization for bargaining, both among officers and between them
and other political actors. We treat officers as motivated by their own interests.
They can ally with other groups just as any other political actor can, but they do
not routinely represent any particular societal interest.

Downloaded from https://www.cambridge.org/core. HarrassowitzN. American Lib Sub Svc, on 12 Aug 2018 at 07:37:04, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.003
Downloaded from https://www.cambridge.org/core. HarrassowitzN. American Lib Sub Svc, on 12 Aug 2018 at 07:37:04, subject to the
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.003
4

Power Concentration
The Effect of Elite Factionalism on Personalization

Islam Karimov, Soviet Uzbekistan’s last Communist Party leader, remained in

power after 1991, to become independent Uzbekistan’s first dictator. The rest
of the Uzbek political elite at independence had also spent their careers in the
Communist Party. There had been no nationalist uprising in Uzbekistan.
Russian-officered Soviet troops were still stationed in the country, and the
KGB managed internal security (Collins 2006, 173). Indeed, many observers
saw the Karimov regime as a simple continuation of Communist Party rule
under a new name.
Yet, even before formal independence, Karimov had eliminated the Com-
munist Party of Uzbekistan (CPU) as the institutional base for the regime and
taken personal control of the security services and high-level appointments
previously controlled by the party. Karimov retained the support of most of
Uzbekistan’s pre-independence elite despite destroying the formal underpin-
nings of their political power because the informal bases of their power
remained intact. Regionally based loyalty networks, usually referred to as clans,
continued to structure political bargaining and decision-making as they had
during communist rule.1
When first appointed, Karimov was weak; he was not a clan leader or an
important figure in the ruling party (Carlisle 1995, 196, 255; Collins 2006,
118–23). The most influential Uzbek clans had consolidated their informal
control during Soviet rule by infiltrating and coopting the CPU, which enabled
them to take over party and government patronage networks along with
different parts of the state-owned economy. Gorbachev’s earlier efforts to clean
up corruption and limit clan power had failed, and he had bowed to political

1
Clans are “informal power networks mobilised to capture the state and its resources in the interest
of the members and leaders of these networks” (Ilkhamov 2007, 70). Note the similarity between
this deﬁnition and Downs’s deﬁnition of parties as teams organized to capture government.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:04, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.004
62 Elite Consolidation

necessity in appointing a new CPU general secretary supported by the most

powerful clans. Clan leaders backed Karimov because they distrusted one
another. Karimov lacked an independent power base and needed the clan
leaders’ support to retain power, so he was expected to be responsive to their
demands (Collins 2006, 122–23; Ilkhamov 2007, 74–76). Clan leaders
“thought of him as their puppet” (Carlisle 1995, 196).
When Karimov was appointed first secretary, he “needed to demonstrate
sensitivity to local elite interests, and to maintain a balance of power amongst
the various significant clan actors . . . [E]ven though [he] did not trust certain
clan or regional factions,” he incorporated “at least token members of each
regional elite into the new government” (Collins 2006, 128). He had to share
the most with the clan leader most eager to replace him, whom he appointed
first as prime minister and then as vice president (Carlisle 1995, 196, 198;
Collins 2006, 129). Supporters of the vice president dominated the Supreme
Soviet.
As of early 1991, Karimov somewhat precariously controlled the CPU and
government by balancing and juggling competing clan interests. He lacked a
Soviet patron, and the Communist Party of the Soviet Union was disintegrating.
Karimov had to choose political and economic strategies to “retain all the
major [clan] players in the pact” (Collins 2006, 193). One observer described
his governing style as “consensual” (Ilkhamov 2007, 75, 78).
All clans included in the pact benefited from the regime, but they nevertheless
competed fiercely over high-level posts and the opportunities to make profits
and strip state assets that came with appointments. The clans’ mutual antagon-
ism provided Karimov with opportunities for power grabs at the expense of
particular clans, despite his need to maintain the support of most of them. On
the heels of the Soviet coup attempt, Karimov banned the CPU and seized its
assets, in part to reduce the resources and power of the Supreme Soviet.
Banning the party stripped many deputies of their privileged access to jobs as
members of the nomenklatura.2 He then created a new government-support
party, the National Democratic Party of Uzbekistan, led by himself. It never
controlled the resources the CPU had, however, or dominated politics and
controlled appointments. In the future, legislative nominations would be vetted
not by the party but by Karimov personally, reducing the power of the legisla-
ture (Collins 2006, 194–95, 253, 257). Outlawing the CPU shifted the balance
of power between Karimov and others in the ruling elite sufficiently for him to
abolish the office of vice president, arrest some of the vice president’s allies, and

2
In communist systems, high-level jobs in the bureaucracy, party, economy, education, and
military were reserved for individuals approved by the Communist Party. The nomenklatura
was the ofﬁcial list of individuals, nearly all of whom were party members, who could be
appointed to jobs at different levels of importance. By ending the party, Karimov opened
recruitment for these jobs to much wider competition.

thus rid himself of his most threatening supporter. Other clans did not defend
those excluded or arrested.
“The result of [Karimov’s] policies from 1990 through 1993 was the gradual
transformation from a communist regime to an autocratic one, in which power
belonged not to a hegemonic party, but to Karimov himself and the clique of
clan elites who surrounded him” (Collins 2006, 198). A feature of the post-
Soviet context that aided Karimov’s power grab is that he was not threatened
militarily. The clans had not penetrated the military stationed in Uzbekistan
or the KGB because Russians controlled both, and none of the clans had its
own militia.
In the wake of the Soviet coup attempt, Karimov increased his control of
security forces. He shifted resources to the presidential guard and the Commit-
tee for Defense, which he had created a few months before to counterbalance
the Soviet troops stationed in Uzbekistan. He also put Soviet military forces
under the highest-ranking Uzbek officer’s command (Collins 2006, 162) and
appointed a close ally from his own clan to lead the internal security agency.
From then on, Karimov kept the presidential guard, army, and internal security
agency “under his close supervision and control” (Collins 2006, 274).
Through the early 1990s, Karimov continued juggling multiple clan support-
ers, mostly by distributing state-controlled economic opportunities among them.
“[W]ith a tenuous political pact supporting him, Karimov had no choice but
to engage in a negotiating process with various factions” (Collins 2006, 257). On
the civilian side, he used his “political budget,” funded largely by the export of
cotton, gold, and oil along with the drug trade, to secure support (Collins 2006,
262–67; De Waal 2015). In contrast to the dictators described by De Waal
(2015), however, he had a near-monopoly on the means of violence. Members
of Karimov’s clan staffed the internal security agency. In the military, Russian
officers were purged and usually returned to Russia, while politically motivated
promotions and dismissals solidified the loyalties of Uzbek officers (Collins
2006, 274).
In the mid-1990s, Karimov began incrementally eliminating some of his
erstwhile allies from the inner circle. “Gradually, he consolidated power under
his personal control and loosened his dependence on his previous allies and
partners” (Ilkhamov 2007, 76). For example, the all-important head of cadre
policy, a representative of one of the most powerful clans, was arrested in 1994.
Karimov transferred the head of the Uzbek KGB, a man from his own clan, to a
less important post in 1995, and created a second security service so that the
two could report on each other (Collins 2006, 263).
Nevertheless, most clans continued to support Karimov in exchange for the
vast economic opportunities made possible by continued state ownership of
much of the economy. The Jurabekov clan linked to Samarkand, for example,
controlled oil and gas, many of the bazaars, and the cotton complex. The
Alimov clan from Tashkent controlled much of the banking system, the Minis-
try of Foreign Economic Relations, the tax inspectorate, the general procurator,

a share of the dollar trade, and most import-export businesses. Less powerful
clans controlled less, and a few were entirely excluded (Collins 2006, 264–68).
The monopoly rents, black market currency dealing, asset stripping, and wide-
spread corruption facilitated by clan control of the economy, however, led to
slow growth, increased poverty, and rising inequality.
By the late 1990s Karimov was “locked in an ongoing struggle to maintain
and increase his own personal autocratic control and to hold together powerful
regional and clan elites without allowing them to strip the state of its capacity to
survive” (Collins 2006, 170). He dismissed or demoted the prime minister,
defense minister, and several members of the inner circle from the Ferghana
network without political consequences. In 1998, Karimov dismissed
Jurabekov, Uzbekistan’s most powerful clan leader, along with many other
members of his network. A few months later, however, Karimov reinstated
Jurabekov after an assassination attempt attributed to him (Collins 2006,
170–71).3 Though Karimov narrowly escaped assassination, the attempt dem-
onstrated the Jurabekov clan’s credible threat to oust the dictator if he failed to
share with them, and he promptly reinstated them. However, the other clans
did not make similarly credible threats and therefore could not prevent power
grabbing at their expense.
As Karimov excluded important members of the original inner circle from
office and benefits, members of his own family took control of key sectors of the
economy. “Step by step, the major export resources were concentrated in the
hands of the central government, under the President’s personal control”
(Ilkhamov 2007, 76). By the early 2000s, the family controlled the major state
telecoms company, gold mining, and part of the oil business (Collins 2006,
170–71). In 2006, Radio Free Europe/Radio Liberty reported that Karimov had
“chip[ped] away at the political and economic might of some of Uzbekistan’s
most influential clans.” Jurabekov was dismissed once more in 2004 and
accused of corruption. This time he did not return to the inner circle. The
defense minister was forced to resign in 2005 and tried for corruption and
abuse of office. About 200 families had grown very rich under Karimov’s
original system of power sharing, but the circle of beneficiaries became smaller
and smaller in the 2000s, as it narrowed to not much more than the Karimov
family.4
Until his natural death in 2016, Karimov remained “a master at maneuver-
ing among the various clans in Uzbekistan and playing them off one another”
(Panier 2016). He retained control by balancing the clans, allowing no single
one to become too powerful, and rotating ministers, governors, and other

3
“Analysis: Uzbek Eminence Falls from Grace,” 2005, Radio Free Europe/Radio Liberty
(February 22), www.rferl.org/a/1057594.html; “Uzbekistan: Islam Karimov vs. the Clans,”
2005, RFE/RL (April 22), www.rferl.org/a/1058611.html.
4
“Uzbekistan: Karimov Appears to Have Political Clans Firmly in Hand,” 2006, RFE/RL (August
31), www.rferl.org/a/1070977.html.

appointments frequently to prevent ofﬁcials from building their own support

networks from which to challenge him (Saidazimova 2005; Ilkhamov 2007, 77;
Panier 2016). Nevertheless, “[c]ompeting regional and clan factions trusted
Karimov more than they trusted each other, and hence preferred to have him
at the center” (Collins 2006, 261).

Uzbekistan’s experience raises the question: How can a political leader who
seemed entirely dependent on the support of his country’s most powerful
political forces concentrate power at their expense and without forfeiting their
support? In this chapter, we explain how this happens. We show how factional-
ism in the ruling group – in Uzbekistan, their division into multiple competing
clans – undermines power sharing and thus facilitates the emergence of one-
man rule. This happens for two reasons. First, disunited ruling groups have
difficulty making credible threats to oust the dictator and therefore cannot
constrain him. And second, most members of the support group remain willing
to support the dictator even when he unilaterally reduces their access to benefits
because they are still better off inside the inner circle than excluded from it. In
contrast, a united seizure group can constrain the dictator’s ability to concen-
trate power, as well as his policy and distributive discretion, because they can
make credible threats to oust him if he fails to share.
The chapter begins with an overview of how bargaining works in dictator-
ships. It describes the central dilemma of new dictatorships: the colossal control
problem caused by handing dictatorial power to one member of the ruling
group. It explains our theory of how this problem affects bargaining between
dictators and other members of the dictatorial elite and describes the central
interests of both.
The second half of the chapter shows how preexisting characteristics of the
seizure group affect its ability to make enforceable bargains with the new
dictator. The dictator has little need to share or consult with his closest
supporters if preexisting factions facilitate bargaining with them separately,
as Karimov could with clan leaders. Bargaining separately induces competition
among faction leaders, which drives down the price dictators have to pay for
support. Based on these insights, we generate expectations about conditions
that facilitate the concentration of power in the hands of one man, which we
call “personalism.” We then explain how these informal bargaining relation-
ships, established during the dictatorship’s first years, can become sticky over
time. Last, we test these ideas using new data and show evidence consistent
with our arguments.

elite bargaining in dictatorships

In autocracies, a small number of regime insiders, usually acting in private
under informal rules, hammer out key decisions about leadership and policy
directions even in regimes with stable, well-developed formal institutions.

The influence and authority of members of the dictatorial elite may be renegoti-
ated frequently during the early years and subject to arbitrary and violent
change. Dictatorial elites may ignore formal rules and institutions if they
obstruct the drive to amass power. Losers in policy debates may be excluded
from the inner circle, demoted, arrested, or even executed. Life in a dictatorial
elite is thus insecure, dangerous, and frightening. Informal procedures may
become institutionalized over time, meaning that they become both more pre-
dictable and costly to change, but for the first months or years after an autocratic
seizure of power, bargaining within the dictatorial elite often occurs in an
environment of contested, changing, and nonbinding institutions (Svolik 2012).
When making decisions about policy, leadership, and institutional choice,
the dictator and members of his inner circle take into account expected effects
of the choice on regime survival, but also how decisions may affect their
individual power, influence, and access to resources. Members of dictatorial
elites live in grim, dog-eat-dog worlds. Taking one policy position can provide
the opportunity to take a bite out of another dog, while taking a different one
could incite the pack to tear you apart. Autocratic policy makers, like demo-
cratic ones, may care deeply about the substance of policy, but they cannot
afford to ignore how their decisions will affect regime survival and their
personal survival as well.
To explain these decisions, we focus initially on the interests of members of
the seizure group. Because the members of groups never share exactly the same
interests, our theory begins with strategic interactions among them. We do not
assume that seizure groups, or the regime elites that derive from them, are
unitary actors because the empirical record shows that discipline among them is
imperfect. How much discipline they can maintain requires empirical investi-
gation. Consequently, neither the dictatorship’s inner circle as a whole nor any
subset of it larger than one member should be assumed to behave as a
unitary actor.
We do assume that members of the dictatorship’s inner circle want to
maintain the dictatorship, which they expect to further their policy goals, as
well as provide opportunities for personal advancement and often enrichment.
However, they also want to increase their personal share of power relative to
others in the inner circle. They must compete with each other for power, not
only to improve their standing in the inner circle or their access to wealth, but
also to maintain their current positions against lower-ranked regime supporters
striving to replace them in the inner circle. An increase in power for one
member of the inner circle comes at the expense of someone else. We see power
as a rank ordering based on politically relevant resources, understood by
insiders even when not perceptible to observers. One insider cannot move up
the rank order without displacing someone else.
Because of the intense competition within the inner circle, we expect the
dictatorship’s most powerful decision makers to consider how all policy,
appointment, and institutional choices might affect their own standing, as well

as regime maintenance. The creation of new formal institutions beneﬁts the

individuals who will lead them and those who will work in new agencies
associated with their implementation. Policy choices also often entail the cre-
ation of new agencies, which, again, benefits those chosen to lead and work in
them. Policies also have distributional consequences that advantage some and
disadvantage others. They thus affect the welfare of the constituents of inner
circle members differently. Consequently, we expect inner-circle members to
favor policies and institutions that both improve their own place in the hier-
archy and increase the likelihood of regime survival. However, both are impos-
sible for everyone in the inner circle since any change that increases the powers
of one member decreases those of others. Bargaining over policy thus has a
noncooperative dimension, and strategic considerations often affect substantive
policy choices (just as in democracies).
In other words, the dictator and inner circle engage simultaneously in two
kinds of strategic interaction: (1) a cooperative effort aimed at keeping all of
them (the regime) in power and (2) noncooperative interactions in which
different members/factions seek to enhance their own power and resources at
the expense of others in the inner circle. Each individual strives to amass
resources and capacities up to the point at which his efforts would destabilize
the regime or lead to his own exclusion from the inner circle.
The dictator has a resource advantage because he has the most direct access
to state revenues and an information advantage because he has access to the
reports of all internal security services. Nevertheless, the dictator faces the same
dilemma as other members of the inner circle: he wants to extend and consoli-
date his control up to the point at which other members of the inner circle
would take the risk of trying to oust him. The struggle over the distribution of
power in a new dictatorship can transform the seizure group from the coopera-
tive near-equals who had plotted the fall of the old regime into competitors in a
vicious struggle for survival and dominance.
We see these incentives as common to all dictatorships, but they play out in
different ways, depending on concrete characteristics of the seizure group that
pre-date the installation of the dictatorship. We argue that preexisting differ-
ences among the groups that initiate dictatorship lead to post-seizure differ-
ences in what kinds of individuals with what interests become members of the
dictatorial inner circle, how the inner circle makes decisions, which policies and
institutions they choose, how they seek to attract members of society as allies,
and how they respond to opposition. Preexisting characteristics of the seizure
group do not determine everything that happens over the course of a dictator-
ship, but they do affect the likelihood that regimes will display specific, often
long-lasting, patterns of behavior. The institutions chosen by dictatorial elites
after they take power also have consequences for subsequent bargaining and
the way dictatorships break down. That is their purpose, after all. Preexisting
characteristics of the seizure group, however, influence the choice of these
institutions.

We thus share with Svolik (2012) the view that all dictatorships face certain
dilemmas, such as the dictator’s temptation to grab more power than his
supporters want to delegate, but we emphasize that these dilemmas can have
different outcomes, depending on characteristics of the seizure group that affect
the bargaining power of different members of the group. In this chapter, we
explain how one preexisting feature of seizure groups, their position on a
continuum from factionalism to unity, affects authoritarian politics.
We provide greater detail in the sections that follow, beginning with the ﬁrst
decision seizure groups confront: selection of a leader.

handing power to a leader

In order to govern, seizure groups must choose a leader (dictator).5 They need a
leader to speak on behalf of the new government, represent it to the populace
and foreign actors, organize the implementation of policies made by the group,
coordinate their activities across agencies and levels of government, mediate
conflicts within the inner circle, act in emergencies, and make final decisions
when opinions in the group are divided. The point of choosing a leader is to
achieve the goals of the group. However, the delegation of the powers needed to
fulfill these responsibilities in the largely institution-free setting of early dicta-
torship causes the interests of the dictator to diverge from those of his closest
allies. While the dictator hopes to defang his allies’ threats to oust him if he
disregards their interests, they seek to hold the new dictator in check. This
happens regardless of whether the dictator and his allies come from the same
ethnic or other close-knit group. This divergence of interests creates the colossal
problem of how to control a leader with dictatorial powers.
When the seizure group delegates powers to a leader, it does not intend to
give him the capacity to choose policies most of them oppose, unilaterally
exclude from the inner circle individuals who helped seize power, or dismiss,
jail, or kill seizure-group members, their allies, and family members. These are
highly visible depredations on the ruling group. The absence of binding limits
and institutional checks on the dictator, however, mean that only credible
threats to oust the dictator deter him from reneging on agreements and abusing
his supporters (Magaloni 2008; Svolik 2012).
The dictatorial elite cannot costlessly dismiss the dictator. On the contrary,
efforts to oust a dictator always involve a high risk of failure, followed by near-
certain exclusion from the ruling group and possible exile, imprisonment,
torture, and/or execution. And yet the dictatorial elite can limit the dictator’s

5
Literal leadership choice often occurs before the seizure of power. Regardless of the timing of
original leadership choice, after the seizure of power groups “choose” leaders in the sense that if
the support of enough members were withdrawn, the leader could not retain his position (Bueno
de Mesquita et al. 2003).

depredations only if they can credibly commit to ousting him if he seizes more
power or resources than they have agreed to transfer.
Seizure groups should anticipate the possibility that the man chosen to lead
could escape their control, since the dictator cannot make credible promises not
to use his powers. Before they take power, some seizure groups try to hem in the
dictator by making rules about who should help him make policy, how these
lieutenants should be chosen, the periodic rotation of leadership, and how they
will handle succession. When plotters come from professionalized militaries,
which tend to be legalistic and rule-bound, they may negotiate quite detailed
arrangements for term limits and consultation over policy choice within the
officer corps (Fontana 1987). These rules can be enforced at the time they are
agreed to because power is dispersed within the group. The man who wants to be
leader must agree to power-sharing arrangements such as regular consultation
or term limits in exchange for the support of other members of the seizure group.
Most dictators, like Karimov, are weak the day they become regime leader.
Karimov needed support initially from several clan leaders and the Communist
Party to stay in power, and thus had to distribute state offices and the resources
they controlled among them. Military dictators are sometimes weak because
they have had to retire from active duty, and thus give up the ability to control
others’ promotions, postings, and retirements, in order to secure the support of
others in the junta for their appointment as leader.
Bargains made when the dictator is weak, however, last only as long as they
are self-enforcing because of the lack of third-party enforcement institutions in
dictatorships (Barzel 2002, 257; Acemoglu, Johnson, and Robinson 2005, 429;
Svolik 2012). The initiation of dictatorship creates immediate opportunities for
changing the relative distribution of power within the seizure group. History is
replete with examples of men who were invited to lead by plotters who believed
they would be malleable figureheads, but who quickly marginalized and some-
times killed those who had expected to call the shots. Plotters sometimes
consciously choose an individual considered uncharismatic and legalistic to
reduce the likelihood of future concentration of power in his hands. It has been
reported that Chilean plotters chose General Augusto Pinochet, a latecomer to
the coup conspiracy, for that reason. He was quick to concentrate power in his
own hands, however. Other examples include Major Mathieu Kérékou, who
was invited by even more junior coup plotters to lead the 1972 seizure of power
in Benin. Within three years, he had excluded the original plotters from gov-
ernment, apparently killing one and jailing several others (Decalo 1979; Allen
1988). After the 1975 coup in Bangladesh, junior officers and ex-officers
released General Mohammad Zia ul-Hak from jail and appointed him Chief
Martial Law Administrator because they feared the rest of the army would not
acquiesce in the coup if a junior officer was appointed. Zia quickly excluded
them from the ruling group, arrested those who tried to oust him for violating
their initial bargain, and executed their leaders for treason (Codron 2007,
12–13).

This problem is not limited to military seizure groups. Malawi’s Hastings

Banda, recruited as a presidential candidate by young civilian nationalists, not
only excluded his youthful colleagues but quickly transformed the elected
government he led into a highly repressive autocracy. Banda, a doctor who
had been working outside the country for more than twenty years, was invited
to lead by young independence leaders who “assumed he could not conceivably
harbor any long-term political ambitions” (Decalo 1998, 59). After the
founding election, he dismissed his young allies from their cabinet posts, forced
them into exile, and purged the ruling party of anyone suspected of challenging
his “unfettered personal rule” (Decalo 1998, 64), which then lasted for more
than thirty years.6
The frequency of this sequence of events suggests that control of the state,
even as rudimentary a state as Malawi’s at independence, gives the paramount
leader a resource advantage over his erstwhile colleagues. The new dictator
finds himself “almost immediately in command of all the financial and adminis-
trative resources of the state” (Tripp 2007, 143). Becoming head of state gives
the new dictator access to revenues – especially from taxes, the export of
natural resources, and foreign aid – far greater than he has had before and
greater than other members of the seizure group can command. This control
endows the new dictator with agenda-setting power when it comes to policy-
making and distributive decisions. Revenues can be shared with the inner circle
or spent by the dictator to buy personal support and security. Further, access to
state revenues gives the dictator substantial control over appointments to state
offices, which he can use to bring loyalists into decision-making positions
and to create state agencies to pursue goals not shared by the rest of the
seizure group.
These revenues can enable the dictator to outmaneuver his allies. In some
circumstances, he can buy the support of some members of the inner circle for
the exclusion of others. He may even be able to buy their support for changes
that further enhance his power at the expense of theirs. In short, a dictator who
was first among equals on the day he was chosen has substantial potential to
grab additional resources and power later.
We refer to dictatorships in which the leader has concentrated power at the
expense of his closest supporters as personalist. The defining feature of person-
alist dictatorship is that the dictator has personal discretion and control over
the key levers of power in his political system. Key levers of power include the
unfettered ability to appoint, promote, and dismiss high-level officers and

6
Of course, the respectable senior leader does not always win these power struggles. General
Naguib, invited by the youthful Free Ofﬁcers to lead the government that would replace the
Egyptian monarchy, was ousted two years later by Colonel Nasser, the plotters’ original leader
(Haddad 1973, 11–42). Regardless of who eventually emerges on top, however, these are
examples of why struggles for power within the dictator’s inner circle often begin soon after
seizures of power.

officials, and thus to control the agencies, economic enterprises, and armed
forces the appointees lead. In such regimes, the dictator’s choices are relatively
unconstrained by the institutions that can act as veto players in other dictator-
ships, especially the military high command and the ruling party executive
committee. Personalist dictators juggle, manipulate, and divide and rule other
powerful political actors. Like all dictators, they need some support, but they
can choose from among competing factions which ones can join or remain in
the ruling elite at any particular time.7 Personalist dictators are thus powerful
relative to other members of the elite, but not necessarily relative to society or to
international actors.
Islam Karimov’s rule exemplified personalism, especially in later years. His
control of Uzbekistan’s political system derived from his appointment powers.
Initially, he had to bargain with several clan leaders over the composition of his
government and the control of all important state-owned enterprises and
government agencies. Over time, however, he achieved much greater personal
discretion over appointments, and most important clans continued supporting
him despite losing some of their influence and access to income.
We argue that discretion such as Karimov’s arises from the dictator’s ability
to bargain separately with supporters, to play them off against each other, to
ally with some in order to damage or exclude others, and to bring previously
excluded groups into the dictatorial inner circle in order to tip the balance of
power in his favor when needed. The concentrated power wielded by person-
alist leaders is thus not absolute but rather depends on the dictator’s ability to
use a changeable divide-and-rule strategy against supporters who could control
or overthrow him if they could unite. The description of Yemen’s Ali Abdullah
Saleh as “dancing on the heads of snakes” captures this understanding of the
personalist dictator as not necessarily the deadliest member of the ruling coali-
tion but rather the one who can stay aloft by pitting some factions against
others in an ever-changing balancing act (Clark 2010).
Personalism can rise and fall during a single dictator’s tenure, for reasons we
describe later. (The data we use to measure personalism are coded yearly to
capture these within-ruler and within-regime changes. They thus differ from
our older regime-type coding, which could not capture these real-world
variations.)

7
Some deﬁnitions of personalism emphasize informal alliances between the dictator and leaders of
ethnic or other kinds of groups and personal loyalties maintained through patronage networks
(e.g., Roessler 2016). We put less emphasis on informality for two reasons. First, personalist
dictators distribute formal ofﬁces in government, the ruling party, and the military to their elite
supporters. The bargains that keep the ruling group together could not be maintained without
access to state revenues, and even rudimentary states have formal governing structures that
provide the resources that hold alliances together. Second, personalized relationships and bar-
gaining can occur within formal institutions such as ruling parties. For example, Stalin had nearly
full discretion over appointments and decision-making during the last decade of his life, but most
of those decisions were made in party committees and implemented by party cadres.

The Dictator’s Interests

Since achieving the post of dictator requires great effort (as well as quite a bit of
luck), we can infer that those who achieve it wanted it. Regardless of whether
the job lives up to expectations, the danger inherent in giving it up predisposes
dictators to try to hang on to power. Twenty percent of dictators are jailed or
killed within the first year after losing office, while another fifth flee their native
countries to avoid such consequences.8
Dictators must fear their closest allies. Their careers can end in two ways other
than natural death: the overthrow of the regime or the ouster of the dictator
despite regime continuity. Members of dictatorial inner circles often lead ousters
even when they also involve popular mobilization. Figure 4.1 shows the fre-
quency of the various events that end dictatorial regimes, a subject we return to in
Chapter 8. More dictatorships end in coups, that is, overthrows by the officers
who were entrusted to defend them, than in other ways. Formerly powerful
members of inner circles, however, have also led many popular uprisings and
opposition election campaigns, the other common ways that autocracies end.
Only about half of dictator ousters accompany regime failures (Geddes,
Wright, and Frantz 2014). The rest of the time, dictators are replaced but the
regime survives. Regime insiders cause nearly all dictator replacements in
surviving autocracies. Coups cause about a quarter of these. Eighty-five percent
of such leader-shuffling coups replace one military dictator with another.
Deaths, party decisions, and term limits enforced by the dictatorial inner circle
account for most of the remaining three-quarters of dictator replacements in

80
77

60
59
Frequency

40
38

20
17 17
10
5
0
Rule change Election Uprising Coup Rebellion Foreign Failed state
figure 4.1 Frequency of events that end dictatorships.
Note: Election includes those in which incumbents lose or do not run.

8
Calculated from Archigos data.

ongoing regimes. It is rare for popular action to oust a dictator without ending
the dictatorship as well.
These basic facts about dictator ousters reveal the potential threat to the
dictator posed by members of the inner circle (Svolik 2012; Roessler 2016). This
threat gives the dictator an interest in closely scrutinizing his allies’ activities and
dismissing, jailing, or killing them if he begins to suspect their loyalty. The dicta-
tor’s interest in limiting threats from the inner circle implies that members of the
dictator’s support coalition today cannot count on being included in it tomorrow.9
To sum up, the average dictator has good reasons for wanting to retain
power since he cannot count on a safe, afﬂuent retirement if he steps down. Nor
can he count on his family being left in peace and prosperity. The individuals
most likely to be able to oust him are members of his inner circle. The dictator
thus needs their support. Because promises can never be completely credible,
however, the dictator has reason to spy on his allies to try to assess whether
they are plotting behind his back. The dictator has every reason to try to build
his own political resources at the expense of his allies, as he can never fully trust
them. In other words, he has strong reasons to violate the implicit leadership
contract by aggrandizing his own power.

The Interests of Other Members of the Inner Circle

Exclusion from the dictator’s inner circle can result in execution, torture, long
imprisonment, property confiscation, exile, and poverty. Family members may
have to bear these costs along with the target of the dictator’s suspicions.
Consequently, constraining the dictator’s ability to exclude members from the
inner circle ranks at the top of its members’ goals. Members of the inner circle
also want to influence policy choices and build their own clientele networks,
which are needed to secure their influence and acquire wealth. Some of them
yearn to supplant the dictator. To accomplish these goals, they need to retain
influence on decision making, and they need to obtain posts and promotions
that entail both some policy discretion and the ability to hire, promote, and do
favors for others. The opportunities available to members of the inner circle
vary with the kind of posts they occupy.
The members of the dictator’s inner circle thus have strong reasons to want
to maintain collegiality in decision-making and the dispersion of resources
within the group, rather than allowing the dictator to usurp policy discretion
and control over top appointments. These aims give members of the dictator’s
inner circle good reasons to create institutions that enforce constraints on the
dictator, and thus to prefer some institutional arrangements to others.
For example, members of the inner circle may demand term limits for the
leader as a way of both limiting his ability to amass powers over time and

9
This statement of the dictator’s interests thus conﬂicts with a central feature of the model
proposed by Bueno de Mesquita et al. (2003).

increasing their own chances of occupying the top post in the future. In
dictatorships led by a military officer, other officers in the inner circle also want
him to retire from active duty, which establishes the credibility of his commit-
ment not to use control over their promotions, retirement, and postings to
concentrate power at their expense (Arriagada 1988; Remmer 1991). If the
dictatorship is organized by a ruling party, members of the inner circle want
party procedures for choosing its executive committee (politburo, standing
committee) to be followed in spirit not just letter; they do not want the dictator
personally to choose members of the party’s decision-making body since that
would mean that he can exclude anyone who might disagree with him.
Members of the inner circle also favor an institutional arrangement that
limits the dictator’s personal control over internal security forces. Preventing
the dictator from gaining personal control over the internal security apparatus
is important to the welfare of members of the inner circle. If a dictator
controls the security police, he cannot credibly commit not to use it against
his allies.
Conflict over the distribution of power between dictators and their support-
ers afflicts all new dictatorships. Many of the power struggles during the first
months and years of dictatorship can be understood as efforts by the inner
circle to control the dictator, efforts by the dictator to escape control, and
efforts by both to institutionalize the relationship in order to reduce potentially
regime-destabilizing conflict between them.
In this environment, everyone’s actions are somewhat unpredictable,
prompting both the dictator and his lieutenants to remain on guard and
trigger-happy. Furthermore, the unreliability of information increases the like-
lihood of misinterpreting the actions of others, opening the way to paranoia.
Despite his information advantage over others in the inner circle, the dictator’s
information about what they really think and what they may be planning
remains limited and unreliable (Wintrobe 1998). For all these reasons, early
periods in dictatorships tend to be unstable, conflictual, and sometimes bloody.

bargaining over the distribution of

resources and power
Earlier studies have emphasized the importance of whether the dictator can
credibly commit to fulﬁll promises he makes (e.g., Magaloni 2008; Svolik 2012;
Boix and Svolik 2013). Less attention has been paid to how much spoils and
power the dictator really needs to share with his closest allies to retain their
support. Here we ask: Assuming the dictator could credibly commit to sharing
power, how much does he need to share? That is, how much power sharing
does his personal survival require?
A dictator can reduce the amount he shares in two main ways: by reducing the
number of supporters with whom he shares spoils and inﬂuence and by reducing

the amount of power or inﬂuence he shares without decreasing the number of

supporters. These two strategies often go together. Reducing the number of groups
represented in the inner circle, and thus reducing the benefits received by the
constituencies they represent, is easier for outsiders to observe than reductions in
influence while the individuals who have lost some power remain part of the
dictatorial elite. We discuss reducing the size of the ruling coalition first.
Seizure groups are often large when they take power. At the time of a coup,
for example, many members of the officer corps must acquiesce in the seizure
for it to succeed, even if they do not actively support it; “authoritarianization”
can occur only after a party has attracted the support of most citizens in an
election. A dictator may not need all this support to survive, however, because
some members of the seizure group may lack the means to oust him.10 This
means that the inner-circle members linked to some constituencies can be safely
shed, and the dictator can keep their “share” (Bueno de Mesquita et al. 2003).
Bargaining between the dictator and others in the inner circle over the
distribution of power and spoils begins when the dictator decides whether to
stick to arrangements initially agreed to or renege on the implicit or explicit
commitment to share and consult. The dictator controls some amount of goods
and powers that both he and other members of the inner circle value highly.
These can include access to material goods, such as the revenues from state-
owned natural resources, as well as control over various aspects of policy and
the choice of personnel to fill high and low offices. These goods are valuable,
both to have and to dole out to clients. The dictator makes an initial decision
about how to distribute these goods to best secure the adherence of needed
supporters. His lieutenants expect to be consulted about these decisions and to
receive a share, but their expectations may not be enforceable.
The dictator’s agenda-setting power gives him an advantage over those who
can only react to proposals (Baron and Ferejohn 1989). Those whose only
options are to accept or reject distributive proposals must make decisions about
whether to continue supporting the dictator – based on a comparison between
what the dictator offers and what they expect to receive if they withdraw from
the ruling group. Once the dictator has demonstrated the way he intends to
handle the resources he controls, other members of the inner circle can acqui-
esce in the distribution he proposes or contest it. If they could replace the
dictator, they might do much better, but if they see the choice as lying between
acceptance of what the dictator offers and exclusion from the ruling coalition
(the consequence of rejecting the offer), they would be better off accepting even
quite a small amount in exchange for their continued support. This logic leads
to the counterintuitive conclusion that members of dictatorial elites may con-
tinue to support the dictator even if they receive only a little in return.

10
Thus we focus here on the groups Roessler (2016) does not consider, those that it is safe to
exclude precisely because they cannot threaten the dictator with overthrow.

The capacity of the dictator’s allies to inﬂuence his distribution decisions

(enforce their expectations) depends on the credibility of their threats to oust
him. The dictator can always be removed by the united action of other elites.
Sometimes he can be removed by small groups of them. Trying to remove
dictators is risky, however, and no one plotting such a course can count on
success. Terrible consequences can follow the discovery of plots. As a result,
fewer allies plot than are dissatisﬁed with their share. The dictator thus reaps an
additional bargaining advantage from the riskiness of plots. The more unlikely
a plot’s success, the larger share a dictator can keep.
To summarize these points, all dictators need some support, which they must
reward, but they need to offer only enough to maintain the minimum coalition
required to stay in power. The other original members of the seizure group, and
the parts of its larger support coalition associated with them, can be excluded
without endangering the regime. There is a strong incentive to exclude them
because the dictator can then keep their “share” or give it to others whose
support he needs more. Remaining inner-circle members want to share spoils
and power, but they still have little bargaining power besides the threat to
replace the dictator. These conditions mean that the dictator can often get away
with keeping the lion’s share for himself, just as the proposer in standard
legislative bargaining games can (Baron and Ferejohn 1989).
This logic thus makes clear why members of authoritarian coalitions often
acquiesce in the concentration of power and resources in the hands of dictators.
Note that although Baron and Ferejohn’s (1989) result does not ﬁt empirical
reality in democratic legislatures very well – that is, prime ministers do not
generally keep the lion’s share of resources – it is eerily similar to the reality of
conspicuous consumption and Swiss bank accounts enjoyed by many dictators.

characteristics that influence the credibility of

threats to oust the dictator
So far we have treated members of the dictator’s inner circle as separate
individual actors. A real inner circle might resemble this image if, for example,
the leaders of multiple clans, parties, or ethnic groups, who had cooperated to
throw out a previous regime or colonial power, formed the new ruling group.
Militaries riven by ethnic, ideological, or personal factions may also contain
many faction leaders who bargain individually rather than being subsumed in a
single uniﬁed military bargainer. Parties colonized by clans, as in Uzbekistan,
can also contain multiple faction leaders who bargain individually on behalf of
their members. So can recently organized parties formed by coopting the
leaders of older rival parties. In these circumstances, the dictator does in fact
bargain with multiple separate actors, and threats to oust him are less credible
because of the high risk of plots involving single factions and the difﬁculty of

uniting the factions for joint action. As a result, the dictator can often concen-
trate resources and power in his own hands, as Karimov did.
Preexisting discipline and unity within the seizure group – from which the
dictatorial inner circle is chosen – tilt post-seizure bargaining against the dicta-
tor, however. Internally cohesive seizure groups can bargain as something close
to unitary actors over issues such as limitations on the dictator’s personal
discretion. They also face fewer collective action problems when it comes to
organizing the dictator’s overthrow. Disciplined unity develops in professional-
ized militaries and parties formed as “organizational weapons”11 because these
institutions transparently link individuals’ future career success to obedience to
superior officers or the party line. Officers are punished or dismissed for
disobeying senior officers, thus ending their careers and livelihoods. In discip-
lined parties, ordinary party members can be excluded for criticizing the party
line, and elected deputies who vote against it may be expelled from the party
and lose their seats. Such incentives are needed to maintain unity within groups.
Many armies and ruling parties lack this degree of internal discipline. In
armies factionalized by ethnic, partisan, or personal loyalties, officers’ career
prospects depend on their faction leaders’ success in achieving promotions and
access to other opportunities. In such a military, lower-ranked officers cannot
be counted on to obey the orders of higher-ranked officers from rival factions.
In parties that have achieved dominance by persuading the leaders of other
parties to “cross the aisle” in return for jobs and other spoils, discipline also
tends to be low. A party history of incorporating most major political interests
into one party tends to result in party factionalization based on ethnicity,
region, policy position, or personal loyalties.
Where dictators have to bargain with an inner circle drawn from a unified
and disciplined party or military, the threat of ouster is more credible and the
price of support higher. Dictators in this situation face groups that, like labor
unions, can drive harder bargains than the individuals in them could drive
separately (Frantz and Ezrow 2011). In these circumstances, dictators usually
find it expedient to consult with other officers or the party executive committee
and distribute resources broadly within the support group.12 In short, the prior
organization, unity, and discipline of seizure groups give dictators reason to
maintain power-sharing arrangements with members of the inner circle.

11
Selznick’s (1952) term for communist parties characterized by “democratic” centralism and
extreme discipline.
12
In a bargaining model, commitment by the seizure group to make decisions as a unitary actor
turns the negotiation between the dictator and his allies into a two-person game. In two-person
games where exclusion from the game is not possible without ending the game (that is, the
dictator cannot retain power if the uniﬁed support group turns on him or if he excludes them)
and the cost of bargaining is the same for both players, the division of the pie will be equal
(Rubinstein 1982).

If the original seizure group includes both a disciplined group and some
additional allies, supporters affiliated with more disorganized groups pose
weaker threats to the dictator and are thus less risky to exclude. As an example
of this process, within months of the Sandinista rebels’ victory over the forces of
Nicaraguan dictator Anastasio Somoza Debayle in 1979, they had eliminated
the democratic reformist (non-Sandinista) members of the broad coalition that
supported the revolution (Gorman 1981, 138–42). Cold War observers noted
that when broadly based coalitions that included a well-organized Marxist
party ousted a government, the non-Marxist members of the coalition were
shed soon after power was secured. The special perfidy of Marxist parties
does not explain this phenomenon, however. It arises from the logic of the
post-seizure situation. Military coup makers and well-organized non-Marxist
parties also excluded their less unified and unneeded supporters once they had
secured power. Non-Ba’thist officers from the intensely factionalized Iraqi
military, for example, led the 1968 coup, supported by the Ba’th Party. To
try to stabilize power sharing between these groups, the plotters chose a
Ba’thist president (regime leader), and the non-Ba’thist coup leaders got prime
minister, minister of defense, and command of the Republican Guard. Within
less than two weeks, however, one of the non-Ba’thist coup leaders had been
persuaded to join the party. The other non-Ba’thist coup leaders could then be
excluded safely. They were forced into exile, leaving the Ba’th Party and its
chosen leader able to consolidate a more narrowly based dictatorship (Tripp
2007, 184–85).
The dictator’s drive to narrow his support base arises from minimum-
survival coalition logic (Bueno de Mesquita et al. 2003). That is, the dictator
wants enough support to survive in power, but no more because support
must be paid for, even if at a low rate, in resources and shared power. The
larger the number of individuals or groups over which resources must be
spread, the less for each one. Because authoritarian governments need less
support to remain in power after they have captured and transformed the
state bureaucracy, courts, military, police, and taxing authority than for the
initial seizure of power, they can get away with excluding some members of
the seizure coalition and large parts of the population from the distribution
of benefits.
The internal cohesion of seizure groups affects bargaining within the inner
circle via two different paths. First, the dictator can bargain separately with
each member of a factionalized seizure group, offering some special deals in
return for siding with him on crucial decisions and inducing members of the
inner circle to compete with each other for resources. If the dictator excludes a
member of a factionalized inner circle, as Karimov did many times, remaining
members are more likely to seize the resources of a fallen comrade and use him
as a stepping-stone to a higher place in the hierarchy than rally to his support.
Overall, members of factionalized seizure groups tend to get less from the

dictator in exchange for their support because their competition with one
another lowers the price they can extract.
Second, factionalism reduces the credibility of threats to oust the dictator if
he fails to share. Factionalized support groups have difficulty organizing to oust
the dictator, and most unarmed factions lack the capacity to oust him on their
own. As we saw in the summary of recent Uzbek history at the beginning of the
chapter, if one faction can credibly threaten the dictator with ouster, it can
enforce its own sharing agreement, but most of the time unarmed factions
cannot. Consequently, the dictator can concentrate more powers in his
own hands.
In this section, we focused on one characteristic of seizure groups, which
influences the credibility of threats by members of the dictatorial inner circle to
oust the dictator: how much unity and discipline had been enforced within the
seizure group before it seized power. When united militaries or disciplined
parties lead authoritarian seizures of power, lieutenants are likely to be able
to resist extreme concentration of power in the dictator’s hands. When, instead,
factions divide the officer corps or parties are recent amalgams of multiple
jostling cliques, they cannot obstruct the dictator’s drive to concentrate power.
Our argument thus suggests an explanation for the personalization of power in
many African countries noted by Africanists (e.g., Bratton and van de Walle
1997) after an initial seizure of power by the military or a transition from
elected government to single-party rule. The newness of parties and the recent
Africanization of the officer corps at the time of independence often resulted in
factionalized party and military seizure groups in the first decades after
independence.

measuring personalism
In the next section, we test the argument about why dictators can concentrate
power in their hands at the expense of their closest supporters. Though anec-
dotal evidence supports the ideas we propose, the absence of a measure of
personalism has hampered our ability to evaluate them systematically in the
past. Here, we leverage new data we collected on various features of dictator-
ship to derive such a measure.
To capture the idea of personalism, we use eight indicators of dictators’
observable behavior (assessed yearly) that we believe demonstrate power con-
centration at the expense of others in the dictatorial elite (Geddes, Wright, and
Frantz 2017; Wright 2018). We create a time-varying index of personalism
using these eight indicators: dictator’s personal control of the security appar-
atus, creation of loyalist paramilitary forces (create paramilitary), dictator’s
control of the composition of the party executive committee, the party execu-
tive committee behaving as a rubber stamp, dictator’s personal control of

appointments, dictator’s creation of a new party to support the regime, dicta-

tor’s control of military promotions, and dictator’s purges of officers.13
The first two reflect the dictator’s relationship with security forces: whether
he personally controls the internal security police (security apparatus) and
whether he has created a paramilitary force outside the normal chain of military
command (create paramilitary). Personal control of internal security agencies
increases the dictator’s information advantage over other members of the
dictatorial elite as well as his ability to use violence against them. The dictator’s
advantage comes not only from his access to the information collected, which
he can keep from other members of the inner circle, but also from his ability to
order security officers to arrest his colleagues. Knowledge provided by security
agencies can help the dictator identify members of the inner circle who might
challenge him. Actions by the dictator that we code as indicating personal
control of internal security include his direct appointment of the head of the
security service (if this appointment appears to ignore the normal military
hierarchy), his creation of a new security agency, and his appointment of a
relative or close friend to lead a security force.
Dictators use paramilitary forces to counterbalance the regular military
when they see it as unreliable. The creation of armed forces directly controlled
by the dictator increases the concentration of power in that it reduces the
regular military’s ability to threaten the dictator with ouster if he fails to share
or consult. In order to identify only paramilitary forces created by dictators to
solidify their personal power, we exclude party militias and those created to
help fight insurgencies. We code both the appointment of a relative or close
friend to command a paramilitary force and the recruitment of a paramilitary
group primarily from the dictator’s tribe, home region, or clan as indicating his
personal control. The forces coded as dictators’ paramilitary forces include
Kwame Nkrumah of Ghana’s President’s Own Guard and Saddam Hussein’s
Republican Guards.
The next indicator of personalism assesses the dictator’s relationship with
the ruling party’s leadership. We code whether the regime leader chooses or
vetoes members of the party executive committee (party executive) in dictator-
ships organized by a ruling party. The dictator has concentrated power, we
argue, if he chooses top party leaders rather than party leaders choosing him. In
communist Hungary, for example, the first dictator, Mátyás Rákosi, began the
regime with a politburo composed of about equal numbers of close allies who
had spent the war with him in Moscow and cadres who had spent the war
underground or in jail in Hungary. Rákosi did not choose the leadership of the

13
Adding to the index additional indicators of personalism that measure whether the dictator
appoints his relatives to high ofﬁce, rule by plebiscite, whether the military is ethnically homo-
genous, and whether the regime leader also leads the support party does not alter the ﬁndings
reported below. A composite measure that includes these additional variables is correlated with
the one used at 0.982.

underground party members, meaning that he did not initially control the
composition of the Politburo. After the seizure of power, however, Rákosi
gradually eliminated those with independent bases of support in Hungary –
life-long dedicated communists who were accused of treason, subjected to show
trials, and sentenced to long prison sentences or execution (Kovrig 1984). By
the early 1950s, Rákosi fully controlled who joined or was dismissed from the
Politburo, and thus the Politburo could not constrain him.
A fourth variable also captures information about the ruling-party executive
committee. It identifies party executive committees that serve as arenas for
hammering out policy decisions rather than as rubber stamps for policy and
personnel choices made by the dictator (rubber stamp). We see discussion of
policy alternatives and disagreements over choices, which are reported in the
media and in secondary sources, as indications that the dictator has not concen-
trated policy-making power. The absence of policy disagreements indicates the
opposite. In North Korea, for example, Kim Il-sung reorganized the Korean
Workers’ Party (KWP) leadership structure at the 1966 Party Congress; by
1968 “he faced no further challenges from within KWP” (Buzo 1999, 34). The
party leadership had been transformed into a rubber stamp.
The variable appointments assesses the dictator’s control over appointments
to important offices in the government, military, and ruling party. To code this
item, we rely on secondary literature such as this statement about Mobutu Sese
Seko, dictator of what is now the Democratic Republic of Congo:
State-party personnel are completely dependent on him for selection, appointment, and
maintenance in power . . . Mobutu constantly rotates the membership of the highest
organs of power. (Callaghy 1984, 180)
The variable new party identifies country-years in which autocracies organ-
ize new ruling parties. We consider the dictator’s creation of a new support
party a strategy for adding personal loyalists to the dictatorial inner circle.
Bringing new members into the dictatorial elite dilutes the power of existing
members (usually those who helped seize power) and increases the weight of the
faction supporting the dictator. We code a new support party if the dictator or a
close ally created a new party after the seizure of power or, in a few cases,
during an election campaign before authoritarianization.14 When a dictator
organizes a ruling party, he chooses its leadership. Such parties rarely develop
sufficient autonomous power to constrain the dictator.
The last two indicators of personalism assess the dictator’s relationship with
the military: whether he promotes officers loyal to himself or from his tribal,
ethnic, partisan, or religious group (promotions) and whether he imprisons or
kills officers from other groups without fair trials (purges). Dictator-controlled
promotions and purges demonstrate the dictator’s capacity to change the

14
Once the dictator or close ally creates a support party, all subsequent years for that leader are
coded as having a new party that was created to support the dictator.

command structure of the military, and thus the composition of military

decision-making bodies. If the dictator can control the composition of the
officer corps, the military cannot make credible threats to oust him if he fails
to share power.
There is substantial overlap among these indicators, both because dictators
who use one strategy for concentrating power in their own hands often use
others as well and because one piece of historical information can sometimes be
used to code more than one indicator. The information that Saddam Hussein
appointed relatives to the military high command and to head the security
apparatus and Republican Guard, for example, demonstrates that he controlled
the internal security forces (security apparatus), personalized a paramilitary
force (create paramilitary), and controlled military promotions (promotions).
Using these eight indicators, we create a composite measure of personalism
from an item response theory (IRT) two-parameter logistic model (2PL) that
allows each item (variable) to vary in its difficulty and discrimination.15 We
transform the scores from this latent trait estimate into an index bounded by
0 and 1, where higher levels of personalism approach 1 and lower levels of
personalism approach 0.
The personalism index differs from the categorical regime-type variables
used in past research (e.g., Geddes 2003). The old measure classified differences
across regimes (that is, spans of consecutive country-years), but the new one is
coded every year in every regime to measure changes over time in the dictator’s
concentration of power. Importantly, this time-varying measure is coded for all
dictatorships – not just those with powerful leaders – to capture differences
between regimes, between leaders in the same regime, and over time during any
individual leader’s tenure in power. This allows us to investigate the gradual
concentration of power by individual dictators (the most common pattern) as
well as occasional reversals.
Figure 4.2 shows the personalism scores for six dictatorships: three com-
munist dictatorships in Asia, all identified by Levitsky and Way (2013) as
revolutionary regimes, and three coded as hybrid regimes by Geddes (2003)
and not considered revolutionary. This latter group has features of personalist
rule as well as features of military and party-based rule, which is the reason they
were coded as hybrids of the three pure types. The personalism scores show
how the concentration of power in the hands of paramount leaders varied over
time in these long-lasting dictatorships.
In the left panel, we see that the personalism score for China (solid line)
reflects the ups and downs in power concentration since the Chinese Commun-
ist Party took power in 1949. The highly collegial communist leadership
immediately after the revolution was followed by the modest concentration of

15
As an alternative, we used principal component analysis (PCA) to extract the ﬁrst dimension
from these variables. This factor, with an eigenvalue greater than 3, is correlated with the
personalism index employed throughout at 0.99.

Personalism in 3 communist regimes Personalism in 3 hybrid regimes

1 1

0.8 0.8
Personalism score

Personalism score
0.6 0.6

0.4 0.4

0.2 0.2

0 0
1950 1960 1970 1980 1990 2000 2010 1950 1960 1970 1980 1990 2000 2010
Year Year
China North Korea Vietnam Indonesia Paraguay Syria

figure 4.2 Illustration of personalism scores.

power in Mao’s hands during the 1950s, Mao’s loss of power after the failure
of the Great Leap Forward, and then Mao’s rapid concentration of power in
the 1960s during the Cultural Revolution.16 The concentration of power
dropped again as Mao’s health failed and normalcy was reestablished in the
1970s. Chinese leadership then became unusually collegial again except for a
short time during Deng Xiaoping’s dominance.17 The North Korean regime
(dashed line) was moderately personalist during its first years in existence under
Soviet tutelage, but Kim Il-sung dramatically increased his control relative to
other ruling party elites in the late 1950s and even more during the 1960s.18
His son and grandson maintained the extreme personalization of power in
North Korea.
In contrast, Vietnam’s (hatched line) first dictator, Ho Chi Minh, never
concentrated great power in his hands. The uptick of personalism in the late
1960s reflects the exclusion of the faction opposed to escalating the war in the
south as Lê Duẩn consolidated his position as successor to Ho Chi Minh in the
years before Ho’s death (Vu 2014, 28–30). The larger surge in personalism
scores from the mid-1970s to mid-1980s reflects the post-war consolidation of

16
Changes in the graph appear slightly later than real-world changes because the data are coded as
of January 1, meaning that if a dictator increases the concentration of power in his hands in one
year, it will first appear in the data set the following year.
17
As we write, China specialists suggest that the current leader, Xi Jinping, is concentrating greater
power in his hands. (See, for example, Lee, 2015.) The data on which the personalism score
depends extend only to 2010, however, so Xi’s term is not shown.
18
To reflect this, the Communist regime in North Korea was classified as a hybrid, dominant-
party-personalist, in the old regime-type coding (Geddes 2003).

power by Lê Duẩn and a handful of close supporters at the expense of other

veteran communists. Lê Duẩn’s appointment of his son to head the secret
police, appointment of other relatives to other important posts, and the purge
of pro-Chinese factions from the party elite that enabled him to control promo-
tions in the military demonstrate his concentration of power (Nguyen 1983,
70–72). The decline in Vietnamese personalism in 1986 coincides with Lê
Duẩn’s death.
In the right panel, we note first that the early years of the dictatorships in
Syria and Paraguay show a pattern similar to that in North Korea and China
while Mao was alive: an initial period of relative collegiality followed by the
rapid concentration of power in the dictator’s hands. This is the average pattern
we find in the data. Note that in all three of these hybrid regimes, first dictators
began their rule with more power than any of the communist leaders because
none of them had to bargain with well-developed, unified party institutions.
Suharto of Indonesia (solid line) faced remarkably little constraint from
other political elites even at the beginning. Not only did he lack a support party
until creating one a few years after seizing power, but the upper ranks of the
officer corps had been decimated by assassinations during a violent uprising not
long before the coup that brought him to power. Consequently, Suharto did not
have to negotiate with other highly ranked officers as most dictators from the
military do. The precipitous drop in the personalism score for the last year of
this dictatorship reflects Suharto’s resignation and his replacement by a weak
protégé who lacked a support base in the military.
In Paraguay (dashed line) and Syria (hatched line), military dictators allied
with preexisting but highly factionalized parties, which they reorganized,
purged, and molded into effective instruments of personal rule over their first
years in power. The Colorado Party in Paraguay was especially useful because,
although hopelessly factionalized at the elite level and thus unable to exert
much constraint on the dictator, it had a well-organized mass base that Stroess-
ner coopted to use for both spying and mobilization against challenges from
fellow officers. The steep drop in personalism followed by reconcentration in
the 1980s reflects the development of competing factions within the Stroessner
ruling group as he aged and members of the inner circle began to battle over
succession. Stroessner briefly reasserted himself, but was then ousted by a
member of his inner circle. The dramatic decline shortly before the regime
ended reflects the tenure of this less powerful successor.
In these cases, the dictator who concentrated great personal power lived for
several decades longer and maintained a high level of personalization, followed
by a precipitous drop when he died, resigned, or was ousted. In Syria, we see a
drop in personalization when Bashar al-Assad succeeded his father in 2000 and
then an upswing as Bashar consolidated his hold on the system. In the other
two, successors were unable to reconcentrate personal power or maintain the
regime. In all cases, the personalism score seems to track historical events well.

Importantly, these time-varying indicators of personalism reﬂect observed

behavior after the seizure of power. The indicators of factionalism within the
seizure group that we employ below in the analysis to explain personalism
reﬂect characteristics of what later became the seizure group measured prior to
the seizure of power. They are thus not endogenously determined by dictators’
strategic behavior aimed at remaining in power.

patterns of personalism
In this section, we use our measure of personalism to describe typical patterns of
power concentration in dictatorships. Personalism scores tend to be low during
the year after seizures of power, as would be expected if most dictators are weak,
like Karimov, relative to other members of the ruling group when dictatorships
begin. Where the dictator can take first steps toward power concentration soon
after seizures of power, however, we expect him to then use his increased
resources to eliminate from the inner circle individuals who have the greatest
ability or disposition to challenge him in the future – as Karimov did. In this way,
he can concentrate more power in his own hands. This strategy is associated with
longer tenure in office for the dictator. First dictators – that is, those who are the
first to assume power after the initiation of dictatorship – with higher personal-
ism scores during their first three years in office retain their positions nearly twice
as long on average as first dictators with low early personalism scores. First
dictators with high personalism scores (top third of the personalism index)
during the first years in power survive 14.7 years on average, while first leaders
with low personalism scores (bottom third) survive only 7.8 years on average.19
Changes in the relative power of inner-circle members can become long
lasting through the replacement of individuals who might potentially have
challenged the dictator with others who lack independent support bases and
are thus more dependent on him. Whatever resolution arises from the earliest
conflict between the dictator and his closest allies increases the likelihood of a
similar resolution to the next one. In other words, if the dictator gains more
control over political resources as a result of the first conflict with other
members of the inner circle, he then has a greater advantage in the next conflict
with them.20 In this way, where steps toward personalization occur soon after
the seizure of power, it is likely to progress further.
In contrast, initial reliance on collegial institutions reduces the chance of
later personalization. Where members of the inner circle have developed the
expectation of participating in key decisions, attempts by the dictator to reverse

19
The median tenure for first leaders with low personalism scores is 4 years; for those with high
personalism scores, the median is 11.5 years. Note that this comparison pertains only to first
leaders who survive at least three years in office, since scores during the first three years were
used to create the comparison groups.
20
Svolik (2012) shows how greater leader power enables the dictator to defeat regime insiders who
try to use rebellion to deter further power grabs.

their policy choices or postpone regular meetings of the collegial decision-

making body become focal points around which it is relatively easy (though
never easy in absolute terms) to organize collective action against the dictator.
In short, we expect the deal agreed to by the dictator and members of the
seizure group in the early months after the seizure of power to shape later
interactions. Whatever pattern of power aggrandizement is established during
the first years of a dictatorship tends to be perpetuated until the first dictator
dies, sickens, or is overthrown.
The replacement of one dictator by another during a single regime often
involves renegotiation of the distribution of power within the inner circle. Those
who yearn to replace the dictator, whether after his death or via violent over-
throw, must promise their colleagues a larger share of power in order to attract
their support, but as with the first dictator, such promises are unenforceable
unless members of the inner circle can both oust him if he reneges and credibly
commit their subordinates to refrain from overthrowing him if he sticks to the
bargain. The main differences between subsequent struggles and the first one is
that members of the inner circle have learned from earlier struggles and may have
developed disciplined, within-regime networks that allow them to bargain more
credibly and effectively with the new dictator. In dictatorships that last beyond
the tenure of the first leader, power relationships between the dictator and the
inner circle thus tend to become somewhat more equal under subsequent leaders.
We assess these expectations in two ways. First, we examine how levels of
personalism change over time for a regime’s first leader relative to subsequent
ones. The first dictator has an advantage in bargaining relative to later ones
because of the inexperience of members of the ruling group. That is, they are
experienced as military officers, insurgents, or party militants, but they have not
usually had experience in the rule-free and dangerous context of bargaining within
the inner circle of a dictatorship. The day after the new regime seizes power, the
new dictator can begin using state resources, appointing officials, establishing
procedures, and issuing decrees. It can take some time for other members of the
inner circle to grasp all the implications of some of the dictator’s initiatives. This
implies that, on average, first dictators have advantages over later ones in
parlaying initial gains in personal power into further increases over time.
Figure 4.3 compares levels of personalism over time for initial regime leaders
with the personalism scores of subsequent dictators in the same regime. The
horizontal axis marks the first three full years each dictator rules, while the
vertical axis shows the predicted level of personalism from a regression model
in which the dependent variable is the measured level of personalism:

Personalism ¼ α0 þ β1 ∗ Durationt þ β2 ∗ FirstLeaderi

þ β3 ∗ ðDurationt ∗ FirstLeaderi Þ þ γt þ εit ð4:1Þ

In this equation, Duration is the natural log of leader years in power; First-
Leader is a binary indicator of whether the dictator is the ﬁrst one after the

Pooled data Leader fixed effects

First leader Subsequent leader First leader Subsequent leader
0.15 0.15
Personalism index

Personalism index
0.1 0.1

0.05 0.05

0 0
1 2 3 4 1 2 3 4
Leader time in power (years) Leader time in power (years)
figure 4.3 The ﬁrst dictator’s advantage in personalizing power.

regime seizes power; γt are five-year time period effects; and i indexes leader
and t indexes years in power. Time-period effects ensure that measurement
error correlated with historical time is not driving the estimates of interest.21
The figures below report the substantive effect of the linear combination of
β2 and β3, which estimates levels of personalism for first regime leaders; and
of β1, which estimates levels of personalism for subsequent leaders. By design,
the approach simply shows the average levels of personalism as years in power
increase, setting the average level in the first year as the baseline (set to zero on
the vertical axis). The right panel of Figure 4.3 shows the result from a similar
model specification but adds one crucial set of controls: an individual-level
fixed effect for each dictator (δi). This allows the model to isolate the changes
over time for each leader, net of any baseline differences between leaders in
different countries or regimes.
The left panel shows the pooled data. During his first three years in office,
the first regime leader, on average, increases the level of personalism nearly
0.15 points on the (0, 1) scale. For subsequent regime leaders, the gains in
personalism are less than half of this. The right panel shows a similar pattern:
the first leader increases personalism in the first three years by almost 0.09
points, while later leaders increase it by less than 0.02 points. The size of these
effects is smaller in the right panel because the model accounts for all

21
If we can better observe manifestations of personalism for more recent periods (e.g., post-2000),
for example, this could systematically bias estimates.

First leader advantage Initial gains advantage

First leader Subsequent leaders High initial gains No initial gains

0.08 0.08

0.06 0.06
Personalism index

Personalism index
0.04 0.04

0.02 0.02

0 0
4 5 6 7 8 9 10 4 5 6 7 8 9 10
Leader time in power (years) Leader time in power (years)
figure 4.4 Personalizing power after the ﬁrst three years.

differences in the average level of personalism for each individual leader

and thus looks only at each dictator’s time-trend in personalism. The descrip-
tive patterns in Figure 4.3 are consistent with the expectation that first
regime leaders have advantages in accumulating power. In the early years of
dictatorship, the first dictator concentrates power faster than leaders who
follow him.
Next we examine what happens after an initial three-year period of regime
consolidation. In this exercise, we look at how personalism evolves in years
four to ten for different groups of dictators. Figure 4.4 shows the pattern of
personalism across all dictators who survive in power at least four years, using
the average level of personalism at the start of year four as the baseline level (set
to zero). The left panel of Figure 4.4 reports the estimates for a model similar to
Equation 4.1 but with leader fixed effects (like the right panel of Figure 4.3).
After a period of initial power concentration, first leaders continue to have an
advantage in accumulating power relative to subsequent leaders: first dictators,
on average, boost personalism scores by a further 0.07 points between years
four and ten, while subsequent leaders increase their power by half as much
during these years. These increases are smaller than those reported in Figure 4.3
because power concentration is more rapid during initial periods of regime
consolidation than during later years.
Another way to analyze personalization after the initial period of consoli-
dation is to compare leaders who successfully personalized during the first three
years with those who did not. For each leader, we construct a variable that
measures whether his score on the personalism index increased a lot during the

first three years he ruled (high initial gains).22 Then we test a regression model
similar to those used above (with leader fixed effects), but compare leaders with
high initial gains in personalism scores with leaders who did not concentrate
power during their first three years. The right panel of Figure 4.4 shows that
leaders who concentrate personal power in their first three years further
increase their personalism scores by more than 0.06 points from years four to
ten. Leaders who failed to accumulate personal power in their first three years
do not make this up later; their gains after the initial period are smaller than
those of leaders who amassed personal power from the outset. This evidence is
consistent with Svolik’s (2012) model of power concentration in which initial
successful power grabs beget more successful power concentration over time.

the effect of factionalism on the

personalization of power
Next we test our explanation of why some dictators can concentrate more
power than others. We look at the effect of the degree of factionalism in the
seizure group before the initiation of dictatorship on how personalism evolves
over time after the group takes power. We expect more unified seizure groups
to bargain more successfully, and thus to limit the accumulation of power in the
dictator’s hands.
Though we lack direct measures of seizure-group unity, we investigate the
effects of two proxy measures. The first is the pre-seizure history of the group
that becomes the dictatorship’s ruling party after the seizure. We posit that a
support coalition organized as a political party either to contest elections or to
lead a revolution before the seizure of power has greater organizational unity,
and can thus more successfully bargain with the dictator, than a support
coalition not organized as a party before the seizure.23
The second proxy for coalition unity is intended to capture the pre-seizure
unity of military seizure groups. For dictatorships that seized power in coups,
we use the first dictator’s military rank before the seizure of power as a proxy
measure of factionalism. The logic is as follows. Junior and mid-level officers
carry out many coups. In countries with relatively unified and disciplined
military forces, however, lower-ranked coup leaders hand regime leadership
to a senior officer after the coup because they do not expect other senior officers

22
Of the 312 leaders who last more than three years in power, 28 percent have high initial gains
during their ﬁrst three years in power. About half (53 percent) of the 312 leaders are ﬁrst regime
leaders.
23
Parties organized to contest elections ran candidates in one or more elections in an earlier
authoritarian or democratic regime. We do not include the parties that were organized as
vehicles for the dictator’s election campaign prior to authoritarianization (e.g., Cambio 90,
organized by Fujimori to manage his presidential campaign in 1990). Vehicle parties are centered
on the leader from their creation and often have little independent organizational existence.

to follow orders issued by junior ofﬁcers. In factionalized armies, however,

multiple hierarchies exist, some of which junior officers lead. When senior
officers lead dictatorships, we cannot be sure whether the military that backs
them is unified, but when junior officers such as Captain Moammar Qaddafi of
Libya or Sargent Samuel Doe of Liberia lead dictatorships, we know that the
military that backs them was factionalized before the coup. The indicator we
use here groups regime leaders ranked major and below in one category and all
those ranked higher in the other.24 It thus distinguishes the most factionalized
cases (only the top 9 percent) from all others.
Seizure of power via popular uprising is another indicator of a factionalized
army. Because popular uprisings are defined as unarmed seizures of power, they
occur only when the country’s military has refrained from using its advantage
in violence to quell the upheaval. When the army is united, either it backs the
incumbent to prevent popular demonstrations from ousting him or it replaces
the incumbent itself. The overthrow of a government by popular uprising
suggests an army divided between government supporters and opponents just
before the regime change, and possibly along other dimensions as well.
We can thus use both seizure of power by popular uprising and the first
dictator’s rank before seizure via coup as indications of pre-seizure factionali-
zation in the military coalition that supports the dictatorship.
The data indicate that more than one-third (37 percent) of dictators are
supported by an inherited revolutionary or electoral party when they seize
power (a sign of unity). Only 15 percent of dictators are supported by a highly
factionalized military.25 These two features are almost mutually exclusive: of
the 280 regimes in the data, only seven have both an inherited party and a
factionalized military.
We expect leaders who bargain with supporters organized in inherited
(electoral or revolutionary) political parties, which should on average be more
united than newly created parties or informally organized coalitions, to be less
capable of personalizing power. By the same logic, leaders who bargain with a
more unified military should be less likely to concentrate power in their own
hands than those who negotiate with a more factionalized officer corps.

24
Colonel was the highest rank in many armies at the time of coups, especially in the 1960s and
1970s, so the meaning of this rank is ambiguous.
25
Note that each of these measures creates a more homogeneous set of cases on one side of the
dichotomy than the other. Dictators in cases with inherited parties must all bargain with inner
circles organized by inherited parties. In the cases that lack inherited parties, some dictators
bargain with newly created parties that tend to lack organizational coherence, but others lack
parties and bargain instead with officers from a disciplined, professional officer corps. In other
words, the cases that lack inherited parties are a mix of cases with factionalized ruling groups
and cases with united ruling groups. Our proxy measure of military factionalism implies that the
militaries identified here as less factionalized include a mix of factionalized and unified military
forces, so we can show the effect of high factionalism but not of high levels of military unity.

Inherited party Not inherited Factionalized military Not factionalized

0.25 0.25

0.2 0.2
Personalism index

Personalism index
0.15 0.15

0.1 0.1

0.05 0.05

0 0
2 4 6 8 10 2 4 6 8 10
Leader duration (years) Leader duration (years)
figure 4.5 United versus factionalized seizure groups.

A first look at the raw data provides some evidence consistent with these
expectations. While roughly 41 percent of first leaders increase personal power
in their first three years, this figure differs considerably depending on the type of
group with which the leader bargains. Less than one-third (32 percent) of first
dictators who face an inherited support party concentrate power in their own
hands, but nearly half (47 percent) of those who do not bargain with an
inherited party do so. Fifty-four percent of first dictators who face a highly
factionalized military concentrate power, while only 38 percent of those who
face a more unified military do so.26
The left panel of Figure 4.5 compares the increase over time in personalism
levels for first leaders who bargained with an inherited party (a more unified
support coalition) and those who bargained with a new party or with support-
ers not organized into a party.27 First leaders who do not bargain with an
inherited party personalize by 0.098 points on average during the first three
years; those who negotiate with an inherited political party, however, increase
their personalism score by much less (0.053 points). By the end of the first
decade in power, those who do not face an inherited party have increased
personalism scores by 0.163 points, and those who bargain with an inherited
party do so by a little more than half that amount (0.089 points). This suggests

26
These differences for inherited party and factionalized military are both statistically significant at
the 0.05 level. Figures are for first regime leaders, excluding subsequent ones.
27
The estimates on the vertical axis reflect the predicted change in the level of personalism from a
regression model similar to Equation (4.1) but with leader and year fixed effects.

that dictators who must bargain with a unified support group can less easily
concentrate personal power than leaders whose supporters are less organized.
The right panel of Figure 4.5 shows the average level of personalism for first
leaders who face a highly factionalized military and those who do not. We see an
even stronger pattern: those who bargain with very factionalized militaries boost
personalism by 0.14 points in their first three years, while those who bargain
with more united militaries increase personalism by only 0.07 points, on aver-
age.28 After a decade in power, dictators facing more unified officer corps have
increased their power by 0.12 points, on average, while those whose military
supporters are more factionalized increase it by twice as much (0.23 points). We
interpret this evidence as suggesting that leaders who initially bargain with a
highly factionalized military have a clear advantage in personalizing power.
To summarize, our empirical analysis shows that first dictators have an
advantage in concentrating power in their own hands relative to later ones.
Moreover, initial gains in power tend to make further gains easier. Dictators
whose efforts to grab power are blocked early on are likely to concentrate
power more slowly if at all later, even if they survive long in office.
We also show that dictators who have to bargain with a more united seizure
coalition face stiffer resistance to concentrating personal power than those who
do not. Inherited parties, we show, limit dictators’ gains in personal power,
as do more unified militaries. In contrast, dictators who bargain with factional-
ized military or civilian supporters have great advantages in concentrating
personal power.

conclusion
All members of the inner circles of dictatorships have common interests in
regime survival but compete with each other over power and resources. Each
individual member, including the dictator, has strong reasons to try to increase
his power and access to resources at the expense of the others. Even if some
individual members do not yearn for the dictator’s job, they must compete in
order to maintain their positions against ambitious regime supporters below
them in the hierarchy. The competition within the inner circle means that for
most purposes dictatorships should not be analyzed as unitary actors. Instead,
we see members of the inner circle as continuously engaged in simultaneous
cooperative strategies aimed at regime survival and noncooperative strategies
aimed at increasing personal power.
The choice of one member of the inner circle as dictator (or the elected
leader’s acquisition of dictatorial powers if the seizure of power is accomplished
via authoritarianization) results in the central political dynamic of authoritar-
ian politics: conﬂict over the distribution of power within the regime’s

28
Remember that “more united” is a mix of fully united and fairly factionalized since we have no
indicator for fully uniﬁed.

leadership group. Once the dictator is chosen, his interests diverge from those
of his lieutenants. Dictators who had been first among equals in a collegial
conspiracy before the ouster of the old regime gain reasons to concentrate
power and resources in their own hands in order to increase their security at
the top. Other members of the inner circle, meanwhile, have good reasons to try
to limit the dictator’s resources and policy-making discretion in order to protect
their own positions and maintain their own influence and clientele networks.
This conflict plays out in different ways, depending on the ex ante factionalism
of the seizure group.
Characteristics of the seizure group that pre-date the establishment of the
dictatorship influence the initial distribution of resources within the inner circle
and what kinds of bargains can be enforced among them. In this chapter we
focused on one ex ante characteristic: the unity or factionalism developed
within the seizure group before they gained power. By unity, we mean that
members of the inner circle can bargain with the dictator as a unitary actor and
thus drive a harder bargain with him. Some military forces are unified by virtue
of enforced discipline and the hierarchical command structure, but others are
factionalized. The same goes for parties. Where the dictator’s supporters are
divided into factions, they are unlikely to be able to make credible threats to
oust the dictator if he fails to share power and spoils. Where, however, they can
behave as a unitary actor, they can more easily act together to oust him and
thus the dictator’s promises to share are credible.
The focus on bargaining highlights the logic behind coalition narrowing in
dictatorships. We argue that if the members of the seizure group have been able
to develop ways of enforcing their own internal unity, dictators’ efforts to
concentrate power tend to fail. In the real world, enforced internal unity
develops in professionalized military forces and highly disciplined parties.
In contrast, where a conspiracy drawn from a factionalized officer corps
or party seizes power, the dictator’s supporters often fail to resist the personal-
ization of rule.29
Authoritarian regimes differ enormously from each other in levels of repres-
sion, distribution of costs and benefits across societal groups, policies followed,
and ideological justification. Nevertheless, the impulse toward personalization
seems to be common in all. The elite bargaining described in this chapter
explains why these processes occur in such apparently different kinds of
dictatorship.
In this chapter we have focused on how the factionalism or unity of the
seizure group affects its ability to oust the dictator and thus the credibility of the
dictator’s promises to share power and spoils. In the next, we focus not on what

29
Analysis in the replication ﬁles shows that dictatorships in the past two decades are increasingly
likely to seize power with a factionalized military and less likely to seize power with an inherited
political party. Together, these trends may explain why we observe an increase in personalist
regimes since 1990 (Kendall-Taylor, Frantz, and Wright 2016).

makes the lieutenants’ threats to oust the dictator if he fails to share more
credible but rather on whether their promises to refrain from ousting him when
he shares are credible. Allies’ promises of support are credible only if their
subordinates obey orders because dictators can be overthrown or assassinated
by small numbers of armed men. We then consider the options available to the
dictator when the promises of support from other members of the ruling group
are not credible. In the process, we explain why seizure groups that did not need
an organized civilian support base to achieve power sometimes later create
mass parties.

Dictatorial Survival Strategies in

Challenging Conditions
Factionalized Armed Supporters and Party Creation

When Benin became independent in 1960, three parties dominated politics,

each rooted in a region and its ethnic groups. The same regional loyalties
factionalized the newly created army. Most officers came from the south. They
had started in the ranks during French rule and were promoted rapidly in order
to indigenize the officer corps at independence. Most enlisted men came from
the north (Decalo 1976, 55–57). As of 1965, when Benin’s second coup
occurred, the army was a few years old and had only 1,700 men. It had
43 indigenous officers and 12 French ones (Bebler 1973, 12–13). During the
first decade after independence, divisions between older officers, who were
rapidly promoted as the new nation built its army, and younger ones trained
in military schools, whose promotions were soon blocked by budget crises,
reinforced and overlaid regional factionalism within the army (Decalo 1976).
After the first coup in 1963, soldiers replaced the civilian president with a
different civilian and returned to the barracks. Officers themselves took power
after the1965 coup. Factionalism undermined Benin’s first military dictator-
ship, which ended with a coup in 1967 that brought to power a new military
regime of mid-ranking officers from a different ethnic group. The second
military dictatorship attempted to deal with army factionalism by appointing
an ethnically balanced cabinet. The new cabinet included captains, lieutenants,
and NCOs in the dictatorship’s inner circle to make sure that all army interests
were represented. The leadership also dismissed some southern senior officers
and promoted some from the north in an effort to equalize opportunities
(Bebler 1973, 20–23). Nevertheless, another faction-based coup ousted this
regime two years later. Officers then tried to reunify the military around an
inclusive decision-making body with collegial leadership, but were unsuccess-
ful. They returned power to civilians in 1970 because they could not find a
successful formula for power sharing within the officer corps.

Benin’s history in the 1960s is a story of military dictators’ repeated failure

to consolidate their rule because the factionalized officer corps could not
provide stable support. Decalo describes the army as a “patchwork of compet-
ing personalist/ethnic allegiance-pyramids centered around officers of all ranks
in which superior rank or authority was only grudgingly acknowledged”
(1979, 234).
A seizure group that includes many members with control over armed force
should be able to achieve an especially advantageous power-sharing arrange-
ment with the leader because their threats to oust the dictator are highly
credible. Responding to these threats, early military dictators in Benin agreed
to oversight by broadly representative groups of officers, but to no avail. These
strategies failed because officers included in the regime’s inner circle could not
prevent rogue coups led by other officers formally subordinate to them. Despite
successive dictators’ efforts to secure stability by consulting with representatives
of many military factions, some officers always remained dissatisfied and quick
to oust the current dictatorship. These failures were caused not by the inability
of dictators’ allies to make credible threats to oust the dictator if he fails to
share, but rather by their inability to make good on promises of support when
he was sharing. In that situation, a dictator cannot make himself safe by sharing
more and more because dictators, like all other political leaders, face budget
constraints.
This destructive game of musical chairs among military factions ended a few
years after the 1972 coup that brought Major Mathieu Kérékou to power.
Kérékou initially followed the same strategy as earlier military dictators. He
dismissed all senior officers, appointed a cabinet of junior officers, and con-
sulted a military ruling council representative of major military factions. Des-
pite Kérékou’s effort to consult all factions, the military remained unable to
provide a stable base of support. Kérékou survived several coup attempts
during his first two years in power (Decalo 1976, 76–84). Then, in 1974
Kérékou began creating an organized civilian support base, the Benin People’s
Revolutionary Party (RPB), to counterbalance the military and help stabilize his
rule. Six “close friends of the president” made up the new party’s politburo
(Martin 1986, 68). Over time, Kérékou gradually increased civilian participa-
tion in inner-circle decision-making. Their support enabled him to remove the
most threatening rival officers from posts from which armed challenges could
be launched (Martin 1986, 68–75). Adding a loyal civilian support group to
balance the factionalized military stabilized Kérékou’s rule. He retained control
until 1990, when widespread popular opposition forced him to democratize.

In this chapter, we explore two theoretical ideas highlighted by Benin’s history.

First, armed supporters can drive a hard bargain with the dictator when they
maintain unity but not otherwise; factionalism prevents them from making
credible promises to support the dictator if he shares. To be viable, a power-
sharing bargain between the dictator and his allies must be credible on both

sides: allies must believe the dictator’s promise to share, and the dictator must
believe the allies’ promises to support him if and only if he shares. The allies’
promise of support is not credible, however, if they cannot commit their own
subordinates to honor the bargains they make. The second theoretical idea is
that dictators can in some circumstances change the balance of forces within the
ruling group by empowering new political actors. Kérékou did this when he
organized a new support party and brought its leaders into the inner circle of
the dictatorship.
In Chapter 4, we analyzed how and why dictators exclude rivals from their
inner circle. Here we focus on the dictator’s strategic inclusion of new players
he expects to be more malleable and less dangerous than the original members.
We also consider the conditions under which he has the ability to exercise
this option.
In the first part of this chapter, we describe the especially difficult situation
facing dictators whose power initially depends on an armed support base, or
what Alex De Waal (2015) calls specialists in violence. In the next, we spell out
our argument about how the interaction of dispersed armed force and faction-
alism in the seizure group can limit the dictator’s options for bargaining with
these supporters.1 The third section considers the dictator’s incentives to bring
into the inner circle political actors who are more dependent on him for benefits
and protection than are specialists in violence (military officers or others who
command armed subordinates). It explains why civilian support groups are less
dangerous to dictators than armed factions, and it discusses party creation as a
strategy for mobilizing organized civilian support to counterbalance unreliable
armed supporters. The fourth section provides evidence that party creation in
dictatorships that lack support parties when they seize power contributes to a
strategy designed to concentrate more power and resources in the dictator’s
hands, reducing his need to rely on members of the original seizure group. The
last sections show evidence that fewer coups against dictators occur after party
creation and that dictatorships that create parties post-seizure last longer than
otherwise similar regimes lacking party organization.

the strategic context

Dictators who achieve power through armed force face an especially difﬁcult
survival problem. Consider the dilemma faced by a dictator who comes to
power in a coup.2 Besides the support of fellow plotters, he probably has the

1
De Waal (2015) focuses on factions based on ethnicity and cases in which all relevant players are
specialists in violence. We analyze the more general situation in which the seizure group is divided
into factions for any reason and some potentially powerful political players lack easy recourse to
violence.
2
A large majority of the dictators who achieve power by force do so via coup, but sometimes
ofﬁcers are handed power during popular uprisings; insurgencies bring some armed groups to

initial support of some civilian elites who were fed up with the ousted incum-
bent but who would lack the capacity to save the new leader from ouster if his
armed supporters turned against him.
At the outset he often has quite a bit of popular support as well, and for the
same reason: dissatisfaction with the overthrown incumbent. This popular
support, however, is unorganized, opportunistic, and superficial. “The people”
or “the street” may support a coup in order to rid themselves of an incompe-
tent, brutal, or disreputable leader, but that support will evaporate if the
economy fails to improve quickly or if any other disaster strikes. Much of the
citizenry can swing rapidly from support to opposition. Egypt’s recent experi-
ence illustrates the volatility of unorganized popular support: Egyptian protest-
ors forced a military-supported dictatorship from power in 2011, but in
2013 helped oust a democratically elected president and return the military
to power.
The coup leader needs widespread support from other officers to survive in
power when popular opposition arises, as it inevitably does. Military dictators
have the support or acquiescence of many other officers at the time of seizure
(otherwise their coup would have failed), but no way of guaranteeing that
support in the future. Even at the beginning, support from the rest of the
military might be quite superficial. As shown in earlier work (Geddes 2003),
the incentives facing military officers when their colleagues initiate a coup
create a first-mover advantage similar to that in “battle of the sexes” games.
This means that if a small group of officers makes a successful first “coup
move,” such as seizing the airport and presidential palace, the rest of the
military tends to go along, whether they sincerely want the intervention or
not. So, the fact that a seizure of power has occurred implies temporary
acquiescence by the rest of the officer corps but very little in terms of sincere
or long-term support. In short, military dictators cannot count on the support
of fellow officers tomorrow even if they have it today, and today’s support may
be shallower than it appears at the time the military seizes power.
Within the dictatorial inner circle, support for the dictator may also be short
lived. Even though the dictator was a brother officer – and often a longtime ally
and friend of other high-ranking officers – the day before he was selected as
supreme leader, their interests diverge after his selection just as those of civilian
dictators and their supporters do. Officers in the support group want to ensure
large military budgets and their continued monopolization of force. They do
not want their budget reduced in order to hire more security police or provide

power; and a few are imposed by foreigners. After insurgencies or foreign impositions, the new
ruling group usually replaces the incumbent officer corps with its own supporters. Thus, when we
speak of officers, we refer not only to the armed forces of an ousted regime that achieved power
for themselves via coup or popular uprising, but also to those who replaced incumbent officers
after insurgent seizures of power. Once the old officer corps has been replaced, “the military” is
the new one imposed after regime change.

patronage jobs for civilian supporters of the dictator. They oppose the creation
of presidential guards or people’s militias to protect the dictator because such
forces challenge their own monopoly of force and the credibility of their threats
to oust the dictator if he fails to share, as well as depleting the budget. They
have an interest in consultation within the officer corps before major policy
decisions. They want to avoid the concentration of power, resources, and
discretion in one man’s hands, as well as the favoritism and deprofessionaliza-
tion within the military that often accompany it. Dictators who succeed in
concentrating resources and discretion in their own hands, that is, in personal-
izing power, threaten both the military as an institution and individual officers,
since they control promotions, postings, forced retirements, and access to profit
opportunities (both legal and illegal). At the extreme, such dictators control life
and death through their personal control of the security apparatus, and military
officers have no special immunity from security police. The same logic applies
to other dictators brought to power by specialists in violence.
Since coups have ousted most dictators, especially those who come from the
military, it is obvious that officers can be dangerous. Weapons, know-how, and
the command of troops are widely dispersed in armies and in some insurgen-
cies. This reality creates a hazardous environment for dictators, because coups
require only a small number of individual plotters to execute them. Indeed,
coups involving fewer than twenty men have occasionally succeeded.3 Only a
minority of coups involve consensus among the whole officer corps. Instead,
small conspiratorial groups of officers carry out most coups using the first-
mover strategy (Nordlinger 1977). In short, many different small groups of
officers could stage a coup with reasonable prospects for success.

the interaction of dispersed arms and factionalism

Though much of the literature on autocracies has emphasized dictators’ cred-
ibility problem,4 it is not the main impediment to successful sharing bargains
when arms are widely dispersed within the ruling group. Military dictators can
increase the credibility of their promises to share in some ways not available to
civilians, for example, by retiring from active duty.5 Once they no longer
command other ofﬁcers, do not determine promotions, and cannot decide
which ofﬁcers will command which garrisons, leaders’ ability to deter coups
depends almost entirely on their ability to satisfy the rest of the military’s policy

3
For example, Master Sergeant Samuel Doe and seventeen men ended the dictatorship of William
Tolbert and the long reign of the True Whig Party in Liberia (Thomson 1988, 44); sixteen NCOs
carried out the 1980 coup in Surinam that ousted then-President Johan Ferrier (Hoefte
2013, 133).
4
For example, Acemoglu and Robinson (2005); Magaloni (2008); and Svolik (2012).
5
Sometimes ofﬁcers are required by the rest of the military to retire before becoming junta
president for exactly this reason (Remmer 1991).

and budget demands. This makes their promises credible. Military dictators can
also increase their credibility by leaving internal security services within the
military chain of command, thus limiting their ability to spy on, intimidate, and
murder other officers.
The more serious impediment to successful power-sharing bargains is that
armed supporters’ promises not to oust as long as the dictator shares are never
completely credible because they cannot always prevent “rogue” coups or
armed ouster by other specialists in violence.6 These are coups by factions,
often led by lower-ranked officers, that could be defeated if the rest of the
armed forces mobilized against them, but the dictator cannot count on the rest
of the army doing so. The first-mover advantage built into the incentives facing
officers means that a faction that makes a credible first coup move without
being met by violent opposition can overthrow the government because the rest
of the armed forces will acquiesce to this coup just as they did to the one that
brought the current leader to power.7 All dictators face some risk from armed
supporters, but the less control commanding officers have over lower-ranked
officers – that is, the less disciplined and unified the armed support group is –
the less ability officers in the inner circle have to make enforceable bargains
with the dictator that would reduce the risk.
An alternative way to express this problem is to note that militaries and
other armed groups are not unitary actors, though some more closely approxi-
mate unity than others. In more professionalized militaries and unified insur-
gent groups, individuals’ future career success is inextricably bound to
following the orders of their commanders. Professional success depends on
obedience to orders from the day that youths enter military school. Subordin-
ates obey superiors regardless of personal, ethnic, or political loyalties. The
likelihood of a rogue coup succeeding is low in unified militaries, and the cost of
a failed attempt very high because serious breaches of discipline end careers.
Coup leaders can also face court-martial, jail time, or execution. In this kind of
military institution, commanding officers can count on lower-ranked officers to
obey orders, which transforms a large group of individuals into something
approximating a unitary actor. Unitary actors can make credible promises.
Factionalized militaries were common between 1946 and 2010, however,
especially in newly independent countries. A factionalized army, “far from
being a model of hierarchical organization, tends to be an assemblage of armed
men who may or may not obey their officers” (Zolberg 1968, 72). In factional-
ized forces, discipline is less predictably enforced because personal, partisan, or
ethnic loyalties can cross-cut military hierarchy. “[S]oldiers often [have] a

6
Thus, for dictatorships with an armed support base, we see the supporters’ credibility problem as
the opposite of that emphasized by Svolik (2012). He sees their threat to oust a dictator who
violates power sharing as lacking credibility, but we believe it is supporters’ promise to refrain
from ousting a dictator who shares that most often lacks credibility.
7
See Singh (2014) for an elaboration of which actions make the ﬁrst move of a coup credible.

stronger sense of commitment to their unit commander than to the army”

(Crouch 1978, 27). Promotion and protection for lower-ranked officers,
NCOs, and soldiers depend on faction leaders, not just on compliance with
orders and military norms. Routes to higher rank other than the slow but
predictable rise via increasing seniority are more available. Consequently,
lower-ranked officers may disobey the orders of higher-ranked officers if they
conflict with those of faction leaders. Support for coups can be a short-cut to
rapid promotion, and punishment for coup attempts is not always severe. Some
coup attempts result only in demotion, apparently because dictators fear the
consequences of imposing harsher punishments in factionalized militaries.
Others result in dismissal, but when changes at the top occur later, as they
often do in countries with factionalized militaries, dismissed officers may be
reinstated.
Officers included in the inner circle in this kind of setting cannot commit
their subordinates to abide by the sharing bargain as they would be able to in a
more unified military institution because faction leaders command the obedi-
ence of their members but do not offer unconditional obedience to hierarchical
superiors. If factions become dissatisfied with their share or oppose a policy
decision, they have a reasonable chance of bringing off a successful coup.
If the dictator has reason to doubt higher-ranked officers’ ability to prevent
factions from launching rogue coups, it makes little sense for him to share
and consult because doing so will not protect him. Instead, he will renege on
sharing agreements and use the resources saved to pursue other strategies for
deterring coups.

the strategic creation of new political actors

Dictators facing a factionalized armed support base use several strategies for
trying to deter coups and thus reduce their vulnerability. They spend heavily on
the military. They promote loyal officers and retire opponents or appoint them
as ambassadors to faraway places. They resist resigning from active service so
that they can maintain personal control of promotions and command assign-
ments. They strengthen security police and try to take personal control of them.
They create paramilitary forces and presidential guards led by relatives and
recruited from their home regions to counterbalance the regular military.
These strategies show how much dictators fear military supporters. Though
these strategies may sometimes be useful, they can also backfire because other
officers try to defend their own positions of power, which depend on the
credible threat to oust the dictator. Officers resent and sometimes resist policies
that undermine that threat. Promotion of the dictator’s friends over the heads of
competent officers violates military norms and leads to anger. Officers who
have been passed over for promotion, dismissed, jailed, or exiled have led many
coups and insurgencies. Officers may also resent the creation of paramilitary

forces. In other words, these strategies are not risk free. They are highly visible
to officers and may trigger the outcome they were designed to prevent.
Bangladeshi dictator General Mohammad Zia ul-Haq, for example, created
a paramilitary force soon after achieving power and remained active duty chief
of staff for several years to try to control the officer corps, but every time he
intervened in promotions, mutinies broke out. He frequently transferred offi-
cers, sent them for new training, and raised salaries (Codron 2007, 14). Despite
these measures and many executions of rebellious soldiers, he “did not manage
to make the military a safe constituency to back his rule” (Codron 2007, 15).
He began organizing a support party “to create a civilian power base when he
failed to achieve the united support of the armed forces” (Rizvi 1985, 226).
Creating a civilian support organization, as General Zia did, is a subtler and
often safer strategy for counterbalancing the military. Paul Lewis (1980) inter-
prets Paraguayan dictator General Alfredo Stroessner’s reorganization of the
Colorado Party into a support vehicle in exactly this way. After describing
Stroessner’s lavish spending on the military and attempts to manage its factions,
Lewis notes:
Other Paraguayan presidents [also] tried to buy military support and surrounded
themselves with trusted officers. In the end, however, they failed to keep the greedier
or more ambitious soldiers in line. Stroessner . . . achieved a real advantage over his
predecessors [by] fashioning . . . a dominant single-party regime, based on a purified and
obedient mass organization . . . This instrument . . . makes it risky and unprofitable for
[officers] to conspire against him. (1980, 124–25).
Samuel Decalo describes the civilianization of the military regime in Niger
similarly: “Faced with continuous factionalism within both the Supreme Mili-
tary Council (CSM) and his cabinet, Kountché progressively disencumbered
himself of his most threatening . . . officer colleagues” (1990, 277). Kountché
could not afford to challenge other officers openly. “[H]e was forced rather to
‘work his way around them by mobilizing the masses’” (p. 278).8 Egyptian
military dictator Gamel Abdel Nasser established the National Union party “to
strengthen his personal power and weaken the RCC [Revolutionary Command
Council, dominated by the Free Officers]” (Perlmutter 1974, 143).
Organizing civilians aims at reducing the dictator’s dependence on the
military. A civilian support base can change the calculations of potential coup
plotters by reducing their chances of a credible first coup move. Civilian
support for dictators, even if superficial and manipulated, can deter coups
because officers do not want troops on their way to encircle the presidential
palace to confront crowds of fellow citizens. As a Guatemalan officer explained
to an interviewer after an aborted attempt to overthrow a military dictator,
“with civilians standing in front of the artillery tanks, the commander didn’t
want to cause civilian casualties” (Schirmer 1998, 218).

8
The quote within the quote is from Jeune Afrique, April 28, 1982.

Many coups are bloodless. That is because potential coup leaders choose
times when they expect little opposition. Since the Russian Revolution, officers
have understood that asking troops to fire on their fellow citizens can lead to
indiscipline, desertions, and mutiny.9 The military can of course defeat
unarmed or lightly armed civilian demonstrators, but orders to beat or fire on
civilians risk provoking defiance among troops, which would undermine the
military institution and officers’ political power base, so officers exercise cau-
tion in what they demand of soldiers. The strategy of organizing a mass civilian
support base – a new support party – helps dictators survive because of officers’
strong preference for unopposed coups. The “fear of having to deal with
massive civilian opposition” deters military plotting (Brooker 1995, 111).
As an example of how deterrence can work in practice, consider this
sequence of events in Paraguay. Two top officers central to General
Stroessner’s military support base publicly criticized his decision to sign a
treaty, which they claimed would compromise Paraguayan sovereignty. Such
open criticism was extremely rare during Stroessner’s rule. Insiders interpreted
it as a sign of widespread military disaffection, indicating danger of a coup to
replace Stroessner. In response, Stroessner mobilized his civilian support
vehicle, the Colorado Party, in a massive campaign. Letters supporting Stroess-
ner poured in to the newspapers. The party organized a pro-Stroessner
demonstration in the capital. They used posters, fliers, full-page newspaper
ads, and sound trucks in every neighborhood to publicize the event. Local
party activists contacted people in person. The party assembled more than
1,400 cars and trucks to transport people in and out of the capital. On the
day of the demonstration, the vehicles deposited people at party headquarters
where they were all given red party T-shirts and large pictures of the dictator to
brandish during the demonstration. The party provided free lunch. All public
officials and their families were required to attend. In these ways, a demonstra-
tion of 50,000 people, “enormous by Paraguayan standards,” turned out to
support the dictator (Lewis 1980, 148). The military dropped its opposition to
the treaty and made no coup attempt (Lewis 1980, 148–50).
The ability to mobilize mass demonstrations and overwhelming votes in
support of the dictator makes organized civilians useful to dictators. Dictator-
ships provide organized civilian supporters with access to mass communica-
tions and government-controlled transportation, as in the Paraguayan example
above. They supply party cadres with the carrots and sticks to make sure that
ordinary people turn out for big demonstrations or vote when told to. Some

9
During the 1956 Hungarian Revolution, “Army units dispatched to the scene not only refused to
attack the demonstrators but handed over their weapons . . . [T]he bulk of the military either
remained inactive or joined the insurgents” (Kovrig 1979, 300–301). Soviet troops had to be
used. Even at Tiananmen Square, where soldiers did ﬁre on demonstrators, troops from the
countryside had to be brought in because ofﬁcers and troops stationed nearby had expressed
sympathy with the demonstrators.

autocratic parties also earn popular support by distributing beneﬁts to ordinary

citizens, providing good economic policy, and making opportunities for
education and upward mobility available to people whose futures looked
bleak before.
Parties’ usefulness to dictators does not depend on supplying benefits to
citizens, however. Many of the toothless parties created after armed seizures
of power are incompetent, abusive, corrupt, or simply inconsequential for most
people. Cadres may sell the things they are supposed to distribute, and they may
use their party positions to exploit their fellow citizens. According to Mobutu
Sese Seku’s Commissioner of Political Affairs, for example, cadres of the party
created after Mobutu’s seizure of power “treat the population with
arrogance . . . [and] love to threaten [people] with arrest for any reason at all,
large or small” (Callaghy 1984, 168). Nevertheless, these party cadres can turn
out masses of citizens for votes and demonstrations of support by threatening
to block their future access to important services or to turn the names of those
who fail to participate over to security police. One of the main tasks of
Mobutu’s party was organizing mass marches to demonstrate the people’s
“unfailing attachment and support for the Father of the Nation” (Callaghy
1984, 324).
Party militants develop a vested interest in the dictator’s survival since the
dictator supplies them with benefits in return for support. The dictator’s control
of state revenues makes this strategy possible. Party officials and activists often
draw salaries. They have preferential access to jobs in the state bureaucracy and
schooling for their children. They have insider opportunities to form businesses
subsidized by the government and to manage or even take ownership of
expropriated businesses and land. Their connections help them to get lucrative
government contracts and profit from restrictions on trade. They have the
possibility of rising in the party to achieve the political power and, usually,
wealth associated with high office. Party militants’ stake in regime maintenance
derives from these advantages. Even where party activists enjoy no current
benefits, their connections open up future possibilities for rewards and upward
mobility (Svolik 2012). These benefits explain why dictators never have diffi-
culty recruiting civilians into their support parties.
Dictatorships often also task party members, especially local party officials,
with reporting suspicious behavior, hostile attitudes, and the presence of
strangers in villages or neighborhoods. Party militias can be used to set up
roadblocks to impede the movement of weapons and to patrol at night looking
for clandestine meetings and other suspicious activity. The effectiveness of such
mass spy networks varies a lot from one dictatorship to another, but in some,
the pervasiveness of spies and informers makes it very hard for potential
plotters to find ways of meeting and communicating with each other. In this
way as well, civilian support parties can reduce the likelihood of coups.
Although many dictators who achieved power without party support create
one after the seizure, most do not. About 40 percent of groups that established

dictatorships after 1946 were organized as parties beforehand. If monarchies

are excluded, military officers led about four-fifths of the nonparty seizure
groups. Of those cases in which groups not organized as parties seized power,
38 percent of the time a party was later created to support the dictatorship and
9 percent of the time the dictator coopted and allied post-seizure with an
existing party that had been organized during an earlier regime. In the latter
scenario, the dictator often reorganized and purged the party, refashioning it
into a personal support machine. In the other 52 percent of cases, the dictator
never created or coopted a party – as would be expected if party creation is
attractive to dictators who depend on the support of an officer corps riven by
factions but not to those whose military support base is more unified and thus
more stable.

evidence that post-seizure party creation aims to

counterbalance factionalized armed supporters
The argument that dictators create parties after seizures of power in order to
counterbalance factionalized armed support groups unable to make credible
sharing bargains implies that a number of relationships should be observable in
the real world. If authoritarian party creation is a strategic choice by dictators
to protect themselves from armed rivals, we should see that dictators themselves
initiate most party creations. We cannot observe that directly because public
announcements may be untruthful or may not make clear who sought to create
this new institution. We should, however, observe the following:
Most newly created dictatorial support parties should be led either by the
dictator himself or one of his relatives or close allies.
If dictators create parties to counterbalance their armed rivals, we should
also find that party creation brings with it lessened military influence on policy-
making. As an example, consider events in the Somalian dictatorship led by
Major General Siad Barre. When a group of colonels ousted the elected gov-
ernment in 1969, they formed the Supreme Revolutionary Council (SRC) of
twenty-five officers to rule the country and invited Barre to lead it. Barre had to
consult with other officers on the SRC to formulate policy. After surviving
several coup attempts, in 1976 the dictator created a new ruling party (the
Somali Revolutionary Socialist Party, SRSP).10 Decision-making was formally
transferred from the military SRC to the seventy-five-man executive committee
of the party, which included Barre’s civilian supporters as well as some of the
officers from the SRC. The SRC was disbanded, ending Barre’s formal

10
Barre’s Soviet allies pressed their standard blueprint, which included a ruling party, on him. He
may have created the party to please Soviet aid providers rather than to consolidate power, but it
nevertheless functioned as other post-seizure dictatorial parties do to help stabilize his regime.

consultation with the military but retaining the most powerful officers in the
party executive committee. After this first step toward military marginalization,
the military had to share decision-making with selected civilians. Barre then
further concentrated power in the party’s five-member politburo, which
included Barre and his son-in-law, who headed the internal security service.
Over time, Barre replaced officers in the politburo with civilians, further mar-
ginalizing the military. Though the SRC was revived in 1980, it functioned
thereafter as a parallel structure to the SRSP, with Barre very clearly the key
regime decision maker (Adibe 1995, 8). In general, we expect that:
Dictators create authoritarian support parties as part of an effort to margin-
alize armed supporters from policy-making.
Elections to confirm the dictator as president of the nation help to strip
armed supporters of their role as king-maker and -breaker. Elections create an
appearance of popular support aimed at undermining the feeling among offi-
cers that what the military gives it has the right to take away. Parties help
incumbents “win” such elections even if no opposition candidates are permit-
ted. Party activists, who are often public employees, spread the regime’s mes-
sages, distribute T-shirts and benefits, hold campaign events, and make sure
that citizens turn out to vote and vote for the right candidate (if they have a
choice). When legislatures have the task of anointing the president, the selection
of legislative candidates achieves high importance. Ruling party executive
committees typically choose candidates, usually in consultation with the dicta-
tor himself. Because parties help dictators control elections, we should expect
party creation to predate elections to confirm the dictator.
Parties should be created before elections that confirm the dictator as
national executive.
If the military were unified and able to bargain effectively, the dictator would
be unable either to redistribute resources toward new civilian supporters uni-
laterally or to add new actors to the dictatorship’s inner circle at will. Conse-
quently, dictators infrequently propose party creation in countries with unified
armed forces.11 It is instructive therefore to observe one of the few times when a
dictator misunderstood the strategic situation and tried to create a party despite
the existence of a fairly unified army. General Gustavo Rojas Pinilla, the leader
of a military regime in Colombia (1952–58), announced the creation of a new
party, the Third Force movement. The idea received an unenthusiastic response
from other officers, and languished unimplemented until the following year,

11
In the small number of instances in which dictators have created post-seizure parties despite
relatively uniﬁed and disciplined militaries (as in Brazil during military rule), enforced leader
rotation and term limits have usually accompanied their creation. These institutions serve as an
insurance policy for other ofﬁcers to prevent the concentration of power in the dictator’s hands
that a party might otherwise facilitate.

when Rojas brought it forward again. But after meeting for several hours with
an incensed group of more than a hundred other officers, Rojas let the party
quietly disappear (Szulc 1959). They had made the threat to oust credible.
If more united seizure groups have greater ability to resist the personaliza-
tion of rule, and hence party creation, then in military-led regimes, we might
also expect to see more parties created in dictatorships led by junior officers
compared with those led by senior officers. If the military is united and its
hierarchy intact, junior officers who lead successful coups will turn power over
to a senior officer immediately. This happens because a senior officer can
command the cooperation of officers who do not share the goals of those
who led the coup and because many officers would object to the violation of
military norms involved in a junior officer becoming president and thus com-
mander-in-chief of higher-ranked officers. The ability of junior officers actually
to take power after a coup suggests severe factionalism. It indicates that
factional loyalties have undermined the military’s conventional emphasis on
hierarchy and discipline.
Military dictatorships led by junior officers should create more support
parties than those led by higher-ranked officers.
Finally, we think it unlikely that party creation would be the only effort a
threatened dictator would make to try to safeguard himself from coups. We
expect that a dictator who fears rogue coups would also invest in internal
security agencies to spy on possible plotters and that he would create new,
more loyal armed forces to protect himself from army attempts to oust him. If
party creation is part of a broad strategy for reducing reliance on the military as
a base of support, then we might expect to see that dictators who create parties
are also more likely to take personal control of the security forces and to
establish paramilitary forces to counterbalance the military than are dictators
who can rely on a united military support base.
Dictator control over internal security services should be more likely in
dictatorships that create a new support party than in those that do not.
The establishment of new paramilitary forces to protect the dictator should
be more likely in dictatorships that create a new support party than in
those that do not.
In what follows, we use our data set (unless otherwise noted) to examine
whether these expectations match empirical reality.

Is Party Creation Usually Initiated by the Dictator?

In three-quarters of dictatorships in which a party was created after the seizure
of power, we ﬁnd that the dictator or a close relative led the newly created
party, as would be expected if he controls party creation. The dictator always

delegated the leadership of the new party to a close ally if he did not keep the
post for himself or a relative.

Is Post-Seizure Party Creation Part of a Military

Marginalization Strategy?
If dictators initiate party creation as part of a strategy to concentrate power,
then party creation should be part of a process of marginalizing other officers
from policy-making. To examine whether the process visible in Somalia is more
general, we investigate how the creation of a new support party influences
bargaining between the dictator and other officers by examining three related
measures of military marginalization: leadership rotation within the military
ruling group, consultation with other officers about policy decisions, and
military representation in the cabinet. Because this argument pertains to
military-led regimes, we test these expectations on dictatorships that gained
power through armed seizures of power (coup, rebellion, uprising, or foreign
imposition) in which the military selects the dictator. We further restrict the
analysis to regimes that did not inherit a regime support party so that we can
examine the extent to which the creation of a new party influences military
marginalization. Among these regimes, 42 percent create a new support party,
while the majority rule without a political party.12
The strongest indicator of collegial decision-making is the regular rotation of
the presidency among officers. We expect party creation to be associated with
less leadership rotation since it helps the dictator to reduce the power of his
armed supporters. Regular leadership rotation is relatively rare: it occurs (at
some point) in only 9 percent of military-led regimes that came to power in
armed seizures (and lacked an inherited support party). We test the relationship
between party creation and leader rotation using a model with regime-case
fixed effects to isolate the influence of creating a new support party.13 The
model thus controls not only for cross-country variation in factors such as level
of development and colonial legacy, but also for regime-specific features such as
how the regime seized power, the prior experience of regime elites, and the
institutional environment in which elites operate. We also control for two
factors likely to influence the creation of new parties: whether the current
dictator is the first leader of the regime, and the change in the international
environment after the end of the Cold War to encourage the creation of
“democratic-looking” political institutions such as parties. Figure 5.1 shows

12
These percentages differ slightly from those above because the universe of cases in which the
percentages are calculated is different. Because military-led regimes that create a new support
party last longer than those that do not, the number of regime-years with a new support party is
larger: 55 percent of military-led regime-years have new parties.
13
This means we are comparing periods before and after new party creation within the same
regime, and then pooling these estimates.

Cold war

New party

First regime leader

–0.1 –0.08 –0.06 –0.04 –0.02 0 0.02

Probability of rotation agreement
figure 5.1 Post-seizure party creation and the rotation of dictatorial leadership.
Note: Estimating sample is military-led regimes with an armed seizure of power and no
prior support party.

that post-seizure party creation (New party) reduces the likelihood of the
regular rotation of the presidency by roughly 2.5 percent for military leaders
who seized power without a party.
We next look at two other measures of military marginalization: lack of
consultation with officers about policy decisions and civilianization of dictator-
ial policy-making. We capture the former using country specialists’ assessments
of whether the dictator consults regularly with other officers. We measure the
latter by looking at the composition of cabinets. Military representation in the
cabinet is measured as whether the most important members of the cabinet –
other than the defense minister – are active duty or recently retired military,
police, or security officers. We define the “most important” ministries as the
prime minister (if one exists), the ministry of interior or state (which in most
countries controls the police, internal security agencies, and voting), and others
that are particularly important in the country context (e.g., the ministry
that deals with oil in oil-exporting countries). These forms of military margin-
alization are quite common: in more than half of these military-led regimes
(56 percent) the dictator makes most decisions without regular military con-
sultation; and in most (77 percent), cabinets have little military representation.
To analyze the effect of party creation on these measures of military margin-
alization, we again estimate a linear probability model with regime-case fixed
effects, and controls for the Cold War period and first regime leader. This
approach compares periods prior to party creation with periods after
party creation within the same regime, and then pools these comparisons
into one average estimate. It accounts for cross-country variation in economic
and cultural factors as well as regime-specific features that affect elite

Not consult 0.47

military

Exclude military 0.16

from cabinet

95% ci
0 0.1 0.2 0.3 0.4 0.5
Probability of military marginalization
figure 5.2 Post-seizure party creation and military marginalization.
Note: Estimating sample is military-led regimes with an armed seizure of power and
no prior support party.

decision-making. Figure 5.2 reports the results, showing that party creation is
associated with increases in both the probability that dictators eschew consult-
ation with other officers and the likelihood that officers are excluded from the
most important cabinet posts.
We also expect that if dictators create parties as part of an effort to margin-
alize the military, they should often be established in the run-up to elections that
confirm the dictator as national leader, either directly or through the election of
a legislature tasked with doing so. To assess this, we again restrict our analysis
to dictatorships that came to power via armed force and lacked a support party
when they seized power. In order to model the creation of a new support party,
the sample includes only the observation years in which the dictatorship lacked
a support party in the previous year.
We use data on national-level elections in which the incumbent, his party, or
his chosen successor appears on the ballot to identify election years. We test the
likelihood of party creation in the year before or year of elections relative to
other years of the same regime.14 As shown in Figure 5.3, we find that dicta-
torships are much more likely to create parties in the year of or before leader
elections (31 percent) than at other times (4 percent). We also tested this
prediction in a logistic regression model with control variables (Cold War,
seizure type, a polynomial of years with no new party, and first leader), with

14
Election data are from NELDA3. We examine only national-level presidential and parliamentary
elections in which the ofﬁce of the incumbent was contested (NELDA 20) and in which the
incumbent or his chosen successor ran (NELDA 21).

Probability of new party creation 30 31

4.2
0
No election Election
figure 5.3 Post-seizure party creation before the leader’s election.
Note: Military-led regimes with an armed seizure of power and no prior support party.

a similar result. Finally, we tested a linear model with regime-case ﬁxed effects
to isolate variation over time within dictatorships. Again, we ﬁnd that dictator-
ships are more likely (18 percent) to create new parties in the run-up to
elections.

Can More United Militaries Deter Party Creation?

To test the hypothesis that a more unified officer corps can better deter party
creation, we use two individual traits of autocratic leaders as proxies for
factionalism/unity within the officer corps: their age when they seized power
and their rank just prior to the seizure of power.15 As explained above, we
think that the pre-seizure rank of the first dictator can be used as a proxy
measure for factionalism. We reason that dictatorships led by highly ranked
officers may depend on either factionalized or united officer corps for support
(since a faction may be led by a high-ranked officer), but that those led by junior
officers all rely on a factionalized military base. The future dictator’s rank
before the seizure of power is exogenous to promotions and other decisions
made by the dictatorship. We use a parallel logic with regard to the age of the
first dictator. Youthful military dictators are either low-ranked officers or
higher-ranking officers in newly created armies. We expect new armies to have
more problems with factionalism and indiscipline. We restrict the analysis to
first leaders in regimes that came to power in armed seizures of power after
1945; there are 135 dictators in this group.

15
Data on leader age from Horowitz and Stam (2014), with updates by the authors to ﬁll in
missing data.

75
% of leaders who created a new party

66.7
61.3
57.7
50

32.7
25 27.3 28.8

0
<=40 >40 low-ranking mid-ranking high-ranking not military
officer officer officer officer
Leader characteristics
figure 5.4 Post-seizure party creation, age, and rank of ﬁrst dictator.

Figure 5.4 shows the share of dictators in each category that created a new
party. The left two bars show that dictators forty years old or younger at the
time they seized power were almost twice as likely to create a new party as
those older than forty.
Next, we divide these dictators into four categories: civilians before seizing
power and military officers of different ranks.16 The four bars on the right show
that two-thirds of low-ranking military dictators create new parties, while just
over one-quarter of mid- and high-ranked officers do. Just over half of the
civilian dictators to whom officers delegate power create post-seizure parties. In
other words, young or low-ranking military dictators, who reflect factionalism
in the armed forces, are even more likely to create parties than civilians.
Next, we examine whether these patterns persist when we control for
potential confounders. To do this, we estimate a model that compares first
regime leaders with one another, while controlling for seizure type (rebellion,
uprising, and coup, with foreign imposition as the baseline category), indicators
for whether the regime before the one being coded was a democracy or a
military regime (with nonmilitary dictatorship as the reference category), and
a variable measuring whether the dictator seized power before 1990. We
cannot test dictator age at the time of seizing power in the same model as

16
High-ranking officers are either generals or colonels in a military where colonel is the highest
officer rank. Mid-ranking officers are colonels in a military with generals; and low-ranking
officers are majors and below. Civilians are simply nonofficers. The cases include some civilian
dictators because occasionally military officers delegate power to a civilian leader.

Age model Rank model

>40 yrs age

Prior democracy

Prior military regime

Cold war entry

Seizure type

Coup

Rebellion

Uprising

Low rank

Mid rank

Not mil. officer

–2 –1 0 1 2 –2 –1 0 1 2
Coefficient estimate
figure 5.5 Effect of age, rank, and previous regime on post-seizure party creation.

variables for officer rank because age and rank are highly correlated: more than
80 percent of low-ranking officers who seize power are aged forty years or
younger, while almost 90 percent of mid- and high-ranking officers are older
than forty.
Figure 5.5 shows the results. The estimates in the left panel show that older
dictators are less likely to create new parties than younger ones, while the right
panel shows that low-ranking officers are more likely than high-ranking officers
to do so. If leader age and officer rank are good proxies for military factional-
ism, then these findings suggest that dictatorships launched by factionalized
militaries are more prone to personalization. Previous democratic experience
reduces the likelihood that a new dictatorial support party will be created, as
does earlier experience of military rule, relative to earlier experience of civilian-
led autocracy.
Finally, as another proxy indicator of less professionalized military forces, we
look at the newness of the military institution, the idea being that discipline and
norms about hierarchy probably take some time to develop. In most previously
colonized countries, the officer corps was created at around the time of inde-
pendence, so we might expect to see more post-seizure party creation in new
nations with new indigenous officer corps. To investigate this possibility, we
examine the calendar time trend in post-seizure party creation by estimating a
nonlinear model with a cubic calendar time polynomial and controls for how the
regime seized power (coup, rebellion, and uprising, with foreign imposition as
the reference category), leader’s time in office (log), and whether the dictator is

0.3
Probability of party creation

0.2

0.1

0
1950 1960 1970 1980 1990 2000 2010
Year
figure 5.6 Post-seizure party creation over time.
Note: Sample is military-led regimes following an armed seizure of power, with no
support party.

the regime’s first leader.17 Figure 5.6 shows that early in the post–World War II
period, when indigenous militaries in many countries were still quite new,
dictators had the highest propensity to create new parties. Party creation drops
to a low point in the mid-1970s, once the period of decolonization was largely
finished, and remains low for the next forty years, with a slight rise as the Cold
War ended and newly independent countries emerged from the Soviet Union.
Some of these countries lacked indigenous officers and had to create new officer
corps at independence just as the earlier wave of newly independent countries
did, and leaders in some of the post-Soviet dictatorships created new parties.

Is Party Creation Part of a Strategy to Reduce the Dictator’s

Vulnerability to Coups?
If dictators organize new support parties as part of a broad strategy to reduce the
likelihood of coups, as suggested above, we would also expect them to engage in
other coup-prooﬁng strategies. Coup-prooﬁng could include taking personal
control of internal security in order to monitor potential plotters. It could also
include establishing paramilitary forces such as presidential guards, which are
often recruited from dictators’ home regions. Such paramilitary forces help to
defend dictators by remaining loyal during coup attempts by regular military

17
For this analysis we want to model variation in party creation among countries in different time
periods, so we do not include country ﬁxed effects.

ofﬁcers. In the dictatorships in which a party was created after the seizure of
power, the dictator also takes personal control of internal security 72 percent of
the time, compared with 27 percent of the time for dictators of regimes never
supported by a party. Dictators who establish new support parties are three times
as likely to establish paramilitary forces to protect them (52 percent) than are
dictators who do not form post-seizure parties (17 percent).
As in the tests above, these relationships hold after we account for cross-
country variation in a model with ﬁxed effects: creating new parties is associ-
ated with a 25 percent increase in the probability of personalizing the security
apparatus and a 33 percent rise in the chances of forming a paramilitary force
loyal to the dictator.
These relationships indicate that dictators initiate party creation as part of a
strategy to reduce dependence on the regular military when their original armed
support base was too factionalized to make commitments of stable support in
exchange for power sharing credible. Party creation after armed seizures of
power thus paradoxically contributes to the personalization of dictatorial rule.
Parties originally organized to lead a revolution or the struggle for independ-
ence may develop enough internal unity and discipline to constrain the dictator,
but parties created by sitting dictators rarely do so because the dictator himself
chooses and dismisses their leaders and controls the resources they need to
maintain themselves.

post-seizure party creation and dictatorial survival

We concur with other analysts in seeing dictatorial ruling parties as autocratic
survival tools (Gandhi and Przeworski 2007; Gandhi 2008; Magaloni 2008).
We believe that post-seizure parties, like inherited parties, prolong dictatorial
survival beyond what it would have been without them, but we see parties
created after armed seizures of power as playing a quite different role in
bargaining among elite actors than that played by parties that lead seizures of
power. Post-seizure parties reduce intra-elite conflict by helping the dictator to
concentrate power at the expense of armed supporters. The reduction in elite
conflict increases dictatorial longevity.
Beatriz Magaloni’s (2008) influential argument suggests that parties extend
the life of dictatorships because they make possible credible intertemporal prom-
ises by the dictator to continue sharing spoils if allies continue supporting him.
She suggests that parties can solve dictators’ credibility problem if the dictator
delegates control over appointments to high offices, including the dictatorship
itself, to the party. When the party controls access to office, the dictator has
reason to fulfill his promises because he knows he can be ousted, and his allies
have reason to remain loyal because they can expect higher offices in the future.
She refers to this form of power sharing as delegation by the dictator to the party,
which implies that the dictator’s own interests are served by it.

That seems a dubious assumption. Such agreements prolong regime survival

while limiting both the dictator’s time in power and his resources while in
power. Thus, they do not appear to serve the dictator’s interests. We suggest
that dictators agree to such arrangements when they are weak relative to the
rest of the dictatorial elite. That is, they agree to limit their own discretion when
their choice is between becoming dictator under constraints and not being
dictator at all. We can identify two empirical conditions that contribute to
the dictator’s relative weakness: (1) very recent accession to leadership, which
we discuss in greater detail in Chapter 8 and (2) the wide dispersion of armed
force among factions within the ruling group. Weakness forces the dictator to
grant this form of power sharing. If power has already begun to be concen-
trated in the dictator’s hands when the party is created, or if creating a new
party helps the dictator reduce his dependence on armed supporters, he has less
reason to delegate powers that may increase regime survival but not his own
time in office. After most seizures of power through force, the dictator does not
delegate control over highest offices to the party.
With the notable exception of the PRI regime in Mexico, nearly all dictator-
ships in which the dominant party actually controls access to high office were
brought to power by parties originally organized before the seizure of power to
win elections, lead revolutions, or fight for independence. In very few cases in
which parties were created after the seizure of power does the party control
access to high office.
Instead, as shown above, the dictator himself or one of his relatives leads the
newly created party most of the time. In 45 percent of regimes in which the
dictator creates a post-seizure support party, he also controls appointments to
the party executive committee.18 Controlling appointments to the party execu-
tive committee is symptomatic of much broader control. A contemporary
observer of Mobutu (who created a post-seizure support party), for example,
reports: “He controls and distributes all offices, all the posts, all advantages
linked to power. All revenue, all nominations, all promotions ultimately depend
on presidential good will.”19 Where the dictatorship’s support party pre-dates
the seizure of power, by contrast, the dictator controls appointments to the
party executive committee in only 28 percent of cases.
The dictator who controls appointments to the party executive committee
not only cannot make credible promises to share; he does not want to do so.
Instead, he tries to keep his supporters insecure about the future so that they
will compete with each other and work hard to demonstrate their loyalty. In
Rafael Trujillo’s Dominican Republic, for example, all officials had to sign
undated letters of resignation before taking office. Legislators who displeased

18
This and the following ﬁgure refer to dictatorships in which the regime gained power in an armed
seizure (coup, uprising, rebellion, or foreign imposition).
19
Jean Ryneman, “Comment le régime Mobutu a sapé ses propres fondements,” Le Monde
Diplomatique, May 1977.

Trujillo simply disappeared from the legislature and sometimes from the world
of the living without anyone seeming to notice. Multiple legislators “resigned”
within a month after nearly every election (Galíndez 1973). Cabinet ministers
could find out they had lost their jobs by reading it in the newspaper (Hartlyn
1998). The statutes of the party Trujillo created after seizing power gave him
the unilateral right to make decisions about who occupied these posts (Galíndez
1973). Under Mobutu, “state-party personnel are completely dependent on him
for selection, appointment, and maintenance in power . . . The powers of
appointment and dismissal that Mobutu wields create constant uncertainty
for all officials, which helps to maintain their loyalty to him” (Callaghy
1984, 180).
In many of the regimes in which the dictator’s intertemporal commitment
problems have not been solved, both dictators and dictatorships nevertheless
last a long time. Mobutu lasted thirty-seven years in a very turbulent political
environment, and Trujillo for thirty-one. Dictatorships that achieved power by
force but then later created post-seizure parties last more than twice as long on
average as otherwise similar regimes without support parties. To sum up our
argument, even parties that do not deliver benefits beyond a relatively small
group and do not control access to highest offices can still prolong both dictator
and regime survival.
Next, we compare the effect on dictatorial survival of ruling parties estab-
lished before and after seizures of power. We test a linear probability model
with country and year fixed effects, controls for regime duration, and indicator
variables for party history: pre-seizure electoral party, pre-seizure rebel party,
no party (reference category), and new party (post-seizure creation). This
analysis accounts for the fact that some regimes were still in power at the end
of the sample period in 2010. The results show the following average yearly
regime collapse rates: 10.1 percent probability of breakdown for dictatorships
without a support party; less than half that, 4.5 percent, for those in which a
party was created post seizure; 3.8 percent for regimes led by parties first
organized to run in elections before the initiation of dictatorship; and only
1.9 percent for regimes led by parties organized to lead insurgencies. Compari-
son between the breakdown rate for regimes lacking support parties and those
that create a party post-seizure provides compelling evidence that party cre-
ation prolongs survival because regimes that lack support parties and those that
create one post-seizure have similar origins: both typically achieved power by
force (most often via coup).20 Apparently, autocracies supported by parties that
have not solved the dictator’s commitment problem – that is, parties that we

20
Regimes that lacked a support party at the time of seizure but later created parties came to power
via a coup in 65 percent of cases; those that did not create a party seized power in a coup in
60 percent of cases. In contrast, only 17 percent of regimes supported by a preexisting party
achieved power via coup.

New
party

Prior
rebel
party

Prior
elected
party

–0.15 –0.1 –0.05 0

Coefficient estimate
No unit effects RE FE

figure 5.7 Parties and regime survival.

Note: Negative coefﬁcient estimates interpreted as a decrease in the likelihood of
collapse, relative to the comparison group: regimes that never have a support party.

know did not control the dictator’s access to ofﬁce because he achieved it before
the party came into existence – nevertheless last quite a long time.
Figure 5.7 displays these same relationships as comparisons among regimes
without support parties and regimes supported by parties created at different
times for various purposes. In all models, dictatorships with support parties are
less likely to end. In models that include ﬁxed effects (circle symbol), post-
seizure party creation is associated with a bit more than a 10 percent decrease in
the likelihood of regime collapse compared with having no support party.
Parties originally created before the seizure of power to run in elections or lead
insurgencies reduce the likelihood of regime breakdown even more, but not by
a great deal more.
These comparisons show that armed forces by themselves often do not
provide a reliable support base for autocracy. Consequently, dictators who
are delegated power by a seizure group that did not need to be organized as a
party in order to achieve power can improve the odds of retaining it by creating
a support party to counterbalance the potential volatility of military support.

the effect of post-seizure party creation on the

likelihood of coups
If our argument about party creation is correct – that it is pursued by dictators
propelled to power by factionalized armed seizure groups in order to lessen
their risk of ouster by force – then we should see fewer coups and coup attempts
in regimes that create parties than in those that opt not to. To evaluate this
expectation, we look at the incidence of coups in dictatorships without support

Coup attempts Successful coups

0.4 0.4
Create new party Create new party
Never create party Never create party

0.3 0.3
Baseline coup risk

0.2 0.2

0.1 0.1

0 0
0 5 10 15 20 0 5 10 15 20
Regime duration Regime duration
figure 5.8 Coups in dictatorships with post-seizure parties or no parties.

parties at the time of the seizure of power. We compare the likelihood of coups
in dictatorships that initially lacked a support party but later created one with
those that never organized supporters into a party at all.
We begin by looking at the differences in the baseline rate of coup attempts –
both failed attempts and successes – between regimes that had created parties
and those that never did.21 Coup risk declines on average during the first few
years in all kinds of dictatorships, as dictators try to remove their least reliable
supporters from troop commands, and the least stable dictatorships collapse
and exit the sample. Figure 5.8 shows that during the first two decades in
power, regimes that never create a party are at greater risk of both coup
attempts (shown in the left panel) and successful coups (shown in the right).
This implies that dictatorships that never create a party are particularly suscep-
tible to coup conspiracies.
Because we have argued that the creation of a mass-based civilian support
party deters officers from staging coups, we expect to see a larger difference in
coup attempts than in successful coups. A comparison of the left and right
panels of Figure 5.8 shows larger differences between the two lines for
attempted than for successful coups. Both panels also show that the coup risk
for dictatorships that have created support parties remains stably lower for
more than a decade after party creation.

21
The baseline coup rate is the ﬁve-year moving average of the number of coups and coup
attempts, by regime duration year, divided by the number of regime-years at risk of having a
coup (for each regime duration year).

In contrast, there is a large spike in coup attempts after the first decade in
power for regimes that never create a party. Typically, armed seizures of power
are followed by retirements of officers ranked above those who take power and
a purge of officers who supported the ousted government. Rapid promotions
for the cohort of the active coup plotters and those just below them follow.
These changes in the officer corps create a cohort of coup beneficiaries, but over
time many of these officers will retire or be dismissed, creating opportunities for
younger officers to gain command of troops. Coups depend on troop com-
mand, and these junior officers become the coup plotters of the future. In this
way, the normal seniority-based promotions inherent in military careers inter-
act with the ambitions and criticisms of junior officers to produce the spike in
coup attempts that begins after about ten years in power unless a well-
established support party deters them.
The difference in trends for regimes with and without ruling parties is
important for showing that party creation has consequences. If the trend lines
moved in similar ways in each set of regimes, we might wonder if preexisting
differences accounted for the difference in coups. However, the trends diverge,
suggesting the limitations in the coup-proofing strategies that can be deployed
by dictators dependent on all-military support bases.
For further evidence that dictatorships with parties created post-seizure are
less vulnerable to coups, we note the success rate for coups conditional on
attempts: while 22 percent of coup attempts in regimes with post-seizure parties
succeed, nearly 29 percent succeed in regimes that never established a support
party. After five years in power, the difference is greater: only 20 percent of
coups in regimes with post-seizure parties succeed, while the share that succeeds
in regimes with no party increases to 35 percent.22 This suggests that once we
account for all the factors that lead to coup attempts in the first place, dictators
who have created a new support party have greater ability to thwart them. This
result most likely reflects the strong correlation between post-seizure party
creation and investment in other coup-proofing strategies, especially the dicta-
tor’s establishment of new paramilitary forces recruited from especially loyal
regions. Republican guards, presidential guards, and other kinds of specially
recruited paramilitary forces are usually stationed near the presidential palace
and tasked explicitly with defending him from armed challenges.
Next, we test a series of parametric models to evaluate whether creating
a party reduces coup risk. We employ a nonlinear estimator with random
effects for each regime. First, we test a specification with a minimum of control
variables: time since last coup (logged), whether the dictator is civilian

22
These differences are statistically significant. During the first five years, however, the difference in
success rate is not statistically significant. Remember that these are the success rates for coup
attempts against dictatorships that achieved power without the support of a party, most of which
are led by the military. The higher success rate found in other studies refers to coups against
democracies and other dictatorships as well (e.g., Singh 2014).

New party

Civilian leader
Regime seizure type
Coup yrs (log)

Rebellion

Coup

Foreign

Uprising

–2 –1 0 1 2
Coefficient estimate
Coup successes Successes with controls
All attempts Attempts with controls

figure 5.9 Post-seizure party creation and coup risk.

(as opposed to military), and how the regime seized power (coup, rebellion, foreign
imposition, or popular uprising).23 The top set of estimates in Figure 5.9, shown as
large diamonds, is from this model, with successful coups as the dependent
variable. Next, we add a set of standard control variables often thought to affect
coup risk: GDP per capita (log, lagged), economic growth (lagged two-year
moving average), oil rents (log, lagged), military expenditure (log, lagged), civil
and international conflict (lagged), and protests (lagged). In the third specification
we return to the minimal model but use all coup attempts (failed and successful
ones) as the dependent variable. In the final model, we add controls to the all
attempts model. In all four tests, post-seizure party creation is associated with a
lower incidence of successful coups; however, the estimate for New Party is only
statistically significant at the 0.10 level for the all attempts models.
To further explore how the creation of new support parties influences coup risk,
we estimate a conditional logit model that compares the coup risk of individual
leaders before and after new party creation.24 This analysis looks only at the

23
The reference category for seizure type is nonviolent, which includes family seizures, authoritar-
ianization, and rule-change seizures.
24
The speciﬁcation includes the time since last coup (logged) and a binary indicator for the Cold
War period.

sixty-seven leaders in dictatorships that initially lacked a support party but later
created one; that is, it looks only at rulers who held power during periods both
with and without a support party. Importantly, by focusing on the variation over
time for individual dictators, we can rule out alternative explanations based on
differences between leaders, regimes, and countries. The results from this test
(not shown) also indicate that coup risk is lower after new party creation.25
Thus far in this chapter, we have combined reshuffling and regime-change
coups when examining how party creation protects the dictator from ouster.
Our theory about why dictators who achieved power by force later create
parties to organize civilian supporters, however, claims that party creation
helps the dictator concentrate personal power over other members of the ruling
coalition. That idea leads to the expectation that new parties should lengthen
dictators’ time in power by deterring coups aimed at replacing leaders (reshuf-
fling coups), but not necessarily those aimed at ending the regime. Members of
the ruling group who want a change in leadership but not the end of the regime
organize most leader-change coups. Regime-change coups, however, are usu-
ally organized by factions of the military excluded from the inner circle of the
dictatorship. If party creation is a dictator’s strategy for increasing his own
power relative to that of other members of the dictatorial ruling group, as we
have argued, it should protect him from leader-shuffling coups but not neces-
sarily from regime-ending coups.
Figure 5.10 shows the results from a series of random effects models similar
to those in Figure 5.9. The first two estimates are from models of leader-
shuffling coups that treat regime-change coups as right-censored events, while
the latter two estimates are from models of regime-change coups that treat
reshuffling coups as right-censored events. These tests indicate that post-seizure
parties are associated with fewer leader-shuffling coups but not with fewer
regime-change coups.26
This evidence is consistent with our theory that post-seizure party creation is
a strategy to protect dictators from their erstwhile allies in the armed forces.
That is, party creation protects dictators from coups led by ambitious regime
insiders eager to take the dictator’s place without ending the regime. Post-
seizure party creation is less reliably helpful for deterring coups aimed at ending
the regime. We know from the findings reported in Figure 5.1 that the creation
of new parties contributes to the survival of regimes as well as individual

25
This result is statistically significant only at the 0.10 level. We also test similar models that divide
the newly created parties into two categories: those led by the regime leader (or close relative)
and those led by someone else. We find that, as expected, only those newly created support
parties led by the regime leader are correlated with lower coup risk. These are the cases in which
party creation contributes most to the personalization of rule. This finding is highly significant
and persists in a similar model that isolates the temporal variation within regimes.
26
These results also hold in a conditional logit model and in linear probability models with regime-
case fixed effects.

New party

Civilian leader
Regime seizure types
Coup yrs (log)

Rebellion

Coup

Foreign

Uprising

–3 –2 –1 0 1 2 3
Coefficient estimate

Leader Leader Regime Regime

shuffle shuffle change change
coups (controls) coups (controls)

figure 5.10 Post-seizure party creation and reshufﬂing versus regime-change coups.

dictators, but the results shown in Figure 5.10 imply that the extra regime
durability associated with party creation comes not through coup deterrence
but through some other mechanism. We discuss this other mechanism, the
civilian side of how newly created parties contribute to regime durability, in
Chapter 6. Those coups that do end the entire regime – rather than just replace
the dictator – occur most often when junior officers from excluded ethnic groups
successfully oust the regime leader and allied senior officers. Thus, the evidence
from differentiating regime-change from leader-shuffling coups suggests that the
kind of parties created by dictators post-seizure do not successfully coopt ethnic
groups excluded from executive power and the senior officer ranks.
To conclude this section, dictators who organize post-seizure parties are less
vulnerable to coups than dictators in regimes unsupported by a ruling party, as
our argument detailing the motivations for post-seizure party creation implies.
Further, the evidence suggests that post-seizure party creation helps deter coups
emanating from regime insiders rather than those plotted by groups of soldiers
excluded from the ruling group.

conclusion
Dictators who achieve power through force of arms can face special difﬁculties
in consolidating their rule because many of their supporters control sufﬁcient

weapons to oust them. This chapter has focused on the consequences for intra-
elite bargaining of the interaction between the dispersion of armed force across
members of the ruling group and the group’s division into multiple factions.
When many members of the seizure group – and the ruling coalition it
becomes – command armed forces sufficient to threaten the dictator with
ouster, they can achieve an effective power-sharing bargain with the dictator
if they can maintain their own unity. If, however, deep factions divide an armed
seizure group, those included in the dictator’s inner circle cannot credibly
commit their subordinates to support the dictator if he shares power and spoils.
Consequently, power-sharing bargains cannot be maintained.
When the dictator cannot secure his hold on power by agreeing to share with
the rest of the seizure group, he is better off keeping a larger share of the spoils
and other benefits of office so that he can invest in other strategies. We suggest
that dictators who depended on armed supporters to achieve power, but who
cannot count on those supporters for holding onto it, often try to counterbal-
ance their armed supporters with unarmed ones. To accomplish this, they
organize civilian support networks and appoint their leaders to the dictatorial
inner circle. The addition of civilian supporters to the ruling group changes
bargaining within the group by increasing the diversity of interests and further
reducing the unity of the inner circle; this in turn undermines the bargaining
power of members of the dictatorial inner circle relative to the dictator. It thus
contributes to the personalization of dictatorial rule. These civilian support
organizations are usually called parties.
The evidence shown in this chapter is consistent with the argument that
factionalism within an armed seizure group increases the likelihood of post-
seizure party creation. We also offer evidence that post-seizure party creation
usually reflects a dictator’s interest rather than the collective interest of the
ruling group: dictators usually assume a new party’s leadership themselves; new
parties tend to be established in the run-up to elections that “legitimize” the
dictator’s occupation of national executive office; and dictators often control
appointments to the executive committees of ruling parties created post-seizure.
We further show that post-seizure party creation is associated with the margin-
alization of military influence within the dictatorial inner circle.
Finally, we provide several kinds of evidence that post-seizure party creation
is an effective dictatorial survival strategy. It is associated with both longer
dictator tenure in office and longer regime survival. Party creation seems to
protect dictators from coups, as would be expected if it were a strategy for
reducing the dictator’s vulnerability to ousters launched from an unreliable
military force. After controlling for other factors known or believed to affect
the incidence of coups, post-seizure party creation is associated with a reduced
incidence of both coup attempts and successful coups.
Further investigation shows that post-seizure party creation affects the inci-
dence of leader-change coups but not regime-change coups. This is what would
be expected if party creation is a strategy used by dictators to safeguard

themselves from armed insiders rather than a group strategy for regime defense.
The ﬁndings in this chapter thus explain how party creation extends the
survival of individual dictators: it is associated with a reduction in the likeli-
hood of coups that would replace one dictator with another insider. They do
not, however, explain how new dictatorial support parties contribute to regime
survival. We turn to that question in the next chapters.
Post-seizure party creation thus seems paradoxical within the usual way
political scientists think about authoritarian parties and support coalitions.
The creation of a mass party through which some beneﬁts are channeled from
the political center to ordinary citizens seems to imply the broadening of the
dictator’s support coalition. At the same time, however, such broadening
among the mostly powerless accompanies a disorganization and eventually a
narrowing of the support coalition among the powerful, as threatening military
supporters are replaced by less powerful civilian ones. In most cases, much of
the military becomes marginalized after post-seizure party creation. Civilian
party leaders handpicked by the dictator replace military members of the
dictatorial inner circle, but they have much less ability to constrain or oust
the dictator than do armed supporters. Post-seizure party creation thus trans-
forms the dictator from a relatively equal bargainer within a group of others
similar to himself into an arbiter among competing support factions, reducing
his dependence on all of them and enhancing his individual discretion over
resources and policy.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:13, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.005
Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:13, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.005
6

Why Parties and Elections in Dictatorships?

implementation, monitoring, and

information gathering
As the initial conflicts within the inner circle become resolved, and elite
decision-making becomes somewhat predictable, the ruling group must turn
its attention to the rest of the country. Most dictatorships face some very basic
problems when it comes to trying to rule the places they have taken control of.
One ordinary problem is the need to motivate the implementation on the
ground of decisions made in the inner circle, especially in new dictatorships
when local officials may have owed their jobs to the ousted government.
Policy implementation requires the cooperation of local officials, who must
be converted to the new order after regime change or recruited afresh, and then
monitored to assure collaboration and prevent abuse of office. Without effect-
ive monitoring, local officials can sabotage policies, abusive ones can motivate
popular opposition to the new regime, and resources extracted from citizens
may stick to officials’ fingers rather than reaching central coffers. Direct moni-
toring requires expertise and resources, however. Many dictatorships lack the
trained and loyal manpower needed to do it and the revenues needed to
pay them.
Information shortfalls are another and related ordinary problem (Wintrobe
1998). The dictatorial elite needs information about who opposes them, how
their policies are working, and what disasters and difficulties afflict people in
different parts of the country. Central leaders have difficulty acquiring needed
information. Local officials engaged in sabotage, theft, or abuse of power will
not provide accurate reports on local conditions. Even loyal and honest local
officials dread being blamed for problems if they reveal them, and everyone
fears retribution for bearing bad news. The potential for violent and arbitrary

129

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
130 Ruling Society: Implementation and Information Gathering

punishment in dictatorships distorts and disrupts the information reaching

regime elites.
Cadres and officials whose futures depend on successful performance of
tasks assigned to them by regime leaders have incentives to report successes
and to hide policy failures and popular opposition. Writing about the last years
of the East German dictatorship, which fell to popular opposition in 1989,
Peter Grieder (2012, 91) reports:
The heavily varnished reports sent to the centre by district and local party officials only
fortified the rarefied fool’s paradise in which party leaders cocooned themselves. Most
functionaries dared not submit truthful reports lest they be blamed for the problems
identified in them. In 1988 and 1989, almost all failed to inform the Politburo accurately
of the deteriorating situation.

Several of the most horrific dictatorial policy failures, such as the famine
caused by the Great Leap Forward in China and the mass starvation during Pol
Pot’s rule in Cambodia, resulted partly from local cadres’ unwillingness to
report appalling policy failure on the ground (Kiernan 1982; Manning and
Wemheuer 2011).
To solve some of their problems with policy implementation, monitoring,
and information gathering, dictatorships establish seemingly democratic
institutions, such as elections, mass parties, and legislatures, as we detail
later. In order to engage ordinary people, institutions dependent on popular
involvement must distribute benefits to motivate citizen participation. In
return for access to benefits, help, and opportunities, citizens join the ruling
party, vote, and demonstrate regime support in other ways. Mass organiza-
tions also facilitate mostly nonviolent forms of coercion and social control.
The monitoring of local officials and information gathering are by-products
of the more visible distributive and mobilizational functions of popular
institutions.
In this chapter, we discuss some of the institutional arrangements used in
dictatorships to enmesh, coopt, and gather information from ordinary citizens.
We first describe the uses of ruling parties to routinize the exchange between the
political elite and citizens of benefits for loyalty and service, while solving some
of the dictatorial elite’s implementation and monitoring problems. Next, we
discuss how legislatures fit into the system of local information gathering and
distribution of resources to citizens. Then we show the way elections incentivize
the extension of patron–client networks down to the village and neighborhood
level and the transfer of information from the grassroots to the center. We
demonstrate that elections, even elections without choice, increase the benefits
that reach the grassroots. This chapter thus focuses mostly on the deployment
of positive incentives to encourage information gathering, monitoring, and
citizen support.
We also explain the staying power of these institutions once they have been
established. The institutions that engage citizens tend to stabilize and become

costly to change, as large numbers of individuals develop an interest in main-

taining them. The tendency of mass institutions toward stabilization contrasts
with the fragility of agreements made within the ruling group, where tempta-
tions to renege on agreements are ever-present. The stability of mass institutions
contributes to regime routinization.

elite competition and institutions that

engage citizens
As members of the inner circle strive to improve their standing relative to others
in the elite, they seek new sources of information and new ways to bolster the
clientele networks that support and depend on them. Formal political insti-
tutions that reach, engage, and collect information from ordinary citizens can
contribute to achieving these goals. Besides newly created ruling parties,
explained in Chapter 5, such institutions include inherited ruling parties, elec-
tions, and legislatures. The institutions that extend distributive, mobilizational,
and information-gathering capacity to the grassroots are run by, and report to,
particular members of the dictatorial elite. They increase the power of the
members of the inner circle who control them because they create resources
that are useful in inner-circle competition. Institution creation generates jobs
for supporters and thus helps the individuals who lead them to build their
individual clientele networks. Institutions also generate new information
streams that benefit those with access to them.
Besides these advantages to particular individuals, outreach institutions are
expected to help secure the regime,1 which is the reason members of the inner
circle who do not benefit directly from them agree to their creation. These
institutions can contribute to regime persistence by distributing benefits to
ordinary citizens – and thus reducing their likelihood of joining opposition
campaigns or uprisings – by increasing the likelihood of discovering plots and
by reducing predatory behavior by local officials.

parties
In Chapter 5, we focused on the use of newly created parties to counterbalance
factionalized military support. Here, we emphasize other uses of parties: their
information-gathering, mobilizational, and distributive activities.
The obvious function of dictatorial ruling parties is to distribute benefits to
the nonelites on whom the dictatorship’s survival depends. The party routinizes
the exchange of material benefits (such as salaries, profit opportunities, favors,

1
A number of studies have shown that dictatorships supported by parties last longer than those
that lack a support party (Geddes 1999; Gandhi 2008; Magaloni 2008). All members of the
dictatorial elite beneﬁt from this.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
132 Ruling Society: Implementation and Information Gathering

special access, various kinds of goods, and services) for the loyalty, effort, and
time of party activists and members. The ruling party organization monitors the
performance of party cadres to make sure the dictatorial leadership is getting its
money’s worth from those to whom it provides benefits – in theory if not
always in practice.
If the party led the seizure of power, then members of the party need to be
rewarded for the risks they took and the privations they suffered to make the
seizure possible. Regardless of how the dictatorial elite achieved power, how-
ever, the dictatorship continues to need active effort by loyalists to handle
policy implementation, monitoring both officials and society, and information
gathering as well as the organization of winning votes and the other displays of
support that deter overt opposition.
Demonstrations to support dictators are rarely spontaneous. They require
the work, organization, and logistical skills of many people, and these people
have to have already developed their links to the rest of the community. It thus
makes sense for dictatorships to use ruling-party networks to orchestrate
support demonstrations. Such demonstrations serve serious purposes. Like
election victories, they signal strength. Outsiders often respond to obviously
orchestrated displays of popular support with puzzlement or ridicule because
they assume the demonstrations aim to fool people about the dictatorship’s
popularity. We think this reaction reflects a misunderstanding. Such demon-
strations show the strength of leaders and the regime in that they demonstrate
the resources and organizational capacity to turn out huge crowds, choreo-
graph their activities in minute detail, and prevent unwanted demands or
unruliness from arising during mass actions that bring many thousands of
people into face-to-face contact where they could potentially share grievances
and plot unrest.
We interpret the over-the-top displays of grief and gargantuan demonstra-
tions after Kim Jong-Il’s death and his son’s succession in North Korea,2 for
example, as a costly signal aimed at two different audiences during a time of
regime weakness. One signal aimed to show foreigners that the population
would defend the regime if it was attacked. The other signaled the resources
and commitment of the faction supporting Kim Jong-Un to members of the elite
who doubted the wisdom of choosing a politically inexperienced twenty-
something as regime leader.
The need to organize demonstrations and election victories makes inter-
mittent demands on party activists, but policy implementation, monitoring,
and information gathering are their everyday tasks. Central leaders expect to
solve the implementation problem by appointing party members to adminis-
trative jobs. Leaders assume that people who helped put them in power share

2
“North Korean Leader Kim Jong-il Dies ‘of Heart Attack,’” 2011, BBC News (December 19),
www.bbc.co.uk/news/world-asia-16239693 (accessed November 20, 2015).

their ideas and interests and will therefore be loyal. The ruling group tries to
assure the loyalty of officials and policy implementation on the ground by
limiting government jobs and official posts to party members. They hope
that shared dependence on the ruling party will align officials’ incentives
with their own and thus prevent sabotage and noncompliance with central
directives.
If the party lacks sufficient educated members, they must continue to rely on
some of the ousted government’s employees but use party militants to monitor
as much of their behavior and performance as they can. Monitoring is not
limited to officials whose loyalty is suspect, however. Ruling parties also try to
monitor their members who hold official positions to prevent shirking, stealing,
incompetence, and other human frailties.
Cadres assigned to locations outside the capital are also expected to monitor
local conditions and local people for signs of disgruntlement or opposition.
Local officials are the main source of information about the grassroots for
leaders in the capital. Officials are required to report on local affairs, expres-
sions of opposition, strangers visiting, movements of people through the area,
and anything else central authorities might consider signs of impending danger
as well. In this way, they contribute to the information available to leaders.
Leaders may also use party loyalists as “a network of unpaid spies and inform-
ers [in order] to keep all potential enemies under surveillance” (Lewis
1980, 150).
Central leaders assign party cadres many other tasks as well, for example,
explaining policy choices to other citizens, and thus building support and
compliance with them. The dictatorial elite expects party activists to “per-
suade the masses to . . . fulfill obligations to the state as well as to comply
with laws and party resolutions. At the same time they are expected to give
higher-level party authorities accurate feedback about both basic needs and
concerns of the population and its reactions to party policies” (Porter
1993, 71).
The ruling party usually controls local governance through its control over
the choice of local officials. Local administration can include law and order,
provision of social services and emergency aid, hiring in local offices and
schools, and allocation of basic infrastructure like electricity and piped water.
If parties are well developed enough to control so many things of importance in
daily life, citizens have strong reasons to cooperate with the ruling party and to
refrain from overt opposition.
Party cadres are supposed to provide sufficiently good local governance and
distribute enough to ordinary people to prevent the mobilization of opposition.
Some mass parties also incorporate large numbers of ordinary people in net-
works that distribute smaller or intermittent benefits. Typically, party networks
distribute whatever benefits the dictatorship makes available to nonelite sup-
porters during election campaigns and via party control of local government.
Party activists trade material benefits to citizens in return for political support,

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
134 Ruling Society: Implementation and Information Gathering

just as in democracies.3 Regime leaders want their parties to develop clientele

networks down to the grassroots in order to make sure goods reach the masses
and information is collected.
These are important and time-consuming tasks. Doing them well requires
effort and skill as well as loyalty. Party cadres must be paid for their time and
effort. The payment of salaries and other kinds of beneﬁts to party cadres builds
and maintains a fairly wide network of people who receive something of value
from the dictatorship and therefore tend to support it. Public employees and
ofﬁcials, who are ruling-party members in most dictatorships organized by
parties, often form the core of popular support for dictatorships. Those
employed by the party tend to be the dictatorship’s most committed supporters
because their own well-being depends on regime survival.

The Limitations of Dictatorial Ruling Parties

Dictatorial ruling parties often come up short with regard to accomplishing all
the tasks central elites assign to them. Many of the reasons for their failure to
perform well lie in human nature. The dictator himself may cause additional
problems with the quality and motivations of party cadres, however.
Because parties have so much mobilizational potential, dictators exercise
vigilance to make sure the party support base remains their own rather than
being used by those with the responsibility for day-to-day party leadership or
their allies. To prevent other members of the inner circle from using party-
organized mass mobilization on their own behalf, dictators often interfere with
party leadership. Most of the time (80 percent),4 the dictator himself leads the
ruling party.
The party strategy of General Francisco Franco of Spain provides an
example of such vigilance and its consequences. Franco delegated the task of
creating a support party from the civilian groups that backed his military
uprising to his brother-in-law, making certain that the party did not become a
challenging center of power during the early days when a chaotic political
environment made that most possible. After its ﬁrst years, when the regime
had stabilized, Franco selected new leaders, reduced the importance of the
party, and halved its budget (Payne 1987). Throughout the long years of his
rule, Franco used the party to counterbalance his military support base, care-
fully preventing either one from becoming the dominant force in the dictatorial
inner circle. By the 1950s, “no one belonged to the party who did not in some
way make a living from it” (Payne 1961, 262). Franco, however, repeatedly

3
Although most dictatorships use a party to organize distribution, dictators who have chosen not
to organize a party can instead use administrative ofﬁcials to distribute state resources to buy the
cooperation of citizens. Several Middle Eastern monarchies have used this strategy.
4
This number differs from the one noted in Chapter 5 because it refers to the proportion of all
dictatorial ruling parties, not just those created after seizures of power.

resisted party leaders’ efforts to rebuild it into a more engaged organization that
could attract idealistic adherents and thus potentially increase its political heft.
The instrumental nature of most party members’ loyalty did not hinder the
party’s ability to turn out huge, cheering crowds for Franco to harangue, an
activity he enjoyed (Payne 1987).
The Spanish example also illustrates a very common reason for the failure of
dictatorial ruling parties to carry out all the tasks assigned to them. People join
them for instrumental reasons. Once dictatorships are established, opportunists
typically swamp even the parties that attracted very idealistic and committed
members before seizing power. Some dictatorships, notably those controlled by
communist parties, invest heavily in screening applicants and monitoring
members’ behavior, but most do not. Parties need not be all-encompassing or
highly disciplined organizations in order to be useful to dictators, and most
dictatorial ruling parties are not. Even if the party cannot be used as an
organizational weapon, it still builds robust patron–client networks linking
regime insiders to party members. The delivery of benefits via these patron–
client networks creates vested interests in regime survival.
These vested interests, however, often fail to reach all areas of the country
because many real-world ruling parties have lacked the number of trained and
disciplined cadres needed to penetrate society effectively, especially during the
first years after seizures of power or following post-seizure party creation. As a
result, the number and location of people incorporated into the dictatorship’s
support network can be quite limited. Moreover, central regime leaders often
lack the ability to monitor the behavior of local party cadres. The task of
distributing goods to citizens creates opportunities for party cadres and local
officials to steal, embezzle, and abuse their power. As we explain below,
elections help to reduce abuses that most dictatorial ruling parties could not
otherwise control.

Why Dictatorial Ruling Parties Persist

Despite these limitations, all dictatorial ruling parties control the allocation of
some benefits that citizens value. Even those that have not successfully
developed grassroots networks create strong attachments with those people
whose jobs or other benefits depend on the ruling party. As a result, if parties
are established, they tend to become self-sustaining. Organizing a ruling party,
allocating resources to it to pay employees and distribute some benefits to
others, and building the networks needed to link the regime inner circle to local
leaders create widespread vested interests in the party’s persistence. Citizens
want to continue receiving whatever benefits the party delivers. Elites who
occupy high offices in the ruling party would be alienated by losing their posts,
which might lead to efforts to unseat the dictator.
In dictatorships organized by parties, all members of the ruling group –
whether officers or civilians – derive some benefits from the party, but one

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
136 Ruling Society: Implementation and Information Gathering

member leads it, and a few control key party resources, most importantly,
appointments to government jobs and party offices. Whoever controls the party
has some influence on membership in the inner circle itself, as well as lower
levels of leadership and vast numbers of jobs in both government and party.
Nor would the dictator usually want to eliminate a counterbalance to the
military. Disbanding the party would leave the military unchecked, and thus
a greater threat to him.
Consequently, once in existence, regime support parties are rarely dis-
banded. Indeed, only 10 percent of dictatorships with support parties later
dissolve them. Where they are disbanded, regime insiders often replace them
with a new party led by many of the same people.
To sum up, parties seem to be an effective way to organize regime supporters
and to routinize the distribution of benefits to them and the gathering of
information from them. From a dictator’s point of view, a party that organizes
quotidian, largely instrumental support may be safer than one that can mobilize
great activism and idealism, unless the dictator himself feels confident about his
ability to control and direct it. Many dictators, like Franco, talk the talk of
heroic party activism, but in practice rely on the government bureaucracy for
making policy and allot to the party only humdrum tasks. Even where ruling
parties have little real capacity to affect policy and leadership choice,5 they tend
to persist because they deliver benefits both to the dictatorship’s inner circle by
increasing regime stability and to party activists and members.

dictatorial legislatures
Some analysts have argued that dictatorial legislatures serve as arenas for policy
bargaining (Gandhi and Przeworski 2006; Gandhi and Przeworski 2007;
Gandhi 2008; Malesky and Schuler 2010). Case studies and interviews with
deputies often suggest a different picture, however. The Indonesian parliament
during the long Suharto dictatorship, for example, “never drafted its own
legislation and . . . never rejected a bill submitted by the executive branch. It
has no say in cabinet appointments [and] little inﬂuence over economic policy”
(Schwarz 2000, 272). In Guinea Bissau, the assembly provided “pro-forma
electoral acclamation of those appointed by the [ruling] party to government
posts. It also [served] as a public forum for the airing of party propaganda and
for the routine yes-voting on party policies” (Forrest 1987, 113). During the
Diori dictatorship in Niger, “the assembly never served as a serious forum for
debate” (Decalo 1990, 257). Ellen Lust-Okar argues that in dictatorial legisla-
tures “competition is not over policy-making. Many (and in some cases most)
policy arenas are off-limits to parliamentarians . . . Rather . . . [legislatures]

5
See Brooker (1995) for a comparison of the ruling parties in a number of party-based autocracies
with regard to how large a role they play in policy choice.

provide an important arena for competition over access to state resources”

(2006, 459).
Interviews with deputies reinforce Lust-Okar’s claim. In response to the
question “To what extent do you believe that the parliament is able to influence
the government?” only 7 percent of Jordanian deputies thought parliament
played a large role (Lust-Okar 2006, 470). An Ivoirian deputy responded to
questions about his activities as follows: “Individual deputies have no business
trying to discuss [policies made by the ruling party politburo] in the Assembly
because rhetoric is a waste of time and could even be harmful” (Zolberg 1964,
282). Another reported that deputies are “concerned mainly with gaining
access to . . . tangible benefits for their constituents” (Zolberg 1964, 283).
Our reading of the evidence is that most authoritarian legislatures play a role
in the allocation of private benefits and local public goods to citizens, but that
they have little influence on policy (Truex 2016).
The most valuable function of legislatures to dictatorships may be that they
incentivize competition among regime supporters for the opportunities the
legislature makes available to deputies themselves. The ruling elite’s decision
to elect a legislature creates highly desirable plums to be distributed to party
militants by the dictator or those to whom he delegates the task of choosing
candidates. Deputies usually receive salaries and other perquisites such as the
use of cars and subsidized housing, as well as access to many kinds of oppor-
tunities and favors that can be used to help their friends, further their careers,
and accumulate personal wealth.6 So, ambitious people have strong reasons to
compete for nominations and office. Ruling elites use this competition to
motivate deputies to extend their distributive networks to the grassroots and
transmit information to the center from the grassroots, as we describe in the
next section.

elections
Most dictatorships led by parties have regular popular elections. Until about
1990, most held single-candidate or single-list elections that gave voters no
choice at all. Even elections without choice involve substantial costs because a
campaign has to be organized to reach all parts of the country, and in most
dictatorships, elites want to be able to claim high rates of voter participation.
The frequency of elections in dictatorships, especially choice-free elections,
raises the question: What are they for?
The dictatorships that held regular choice-free elections included the com-
munist regimes and a number of one-party dictatorships in developing coun-
tries. Before the end of the Cold War, only about one-quarter of party-led

6
See Zolberg (1964, 192–93), Lust-Okar (2006), and Blaydes (2011) for discussions of the beneﬁts
of elective ofﬁce.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
138 Ruling Society: Implementation and Information Gathering

autocracies allowed opposition in legislative elections. The proportion that

allows some opposition has risen to nearly two-thirds (61 percent) since the
end of the Cold War.
When voters have some choice but the regime outlaws important parties,
restricts suffrage to prevent substantial numbers of people from voting, or tilts
the electoral playing ﬁeld in ways that give the ruling party substantial advan-
tages, we label elections semi-competitive. Dictatorships have inventive ways to
bias election outcomes. Current semi-competitive electoral systems include
those that:
Permit all opposition parties to compete but use control of the media,
interference with opposition campaigning, fraud, violence, and large-scale
state spending to bias outcomes
Permit some parties to compete but not others (e.g., regimes that outlaw
popular Islamist parties while allowing secular opposition)
Permit no opposition parties but allow independents to contest elections
Permit competition among ruling-party candidates but not opposition
parties.
Terms such as electoral authoritarian, competitive authoritarian, and quasi-
democratic refer to dictatorships that hold regular semi-competitive elections,
or sometimes a subset of such systems.7 Since the end of the Cold War,
international donors have tied foreign aid and other resources to holding
elections that allow some competition. Many dictatorships also receive help
in paying for elections. Indeed, some observers have suggested that aid offered
to induce holding multiparty elections has become an important source of illicit
wealth for the dictator’s cronies in some regimes.8
Figure 6.1 shows the per capita amounts of aid (in constant US dollars) going
to dictatorial regimes in which the paramount leader had won a semi-
competitive multiparty election (solid line), a one-party election (dashed line),
or no election (dotted line).9 The last category includes regimes such as com-
munist dictatorships, where ruling party elites choose regime leaders (though
legislatures were usually elected in single-party elections); about half of
military-led regimes; and monarchies. Until the early 1990s, dictatorships that
held uncontested executive election rituals, such as Egypt during much of the
time after 1952, received the most aid per capita. Post–Cold War, however,

7
See Ezrow and Frantz (2011) for a review of the terms used to capture different types of electoral
dictatorships.
8
Wiseman, for example, sees “a real danger that foreign assistance for elections can simply
become a new avenue for personal accumulation by state elites; in practice a new type of ‘rent’”
(1996, 940).
9
Aid per capita is measured using a three-year moving average. We show the median instead of
the mean level to reduce the inﬂuence of outliers. Substantial Western (US) aid ﬂowed to
Afghanistan, Iraq, and Pakistan from 2001 to 2010.

80
Aid per capita (median)

1960 1970 1980 1990 2000 2010

Year

Multiparty election One party election Not elected

figure 6.1 Foreign aid and the election of dictators.

dictatorships that hold semi-competitive executive elections, and can thus claim
to be taking steps toward democracy, receive the most aid.10
This pattern reflects two changes that occurred in the late 1980s and early
1990s: donors withdrew aid from dictatorships that refused to hold multiparty
contests, and many autocracies held multiparty elections for the first time.
Substantial research suggests that changing donor behavior prompted many
dictatorships to allow multiparty contests.11 Sometimes these elections led to
electoral defeat for the dictator and democratization. But often, dictators
successfully navigated the imposition of multiparty elections and remained in
power. Andrea Kendall-Taylor and Erica Frantz (2014) show that since the end
of the Cold War, dictatorships increasingly rely on pseudo-democratic insti-
tutions and that this accounts for part of the global increase in autocratic
survival rates.
Even when some competition is allowed, dictatorial elections rarely deter-
mine either who rules or the government’s policy strategy. Their prevalence,
even before the end of the Cold War and despite the risks they involve, suggests
that they have other important functions, however. In this section, we highlight
two of these functions. First, presidential elections are costly signals of the
dictator’s strength, aimed at deterring elite defections from the ruling coalition
(Geddes 2005; Magaloni 2006). Second, legislative and local elections create
incentives that might not otherwise exist for party officeholders and cadres to

10
Levitsky and Way (2010) show strong evidence that semi-competitive elections do not imply that
democracy lurks around the corner.
11
See Dunning (2004), Wright (2009), Bermeo (2011, 2016), and Dietrich and Wright (2015).

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
140 Ruling Society: Implementation and Information Gathering

extend their patron–client and information-gathering networks to the grass-

roots, which helps authoritarian regimes to survive.
Big election wins for the dictator demonstrate his capacity to run a successful
mobilization campaign, which deters both elite and mass opponents from active
opposition by raising the perceived probability of regime survival. If an incum-
bent victory appears inevitable, potential opponents are less willing to shoulder
the costs of overt opposition. Wins against opposition generate the strongest
signals of regime strength, but even elections without choice show that the
dictatorship has the resources and organizational capacity to ensure mass
voting. To be effective in deterring opposition, vote and turnout counts need
to be reasonably honest (Magaloni 2006). However, dictatorships can appar-
ently tilt the playing field by controlling the media, concentrating state
resources on supporters, and harassing, threatening, jailing, or beating up
opposition activists without undermining elections’ usefulness in deterring elite
defection. In contrast, visibly fraudulent vote counting has sometimes set off
explosions of popular hostility, bringing down dictatorships (Tucker 2007;
Rozenas 2012).12
Most dictatorships that hold executive elections also elect a legislature and/
or local officials as well, usually at the same time. Although ruling parties and
dictators seldom lose semi-competitive national elections, individual ruling-
party candidates for the legislature or local office do sometimes lose. Moreover,
competition for ruling-party nominations is often far more fierce than partisan
competition. Even where a dictatorship holds choice-free elections, ambitious
party militants compete with each other for party nominations. In the Ivory
Coast under one-party rule, for example, about ten candidates competed for
each party nomination. “Although electoral competition has been eliminated,
in the realm of recruitment [i.e., nominations] competition still exists . . . This
has insured that the system remains responsive to the demands of various
groups in the population” (Zolberg 1964, 272).
The real possibility of losing in either the competition for nomination or
semi-competitive elections creates incentives for elected officials to deliver
benefits to citizens, as Aristide Zolberg observed. During the decades when it
held no national elections, the Kuomintang (KMT) regime in Taiwan
“promoted electoral participation at the local level, using elections to pressure
party cadres . . . Chiang [Ching-kuo] openly stated that the threat of electoral
punishment was necessary to force cadres to jettison ‘old conceits,’ internalize
new attitudes, and consolidate the party’s broadening social base” (Greitens
2016, 101).

12
Examples include the ousters of Slobodan Milošević in Serbia and Eduard Shevardnadze in
Georgia, after the announcement of election results seen as fraudulent (Kuntz and Thompson
2009). Perceptions of election fraud have also led to massive unsuccessful protests that were
nevertheless costly to the ruling group, as in Iran in 2009 and Venezuela in 2013.

Party leaders’ nomination decisions depend on the potential candidate’s

behavior in his previous office or as a party organizer, as well as loyalty and
other factors. In order to hang on to their posts, officials in most dictatorships
have to ensure high turnout even in elections without choice and substantial
majorities for the ruling party when voters have choices. Low turnout, low vote
shares, widespread absenteeism from pro-regime events, opposition demonstra-
tions, and other manifestations of citizen anger in their districts alert central
leaders to the faults of local officials and reduce their chance of renomination.
In addition, any sort of electoral competition, whether with tame opposition
parties or other candidates from the ruling party, gives voters the opportunity
to throw out incumbents who have disappointed or exploited them.
Competition thus puts pressure on local officials and deputies to refrain from
exploitation and brutality toward constituents and compete on behalf of their
areas in the national scramble for schools, clinics, paved roads, and whatever
else is being given out. They need to deliver some benefits and provide some
local public goods in order to ensure turnout, votes for the ruling party, and
participation in ruling-party rallies. The need to deliver benefits to voters gives
ambitious officials and candidates incentives to convey the needs and problems
of their districts to central leaders and build clientele networks to reach the
grassroots. Legislative and local elections can thus improve the dictatorship’s
information-gathering capacity and add many more people to the group receiv-
ing benefits from the dictatorship.
Elections help dictators overcome their problem monitoring local officials as
well. “[E]lections are a way of obtaining information . . . [T]hey provide an
occasion for inspection of the party structure at the local level” (Zolberg 1964,
272). The failure of local people to turn out or vote for ruling-party candidates
alerts regime leaders to officials’ shirking or bad behavior and initiates investi-
gation of local problems. By creating these incentives, local and legislative
elections partially substitute for monitoring local officials, with citizens’ votes
serving as “fire alarms.”13 Elections thus give local officials self-interested
reasons to behave as regime leaders want them to. As several scholars have
noted, ruling elites have an interest in controlling predation by local officials
because it contributes to popular opposition to the regime.14
In other words, elections help align the incentives of deputies and local
officials with those of the center by giving them strong reasons to limit leakage
in the transfer of benefits from the center to their regions, treat constituents

13
McCubbins and Schwartz (1984) coined the term “fire alarm” for the use of complaints or
appeals to alert principals to the misbehavior of their bureaucratic agents in situations in which
continuous monitoring would be costly.
14
Note the number of analyses of China’s ruling party’s use of citizen petitions and internet
freedom to complain about officials as a means of monitoring local officials (e.g., Paik 2009;
Lorentzen 2014). China is one of the few contemporary party-led dictatorships that holds no
direct elections above the village level.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
142 Ruling Society: Implementation and Information Gathering

decently, and convey information about the needs and grievances of voters to
central decision makers. After an election, the knowledge that they will face
renomination and reelection gives legislators and local elected officials reasons
to reinforce their clientele networks, compete for resources from the center for
their districts, and do favors for local notables. Local officials’ need to attract
nominations and votes at regular intervals reduces the temptation to hide local
complaints from the center, steal goods intended for distribution, abuse their
authority over people in their districts, and exploit those dependent on them. In
other words, even uncompetitive elections help to limit principal-agent prob-
lems between central authorities and local officials in contexts in which moni-
toring would be both expensive and often ineffective.
In regimes with uncompetitive elections, information about local officials is
conveyed mostly by real (as opposed to reported) turnout (Magaloni 2006). In
most dictatorships, low turnout means either that local officials and activists
have not done their job of making sure that people go to the polls or that people
are hostile enough to risk penalties for failing to vote.15
Elections, even if uncompetitive, thus provide regular opportunities for the
generation of information useful to regime elites about how local officials have
performed, whether policies work on the ground, and levels of disenchantment
among citizens in different areas of the country, as a number of analysts have
noted (Gandhi and Lust-Okar 2009). We think that information about lower-
level officials and policy implementation has more value to leaders than the
distorted information elections provide about popular opposition to the dicta-
torship. We do not consider elections good sources of information about citizen
opinion because votes depend on which alternatives exist, whether people see
ending the dictatorship as possible, and other strategic considerations. Besides,
dictatorships need daily information about opposition, not information at
multiyear intervals. This is the reason they invest so heavily in internal security
agencies (discussed in Chapter 7).
The benefits that ordinary citizens receive because of elections are
by-products of competition among politicians in dictatorships, just as in dem-
ocracies. Most of this distribution flows through politicians’ clientele networks,
happens in person at campaign events, or involves the politically motivated
allocation of local public goods, so it is tangible and visible to recipients and
those close to them.
Election campaigns are a predictable time when citizens can expect to receive
something, beyond whatever services and public goods they usually enjoy, in
exchange for their votes. During election campaigns, many people who have no
interaction with officials or party activists most of the time receive extra food

15
There are a few dictatorships in which leaders seem to encourage turnout only among those
whose votes are easiest to manipulate, e.g., during Mubarak’s rule in Egypt (Blaydes 2011). As
an example of how seriously most dictators take turnout, Cameroonian President Paul Biya has
dismissed cabinet ministers to punish them for low turnout in their constituencies (Jua 2001, 39).

and entertainment in autocracies, just as they do in many democracies. For

example, Egyptian calorie consumption rose during election campaigns in the
Mubarak dictatorship (Blaydes 2011). Belarusian president Aliaksandr Luka-
shenka “fixed elections, but still spent money on his campaign as if he were part
of a real contest” (Wilson 2011, 196). He staged a six-week pop music tour to
entertain voters during campaign events in 2010 (Wilson 2011, 219). Salva-
doran military dictator Julio Rivera “campaigned as hard as, or perhaps harder
than, he would have had he been opposed” (Webre 1979, 47). Paraguayan
dictator Alfredo Stroessner “stumps the country as though he were in a real
race . . . All his appearances are surrounded by a great deal of hoopla – fiestas,
dances, barbecues, parades . . . and stirring polkas praising Stroessner’s deeds”
(Lewis 1980, 106).
As with other institutions that engage ordinary people, elections tend to
become routinized and predictable over time, even though dictators may have
initially introduced them only to placate foreign donors. The ad hoc referenda
held after violent seizures of power, which new dictators never lose, are seldom
more than opportunistic attempts to use the appearance of popular support to
give pause to opposition elites, international critics, or ambitious rivals in the
inner circle. Once elections have occurred a couple of times, however, eliminat-
ing or postponing them has a much different meaning than failure to hold
elections in the first place.
Because successful elections are a signal to potential elite defectors of the
regime’s invincibility, postponing scheduled elections signals a regime or dicta-
tor in difficulties. Postponement implies that regime leaders fear defeat.
A postponed election has the same political effect as the announcement that
the dictator has suffered a mild stroke: it sets off a covert struggle for power
among potential successors, increases plotting, and motivates efforts to cooper-
ate among regime opponents. Indeed, irregular elections (which include those
held after a postponement)16 are twice as likely to be followed by regime
collapse as regular ones, and leaders who preside over irregular elections are
more than twice as likely to face a bad post-exit fate (death, imprisonment, or
exile) compared with leaders who hold regular elections. For these reasons,
dictatorships rarely postpone scheduled elections, even when growth has stalled
or other problems reduce their ability to control election outcomes. Only
8 percent of dictatorships that held one executive election failed to hold others.
If they fear losing elections, dictators have a number of tools safer than
postponement at their disposal. They can alter electoral rules, increase their
control over media, and disqualify opposition candidates, usually without
serious consequences.

16
Unfortunately, available data do not allow us to distinguish postponed elections from other
irregular ones.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
144 Ruling Society: Implementation and Information Gathering

Tests of the Effects of Elections

In what follows, we test some implications of the argument that dictatorships
use elections to incentivize the extension of their information-gathering and
distributive networks to the grassroots. We reason as follows. If elections are a
costly signal of the dictator’s invincibility, we would expect to see central
government spending increase during election years as officials demonstrate
their ability to organize and mobilize the masses. If sending a costly signal was
the only purpose of elections, however, we would expect to see much higher
spending during semi-competitive election years (when a costlier signal is sent)
than during elections without choice. On the other hand, if an important
function of elections is to incentivize local candidates to extend distribution
to the grassroots, we would expect to see high spending during all elections,
regardless of competition because distribution to the grassroots is costly. Our
second test compares health spending in countries with dictatorships that hold
either uncompetitive or semi-competitive elections with those in which no
elections are held. If elections incentivize spreading benefits to the grassroots,
we should see evidence that health spending reaches ordinary people. To test
this expectation, the third test looks at whether elections actually affect citizen
welfare. We use child mortality as an indicator of welfare.
We first compare government spending in election and nonelection years. If
elections motivate additional distribution to citizens, government spending
should rise in election years. Other analysts have found evidence in a number
of specific dictatorships of a political-business cycle in which public spending
rises before semi-competitive elections (Heath 1999; González 2002; Magaloni
2006; Pepinsky 2007).17 To investigate whether this is a general phenomenon,
we compare average spending in election years with spending in nonelection
years in dictatorships with support parties.
To pinpoint the electoral mechanism, we identify the years of first-round
presidential and/or legislative elections in which the incumbent party, regime
leader, or a hand-picked successor was on the ballot, relying on data from the
NELDA project (Hyde and Marinov 2012).18 These elections include both
uncompetitive and semi-competitive events.

17
The foundational literature on political budget cycles focuses on democracies, positing that
myopic voters use retrospective voting to sanction incompetent politicians. Knowing that voters
do this prompts incumbents to pursue expansionary economic policy prior to elections
(Nordhaus 1975). The evidence for political budget cycles in advanced industrial democracies
is weak, but political budget cycles are more prevalent in new democracies (Block, Ferree, and
Singh 2003; Brender and Drazen 2004; Shi and Svensson 2006). Analysts suggest that lack of
transparency, high personal rents for retaining ofﬁce, and a large share of uninformed voters
provide stronger incentives for incumbents to pursue ﬁscal manipulation before elections. The
same conditions characterize many dictatorships.
18
We use nelda20, nelda21, and nelda22.

We restrict the analysis to regularly scheduled elections, excluding any that

had been scheduled but then postponed or canceled (nelda1 and nelda6). This
qualification is important because the timing of both elections and spending
decisions may be determined by unmodeled factors such as political instability,
international shocks, or economic crises. For example, a rise in the inter-
national price of a country’s main export might motivate a dictator to call
elections while times are good and also encourage increased public spending.
We want to exclude the possibility that dictators time elections to surf waves of
popularity rather than manipulating spending at election time (Kayser 2005).
The possibility of opportunistic election timing is of special concern for irregu-
larly held elections when, by definition, the election date is not fixed ex ante, so
we exclude those. In the sample period, 1961–2010, there are 174 regular
election years.
To investigate whether elections influence the distribution of benefits, we
examine general government spending, measured as logged per capita expend-
itures. We include control variables for decade, leader duration (logged), the
age dependency ratio, trade (% GDP), civil and international conflict, GDP per
capita (logged), population size (logged), and oil rents per capita.19
Standard empirical specifications in the political business cycle literature
employ a lagged dependent variable model (Brender and Drazen 2004; Alt
and Lassen 2006; Shi and Svensson 2006; Hyde and O’Mahoney 2010). The
dependent variable is typically a change variable as a proportion of the total
economy, for example, △Budget/GDP. The main independent variable is a
dummy for pre-election, election, or post-election year. We depart from this
specification in three ways, while retaining the lagged dependent variable and
change in spending as the dependent variable. First, we purge the dependent
variable of the GDP measure in the denominator to allow for a more transpar-
ent test of the election effect. A measure of economic size and population are
included as right-hand-side variables to ensure that the estimated effect of
elections is conditioned on country size. Second, we estimate an error-
correction model that allows for a more general test of both the long- and
short-run impacts of elections. Empirical models that include only a dummy
variable for elections on the right-hand side of the equation estimate the total
effect of elections without separating short-term changes from long-run equi-
librium relationships (De Boef and Keele 2008). Finally, because of the paucity
of data on budget balance in dictatorships, we estimate the effect of elections on
spending, where the dependent variable is the change in ln(Expenditure)

19
Data on age dependency ratio, trade, GDP, and population are from the World Development
Indicators (2015). Conﬂict data are updates to Gleditsch et al. (2002), and oil rents data are from
Wimmer, Cederman, and Min (2009).

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
146 Ruling Society: Implementation and Information Gathering

All incumbent elections Elections with and without opposition

8 8
% change in Govt. spending

% change in Govt. spending

4 4

2.1 1.9 2.1

.3 .3
0 0

–4 –4

Election year t+1 t+2 No opposition Opposition contested

Years since election Election year

figure 6.2 The electoral spending cycle in dictatorships.

measured in constant dollars, using data from the World Development Indica-
tors (2015).20 The speciﬁcation is:
Δlnð Exp Þ ¼ lnð Exp Þt−1 þ ΔE þ Et−1 þ ΔX þ Xt−1 þ δt þ ϵi,t ð6:1Þ
where E is a binary indicator for election year, X is a vector of control variables,
and δt are decade effects. We estimate this equation with a linear model that
corrects for heteroskedasticity and autocorrelation in the errors.
The left panel of Figure 6.2 shows the results. The plot shows that dictator-
ships, on average, increase spending during election years, as expected. Expend-
itures move closer to the long-run equilibrium levels in the years following
elections, though there is still some increased spending after elections, as would
be expected if much of the extra expenditure in election years is for construc-
tion, infrastructure, and local public goods, which often have costs extending
over multiple years. These ﬁndings are consistent with both the claim that
dictatorial elites use spending and mobilization during election campaigns to
signal invincibility and the claim that dictatorships distribute to the grassroots
during elections.
Next, we compare changes in spending at the time of elections with and
without opposition participation. If authoritarian elections are only a way for

20
The main results for government spending are robust in specifications without control variables,
with regime-case fixed effects and AR(1) errors, and with year effects, as well as alternative error
estimators. We drop observations associated with the budget crisis in Zimbabwe in 2008, and
the subsequent fiscal rebound the following year, because these are clear outliers in the spending
data. We also drop the election in Senegal in 1968 because this a clear spending outlier. Adding
these observations back into the estimating sample yields a stronger result than that reported in
the right panel in Figure 6.2.

the dictatorship to demonstrate its support relative to the opposition (in order
to deter elite defections) or to identify regions of opposition party strength, we
would expect spending to increase before contested elections but not before
uncontested ones. If, however, elections help dictatorial elites to align the
incentives of local officials with their own, then we would expect to see
spending increase during elections regardless of opposition participation.
Because potential ruling-party candidates have to compete for nominations
even when elections are uncompetitive, we argue that uncompetitive elections
motivate candidate behavior and government spending similar to that of can-
didates who face opposition party competition.
The right panel of Figure 6.2 shows the results from a test that splits
regularly scheduled elections into two groups: those in which the dictatorship
allowed opposition participation and those in which no opposition contested
the election.21 The plot shows that dictatorships increase spending during both
types of elections. The estimates for each type of election are roughly the same
size as the estimate that pools both groups of elections together. However, the
estimates for contested and uncontested executive elections are not statistically
significant because there are fewer observations in each category when the
sample is divided. To further explore this finding, we also tested models that
separate all dictatorships with legislatures that include some opposition from
those without opposition (or with no legislature). In these tests (not shown), we
find that dictatorships that allow no competition at all still boost spending
during election years.22
The evidence about government spending thus suggests that both semi-
competitive and uncompetitive elections motivate increased effort by officials
to reach citizens with benefits. Through this mechanism, dictatorships can use
elections to monitor the behavior of local officials and buy support from
citizens.
The argument that elections help to monitor the behavior of local officials
implies that officials engage in less theft and abusive behavior in dictatorships
that hold elections, and therefore that citizen welfare would improve. Next, we
investigate this implication. We assess the effect of authoritarian elections on
citizen welfare. These tests focus on the average effects of different election
rules, not on cycles. The point of this examination is to compare the effects of
semi-competitive and uncompetitive elections, relative to holding no elections,
on the distribution of welfare-enhancing goods to citizens. If elections

21
Opposition-contested elections, as deﬁned here, are those that meet the following three criteria:
(1) an opposition party exists to contest the election (nelda3), (2) more than one political party is
legal (nelda4), and (3) there is a choice of candidates or parties on the ballot (nelda5). All other
elections are considered uncompetitive. There are 67 uncompetitive elections and 106
opposition-contested elections in the sample (1961–2010).
22
Replication ﬁles also report a test that looks separately at the years 1990–2010, in which
uncompetitive elections remain associated with an increase in government spending.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
148 Ruling Society: Implementation and Information Gathering

Child mortality Health spending

Semi-competitive Semi-competitive
electoral regime electoral regime

One-party/candidate One-party/candidate
electoral regime electoral regime

Regime duration Regime duration

GDPpc (log) GDPpc (log)

Legislature Legislature

Monoethnic party Monoethnic party

Multiethnic party Multiethnic party

–1.5 -1 –0.5 0 0.5 –0.1 0 0.1 .2

Coefficient estimate Coefficient estimate

Time trend, Add GDP, Add party Time trend, Add GDP, Add party
Region effects conflict Region effects conflict

figure 6.3 Dictatorial elections and health outcomes.

encourage politicians and ofﬁcials to extend their patron–client distribution

networks to the grassroots, as we have argued, we should see better welfare
outcomes associated with elections.
This analysis relies on child mortality rates, which are a good indicator of
overall popular well-being, and government spending on health care as a
measure of government effort.23 It compares dictatorships with unelected
leaders with those that hold semi-competitive executive elections and those that
hold one-candidate elections.24 The left panel of Figure 6.3 shows child mor-
tality rates, and the right panel shows central government spending on health.
The top coefficient in each cluster was generated by a model that contains only
controls for regime duration (logged), geographic region, and time period fixed
effects, to capture world time trends in child mortality and health costs.25 The
second estimate adds GDP per capita (in constant dollars, logged) and a
measure of the time since the last high-intensity conflict as controls.

23
IMF data on health spending, available from 1985 to 2009, come from Clements, Gupta, and
Nozaki (2011). Data on the under-ﬁve mortality rate (U5MR) come from the Institute of Health
Metrics and Evaluation and span the years from 1971 to 2010.
24
Dictatorships that hold semi-competitive executive elections nearly always also hold competitive
legislative elections, whereas quite a few dictatorships with semi-competitive legislative elections
do not have contested executive elections, so we use executive elections for this test. The sample
used for these tests is all dictatorships: 1971–2010 for child mortality and 1985–2009 for health
spending.
25
There are eighteen region effects corresponding to the Global Burden of Disease categories from
the Institute for Health Metrics and Evaluation. We do not include country effects because
84 percent of the variation in the child mortality rate is cross-sectional. However, a bivariate
model with country effects yields similar results.

The third estimate adds information about legislatures and regime support
parties. It includes an indicator variable for the existence of an elected legisla-
ture and indicators for ruling parties with support from just one ethnic group
(monoethnic) and those that are supported by multiple ethnic groups (multi-
ethnic). The reference category for the party variables is dictatorships without a
support party. If elections in this specification improve child mortality rates, it
means that elections affect welfare beyond any effect of legislative bargaining
directly on welfare or on economic performance (which would affect welfare
indirectly) and any advantage that dictatorships organized by parties have in
the delivery of services. Recall that our argument about the effect of elections on
the incentives of officials implies that more benefits should actually reach
citizens if the regime holds elections. Since the vast majority of dictatorships
that hold executive elections also have ruling parties, this is a hard test.
Figure 6.3 shows that child mortality rates are lower and health care spend-
ing higher in dictatorships in which executives face semi-competitive elections
compared with those in which they face no elections, regardless of the specifi-
cation. Even after controlling for the effects of legislatures and ruling parties,
elections improve welfare, as our argument predicts.
Our interpretation of the effect of uncontested elections must be a bit more
tentative because changing the specification affects results, but models that
account for level of development and conflict (bottom two coefficients in each
cluster) suggest that child mortality rates are also lower and health spending
higher in dictatorships with uncontested elections than in those without elec-
tions. The pattern shown here is what would be expected if even uncompetitive
elections create incentives for officials to attend to the welfare of ordinary
people, as we have argued.
The main results with regard to legislatures and parties are that they make
no independent contribution to child welfare or health spending when elections
are controlled for. Indeed, ethnically exclusive dictatorial ruling parties are
associated with higher rates of child mortality than exist in dictatorships
without parties. This finding suggests that when the leadership of a dictatorial
ruling party excludes some or most of a country’s ethnic groups, welfare-
enhancing benefits go only to included groups. We interpret the nonsignificant
coefficient for multiethnic ruling party as meaning that although ruling parties
distribute benefits, in the absence of elections, party cadres lack incentives to
extend distribution to poorer citizens.
We find no support for the claim that dictatorial legislatures serve as fora for
policy bargaining (Gandhi and Przeworski 2006, 2007; Gandhi 2008). If
bargaining in legislatures contributes to economic performance, as analysts
have proposed, it should lower child mortality rates, but we do not see evidence
of that. We take this to indicate that in most authoritarian legislatures, little real
policy bargaining occurs, or at least that whatever bargaining occurs has little
effect on public welfare. Dictatorships do negotiate with private economic
interests, of course, but primarily through informal personal contacts with

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
150 Ruling Society: Implementation and Information Gathering

business people, as documented in the many empirical studies of state–business

relationships in autocracies (e.g., MacIntyre 1994; Doner and Ramsay 1997;
Schneider 2004).
One of the reasons welfare-oriented policy bargaining rarely occurs in
dictatorial legislatures is that potentially challenging mass-based opposition
movements, of the kind discussed by Gandhi and Przeworski (2006, 2007),
have seldom had enough representation in dictatorial legislatures to bargain
there effectively. In less than a quarter (23 percent) of authoritarian legislatures
has the opposition held more than 25 percent of seats, as would be required in
order to bargain with any chance of success.26 Moreover, several of the dicta-
torships that allow the largest number of opposition deputies tolerate only
moderate opposition parties while excluding more popular and challenging
parties (Lust-Okar 2005). Rather than bargaining with the largest opposition
group, some autocrats survive by keeping the most challenging opposition
divided from more easily coopted moderates who are allowed legislative seats
and the private benefits that go with them. The underrepresentation of oppos-
ition in dictatorial legislatures may be the reason for our finding that distribu-
tive effort undertaken in the context of elections has more effect on popular
welfare than legislatures do.
In sum, the findings offered here suggest that though dictatorial elites use
elections strategically for the survival benefits they confer, elections also bring
some benefits to citizens as a by-product. Post–Cold War foreign aid may thus
have made some contribution to improving the lives of people living in dicta-
torships through incentivizing dictators to distribute more at election time,
regardless of whether it has been effective in encouraging democratization.

conclusion
Once elite bargaining has evolved into somewhat predictable patterns, dicta-
torships often face problems with the implementation of their policies, moni-
toring local officials to prevent theft and abuse of office, and gathering
information from the grassroots. Many dictatorships use institutions that
engage ordinary citizens to help solve these problems.
In dictatorships with support parties, central leaders generally assign respon-
sibility for gathering information about local conditions and opinions, as well
as implementing regime policies, to party members who serve as local officials,
civil servants, or managers and employees in state-controlled firms. Leaders
expect party cadres to explain policies, build support for them, and prevent
ordinary people from ignoring or sabotaging central directives, while also
distributing positive inducements for cooperation. They also expect officials

26
Prior to 1990, this ﬁgure was 16 percent. In the subsequent two decades, this ﬁgure jumps to 38
percent.

and cadres to send information about local problems, how policies work on the
ground, and signs of opposition to central authorities. Often, however, regime
elites lack the capacity to monitor the behavior of party cadres and officials. As
a result, they may contribute less to implementation and information gathering
than regime officials had hoped. Instead, they may steal benefits earmarked for
citizens, demand bribes, seize land and other assets from local residents, sabo-
tage or distort policy implementation, falsify information about local condi-
tions, and in other ways abuse their positions of power.
Local and legislative elections partially compensate for regime leaders’
limited ability to monitor their local agents, which may be the reason nearly
all dictatorships supported by parties hold elections. Low votes for regime
candidates or low turnout can serve as “fire alarms” to alert leaders to espe-
cially incompetent, abusive, or corrupt local officials. The need to achieve
winning votes and high turnout motivates legislators and local officials to
distribute most benefits provided by central leaders to constituents rather than
diverting them, to try to acquire resources from the center for their areas, and to
convey information about local problems to top officials in order to secure aid.
Even where no partisan competition is allowed, there is competition among
potential candidates for ruling-party nominations. Nominations depend on
earlier success in mobilizing citizens, which gives legislators and elected local
officials reasons to care about the goodwill of constituents.
Legislative and local elections thus align the incentives of mid- and lower-
level officials with the needs of top leaders. Regime leaders need reasonable
levels of competence and honesty in local officials to prevent mass alienation
and potential opposition mobilization. They also need information about prob-
lems and local disasters so that they can respond effectively. Elections give local
politicians, job-holders, and the regime’s lower-level officials reasons to develop
clientele networks that reach to the grassroots to distribute resources in
exchange for votes, and to bring information about local needs to the attention
of central policy makers. By creating broader vested interests in regime survival,
a more effective system of distribution to the grassroots, and additional infor-
mation for regime insiders, elections aid regime survival.
Elections in which voters “choose” the dictator serve a different purpose,
however. The dictator need not monitor himself, but nevertheless nearly as
many dictatorships schedule regular executive elections as regular legislative
elections. From the dictator’s point of view, executive elections help deter elite
defections, which are the most serious threat to both the dictator’s and the
regime’s survival (Svolik 2012; Roessler 2016). Executive elections demonstrate
the dictator’s control over the resources needed to hang on to power. They
show that he can mobilize very large numbers of people and demonstrate his
command of a nation-wide network of political activists. They aim to show
potential defectors that it would not be easy to organize enough support to
replace the dictator. If campaigns are well run, potential defectors cannot tell
how much popular support is genuine, but they can observe the enormous

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:38:09, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.006
152 Ruling Society: Implementation and Information Gathering

deployment of human and material resources. To achieve this mobilization,

dictators distribute even larger quantities of goods than do dictatorial
legislatures.
A number of previous studies have shown that dictatorships led by parties
and those that have legislatures last longer than dictatorships that lack these
institutions. In the current chapter we show evidence consistent with the claim
that elections are the institutional mechanism that links parties and legislatures
with regime durability. We show, first, that elections are associated with
increased government spending, as would be expected if dictatorial govern-
ments distribute various kinds of benefits and local public goods during
campaigns.
Increased distribution leading up to elections would be expected no matter
what the reason they hold elections. If the only purposes of elections were to
deter elite defections or to provide information about where opposition exists,
we would expect to see much more public spending during semi-competitive
elections than during elections without choice. Choice-free elections are a
weaker signal of regime invincibility and convey little information about
opposition. In order to test whether elections also help align the interests of
central and local officials, our second empirical test compares the effect of
elections with and without choice on public spending. We find that the increase
in spending is about the same for both kinds of elections. We interpret this
finding as support for the claim that all elections, even uncompetitive ones,
incentivize deputies and local officials to deliver benefits to the grassroots
because they entail competition for party nominations even if there is no
competition in general elections.
As a further test of the argument that elections incentivize politicians to
reach the grassroots with benefits, we look at whether dictatorships that hold
elections spend more on health than dictatorships that do not, and at whether
health outcomes actually improve. We show that dictatorships that hold semi-
competitive elections spend more on health care than do dictatorships that hold
no elections, and that they have lower rates of child mortality. Even uncompeti-
tive elections are associated with higher health spending and lower rates of
child mortality than exist in dictatorships that hold no elections. We interpret
these findings as further support for the argument that competition for ruling-
party nominations gives legislators and local officials reasons to attend to their
constituencies even when there is no partisan competition. The lower rates of
child mortality suggest that local officials in systems with any kind of elections
are more likely to play the role regime leaders assign to them in the distribution
of benefits than local officials in systems without elections. In other words,
elections partially substitute for direct monitoring of local officials by central
leaders. Including controls for the existence of a legislature and ruling party
does not change these results, and legislatures make no independent contribu-
tion to these outcomes.

Elections for legislatures and local offices thus help protect dictatorships by
providing a means of periodic monitoring to detect predatory behavior, theft,
and incompetence in lower-level officials. At the same time, they give officials
and deputies strong reasons to report information about local problems and
discontent to central leaders, to lobby for benefits for their regions, and to
distribute some of the benefits they acquire to local people. In these ways, mass
institutions help the dictatorship located in the capital to extend its policies and
governance to all parts of the country.

Double-Edged Swords
Specialized Institutions for Monitoring and Coercion

Dictatorial control rests on the threat of force. To make that threat credible to
potential opponents, dictatorships need visible coercive institutions, that is,
armies, internal security police, and other armed forces even if they face no
threats from beyond their borders. For those dictatorships that relied on these
forces to establish their rule in the first place, the credibility of the threat of
further violence exists from the beginning. Even dictatorships established
peacefully, however, deploy the threat of force and use some violence against
opponents. The worst human rights abuses often happen during the first years
of dictatorship, when the ruling group is unsure of its grip over the country
and riven by internal power struggles. Violence may also be worse because
new rulers lack the information needed for carefully targeted coercion
(Greitens 2016). During these years, they often establish new security agencies
and paramilitary forces and increase military budgets.1 Dictatorships use these
forces against people they suspect of opposition, but investment in them also
increases the credibility of threats of violence and thus can deter further
opposition.
Once the dictatorship has established a ruthless reputation and conflict
within the ruling group has settled down, it usually relies less on overt coercion.
Most repression is preventive, in the sense that it aims to block the dissemin-
ation of negative information about conditions in the country and members of
the dictatorial elite, to discover potential plots and movements before they
become organized enough to challenge the dictatorship, and to undermine
and disperse groups that share critical information or opposition points of view
(Dragu and Przeworski 2017). In other words, though dictatorships that have

1
See Greitens (2016) for descriptions of the reorganization and creation of new security forces in
Taiwan, the Philippines, and South Korea soon after the establishment of new dictatorships.

154

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
Specialized Institutions for Monitoring and Coercion 155

lasted a long time usually rely less on overt violence, they continue to invest in
coercive institutions to deter and undermine opposition.
The decline in overt violence that usually accompanies the routinization of
dictatorial rule might be caused by its high cost. Coercion uses up resources that
could be spent on other things, and overreliance on it drives both elite and mass
opposition underground, worsening many of the information problems that
beset dictatorships.2
Increasing the size and political importance of the coercive institutions can
also be dangerous to dictators and their closest supporters. Security services and
armed forces are serious weapons that are difficult to keep under control
(Svolik 2012; Dragu and Przeworski 2017). In this chapter we examine what
determines who aims these weapons at whom.
We first describe the various security agencies, paramilitary forces, and army
that typically guard a dictator and regime. We provide information about what
the different forces guard against, how they interact with each other, and the
difficulties of controlling them in an environment that lacks third-party enforce-
ment of legal rules.
Internal security agencies are specialists in preventive repression (Dragu and
Przeworski 2017). They spy on and intimidate potential dissidents, and they
identify suspected opponents of the dictator inside the inner circle and the
military. The main purpose of these agencies is to gather information that can
be used to devise strategies for preventing overt opposition. Paramilitary forces
have a narrower role. They defend the dictator and dictatorship from military
coups and other armed assaults. The army defends the dictatorship from
invasions, armed rebellions, and, occasionally, popular upheavals. As the tasks
of the different coercive institutions make apparent, dictators use them not only
to protect against different kinds of threats but also to protect against each
other because the dictator cannot count on controlling them. The more power-
ful and effective the force, the more dangerous it could be to the dictator if it
turned against him.
Dictators try to increase their own chances of survival by establishing
multiple armed forces to spy on, compete with, or counterbalance each other
(Quinlivan 1999; Haber 2006; Greitens 2016, 13–14, 79–80). They also try to
take personal control of security agencies and promotions in the army, and to
recruit paramilitary forces from especially loyal citizens. Other members of the
dictatorial elite, however, resist these efforts as part of their general attempt to
retain the capacity to threaten the dictator with ouster, and thus to limit his
discretion.

2
Since Wintrobe (1998), theories of autocracy have claimed that coercion is more costly than
cooptation. Though we know of no comparative studies of relative costs, we think this claim is
likely to be correct.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
156 Ruling Society: Implementation and Information Gathering

internal security agencies

Although dictatorships may find periodic indirect monitoring (as described in
Chapter 6) sufficient for controlling local officials, they invest much more in
specialized institutions to gather information continuously about the loyalty
and job performance of high-level officials, military officers, and other members
of the elite, including those in the inner circle. These internal security agencies
also spy on the opposition, but since the most serious threats to regime and
dictator usually arise from individuals within the dictatorial elite, much of the
surveillance focuses on them. Internal security police not only provide their
bosses with information about potential challengers, but also use manipulation
of information, censorship, intimidation, beating, torture, imprisonment, death
threats, and murder to deter challenges. These tactics increase the costs of
opposing the regime or dictator. Pervasive spying hinders opposition collective
action by heightening the risk and potential cost of expressing critical opinions,
thus making it harder for opponents to identify each other.
Most of the time, the internal security forces that handle day-to-day moni-
toring of political activity are civilian agencies, not military. They spy directly
on regime officials, party leaders and cadres, civil servants, military officers,
managers of state-owned firms, professors, teachers, journalists, and union
leaders (e.g., Callaghy 1984, 292; Soper 1989; Micgiel 1994, 95; Schirmer
1998, 175; Tismaneanu 2003, 146; Dodd 2005, 79–83; Barany 2012, 214).
In other words, they spy on anyone who might have the capacity to influence
others or sabotage the implementation of regime policy, as well as on oppos-
ition leaders. They may also use large numbers of internet monitors and paid,
blackmailed, or voluntary informers to report on ordinary citizens they could
not otherwise observe (e.g., Lewis 1980, 150; Peterson 2002).
The East German Stasi, because they kept such extensive records, provides a
window on both the costliness of coercion and how dictatorships actually
deploy it. At its peak, the Stasi employed 100,000 people (Peterson 2002,
26–29). If informers are included, there was one spy for every sixty-six citizens
(Koehler 1999, 9). Regime leaders put highest priority on monitoring elites. The
Stasi placed informers in ministries, the Planning Commission, the army, and
the Stasi itself (called the Unknown Colleagues). Agents also monitored com-
pliance with the economic plan and performance in state-owned firms and
collective farms, reporting incompetence, drunkenness, adultery, and inaccur-
ate production reports as well as dissent (which was much less common than
human frailty). In addition, the ruling party had its own separate security
network of 44,000 functionaries in economic enterprises to monitor
performance.
In contrast to many dictatorships, East German leaders put massive
resources into monitoring ordinary citizens as well, because of the special
problems caused by the existence of the very successful West German economy
next door. The Stasi had internment camps in twenty-four locations. Its

domestic counterintelligence sector included 4,400 agents just for inspecting

mail. Every post office had Stasi officers who opened all letters and packages
sent to or from noncommunist countries. Six thousand agents tapped tele-
phones. Eighteen hundred were assigned to combat underground political
movements, each of whom supervised at least thirty informers. And of course
thousands of agents guarded the borders (Peterson 2002, 24–25). East
Germany is an extreme case, both because of the regime’s great vulnerability
and because they had the resources and human capital to man such an extensive
operation, but building effective security services requires significant resources
everywhere.3
Beyond resources, it also takes time and training to build reliable internal
security services. In Poland, where the communists knew that extensive coer-
cion would be required to maintain the regime, the Soviet Union began training
Polish recruits in schools supervised by the Soviet People’s Commissariat for
Internal Affairs (NKVD) in 1940, long before World War II ended. Once the
Polish communist party had established itself in Lublin in 1943, it began
sending hundreds at a time to NKVD officers’ schools, even before the provi-
sional government was formed (Micgiel 1994, 94–95; Iazhborovskaia 1997).
The willingness of both the Soviets and the Poles to expend the resources
required for this kind of training indicates the importance communist leaders
gave to it. Resources desperately needed for the war were diverted to pay for
schools large enough to accommodate thousands of East European students,
while manpower and highly trained officers were diverted from the battlefield
as war raged.
Of course, post-war Poland and East Germany are not typical, but they
provide a sense of the cost and difficulty of creating an effective internal security
service. Most dictatorships do not begin with military occupation by a more
powerful dictatorship famous for its hyperdeveloped internal security agencies.
In the aftermath of most seizures of power, dictatorships lack the trained and
loyal practitioners an effective internal security service would require. Sheena
Greitens’s data on the ratio of internal security personnel to population in nine
countries shows average ratios of 1:124 for North Korea (1:40 if informants are
included), the most intensively policed country for which we have information,
1:5,090 for Saddam Hussein’s Iraq, and 1:10,000 for Chad between 1982 and
1990 (2016, 9).
It usually takes a few years for new dictatorships to reorganize or build a
loyal political police force. Even military-led dictatorships can lack effective
internal security forces initially, despite their expertise in the use of coercion. In
dictatorships launched by military officers, on day one each service typically
controls a separate police apparatus located within its regular chain of

3
East Germany was extreme but not unique. In Ceausescu’s Romania, besides the other tasks of
security agents, they collected handwriting samples from all residents and registered all type-
writers and copy machines (Soper 1989).

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
158 Ruling Society: Implementation and Information Gathering

command and oriented toward preventing infringements of military discipline.

These agencies can be reoriented toward policing subversion among civilians,
but concerns about turf and disagreements about how to handle opponents can
hinder their unification and transformation into obedient instruments of the
dictatorial elite. A further problem with using troops for subversion control is
that many soldiers resist assignments to repress unarmed citizens (Janowitz
1977; Nordlinger 1977; Barany 2011). Soldiers often share the grievances that
motivate opposition demonstrations and may balk at orders to fire partly for
that reason.
Because of these problems, even military-led dictatorships typically call on
the army for protection from domestic opponents only when police and para-
military forces have failed to control opposition demonstrations or when armed
insurgents challenge them. Using the military to control demonstrations can
backfire. The Egyptian dictatorship begun by the Free Officers’ coup in 1952,
for example, used security police and paramilitary forces against demonstrators
during the Arab Spring in 2011. When that failed to disperse them, the army
was called out, but did not attack demonstrators. Instead, the Supreme Council
of the Armed Forces, which had been a key pillar of regime support, forced the
dictator to resign and took control of the country. “[E]ven had the generals
been willing to shoot demonstrators, many officers and enlisted men would
probably have refused to obey such an order” (Barany 2011, 32–33). Paramili-
tary forces are more likely than regular soldiers to shoot unarmed protesters
when called upon to do so (Barany 2011; Nepstad 2015).
Other dictatorships, whether led by officers or civilians, have had similar
experiences. As a number of observers have noted, autocratic regimes fall to
unarmed popular opposition when generals refuse to use troops to repress
demonstrators. President Ben Ali of Tunisia fled into exile when the army
chief-of-staff refused to deploy troops to disperse protestors (Barany 2011,
30–31). In Georgia, then-President Eduard Shevardnadze withstood weeks of
massive demonstrations, but resigned at the end of the day when “one by one
the heads of police departments and army units declared” their unwillingness to
continue defending him (Wheatley 2005, 185). In the last months of Duvalierist
Haiti, Tonton Macoutes and the police beat, arrested, and murdered ram-
paging protesters, but top Duvalierist army officers refused orders to shoot
them (Abbott 1988, 306). Even when officers’ interests link them firmly to the
regime, as they did in the Haitian and Egyptian examples, they resist using
troops to repress demonstrators if they fear soldiers would refuse to fire or
would hand their guns to protestors (Pion-Berlin and Trinkunas 2010; Dragu
and Lupu 2017).4

4
There are obvious exceptions to this statement, especially during civil wars and violent ethnic
mobilization, when soldiers drawn mostly from some ethnic or regional groups ﬁght insurgents
from other groups. See Roessler (2016) for examples.

The discomfort many officers and soldiers feel about using violence against
unarmed fellow citizens partly explains the military’s unwillingness to put
down the protests that overthrew these and several other dictatorships (Haddad
1973, 195; Woodward 1990, 164; Kuntz and Thompson 2009; Barany 2011).
As a result of these attitudes, military-led regimes often create civilian security
services to deal with internal opposition, just as most party-led dictatorships do
(Plate and Darvi 1981).5 Some also sponsor informal armed civilian forces
(e.g., death squads) to further deter “subversion.”
Party-based seizure groups that achieve power via authoritarianization or
popular uprising may also initially lack loyal security services. On assuming
dictatorial powers, they must decide what to do with the preexisting military
and police forces, which may reflect interests and ethnic groups opposed to
those that dominate the ruling party. Unreliable police forces reduce the new
elite’s ability to induce cooperation from the population, and the military tends
to incubate coup attempts rather than keeping the dictatorship safe. Many
party-led regimes create new party-controlled internal security services after
authoritarianization.
Parties that have led insurgencies, in contrast to other party-based dictator-
ships, usually have cadres specialized in internal security before seizing power.
Since successful insurgency requires the maintenance of organized networks for
recruiting manpower, extracting resources, training soldiers, and disciplining
dissidents and “shirkers,” insurgent groups often have the monitoring and
coercive capacity needed to spy on opponents and those who occupy influential
positions while ensuring citizens’ cooperation with the seizure group’s project.
Parties that have fought lengthy civil wars are likely to have the organization
and personnel to take over and/or replace preexisting state coercive institutions
rapidly and thoroughly. As a result, pre-seizure security organs created by
insurgents can be carried over into the new regime.
To sum up this section, most dictatorships not brought to power by insur-
gency initially lack their own loyal and effective internal security agencies. Since
police forces carried over from the ousted government tend to have inadequate
resources, loyalty to the new rulers, and commitment to the new task, dictator-
ships often establish new security forces staffed by loyalists. It takes resources
and time to build a dependable internal security force because of the need for
trained and committed personnel to staff it.
The creation or strengthening of internal security services can increase the
information available to the inner circle, not only about potential opponents,
but also about each other. This is what makes them potentially dangerous to
members of the inner circle as well as their enemies.

5
Janowitz (1977) ﬁnds that men employed in various internal security and national police services
outnumber those in the army in a sample of military-led dictatorships.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
160 Ruling Society: Implementation and Information Gathering

The Dangers of Personalized Control of Internal Security

Societal opposition may first motivate the establishment of new internal secur-
ity agencies. Members of the seizure group usually support the creation of a
strong security apparatus to protect the regime from opposition threats. These
institutions, however, bring real risks for the dictatorial inner circle if the
dictator comes to control them personally. A dictator who commands the
security apparatus and has access to the information it collects can eavesdrop
on the private conversations of his closest collaborators, making possible pre-
emptive strikes against anyone suspected of disloyalty or too much ambition.
Meanwhile, the dictator’s plans for humiliating, discrediting, arresting, or
executing his erstwhile colleagues remain secret until he acts.
Reports from the security services help the dictator identify which members
of the inner circle might challenge him, and hence which he might want to
exclude preemptively, worsening the inner circle’s difficulty in enforcing limits
on the dictator. The dictator’s advantage comes not only from his access to the
information collected, but also from his ability to order security officers to
arrest his colleagues. If the dictator can, in effect, choose members of the inner
circle by excluding anyone he wants, then he has become a free agent with
immense resources and murderous powers.
The dictator’s information advantage may also allow him to hide some
transgressions, as Svolik (2012) argues. We do not see the dictator’s ability to
hide overstepping his delegated powers as a major cause of the inner circle’s
inability to hold him accountable, however, since transgressions often involve
appointments, dismissals, arrests, and arbitrary policy choices, all of which are
highly visible. Instead, we see the inner circle’s main difficulty as arising from
the riskiness of trying to punish overreaching dictators. The dictator’s control
of security forces can make talking about ousting the dictator so dangerous that
party executive committees with the formal power to replace dictators never
discuss doing so and plots are never developed. If members of the inner circle
cannot oust the dictator for overstepping agreed-on limits, they cannot enforce
constraints on him even when they observe violations.
If both the dictator and his lieutenants understand that plots are unlikely to
succeed, the dictator will concentrate additional resources and power in his
hands, and his lieutenants will acquiesce. If the likelihood of ouster is very low,
lieutenants are better off as marginal members of the dictator’s coalition than as
ex-members.
With so much at stake, members of the inner circle who understand the
situation they face should try to prevent the dictator from establishing personal
control over the security apparatus in order to safeguard their ability to limit his
discretion. Usually a unified seizure group can in fact prevent the dictator from
taking over internal security in the same way that they can prevent other power
grabs, though specific circumstances can sometimes give dictators unexpected
opportunities.

In regimes led by more professionalized military forces, juntas are usually

able to prevent the personalization of internal security. Among the South
American military regimes in Argentina, Uruguay, Brazil, Ecuador, Peru, and
Chile during the 1960s to 1980s, for example, only Pinochet in Chile usurped
personal control of security forces (Remmer 1991).
Pinochet’s control seems to have been an unforeseen consequence of a
consensual junta decision taken to solve a different problem. In the months
immediately after the 1973 Chilean coup, the junta gave provincial military
commanders responsibility for internal security, including the arrest, trial, and
punishment of “subversives,” in their regions. Each commander handled these
responsibilities according to his own views about the legal norms in force,
which led to wide variation in the treatment of the accused. The Chilean
military, with no authoritarian experience during the four decades before the
coup, had virtually no past involvement in subversion control, and some
officers found the new role repugnant. Pinochet and his allies favored a harsh
and violent anti-subversion policy, but some of the provincial commanders
refused to ignore the democratic legal norms still formally in force despite
pressure from Pinochet and other military hard-liners. These disagreements
caused conflict and disunity among high-ranking officers (Policzer 2009).
To maintain the unity and discipline of the officer corps in the face of deep
disagreements over how to treat opponents, junta leaders decided to centralize
subversion control in an agency outside the regular chain of command of the
armed forces. In other words, junta leaders created a new security agency
outside the military chain of command in order to remove subversion control
from the hands of legalistic officers and reduce disagreements that were under-
cutting unity in the army (Policzer 2009).
This decision inadvertently gave Pinochet the opportunity to appoint the
new agency’s top leadership, and thus to control it (Policzer 2009). It resolved
the problem of conflict within the officer corps but gave Pinochet the capacity
to use the security agency to spy on and coerce officers as well as other citizens,
changing the balance of power within what had started out as a relatively equal
military junta. Thus, it seems that the Chilean military’s lack of recent authori-
tarian experience made it more vulnerable to the personalization of dictatorial
power than other professionalized armed forces in the region, most of which
had more recent experience governing.
Where a dictator gains personal control of the security apparatus, he has
taken a giant step toward the personalization of rule, even in countries with a
united military or disciplined ruling party. In such settings, control of the
security apparatus may not be enough to personalize the dictatorship fully,
but it moves in the direction of power concentration, as in Chile.
In contrast to the South American military dictators who came from rela-
tively professionalized military forces, dictators who were creating a new officer
corps as they seized state power, such as Anastasio Somoza García in
Nicaragua, Mobutu Sese Seko in the former Zaire, and several other early

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
162 Ruling Society: Implementation and Information Gathering

African military dictators, faced few impediments to personal control over

security forces. Top officers in these armies often owed their recruitment and
promotions directly to the dictator, making them unlikely to contest his
appointments to security agencies or constrain his decisions in any other realm.
Dictators seek personal control over internal security in order to protect
themselves from ouster, but the personalization of internal security is associated
with greater regime longevity as well as more stable dictator tenure. Our data
indicate that regimes led by dictators who exercise personal control over
security forces last on average seven years longer than similar regimes in which
they do not.6 We interpret this increase in regime longevity as a result of
reductions in plotting and intra-elite conflict when the dictator controls security
forces, because other members of the inner circle understand that plots are
unlikely to succeed. Intra-elite conflict is one of the main causes of authoritarian
breakdown (see Chapter 8).
To sum up this section, dictatorships need internal security services to spy on
potential opponents and deter overt opposition. Security agencies can also
threaten members of the ruling elite, however. If the dictator secures personal
control over internal security forces, then he can also spy on and monitor other
members of the dictatorial inner circle. Control over internal security provides
the dictator with a major information advantage relative to others in the
dictatorial elite, which reduces the likelihood that plots can be kept secret and
thus diminishes their chance of success. Control over the security service also
means that the dictator can order agents to arrest, interrogate, torture, and
execute his inner-circle colleagues. The possibility of such punishments obvi-
ously changes the distribution of power within the inner circle, rendering
constraints on the dictator’s discretion unlikely if not impossible.
If the dictator gains the advantage conveyed by direct control over the
security apparatus, he will want to hang on to it, and other members of the
inner circle would face great risks if they tried to remove it from his control.
Between 1946 and 2010, only 3 percent of dictators who gained control of the
security apparatus lost it later.

the army: bulwark of the regime or

incubator of plots?
The army serves as the dictatorship’s defense against foreign invasion, insur-
gency, and popular uprisings that the police and security troops have failed to
suppress. Dictatorships need to maintain armies to defend them against armed
challenges and make the threat of violent repression credible, but ofﬁcers’

6
Regimes without this feature last fourteen years, on average, while those with it endure for
twenty-one years. This pattern holds in standard duration models of regime failure.

training, access to weapons, and command of armed men mean that they can
more easily overthrow the dictator than can others.
Some analysts see the uniformed military as the key component of the
opposition-control apparatus in dictatorships (Svolik 2012; Debs 2016). We
find that view problematic. Although officers have played this role in some
Latin American military regimes, dictators have not usually used military
officers to spy on, intimidate, intern, torture, and murder suspected opponents.
Dictators sometimes call on the military to repress particularly threatening
demonstrations, but not for routine subversion or crowd control.
Instead, the officer corps often nurtures challenges and plots, as the figures in
previous chapters on the frequency of military coups show. There are many
reasons for plots. Officers commissioned before the seizure of power may have
received their promotions because of loyalty to the leaders ousted by the new
regime and may want to reverse the seizure of power. Officers may come from
ethnic groups or political factions not represented in the dictatorial elite and
resent disadvantages for their group. They may be angered by slow promotions,
low pay, or poor living conditions. They may simply be ambitious to rule.
Consequently, dictators have more often felt the need to protect themselves
from potentially disloyal officers than been certain enough of officers’ devotion
and reliability to use them in subversion control. Philippine dictator Ferdinand
Marcos, for example, told reporters that what he feared most was overthrow by
the military. He imagined being assassinated during a coup, with US complicity,
like South Vietnamese dictator Ngo Dinh Diem (Greitens 2016, 126).
This is the reason internal security police spy on officers, and it is the reason
that some dictatorships spend substantial resources to indoctrinate soldiers and
impose commissars with the power to countermand officers’ orders in military
units (Fainsod 1967, 468–81; Bullard 1985, 65–83; Greitens 2016, 89–90,
97–99). A few dictators have even stored ammunition in areas inaccessible to
the army. The Duvaliers of Haiti kept the entire national arsenal, even heavy
weaponry, in the basement of the National Palace. They put up with explosions
that required rebuilding three times rather than allow the military access to
weapons (Crassweller 1971, 317–18).
Since the means of overthrowing the dictator are always at hand for officers
(unless weapons are locked up), their critical opinions or divergent ethnic
loyalties can be dangerous to dictators even if they were allies at the time of
the seizure of power (Roessler 2016). Crafting strategies to maintain military
acquiescence is a challenge all dictatorships face. We described some of the
strategies used in military-led regimes in Chapter 5. Here we consider some less
nuanced strategies used by civilian as well as military dictatorships.

Counterbalancing
Many dictators (63 percent) try to meet the challenge of a possibly untrust-
worthy military by creating presidential guards or other kinds of paramilitary

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
164 Ruling Society: Implementation and Information Gathering

forces outside the regular military chain of command. The high frequency
with which dictators establish new paramilitary forces indicates how little
most of them trust regular officers to defend their rule. Dictatorships have
created three main kinds of paramilitary force: armed civilian forces to help
the regular military fight insurgents (for example, Orden in El Salvador);
party militias, which are typically armed youth adjuncts of the ruling party;
and paramilitary forces recruited from groups thought to be especially loyal
to the dictator himself and expected to defend him from coups led by regular
army officers. We refer to the third kind as loyalist paramilitary forces to
distinguish them from the others. Dictators expect loyalist paramilitary forces
to be more reliable than the regular military because they recruit them from
partisans, co-ethnics or both, while highly ranked regular officers would
often have been recruited and promoted before the dictatorship began and
thus represent a broad range of ethnicities, regions, religions, and partisan
leanings.
Paramilitary loyalty is reinforced if the new force is closely identified with the
dictator because of shared ethnicity or partisanship and would thus likely be
disbanded if the dictator fell. If that happened, its officers would lose their
special privileges and face possible prison or exile. In other words, loyalist
paramilitary officers stand or fall with the dictator (Snyder 1998; Escribà-Folch
2013). In contrast, senior officers in the regular military are more likely to turn
their backs on the dictator during periods of crisis because they typically have
fewer ties to the ruler and a more developed “corporate” identity linked to
defending the state, rather than the particular leader. Senior officers in the
regular military often survive successive dictators in the same regime with their
ranks intact, and they may even survive regime collapse, especially if they cause
it. The fates of regular military officers are not routinely linked to the
dictator’s fate.
Dictators establish loyalist paramilitary forces to change the assessments
of regular army officers about the likely risk of failure for coup attempts. To
enhance the deterrent power of paramilitary forces, the dictator may buy
them better arms and training than the regular army. He may schedule
ostentatious parades of uniformed paramilitary forces and their weapons.
Loyalist paramilitary units are often stationed in or near the dictator’s
residence as a specialized and highly visible defender of the dictator and
dictatorship.
We see the coup deterrence value of loyalist paramilitary forces as arising
from their staffing by officers and men from groups identified with the dictator
who expect to share his fate if he is overthrown. An alternative explanation for
why the establishment of paramilitary forces reduces the likelihood of coups is
that the existence of multiple independent armed forces in the dictatorship
raises the collective-action costs of committing to and executing coups
(Quinlivan 1999; Böhmelt and Pilster 2015). In this view, coups do not succeed
unless officers from both the regular military and the paramilitary go along

with them.7 This argument simply applies the well-known logic of collective
action: the more independent actors are needed to accomplish a collective goal,
the greater the difficulty of doing so. As obstacles to cooperation increase,
coups should decrease (Singh 2014).
Dictatorships sometimes also create party militias to help defend themselves.
On average, party militias cost less and receive much less training than presi-
dential guards and other forces stationed near the dictator’s residence. In a few
well-known instances, party militias were well funded and grew into very well-
armed and important elite political players, as during the Banda dictatorship in
Malawi and Qaddafi’s rule in Libya. In Malawi, the militia reported directly
“to the president on the mood in the countryside and on all significant new
arrivals in every village in the country” (Decalo 1998, 86). In these cases,
militias effectively armor the dictator. Most of the time, however, militias have
a less central role. They engage large numbers of young men, provide them with
light weapons and rudimentary training, and assign them tasks such as keeping
order and rooting out subversion among students or ordinary citizens.
Members of militias are expected to inform on their neighbors, patrol problem
areas, help to mobilize others for regime projects, and train as auxiliary
national defense forces in case of invasion or rebellion.
Regime insiders create party militias to coopt the part of the population
(young men) most likely to lead popular opposition by giving them a stake in
regime persistence. Militias are supposed to help with internal security while
delivering benefits to those who join them – at low budgetary cost, since they
usually receive no salaries. The benefits to militia members include the power to
order others around, recognition from political leaders, camaraderie, and infor-
mal opportunities to steal, extort, and demand sex from those they police.
Militia members have a lot of discretion over who they stop, search, and
demand fines or payments from, as well as how much force to use, which
naturally leads to abuses. They often become undisciplined and venal,
making use of the powers granted them to pursue their own ends. In
Congo/Zaire, for example, Disciplinary Brigades set up unauthorized barri-
cades to “stop and harass people indiscriminately, demand ‘tips,’ and illegally
detain people” (Callaghy 1984, 293). Like death squads, these forces may
commit many human rights abuses without making the dictatorship safer
because regime leaders cannot control their behavior. In practice, popular
militias have often caused opposition to dictatorships because of arbitrary
acts of violence, theft, and intimidation, and they have eventually been
disbanded or integrated into the regular military as a means of imposing
discipline in a number of cases.

7
Because counterbalancing creates an additional armed actor within the regime, it opens up the
possibility that these actors will take different sides during periods of unrest even if one does not
automatically side with the dictator (Morency-Laﬂamme 2017).

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
166 Ruling Society: Implementation and Information Gathering

Controlling Military Leadership

Many dictators also try to control their military forces through direct interfer-
ence with promotions. The source of officers’ autonomy from the dictator, if it
exists, is the professional norms that determine promotions, recruitment, and
retirement. Meritocratic recruitment and promotion based on competence and
seniority mean that career success in the military does not depend on political
loyalty. As long as recruitment to the officer corps is open to talent and
promotions depend on performance and years of service, ethnic and partisan
groups different from the dictator’s will be represented in the higher ranks of
the officer corps as will a range of opinion. If, however, the dictator can
override professional norms to recruit, promote, and retire individual officers
on the basis of loyalty or ethnicity, the officer corps cannot maintain its
autonomy because the dictator can reserve high ranks for loyalists.
Nearly all dictatorships (and many democracies) use loyalty as one criterion
for highest-level promotions, but in some, the existing officer corps is all but
destroyed by purges, forced retirements, and promotions aimed at ending
military autonomy. At the extreme, insurgents may replace the whole military
they have fought against with their own officers and troops when they take
power. The decimation of the officer corps was almost as severe after the
communist takeovers in Eastern Europe. Many officers were imprisoned or
murdered in order to keep communist party rule safe from potential military
intervention.
Some dictators see the promotion of co-ethnics as the best strategy for
achieving interest congruence between officers and themselves. Reserving the
most sensitive military posts for clan or even family members, as in Syria under
the Assad family, a number of monarchies, and Iraq under Saddam Hussein,
takes this strategy to an extreme. Using ethnicity or partisanship as the main
criterion for promotion increases the value to co-partisan/co-ethnic officers of
the dictatorship continuing to rule. As with limiting the dictatorial inner circle
to co-ethnics, discussed above, this strategy aims to reduce factionalism, and
thus decrease the likelihood of rogue coups, while increasing interest congru-
ence between soldiers and political leaders (Roessler 2016). The politicization
of promotions boosts the power of loyal officers while marginalizing or retiring
those thought less reliable. For officers disadvantaged by partisan promotions,
however, this strategy makes opposition more attractive.8
The extreme examples of dictatorial interference in the military are possible
when the armed forces backing the dictator (e.g., the Soviet occupation army in
Eastern Europe or victorious insurgent forces) can defeat the regular army he
wants to subjugate. A dictator who lacks his own military force, however,

8
See Roessler (2011, 2016) for an analysis of how excluding ethnic rivals from the inner circle to
lower coup risk can increase the chance of civil war.

would face the immediate threat of a coup if he tried wholesale interference

with promotions and retirements. Promotions are central to the career interests,
day-to-day well-being, and future prospects of officers. One of the attractions of
careers in the military is the clear, predictable career path and the opportunity
for upward mobility it provides to young men who lack political or social
connections. Political or ethnic favoritism in promotions threatens career
expectations and is immediately visible to other officers. It is a relatively easy
grievance to organize around because it affects not only personal welfare but
also military professionalism and esprit de corps, which even officers who are
not personally disadvantaged may feel strongly about.
Officers passed over for promotion, forced into retirement, or dismissed
have led quite a few coups, and soldiers loyal to jailed former commanders
sometimes free them as the first stage of a coup. Dismissed and jailed former
high-ranking officers have led a number of insurgencies as well (Roessler 2016).
So, dictators have to make careful strategic calculations before trying to take
control of the officer corps through recruitment, promotions, and dismissals.
The establishment of a countervailing paramilitary force increases the feasi-
bility of interfering with promotions because it can reduce the odds of a
successful coup, though the creation of paramilitary forces itself carries some
risk. The diversion of scarce resources to pay and arm them reduces the benefits
that can be allocated to the regular military, and descriptions of why coups
occurred sometimes also mention officers’ complaints about the better weapons
and nice uniforms of paramilitary troops (Nordlinger 1977).
The implications of creating a new presidential guard or other paramilitary
force, however, may be less initially obvious to regular officers than interference
with promotions. These new forces are usually announced with patriotic fan-
fare, and they may at first appear to add new units to the regular forces. Thus,
they can seem to strengthen the military rather than challenging its monopoly
of force. Because the consequences of creating new paramilitary forces are less
obvious, dictators may establish them before they try to interfere with promo-
tions. Once a well-funded paramilitary unit guards the presidential palace,
prospects for successful coups decline, and officers may plot fewer attempts
because they fear violent defeat by the new armed force. The establishment of
paramilitary forces may thus embolden dictators to interfere with promotions
in the regular army.

the relationship between counterbalancing

and interference
One implication of this discussion is that dictators who counterbalance the
regular army by creating loyalist paramilitaries will then be more likely to
interfere in the military’s promotion decisions. We evaluate whether the suc-
cessful pursuit of one strategy for reducing the potential threat from the

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
168 Ruling Society: Implementation and Information Gathering

military – counterbalancing – paves the way for the successful pursuit of

another – interfering with promotions. As with other power grabs by dictators,
we expect the success of one to increase the likelihood of the next one in the
kind of spiraling consolidation of personal dictatorial control described in
Chapter 4.
Our argument about the linked fates of dictator and paramilitary forces
recruited from his group also implies that loyalist paramilitary units should
deter coups more effectively than paramilitary forces created for other pur-
poses, such as to keep youth loyal or to battle insurgency alongside the
military. If loyalist paramilitaries more effectively deter coups, they would
have a larger effect on dictators’ propensity to manipulate military promo-
tions than would other kinds of paramilitary. In contrast, claims that para-
military forces reduce the likelihood of coups by increasing the number of
independent armed forces that must agree to oust the dictator imply that any
kind of paramilitary force would have the same effect, both on the incidence
of coups and on interference with promotions. Our data on different types of
paramilitary forces allow us to test these two logics: (1) the linked fate story,
which implies that loyalist paramilitaries should increase the likelihood
of interference with promotions but that other types of paramilitary
force should not and (2) the collective action logic, which implies that all
kinds of paramilitary should increase the likelihood of interfering with
promotions.
To classify paramilitary forces, coders first noted the creation or existence of
armed forces that were both outside the regular military chain of command and
not part of the regular police or internal security service. Once such a force had
been identified, it was coded according to the following rules:
Coded 2 if paramilitary forces are created to fight on the government’s side
during civil wars or insurgencies (e.g., anti-insurgent forces in El Salvador or
Thailand) or to carry out other tasks the military or security service wants
accomplished.
Coded 1 if a party militia or other irregular armed force organized by the
regime support party has been created.
Coded 0 if the regime leader creates a paramilitary force, a president’s guard,
or new security forces apparently loyal to himself.
Use this code if a military or paramilitary force has been recruited primar-
ily from the regime leader’s tribe, home region, or clan; if they report to
him; or if they are newly garrisoned in the presidential palace.
The same code was then used for every subsequent country-year until the
paramilitary force was disbanded or integrated into the regular military or until
the regime ended.
Some dictators created more than one loyalist paramilitary force. In these
cases, the country-year was coded as “0” after the first was created and
continued to be “0” in subsequent years until all loyalist paramilitaries were

disbanded or the dictatorship lost power. For country-years in which more than
one kind of paramilitary force existed, the lowest applicable code was used.
To test our arguments, we examine two related forms of meddling in the
military’s leadership: purges (defined as arrests, executions, or murders) of
senior officers and dictatorial control of promotions. Our argument implies
that these kinds of interference should be more likely to occur when the dictator
has already established a loyalist paramilitary force.
We coded a purge when one or more military or security service officers were
jailed or executed without reasonably fair trials or was murdered. Coders were
instructed to rely on country specialists and/or journalists for judgments about
whether arrests for plotting or treason trials were government responses to real
events. They were told not to code a purge if the evidence indicated that a
person tried for treason really did try to overthrow the government by violent
means and was given a reasonably fair trial.
Dictatorial promotion strategies were coded as follows:
Coded 2 if country specialists do not report that the regime leader promotes
officers loyal to himself or from his ethnic, tribal, regional, or religious group
or that he forces officers not from his group into retirement for political
reasons
Coded 1 if country specialists report promotions of top officers loyal to the
regime leader or from his group, but not widespread use of loyalty as a
criterion for promotion or retirement.
Coded 0 if country specialists report promotions of large numbers of officers
loyal to the regime leader or from his group, or large numbers of forced
retirements.
The same historical event might, in some cases, serve as the basis for coding
both purges and promotion practices. For example, the communists in Hun-
gary both arrested (purged) many senior officers when they took over and
began interfering with promotions at the same time. For this reason, we test
separate models for each of these forms of interference with military leadership.
Importantly, however, the creation of paramilitary forces is not only conceptu-
ally distinct from interference in the regular military, but also coded in our data
as a distinct phenomenon in the historical record.
The data on military purges and promotion practices are constructed such
that once we find evidence that the dictator pursues either of these strategies for
the first time, the variable is coded the same way for all subsequent years while
he remains in power unless a change in strategy is reported in the country
specialist literature.9 This coding procedure assumes that a dictator who can

9
For example, immediately after the Communist victory in China, Mao controlled some top
promotions, but the revolutionary army included many officers with views very different from
Mao’s, and military leadership was very decentralized. The first years after the seizure of power
are coded as “1.” In 1954, however, under Soviet influence, Peng Dehuai was appointed defense

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
170 Ruling Society: Implementation and Information Gathering

interfere with promotions in one year could probably do so again but may not
be observed doing so because he has already put the officers he wants in place.
The dictator’s control over the military tends to persist over time because the
officers he promoted retain their posts or are further promoted, and they in turn
promote their allies.
To test whether loyalist paramilitaries increase the likelihood of interfering
with military leadership, we restrict the analysis to years in which the dictator-
ship does not interfere with military promotions, and test whether dictators
with loyalist paramilitaries in those years are more likely to interfere with
promotions in the regular military the following year. This design means we
are testing whether dictators with loyalist paramilitaries initiate interference in
the military more quickly than dictators who lack them, given that the dictator
has not yet interfered in these ways.
In the full data set, loyalist paramilitaries are more common (35 percent)
than party militias (18 percent) and irregular forces created to fight alongside
the regular military (9 percent). However, in the samples we employ that
restrict the analysis to years prior to military interference, the incidence of
different types of paramilitaries is more evenly distributed. Prior to observing
promotion interference (921 observations), loyalist paramilitaries (16 percent)
are less common than party militias (19 percent) but still more common than
anti-rebel paramilitaries (10 percent). Prior to observing the first military purge
(2,555 observations), loyalist paramilitaries (27 percent) are the most common,
with party militias (20 percent) and anti-rebel paramilitaries (9 percent) less
prevalent.10
We begin the analysis by comparing the baseline probabilities in Figure 7.1.11
The plots in the top two panels of the figure show the probabilities of dictatorial
interference with promotions and military purges, both with and without an
existing paramilitary force of any kind. Just under 5 percent (0.046) of dictators
who lack a paramilitary force initiate interference with promotions, while just
under 8 percent (0.077) of those supported by paramilitary forces of one kind
or another do so. Dictators unsupported by a paramilitary force begin a purge

minister and began intensive professionalization of the armed forces, which included limits on
Mao’s interference (Whitson 1973, 98–100). From 1955 until 1959, when Mao succeeded in
purging Peng Dehai from the regime’s inner circle, Chinese promotions are coded as “2” because
Mao had little influence on them.
10
The data are constructed so no observations were coded with more than one type of paramili-
tary. Therefore, colinearity issues do not arise in the analysis that includes all three types of
paramilitaries in the same specification.
11
The analysis examines pooled leader-year observations when the dictator did not interfere the
previous year. There are 164 leaders in 74 countries (921 observations) in the analysis of
promotions; this excludes leaders who interfered in military promotions in their first year in
power. The purge analysis examines 340 leaders in 115 countries (2,555 observations), again
excluding leaders who purge the military in their first year in power. It is more common for
leaders to interfere in promotions than to purge the military during their first year.

Promotion Purge
0.1 0.1
Probability of Inteference with Promotions

0.08 0.08

Probability of Purge
0.075

0.06 0.06

0.04 0.044 0.04

0.031
0.02 0.02 0.025

0 0
No paramilitary Any paramilitary No paramilitary Any paramilitary

Promotion Purge
Probability of Interference with Promotions

0.1 0.1
0.096

0.08 0.08
Probability of Purge

0.06 0.06

0.051
0.04 0.04
0.038

0.02 0.02 0.025

0 0
No Loyalist Loyalist No Loyalist Loyalist
paramilitary paramilitary paramilitary paramilitary

figure 7.1 Paramilitary forces and interference in the army.

of ofﬁcers in 2.5 percent of observed years, while those supported by one

initiate a purge 3.1 percent of the time. These percentages may seem low, but
recall that all country-years in which dictators had previously interfered with
promotions or purged ofﬁcers have been excluded from the data set. So, these
are estimates for the beginning of dictatorial efforts to control military leader-
ship in countries in which the dictator has previously not interfered.
The lower two panels compare dictators who lack loyalist paramilitaries
with those who have them.12 Just over 5 percent of dictators unsupported by
loyalist paramilitary forces begin interference in the military’s promotion prac-
tices, while almost 10 percent of dictators who can count on the support of a

12
The group that lacks loyalist paramilitary forces includes dictatorial country-years in which
party militias and/or anti-rebel irregular forces exist plus country-years that lack any kind of
paramilitary.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
172 Ruling Society: Implementation and Information Gathering

Promotions Purges

Loyalist paramilitary Loyalist paramilitary

Leader duration (log) Leader duration (log)

Military leader Military leader

Leader age Leader age

Party paramilitary Party paramilitary

Anti-rebel paramilitary Anti-rebel paramilitary

Int'l conflict Int'l conflict

Civil conflict Civil conflict

–1 0 1 2 –1 0 1 2
Coefficient estimate Coefficient estimate

figure 7.2 Loyalist paramilitary forces and interference.

loyalist paramilitary force do so. The relative increase from the baseline prob-
ability of 5.3 percent is more than 80 percent. For purges of officers, we find
that 2.5 percent of dictators unsupported by loyalist paramilitary forces initiate
a purge of senior officers, while 3.8 percent who can depend on loyalist
paramilitary support do so. The relative increase for purges is 50 percent.
Next we test a series of logistic regression models that control for the length
of time the dictator has been in power.13 The first specification, shown by
squares in Figure 7.2, includes only the indicator variable for the existence of
a loyalist paramilitary force and the natural log of leader time in power. The
next set of specifications, shown as triangles, adds two variables that pertain to
individual dictators: their age and whether they were military officers before
becoming regime leaders. Next, we add indicators for other types of paramili-
tary forces (party militias and anti-rebel irregular forces). The final specifica-
tion, shown as diamonds, adds indicators for civil and international conflict.
The left panel of Figure 7.2 shows the results for initiation of interference
with promotions in the regular military, while the right panel shows results for
first military purges. In all tests, the coefficient for loyalist paramilitary is
positive and significant, indicating that these specific paramilitary forces are
associated with a higher likelihood of interfering with leadership in the
military.14 Models with a full set of covariates (depicted by diamonds) indicate

13
This design mimics standard hazard models; we control for log duration in the reported analysis
and show in replication files that the results remain when using duration time polynomials
instead.
14
Replication files show that this result is robust to including indicators for how the regime seized
power and structural covariates (GDP per capita, population, oil rents, and growth). Models
with leader-specific random effects yield similarly sized coefficient estimates but larger estimates
of the variance in the purge model.

that loyalist paramilitary forces are associated with a 5.7 percent increase in the
probability of beginning to interfere in military promotions and a 1.7 percent
increase in the likelihood of starting a military purge, which is roughly similar
to the increases depicted in the lower two panels in Figure 7.1.
The models in Figure 7.2 also suggest that dictators are more likely to
interfere in military leadership earlier in their tenures, when they are younger,
and when they came to the dictatorship from careers in the military. Finally,
participation in an international conflict increases interference in military lead-
ership, but the estimates are significant only at the 0.10 percent level.15
Importantly for our argument, the estimates for other kinds of paramilitary
forces are not statistically significant, suggesting that these forces do not
encourage the dictator to grab more power from the regular military. In the
replication files, we test similar models for party militias and anti-insurgent
irregular forces separately and again find no consistent association between
them and initiation of interference in regular military promotions or purges.
These findings provide initial evidence that loyalist paramilitaries help dictators
to consolidate power over the regular military by interfering with their promotion
procedures and purging senior officers of doubtful loyalty. We find no evidence
linking paramilitary forces of other kinds, namely, party militias and anti-insurgent
irregulars, to interference with the regular military. This suggests that counterbalan-
cing is not simply a numbers game in which dictators need only increase the number
of security veto players to increase plotters’ collective action problems in order to
protect themselves. Rather, we suggest that the informal personal ties between the
dictator and the loyalist paramilitary he creates by recruiting from co-ethnics,
co-partisans, or his home region allow him to interfere with impunity in the internal
workings of the regular military.16 When regular army officers know that loyalist
paramilitary forces will defend the dictator to the bitter end because their own
futures are tied to his fate, they are less likely to attempt coups to stop the dictator’s
manipulation of promotions or purges of senior officers. Regular officers’ reluctance
to risk punishment for failed coups emboldens dictators to interfere more.

conclusion
Coercive institutions in dictatorships maintain the credibility of the dictator’s
threat to use force against challengers and opponents. Most of what such
institutions do is preventive rather than active repression. They deter overt
opposition and monitor elites and ordinary citizens in order to interrupt plots
and opposition movements before they have even been organized.
Internal security services focus much of their effort on high-ranking and mid-
level elites. They gather information about party ofﬁcials, military ofﬁcers,

15
We caution against interpreting the estimate for international conflict because identification
relies on a very few “positive” cases in each model.
16
This set of findings is consistent with Morency-Laflamme’s (2017) analysis of Benin and Togo.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:17, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.007
174 Ruling Society: Implementation and Information Gathering

high-level civil servants, and administrators, in addition to union leaders,

teachers, journalists, opposition leaders, and anyone else able to undermine
policy implementation or plot the dictator’s ouster. Security agencies combine
spying with the authority to arrest, intimidate, interrogate, and execute people
thought to represent a threat to the regime or dictator. Some security services
also employ informers to spy on ordinary people, but this level of surveillance
requires more financial resources than many dictatorships have. The internet
has increased the feasibility of monitoring ordinary citizens and interfering with
their access to information and ability to express opposition, however.
The coercive institutions of dictatorships are weapons that can be used by
the dictator to undermine or even murder other members of the dictatorial elite
if he can gain full control over them. Other members of the elite, however, have
strong incentives to resist the personalization of control over security services
and the dictator’s efforts to undermine the autonomy of the army.
If the dictator has secured personal control of the internal security apparatus,
he has a major advantage in the power struggle with other members of the inner
circle. Control of internal security tilts the distribution of two crucial resources,
information and capacity for violence, in the dictator’s favor and thus reduces
the likelihood that other elites could oust him or constrain his behavior via
credible threats to oust. The dictator’s control of internal security amounts to a
major step in the direction of the personalization of rule, and other members of
the inner circle can rarely reverse it within a dictator’s lifetime.
Armies are the last defense of dictatorships threatened by rebellion or upris-
ing, but since military coups are the most frequent means of ousting them,
dictators often fear their armies. Consequently, many dictatorships create
paramilitary forces to defend against coups and counterbalance the regular
army. We term paramilitary forces recruited from ethnic, partisan, or regional
groups closely tied to the dictator “loyalist” to distinguish them from other
irregular armed forces such as party militias and anti-insurgent forces.
As a further safeguard against the army, dictators also want to manipulate
recruitment, promotions, and retirements to favor loyal officers, and they may
want to purge, arrest, and execute officers whose loyalty they suspect. The main
impediment to interference with military leadership is the dictator’s fear of
coups, which officers may risk in order to safeguard the army’s autonomy
and their own careers and futures. Because the establishment of a loyalist
paramilitary force increases the risk that coups will fail, dictators supported
by them can interfere with military leadership with greater impunity.
These strategies for keeping armed supporters in check work some of the
time, but officers can at times fight back, and other members of the dictatorial
elite may support them. The establishment of paramilitary forces and the
politicization of promotions have been mentioned as causes for a number of
coups (as discussed in Chapter 3), and purged officers have also gone on to lead
insurgencies in some instances (Roessler 2016). Such a backlash can lead to
authoritarian breakdown, a subject to which we now turn.

Why Dictatorships Fall

The Egyptian monarchy fell to a coup led by Colonel Gamal Abdel Nasser and
the Free Officers in 1952 (Haddad 1973, 7–23). The dictatorship established by
Nasser lasted until 2012, when popular protests forced its final attenuated
incarnation to allow a fair election. Massive, largely nonviolent demonstrations
forced the resignation of the regime’s fourth dictator, Hosni Mubarak, in 2011,
but the officer corps, a pillar of the regime since 1952, remained in control of
the country until their candidate lost the 2012 election. This election ended
Egypt’s second modern authoritarian regime. The new democratically elected
regime lasted only until another military coup ended it in 2013.
Egypt’s experience illustrates that dictatorships can fall in a variety of ways,
ranging from coups carried out by a handful of officers to election losses and
popular uprisings. In some instances, the end of a dictatorship leads to democ-
ratization; in others, a new autocratic leadership group takes over, as in 1952
Egypt, bringing with it new rules for making decisions, a new elite group, and a
different distribution of benefits and suffering. This chapter investigates how
and why dictatorships end, as well as why democracy sometimes follows, but
often does not.
We begin with some basic information about how dictatorships end. Next,
we describe the decision calculus facing regime insiders as they decide whether
to desert a dictatorship, followed by a discussion of how economic and other
kinds of crisis can alter their assessment of the costs of opposition. We then
analyze how characteristics of the dictatorship itself contribute to destabilizing
conflict within the inner circle. In the sections that follow, we investigate how
specific pre-seizure-of-power features of the group that established the dictator-
ship influence the way the dictatorial elite responds to challenges and the
opposition they cause. We explain why some kinds of dictatorship survive
crises better than others and, finally, why some tend to exit peacefully through

177

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:33:58, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.008
178 Dictatorial Survival and Breakdown

fair, contested elections when faced with strong opposition, while others grip
power with tooth and claw until violently cast out.

how dictatorships end

Potential opposition leaders exist at all times in all dictatorships. They may
come from groups outside the regime’s distributive network, but they often
come from groups currently or formerly allied with the dictator. Because ties to
the dictatorship often entail resource and network advantages, it is easier for
those linked to it to challenge the dictator and regime than it is for those
without such resources (Svolik 2012; Roessler 2016). Overthrows led by cur-
rent or former regime allies are thus common.
Citizens living under dictatorship, whether regime insiders or not, must
decide whether to challenge it and, if so, the method for taking action. Potential
opposition leaders assess their options carefully, taking into account whether
they have the support and resources they need to mount a particular type of
challenge successfully, as well as the risk and potential cost of failure.
These strategic choices depend on the costs and difficulties of creating and
maintaining different kinds of opposition in varying circumstances. Some
methods of ending dictatorships, such as election campaigns, require much
more overt opposition support than others, such as coups. In 1952 Egypt,
many people felt angry about British control of the Suez Canal, but citizen
anger was not aimed at ousting the king. Officers chose a coup as their method
of overthrow because they had a comparative advantage in the deployment of
force, but also because a coup had a reasonable chance of success at a time
when organizing sustained popular opposition would probably have been
impossible. Political and economic circumstances also influence when potential
opposition leaders can attract enough support to succeed. The Free Officers
chose a time when many Egyptians felt their government had failed to defend
their interests, and when the government itself was in chaos. This timing
reduced the chance that citizens would defend the king and that the government
would respond to the military challenge quickly or coherently.
Potential opposition leaders’ choices about how to try to end dictatorships
cause regime breakdowns to unfold in the patterns that we label coups, popular
uprisings, opposition election victories, and so on. Each of these labels identifies
not only the form an attempt to overthrow a dictatorship takes but implicitly
the kind and approximate number of people involved in the effort. “Coup”
labels an ouster of political leaders carried out by military defectors from the
regime. Coups require the voluntary cooperation of only a few officers because
lower-level officers and soldiers usually obey orders. They typically take only a
day or two, and usually no more than a few people are killed during them. They
are thus relatively easy for officers to organize. “Popular uprisings” require the
voluntary cooperation of many more people, usually civilians; they can last
from a few days to many months; and they often result in quite a few deaths

% of regime collapse events

0.35 0.26 0.17 0.08 0.08 0.04 0.02

Coup Election Popular Insurgency Insider Foreign State

uprising rule change imposed dissolution*
figure 8.1 How autocratic regimes end.
*State dissolution includes cases in which a previously independent state was subsumed
into another one, cases in which a state broke up into constituent parts, and cases in
which a dictatorship ended but was not immediately followed by a government that
controlled a substantial part of the state’s territory.

and injuries even though protesters are generally unarmed. “Insurgencies” also
require the cooperation of substantial numbers of people, but involve confron-
tation between armed opposition and the military forces defending the dicta-
torship. Consequently, many more are likely to be killed. Insurgencies can last
from a few days to many years, and during that time, insurgent leaders must
ﬁnd ways to arm, train, and feed their troops. So, this strategy requires a much
more substantial commitment from participants than popular uprisings.
Finally, a successful opposition election campaign requires cooperation from
even more people than uprisings or insurgencies, usually near 50 percent of the
adult population, but participation is relatively costless compared with other
methods of ousting dictatorships.
Figure 8.1 shows how dictatorships ended between 1946 and 2010. It
reveals that more dictatorships fell to coups than to other kinds of challenge.
The second most common means of ending dictatorships is elections won by
someone not supported by the dictatorship.1 Overthrow by popular uprising
occurred less often before 1990, but has become much more frequent since the

1
Regime-ending elections include both semi-competitive elections that the dictatorial ruling party
expected to win but did not and fair elections that dictatorships that had agreed to step down
organized as a means of determining who would succeed them. The latter are usually organized
by military dictatorships.

0.3
Coup attempts per year

0.2

0.1

0
1950 1960 1970 1980 1990 2000 2010
Year

No election/one-party election Multiparty election

figure 8.2 Semi-competitive elections and coup attempts in dictatorships.

end of the Cold War. Insurgencies toppled some dictatorships, as did foreign
invasions. Some ended when regime insiders changed the basic rules governing
leadership choice, e.g., adopted universal suffrage, transforming oligarchy into
democracy.
Coups are the most common method of ending dictatorships for the same
reason that they are the most common means of initiating them: coups require
the cooperation of the fewest individuals, and soldiers have weapons. Though
coups are easier to coordinate than other forms of regime ouster, however, they
entail risks. About half of coups fail (Singh 2014). Failed coups can lead to
dismissal from the army, imprisonment, and execution for treason, so the cost
of failure for the top ofﬁcers involved can be high.
The high cost of failure and the irreducible element of luck in whether
coups succeed lessen the appeal of plotting where alternative mechanisms for
ending dictatorships exist. Coup attempts have declined since the end of the
Cold War, when more dictatorships began to allow semi-competitive elections
in which multiple parties can compete (Kendall-Taylor and Frantz 2014).
Figure 8.2 shows the relationship between semi-competitive elections in dicta-
torships and the frequency of coup attempts from 1946 to 2010. The vertical
bars show the yearly rate of coup attempts in dictatorships that either hold
no elections or hold elections that offer voters no choice. The solid line shows
the attempted coup rate in dictatorships that hold semi-competitive elections.
As the graph demonstrates, coup attempts tend to be more common in
dictatorships that do not allow the opposition to compete in elections – as
expected if potential opposition leaders chose methods for trying to end

dictatorships strategically.2 Where the opposition can participate in elections,

however unlikely they are to win, regime opponents are less likely to take the
risk of a coup attempt.
In sum, potential opposition leaders try to end dictatorships using the means
of doing so with the best chance of success – given their resources and support
networks – and the lowest potential cost for failure. They seldom indulge in
quixotic acts. They also choose times to act when they expect substantial
support from others for regime change. We discuss the dynamics that underlie
potential opposition leaders’ decisions to take action against the dictatorial elite
in the section that follows.

individual support and opposition

Dictatorships provide substantial benefits to regime insiders and some citizens,
as described in Chapter 6, but they also survive because most of the time those
who do not receive benefits avoid the costs of opposition. Opposition increases
when the benefits that individuals usually receive have declined, when the risk
of opposition has fallen, or both. In order to build a foundation for analyzing
the effects of specific events and institutions on the survival of dictatorship, this
section develops an abstract description of individuals’ decision calculus as they
decide whether to oppose the dictatorship.

The Interests of Regime Insiders

Ordinary citizens can play a large role in ending dictatorships, but they rarely
do so without leadership, and the individuals who lead them often spent earlier
stages of their careers as supporters of the dictatorship. This is so partly because
the dictatorship is often the only game in town for the politically ambitious and
partly because participation in the dictatorship is the best way to gain access to
the resources and build the clientele networks that facilitate opposition mobil-
ization later. When members of the inner circle defect, they can often take with
them the clientele networks originally built within the ruling group using state
resources (Garrido 2011).
Members of the inner circle will defect to the opposition if:

2
The attempted coup rate is the average number of attempted coups per year. In regimes lacking
semi-competitive elections, the attempted coup rate is 14.0 percent; in those holding semi-
competitive elections, the attempted coup rate is 9.6 percent. During the pre-1990 period, the
attempted coup rate is 50 percent higher in regimes lacking semi-competitive elections than in
regimes with semi-competitive contests. But the gap narrows during the two decades from 1990 to
2010: the coup rate is only 25 percent higher during this period.

where:
p is the probability the status quo regime will survive
q is the probability of maintaining/acquiring office in the status quo
regime
r is the probability of achieving office in the new regime (for a partisan
of the new)
s is the probability that a partisan of the old regime achieves/retains
office in the new regime
q>s
t is the probability of the new regime turning out as anticipated3
Oj is the benefit of office to a partisan of the status quo in the status quo
regime
Ok is the benefit of office in a new regime to a partisan of the new
regime
Ol is the benefit of office in a new regime to a partisan of the old regime
Oj > Ol
C is the cost of opposing the status quo regime.
The term qOj captures the idea that the better the posts individuals have or
expect, the more they would have to lose by defecting, but also that no one in a
dictatorship can be sure of occupying the same post or receiving the same
benefits tomorrow that they receive today. Dictatorships lack enforceable
contracts to ensure that promises of future rewards will be kept or that legal
provisions that have governed past events will govern future ones. Potential
defectors must assess their future prospects based on incomplete but well-
informed insider information.
If, however, a potential defector has been excluded from the inner circle or
has failed to receive a hoped-for post and does not expect it in the future, then
qOj becomes zero, substantially reducing the cost of defection. A member of the
elite’s failure to receive a hoped-for office can trigger defection – as happened
when leaders of the Mexican Partido Revolucionario Institucional (PRI) nom-
inated someone else as the party’s presidential candidate, and Cuauhtémoc

3
We assume that if the new regime turns out different from what those who invested to bring about
regime change had hoped (e.g., if a popular uprising results in a new dictatorship rather than
democracy), then those who hoped for a different outcome receive zero. We do not assume a
benefit from the fall of the status quo regime for individuals excluded from the new ruling group
because the new regime could turn out worse than the status quo regime (e.g., Qaddafi’s
dictatorship was worse than the monarchy it replaced for most Libyans, and Qaddafi’s regime
may have been better for most Libyans than the failed state that followed its overthrow).

Cárdenas left the party in order to organize a new opposition party to challenge
the PRI dictatorship. Loss of current positions of power can also result in
defection. After the Iraqi coup of 1958, Brigadier-General ‘Abd al-Karim
Qassem became commander-in-chief of the armed forces, prime minister, and
defense minister in the new dictatorship. His closest collaborator, Colonel ‘Abd
al-Salam Aref, became deputy prime minister, interior minister, and deputy
commander-in-chief of the armed forces, but when disagreements arose,
Qassem excluded Aref from the inner circle by naming him ambassador to
Germany. Rather than accepting the sinecure, Aref defected and returned to
Iraq to challenge Qassem’s leadership. Although he was arrested and served
time in prison, a few years later Aref led the coup that overthrew Qassem (Dann
1969, 20–32, 37–41, 77–89, 363–72).
Cárdenas and Aref are examples of members of dictatorial elites who
defected after events that caused a reassessment of their likely future in the
dictatorship. Most members of the PRI remained loyal in 1987, however, as did
many Iraqi officers during Qassem’s rule. The dictators’ supporters defect only
if they think their chances of achieving their ambitions by joining the oppos-
ition outweigh the future benefits of loyalty to the current ruling group. The
term (1−q)~Oj captures the idea that loyal members of the dictatorial elite
continue to receive benefits (such as the German ambassadorship) as long as
the regime survives, even if they lack high-level posts.
Insiders also decide to defect when events lead them to believe that the
regime may fall, lowering their assessment of p, the perceived chance of regime
survival. If the regime is going to collapse, those who jump ship earlier have
better prospects for achieving respected positions in the opposition than those
who cling longer to the old order. Because no one knows whether and when the
regime will fall, however, one insider may defect while others see defending the
dictatorship as the more sensible strategy.
In some situations, members of the status quo ruling group may continue to
occupy valued posts after regime change. If democracy replaces the dictator-
ship, for example, the former ruling party often becomes a viable party in the
new regime, and some of its members may be elected to Congress or continue to
occupy high-level bureaucratic posts. Military officers also often retain their
posts after democratic transitions, even when commanding officers are forced
to retire or prosecuted. The term (1−p)sOl captures the possibility of continued
benefits for members of the old status quo ruling group after regime change. If
individuals expect continued benefits from remaining members of the old ruling
party or clique even after regime change, they have no reason to defect.
The terms on the right-hand side of the inequality are analogous to those on
the left, but refer to expected benefits under a different regime. The main
difference between left and right is that the right-hand term includes C, the cost
of opposition, and t, the probability that the new regime really turns out the
way its supporters claim or hope it will. The cost term is included because
opposition to dictators is invariably costly, and that cost must be borne

regardless of whether the opposition ultimately succeeds. The cost is likely to be

greater if the old regime survives, captured by the term p, which would be low if
the regime is expected to collapse and very high in stable autocracies.
The probability that the new regime will function as anticipated, t, is
included to reﬂect the uncertainty that exists about what will happen after
regime change. In order to attract citizen support, those who lead opposition
parties, uprisings, or insurgencies must promise democratization regardless of
what they intend or what is likely. The ubiquity of such promises suggests that
opposition leaders understand this. The outcome of successful ousters is uncer-
tain, however, for both elites and citizens, as they cannot be sure that they will
receive promised beneﬁts from the change. In fact, they cannot be certain that
the new regime will be more democratic, or better in any other way, than the
old one (Bueno de Mesquita et al. 2003). Slightly more than half of authoritar-
ian breakdowns since World War II were followed by new dictatorships, not
democratization (Geddes, Wright, and Frantz 2014).

The Interests of Others

Lower-level regime insiders face a similar decision calculus. Ordinary citizens’
decision-making differs from that of regime insiders in that they have little
expectation of achieving office in the current regime or a future one. So their
decisions about joining the overt opposition depend on whether they are
receiving benefits from the current dictatorship, their assessment of the prob-
ability that it will persist, the benefits they expect from an alternative to the
status quo, and the cost of opposition.
Potential opposition leaders unaffiliated with the dictatorship share a similar
decision calculus with ordinary citizens, except that they have a good chance of
securing office under the future regime. Their calculation would therefore
include terms like those on the right-hand side of Equation 8.1.
If citizens receive a stream of benefits under the status quo (e.g., a salary or
advantages associated with a clientele network linked to regime leaders), they
are unlikely to participate in a movement to unseat the dictator. Mass actions to
remove dictators most often occur when economic misfortune or policy failure
prevents regime insiders from delivering benefits, economic growth, and every-
day services to citizens (Bratton and van de Walle 1997).
A dictatorship’s reputation for zero tolerance of opposition and for draco-
nian punishments deters the expression of opposition by raising C. Effective
security services, pervasive networks of informers, and intrusive mass-level
institutions for social control raise the likely costs of opposition. Citizens are
more likely to join the opposition, however, if large numbers of others have
joined, both because large numbers increase the likelihood that the regime will
fall and because the presence of large numbers indicates a lower risk of
suffering punishment for joining (Kuran 1989, 1991; Lohmann 1994). Unless

the cost of opposition is near zero, however, as in secret-ballot elections, most

will not participate because they will receive any beneﬁts of regime change
whether they participated or not.

Perceptions about Regime Survival

Dictatorships rarely end unless many citizens oppose them, but they often fail
to end despite widespread citizen opposition, especially when people keep their
opposition private, as they do most of the time (Kuran 1989). As Adam
Przeworski (1986) explained long ago, in stable dictatorships most people seem
to support the regime, but after the dictatorship falls, almost everyone seems
to have wanted regime change. This kind of process is called a tipping
phenomenon.
Perceptions about the likelihood of dictatorial collapse follow the same
tipping logic. Before the dictatorship faces serious challenges, individuals inside
and outside the dictatorial elite expect it to survive. They adapt their behavior
to do the best they can within the current political system and, even if they hate
the dictatorship, see little point in shouldering the costs of overt opposition.
Once challenges begin to reduce perceptions of dictatorial invincibility, how-
ever, assessments can cascade downward until nearly everyone believes the
regime will fall.
Understanding that a minor challenge can precipitate a dramatic downward
rush in assessments of regime strength, some members of the dictatorial elite
may defect at this point in order to establish reputations with the opposition in
anticipation of regime change, but most will remain loyal. In this way, some
politicians who spend their early careers as loyal servants of the dictator can
become stalwarts of democratic politics during transitions – joining those who
defected after loss of ofﬁce.
Some citizens may also then make their opposition public when the ﬁrst
challenge emerges, but most will play it safe. As soon as some people begin to
express opposition, however, others recalibrate the likelihood of dictatorial
survival, leading more citizens to express opposition, and so on. As citizens
update their estimates of the prospects for regime survival, more and more
become comfortable expressing their discontent in election campaigns or dem-
onstrations. If the dictatorship can crush or defuse early demonstrations, per-
ceptions about likely survival may increase again. If not, however, they can
drop sharply. Eventually, those who sincerely support the dictatorship begin to
see costs associated with such support and keep quiet about it, contributing to
the perception that the regime is done for.
In this way, perceptions about future regime survival affect the willingness
of members of the dictatorial elite to defect and of citizens to express
overt opposition by changing their calculus of p, the probability of regime
survival.

the effect of crisis on decisions to oppose

the dictatorship
Many observers believe that economic crisis contributes to authoritarian break-
down. Here, we focus first on the mechanism through which externally induced
economic problems might threaten regime survival. As an example of an
external crisis, consider a fall in the international price of a country’s most
important export. Typically, such a decline would reduce the growth rate,
affecting the benefits available to citizens, and reduce government revenues.
Reduced revenues could result in downward pressure on public employment,
wages in the public sector, spending on schools and health care, state invest-
ment, and all the benefits that dictatorships provide for their supporters. After
the international price of oil dropped in 2013, for example, Venezuela’s
reduced ability to pay for imports led to empty shelves in many grocery stores.
These included the centerpiece of the Chávez/Maduro regime’s most popular
program for channeling benefits to supporters, the state stores that sell food at
subsidized prices in poor neighborhoods. So, not only was the benefit stream
provided by the dictatorship for all citizens reduced by the price shock, but the
ruling elite even lost some of its ability to deliver special advantages to its most
loyal supporters. These problems have worsened over time.
What happened in Venezuela is an extreme example of what happens after
price shocks and other economic crises. For those regime beneficiaries who
occupy posts in government, such challenges reduce the expected value of future
benefits from office, Oj, as wages fall. The likelihood of being appointed to a new
and/or better post or retaining the one held, q, also declines because of budget
pressures to reduce public employment, as do expectations about future benefits
from the regime in case one fails to achieve or retain a particular job, ~Oj.
For ordinary citizens, such crises reduce the expected benefits from the current
government, including expectations about future economic growth.
These reductions increase the attractiveness of joining the opposition, for
both regime insiders and ordinary citizens. If individuals see that many others
have become visible regime opponents, as often happens during economic
crises, their assessment of the probability of regime survival, p, also falls.
Through these two mechanisms, overt opposition can grow and the likelihood
of regime survival decline.
Dictatorial policy makers can mitigate or exacerbate the effects of exogenous
crises by their actions, just as democratic governments can. They make politic-
ally motivated choices not only about policy responses but also about which
parts of the population will bear the largest cost of the downturn. Members of
the dictatorial inner circle try to distribute the costs of the crisis so that they fall
most heavily on those with the least capacity to destabilize the regime, usually
the less organized and economically weaker parts of the population. The
Chávez/Maduro dictatorship, for example, closed many subsidized grocery
stores, but not those in politically volatile Caracas.

economic crisis and breakdown

Despite the belief that economic crisis destabilizes dictatorships and the plausi-
bility of the scenario described, some careful statistical studies have shown little
effect (e.g., Przeworski et al. 2000). We suggest that the contradictory results
found in large-N studies are caused by differences in the capacity of dictator-
ships to respond to economic crisis by redistributing the costs in politically
effective ways. In other words, we think that economic crisis is more destabil-
izing for some kinds of dictatorships than others, leading to weak or contradict-
ory results when all dictatorships are lumped together.
Regime elites, we suggest, can decrease popular opposition by building the
kinds of organizational infrastructure that can deliver help after economic
crises or natural disasters, and can prevent regime insiders from stealing aid
meant for disaster victims. The party-based networks of officials and supporters
built to incorporate more citizens into the dictatorship’s support base and
facilitate the collection of information about events and attitudes at the grass-
roots, described in Chapter 6, can also be used to deliver disaster aid or
redistribute the costs of economic crisis. We believe that the extensive patron–
client networks developed in party-led dictatorships help regime elites to con-
tinue distributing to those most essential for their survival while shifting the
burden of the crisis to other, politically weaker parts of the population. During
economic crises, ruling parties with wide-ranging distributive networks can
help prevent benefits from the status quo regime from dropping too far among
those citizens most capable of overthrowing the dictatorship.
All dictatorial support parties try to coopt citizens, but we believe that those
that had established extensive patron–client networks before the seizure of
power have an advantage over parties created after seizures of power. Their
advantage arises from the necessity of developing relationships between central
party leadership, local party leaders, and people living in different areas in
order to survive while out of power and, often, subject to repression. We expect
these relationships to have been especially strong where the party needed to
exchange protection or other benefits for manpower and resources in order to
maintain an insurgency or where it needed to exchange goods and services for
votes in competitive or semi-competitive elections. Because of their more
developed and extensive patron–client networks, we expect such support
parties to contribute to regime durability during crises.
In contrast, dictatorships that have failed to build such penetrating party
networks may lack the means to respond effectively when natural disaster
or economic crisis strikes. Even a manageable natural disaster or economic
challenge can become calamitous for the dictatorship if it fails to respond
effectively.
In the next section we test whether institutions associated with extensive
patron–client networks insulate dictatorships from the destabilizing effects of
economic crisis.

Empirical Approach
To investigate how economic crisis affects regime stability, we create a
binary variable that flags observations if the country has experienced nega-
tive growth during the previous two years, which we call crisis. If the lagged
two-year moving average of economic growth (per capita) is less than −2
percent, crisis takes the value of 1, and zero otherwise. We create a similar
indicator for good economic times, which we call boom; it is coded 1 if
growth (per capita) is greater than 5 percent.4 Separating out the two
extremes of economic growth allows for a more transparent test of the
proposition that economic crisis undermines authoritarian rule. If we
simply examined the effect of economic growth as a continuous variable,
we would not be able to assess the potentially destabilizing influence of
unusually poor growth easily, or the potentially stabilizing effects of espe-
cially good economic times.
As in earlier sections, we use an exogenous indicator of political institutions
that encompass citizens in well-organized networks, inherited party, which
identifies regimes led by a political party that was organized either to lead an
insurgency or to participate in elections before the authoritarian seizure of
power. This variable does not vary over time within particular regimes since
it measures a pre-seizure characteristic of the party that later became the
dictatorial ruling party.
To estimate the effect of economic crisis on regime breakdown, we test a
linear probability model with country fixed effects and time period effects,
while controlling for duration dependence with a cubic polynomial. This
approach allows us to model all country-specific characteristics, such as geog-
raphy, colonial history, and religion, without dropping dictatorships that
remain in existence throughout the time period covered by the data, such as
the Communist Party regime in China or the monarchy in Saudi Arabia.
We use a minimum number of control variables because many factors that
influence regime stability, such as protest and civil war, are post-treatment
phenomena; that is, economic crisis may cause protest or insurgency. Even
international war can result from a dictator’s initiation of conflict to divert
citizens from economic distress. Therefore, we include only three potential
confounders: prior experience of democracy, whether the dictator was a rebel
leader before the seizure of power, and whether he was a member of the
military. These variables ensure that the inherited party variable is not simply
picking up the destabilizing effect of earlier democratic experience or the
stabilizing influence of revolutionary party organization. First, we test the
following specification:

4
Crises occur in roughly 17 percent of observations, while booms occur in 19 percent.

No interaction Interaction included

0.034 0.057
Crisis

–0.047
Inherit × Crisis

–0.06 –0.046
Inherited party

0.011 0.03
Boom

–0.035
Inherit × Boom

0.026 0.025
Prior democracy

–0.021 –0.022
Military leader

–0.019 –0.02
Rebel leader

–0.05 0 0.05 –0.05 0 0.05

Coefficient estimate
figure 8.3 Economic crisis, party networks, and authoritarian breakdown.

PrðY t ¼ 1jY t−1 ¼ 0Þ ¼ α0 þ β1 ∗ Inherit it þ β2 ∗ Crisis þ β3 ∗ Boom

þβ4 ∗ Rebel leader þ β5 ∗ Military leader þ β6 ∗ Prior Democracy þ δi þ ηt þ ε
ð8:2Þ
where δi are country effects and ηt are five-year time period effects. Then we
include an interaction between crisis and inherited party as well as one between
boom and inherited party. This allows us to test whether the effect of crisis
varies depending on whether an inherited support party leads the dictatorship.
Figure 8.3 reports the results. When no interaction is included in the specifi-
cation (left side), the estimates for crisis are positive and significant, indicating
that economic crises are associated with authoritarian breakdown. Further, the
estimate for inherited party is negative and significant, showing again that such
parties contribute to regime durability. This result suggests that the patron–
client networks established in long-lasting parties, the ability to cope with
succession (also a feature of inherited parties), or both contribute to regime
resilience. This model does not show whether such parties specifically help
dictatorships weather economic crises, however.
The second model (right side) does that. Including an interaction term allows
us to examine the effects of patron–client networks that reach ordinary citizens
on the likelihood of breakdown during or soon after economic crisis. The

estimate for the effect of the interaction between crisis and inherited party is
negative and statistically signiﬁcant. This means that economic crises are less
likely to destabilize dictatorships led by parties that have developed extensive
patron–client networks.5 The positive estimate for crisis alone means that
economic downturns increase the likelihood of collapse in dictatorships that
lack extensive party networks. In short, the data analysis conﬁrms the argument
that a well-organized ruling party can help dictatorships survive economic
crises.

power concentration and regime survival

External crises are not the only challenges dictatorships face. Conflict within
the inner circle can threaten regime survival even without other challenges. As
Guillermo O’Donnell and Philippe C. Schmitter (1986) noted decades ago, the
collapse of authoritarian regimes can begin with factionalism and disagree-
ments among the dictatorship’s most powerful decision makers. Insider conflict
can motivate the dictator to demote or exclude members of the inner circle,
potentially weakening it; alternatively, losers in factional struggles may defect
from the regime in order to lead opposition to it. If elite conflict becomes public,
it can also change citizens’ perceptions of the likelihood of regime survival, p,
increasing their willingness to undertake overt opposition. Intra-elite struggles
can thus make the regime seem weak to outside opponents, emboldening latent
opposition to mobilize against it, as well as directly causing elite defections.
Elite conflict often involves challenges to the incumbent dictator. In
Chapters 4 and 5 we showed that early leadership conflict can result in either
power sharing between the dictator and his closest allies or the concentration of
power in one man’s hands. Chapter 5 focused on the special difficulty of
enforcing power-sharing bargains among members of armed seizure groups.
Now we investigate how those earlier outcomes affect longer-run regime resili-
ence and the ways dictatorships break down.
We focus on how the concentration of power can reduce subsequent conflict
within the dictatorial inner circle. In order to concentrate power, dictators
engage in strategies that end up reducing internal differences. They purge
members of the inner circle most inclined or most able to challenge them, while
frightening the rest into quiescence. Over time, they develop ways to control the
futures, lives, and welfare of other members of the dictatorial elite. Spy agen-
cies, as noted in Chapter 7, often report the contacts and conversations of
members of the inner circle to them, making possible preemptive strikes against
anyone suspected of critical thoughts, let alone deeds. When dictators control
appointments to top offices and the political police report directly to them, open

5
The estimate for the linear combination of crisis plus the interaction is 0.010 and not statistically
different from zero.

disagreement and criticism of the dictator disappear from the corridors of

power. Plotting becomes riskier. The dictator’s ability to use rewards, priva-
tions, and violence against his closest allies prevents most inner-circle conflict
during his lifetime.
The strategies that dictators with concentrated powers use to keep them-
selves safer can undermine prospects for regime survival after the dictator dies,
however. Typically, personalist dictators exile, jail, or execute the country’s
most able politicians, administrators, and officers in order to reduce the likeli-
hood that they will eventually lead opposition. Such dictators value loyalty
more than competence when promoting officers or making administrative
appointments, which undermines the military as a fighting force and the bur-
eaucracy as a reservoir of technical skills, further reducing the likelihood of
challenges. The dictator’s worry about survival often also leads to the rapid
rotation of officials through offices and locations and of military officers
through commands, which reduces their ability to build loyal political networks
of their own that might serve as the core of a plot or opposition movement,
but also undermines their ability to develop experience and expertise. These
strategies reduce the competence and organizational resources of potential
successors.
The dictator’s strategic use of corruption can also damage prospects for
regime survival after his death. Corruption tends to be higher in more person-
alized dictatorships than in those in which power is less concentrated (Chang
and Golden 2010). The dictator’s personal corruption arises not only because
he needs resources to buy political support but also from his knowledge that he
and his family are likely to lose their in-country assets and wind up in exile
(at best) if he loses power. So he needs a substantial insurance fund held safely
outside the country.
The dictator’s lieutenants engage in corruption for the same reason; in case
the regime falls or they fall out of the dictator’s favor, they will need funds to
support their families in exile. The supporters of personalist dictators are
vulnerable to loss of office, arrest, exile, and murder if the dictator becomes
suspicious about their loyalty. So they need insurance funds. And if the dictator
rotates supporters rapidly through offices, they have incentives to amass these
funds quickly, while they can. Few supporters of personalist dictators can
continue political careers post-dictatorship, another reason to get it while they
can (Bratton and van de Walle 1997).
Facilitating lieutenants’ corruption can be part of the strategy of power
concentration. Dictators who allow rampant corruption often use their security
police to collect information about it so that officials who displease them for
any reason can be humiliated, deprived of office, and jailed when their corrup-
tion is “discovered.” Even though nearly all officials engage in corruption and
everyone knows that everyone does it, corruption scandals erupt frequently in
personalized dictatorships, as the dictator uses this strategy to keep his lieuten-
ants on their toes, off-balance, and insecure.

The strategies dictators with concentrated powers use to maintain control

reduce destabilizing inner-circle conflicts during their lifetimes, but also increase
the difficulties anyone who follows them in office will face. If the dictator dies
or is assassinated, his disarticulated supporters, accustomed to competing
against each other for his favor, have difficulty overcoming the collective action
problem of regime maintenance. The dictator’s hollowing-out of regime insti-
tutions leaves them incapable of shaping bargains between the dictator’s suc-
cessor and other regime insiders. As a result, like the initial bargaining period
following the seizure of power, uncertainty is pervasive and the risk of regime
collapse is high.
These features of personalized dictatorship lead us to expect that, all else
equal, regime breakdown would be less likely during the lifetimes of dictators
with concentrated power, but would become more likely after their deaths.
The overall effect of personalism on regime survival, however, depends on
how long an otherwise similar dictatorship would have been expected to
survive under more collegial leadership. If we would expect a regime to
survive only a few years, then increasing its resilience during the dictator’s
lifetime would, on average, increase its duration. If, on the other hand, a
dictatorship has the characteristics that would lead us to expect survival
for many decades, then the increased vulnerability to breakdown after the
first dictator’s death associated with personalism would likely reduce its
expected duration.
The two regime characteristics with well-established effects on the duration
of dictatorships are dominant-party rule, which tends to increase it, and
military rule, which tends to decrease it. Party-led dictatorships can enforce
norms about the selection of leaders better than other dictatorships can, and
thus are more capable of managing succession without crisis. Yet party
constraints on dictators are precisely the things a dictator intent on concen-
trating power in his own hands wants to change. We therefore expect the
personalization of rule to increase the likelihood of the dictator retaining
office until he dies, but to reduce regime longevity in dictatorships organized
by parties.
In contrast, military-led dictatorships tend to replace dictators frequently via
sometimes-violent coups. The dispersal of arms within the dictatorial elite in
military-led regimes encourages recurring conflict within the inner circle. This
instability can in turn prompt officers to return to the barracks if they believe
conflict threatens the military’s unity (Geddes 1999). In military-led regimes, we
therefore expect personalization to extend not only the dictator’s survival but
also the regime’s because the average collegial military regime has a shorter
lifespan than the average dictator.
To test these ideas, we examine the direct effect of the personalization of
power on dictatorial resilience and how personalism interacts with exogenous
characteristics of the seizure group that tend to be carried over into the
dictatorship.

Data and Measurement

To carry out this investigation, we use the time-varying measure of personalism
introduced in Chapter 4. As a reminder, it is a composite measure of person-
alism from an item response theory (IRT) two-parameter logistic model (2PL),
rescaled on the [0,1] interval, where higher levels of personalism approach
1 and lower levels of personalism approach 0. The items used to derive the
latent estimate capture both personalization of the supporting political party
(rubber stamp, party executive committee, appointments, and new party) and
the leader’s consolidation of power over the military and security forces (secur-
ity apparatus, paramilitary, promotions, and purges). This time-varying indica-
tor measures differences in power concentration between regimes, between
leaders in the same regime, and over time during any individual leader’s tenure
in power. It thus allows us to investigate how personalism influences regime
stability in different contexts.
Because our indicator of personalism “measures” post-seizure behavior, we
cannot rule out the possibility that dictators pursue these strategies after con-
sidering their prospects of survival, which we cannot observe. Indeed, we
believe that this is exactly what they do. However, our goal here is not to show
that personalism causes regime survival but rather to examine whether the
concentration of personal power in the leader’s hands correlates with regime
longevity and whether exogenous traits – such as military rule or dictatorial
leadership organized by a political party – shape this relationship.
To review the measurement of the exogenous seizure group characteristics
we focus on in this section:
A dictatorship with an inherited support party (inherit) is one in which the
ruling party was originally organized during an earlier regime either to
participate in (democratic or autocratic) elections or to lead an armed
insurgency. While this measure only uses information about the seizure
group prior to gaining power, it nonetheless identifies most of the same
country-years as the older regime-level indicator, dominant party, proposed
by Geddes (1999, 2003) and updated by Geddes, Wright, and Frantz
(2014).6 The older regime-type coding tried to distinguish dictatorships in
which the ruling party was strong enough to constrain the dictator from
those in which it was not. The latter were labeled personalist. Thus, the cases
with inherited parties, though measured before the seizure of power, are
party-led dictatorships in which the level of personalism tends to be rela-
tively low. This may increase the difficulty of showing how personalism
affects the survival prospects of such regimes. We note, however, that

6
Eighty-four percent of regimes coded as dominant party by Geddes, Wright, and Frantz have an
inherited support party, and 68 percent of regimes (but 80 percent of observations) with inherited
parties are coded as dominant party.

though dictators have greater difﬁculty concentrating power when an

inherited party organizes the regime elite (as shown in Chapter 4), some
dictators – e.g., Stalin, Mao, and Kim Il-sung – have nevertheless succeeded
in doing so.
We define a military-led regime (military-led) as one in which the first
dictator was an active-duty or recently retired member of the military of
the regime that governed immediately before the seizure of power.7 Dictators
whose military titles were earned in the insurgency that brought them to
power are coded as insurgents, not military, and therefore the regimes they
lead are not considered military-led. This operationalization, again, uses
only information from before the seizure of power and thus does not
contain information about the behavior of the regime or its leader once in
power. Military-led simply identifies dictatorships initiated by military
seizure groups.
Military-led does not distinguish between more and less personalized
military-led regimes, as did the older regime-type coding. All of the regimes
coded by Geddes, Wright, and Frantz (2014) as “military” have a first military
leader, but only a little more than half (56 percent) of dictatorships with a first
military leader were coded as “military” in the older data. The coding rules for
that data set limited the term “military” to regimes in which the officer corps
could constrain the dictator’s discretion.8 Other dictatorships led by officers
were coded as “personalist.” In the data we use now, personalism can be
measured as a separate time-varying characteristic measured yearly.
By using measures of military-led and inherited party that rely only on pre-
seizure characteristics of the first leader and the seizure group, we can examine
how the post-seizure behaviors that we identify to measure personalism influ-
ence regime survival in different contexts. We have constructed two ways of
defining the autocratic context (military-led and inherited party) that are
exogenous to regime survival. That is, dictatorial leaders’ strategies for
avoiding regime collapse cannot have influenced the formation of a political
party organized before the seizure of power. Nor can it have affected whether
military officers first seized power.
Because we define military-led regimes and inherited-party regimes in this
way, we can use them as stand-ins for “political institutions” in the Northian
sense of constraints or norms that shape political behavior and thus structure

7
If the ﬁrst leader of the regime is not a member of the military but a subsequent leader comes from
the military, we do not consider this regime military-led because choosing a subsequent leader of a
certain type may be a by-product of attempts to enhance regime longevity. In the main estimating
sample of the 270 regimes in 117 countries, 38 percent of regimes have an inherited party and half
are military-led.
8
See Geddes, Frantz, and Wright (2014) for a discussion of different meanings of the term
“military rule” and the conceptual underpinnings of different ways of coding military-led
dictatorships.

equilibrium outcomes (North 1990). These institutional features thus circum-

vent the Rikerian objection to institutional analysis (Pepinsky 2014) because
dictatorships cannot change these characteristics after seizing power.9

Empirical Approach
We examine the effects of personalism and exogenous characteristics of the
seizure group on autocratic breakdown using two types of estimators. First, we
test a logistic regression model with control variables, including decade dum-
mies to capture trends in autocratic survival across time and regime duration
polynomials to model time dependence in the data. The duration polynomials
allow the logistic regression to mimic standard survival approaches, such as the
Cox proportional hazard model (Beck and Katz 1995; Carter and Signorino
2010). For this estimator, we model heterogeneity across regimes using regime-
level random effects. Second, we estimate a linear probability model with
country and year fixed effects. This estimator allows us to incorporate fixed
unit effects without dropping countries that do not experience regime change in
the period from 1946 to 2010. With each estimator we cluster the standard
errors by regime.
We control for confounders thought to affect regime breakdown: log GDP
per capita, log oil rents per capita, conflict (civil and interstate), and prior
democracy.10 Further, we include a variable that indicates whether the regime
originated in revolution, from Levitsky and Way (2013), who argue that post-
revolutionary regimes are especially durable. Including this variable is import-
ant because we want to know whether our findings hold even after we account
for the revolutionary origins of some of the most durable autocracies of the
twentieth century.11 In other words, we want to make sure that the effect we
show for inherited parties is not entirely due to revolutionary parties.
Finally, each specification includes binary indicators for military-led regime
and inherited party regime. These are not mutually exclusive categories because
some military-led dictatorships seized power with the aid of inherited political
parties. After estimating a specification without interaction terms, we report
estimates from separate specifications, one of which includes an interaction
between military-led and personalism and the other an interaction between
inherited party and personalism. We use these interactions to show that the

9
There may, however, be unobserved factors that cause selection into military-led or inherited
party regimes that also cause regime (in)stability. We rule out unobserved country-specific
sources of spurious correlation by employing country fixed effects estimators.
10
We also tested models with two potential post-treatment variables, economic growth and anti-
government protest, with similar results.
11
We drop this variable with the fixed effects estimator since only ten countries have periods of rule
under both a revolutionary regime and a nonrevolutionary regime. Results including this
variable, however, remain consistent.

RE, nonlinear FE, linear

Personalism

Pers. × Military

Pers. × Inherit

Military-led

Inherited party

GDP pc (log)

Civil war

Revolut. regime

Prior democracy

Oil rents

Int'l war

–4 –2 0 2 4 –0.2 –0.1 0 0.1 0.2

Coefficient estimates

No interaction Personal × Military Personal ×Inherit

figure 8.4 The effect of personalism on authoritarian breakdown.

FE, ﬁxed effects. RE, random effects.

effects of the concentration of power in a dictator’s hands differ depending on

whether his initial support base was rooted in the military or in an
inherited party.

Results
Figure 8.4 shows the results. Estimates from the first specification (top line in
each cluster) in the left panel do not include either interaction term and thus
simply estimate the average effect of personalism across all dictatorships in the
sample. The coefficient for personalism is negative and statistically significant,
suggesting that more personalism on average is associated with a lower

probability of regime breakdown. The estimate for inherited party is also

negative and significant, indicating that inherited parties prolong dictatorships,
all else equal.12 That for military-led regime is positive and significant, consist-
ent with prior research showing that military regimes tend to be fragile (Geddes
2003). Like previous studies, our results link specific autocratic institutions to
greater durability (Geddes 1999; Gandhi 2008; Magaloni 2008). Unlike previ-
ous studies, however, we are able to rule out the possibility of reverse causation.
That is, the possible tendency of more durable dictatorships to adopt party
institutions (Pepinsky 2014) could not cause the relationship because the meas-
ures we use to capture institutions pre-date the existence of the dictatorship.
The next two models (identified by triangles and squares) include interaction
terms, which estimate the effect of increases in personalism in military-led
dictatorships (triangles) and in those led by inherited parties (squares). These
estimates show statistically significant effects in opposite directions. The esti-
mate for the effect of increases in personalism in military-led regimes (the linear
combination of military personalism plus personalism) is negative and sig-
nificant. In contrast, the estimate for the effect of increases in personalism in
party-led regimes (the linear combination of inherit personalism plus person-
alism), while negative, is not statistically significant.
The negative estimate for personalism in military-led dictatorships means
that as leaders in these regimes concentrate more power in their own hands, the
regimes they lead become less vulnerable to overthrow. This result makes sense
because for military dictators, power concentration involves building up non-
military security forces such as internal security police, who usually spy on
officers as well as civilians, and paramilitary forces recruited from regions and
ethnic groups especially likely to be loyal to the dictator.13 These forces loyal to
the dictator reduce the ability of officers in the regular military to oust the
dictator and thus also to force him to consult with them. Power concentration
also often includes a gradual change in the dictator’s support base from
primarily military to greater reliance on organized civilian, ethnic, or even
family networks. A dictator who succeeds in replacing some of his initial
military support with civilians has reduced the proportion of members of the
inner circle who control the arms and men needed to replace him at relatively
low cost. He may at the same time replace officers from multiple regions and
ethnic groups with co-ethnics, co-regionalists, or family members, thus further
reinforcing the likelihood of future loyalty.
The positive coefficient for military-led by itself in the model that includes
the interaction indicates that more collegial military regimes (those with the

12
Recall that the speciﬁcation includes regime-case random effects as well as a control for revolu-
tionary regimes. The result for inherited parties remains after dropping revolutionary regimes
from the estimating sample.
13
Remember that control of security forces and the creation of loyal paramilitary forces are two of
the indicators that go into the measure of personalism.

lowest concentration of power in the leader’s hands and lowest personalism

scores) tend to be short-lived. This result reinforces earlier findings about
collegial military rule and statements above about the fragility of dictatorships
in which armed force is widely dispersed within the ruling group. In more
collegial military-led dictatorships, security services often remain within the
military chain of command, and paramilitary forces are not usually created
unless the military wants help fighting insurgencies. Thus in collegial military-
led dictatorships, the dictator may personally control no armed forces to defend
him if other officers decide to remove him.
In contrast to the findings about personalization in military-led regimes, the
results show that power concentration in dictatorships led by inherited parties
makes them more vulnerable to breakdown. The coefficient for inherited party
by itself indicates that party-led regimes with relatively collegial inner circles are
more durable than other kinds of dictatorship. The coefficient for the inter-
action between inherited party and personalism means that the personalization
of party-based rule decreases its durability. This reflects the difficulty of main-
taining personalized rule after the death of the individual who has concentrated
vast powers in his own hands even in party-based regimes.
The control variable estimates are stable across specifications and in the
expected directions: oil-rich autocracies and revolutionary regimes are more
stable, while dictatorships in countries with a prior history of democracy and
those experiencing conflict are more likely to collapse.
Next, we turn to the estimates from the specifications with fixed effects,
shown on the right side. Recall that this estimation approach accounts for all
unobserved country-specific factors – such as religion, state strength, colonial
legacy, ethnic fractionalization in society as well as in the military at the time of
independence, history of democratic experience, geographic region, terrain, and
climate – that might influence the propensity for either a group led by officers or
an inherited party to seize power. The first specification (top estimate in each
cluster) excludes interaction terms. The estimate for personalism in this model
is negative and statistically significant, indicating that even when we look only
at variation over time within countries, concentrating power in the hands of the
dictator increases regime longevity on average.
The estimate for military-led regimes is positive but very close to zero and
not statistically significant. Because the group of military-led regimes includes
about equal numbers in which the dictator concentrates power (what Weeks
[2014] calls “strongman” regimes) that last much longer than other military-led
regimes and those in which he does not (what she calls “juntas”), the near-zero
estimate for military-led in a fixed effects specification should not be surprising.
The estimate for inherited party is still negative and statistically significant,
providing strong evidence that dictatorships led by inherited parties are more
durable, on average, than those that lack such parties. Remember that we have
coded this regime characteristic based only on pre-seizure information, and
thus it is not contaminated by the strategic maneuvers of regime elites trying to

retain power after they have seized it.14 Further, because we have estimated the
effect of this feature of dictatorships in a fixed effects model, we can rule out the
possibility that relatively time-invariant factors specific to individual countries –
such as the historical legacy of strong states (Slater 2010) or aspects of political
economy that shape dictators’ or elites’ political preferences (Pepinsky 2014) –
explain the finding.
In the second fixed effects specification (middle estimate in each cluster), we
include the interaction between personalism and military-led regimes. The
estimate for military-led regimes alone is positive and statistically significant
(though only at the 0.10 level). The positive estimate suggests that military-led
regimes in which the dictator has not concentrated power in his own hands are
roughly 5 percent more likely to collapse in a given year than collegial, civilian-
led dictatorships.
The estimate for personalism alone is negative and statistically different from
zero. This suggests that personalism stabilizes authoritarian rule in dictator-
ships initially led by civilians (when inherited party is controlled for). The
civilian-led dictatorships that lack inherited parties include monarchies, some
post-Soviet dictatorships in which the dictator juggles multiple, often short-
lived parties, the Sukarno regime in Indonesia, two brief Ecuadoran dictator-
ships led by Velasco Ibarra, and a few others. These regimes tend to be highly
personalized.
The estimate for the interaction between personalism and military-led
regimes is strongly negative and significant, confirming the earlier result that
the regime-prolonging effects of personalism are greatest in military-led
regimes.
Finally, the last model reported on the right side includes the interaction
between personalism and inherited party. The estimate for this term is positive
and statistically significant, meaning that personal concentration of power
makes dictatorships led by inherited parties less resilient. The estimate for
personalism alone is negative and significant, indicating that power concen-
tration strongly stabilizes regimes that lack an inherited party. The estimate for
inherited party alone is negative and significant, indicating that regimes led by
inherited parties with relatively collegial leadership are the most durable.
Combined, these results confirm the pattern identified by Geddes (1999), in
which what she labeled “military regimes” (roughly equivalent to less person-
alized military-led regimes here) are least resilient; “dominant-party regimes”
(equivalent to less personalized regimes led by inherited parties here) survived
longest; and “personalist regimes” (a combination of personalized military-led
and personalized civilian-led) occupied the middle ground.

14
In addition, there are no “hybrid regimes” in this analysis (Pepinsky 2014, 641). We have simply
used a concrete, observable feature of the pre-seizure history of the regime support party to
operationalize what we believe is a theoretically important concept.

0.15

0.1
% change in failure rate

0.05
0.046

0
0.006

–0.03
–0.05

–0.1 –0.093

95% ci
–0.015

Military-led Military-led Inherited party Inherited party

low personalism high personalism low personalism high personalism

figure 8.5 The effects of personalism in dictatorships with different leadership

conﬁgurations.

Figure 8.5 summarizes the ﬁndings for personalism in dictatorships with

different leadership configurations, using the estimates reported in the two fixed
effects models in Figure 8.4. Military-led regimes with the lowest personalism
score are almost 5 percent more likely to fail in a given year than civilian
regimes. Military-led regimes with the highest personalism scores, however,
are less likely to fail than civilian regimes. Turning to regimes with inherited
parties, we see the opposite pattern. Those with low personalism scores are
9 percent less likely to collapse than regimes without inherited ruling parties,
while regimes with inherited parties but high personalism are no more or less
likely to fail than regimes without inherited support parties. This figure illus-
trates the very different effects that concentration of power in the dictator’s
hands has on regime survival, depending on other inner-circle characteristics: it
enhances regime survival in military-led dictatorships but decreases it in dicta-
torships led by inherited parties.
This pattern of results reflects the consequences for regime durability of the
modes of bargaining and handling intra-elite conflict in dictatorships led by
different kinds of groups. Regimes based on collegial military rule tend to break
down easily, both because the dispersal of armed force encourages intraregime
conflict and because officers sometimes choose to return to the barracks if
internal conflict threatens military unity (Geddes 1999). The personalization
of military-led regimes reduces intra-elite conflict by concentrating power in
one man’s hands and thus tends to prolong regime survival during that man’s
lifetime. If military seizure groups cared only about regime survival, more
officers would consent to the personalization of rule. They also care about the
integrity of the military as a fighting force, however, and their own standing

within the regime, both of which pull them toward maintaining greater
collegiality.
In contrast, the personalization of party-based dictatorships increases their
vulnerability to breakdown. As we show later, personalization undermines the
ruling party’s ability to cope with succession. Many of the strategies dictators
use to concentrate power in their own hands reduce prospects for regime
survival after their deaths.
Overall, these ﬁndings suggest that personalism erodes the institutional
framework – regardless of whether the framework is provided by inherited
parties or military institutions – through which elites bargain over power with
the dictator. How personalism inﬂuences regime survival differs in these two
contexts because, in the absence of personalized power, the dispersal of arms
and norms of obedience inherent in military institutions tend to shorten regime
duration, while cohesive party institutions stabilize and lengthen dictatorships
by improving the elite’s ability to handle leadership succession.

leadership changes and regime breakdown

In the full sample of cases, nearly half (47.5 percent) of all leader exits coincide
with regime collapse, which suggests that leadership transitions pose serious
challenges for autocratic stability.15 Here, we investigate how the personaliza-
tion of power affects the likelihood of regime survival in the immediate after-
math of a dictator death or ouster.
Because of the threat to regime survival caused by elite disunity, whenever
some members of the inner circle organize to remove a dictator, they risk not
only their lives and livelihoods if they fail, but also the survival of the regime
regardless of whether they fail. Those whose posts and beneﬁts depend on the
threatened dictator may mobilize against the challengers, especially if they
expect to be purged along with the dictator – as in many cases they do. Even
if the dictator is successfully removed, a deep split within the regime elite makes
the new dominant faction vulnerable to power grabs by other insiders ambi-
tious to concentrate resources in their own hands, as well as to external
opposition.
The Dominican Republic’s experience after Rafael Trujillo’s assassination
illustrates what tends to happen. Trujillo’s right-hand man, the puppet civilian
president Joaquín Balaguer, succeeded Trujillo smoothly, but a three-sided
power struggle began immediately among what had been the main beneﬁciaries
and supporters of the Trujillo regime: the military, Trujillo’s initial base of
support, which saw itself as his natural heir; Trujillo’s extended family, which
had gained control of most of the Dominican economy under Trujillo and

15
We restrict our analysis to regime leaders who held power on January 1 and thus exclude most
leaders who held power for less than a year.

wanted one of their own to lead the regime to safeguard their interests; and
Balaguer, who led the ruling party and controlled government administration.
These three groups failed to achieve a power-sharing agreement, and no one
group could dominate the others, leaving the dictatorship vulnerable to outside
opposition. A minority faction of the military that favored democratization
seized power and ended the dictatorship less than a year after Trujillo’s death
(Wiarda 1975, 263; Hartlyn 1998, 96).
The difficulties the Trujillo inner circle faced in trying to reconsolidate the
regime under a new leader are not unusual in personalized dictatorships.
Figure 8.6 shows how increased personalism affects the probability that the
dictator’s removal from office coincides with regime collapse. The analysis
looks at the 465 dictator exits in the data set. We report the results of four
specifications. The first simply controls for how long the regime and the dicta-
tor have lasted up to the time of the leader’s exit. The second adds calendar time
to control for world trends; and the third adds a battery of control variables
shown in other research to affect dictatorial resilience: GDP per capita, oil
rents, protest, civil conflict, inherited party, military-led, and revolutionary
party. The last specification excludes monarchies to check the suspicion that
monarchies, which tend to be long-lived and personalistic, might be driving
results. Depending on the specification, the estimate for the effect of personal-
ism shows that increasing the personalism index from its lowest to highest value
increases the probability that dictator exit coincides with regime collapse by
more than 50 percent. In short, replacing the dictator, whether because of death
or ouster, without destabilizing the regime becomes substantially less likely as
the dictator concentrates personal power in his hands.
Two different processes contribute to the result in Figure 8.6. First, as in the
post-Trujillo example, regime elites in personalist dictatorships have difficulty
cooperating to maintain the regime after the dictator is gone. This difficulty
arises from the dictator’s strategy of negotiating separately with different
support factions and the weakness of institutions within which bargaining
and policy choice formally occur. Second, personalist dictators’ strategies for
securing their own safety result in fewer ousters by regime insiders and conse-
quently more by regime outsiders, who generally seek to end the regime as well
as ousting the dictator. In the next section we focus on a smaller sample of cases
that excludes those in which the opposition intended to overthrow both the
dictator and the regime: natural deaths of dictators.

Death of the Dictator and Regime Survival

Unlike other kinds of leadership change, a dictator’s natural death in office is an
exogenous challenge to the regime he led. Regime weakness or inner-circle
conflict does not cause it, and thus cannot explain why dictatorships are
vulnerable in the wake of it. Some dictatorships can better handle a leader’s
death than others, however. In the next section, we test the argument that
personalized dictatorships have greater difficulties surviving succession than

Base model
Leader Add period FE
duration Add covariates*
Drop monarchies

Regime
duration

Personalism

–0.2 0 0.2 0.4 0.6 0.8

Coefficient estimate
figure 8.6 Probability that dictator exit coincides with regime collapse.
Note: *Added covariates: growth, GDP pc, oil rents, protest, civil conﬂict, inherited
party, military-led, revolutionary regime

those with more collegial leadership by investigating what happens after the
natural death of a dictator. We show that concentration of power in the
dictator’s hands undermines the dictatorial elite’s capacity to retain its hold
on power after the dictator’s natural death.
To evaluate the effect of power concentration on the capacity of dictator-
ships to handle succession after a leader’s death, we examine the relationship
between personalism, measured as its average level during the three years prior
to the first dictator’s death, and the square root of the number of years the
regime survived after it.16 The data include all forty regimes in which the first
dictator died a natural death in office. The left panel of Figure 8.7 plots these
two variables against each other to reveal a strong negative relationship:
regimes in which the first dictator had concentrated more power collapse much
sooner after his death than regimes in which the first leader accumulated less
personal power. The right panel shows the same relationship after conditioning
on a number of other factors: calendar time, regime duration up to the time of

16
The post-death regime duration variable is skewed, so we use the square root of duration in this
analysis. Tests for normality indicate that the square root transformation results in a less skewed
distribution than the untransformed number of years or the log. For regimes with leader death
before the regime’s third anniversary, we take the average level of personalism during the tenure
of the leader. The right panel of Figure 8.7 is an added-variable plot from a kernel regression.

Unconditional Conditional on covariates

8 4

e(Post-death regime duration (sqrt))

Post-death regime duration (sqrt)

6 2

4 0

2 –2

0 –4
0 0.2 0.4 0.6 0.8 1 –0.5 –0.25 0 0.25 0.5
Personalism index e(Personalism index)

figure 8.7 Effect of personalism on capacity to handle succession.

the first leader’s death, whether the first dictator came from the military, and
the average oil rents accruing to the regime during the first leader’s time
in power.
Next, we extend the analysis to all 61 dictators whose tenure ended in
natural death. This group includes the 40 first regime leaders in the prior
analysis as well as 21 subsequent leaders who died natural deaths. We examine
how personalism during the time just before the dictator’s death influences later
regime stability in a standard empirical model of regime failure. This analysis
differs from prior analyses in that we look only at the 52 regimes in which at
least one dictator died naturally while in office, and we examine only how long
regimes last after leaders’ deaths.
Since we examine only regimes in which a dictator left office as a result of
natural death, we need not worry that the death reflects strategic efforts to
constrain or oust the leader in order to preserve or destabilize the regime. The
design excludes cases such as the reshuffling coup that ousted the senile
Tunisian leader in 1987 or the rebellion that forced out the cancer-ridden leader
of Zaire in 1997. In cases like these, we do not observe natural death in office
precisely because actors who saw a leader nearing death removed him in order
to preserve the regime in Tunisia and to end it in Zaire. By examining regime
duration post–natural death, our design looks only at the cases in which
the level of personalism (during the years just before the death) is plausibly
exogenous since new leaders cannot alter what happened in the past (Jones and
Olken 2005).
Mimicking a survival model, we test a binary cross-section time series model
with regime failure as the dependent variable and log regime duration to
account for duration dependence (duration after leader death). The explanatory

Duration after
leader death (log)

Duration at
leader death (log)

Personalism

GDP pc (log)

Oil rents (log)

Military-led

Inherited party

–0.5 0 0.5 1 1.5

Coefficient estimate

Base model Add GDP, oil Add military, party

figure 8.8 Personalism and post-death regime survival.

variable of interest is the average level of personalism in the regime during the
three years before the dictator’s natural death, as above. The covariates in the
base model include decade dummies and the log number of years the regime
had been in power at the time of the leader’s death (duration at leader death).
Decade dummies account for global temporal patterns in the rise and fall of
dictatorships, while duration at leader death accounts for the possibility that
regimes that have existed for a long time before the dictator dies may be more
resilient than young ones.
The top estimates in each cluster in Figure 8.8, depicted as diamonds, show
the results from this test: personalism during the time just before a dictator’s
death increases the likelihood of regime collapse afterward. Next, we add two
structural covariates to the model, GDP per capita and oil rents. The result for
personalism remains. Last, we add two measures of exogenous institutions
discussed throughout this chapter, indicators for military-led and rule based
on an inherited party.17 Again, the result for personalism persists. Military-led
regimes are more vulnerable to breakdown in this sample as in other analyses.

17
Results in the replication ﬁles show that the estimated effect of personalism varies (i.e., there is a
nonproportional hazard): as regimes survive longer after the natural death of a leader, the
marginal effect of his pre-death level of personalism declines. Thus, personalism has the strongest
effect soon after the natural death of a dictator, as intuition would suggest. If readers want to

The ﬁndings in Figures 8.7 and 8.8 are consistent with the claim that
personalism is destabilizing during successions. Importantly, the measure of
personalism used in these tests is exogenous to the behavior of dictators who
follow after the death of a previous leader. Instead of capturing the potentially
strategic behavior of post-death leaders, the personalism variable used here is
constructed using only information about the dead leader’s behavior while in
ofﬁce. It is thus a proxy for the institutional environment bequeathed to new
dictators after the natural death of their predecessors.
Up to this point, we have investigated some of the causes of dictatorial
collapse. We turn now to what happens after dictatorships fall.

the dictator’s future and the likelihood

of democratization
What happens after a dictatorship falls depends to a considerable extent on how
the dictator and his closest allies respond to angry citizens, embittered ofﬁcers,
and demanding foreigners when they face regime-threatening challenges. In this
section, we investigate one of the factors that determines those responses: the
expectations of dictators and their closest supporters about what will happen to
them and their families if the regime falls, that is, the right-hand side of Equation
8.1. The costs of losing power vary across dictatorships and across individuals
within dictatorial elites. These differences affect the willingness of dictators and
other members of the dictatorial elite to negotiate peaceful transitions when
regime survival appears doubtful. The willingness to negotiate in turn affects
how dictatorships end and what kind of political system follows them.
Negotiation to end a dictatorship empowers parts of the opposition committed
to democratization and aims to devise a peaceful means of choosing the speciﬁc
individuals to whom power will be transferred. For these reasons, negotiated
regime transitions tend to end in competitive elections and to result in democracy.
By contrast, the dictator’s refusal to negotiate increases the likelihood that regime
opponents will resort to force to unseat him. Forced ousters, in turn, reduce
prospects for democratization (Geddes, Wright, and Frantz 2014).

The Costs of Losing Power

For dictators who have concentrated great power in their hands, losing ofﬁce
can be personally disastrous. Earlier research using regime-type data shows that
dictators who concentrate more power in their hands while they rule face a
higher probability than other ex-dictators of exile, imprisonment, execution,

interpret this substantively, the pattern suggests that dictatorial elites may be able to consolidate
against the destabilizing effect of past personalism over time, if they can survive the ﬁrst year
or two.

and assassination after ouster, even if democratization follows their overthrow

(Geddes, Wright, and Frantz 2014). In the abstract terms of Equation 8.1, the
personalization of power causes the dictator’s expectations about future bene-
fits under a different regime to drop very low. Because of their greater chance of
a bad fate after stepping down, dictators who have amassed great personal
power tend to resist negotiations with the opposition that might lead to peace-
ful transition and instead cling to power until violently overthrown (Geddes,
Wright, and Frantz 2014).
Characteristics of the seizure group also affect the costs of losing office for
the dictator and his close allies. We turn now to an examination of these costs
and how personalism interacts with these traits to influence the chance of
democratization.

Costs of Losing Power in Military-Led Dictatorships

The cost of returning to the barracks for most military officers is low if the
regime democratizes. They simply continue their chosen careers. Sometimes the
dictator himself and a handful of other top officers are forced into exile or
prosecuted for human rights abuses, but such punishments have rarely
extended to the rest of the officer corps after a peaceful transition to democracy.
Violent overthrow can lead to much higher costs for officers if insurgents,
foreign invaders, or a rebellious faction of the military defeats them, especially
if the violent overthrow results in a new dictatorship. After violent overthrows,
top officers of the ousted dictatorship are often jailed or exiled and sometimes
executed; insurgents may replace the entire officer corps. Ghana’s multiple
experiences with military rule fit this pattern. Ghana has experienced three
negotiated democratizations after military interventions; the most serious pun-
ishment imposed on outgoing officers after democratization was the forced
retirement of a few at the top. In contrast, after Ghana’s one experience with
the violent ouster of a military regime, the mutiny led by Flight Lieutenant J. J.
Rawlings in 1979, a number of officers were executed (Singh 2014). Recent
events in the Middle East show the same pattern. After the 2012 democratiza-
tion in Egypt, the old dictator was arrested for corruption and a few top officers
were forced to retire. The rest of the military remained intact and, indeed, able
to seize power again a year later. In Libya, however, where the old dictator was
violently defeated by a combination of insurgency and foreign intervention, the
dictator was killed and his army disbanded.
Officers who have served in military governments should thus prefer democ-
ratization if they fear regime breakdown, since their prospects for punishment
are higher in a future dictatorship than in democracy (Geddes, Wright, and
Frantz 2014; Debs 2016). If they prefer democracy for these reasons (or others),
they should negotiate their extrication when p, the likelihood of regime sur-
vival, has fallen. The incentives facing the military dictator, however, can differ
from those facing other officers. The dictator faces a higher risk of punishment

than do others, so he may try to cling to power when other officers want to
negotiate extrication.
As a result of these differences in interests, collegial military-led regimes, in
which other officers can constrain the dictator’s choices or remove him if he
makes choices they oppose, are more likely to negotiate a return to the bar-
racks. When the dictator resists a return to the barracks, other officers can force
him from office. The replacement of a military dictator by a faction intent on
democratization has occurred many times since 1946. Hard-line Argentine
General Galtieri’s replacement by the moderate General Bignone in 1982 is
one example. After their humiliating defeat in the Malvinas/Falklands conflict
and massive protests against military rule, the military junta replaced Galtieri
with Bignone, oversaw competitive elections in 1983, and rushed a transition to
civilian government. Other officers also forced Colombian General Rojas
Pinilla to resign after a popular uprising, so that his replacement could oversee
elections and a transition to democracy (Martz 1962, 249–53).
In these examples and many others, a military faction responded to wide-
spread popular opposition to military rule by ousting the dictator clinging to
power, and then organized a transition to democracy via competitive elections.
These are cases in which other officers imposed their own interests on military
dictators who were trying to avoid the potential costs of losing office – and did
so through means visible to observers.
In contrast, military-led regimes in which the dictator has concentrated great
power in his own hands and can therefore act to further his individual interests
often refuse to negotiate. Further, when they do negotiate, they may later renege
on power-sharing agreements. Examples include Colonel Mobutu Sese Seko in
Zaire/Congo, who agreed to power sharing with the opposition in 1992, but
retained control of the army and ruling party, which enabled him to break his
promises and reconsolidate his position as dictator (Schatzberg 1997, 70).
Similarly, ex-Sargent Gnassingbé Eyadéma in Togo “surrendered power” in
1991 to the interim prime minister who had been selected by a national confer-
ence,18 but was able to use the military to wrestle control back into his own
hands by the end of the year (Press 1991). Dictators who have concentrated great
power in their hands tend to resist losing office to the bitter end. Insurgents
forced Mobutu from office five years after he broke his promises about power
sharing. Eyadéma remained in power until he died of natural causes more than
ten years after breaking his agreement with the opposition. He was succeeded by
his son, who still rules as this is written.
In other words, in military-led dictatorships in which one officer has concen-
trated immense discretion in his hands, other officers have lost the ability to
pursue their own interests in a peaceful return to the barracks if regime survival
is threatened. In more collegial military regimes, however, the interests of

18
“Togo’s President Agrees to Yield Power to a Rival,” New York Times, August 28, 1991.

officers other than the dictator tend to prevail, which increases the likelihood of
a negotiated transition and democratization.
Earlier research confirms that collegial military regimes are more likely to be
followed by democracy than are regimes led by officers who have concentrated
greater power in their hands (Geddes, Wright, and Frantz 2014). This relation-
ship could be spurious, however. If more professionalized military institutions
can better obstruct the personalization of rule, and richer countries tend to have
more professionalized military institutions, then wealth may be the underlying
reason for the correlation between collegial military rule and democratization.
To rule out this possibility, we conduct more rigorous tests of this relationship
below (see Figure 8.9).

The Costs of Losing Power in Party-Led Dictatorships

The minimum cost of ending dictatorship for members of a party-based dicta-
torial inner circle is that they must compete with other parties for the benefits
associated with rule rather than having a monopoly. If they lose elections, they
lose automatic access to state resources for personal consumption or use in
election campaigns, and they lose opportunities for corruption and business
advantages associated with their political connections. Party cadres may also
lose their government jobs. So the losses can be substantial, but loss of office
does not usually lead to jail or execution except sometimes for top leaders.
As with military regimes, former members of the dictatorial government are
better off under a subsequent democracy than under a new dictatorship
(Geddes, Wright, and Frantz 2014). Members of the inner circle may therefore
want to increase the chance of a peaceful, nonviolent regime change by negoti-
ating with the opposition and agreeing to reforms that increase the fairness of
elections, but dictators, who face the possibility of jail or assassination if they
step down, often refuse to negotiate. In contrast to the many coups that have
ushered in democratic transitions when military dictators opposed comprom-
ise, however, civilian members of the elite surrounding a dictator who refuses to
negotiate have more difficulty ousting him in order to oversee a peaceful
transition. As a result, democratization is less likely to follow party-led
dictatorships than those led by military officers.
The dictator’s concentration of power in party-led dictatorships further
reduces the likelihood of democratization. Officials of personalized party-led
dictatorships face less difficult futures than do top leaders but are more likely to
be politically marginalized and deprived of economic assets after democratiza-
tion than are officials from dictatorships with more collegial decision-making.
As Michael Bratton and Nicolas van de Walle note:
Recruited and sustained with material inducements, lacking an independent political
base, and thoroughly compromised in the regime’s corruption, they are dependent on
survival of the incumbent. Insiders have typically risen through ranks of political service

and, apart from top leaders who may have invested in private capital holdings, derive
livelihood principally from state or party ofﬁces. Because they face the prospect of losing
all visible means of support in a political transition, they have little option but to cling to
the regime, to sink or swim with it. (1997, 86)
This difference in future prospects between ofﬁcials of regimes in which the
dictator has concentrated great power and those in which a party organization
maintains a degree of power dispersal arises from the dictator’s control of
appointments in the former. The personalistic dictator’s exile, imprisonment,
or execution of the most able and popular politicians in the ruling party, while
at the same time making appointments based on loyalty alone, reduces the
likelihood that the party will be able to transform itself into a successful
competitor if democracy succeeds the dictatorship.
These considerations mean that party-led dictatorships are less likely to
negotiate transitions than collegial military-led dictatorships, especially if
power is concentrated in the hands of the dictator.

Costs of Losing Power for Monarchs

Twelve monarchs have lost dictatorial control since 1946, three of them in
Nepal.19 Coups ended seven of the monarchies, popular uprisings three, and
insurgency one. Seven authoritarian monarchies remain in countries with more
than a million inhabitants, all but one in the Middle East or North Africa. In
only one instance was a transition to constitutional monarchy and competitive
elections negotiated: in Nepal in 1991. In this instance, the constitutional
monarchy survived until a later king usurped unconstitutional powers in
2002. The Nepali monarchy was abolished a few years later after a popular
uprising. Of the monarchs ousted by coup, popular uprising, or insurgency, one
was murdered along with his family during the coup, one died in prison
afterward along with several family members, and the others were exiled. Much
of their property was conﬁscated. None survived as constitutional monarchs in
democracies.
In short, since 1946 monarchs ousted by force have not enjoyed a peaceful,
economically secure retirement in their own countries, and other members of
their families have also faced punishments and exile. Since these numbers are
small, conclusions have to be tentative, but these experiences suggest that
monarchs and their families have a lot to lose if they are forcibly ousted. It
might seem that a ruling family’s best strategy in the face of opposition or
demands for democracy would be to pursue gradual democratization (that is,
steps toward constitutional monarchy as practiced in Europe), which might
safeguard their lives and quite a bit of their status and property.

19
This number does not include a few monarchs who brieﬂy held formal power but never actually
ruled, such as King Michael of Romania during Soviet occupation.

The majority of the monarchies ousted by coup, however, appeared to

be following exactly that strategy. They held regular semi-competitive
parliamentary elections. Elites controlled these parliaments, as occurred before
full democratization in Europe, and monarchs controlled substantial areas of
policy. These regimes might have gradually become fully constitutionalized and
democratic over time as some European monarchies did, but they were over-
thrown before that could happen. These experiences suggest that the strategy
that worked for a number of European monarchies may not be available to
contemporary monarchs.
In contrast to much of Europe when parliamentary supremacy was being
imposed on monarchies, nearly all countries that achieved independence after
World War II (which includes most contemporary authoritarian monarchies)
created standing armies more or less immediately. Coups carried out by small,
recently created armies occurred within ten years of independence in a quarter
of countries that gained independence after World War II, including a number
of monarchies. Once a country has a professional army (that is, one not raised
by tribal levies or maintained by regional aristocrats) and an officer corps open
to the middle class, monarchs apparently become susceptible to dissatisfied
officers just as other rulers are. This may have reduced the feasibility of
incremental democratization strategies for monarchs. It should perhaps not
be surprising, then, that most contemporary monarchies have opted to rely
on repression and cooptation through distribution rather than steps toward
democratization when faced with opposition demands.
To sum up our argument about prospects for democratization, the costs of
losing power for dictators and their closest collaborators influence their willing-
ness to negotiate stepping down once the dictatorship’s prospects look dim. When
dictators negotiate an exit, authoritarian breakdown usually results in democra-
tization. When the dictatorial elite circles the wagons, however, and fights until
the end, the fall of one dictatorship is likely to coincide with the beginning of a
new one. In the next section, we test some of the implications of this argument.
We investigate, first, the effect of the personalization of dictatorial rule on pro-
spects for both democratization and a peaceful transition. Second, we show that
the effect of personalization on the likelihood of democratization varies depending
on whether dictatorial elites come from the military or a ruling party.20

the effect of personalization on prospects

for democracy
Figure 8.9 shows how the personalization of power inﬂuences the chance of
democratization and a peaceful transition. For these tests, we estimate kernel
regression models, including common covariates of democratization: GDP per

20
We cannot test claims about monarchies because their number is too small.

Democratization Armed collapse

0.3
0.3
Change in probability of democratization

Change in probability of armed collapse

0.2
0.2

0.1
0.1

0
0

–0.1 –0.1

–0.2 –0.2

–0.3 –0.3

–0.4 –0.4
1950 1970 1990 2010 1950 1970 1990 2010

Year Year

figure 8.9 Personalism and democratization.

capita, regime duration, a Cold War dummy, civil conﬂict, prior democracy, and
regime institutions (military-led, inherited party, personalism). The left panel
shows that personalism lowers the chances of democratization during the entire
sample period, conditional on old regime breakdown.21 The estimates on the
vertical axis correspond to the difference in the probability of democratic transi-
tion (given old regime collapse) between dictatorships with collegial leadership
(0 on the personalism index) and those in which the dictator has concentrated
power (1 on the personalism index). While not shown, the control variables
yield intuitive results as well. Wealthier dictatorships, military-led regimes, and
those that were preceded by democracy are more likely to democratize. New
autocracies were more likely to replace dictatorships that collapsed during the
Cold War than those that have broken down more recently. The substantive
effect of personalism is larger than that of any covariates except the Cold War.22
These results suggest that concentration of power in dictators’ hands impedes
democratization in the wake of authoritarian breakdown.
One of the mechanisms through which we believe personalism affects the
likelihood of democratization is that dictators who have concentrated power in
their hands resist negotiating peaceful transitions, which results in more of
those that eventually do fall being ousted by force. In the right panel of
Figure 8.9, we show the results of an investigation of this claim. The dependent
variable in the right panel is forcible regime overthrow (conditional on the
occurrence of regime collapse). Personalism increases the likelihood that the

21
The plot depicts the nonlinear ﬁt for the point-wise derivatives for personalism, over time.
22
The standardized coefﬁcient from a linear model is also larger than that of any other variable
except time period. Adding personalism to a logit model increases the area under the curve from
0.809 to 0.835.

dictatorship ends in violence. Paralleling the results for democratic transition in

the left panel, model results also suggest that dictatorships in wealthier coun-
tries, military-led autocracies, and those with a prior history of democracy are
less likely to end violently (and thus more likely to have negotiated transitions).
Dictatorships that collapsed before 1990 were more likely than those that have
ended since then to hang on until forced out by armed opponents.
In sum, personalist dictators tend to resist negotiating transitions, possibly
because they face high risks of post-exit punishments. Their lieutenants also
have more to lose from regime breakdown than do high-ranking supporters in
more collegial dictatorships. And of course, it is more dangerous for members
of a personalist dictator’s inner circle to try to oust the dictator or negotiate
with the opposition behind the dictator’s back. These differences in the costs of
regime breakdown to high officials of personalized dictatorships help explain
why personalist rulers fight to hold on to power even when the fight looks fairly
hopeless – as in Syria after 2011 – and why, in turn, their opponents often use
violence to try to force them out. Recall that contemporary Syria scores almost
as high on our measure of personalism (shown in Figure 4.3) as North Korea.

Personalism in Military-Led Regimes

Past research has shown that military regimes are more likely than other
dictatorships to end with democratization (Geddes 2003; Debs 2016; Kim
and Kroeger 2017). In this section, we investigate the effect of personalism
within this subset of dictatorships by adding an interaction term between
personalism and military-led rule to a model of democratic transition.23 We
report the results from a series of linear models with controls for duration
dependence in Figure 8.10. First, we test a specification that includes only an
indicator variable for military-led regimes, the measure of personalism, and an
interaction between the two. Next, we test a specification with country and time
period fixed effects, and finally a specification with an assortment of control
variables: economic growth, GDP per capita, conflict, oil rents, prior democ-
racy, and revolutionary regime, as coded by Levitsky and Way.
In all specifications the estimate for military-led regimes is positive and
significant, indicating that collegial – or less personalized – military regimes
are more likely to democratize than similarly collegial civilian-led dictatorships.
This finding confirms earlier research (see especially Kim and Kroeger 2017).
The estimate for the interaction term between military-led and personalism is,
as expected, negative and significant: as power becomes more concentrated
in a military dictator’s hands, democratization becomes less likely. Highly

23
Recall that military-led is measured before the seizure of power and is thus exogenous to the
dictator’s behavior once in ofﬁce.

Military-led

Personalism

Pers. × Military

Civil conflict

Int'l conflict

Prior democracy

GDP pc

Growth

Oil rents

Rev. regime

–0.1 –0.05 0 0.05 0.1

Coefficient estimate

No FE FE FE + covariates

figure 8.10 Personalism, military rule, and democratization.

personalized military-led regimes are thus no more likely to democratize than

civilian dictatorships.
To conclude this section, on average, military-led dictatorships are more
likely than those led by civilians or monarchs to negotiate exits from power and
thus more likely to end in democratization. When a military dictator has
concentrated great power in his hands, however, other ofﬁcers cannot impose
their preferences on the dictator. Consequently, he tends to resist negotiation
with the opposition and, if he is ultimately ousted, to be overthrown by force by
a group that establishes a new dictatorship.

conclusion
Most dictatorial regimes end when coups replace them, incumbents agree to
fairer elections and lose, a popular uprising forces incumbents to resign, or an
insurgency defeats them in battle. Each of these events means that important
political actors not only oppose the dictatorship but consider it worthwhile to
take potentially dangerous and costly public action against it. Latent opposition
is widespread in many dictatorships, but most of the time opponents have
strong reasons not to plot or engage in public expressions of discontent. In this

chapter, we investigated some of the factors that can transform latent oppos-
ition into overt political activity and some characteristics of the dictatorial elite
that increase its vulnerability to ouster.
The chapter began with a schematized description of the incentives facing
dictatorial elites and others as they contemplate opposition. The schematization
shows the interplay among the benefits individuals receive from the dictator-
ship, their expectations about future benefits under a different regime, their
perceptions about the likelihood that the current regime will fall, and the cost of
overt opposition. This abstract version of the situation facing political actors in
dictatorships helps situate different real-world events and crises in relation both
to each other and to the likelihood of regime change. The remainder of the
chapter describes how various events and regime characteristics can change the
decision calculus of individuals and thus the likelihood of authoritarian break-
down and what follows it.
Events such as economic crises and natural disasters typically reduce benefits
for both citizens and elites. All else equal, a decrease in benefits would increase
public opposition and the likelihood of regime collapse. The dictatorial elite
may, however, be able to distribute the costs of crisis to shield the individuals
most able to threaten regime survival, and may also deliver help effectively.
When dictatorships can do these things, they tend to survive. They have greater
ability to respond to crises effectively if leaders have previously built patron–
client networks that reach ordinary people. An effective response is also more
likely if sufficient discipline has been enforced within the ruling group to
prevent the theft of relief supplies and other benefits meant for people afflicted
by the crisis.
Our analysis shows that dictatorships led by parties that began as insurgent
or electoral organizations in the political systems that pre-dated their seizure of
dictatorial power survive longer than dictatorships in which regime elites
created parties after achieving power or remained unstructured by a party.
Importantly, dictatorships led by inherited parties are less affected by economic
crises than other authoritarian governments. We interpret this finding as mean-
ing that patron–client networks that encompass a substantial part of the
citizenry help to perpetuate dictatorships, not only by including more people
in routine distribution during normal times, but also by maintaining the organ-
izational resources needed to manage crises.
Not all challenges to dictatorial survival arise in the external world. Conflict
within the inner circle of dictatorships can also precipitate regime breakdown.
Our investigation shows that power concentration in the dictator’s hands
increases the durability of military-led dictatorships, which otherwise tend to
be relatively short. In collegial military regimes, wide dispersal of arms results in
frequent inner-circle conflicts over leadership, which can destabilize the regime.
The concentration of power in the dictator’s hands limits the ability of other
officers to overthrow him, and thus increases regime stability during the dicta-
tor’s lifetime, as well as leader security.

Civilian-led regimes experience less leadership conﬂict on average. The

ouster of a dictator by a civilian ruling group requires coordinated action by
most of the inner circle (Svolik 2012), which is harder to organize than a coup.
This is one reason civilian-led regimes tend to survive longer. More collegial
dictatorships led by inherited parties tend to be even more stable. They rely on
the ruling party’s executive committee for high-level decision-making, so policy
choices tend to have wide backing from the inner circle, reducing conflict.
Members of the inner circle also have a lot of experience bargaining with,
and a lot of knowledge about, one another. These characteristics and experi-
ences contribute to orderly successions and thus, on average, to highly durable
dictatorships.
In contrast to the orderly successions characteristic of relatively collegial
dictatorships led by inherited parties, the death of a dictator who has concen-
trated great power in his hands often leads to regime crisis. Dictatorships are
less likely to fall in the aftermath of the dictator’s natural death than after a
coup or other violence ouster, but much more likely to fall than during periods
of stable leadership (Kendall-Taylor and Frantz 2016). Besides the conflict over
who will succeed the old dictator, succession usually results in a period of
uncertainty and struggle over what norms will govern elite bargaining under
the new dictator. The high stakes and extreme dangers involved in choosing the
next dictator increase the likelihood of intra-elite conflict, and thus the chance
of both regime collapse and the purge of parts of the ruling group even if the
regime survives.
Besides affecting how long dictatorships survive, the concentration of power
in one man’s hands also affects the likelihood of a peaceful, negotiated transi-
tion, once survival appears unlikely. Dictatorships with more collegial decision-
making tend to negotiate their extrication from power when they doubt their
ability to hang on to it, but personalized dictatorships often fight to the bitter
end. Two factors contribute to this difference. First, dictators who have concen-
trated great power in their hands have more reason to fear jail, execution, or
assassination after ouster than do more consultative dictators, regardless of
whether they are officers or civilians. Second, dictators with more power
concentrated in their hands can exclude from the inner circle, arrest, or kill
allies who want to negotiate with the opposition, and in this way retain a
monopoly over decisions about how to respond to challenging situations. In
more collegial regimes, in contrast, dictators have to bargain with other
members of the inner circle over whether and how to negotiate. Dictators’
top supporters generally have less to worry about post-exit than dictators
themselves (Albertus and Menaldo 2014), so they tend to favor negotiation
when they fear regime collapse.
In regimes led by a military junta rather than a strongman, a dictator who
refuses to negotiate when the rest of the junta favors it is likely to be ousted and
replaced by another officer who favors a return to the barracks. The ouster of a

hard-line dictator by a faction that favors negotiated transition has happened

many times near the end of military-led regimes.
Members of ousted dictatorial elites are usually better off under democracy
than in hostile new dictatorships. Consequently, when they negotiate stepping
down, they rarely resist democratization.
These differences among members of dictatorial elites in the costs and risks
of transition result in a tendency for junta-led dictatorships to negotiate peace-
ful transitions to democracy via fair elections to choose new civilian rulers.
Civilian-led dictatorships with collegial leadership are less likely to exit peace-
fully via elections than junta-led regimes, but more likely to do so than civilian
regimes in which the dictator has concentrated great power in his hands. Where
a transition occurs peacefully via negotiation and elections, the immediate
outcome is usually democracy. Where, however, the dictator retains his steely
grip until forceful overthrow, democracy is less likely to follow. Ouster by
coup, insurgency, or foreign invasion leads to control by the leaders of the
group that led the armed overthrow. Such leaders nearly always promise
democratization, but deliver on their promises only some of the time.

Conclusion and Policy Implications

Between his accession to full power in 1979 and the US invasion of Iraq in
2003, Saddam Hussein concentrated immense discretion over policy, person-
nel, and life itself in his own hands, undermining the Iraqi military’s profes-
sionalism and ability to fight in the process, and the capacity of the Ba’th Party
to run a government. His drive to concentrate power began immediately. Less
than two weeks after replacing the previous dictator, Saddam announced the
discovery of a plot involving top regime officials; a special party court met,
judged twenty-two guilty, and sentenced them to execution. Those executed
included five members of the dictatorship’s sixteen-man inner circle, the Revo-
lutionary Command Council (RCC). Saddam and other remaining members of
the RCC carried out the executions in person. Up to 500 other party members
were executed along with numerous military officers (Amnesty International
1980, 331, 336; Farouk-Sluglett and Sluglett 1987, 209–10; Tripp 2007, 214).
Needless to say, when the dictator can murder other members of the inner
circle, they have little ability to constrain his decision-making or enforce power
sharing.
Before taking the top post, Saddam had already gained control of the
regime’s extensive internal security forces and overseen earlier purges of the
military and party. Nevertheless, at each subsequent crisis, Saddam purged
more people in order to further narrow decision-making circles and concentrate
ever more control in his hands. After terrible losses during the Iran-Iraq War in
the early 1980s, Saddam reshuffled the RCC, the Ba’th Party Regional Com-
mand (the party’s executive committee), and the cabinet to eliminate everyone
except his relatives, closest allies, and protégés. He appointed close relatives,
including sons, sons-in-law, and several half-brothers, to key security and
ministerial posts (Brooker 1997, 119; Tripp 2007, 228, 244).
As soon as the war ended in 1989, Saddam again turned his attention to
reducing potential threats from the officer corps – which he had had to rebuild

218

to prevent defeat by Iran – through retirements, demotions, “accidents,” and

arrests (Tripp 2007, 240–45). In 1995, the Economist explained why the army
was unlikely to oust Saddam: “the army is demoralized, barely a serious
fighting force; its senior officers have been switched, fired, executed or so tarred
with Mr. Hussein’s brush that they have no future outside his orbit.”1 Even
relatives had become unsafe by the mid-1990s, when two sons-in-law (high-
ranking officers holding very powerful posts) were murdered along with many
members of their families.
In short, Iraq under Saddam Hussein was an extreme example of what we
have labeled personalist dictatorship. Our analysis in Chapter 8 of how per-
sonalized dictatorial rule influences the likelihood of democratization implies
that if foreigners intervene militarily to remove a personalist dictator like
Saddam Hussein or Moammar Qaddafi of Libya, the intervention is unlikely
to result in democracy. Not only do new dictatorships often follow the violent
ouster of personalist regimes, but civil war and state disintegration follow these
kinds of authoritarian breakdown more often than other kinds (Geddes,
Wright, and Frantz 2014).

If the goal is democratization, our analysis suggests that violent foreign

intervention to overthrow dictators similar to Saddam Hussein has only a
modest chance of success. In the remainder of this conclusion, we summarize
our ﬁndings chapter by chapter and then discuss the policy implications (in
boxes) that follow from our results. Though no one can know for certain what
consequences a policy choice will have since many factors contribute to ﬁnal
outcomes, the evidence we provide in this study enables a probabilistic predic-
tion. We think that the academic and policy communities should seek to
develop “evidence-based” policy guidelines modeled on those that inform many
decisions in medicine, and we see our study as a step toward doing that.
Chapter 2 focuses on the initiation of dictatorship and the situation that
faces new autocrats immediately after the seizure of power. Most post-1946
dictatorial seizures of power replace other dictatorships, but about 30 percent
replace democracies. The vast majority occur when many citizens are fed up
with the incumbent, whether democratic or not. Those who overthrow govern-
ments often promise democracy and other desirable goals such as growth and
an end to corruption, but they rarely deliver.
Military factions or political parties initiate most dictatorships. The former
usually seize power via coups. The latter most often either authoritarianize a
democratic government they already lead or seize control via armed rebellion.
During the whole post–World War II period, foreigners installed as many
dictatorships as did homegrown insurgencies, but foreign imposition of dicta-
torship has become less frequent during recent decades. Since the end of the

1
“Saddam Sacks a Henchman,” Economist, July 22, 1995, 46.

Cold War, authoritarianization has grown more common, while military coups
and foreign interventions have decreased.
Those who oust governments rarely have detailed plans about what they will
do once they have control. This is especially true for forcible seizures of power
such as coups and insurgencies. The need for secrecy before coups prevents
wide consultation about policy during the plotting stage, and the lack of
government experience of most coup and insurgent leaders limits their ability
to foresee the decisions they will face once in office.
Consequently, most dictatorial seizures of power are followed by a period of
chaos, uncertainty, and conflict within the ruling group. They argue over what
to do, who will lead, and how much power the new dictator will have relative
to others in the dictatorial elite. This situation makes it hard for foreign policy
makers to figure out a best response. Even participants may not be able to
predict how the new dictatorship will shape up and what policies it will follow.
Because poor incumbent performance partly motivates dictatorial seizures of
power, and plotters strategically time them to coincide with widespread popu-
lar disenchantment, the new elite has to grapple immediately with the serious
problems that undermined the old regime. New dictatorial elites often take
power during economic crises, and their first order of business may be the
urgent search for foreign aid. The policy implications that flow from Chapter 2
are outlined in the box below.

The difﬁculties and uncertainty new dictatorships face mean that

international actors can exercise more inﬂuence immediately after a
seizure of power than later, after the seizure group consolidates its
control. Early on, strategies are unsettled, but later, vested interests will
have developed. In cases where the leadership position remains contested
after the seizure, foreign policy makers may be able to disadvantage
hard-line or extreme factions by promising a more positive future
relationship should the moderate faction secure control.
International actors may even persuade members of the seizure group
to hand power to a neutral political actor instead of holding onto it. This
kind of strategy seems to have guided recent international responses to
some coups. During the last two decades, members of the international
community have been quick to pressure some coup makers to return
power to civilians, and some of these interventions have persuaded the
military to give up direct power. In Honduras in 2009 and Mali in 2012,
for example, ofﬁcers turned power over to civilian interim governments
after becoming convinced that aid and international recognition would
be withheld until they did so.
Members of the international community have been slower and less
united in condemning the authoritarianization of democratic politics led

by elected leaders such as Recep Tayyip Erdogan of Turkey. After a

failed coup attempt in 2016, Erdogan ordered the arrest of about 50,000
people, including much of the opposition. The following year, a
constitutional revision granted Erdogan quasi-dictatorial powers.2 It is
harder to achieve international consensus about such democratic
breakdowns because regime transition often occurs incrementally.
Knowing that the authoritarianization of previously democratic
governments has been one of the more common means of initiating
dictatorship (responsible for 16 percent of new dictatorships since World
War II), and has become more common since the end of the Cold War,
might help policy makers agree to withhold aid after explicit
authoritarianizing actions such as Erdogan’s.

Military ofﬁcers initiate most dictatorships. In Chapter 3 we investigate the

conditions associated with coups that install new dictatorships. We find, con-
trary to several popular theories, no relationship between coups and wide-
spread popular opposition to incumbents, mass political mobilization, or
inequality. Instead, our findings suggest that officers pursue their own interests
when deciding whether to intervene in politics. We find that soldiers are less
likely to seize control when they come from the same groups and share interests
and ideas with political leaders. We confirm other research showing that coups
are more likely in poor countries and that they were more frequent during the
Cold War than since it ended.

Our findings raise questions about the wisdom of using military aid to
support one side or another in the political conflicts of other countries.
Resources meant to shore up incumbents may instead contribute to the
fulfillment of officers’ unlawful political ambitions. US military and other
forms of aid played a large role in the buildup of deadly and extensive
security services in numerous Cold War allies, while the Soviet Union
and China contributed to similar buildups among their allies.
Competition between the United States and the Soviet Union also
motivated US support for coups that ousted leftists and Soviet support
for coups against conservatives. In that context, coups were frequent,
and military dictators ruled many countries.
During recent decades, most coups have faced international
disapproval and sanctions, though the al-Sisi coup in Egypt is a

2
“Turkey Is Sliding into Dictatorship,” Economist, April 15, 2017.

conspicuous exception. The sharp decline in the incidence of coups since

the end of the Cold War suggests the responsiveness of ofﬁcers to
international pressure. Our ﬁndings, along with those of others, imply
that the decisive withdrawal of foreign support after coups can reverse
military seizures of power. We believe that other potential military
interventions have been deterred by knowledge of the hostility of
international aid givers toward coups and that this is one of the reasons
for the substantial decrease in coups since 1990 (Marinov and Goemans
2014).

During the consolidation stage that follows the initiation of a new dictator-
ship, members of the new dictatorial elite bargain and fight over how to
distribute power among themselves. Struggles among members of the dictator-
ial elite during the first months and years of the dictatorship determine the
degree to which the dictator can concentrate control in his own hands – as
opposed to sharing power within a collegial inner circle.
Chapter 4 begins the explanation of why some dictators can concentrate so
much power in their hands while others cannot. After the seizure of power, the
ruling group must choose one member as leader if they have not already done
so. All members of the dictatorial elite want to maintain the dictatorship, but at
the same time, they want to increase their own power and access to resources
relative to others in the inner circle. The day he is chosen, the dictator has little
more power than his colleagues, but that can soon change because dictatorships
lack third-party enforcement of bargains and contracts. Once the man who is
delegated leadership powers has the additional resources that control of the
state gives him, other members of the dictatorial elite may find it difficult to
enforce limitations on the dictator’s discretion agreed to earlier.
The lack of binding third-party enforcement in dictatorships means that the
only way the rest of the elite can constrain the dictator and enforce consultation
during decision-making is by credibly threatening to oust him if he fails to share
power (Magaloni 2008; Svolik 2012). A unified and disciplined elite can
bargain effectively with the dictator because their unity reduces collective action
problems, and thus they can make credible threats to oust if the dictator usurps
more power than was delegated to him. A factionalized, divided, or undiscip-
lined seizure group, however, cannot create the kind of inner circle that can
bargain effectively with the dictator. Instead, the dictator can bargain with each
faction separately. Factions compete for the dictator’s favor, driving down the
price he has to pay for support. Factions also increase the collective action
problems involved in trying to replace a dictator, which reduces the credibility
of threats to oust.
Bargaining between the dictator and a factionalized inner circle can result in
narrowing of the dictatorship’s support base and the concentration of great

power and discretion in the dictator’s hands, which we label personalism. If the
dictator does not need the support of everyone who joined in the effort to
overthrow the old regime in order to hang on to power, he can exclude some
members of his initial support coalition and keep their share of spoils or
distribute them to remaining supporters. This is the reason we often see some
of the supporters of a seizure of power jettisoned from the dictatorship’s
support coalition within the ﬁrst year or two. Those thrown over the side tend
to be those most likely to disagree with the dictator over policy or to challenge
his right to rule.

Because dictators often exclude some members of the initial inner circle
and narrow their initial support coalition, international observers should
not view the inclusion of moderates or pro-democracy figures in the first
authoritarian cabinet or command council as guarantees of moderation.
Dictators strategically include such figures precisely in order to reassure
foreign lenders, investors, and aid givers while also allaying the fears of
influential domestic actors, but these individuals may be the first to go
once the dictatorship seems a little safer.

Unity and discipline tend to be higher in seizure groups whose organi-

zation long pre-dates the establishment of dictatorships. At the moment of
seizure, the military that takes power in a coup could be united and highly
disciplined or factionalized by ethnicity or competing partisan loyalties. Long-
established military forces tend to be less factionalized than recently created
ones. The party that seizes power could be a highly disciplined “organizational
weapon” or a loose coalition of parties and groups put together to compete in
the last fair election before authoritarianization. These group characteristics
develop before the seizure of power. In social science terminology, they are
exogenous to the dictatorship. Consequently, these seizure-group characteris-
tics can help to explain things that happen later during dictatorships because
they could not have been caused by decisions made by the post-seizure
dictatorial elite.
The empirical section of Chapter 4 shows that factionalism within the
seizure group before the establishment of the dictatorship is associated
with greater concentration of power in the dictator’s hands post-seizure.
We ﬁnd that ruling parties formed before the dictatorship to lead insur-
gencies or compete in elections inhibit the later personalization of dicta-
torship and that dictators from more undisciplined militaries are more
likely to concentrate power in their own hands than other military dicta-
tors. Dictators who have taken initial steps toward personalization during
the ﬁrst three years of rule tend to concentrate further power in their hands
over time.

The personalization of dictatorship has dire consequences for people

living under it, as well as their neighbors. As a number of studies have
shown, the personalization of dictatorial rule is associated with worse
governance, more erratic and aggressive international behavior, and
enhanced prospects for violence during and after regime overthrow
(Frantz and Ezrow 2011; Weeks 2012). The concentration of power
leads to erratic decision-making and international adventurism, which
results from the absence of consultation within the regime leadership and
the unwillingness of the dictator’s terrified assistants to give him accurate
information either about other countries’ likely responses or about their
own country’s capacities.
Personalist dictators replace subordinates at will, often isolating
themselves from information and advice that might have dissuaded them
from engaging in risky and ultimately costly acts of international
belligerence. Saddam Hussein, for example, consulted only his son-in-
law (a general) before the final decision to invade Kuwait. The army chief
of staff, whom Saddam had consulted earlier, had told him that an
invasion would lead to war with the United States, which Iraq would
lose. So, Saddam dismissed him (Ashton 2008, 266–69).
Given the consequences of personalist rule, we think members of the
international community concerned with development and peace should
do what they can to inhibit its emergence. An obvious implication is that
policy makers should take an especially discouraging stance toward
military dictators ranked low before the seizure of power because the
choice of a low-ranking officer to become dictator indicates a
factionalized military seizure group. Even if we believe that some
circumstances legitimate support for military coups, we should avoid
supporting post-coup dictatorships led by junior officers (such as
Captain Moammar Qaddafi or Liberia’s Sergeant Samuel Doe), who
have the potential to become some of the most violent and erratic
dictators.
The findings in Chapter 4 also imply that incremental
authoritarianizations led by parties organized during the president’s
election campaign, like the ones that backed Hugo Chávez in Venezuela
or Alberto Fujimori in Peru, are likely to turn out worse than
authoritarianizations led by longer-established parties. Policy makers
should be especially quick to respond to actions such as arrests of
opposition leaders when carried out in authoritarianizing democracies
led by recently created parties unable to constrain their leaders.

Chapter 5 highlights the effects of a second characteristic of the seizure

group that pre-dates the dictatorship and shapes bargaining between the

dictator and his supporters: the dispersal of arms across members of the group. In
this chapter, we focus on dictatorships brought to power by armed groups. Most
armed groups achieve power in coups, but some lead insurgencies or are imposed
by foreigners. In armed seizure groups, many individual commanders could
potentially oust the dictator. If the group is also unified and disciplined, then
the relative ease of ousting the dictator gives them strong bargaining power, and
they can establish a power-sharing agreement with the dictator. Governments
with this form are sometimes labeled juntas. If, however, the dictator’s armed
supporters are divided into factions, then superior officers may not be able to
make credible promises to support the dictator if he shares because they may not
be able to commit subordinates in other factions to refrain from rogue coups.
In this situation, power-sharing bargains are not credible and therefore cannot
secure the dictator’s position, so he has no incentive to share with other officers.
Instead, he must try to find different strategies for increasing his security. Among
those strategies, we suggest that the creation of a civilian support party can
counterbalance the dictator’s unstable military support base. We argue that this
strategy may be safer than creating counterbalancing paramilitary forces (though
many dictators do this too) because officers see civilians as less threatening to their
prerogatives and monopoly of force than armed groups. The creation of a civilian
support base helps defend the dictator from coups because officers try to avoid
confrontations between troops and crowds of civilians, and parties are good at
mobilizing crowds into the streets when dictators need shows of support.
Our empirical investigation supports this argument. Dictators create civilian
support parties to counterbalance and marginalize their original factionalized,
armed support base. Post-seizure party creation helps dictatorships that seized
power by force survive longer. Dictatorships that lack support parties face a
10 percent chance of breakdown each year, while similar dictatorships in which
a party was created post-seizure have less than a 5 percent chance of break-
down per year.
“Civilianization” is the term used to describe the replacement of a military
ruling council by a mostly civilian party-led ruling body under the same dicta-
tor. The dictator may formally retire from the military as well. Such systems are
usually augmented by controlled elections for a tame legislature and some form
of election to legitimate the dictator. Dictators who have civilianized their
regimes usually claim to be democratizing, and observers may be confused
about whether these are steps toward democracy. Our findings show that, on
the contrary, civilianization substantially prolongs the survival of dictatorships.

The main policy implication of Chapter 5 is thus clear: members of the

international community should not view the civilianization of military-
led regimes as an indication of impending democracy. As long as the

dictator himself remains in control, they should not reduce any costs they
have imposed on the dictatorship. Post-seizure party creation, and
civilianization more generally, are dictatorial strategies for concentrating
more power in the dictator’s hands at the expense of wider consultation
within the officer corps, while prolonging dictatorship.
A large majority of armed seizure groups are composed of military
officers, meaning that officers lead the bulk of dictatorships discussed in
Chapter 5. Several studies have shown that dictatorships led by
somewhat collegial groups of officers (juntas) end sooner than other
kinds of dictatorship. Not only are they relatively short-lived, but they
are more likely to negotiate a peaceful return to the barracks via fair,
competitive elections than are other kinds of dictatorship. In short, junta-
led military regimes are more likely to democratize after shorter periods
of rule than personalized military-led dictatorships or civilian-led
dictatorships. For international policy makers concerned about
democratization, it is thus better to encourage a return to the barracks
rather than multiparty elections in which the dictator competes. Many
dictators have found ways to win multiparty elections.

In order to extend their control beyond the apex of the political system,
dictatorial elites must find ways to monitor lower-level officials and gather
information about all parts of the country. These are difficult problems in
dictatorships. Local officials may oppose the new dictatorship, or they may
simply be motivated by self-interest. They may sabotage official policies, abuse
the people who depend on them for services, and steal from both ordinary
people and the government. Accurate information gathering poses problems
even if the dictatorship replaces all officials with individuals who support it
because the future careers of officials may depend on things going well in their
districts. Officials may fear bringing the dictator bad news. In Chapter 6, we
consider some of the institutions dictatorships use to monitor officials and
gather information.
The common institutions that link citizens to political leaders in
information-gathering, mobilizational, and distributive networks are ruling
parties, elections, and legislatures. These institutions incentivize the extension
of patron–client networks from the center to ordinary citizens and the trans-
mission of information from the grassroots to the center.
Dictatorial ruling parties link central elites to large numbers of public
employees and local officials via a patron–client network that distributes jobs
and other advantages and opportunities in return for loyalty and effort on
behalf of the party. These secondary elites administer central and local govern-
ments and distribute whatever benefits citizens receive from the center. To
survive, the dictatorship needs local officials to refrain from sabotaging policies

and stealing from those they govern. One element in the dictatorship’s strategy
for maintaining good behavior among its officials is to limit employment to
those thought to be loyal because party officials have vetted them. Ruling
parties control access to government jobs in nearly all dictatorships organized
by parties.
In most dictatorships, however, it is easy to join the ruling party. Even where
the party requires ideological knowledge and a period of probation to demon-
strate commitment before membership is granted (like the communist parties or
the Iraqi Ba’th), most observers report that opportunists outnumber those
committed to party ideals. Apparent ideological commitment or party loyalty
has failed to restrain self-interested behavior by officials.
So, top leaders in dictatorships need ways to monitor officials. Some dictator-
ships, such as the communist regime in East Germany, have used very extensive
secret police networks for this, but most dictatorships lack the resources neces-
sary for such individualized surveillance. We argue in Chapter 6 that elections for
legislative assemblies and local government create incentives that induce officials
and elected representatives to behave better than they otherwise would. We
believe that this substitute for direct monitoring is sufficiently valuable to dicta-
torial leaders to make elections worth their expense and risk.
Dictatorial ruling parties rarely lose elections, but individual candidates for
legislative and local office lose more often. When officials know that they will have
to run in semi-competitive elections before too long and that they could lose, they
have reasons to extend their patron–client networks all the way to the grassroots,
distribute as much as they can to the people who will vote, restrain their own and
others’ predation on the powerless, and convey information about local needs and
problems to the center in the hope of getting help. Even where popular elections are
uncompetitive, officials and deputies have to compete for party nominations.
Elective offices come with many benefits, so competition is stiff. In this way,
elections and nominations take the place of direct monitoring by central party
officials. They give local officials strong reasons to behave as the center wants them
to, and they supply the punishment for not behaving that way in the form of loss of
valuable offices and the salaries and perquisites that go with them.
We see executive elections as having a function different from legislative and
local elections. The main purpose of executive elections is to demonstrate to
potential rivals within the inner circle of the dictatorship that the dictator and
his faction have the resources and organization to be unbeatable, and in this
way to deter elite defections (Magaloni 2006). Other elites would not be
deterred by fraud. Instead, the ruling faction demonstrates its resource advan-
tage by staging an expensive, nationwide campaign; distributing things of value
to many citizens; and often manipulating the economy to give the impression of
prosperity. Overwhelming votes for the incumbent show that opposition chal-
lenges have little hope of success.
The loyalty, monitoring, and information that elections foster must be paid
for. In Chapter 6, we show that government spending in dictatorships rises in

election years. A side beneﬁt of regular elections is that some spending reaches
ordinary people that might not have done so in the absence of ofﬁcials’ concern
about winning future nominations and elections. We show that infant mortality
is lower on average in dictatorships that hold regular elections, even when
elections allow no choice.

A policy implication of Chapter 6 is that conditioning aid or loans on

semi-competitive elections may help improve popular welfare by
incentivizing local ofﬁcials to work to make beneﬁts and services
intended for ordinary citizens available to them. Although the
introduction of semi-competitive elections while the dictator remains in
control may have little effect on near-term prospects for democratization,
such elections do seem to give dictators reasons to extend welfare-
enhancing services to more people.

In contrast to Chapter 6’s focus on democratic-looking institutions for

indirect monitoring and information gathering, Chapter 7 examines the coer-
cive institutions that dictatorships use directly and openly for information
collection, deterring overt opposition, and protecting themselves from over-
throw. New dictatorships often create or revamp internal security forces to
increase their capacity to ferret out plots, intimidate opponents, and impose
extrajudicial punishments on those suspected of opposition. They also often
raise military pay and promote officers to deter potential opposition in
the army.
Internal security forces focus most of their attention on powerful people
because they are most likely to be able to oust the dictator. They thus spy on
high party officials, military officers, high-level bureaucrats, elected officials,
union leaders, teachers, professors, journalists, and others with potential for
mobilizing opposition, persuading others, and spreading discontent. Where the
dictator has managed to take personal control of internal security, he can use it
to spy on members of the dictatorial inner circle and to arrest or kill anyone he
sees as a potential challenger. We find that about two-thirds of dictators have
personal control over internal security.
In addition to security police, dictatorships also need armies to protect them
against armed attacks from insurgents and external enemies. Usually, dictator-
ships inherit the armies of their predecessors, so they cannot count on their
unconditional loyalty. If past recruitment has been open to talent and promo-
tions based on seniority and merit, then the officer corps will include men from
the various ethnic, regional, religious, and partisan groups that exist in the
country. This diversity of backgrounds and underlying interests helps explain
why armies so often incubate plots, even when the dictator is himself an officer.
If the army retains professional norms, including standard promotion rules, it

retains some autonomy from the dictator, and the dictator has reason to fear it.
The army is the political actor most likely to be able to credibly threaten the
dictator with ouster if he violates sharing agreements.
In response to fear of the army, dictators establish loyalist paramilitary
forces recruited from groups especially close to them to counterbalance the
regular army and defend them from coups. Paramilitary forces can thus under-
mine the army’s ability to constrain the dictator. Dictators also interfere with
military promotions and purge distrusted officers to render the military less
dangerous, but this kind of interference can itself motivate coups by officers
trying to defend professional autonomy or their own career interests. So dicta-
tors have to calculate their strategy toward the military with care. We show that
dictators with paramilitary forces to defend them from coup attempts are more
likely to interfere with promotions and purge officers.

Our analysis suggests that international actors should avoid providing

kinds of aid that can be used to build up internal security police.
Engorged security police enhance the dictator’s concentration of powers
and unrestrained policy discretion. Foreign policy makers should
probably avoid providing support for political police in dictatorships in
general, but that should be especially true if the internal security police
report directly to the dictator, and thus enhance his capacity for arbitrary
and violent action.

Chapter 8 investigates the breakdown of dictatorships. Exogenous shocks

such as international economic crises, natural disasters, and the death of the
dictator can challenge the survival of dictatorships. However, regimes estab-
lished by seizure groups with some traits (for example, those organized by
inherited parties) are more resilient in the face of crises than others. These
exogenous traits interact with institutions and strategies chosen after the seizure
of power to enhance or limit the dictatorship’s vulnerability to breakdown. In
this chapter, we show the consequences for regime survival of some of the
decisions made earlier to enhance the power of particular actors relative to
others in the dictatorial elite (and described in Chapters 4–7).
The chapter focuses ﬁrst on describing how dictatorships break down.
Coups, elections, and popular uprisings are the most common ways of ending
dictatorships. Coups still oust more dictatorships than other methods, though
their incidence has fallen since the end of the Cold War. Almost as many
dictatorships end in contested elections as in coups. Unarmed popular uprisings
have become a more frequent means of deposing dictatorships since the end of
the Cold War.
Most of Chapter 8 investigates the reasons for the breakdown of dictator-
ships. Popular hardship can motivate citizens to take the risk of demanding

regime change, but some dictatorial elites respond more effectively to crisis than
others. We show that economic crisis increases the chance of regime failure in
dictatorships not organized by inherited parties, but not in those led by parties
originally organized to compete in elections or lead insurgencies. We interpret
these findings as evidence that the distributive networks developed by parties
that had to work to attract followers during their formative stages help dicta-
torships survive economic hardships. Dense, preexisting networks reaching into
the grassroots increase the ability of the dictatorial elite to respond effectively to
crises. Party networks can organize a safety net or redistribute costs to protect
the citizens most capable of threatening the dictatorship.
We also explain how the leadership configuration arrived at through the
bargaining described in Chapters 4 and 5 affects both the likelihood of regime
breakdown and whether democratization results from it. We investigate the
consequences of the concentration of power in the dictator’s hands on the
durability of autocracies and what kind of regime tends to follow them.
Concentration of powers interacts with exogenous seizure-group character-
istics, resulting in different outcomes, depending on other traits of the leader-
ship. We show that the personalization of power in military-led regimes tends
to increase their durability because it reduces leadership conflict, which other-
wise tends to be high. Seizure groups organized by inherited parties, however,
are weakened by the personalization of rule because it increases their vulner-
ability to succession crises.
Leadership succession is an inherent weakness of dictatorship. Otherwise
invincible ruling groups may disintegrate during succession struggles. These
struggles are a time of uncertainty and intense bargaining even in regimes with
well-institutionalized succession rules, but they are fraught with fear and real
potential for individual and regime disaster in those that lack such institutions.
Dictatorships in which the leader has amassed great personal power rarely have
binding succession institutions, because such leaders put a high priority on
eliminating them. Instead, personalist dictators may refuse even to identify a
successor, though some groom sons or other close relatives to succeed.
We show that succession is more challenging for dictatorships in which the
dictator has concentrated power than in those with more consultative decision-
making. The likelihood that the dictator’s exit, whether by natural death or
ouster, coincides with immediate regime collapse increases by roughly 50 per-
cent as the personalism score varies from its lowest to highest value. As a
further examination of the effect of personalization on the dictatorship’s cap-
acity to deal with succession, we look at what happened after the deaths of all
dictators who died of natural causes while still in power. Regimes in which the
dictator had amassed substantial personal power (measured during the three
years before his death) are less likely to survive after he dies than regimes with a
more collegial leadership.
Since the timing of natural death is not controlled by the dictator or members
of the inner circle, our finding can be interpreted as showing a causal link

between personalism and regime durability: personalization before the dicta-

tor’s death decreases the ability of successors to reconsolidate the regime under
new leadership. Dictatorships led by inherited parties with collegial leadership
handle succession relatively well, while even inherited parties have difficulty
with succession when the dictator has concentrated great power in his hands.
Personalism also influences how dictatorships break down, and what is
likely to follow their collapse. We find that concentration of power in the
dictator’s hands reduces the likelihood of nonviolent regime change. We have
shown elsewhere that coerced or violent transitions increase the likelihood that
a dictatorship is replaced by a new dictatorship rather than a democracy
(Geddes, Wright, and Frantz 2014).
What happens after a dictatorship falls depends to a considerable extent on
how the dictator and his closest allies respond to the challenges they face from
angry citizens, embittered officers, and foreigners pressing for change before the
regime falls. We investigate one of the factors that determines those responses:
the expectations of dictators and their closest supporters about what will
happen to them and their families after they leave power.
We observe that military dictators are sometimes prosecuted for human
rights violations and sometimes forced into exile, but other officers can usually
continue their military careers untarnished after returning to the barracks.
Consequently, collegial military-led dictatorships tend to negotiate in response
to challenges because most officers do not fear returning to the barracks, and
they have the weapons and control over men needed to replace dictators who
resist negotiation. Negotiated transitions tend to lead to democratization.
In personalized military-led dictatorships, however, the dictator himself
makes decisions about whether to negotiate, ignoring the interests of other
officers. Other officers have less capacity to oust a military dictator with
concentrated power because of his control over the security police and, often,
loyal paramilitary forces as well. Since the dictator faces a higher probability of
punishment post-exit, he tends to refuse to negotiate and to hang on until
forced out. Consequently, personalist military-led regimes are less likely to
democratize.
Personalization of civilian-led regimes also lowers prospects for democra-
tization. Personalist dictators purposely undermine national institutions such as
the military, government administration, and the ruling party that might serve
as springboards for opposition challenges. A by-product of these strategies is
that the individuals who might otherwise have led a movement for democra-
tization may no longer live in the country (or at all), and the institutions that
often incubate new opposition organizations may have been destroyed. Fur-
ther, the dictator’s marginalization, exile, imprisonment, or execution of the
most able and popular politicians in the ruling party reduces the likelihood that
the party will be able to transform itself into a successful competitor if democ-
racy succeeds the dictatorship. As a result, other members of the dictatorial elite
have reason to remain loyal to the dictator as he refuses to negotiate a peaceful

transition and instead clings to power with tooth and claw until the bitter,
violent end.
For all these reasons, democracy is less likely to follow personalist rule than
other kinds of dictatorship. We ﬁnd that concentration of power in one man’s
hands reduces dictators’ willingness to negotiate peaceful exits. Personalist
dictators often hang on until forced out by violence. Violent regime breakdown,
in turn, stacks the deck against subsequent democracy, regardless of whether
foreign invaders, domestic insurgents, or army ofﬁcers led the overthrow.

We and other researchers have amassed many different kinds of evidence

that personalist dictators wreak havoc in their own countries, threaten
neighbors, and set the stage for a renewal of dictatorship after they fall.
In short, dictatorship causes more damage when one man can deploy
vast, arbitrary powers of life and death. The principal policy
recommendation implied by this research is that international policy
makers should avoid contributing to the personalization of dictatorial
rule, even if current security interests suggest supporting such dictators
against neighboring autocrats.
The almost uncritical support of “our” dictators ended with the Cold
War, but decisions about which dictators to provide with loans and aid
still depend heavily on strategic interests. We do not suggest that strategic
interests be ignored, but rather that policy makers should also take into
account the extent of concentration of power in the dictator’s hands
because dictators with unlimited policy discretion can switch sides easily
and unpredictably, using the very weapons provided by their allies to
turn against them later.
While aid and loans are positive forms of intervention that can be used
to shape the behavior of leaders in other countries, sanctions and military
intervention try to change behavior by imposing costs. We suggest that
decisions about economic and military intervention to destabilize
dictatorships should be informed by realistic assessments of whether the
intervention is likely to succeed and what will happen if the dictator falls.
Escribà Folch and Wright’s (2015) ﬁnding that economic sanctions
destabilize personalist dictators only in countries that lack substantial oil
exports implies that sanctions aimed at bringing down dictatorships
should be used only against personalists without oil since sanctions
against other kinds of dictatorship are likely to cause popular suffering
without bringing down the dictatorship.
Before using sanctions against personalist dictators without oil,
however, we need a careful, evidence-based assessment of what kind of
government is likely to replace such a dictatorship brought down by

foreign intervention. In the post–Cold War period, democracy is a little

more likely than new autocracy to follow the overthrow of personalist
dictatorships (Geddes, Wright, and Frantz 2014). Nevertheless, the
likelihood of a new dictatorship is quite high, and the likelihood of civil
war or disintegration into chaos, though low, is not insignificant. The
four dictatorships in which the United States has intervened militarily
since 2001 – Afghanistan, Iraq, Libya, and Syria – ranged from
somewhat to extremely personalized before the interventions.3 These
military actions ended the dictatorships in Afghanistan, Iraq, and Libya,
but none of the countries has democratized. Instead, they are among the
places most likely to be home to terrorist organizations that target
Western interests. At the time of writing, insurgents continue to control
portions of each of these countries, and further disintegration seems
possible.
Observation of events suggests that the more arbitrary, violent, and
paranoid the personalist dictator is, the more likely an overthrow will
result in another autocracy, civil war, or failed state. This is so because
more paranoid dictators destroy more of the human talent and
institutional competence in their countries, leaving them harder to
govern in the future and more vulnerable to the breakdown of basic state
services and the disintegration of order. In other words, the more the
dictator “deserves” to be ousted, the more likely his ouster is to make
conditions even worse for citizens in his country.
Quite a few personalist dictatorships rely on narrow ethnic, clan, or
religious groups for support, as the Taliban, Saddam, and Qaddafi did,
while Assad still does. We believe such dictatorships are especially likely
to experience bloody transitions and violent, unstable futures.
A dictatorship based on a narrow ascriptive category is more likely to
continue fighting to survive despite low odds of winning, as we have seen
in Syria in recent years. When some members of the ruling minority have
grown fat at the expense of the rest of the population, popular hostility
toward the dictatorship tends to spill over onto their nonelite co-ethnics.
All members of the minority may then fear sharing the post-dictatorship
fate of the elite, so they become willing to defend the dictatorship even at
great cost.
If the dictatorship is eventually ousted, the government that follows it
may have difficulty leading a return to normalcy because minority groups
that lose power along with the dictator have reasons, and often the

3
We have not included the countries in which military involvement was limited to drone strikes or
support for military interventions led by allies, but most of them also ﬁt the pattern described.

resources needed (because of the beneﬁts and favoritism they experienced

during the dictatorship), to organize warfare against the government that
replaces them. This adds to the dangers faced by political leaders and
citizens after transition.
The survival strategies typically used by personalist dictators also
imply some further post-ouster consequences that well-intentioned
foreigners considering intervention should ponder. During his years in
power, the dictator with great power concentrated in his hands makes
every effort to eliminate the most competent and politically gifted of his
fellow citizens, whom he views as potential rivals. Some are killed and
others jailed for years. At best, they will have spent long years in exile by
the time the dictator falls. They will thus lack the networks of allies and
mutual assistance that politics depends on, especially in places where
rules are not enforced and institutions have been hollowed out. The
dictator’s effort to rid himself of rivals will limit the pool from which new
leaders can come after his ouster and reduce the political resources they
have to draw on. Consequently, the first post-personalist political leaders
may lack the skills and political networks needed for a transition to
competent, nonviolent government.
The destruction of both political and civil society institutions under
personalist rule also leaves nations that have endured it with little human
infrastructure with which subsequent political leaders can build stable
government. Personalist dictators control appointments and promotions
in all important areas of government and the security forces as well, and
they promote on the basis of loyalty rather than competence or expertise.
Over time, such appointment strategies not only reduce the talent of the
government led by the dictator but can also strip the country of its ablest
and best-educated citizens, many of whom flee rather than face the
dangers, uncertainty, and limited opportunities available in the
dictatorship. The decimation of institutions under personalist rule often
includes the military and ordinary police, leaving them subsequently
incapable of maintaining order or defending the new government from
violent attacks.
In the frightening dog-eat-dog environment fostered by personalist
dictatorship, interpersonal distrust aids survival. Only family or clan
members can be counted on for help, which reinforces people’s sense of
responsibility toward extended family members. This intensified familism
tends to persist after the fall of the personalist dictator and to undermine
the neutrality and competence of government agencies, including the
police and military. It affects hiring, promotions, and the delivery of
government services, including safe streets.

For all of these reasons, the governments that follow personalist rule
often lack competent, unbiased personnel. The dearth of capable
personnel makes it difficult for an occupying army to deliver ordinary
government services such as electricity, water, and garbage collection.
The politicization of the police and security services under the
dictatorship may mean that they have not been trained for the neutral
maintenance of order and safe streets. Citizens may hate and distrust
them. Foreign occupiers may undertake training new police and military
forces, but they will have difficulty overcoming the distrust and
intensified in-group loyalties developed during the dictatorship.
Corruption is usually high in personalist dictatorships. When
corruption has become expected, it does not disappear just because
political leadership has changed. No one knows whether the new
government will be more honest than the last or whether it will survive.
So, the get-it-while-you-can logic tends to persist.
The intervener is likely to be blamed for the political violence and
disorder that follow intervention. In these circumstances, citizens who
welcomed foreign intervention as a means of getting rid of a hated tyrant
may very quickly come to revile the foreigners and the government allied
with them for failing to keep them safe and foster the return of normal
life. These attitudes undermine the occupier’s efforts to promote the
country’s development into a stable ally willing to protect the
intervener’s economic and security interests. Governments contemplating
intervention to rid other countries of personalist dictators should
contemplate these likely outcomes.
These are probabilistic statements. We do not claim that foreign
interventions against terrible dictators never improve things. We think
the Vietnamese intervention to remove Pol Pot in Cambodia definitely
improved life for Cambodians.
But we do urge policy makers not to assume that nothing can be worse
than the current dictatorship or to make decisions under the illusion that
democracy is somehow the “natural” consequence of freeing people from
a tyrant.

Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:22, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.009
Downloaded from https://www.cambridge.org/core. YBP Library Services, on 12 Aug 2018 at 07:39:22, subject to the Cambridge Core terms of
use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316336182.009