Determinants of Government Size:

The Capacity for Partisan Policy under Political Constraints.

1 Political Determinants of Government Size

The size and scope of government is a recurring theme both in public and scientific
discourse. Most of these debates center around the proper division of labour
between the market on one side and the state on the other. Politicians, journalists,
interest group leaders and also scientists often disagree with respect to how much
the state should be involved in or interfere with economic matters. One opinion
states that the state should restrict itself to its classic and basic tasks of domestic
and foreign security because its interference with the economy would only lead to
the distortion of market forces and thus to economic inefficiencies. On the other
hand, proponents of a strong state argue that such an involvement is necessary to
counteract the malfunctions and externalities a completely free market would
create.

No matter which position is taken, one basic, although rather implicit, assumption
of both normative prescriptions is that the level of government size can be, and
indeed is, to a large extent purposely changed by government. The main focus of
this thesis will be to explore the validity of this assumption. Generally, left parties
like social-democratic and labour parties are associated with a preference for more
interventionist policies and a greater scepticism towards market-mechanisms than
Christian-democratic or even conservative or liberal parties.
 Political Determinants of Government Size 2

Most importantly, it takes into account two factors which potentially mediate the
effects of government ideology on government size. Both have their origin in
structural features of the political system. The first factor are veto players whose
agreement is necessary for policy change (Tsebelis, 1995, 2002). The number of
veto players is at least partly determined by the characteristics of political
institutions. Depending on election system, regime type, and other institutions of
the political system, the number of veto players varies across countries. This has
consequences for the relationship between partisan government and government
size. It is reasoned that veto players blur partisan effects on policy output since
incumbent parties have to make compromises with other actors in the political
system (Schmidt, 2001).

The second constraining factor are interest organizations. Depending on the

structure of the interest group system and its relations to state institutions, interest
associations can have substantial influence on policy outputs. How this influence is
transmitted can range from simple pressure on politicians and bureaucrats to fully-
fledged “policy concentration” (Lehmbruch, 1984: 62). The latter describes the
cooperative formulation and implementation of policies by the state and powerful
interest associations on the system level. In such a situation, the incumbent
government does not only have to engage in compromises with veto players but
also with organizationally strong business confederations and trade unions. The
paper argues that this is a further hindrance to government in directly realizing its
preferences regarding government size.

Theoretically, the analysis takes the perspective of the leading government party,
for which both veto players and corporatist interest groups are context factors. It is
hypothesized that the impact of its ideological stance is most pronounced in
political systems where it does not face such political constraints on its discretion.
Methodologically, the relationship between party ideology and government size is
seen as being conditional on the number of veto players and the degree of
corporatism in the interest group system. In regression analysis, conditional effects
are appropriately modelled by interaction terms. Although this seems
straightforward, the exploration of these interacting effects has been largely
neglected in previous research on government size.

Especially the constraining effect of political institutions has often been

acknowledged, but modelled as an independent term in the analysis with a
hypothesized negative effect. While this negative relationship has often found
empirical support, it is mainly due to the enduring growth of government size for
the time periods usually examined. As the public sector is more and more cut back
in the eighties and nineties, a negative impact actually means that institutional
constraints furthered this retrenchment. But counter-majoritarian institutions are
supposed to hinder policy change in any direction. Thus, from a theoretical
perspective a hypothesis about an independent effect of veto players is
inappropriate. Since corporatist organizations are assumed to have a similar
mediating impact on the effect of government ideology, modelling these factors
through interaction terms results in a closer theory-model fit.

Another methodological improvement regards the use of an ideology indicator that

varies not only across countries, but also over time. Previous studies usually used
simple classifications dividing parties into broad party families like left, right, and
liberal parties, and weighted them by the number of cabinet or parliamentary seats
held. The variation in these indicators solely results from differences in the
composition of cabinet or parliament. Ideology is implicitly assumed to be constant
within party families as well as over time. This is a rather unrealistic assumption
which gets more implausible the longer the time period and the larger the country
sample under study. Furthermore, these classifications are usually based on
subjective judgements which are likely to evaluate party ideology with regard to
actual policy output rather than party preferences (Budge & Bara, 2001: 12). In
contrast, the ideology indicator used here is objectively derived from election
programs of parties. Although this also has some shortcomings as discussed
below, it is still preferable to traditional measures.
 Political Determinants of Government Size 4

The remainder of this paper is organized as follows: The next chapter examines
the concept of government size in general, clarifies what is meant by the term in
the context of this study, and discusses some measurement problems. The extent
of economic activity of government is the main focus of the analysis and it is
argued that among widely available indicators, public consumption expenditure
and public employment most closely reflect this kind of pursuit. The chapter closes
with a description and illustration of the dynamics of these indicators in general
and with respect to single countries. Although some general trends are visible, it
becomes obvious that large parts in the dynamics of government size are
dependent on country specific differences in economic and political factors.

Chapter 3 reviews the literature that previously proposed such factors and tested
their impact empirically. It concentrates mainly on papers that made some
argument related to the major hypotheses in this analysis. Special attention is
given to studies that examined the impact of partisan politics, institutional
constraints, and interest groups on government size. Theoretical and
methodological issues are discussed as well as some results of the analyses.
Although the literature review focuses on the political factors of main interest here,
it is also used to detect other factors commonly associated with the size of
government. The identification of these alternative or complementary explanations
is helpful for a correct specification of the statistical model which is developed in
the latter part of the paper.

The theories to be examined in the analysis are discussed in more detail in chapter
4. To a large extent, the chapter also focuses on the theories directly related to the
main research question. In the first section, the “parties do matter” thesis is more
closely reviewed and its shortcomings discussed. It is concluded that partisan
theory (Hibbs, 1977, 1992) is most plausible in Westminster style democracies,
that is where governments enjoy a very high institutional power for action and the
interest group system is organizationally dispersed. The theory on veto players
(Tsebelis, 1995, 2002) is endorsed as a promising approach to account for
institutional constraints in a consistent way across countries. To account for the
impact of interest groups, the concept of corporatism (Schmitter & Lehmbruch,
1979; Lehmbruch & Schmitter, 1982) is discussed and which of its features should
result in hindrances to government discretion is made clear. At the end of the
chapter, sociodemographic and economic theories recognized as influential on
government size in previous research are briefly described.

Chapter 5 is devoted to the selection of the data, the operationalization of

theoretical concepts, and the explicit stating of testable hypotheses for the main
theories. The selection of the sample was to a large extent governed by data
availability for the major variables in the study. Although this is obviously not ideal,
it was the only practical solution. A description of the steps involved in reaching the
sample guarantees the transparency of the procedure. The operationalization of
the dependent variables and the political factors is described and advantages and
deficiencies noted. Overall, corporatism stands out in that it is most problematic to
quantify reliably and unambiguously, whereas the indicators for ideology and veto
players are comparatively precise and inter-subjectively replicable. The control
variables representing other theories are introduced in the last section of the
chapter.

The statistical model for the analysis is derived in chapter 6. It is argued that TSCS
regression has substantial advantages over its cross-sectional or longitudinal
alternatives, but also severe limitations. Each of these problems are discussed and
it is noted how they are dealt with in the current analysis. Two further important
issues regard the interpretation of the results of the statistical analysis. First, since
the main hypotheses of this paper are tested through interaction terms, their
meaning and the derivation of conditional effect sizes is examined. Secondly, the
use and value of significance tests are discussed. Although these tests are also
applied in this study, it is acknowledged that their importance is unduly
 Political Determinants of Government Size 6

exaggerated and more attention has to be given to the actual effect sizes and to
the robustness of the findings.

The results of the analysis are presented in chapter 7. It is divided into two main
parts, each devoted to one dependent variable. The first part examines the
determinants of government consumption expenditure and the second part of
government employment. Each model is first calculated using all observations over
the whole period. To investigate the stability of parameter estimates, the
regressions were rerun for two sub-periods. After a discussion of the general
results, the interaction effects detected are examined more closely. Subsequently,
the validity of the assumptions of the model is discussed and the robustness of the
findings to the exclusion of single countries inspected. Especially the latter
procedure turned out to be crucial for judging the accuracy of conclusions reached
from the results of the analysis. The chapter concludes with a summary and
comparison of the results for the two dependent variables.

Finally, chapter 8 explores what factors might be responsible for differing findings
and how far the results of the two analyses can be transferred to the underlying
concept of public sector size. The chapter closes with some tentative conclusions
drawn from this discussion.
 References 111

2 The Size of Government

Before entering into the theoretical debate about the determinants of government
size it is necessary to clarify what is actually meant by the term in the context of
this study. The first section in this chapter discusses the ambiguities associated
with government or public sector as a theoretical concept. The second section
describes some problems in empirically delimiting its scope and size. Section 2.3
gives a descriptive overview of government size, its development over time, and
its variation across countries, which will more vividly illustrate the importance and
appeal of the topic for empirical research.

2.1 Concept and Definition

Given the multitude of quantitative studies on government size, there is

surprisingly little discussion on the demarcation of government and its empirical
measurement. As Rose (1983: 157) noted, before engaging in the analysis of the
causes of government growth, it is necessary to develop an a priori concept of
government. Only with such a prior idea of government can existing government
statistics be examined to the extent they provide adequate indicators for the
measurement of government activity.

Quantitative analyses usually equate government size with readily available

government statistics such as expenditure, revenue or employment, without further
elaboration on the validity of those indicators. By starting with a definition of
government and a discussion of its delineation, an examination of the
appropriateness of the indicators used and a revelation of their shortcomings is
inevitable.

Lane (2000: 15) proposed “State general decision making and its outcomes” as
the most general definition for government. A similar definition was advanced by
Peters and Heisler (1983: 184), but with a stronger focus on steering. According to
them, “Government is … the institution that imparts direction to its society by
various means of collective decision making and exercises the state’s authority on
a daily basis.” In so far the latter definition gives the impression of government as
a unitary, centrally organized decision body, it is inadequate. Many modern states
exhibit a federalist state order where decision making is, at least in certain areas,
highly decentralized among regional units. And even in formally unitary states,
semi-public agencies, publicly owned companies or regional administrations still
have considerable discretion in their actions. Overall, Peter and Heisler’s definition
lays too much stress on authoritative decisions which are not necessary for many
activities pursued by the state, especially in economic matters.

But it points to government activities usually disregarded in measures of

government size. Governments consume goods and services, allocate or
redistribute resources, and collect taxes, all of which can be measured in
monetary terms. But the costs and benefits of regulations, of indirect subsidies like
tax allowances, and its contracting powers as employer and consumer allow
government substantial influence over economic resources with relatively little
reflection in expenditure or employment data (Peters & Heisler, 1983: 183-186). In
short, many modes of government influence on the economy are not adequately
represented by monetary or employment indicators.

Nevertheless, Lane’s definition is preferable in that it is sufficiently general to be

applied across countries and over time. But a functional restriction is imposed on it
for the purpose of this study. The focus here lies on government activity in
economic terms, since ideological differences between governments should
become most visible here. The division of labor between the state and the market
is a major dividing line between left and right parties. Often, the economic part of
state activity is more narrowly described as public sector (Peters & Heisler, 1983:
186), so the two terms will be used interchangeably.
 References 113

If we are content with the definition of government as state general decision

making and its outcomes with regard to economic matters, the next step is to
render it more concrete by a closer specification of what constitutes the state. As
noted above, the state is not a single unitary actor but rather a system of
interrelated institutions and organizations. Obviously, it includes governments at all
levels, be it central, state, or local governments. The distinction between public
and private is not so clear-cut with regard to other cases and it is more and more
blurred with advances to improve the efficiency of public services. There are firms
with different degrees of government ownership or control, like in
telecommunication, railway, or post; privately organized firms whose primary client
or purchaser is government, like in the defence industry; or organizations with
special mandates or licenses from government, like public TV stations or semi-
public organizations “selfregulated” by interest groups (Peters & Heisler, 1983:
183-186).
Generally, a measure of public sector size should include all economic activities of
organizations for which the government has a substantial direct or indirect say in
decision making. In cases were an organization is completely owned, controlled,
and operated by government, it is clearly a public organization. But given the
variety of hybrid publicprivate organizational forms in the mixed economy, an
attempt to delineate what constitutes public and what private activities in general
seems very difficult, if not impossible. The next section introduces the measures of
government size used in the analysis and discusses how they perform with regard
to this and related problems.

2.2 Measuring Government Size

The most commonly used indicators of government size are expenditure measures
derived from national accounts. Often simply total government expenditure is used
to signify the size of the public sector. This might be appropriate in some
instances, depending on the aim of the study, but here a more disaggregated view
is in order. Several components of total government expenditure do not directly
reflect economic activity by the state or do not relate to the theoretical argument
made in this paper. Thus, the choice of the financial indicator was governed by a
trade-off between most closely reflecting the definition of government size as given
in the last section and the association to the theoretical argument about partisan
politics.

The major parts of overall government spending can be classified into capital
formation, subsidies, social transfers, military expenditure, interest payments, and
civilian consumption expenditure (Cusack & Fuchs, 2002: 11). On the whole, it
seemed that a focus on civilian consumption expenditure is most appropriate.
Social transfers and interest payments do not claim any economic resources; they
are just redistributions (Gemmel, 1993: 2). This is also true for subsidies, but they
can be seen as a device of governments to influence economic activity indirectly.
Nevertheless, like in the case of capital expenditure, it is not clear how
governments with different ideological positions relate to this component. Anyway,
the underestimation of economic activity due to the neglect of capital spending and
subsidies is small, since these components constitute only minor parts of overall
outlays (see Cusack & Fuchs, 2002: 11-14). Military spending is also excluded on
theoretical grounds, since it has been argued that the international security
environment rather than ideological factors is its main driving force (Blais et. al.,
1996).
Overall, ideology is most likely to show its effect on civilian consumption
expenditure, which measures the direct economic involvement of government as a
producer and purchaser of non-military goods and services. Besides traditional
state functions like public safety and administration, it mainly includes spending on
education, health care, child care, and other welfare provisions by the state, which
is seen as a major domain of left parties. Thus, using civilian consumption
expenditure in the analysis allows for a more powerful test of the partisan
hypothesis. Furthermore, it avoids a problem associated with total government
expenditure in percent of gross domestic product (GDP) as a measure of
government size, which is not a “real” ratio measure and overstates the size of the
public sector (Warlitzer, 1999: 5). Only government consumption expenditure is
included in the calculation of society’s total economic output.
 References 115

A further advantage of using consumption expenditure is that it is most directly

comparable to a non-monetary indicator of government size, namely government
employment. A major part of consumption expenditure consists of wages and
salaries to public employees and public employment is another measure of the
direct economic involvement of government. The use of two indicators for public
sector size in the analysis and the comparison of the results allows for judging the
certainty for which the findings can be prescribed to the underlying concept.

Both measures are originally drawn from National Accounts as published by the
Organization for Economic Development and Cooperation (OECD), which are
often referred to as the most comprehensive and internationally comparable
source on government and economic statistics (e.g. Saunders, 1988: 272).
Nevertheless, there are some classification and measurement problems that have
to be addressed. Both indicators under-represent the actual involvement of
government in economic matters. Although the definition of “general government”
in national accounts, on which both measures are based, encompasses central,
state, as well as local governments, it disregards all activities of public enterprises.
Even if an enterprise is completely owned by the state, its activities will not be
reflected in general government accounts if it engages in market-oriented activities
(Cusack, 1991: 4). Furthermore, any qualitative devices of government to
intervene in the economy, as described in the last section, are not or only to a little
extent mirrored in these indicators. Since regulation is a major instrument of
economic government policies, this is another deficiency of the indicators with
regard to the definition of government size.

In addition to these conceptual problems, Florio’s (2001) recent analysis uncovers

remarkable inconsistencies in public sector data provided in OECD National
Accounts. The data on OECD countries should in principle be comparable since
they are all based on the standardized United Nations System of National
Accounts (SNA). But as Florio (2001) shows for several items of public sector
data, some countries show figures that are logically impossible. He concludes
that most of these inconsistencies have their source in misinterpretations of SNA
definitions by the statistical offices of member states, from which the OECD
collects its data.

To sum up, there are serious problems regarding the validity and reliability of
government expenditure and employment statistics as indicators for the scope of
government. Despite these apparent shortcomings, the data published by the
OECD is still the most reliable and comprehensive source on internationally
comparable statistics available. Any kind of social research exhibits its distinct
uncertainties which have to be reported and taken into account. Thus, when
interpreting the data and the results of the analysis, it has to be kept in mind that
these indicators do not give a full picture of public sector activities and any
conclusions are principally confined to the narrower concepts these indicators
directly represent. Having discussed the problems associated with civilian
government employment and consumption expenditure as measures of public

Source:

url: https://d-nb.info/1079554270/34

Retrieved: August 27, 2020