Case Studies

Kåre Vernby

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 1 / 44

Lecture Outline

What is a case study?

Types of case studies
What is a case study good for?
Case Studies and Single-Outcome Studies
Exam question examples

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 2 / 44

What is a case study?

Judging by the literature, it is unclear what a case study is

Some claim that it does not involve quantitative analysis
Others that it strives towards a comprehensive description of a
phenomenon
Yet others equate it with some data-gathering technique (e.g.
participant observation)
...and so on

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 3 / 44

What is a case study?

Gerring (2007) attempts to construct a definition that resonates with

established usage
Definition: A case study investigates the properties of one or a few
cases intensively
A case, in turn, is a spatially delimited phenomenon observed at one
or more points in time
Examples of cases may be: particular individuals, neighborhoods,
municipalities, regions, nation-states, continents
They may also be: local, national or international organizations as
well as public policies, norms, cultures

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 4 / 44

What is a case study?

When is a study no longer a case study?

When the cases become so many that...
...it is no longer possible to investigate each intensively
...when the focus shifts from the properties of single cases to those of
a sample of cases
When this point is passed, the study becomes what Gerring calls a
cross-case study

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 5 / 44

What is a case study?

Another attribute of the case study is that cases are seen as instances
of a broader population
In a thesaurus, a case is usually is defined as “an instance of
something”, “a situation of a particular type”, “an example of
something occurring”
A case is “a case of something”
This is why Gerring (2007) distinguishes between case studies and
single-outcome studies (more on this later...)

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 6 / 44

What is a case study?

Notice also that our description of our case defines what the broader
population of interest is
Say you study the introduction of women’s suffrage in Sweden in
1919 and someone asks you what this is “a case of”
If you say it is a case of democratization the population is larger
than...
...if you say that is a case of suffrage expansions
...or if you say that it is a case of suffrage expansions to women
Notice also that suffrage expansions can occur without leading to full
universal suffrage

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 7 / 44

Types of case studies

The defining feature of a case study is that it studies one or a few

cases
But what types of case studies exist?
In Gerring’s typology this is determined by whether you...
...study spatial OR temporal variation OR both
...and whether you study one OR a few cases

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 8 / 44

Types of case studies

Juan Linz’s “red menace” thesis of interwar fascism as a running

example:
Fascism was more successful in those societies in which the
bourgeoisie had been deeply scared by revolutionary
attempts, however unsuccessful, and where the labor
movement held on to a maximalist revolutionary rhetoric
(Linz 1976)

One hypothesis that follows is that worker radicalism and mobilization

should be followed by fascist mobilization

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 9 / 44

Types of case studies

Diachronic single-case study: Study of over-time variation in one

case
Example: In the years 1910s, Italy experienced major strikes and saw
increasing electoral support for the Italian Socialist Party
Socialists also pushed for radical reform, such as land collectivization
Italian Communist Party was formed
These developments were followed by the formation of Italian Fascist
Party in 1921 and the March on Rome in 1922 that brought Mussolini
to power

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 10 / 44

Types of case studies

Synchronic single-case study: Study spatial variation within one

case
Example: In the year of 1921, the Italian Socialist Party was much
stronger in northern and central regions (Emilia Romagna, Piedmont,
Lombardy, Tuscany) than in southern Italy
Northern and central regions also experienced the largest share of
strikes the same year
At the same time, fascism was more mobilized in northern and central
regions but not in the south

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 11 / 44

Types of case studies

Dynamic single-case study: Study spatial and over-time variation

within one case
Example: In the 1910s, Italian Socialists increasing electoral strength
in northern and central regions
These regions also saw a strike wave towards the end of the 1910s
The primary regions in which fascist mobilization occurred in the early
1920s were the same

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 12 / 44

Types of case studies

Comparative method: Study spatial variation between a few cases

(large territorial units)
Example: While the previous single case was “Italian fascism” we
might now compare this to the case of “Swedish fascism”
In the parliamentary elections of the 1920s, the main leftist party was
the reformist Social Democrats
In Italy, the main leftist party was the more radical Italian Socialist
Party
The first Swedish fascist party was started in 1924 but it never
achieved any electoral success (and neither did any of the subsequent
fascist parties)

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 13 / 44

Types of case studies

Comparative historical method: Study spatial variation between

and over-time variation within a few cases (large territorial units)
Example: Comparing some cases of fascism over time
In Italy, we know that over-time variation is consistent with the “red
menace” thesis
Did the Spanish Falangists mobilize as a reaction to labor
mobilization and radicalization?

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 14 / 44

Types of case studies

The point here is not to examine the veracity of the “red menace”
thesis
I have sought to give a rough idea of different strategies for testing
the hypothesis
The point is that these five types are all considered to be variations of
the case study
We will return to the relative advantages of some of these strategies
tomorrow

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 15 / 44

Why Case Studies?

Hypothesis generation vs. testing: Case studies are thought to

have an advantage in hypothesis generation
Cross-case studies are thought to have an advantage in hypothesis
testing
A case study allows one to have a more detailed look at a
phenomenon
With this focused approach, it is easier to gain novel insights (new
hypotheses and discoveries)
"Lightbulb moments arise from a close engagement with the
particular facts of a particular case." (Gerring 2007, 40)

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 16 / 44

Why Case Studies?

Internal vs. external validity: Case studies are thought to have an

advantage in internal validity
Cross-case studies are thought to have an advantage in external
validity
Internal validity (loosely): Have you correctly described what is going
on in the case(s) you study?
Extrenal validity (loosely): Have you correctly described what is going
on in the population of which you study only some fraction of cases?

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 17 / 44

Why Case Studies?

When including only one case or a small number of cases in your

study it is sometimes easier to e.g. measure democracy
Example: Suppose you want to measure whether there have been free
and fair elections
A close study of one particular election will allow you to ascertain the
fairness of the election from more angles
You have the time to utilize a greater number of independent sources
to get at the nature of the election

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 18 / 44

Why Case Studies?

Insights about causal mechanisms vs. causal effects: Case

studies are thought to have an advantage in studying mechanisms
Cross-case studies are thought to have an advantage in establishing
causal statements
Causal statements: Statements of the type “a change in X causes a
change in Y ”
Causal mechanism: Answer to questions of the type “How come X
causes Y?”

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 19 / 44

Why Case Studies?

Even if we have convincingly shown that “the democratization of

Western Europe caused inequality in this region to decrease” we have
not explained why this is so
To understand whether this is explained by less economically
privileged groups voting for parties that promise redistribution of
economic resources we need to study the mechanism
To study this, we need detailed studies of historical documents and
election statistics
As a practical matter, this cannot be done for more than one or a few
cases

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 20 / 44

Why Case Studies?

Deep vs. broad scope of proposition: Case studies are thought to

have an advantage in studying deep propositions
Cross-case studies are thought to have an advantage in establishing
broad propositions
A case study can be more intensive
Why and how, exactly, did country A democratize?
Why did it happen this time, and not before?

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 21 / 44

Why Case Studies?

Heterogeneous vs. homogeneous population of cases: Case

studies are thought to have an advantage in studying heterogeneous
populations
Cross-case studies are thought to have an advantage in studying
homogenous populations
A concept may be difficult to apply to diverse settings (the problem
of conceptual stretching)
A change in causal factor X may have different effects on the outcome
Y in different contexts (the problem of heterogenous causal effects)

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 22 / 44

Why Case Studies?

A study of fascist parties in government necessarily involves only a

few historical cases unless we stretch the concept too far
Causal effects may also be heterogeneous:
Perhaps democratization only causes inequality to decrease in
contexts where there exist well-organized pro-worker parties?
Perhaps democratization only causes inequality to decrease in
contexts where there exists a close cooperation between
well-organized pro-worker parties and labor unions?

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 23 / 44

Why Case Studies?

Strong vs. weak causal effect: Studying weak causal effects is

difficult in case studies
In Cross-case studies can detect weaker causal effects
Strong Effect = Impact: The causal factor explains a lot of the
variation in the outcome of interest (see lecture 3)
An example of a weak effect is that hypothesized between democracy
and growth

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 24 / 44

Why Case Studies?

Is variation rare vs. common: Case studies are advantageous where

there is little variation in X and Y between or within cases
For example, there is little variation between cases in coups,
revolutions and deaths of leaders
And there is little variation within cases in e.g. electoral system
change and democratization

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 25 / 44

Why Case Studies?

Is data readily available or not: Case studies have an advantage

when data is scarce or hard to come by
The more time needs to be spent on information collection, the fewer
cases can be studied
Suppose you are interested in a hard to measure concept such as
legislators’ preference orders over a number of policy proposals
Only an in-depth study utilizing a broad range of evidence could hope
to accomplish this

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 26 / 44

Case Studies and Single-Outcome Studies

Recall that one important attribute of the case study is that cases are
seen as instances of a broader population
A case is “a case of something”
In a single-outcome study the phenomenon of interest is not seen as
an instance of some greater population of cases
No ambitions beyond describing or explaining the phenomenon of
interest

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 27 / 44

Case Studies and Single-Outcome Studies

A study of Italian fascism is a case study if the researcher is

interested...
...in European interwar fascism more generally
A study of Italian fascism is a single-outcome study if the researcher
is interested...
...only in describing or explaining Italian fascism

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 28 / 44

Case Studies and Single-Outcome Studies

Why might we disapprove of single-outcome studies?

They violate the idea that science should explain much with little
They violate the idea that we can learn from experience
More general causal arguments are more falsifiable, because they have
more implications

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 29 / 44

Case Studies and Single-Outcome Studies

Why might we be interested in single-outcome studies?

Some phenomena are intrinsically important
The rise to power of Mussolini or Hitler
The laggard status of the American welfare state
The outbreak of WWI or WWII
The Chinese Civil War that led to the formation of the People’s
Republic of China

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 30 / 44

Case Studies and Single-Outcome Studies

There are also situations where the potential population of cases

appears too heterogenous to allow for any general statements
Potential examples: World wars, major revolutions, assassinations of
executives
Or, when comparable information on the potential universe of cases is
limited
Potential example: Espionage against foreign powers

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 31 / 44

Exam Question Examples: Political Parties

Suppose we define political parties as follows:

A political party is defined as an organised group of people
with at least roughly similar political aims and opinions, that
seeks to influence public policy by getting its candidates
elected to public office.

What is a) the term b) the intension/attributes and c) the

extension/phenomena
Evaluate this concept from the perspective of the criterion of
differentiation

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 32 / 44

Exam Question Examples: Political Parties

Conceptual elements (except that of operationalization):

Term: The word itself
Attributes/Intension: The attributes that define the concept
Phenomena/Extension: The referents of the concept
One of Gerring’s criteria for concepts:
Differentiation: Is it possible to distinguish the concept from other
concepts?

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 33 / 44

Exam Question Examples: Political Parties

Answer sketch:
Term: political party
Attributes/Intension: “organised groups of people with at least
roughly similar political aims and opinions, that seeks to influence
public policy by getting its candidates elected to public office”
Phenomena/Extension: The organizations that consist “of people
with at least roughly similar political aims and opinions, that seeks to
influence public policy by getting its candidates elected to public
office”
Differentiation: It is problematic to the extent that it fails to
distinguish parties from other organizations that also seek to elect
candidates (e.g. interest groups, PACs)

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 34 / 44

Exam Question Examples: War

Carl von Clausewitz defined war as “an act of violence intended to

compel our opponent to fulfil our will.” It is like a duel, but on “an
extensive scale.”
Suppose we wish to measure this concept of war using one of the
following operationalizations:
1. Militarized conflicts between recognized state system members
with battle deaths totaling at least 1,000 every year of the conflict.
2. A war as exists between states whenever one country declares that
a state of war exists with the country.

Which measure is a) more reliable and b) more valid? Explain your

answer!

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 35 / 44

Exam Question Examples: War

Definitions:
Reliability: The level of random error in an attempt to operationalize
a concept
Validity: The level of systematic error in an attempt to operationalize
a concept

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 36 / 44

Exam Question Examples: War

Answer sketch:
a) Alternative 2 is likely to be more reliable because it likely to be
easy to find reliable information on declarations of war. Alternative 1
involves a lot of difficult issues such as the ambiguity of battle deaths
and the problem of finding reliable records of battle deaths
b) Alternative 1 is likely to be more valid since it more closely
matches Clausewitz’a provision of “an extensive scale”. Declarations
of war can be issued without even resulting in any battles.The fact
that battles have occurred and people have died is also evidence that
the intention has been “ to compel our opponent to fulfil our will.”
Declarations of war occur w/o this intention.

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 37 / 44

Exam Question Examples: Natural Disasters

Suppose that someone draws on recent events in the United States,

where the hurricane Sandy is said by some to have bolstered support
for Obama
They formulate the following hypothesis:
Natural disasters tend to increase the support of the incumbent
government

Evaluate this causal argument from the perspective of a)

manipulability and b) independence. Is it problematic or not?
Explain!
c) Suggest a causal mechanism that could explain why incumbent
support increases in the wake of a natural disaster

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 38 / 44

Exam Question Examples: Natural Disasters

Definitions:
Manipualibility: The causal factor (“X”) should ideally be
manipulable (by someone)
Independence: The causal factor (“X”) must be partially
independent of other causes of the outcome (“Y”)
Mechanism: A good causal argument should stipulate the
mechanism by which a change in X generates a change in Y

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 39 / 44

Exam Question Examples: Natural Disasters

Answer sketch:
a) It problematic, since it is difficult to see how we can manipulate
extreme weather phenomena, at least in the short- to medium-term
b) It is not so problematic, since it is difficult to see that natural
disasters are the consequence of other causes of incumbent support
c) The causal mechanism might be formulated several ways:
M1: Voters tend to see a competent handling of a crises as a sign of
the incumbent’s competence
M2: During periods of severe threat, voters seek security by
upholding the status quo

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 40 / 44

Exam Question Examples

Suppose we want to make a causal analysis of this hypothesis

We study whether Obama’s share of the 2012 vote rose more (or, at
least, fell less) as compared to the 2008 in areas that were hit by
Sandy
Evaluate this idea for a causal analysis from the perspective of a)
exogeneity b) proximity and c) independence. Is it good or bad?
Explain!

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 41 / 44

Exam Question Examples

Suppose we want to make a causal analysis of this hypothesis

We study whether Obama’s share of the 2012 vote rose more (or, at
least, fell less) as compared to the 2008 in areas that were hit by
Sandy
Evaluate this idea for a causal analysis from the perspective of a)
exogeneity b) proximity and c) independence. Is it good or bad?
Explain!

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 42 / 44

Exam Question Examples

Definitions:
Exogeneity: Our causal factor (‘X’) should not be affected by our
outcome (‘Y ’)
Proximity: It is (often) easier to detect the effect of a causal factor
X on Y if Y is observed relatively closely after Y
Independence: The studied cases, if one studies several, should be
independent of each other

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 43 / 44

Exam Question Examples

Answer sketch:
a) It is not likely that being hit by hurricane Sandy is a consequence
of being an area that supports Obama (Unless you believe that
hurricanes are God’s punishment for being a supporter of Obama)
b) Since the hurricane hit a short time before the 2012 election this
should not be a problem
c) Potentially, there exists a large problem of spill-overs here so that
areas cannot be considered independent. For one, people in
unaffected areas might work or know people in affected areas.
Second, national media reports to all voters about the events and
comment on how the crisis was handled

Kåre Vernby (Uppsala universitet) Case Studies November 10, 2013 44 / 44