Types of Case Studies Single Case Study

advertisement
Research Method II
Case Studies
Jaechun Kim

Today, we will discuss case studies…

As many quantitative works use statistics,
many qualitative studies use case studies…

Why do we conduct case studies?

Practical reasons


Many IR research questions are not amenable to statistical
methods… Why?
Problem of Small-N… What does it mean?





Problem of too many variables and too few cases  Difficulty of
conducting large-N statistical studies…
e.g., causes of great wars; hegemon and openness of trade… cf.
economics
Rule of thumb - N>20
Problem of quantifying key variables… e.g., war
Still, well-conducted case studies > poorly conducted
statistical studies…


Advantage: You can be more attentive to the details overlooked
by large N-studies
Inherent limitations, thouogh – you’ll never have the controls of
statistical methods… (Lijhphart 1971)

Statistical method requires special training
and talent…

 Many works in IR adopt case studies
methods…

Types of Case Studies

Single Case Study (or Case Study Method (Lijhphart))

A fundamental problem of single case study (if you want to
make causal inferences)?




Number of cases (or observations) > number of variables
Otherwise, indeterminate research design
Can’t even test a single hypothesis…
Heuristic value





A careful in-depth study of a single case can produce good description,
which was not available before…
That is, a single case study can produce a very informed and focused
description of a single event…
Sometimes, “good description is better than bad explanation…” (KKV)
Basic data-gathering operation can contribute indirectly to the theory
building or testing…
All in all, avoid atheoretical case study; cases are selected because of
an interest in case per se  case studies of this sort have limited
values…!

Your single case study can be a part of
multiple case studies… of on-going research…
e.g., Hypothesis: democracies seldom loses war
they initiate
 Cases: Falkland war, First Gulf War, etc.
 Science is cumulative endeavor; cumulative nature
of social science…
 Single study can never debunk a hypothesis,
although it can weaken it… !!!



Value of crucial (hard) case study! Think
about Milner’s piece…
Hypothesis confirming (evaluating) and
infirming (generating) case studies (Lijphart)


You can use single case study to generate a
certain hypothesis that has not been
discovered previously…
Deviant case analysis
Studies of single cases that are known to deviate
from established generalization
 May weaken the original hypothesis (because the
case is hypothesis infirming to a certain extent)
 But your cannot overthrow the original hypothesis…
 You can suggest a modified hypothesis that may be
stronger than the original one!


Rule #1, when n=1
Make as many observations as possible within the
case
 It’s not the number of case, but the number of
observations…
 e.g., Kohli (1987) p. 146


Process-tracing (George and McKeown, 1985)


A way of making many observations…
Essay question for next week: What are the major
elements and usages of process-tracing? What are
the advantages and limits of it? Please explain with
at least two existing exemplary researches that
utilized process tracing.

Comparative Case Studies or
Comparative Method (Lijphart)
Why compare?


Can we measure democracy objectively?
Rationale of “Comparative Politics”

To answer a certain research question!

Please try to maintain analytic focus when you do
comparison… !!


Method of Similarity (or Agreement) (John Stuart
Mill) or Most Different Systems Approach
(Przeworski and Tune)



Researchers find a set of cases as diverse as
possible and try to detect what they have in
common….
e.g., “regime change” literature in the 1980s and
early 1990s…
What is the problem of this, if you want to make
causal inferences (or evaluate causal hypotheses)?

Control?

Method of Differences (JSM); Most Similar Systems
Approach (P&T); Comparable Case Strategy (Lijphart)






Allows researchers to control other confounding variables, thereby
enabling to evaluate causal hypothesis
e.g., Presidentialism vs. Parliamentarism; Hypothesis: Presidentialiam 
Ineffective Policymaking
Cases selected?
Cross-sectional comparison (think about Kohli (1987)!); Time-series
comparison (comparing the same unit!)
Cross-states comparison
Remember? Assumption of Unit Homogeneity or its less stringent version,
Constant Effect Assumption…

Problem of over-determination associated with
most comparable case studies
What is it?
 It’s difficult to find sufficiently similar cases, therefore
differences will always over-determine the dependent
variable…
 JSM advised against it! What about KKV?


True that sufficiently similar cases can be
seldom found, but this problem can be
alleviated by smart selection of cases. For
instance, think about Kohli (1987)!

He studied and compared sub-national units to
verify his causal hypothesis…
Download