Case study Petr Ocelík and Lukáš Lehotský Approach, not method Cross-case studies not sufficient • Incredible amount of models available • Complicated definition of interaction among variables • Assumptions over causal mechanism • Equifinality of causal effects • Statistical significance is arbitrary (𝑝 ≤ 0.05) • … Case study • One of the most frequent approaches associated with qualitative research • Detailed analysis of just one/a few cases – goal to produce holistic assessment of the complexity of the case • Cases ≠ observations • Causal vs. non-causal (?) • Qualitative vs. quantitative (?) • Deduction vs. induction Case study paradox • Frequently used and well-established • But! • Avoiding a proper statement of method • No specific and comprehensive “case-study methodology” in place What is a case? Case • Bounded empirical phenomenon • Instance of wider population of similar phenomena • Spatially delimited phenomenon observed at a single point in time/over a time period • Comprises the type of phenomenon that an inference attempts to explain • Case vs. observation • What is the phenomenon a case of? • Nation state What is a population? Defining population • Important to define scope conditions – delimit the boundaries of the domain/population • Scope conditions affect the outcome • Validity of inference beyond scope conditions is not necessarily relevant • Scope conditions need to be conceptualized • Spatial boundaries • Temporal boundaries • Nation state How many cases? Number of cases in designs Spatial variation Temporal variation Absent Present One case Absent Single-case (diachronic) Within case Single-case (synchronic) Single-case (synch. & diach.) Few cases Within and across cases Comparative method (synchronic) Comparative-historical (synch. & diach.) Many cases Across cases Cross-sectional Time series cross- sectional Within and across cases Hierarchical Hierarchical time series Research questions - example • What explains welfare state development within the OECD? • What explains welfare state development within the OECD after Cold War? • What explains variation in U.S. welfare spending over time? • What explains variation in U.S. welfare spending across states? • What explains the relatively weak American welfare state? How many observations? Number of observations • Number of observations (question of 𝑁) distinguishes (?) case study from cross-case analysis • Large 𝑁 • Can’t be handled in qualitative manner • Present in cross-case studies • Implies number of variables that may be tested (lin. regression?) • “Small” 𝑁 • Allows for both quantitative and qualitative research • Allows to gain insight • Don’t have to be very small in fact Cross-case study Observation Variable X1 X2 X3 X4 X5 Y Population Sample Case 1 Obs. 1 Case 2 Obs. 2 Case 3 Obs. 3 Case 4 Obs. 4 Case 5 Obs. 5 Case 6 Obs. 6 Case 7 Obs. 7 Case 8 Obs. 8 Case study Observation Variable X1 X2 X3 X4 X5 Y Population Sample Case 1 Obs. 1.1 Obs. 1.2 Obs. 1.3 Obs. 1.4 Case 2 Obs. 2.1 Obs. 2.2 Obs. 2.3 Obs. 2.4 Cross-case vs. case-study Study Subjects Cases Observations Analysis Population The American Voter (Campbell et al., 1960) Citizens of the United States 1000+ (individuals) 1000+ Quant (cross- case) Americans The People’s Choice (Lazarsfeld 1948) Citizens of Erie County, OH 600 (individuals) 2000 Quant (cross- case) Americans Middletown (Lynd and Lynd. 1929/1956) Citizens of Muncie, IN 1 (cities) 300+ Quant & Qual American cities Political Ideology (Lane 1962) Working men of “Eastport” 15 (individuals) 15 Qual American working class Research design choices Research goals’ affinity Case study Cross-case research Theory/hypothesis Generating Testing Validity Internal External Causality Mechanisms Effects Argument Deep Wide Relation to theory • Atheoretical (?) • Exploratory research, case-specific research • Building hypothesis/theory • Inductive approach – generalizing knowledge about certain class of phenomena • Modifying hypothesis/theory • Sharpening/refining the hypothesis • Testing hypothesis/theory • Deductive approach – “applying” the theory on a case Relation to theory Moment the hypothesis is formed After empirical analysis Before empirical analysis Existence of theory for grounding the research No Building hypothesis Yes Modifying hypothesis Hypothesis testing Theory building vs. testing Theory building Phenomena CasePopulation Theory testing Theory CasePopulation Research goals’ affinity Case study Cross-case research Theory/hypothesis Generating Testing Validity Internal External Causality Mechanisms Effects Argument Deep Wide Research goals’ affinity Case study Cross-case research Theory/hypothesis Generating Testing Validity Internal External Causality Mechanisms Effects Argument Deep Wide Causal effects vs. mechanisms Cause Effect Cause I1 I2 I3 I4 Effect Causation • Covariational • Change in X leads to change in Y • Correlation logic • Symmetric – if positive correlation: • The higher the X, higher the Y • The lower the X, the lower the Y • Two measurements • Difference in degree (strength of effect) • Difference in kind (qualitatively different score) Causation • Set-theoretic • Asymmetric causation • Cause/condition – outcome • Sufficiency • Y is present if X is present • Y may appear without X – but if X appears, Y is present as well • Necessity • Y is present only if X is present • Y appears only if X appears Research goals’ affinity Case study Cross-case research Theory/hypothesis Generating Testing Validity Internal External Causality Mechanisms Effects Argument Deep Wide Generalization • Concepts, hypotheses, and theories do emerge from empirical studies. Case study seems to be the basis of many of them • Hypothesis/theory testing is a form of gaining general knowledge • Generally possible within population (knowledge applicable across the class of phenomena) Single-case studies Many variables, few observations • 𝑁 = 1 • King, Keohane, Verba • Social reality complex – many factors contribute to the explanation – alternative explanations • Measurement error • Stochastic error Solutions • Select lower number of variables • Aggregate • Limit analysis to few variables • Increase number of cases/within-case observations Single-case study • Gerring argues not a case study – single-outcome study/single-observation study instead • Question of causality/inference in single-outcome • Nested analysis • Most-similar analysis • Within-case analysis • Yin argues single-case designs are meaningful only after case selection (critical, unusual, common, revelatory, longitudinal)