Questions: Comparative Historical Methods: Case Selection and Process Tracing

Question 1

A researcher studying why revolutions succeed selects four historical cases — all successful revolutions — for her comparative study. What is the fundamental methodological problem?

• A. Four cases is too small to draw any meaningful comparisons in historical research
• B. By including only successful cases, the researcher has no variation on the dependent variable, making it impossible to identify what distinguishes revolutionary success from failure
• C. Comparative historical methods require cases from the same time period to hold historical context constant
• D. The most-similar systems design requires that selected cases differ on the outcome variable

Answer: (B)

Selecting cases only on the dependent variable is a classic error in comparative research. If every case is a successful revolution, the researcher cannot determine which factors are necessary or sufficient for success — those same factors might also be present in failed revolutions. Without cases of revolutionary failure, there is no contrast that reveals what makes the difference. This is selection bias: the sample is chosen in a way that systematically excludes the variation needed to answer the research question. The fix is to include cases with different outcomes.

Question 2

In the most-different systems design (MDSD), cases are selected because they:

• A. Are similar on most background characteristics, with variation only on the key independent variable
• B. Differ on nearly every background characteristic except the outcome and one shared causal condition
• C. Represent the full range of the dependent variable across geographically diverse settings
• D. Maximize within-case variation over time to enable longitudinal comparison

Answer: (B)

MDSD selects cases that differ on almost everything — geography, culture, economy, history — except the outcome of interest and one shared condition. The logic is: if very different cases share an outcome and one common factor, that factor is strongly implicated as a cause. It operates as an informal most-different controlled comparison: when everything else varies and one thing is constant alongside a shared outcome, the constant factor gets explanatory credit. This contrasts with MSSD, which holds most factors constant and varies only the suspected causal factor.

Question 3

Process tracing examines causal mechanisms within a single case by tracing the step-by-step sequence of events connecting cause to outcome, and it is most powerful when combined with cross-case comparison.

• T. True
• F. False

Answer: True

Correct. Process tracing and cross-case comparison are complementary: comparison identifies which factors co-vary with outcomes across cases, while process tracing confirms that the hypothesized mechanism actually operated within specific cases. Cross-case comparison without process tracing risks identifying spurious correlations. Process tracing without comparison risks overfitting an explanation to a single case. Together, they provide mechanistic, historically grounded causal inference that neither tool achieves alone.

Question 4

In Bayesian process tracing, a 'hoop test' that is passed confirms the causal hypothesis because it is sufficient evidence for the theory.

• T. True
• F. False

Answer: False

A hoop test is necessary but not sufficient. Passing a hoop test means only that the evidence is consistent with what the theory predicts — which is expected if the theory is correct. Failing a hoop test eliminates the hypothesis. Sufficient evidence comes from 'smoking-gun tests': evidence that only your theory predicts and that you actually find. If a finding would be highly unlikely under any alternative explanation but is predicted by your theory, it provides strong confirmation. The distinction between necessary and sufficient evidence is central to rigorous process tracing.

Question 5

Why is cross-case comparison without process tracing insufficient for making strong causal claims in comparative historical research?

Model answer: Cross-case comparison can identify factors that co-vary with outcomes across cases, but it cannot confirm that those factors actually caused the outcomes through a specific mechanism. A correlation across cases could be spurious — both the suspected cause and the outcome might be driven by a third, unobserved factor. Process tracing is needed to show that the causal mechanism actually operated step-by-step within cases, producing observable evidence at each stage of the causal chain.

The classic example: suppose revolutions succeed in countries with high literacy rates and fail in countries with low literacy rates. Cross-case comparison identifies this correlation. But process tracing asks: did literacy actually matter, and how? Did literate populations organize differently, communicate grievances more effectively, or coordinate collective action better? Or did literacy correlate with some other factor (e.g., urbanization, prior state weakness) that was the real cause? Process tracing answers this by looking inside the cases for evidence of the mechanism — making comparative historical analysis genuinely causal rather than just correlational.