Beach and Pedersen 2016: Congruence Methods

chapter 8, from: Beach, Derek, and Rasmus Brun Pedersen. 2016. Causal Case Study Methods: Foundations and Guidelines for Comparing, Matching, and Tracing. Ann Arbor, MI:University of Michigan Press

1min assessment
writing style: wordy, repetitive

text is a description of "congruence" method: The summary is concise and worth reading. The rest of the text adds little to it.
 * how it's different from process tracing: examination of causal mechanism is less fine-grained
 * how to use it: look at cases, which are typical (Cause and Outcome present) and empirically rich

Summary
congruence method often confused with process tracing method

Commonalities: both use within-case, mechanistic evidence (evidence which "enables within-case inferences about causal mechanisms"), in other words, they both derive hypotheses ("propositions") from a theory linking Cause and Outcome ("C -> O") and look for confirming and disconfirming evidence by in-depth examination of cases; also, both do not compare to counter-factual conditions.

Differences: whereas process tracing "explicitly unpacks mechanisms, in particular detailing the dynamic, productive elements of the causal process that links parts together" (p272), the congruence method does not "explicitly theorize" the "actual causal mechanism"

Author claims that there is "a clear cutoff" on the continuum between the "system understanding" (process tracing) and the "minimalist understanding" (congruence) of causal mechanisms, but only give a hazy explanation (p272, top). It seems that process tracing is testing "each [and every] part of a mechanism" (p273), while congruence tests only some of the steps.

Division of labor of the two methods possible: First, for testing if theory is plausible at all: congruence. The process tracing for more rigorous assessment (p272, no details given).

Two types of tests for congruence method: .
 * singular test: single proposition is assessed multiple times (across time or space)
 * cluster test: multiple "non-overlapping" propositions
 * author claims that here we are capturing more of the causal mechanism (thus the black box becomes a "grey box"), so this is closer to process tracing

.

Usage of congruence method:

 * when explaining cases of intrinsic importance:
 * put together theories eclectically, that is, as needed for the explanation of case
 * if in part predicting different outcomes, be explicit about those predictions and the evidence expected and eventually found
 * (I would add: theorize scope conditions, that is, when which of the theories is applicable)
 * when developing a theory:
 * choose typical cases (cause and outcome present)
 * come up with "rough empirical narrative about what 'happened' in a case"
 * then more systematic, still probing in a "far-reaching" manner for alternative explanations
 * to refine a theory: compare typical case with a deviant case in which the outcome is present without the cause; identify alternative causes for the outcome. But for those deviant cases were the cause is present without the outcome, use process tracing (start with cause, and trace until causal mechanism breaks down) (no reason given for why this is superior, p293)
 * when testing a theory:
 * don't look at deviant cases (meaning either cause or outcome absent), but use typical cases (cause and outcome present), in other words: select on independent and dependent variable (?!)
 * choose cases with rich available empirical evidence
 * first choose cases where alternative possible causes are not present; later also look at those (to see if the mechanism changes for them)

Questions

 * Is the author seriously discouraging researchers from looking at cases in which both cause and outcome are at first sight not present? (he calls them "irrelevant cases") Did I overlook sth?