Case Studies and Theory Development in the Social Sciences - Alexander L. George [73]
CAUSAL MECHANISMS AND THE COMMITMENT TO MICROFOUNDATIONS
Our definition of causal mechanisms raises the question of whether explanation via causal mechanisms, even if these are defined on the ontological rather than the theoretical level, is different from explanation via the D-N model, in which an outcome is explained if it is shown that it should have been expected under the circumstances. How is it different to say that outcomes were generated than to say that they were to be expected? The essential difference between the D-N model and explanation via causal mechanisms is that the D-N model invokes only one aspect of causality, the outcomes or effects of putatively causal processes. The D-N model also relies only upon two of the many sources of inference that David Hume identified: constant conjunction (or a positive correlation between the appearance of the hypothesized cause and the observed effect); and congruity of magnitude between purported causes and observed effects (a positive correlation between the magnitude of the hypothesized cause and that of the designated effect).
Gary King, Robert Keohane, and Sidney Verba pose a view of explanation in Designing Social Inquiry (DSI) that differs from our own in a subtle but important way by emphasizing the importance of causal effects—or the changes in outcome variables brought about by changes in the value of an independent variable—over that of causal mechanisms. These authors argue that the definition of causal effect is:
Logically prior to the identification of causal mechanisms… . We can define a causal effect without understanding all of the causal mechanisms involved, but we cannot identify causal mechanisms without defining the concept of causal effect… . We should not confuse a definition of causality with the nondefinitional, albeit often useful, operational procedure [process-tracing] of identifying causal mechanisms.272
This view risks conflating the definition of “causality” with that of “causal effect.” The definition of causal effect is an ontological one that invokes an unobservable counter-factual outcome: the causal effect is the expected value of the change in outcome if we could run a perfect experiment in which only one independent variable changes. Statistical tests and controlled case study comparisons are operational procedures for estimating causal effects across cases. Usually, in the social sciences, these procedures are employed in nonexperimental settings that can only approximate the logic of experiments. Similarly, a “causal mechanism” invokes an ontological causal process, and process-tracing is an operational procedure for attempting to identify and verify the observable within-case implications of causal mechanisms. Consequently, this passage of DSI compares apples and oranges in juxtaposing an ontological notion (causal effects) and an operational procedure (process-tracing) rather than comparing ontology to ontology or procedure to procedure.
Albert Yee has made an opposite and equally fruitless assertion that causal mechanisms are “ontologically prior” to causal effects because one cannot have a causal effect without an underlying causal mechanism.273 Such arguments are true but trivial, as they divert attention from the key point that causal effects and causal mechanisms are equally important components of explanatory causal theories. The more productive question is whether case studies have comparative advantages in assessing causal mechanisms within the context of individual cases, while statistical methods have comparative advantages in estimating the causal effects of variables across samples of cases.
A more radical