“Context” is important, but (almost) useless if used as a causal category

When making causal (or descriptive) inferences, it is important to think about the context within which the causal relationship is expected to hold because it probably does not hold universally and, possibly, only in a limited setting. Falleti and Lynch have written an excellent article about “context” (I drop the “” now) in relation with causal mechanisms. Personally, I do not like the use of context as a category in causal inference because its meaning its ambiguous. A shared understanding of context” probably underlies its usage in everyday language, but not so much in science.

Context can be defined as synonymous with scope conditions because they define the scope of a causal argument (discussed formally by Walker/Cohen and more accessibly by Foschi). Scope conditions have one important characteristic (as I understand them): One does not make a strong claim that the causal relationship does not hold when the scope conditions are not met. When one tests a hypothesis of economic voting in economically developed democracies (scope condition), one expects the hypothesis to be found confirmed in a sample of developed democracies. One believes that the hypothesis does not hold in economically non-developed democracies, but this is only an assumption or belief and not a clear-cut, formal hypothesis. If the hypothesis on economic voting is only valid for developed democracies, the scope condition seems to be needed and has been correctly specified. If the economic voting hypothesis also finds support among non-developed democracies, the scope condition was not needed, but this is not a problem and would only show that one was too cautious in specifying the scope condition.

This is different when context is understood as making an argument about moderation. Imagine you test a hypothesis of economic voting in developed democracies and you expect that the state of the economy only has an effect on the voting decision if there is an economic crisis. “economic crisis” then is a moderator of the main effect captured by the hypothesis; if there is a crisis, then economic voting works differently than when there is no crisis. Theorizing a moderator means to make a stronger claim because one specifically expects when the causal relationship holds and when it does not hold. A hypothesis about a moderator can be disconfirmed when its presence and absence does not make a difference to the causal relationship of interest. An argument about a scope condition can also be wrong, but it cannot be disconfirmed in a narrow hypothesis-testing sense because scope conditions are not specified as hypotheses in the first place. If scope conditions and moderators would not be different in this regard, we could drop one of the two concepts without any theoretical and conceptual loss because their meaning and implications would be identical.

Consequently, there is a context within a context when one specifies scope conditions and theorizes a moderating relationship (which is the rule for empirical Qualitative Comparative Analysis that works with scope conditions (it should, see chapter 2 here) and usually finds evidence for conjunctions, which are equivalent to moderating relationships). When one refers to “context” in empirical research without further clarification, it is positive because it shows that one pays attention to the scope of a causal argument. However, it leaves the reader in the dark on whether context is meant as a scope condition or an argument about moderation. This is important to know because they have to be interpreted differently and have different implications for follow-up research.

About ingorohlfing

I am a political scientist. My teaching and research covers social science methods with an emphasis on case studies, multi-method research, and philosophy of science concerned with causation, and causal inference. Substantively, I am working on party competition and parties as organizations.
This entry was posted in Uncategorized. Bookmark the permalink.