The most difficult inference problem in the social sciences involves the sorting
out of directions of causation with non-experime
ntal data. In our case, we want to know
how religiosity affects aggregate economic performance, but we also have to worry about