Reply #27: Let me qualify my previous post [View All]

[Euromutt]

Research involving econometric modeling, by its nature, is retrospective; it looks back on data that has been compiled previously, usually by people other than the researchers themselves. Retrospective studies certainly have their place, such as case control studies in medical science, but only very rarely do they produce any conclusive evidence.

Retrospective studies are conducted to see if an association can be found between a phenomenon and a possible cause, and usually they produce such an association. But because the data wasn't gathered under controlled conditions, it's practically impossible to control for every possible variable, which means the associations produced are likely to be spurious to some extent. So competent, responsible scientists use the results as a basis for a subsequent prospective study under more controlled conditions. (These are more expensive to conduct than retrospective studies, obviously, which is why retrospective studies is used for initial exploratory research.) In prospective studies, the associations found by retrospective studies generally turn out to be much weaker, or they disappear entirely. So a retrospective study by itself isn't very good evidence, unless and until its findings have been confirmed by a subsequent prospective study.

Incidentally, if you look at the public health literature on firearms, you see it consists almost entirely of retrospective studies, with no follow-up prospective studies. Several decades of one retrospective study after another, but for some reason, nobody's ever had the bottle to see if the findings of all those studies hold up under a prospective trial. Makes you wonder why, doesn't it?