You want to find empirical studies that show free trade to be harmful to free-trading nations? No problem; you can find them. You want to find empirical studies that show government stimulus spending to be a sure-cure for what ails a slumping economy? There are plenty of such data-rich studies out there. You want to find empirical studies that show that violent crimes aren’t deterred by the death penalty? Not a problem. You want to find empirical evidence that increased rates of handgun ownership increase citizens’ likelihood of dying of gunshot wounds? Many such studies are available.
You can also find plenty of empirical studies showing the opposite of what is shown by all of the above studies. And these other studies are, as a group, no less carefully done than are the studies that they contradict. And these other studies, also, are done by scholars no less credentialed and no less objective than are those scholars who produce the contrary findings.
That’s the reality of the social sciences. It’s not an exercise in simple observation of simple and self-defining facts, only one or two of which change at any time.
Therefore, theory is important. Among other roles, theory directs our attention to what patterns to look for, and helps us to better understand what empirical findings warrant our suspicion more than others. Obviously, theory should never be used as dogma to prevent our learning from careful empirical studies. Nor, however, should well-accepted and coherent theories be tossed aside simply because a handful of people produce a few studies that are inconsistent with that theory – especially if other careful empirical studies support the theory.
So while it’s always a good instinct to ask “What do the data say? What does history tell us about this matter?”, it’s just as scientifically naive to ridicule thoughtful discussion of theory (including discussion of pitfalls in interpreting data) by suggesting that the discussion is useless because it presents no data as it is to suggest that theory should never be subjected to empirical tests.
via Where Are My Data?!.