Ekonomija da ni znanost?

Adam Ozimek se je v Forbesu razkuril nad Pascalom Gobryjem, ki trdi, da ekonomija ni znanost, ker ne uporablja poskusov z naključno oblikovanimi vzorci (randomized trials), ampak se ekstenzivno poslužuje matematike, težko razumljivega žargona in znanstvenih revij. Ozimek vrača, da potemtakem tudi astronomija ni znanost, saj se poslužuje podobnih metod in njenih spoznanj tudi ni mogoče replicirati.

So does he think economics would be improved without the use of data, statistics, theory, and peer reviewed journals? Basically that would turn economics into nothing more than free-wheeling, groundless, story telling about the economy. You can find this kind of economic writing all of the place now, for example in the panicked moralizing of hyperinflationists, and the zany claims of gold bugs. Without tools that economists use we’d simply have stories versus stories instead of actual evidence. To argue that economics should be nothing but randomized experiments and free wheeling punditry would throw out a lot of awful useful research.

Overall, I disagree with the notion that a science needs to be stricly randomized controlled trials. But more importantly, for people who understand empirical methodology, the question “is this science?” is basically useless. Instead we ask detailed questions about causality, robustness, bias, heterogeneity, external validity, etc. In doing so we learn a lot about how much weight we should put on various studies and estimates. And by the way, a lot of the “math, impenetrable jargon, peer-reviewed journals” that Gobry finds so cargo-cultish is quite useful here.

Instead, whether something is “science” is most useful to convey to the layperson how much weight they should put on various research. And to this end I think Gobry is completely wrong that “randomized trials good, non-randomized trials bad” is a useful heuristic. After all, there are plenty of randomized trials done with 12 people in laboratory conditions that have little to do with reality that should receive little weight in our beliefs. And there are large scale, well-done econometric studies that should be given a fair amount of weight in our beliefs. Whether something is a randomized trial or not is but one dimension among many in determining how we should think about it. If you need to boil what is scientific down to a single heuristic, a better criteria than “randomized trials or not” is whether it is falsifiable. You can falsify a lot without randomized trials. The long-run Phillips curve, the idea that productivity shocks can fully explain the business cycle, the claim that the gold standard will lead to economic growth, and a litany of other zany ideas have been thoroughly debunked by economists using tools that Gobry finds unscientific. While there may not be as many economic claims that we know are true, there are certainly a good amount that we know are false.

Vir: Adam Ozimek, Forbes

%d bloggers like this: