Monday, September 05, 2005

Over at Marginal Revolution, Alex Tabarrok has a post entitled "Why Most Research Findings are False." His source is a research paper which boldly claims "it can be proven that most claimed research findings are false."

It turns out that "it can be proven that X" means "I can write down a model that exhibits feature X, and give you a couple of general reasons why my model is not so bad." Which, if you like writing down models, results in a lot of "proofs" - the results of which can be presented without any qualifications (I am referring here to Tabarrok's summary).

This, in my experience, has been rather typical of economics papers (though the paper in question is published in a medical journal). Assume some model, give some qualitative justifications for this or that feature, derive a result, and claim that this justifies policy Y. The fact that the model does not claim to be a close approximation of reality does not usually prevent people from advocating practical policies based on this thought experiment.

I can easily for example write down a model that has the majority of research findings being true (it would involve every claim followed up in multiple subsequent studies). This gives me just about as much claim to "truth" as the paper Tabarrok summarizes.

Keep this in mind while reading over Tyler Cowen's assessment of Chris Mooney's The Republican War on Science. Cowen comments:

I agree with most of the arguments but would have called it The Political War on Science. Democrat politicians are excessively enamored of government regulation, for instance, and many of them do not pay enough attention to incentives. (Admittedly these issues are not as clear cut as the theory of evolution; Mooney in fact suggests a scientific approach will lead to more regulation.) The left often treats human beings as excessively malleable.Both Carter and Clinton committed some gross errors out of self-deception; they violated the simple principle of dominance rather than any complicated scientific hypothesis. (What exactly should count as an error of science?)

Allow me. Science is the reasoned investigation of nature driven by the scientific method. An error in science involves basing a policy on reasons that go against accepted scientific theories. Obvious, one would think.

None of the other examples Cowen cites involve science. Economics is not a science, contrary to Cowen's previous reflections on the topic. No it does not matter that economics uses math or that economics uses "real" data. Unless you are able to test your predictions, you are not a practitioner of the scientific method. Nor does gathering some laboratory subjects and having them trade money count - one cannot say what changes are brought about by having a laboratory experiment (as opposed to what people would do in the "real world") and how far one can generalize based on the idealized laboratory games.

But now, why is my definition of science the correct one? Cowen might say that a sufficient criterion for being a science is dealing with "empirical" data. Who is to say that he is wrong?

Well any encyclopedia for one thing. For another, Cowen and other economists are trying to ride the coat-tails of the success of the hard sciences. Look around you and you will see repeated success of science - from the computer you are reading this on to the myriads of electrical devices all around your home. Calling economics a science has the implication that we should have roughly as much confidence in the theories currently believed by economists as in, for example, the theory of gravity or quantum theory. Considering how thoughroughly the former have been tested, and the natural limitations to testing economic theories, I consider this to be ridiculous.


At 3:08 AM, Blogger angela said...

awww, im sorry your only fan is a spammer. my poor yankee.

At 7:06 AM, Anonymous Anonymous said...

you'll enjoy the article by Daniel
Dennett in the NYT on Aug 28,
entitled "Show Me The Science".


Post a Comment

<< Home