Monday, July 12, 2004

Is Economics a Science?

Tyler Cowen of Marginal Revolution asserts that economics is a science (see here and here). Normally, I would ignore such an assertion, but when a thoughtful blog like The Right Coast points to it, uncritically, I am compelled to comment on it.

Cowen is not alone in claiming that economics is a science, of course. But his claim is especially fatuous; to wit:
We produce empirical knowledge which is subject to process of testing, broadly interpreted, and feedback....We even now have controlled experiments. And look at some of our competitors. String theory is not yet empirical. Environmental science and ecology are rife with ideology. Astronomy doesn't have controlled experiments. And isn't chemistry just plain outright boring? There is plenty of empirical economics I don't trust, but usually it is for quite hackneyed reasons (e.g., data mining), rather than for "intrinsic to economics" reasons.
I'll come back to Cowen's first point, that economics produces empirical knowledge, etc. As for the other points, here goes:

• A controlled experiment involving human behavior doesn't yield valid results if its subjects know they're participating in an experiment or if their environment is manipulated for the purpose of an experiment.

• So what if string theory isn't yet empirical? It's a hypothesis that may, someday, be tested by practitioners of a bona fide science: physics.

• So what if astronomy doesn't have controlled experiments? A science can be a science without controlled experiments.

• The boringness of chemistry is in the mind of the contemplator. And boringness, of course, is neither here nor there when it comes to science.

• Not trusting "plenty of empirical economics" is a valid instinct. (See my post on economic forecasting, for example.)

Before I attack -- no, address -- Cowen's claims about the "empirical knowledge" produced by economics, I want to be clear about the definition of science. Wikipedia says this:
Science is both a process of gaining knowledge, and the organized body of knowledge gained by this process. The scientific process is the systematic acquisition of new knowledge about a system. This systematic acquisition is generally the scientific method, and the system is generally nature. Science is also the scientific knowledge that has been systematically acquired by this scientific process.
The scientific method, according to Wikipedia, amounts to this:
The essential elements of the scientific method are iterations and recursions of the following four steps:
1. Characterization [or observation]
2. Hypothesis (a theoretical, hypothetical explanation)
3. Prediction (logical deduction from the hypothesis)
4. Experiment (test of all of the above)
All of this is overlaid by publication, criticism, and argument. Science cannot be like "cloistered virtue, unexercised and unbreathed, that never sallies out and sees her adversary..." (John Milton, a speech to the parliament of England, 1644, found here). Granted, economists publish a lot, and their lives seem to revolve around criticism and argument. Big deal. The same is true of The New York Times.

So let's look at Cowen's assertion that economists "produce empirical knowledge which is subject to process of testing, broadly interpreted, and feedback" -- which I take as a clumsy attempt to say that economists follow the scientific method. That's precisely where Cowen and others who claim that economics is a science are wrong.

Remember what I said about controlled experiments in economics: They're meaningless. Therefore, the only valid way to test a hypothesis in economics is 1) to make a prediction of future economic behavior that is based on all data from non-experimental events that are relevant and available at the time of the prediction and 2) to test that prediction against data for non-experimental events that occur after the date of the prediction. Any other approach (e.g., "screening" available data or using data for past events to test the validity of a predictive model) is unscientific. (The words "cheating" and "manipulation" come to mind.)

Let's see if Cowen adduces valid examples of the scientific method in his eight empirical propositions about economics. He says, reasonably enough, that if economics is a science "we should expect to see empirical progress." He then lists eight "issues where I (and many others) have been swayed by the data," which he states in the form of untrue claims in which he no longer believes. Here they are, followed in turn by my comments.
1. We can both control the price level and keep interest rates stable by targeting the monetary base. Twenty years ago I believed this, but even the Swiss have not stuck with monetary targeting. A better solution is to broadly target the price level but allow for mild inflation.
This is a case where "science" learns what "real people" already know, as in the case of the recent "discovery" that dogs understand words. Ratifying lessons from experience is more bookkeeping than science.
2. Minimum wage boosts will generally put many low-skilled workers out of work.
I addressed this one in recent posts, here and here. Cowen's co-Volokh Conspirator, Jacob Levy, has more to say here. In any event, Cowen's sources (here) are hardly compelling. The bottom line is akin to a shoddy defense summation. In the exact words of one of Cowen's sources: "Now that we've re-evaluated the [spurious: ED] evidence..., here's what most labor economists believe: The minimum wage kills very few jobs [one, two, thousands?: ED], and the jobs it kills were lousy jobs anyway [says the well-fed economist: ED]. In other words, "My client isn't guilty, but if he is guilty he isn't very guilty."
3. Investment is highly elastic with respect to observed changes in real interest rates. I've seen a few good studies that generate significant elasticities, typically using taxes as an exogenous instrument. But more often than not you can't get this result.
If the "good studies" are good, the other studies must be "bad". This isn't science, it's a fishing expedition.
4. Free capital movements for developing countries should usher in macroeconomic stability. Ask Argentina, Thailand, and Indonesia. Sometimes this proposition will be true, it is simply not as true as we once thought. If you don't do all your reforms to perfection, and perhaps even if you do, international capital markets may put you through the wringer.
This has elements of the scientific method, in that it represents the refinement of a model based on experience. But the resulting model is merely qualitative; it has no predictive power.
5. Immediate privatization is more important than establishing the rule of law [in ex-Communist countries]. Arguably the jury is still out on this one. We haven't observed the other sequencing in many cases (when has rule of law come first?) and thus we do not have the relevant counterfactual. But privatization alone is less effective than we used to think, pick almost any ex-Communist country as an example.
First, the proposition isn't a valid scientific hypothesis because neither "privatization" nor "the rule of law" can be quantified. Second, if the hypothesis was widely believed among economists, that only goes to show the naiveté of economists.
6. It is relatively easy for a disinflation to be credible, provided the government sticks to its guns.
This is another untestable hypothesis, which Cowen "disproves" by referring to a theoretical analsyis. Wouldn't it be great if science were always so easy?
7. Fairness perceptions, envy, and a stubborn attachment to the status quo have little to do with nominal wage stickiness. OK, this one remains up for grabs. But the evidence is mounting in favor of the importance of fairness perceptions; furthermore this is strongly consistent with my real world experience.
So, the original hypothesis -- if you can call it that -- hasn't been disproved, after all. But that doesn't matter, because if you can disprove a scientific hypothesis with a theoretical analysis (as in number 7), I guess you can disprove it with a few "real world" observations.
8. Human beings maximize expected utility in the same way, regardless of context. But now, alas, I despair as to how general a science economics can ever become.
My point, exactly.

Economics will approach being a science only when its practitioners finally make an unambiguously scientific breakthrough, such as accurately and consistently predicting near-term changes in GDP. Barring that, economics will remain what it is today: a cacophony of competing untestable hypotheses, shaped by ideology, and justified by decorative mathematics and selective statistics.