Thursday, September 8, 2011

Of hurricanes and economic equilibrium

Economists generally interpret economies and financial markets as systems in equilibrium or at least close to equilibrium. The history of economics has revolved around this idea for more than a century since the early theories of Leon Walras and Stanley Jevons. More recently, the famous Efficient Markets Hypothesis (EMH) has been the guiding theme of financial market theory. I've written earlier about the many shortcomings of this hypothesis, but of course EMH enthusiasts often point to evidence in favour of the hypothesis. For example, it's hard to beat the market and difficult to find any predictability in market movements. This seems to point to something like information efficiency -- all public information already being reflected in prices.

The EMH runs into trouble in the form of many exceptions to market unpredictability, so-called "anomalies," such as those identified by Andrew Lo of MIT. His 1990 book A Non-random Walk Down Wall St., written with Craig MacKinlay, documented a number of persisting patterns in market movements. For example, he and colleagues found in a study of price movements over a period of 25 years that the present returns of lower valued stocks were significantly correlated with the past returns of higher valued stocks. This means that by looking at what has happened recently to higher priced stocks, investors can in principle predict with some success what will happen to the future prices of cheaper stocks. This clearly contradicts the efficient market hypothesis.

But there's a way out for the EMH enthusiast, who typically responds by saying something like "Of course, there may be small deviations from equilibrium from time to time, but the market will then act to eliminate this deviation. As arbitragers move to profit from it, their actions will tend to wipe out the deviation and its associated element of predictability, restoring the efficient equilibrium after a short time."

The EMH defender can then even claim that this insight is precisely where the EMH displays its greatest power, as it makes specific predictions which generaly turn out to be true. In the case of the "anomaly" identified by Lo, for example, the EMH enthusiast would point out that this particular predictability has disappeared over the past 20 years just as the EMH would predict. By 1997, other researchers revisiting the phenomenon needed high frequency data even to find it at the level of minute-by-minute returns on the New York Stock Exchange. By 2005, when physicists Bence Toth and Janos Kertesz looked at the effect again, they found it had completely vanished. Conclusion: not only is the EMH safe from the criticism, it's a shining example of a theory which makes accurate predictions -- and as every economist knows, this is the sole measure by which to judge the value of a theory, as Milton Friedman argued in his famous 1966 essay The Methodology of Positive Economics.

There are many dubious elements to such a claim, of course. For example, the EMH doesn't make any specific predictions about time scales, so this prediction is a rather weak one. As I noted in my earlier post on the EMH, some similar anomalies have persisted for 20 years and haven't gone away.

But I want to step back and consider something more seriously wrong with the idea that the progressive disappearance of a pricing anomaly is evidence for the "increasing efficiency of the market." This I think is not correct. To see why, it will help to develop an analogy, and the recent hurricane Irene, which swept northward along the coast of the US, provides a good one.

Suppose that some atmospheric theorists developed the Efficient Atmospheres Hypothesis (EAH) which asserts that planetary atmospheres in general -- and the Earth's, in particular -- are always in a state very close to equilibrium with the air resting in a calm state of repose. Given the total amount of air, its density and the force of gravity, the theorists have even been able to predict the air pressure at sea level and how it should fall off with altitude. The EAH works pretty well in explaining some of the most basic aspects of the atmosphere. But the theory also makes a more controversial claim -- that, in fact, the air pressure
at any two places in the atmosphere (at the same altitude) should always be identical, or at least very nearly the same. After all, the EAH theorists argue, if there were pressure differences, they would create winds carrying air and energy from the higher pressure zone toward the lower pressure zone. That flow of air would lower the pressure in the former place and raise it in the latter, eventually bringing those two pressures back into balance.

In other words, any momentary imbalance in air pressure should create forces which quickly act to wipe out that difference. Hence, the pressure everywhere (at the same altitude) should be identical, or almost identical.

Now, critics of EAH might say this is crazy. After all, we observe winds all the time, and sometimes great storms. Just last week high winds from hurricane Irene caused massive flooding up and down the eastern US. Isn't this obvious disproof of the EAH? On the contrary, the EAH theorists might respond, these observations actually provide further evidence for the theory. OK, they concede, the atmosphere does sometimes depart a little from perfect equilibrium, but the hurricane is clear evidence of how normal economic forces drive the system back into equilibrium. After all, if you look at the data on the hurricane after it passed by the US, you'll see that the immense pressure differences within it were slowly eroded away as the hurricane dissipated energy through its winds. Eventually Irene petered out altogether as the atmosphere was restored to equilibrium. This is simply another victory for the EAH; indeed, it predicted just this ultimate fate for such a disturbance.

In the setting of atmospheric physics, of course, no one would take the EAH seriously. It's obviously crazy to claim that the gradual disappearance of Irene was evidence for the "increasing efficiency" of the atmosphere, or (equivalently) for its return to the equilibrium state. Irene was one very much out-of-equilibrium disturbance, but it's disappearance says nothing about whether the atmosphere as a whole has become closer to equilibrium. As Irene disappeared, hundreds of other storms, and perhaps other proto-hurricanes, were being stirred up elsewhere on the planet. Storms are always fading away and others are always growing in the great chaotic maelstrom of the atmosphere, and all of this reflects its condition as a system driven out of equilibrium by energy from the Sun. To show that the atmosphere was actually moving closer to equilibrium over time would require some global study of storms and winds and air pressure differences to see if they were in some general sense progressively getting smaller. What happens to one storm is actually quite irrelevant.

Returning to the case of financial markets, the same must be true of a single anomaly such as the one identified by Lo and MacKinlay. Over twenty years, this one has slowly disappeared, as good empirical work has shown. But this doesn't really tell us anything about whether the market as a whole is getting more or less efficient, closer to or further away from equilibriuim. That's an entirely different question.

I've belabored this analogy because I think it helps to illustrate an important point. Arguments over the EMH often center on highly technical analysis of statistical correlations and market predictability or the lack of it. Yet often such analysis can be correct in their technical detail, but not actually support the larger claims made by the people doing the analysis.

I think the analogy also helps to bring into focus the shady deviousness of many arguments about market efficiency based on the EMH. In a talk from several years ago, soon after the crisis, Jeremy Siegel of the University of Pennsylvania addressed the question of whether the crisis had shown the EMH to be false. The EMH asserts, of course, that markets work very efficiently to exploit the wisdom of crowds to funnel savings into investments giving the best long run returns (rather than, say, fueling a financial bubble based around mortgages bundled opaquely into crappy CDOs). But no, he asserted, the financial crisis hadn't shown anything to be wrong with the EMH, indeed it offered further evidence for the EMH because (I'm paraphrasing from memory) "everything that happened did so for sound economic reasons." In other words, when people finally understood the true nature of those CDOs, their values plummeted, just as the EMH would predict.

This is like an Efficient Atmospheres Hypothesis theorist saying, first, that "a hurricane could never come into existence, because the atmosphere is efficient and in equilibrium, and the forces of physics act to keep it there," and then, in the aftermath of a hurricane, saying "see, I was right, the atmosphere worked to restore equilibrium just as I said." 

It's a simple case of using the theory when it fits the data, and ignoring the data when it doesn't fit the theory.