Why rational expectations models can be wrong

Washington University professor David Levine has a pair of articles (article 1, article 2) in the Huffington Post about why financial crises can't be predicted, and why rational expectations theories are the only good theories in economics. Although I disagree strongly on both counts, Levine's arguments are very thought-provoking.

(As an aside, just to get this out of the way, Levine uses the Heisenberg Uncertainty Principle as an analogy to explain why observer effects make non-rational economic theories fail. This is not a great analogy, since the Heisenberg principle can only be related to observer effects in certain specific situations; also, the Heisenberg principle deals with conjugate pairs of observables, which are different from the kind of observer effects Levine is talking about. This isn't super-relevant, just remember not to get your physics intuition from interdisciplinary analogies. But anyway, back to Levine's ideas...)

Levine's first argument is that financial crises, by their very nature, can't be predicted:

Take an example: how we might predict stock market crashes? Suppose that two behavioral psychologists, call them "Kahneman and Tversky," produce a model of "cognitive biases" that predicts when crashes will occur. The model tells us that the stock market will crash on October 28. Since the model is reliable and has a perfect track record, we naturally believe this prediction. So what would you do? You would sell all your stock on October 27. But of course if enough people do this the stock market will crash on October 27 and not October 28. So this apparently reliable model will be proven wrong.

This is similar to an argument made by Levine's colleague Stephen Williamson. I've addressed this type of argument in the past, but here is a brief recap:

Conditional predictions are different than unconditional predictions. If your modeling goal is to say with confidence that "A financial crisis will occur at 9:01 A.M. on February 17, 2012," then Levine is right; you are probably not going to succeed. However, if your modeling goal is to say: "Unless X policy is taken first, a financial crisis will occur at 9:01 A.M. on February 17, 2012," then you have a chance of succeeding. Why? Because unless investors can predict whether policy X will be taken, then knowing that the model is correct doesn't allow them to make riskless profits. And of course, conditional predictions are the kind policymakers usually care about. So Levine is wrong in cases where policy is decisive.

As a side note, even in cases where Levine is right, this does not make modeling crises a useless exercise. It may be that increasing our understanding of the causes of crises leads to a decrease in the amount of crises rather than an increase in our ability to predict the timing of the crises that do occur. But that's fine! It means that the benefit of better crisis modeling accrued to society instead of to the modelers. That just means that research into the causes of crises generates a positive externality, and should therefore be subsidized by the government and/or some other collective public-goods provision mechanism.

And it means that critics of the econ profession who say "You economists didn't pay enough attention to models of financial crises" can still be right, even if Levine is right!

Levine's next argument is that rational expectations models of the economy are the only models that can be right over the long term:

In simple language what rational expectations means is "if people believe this forecast it will be true." By contrast if a theory is not one of rational expectations it means "if people believe this forecast it will not be true." Obviously such a theory has limited usefulness. Or put differently: if there is a correct theory, eventually most people will believe it, so it must necessarily be rational expectations. Any other theory has the property that people must forever disbelieve the theory regardless of overwhelming evidence -- for as soon as the theory is believed it is wrong.

But notice the logic behind this idea. It assumes that economic theories penetrate the public consciousness and are used by economic actors. Levine postulates that over time, as non-rational-expectations theories are shown to be wrong, economic decision-makers (i.e. you, me, and other folks) will gravitate toward rational expectations theories. This will mean that in the long run, only rational expectations theories will describe the economy. But it also must mean that in the long run, only rational expectations theories will be believed by the public. So if Levine is right, we should see more and more regular people gravitating toward the belief that other people are rational.

Do we, in fact, see this? I am not sure we do. On one hand, index funds are becoming more popular, signifying greater public acceptance of the Efficient Markets Hypothesis. On the other hand, the recent recession probably decreased faith in rational-expectations models of the macroeconomy, both among economists and among the public at large.

So if we don't see people accepting rational expectations theories, then the theory/behavior convergence that Levine talks about is not happening. Meaning that rational expectations theories may well be wrong.

Note that one easy way for rational expectations theories to be wrong is for the average person to simply be dumber than the people making the theories. For example, it's easy to reliably predict the mistakes of rats running in mazes, simply because rats aren't smart enough to learn our theories. Now, obviously, humans are not that far apart in intelligence, but if the difference between different humans is big, it could mean that behavioral theories could consistently be right.

(Does this make behavioral economics elitist? Maybe. But not necessarily. It could be that theorists are smart when making their theories but are dumb at other times. I can personally attest to the fact that this really does happen!)

In his second article, Levine asks how long it takes rational expectations to prevail, and cites the actions of the passengers on Flight 93 on 9/11 as an example of a quick rational response to new information (in this case, the info that hijackers would use planes as weapons). But does this mean that crowds are always rational? There is ample evidence for inefficient herd behavior in the psychological literature.

Anyway, Levine's logic in arguing for the supremacy of rational expectations is subtle and powerful. But I think reality is even more subtle, and even more powerful.