Monday, September 28, 2009

On Keynes, Lucas, and Rationality

. Monday, September 28, 2009

Brad DeLong slaps Robert Lucas across the face. Hard.

Lucas seems to miss the entire big point. Black-Scholes tells you how to do things only if the trades of the non-von Neumann-Morgenstern agents in the economy--that's you, Bob--cancel each other out. If they don't cancel each other out--if there are, as you say, "millions of people like me," then a whole bunch of banks running off of Black-Scholes and similar models without taking account of your existence are going to create a hell of a lot of systemic risk, and then 10% unemployment.


Ouch. For full context, read the full post (not long). A "von Neumann-Morgenstern agent" is the same creature as the one often called Homo Economicus: fully rational, forward-looking, risk-neutral, and possessing something close to perfect information (in aggregate, at least). And if you like that, lookit what DeLong did to Ed Prescott.

In truth, however, Lucas acted rationally in the situation DeLong is referring to. A von Neumann-Morgenstern agent would sell off his equities sometime last fall and buy Treasuries, and then convert back to equities as soon at the equity markets stabilized and started recapturing their lost value (remember all the Wall Street talk about "bargain buys"?). In other words, when equities have a negative expected value over the short run but a positive expected value over the medium-to-long run it is perfectly rational to hold cash in the short run but equities in the medium-to-long run. This is even more true if the short run expectation is for zero (or even negative) inflation but the medium-to-long run expectation is for higher inflation (as DeLong suggests). There is no indication that Lucas, and/or the millions like him, did not do precisely that. Actually, the fact that the S&P 500 is now ~ 35% above its 2009 low is evidence that that is exactly what happened.

If this were not true, then all of Keynes (except maybe chapter 12) would be wrong. All of Keynes' major insights -- the importance of aggregate demand, the paradox of thrift, the difference between short-run and long-run outcomes, the danger of the liquidity trap -- depend on rational expectations. Without rational expectations there is no Keynesianism. The whole point of Keynes is that short run expectations can be different from long run expectations, and that both are important. A Keynesian like DeLong ought to be able to make that distinction.

The problem with Black-Scholes, e.g., isn't that it depends on rational agents; the problem is that it depends on historical volatility. In a panic, historical volatility is basically meaningless: the basic feature of a panic is that we're not operating in the fat middle anymore, but rather in the thin tail. Unfortunately it's impossible to predict future panics with any model based on history, since they are irregular events outside the historical norm. You can't model outliers (much to Alex's dismay). That's why it's an outlier.

Keynes doesn't help us here: he never tried to model outliers, nor did he encourage others to try. Instead, he proposed some mechanisms for trying to cope with them. It's like a researcher developing a treatment for a virus: she's not trying to find a vaccine; she's trying to find ways to deal with the disease after it's taken hold. Sure, the virus may eventually go away on its own as the immune system fights back, but why wait if we can help the process along? That's what Keynes was trying to do: not prevent the disease, but rather speed up the recovery.

DeLong thinks that Keynes' prescriptions work, and they are worth the cost in terms of debt and (possible) future inflation. Lucas (and Prescott) thinks those prescriptions don't work, or even if they do they aren't really worth the cost since fiscal policy tends to only be counter-cyclical in the valleys. That's the argument, and each side has empirical studies supporting their views.

But let's not make this about rationality. It isn't.

0 comments:

On Keynes, Lucas, and Rationality
 

PageRank

SiteMeter

Technorati

Add to Technorati Favorites