In comments, Mike linked to one of his posts from a couple months ago where he quotes William Easterly:
The most important part of the much-maligned Efficient Markets Hypothesis (EMH) is that nobody can systematically beat the stock market. Which implies nobody can predict a market crash, because if you could, then you would obviously beat the market. This applies also to other asset markets like housing prices.
Mike goes on to describe a paper called The Limits to Arbitrage. Under the realistic assumptions that only big players arbitrage and there is an agency problem where those big traders have information about fundamental prices that the people that hire them do not ((Realistic assumptions, yes, but how sensitive are the results to these assumptions? and how would they interact with other behavioral assumptions (e.g. why did principals hire the agents in the first place)? what happens as the arbitrageur uses more and more of his own money?)), even these big traders can’t stay solvent long enough. Thus, Easterly’s contention is false.
I don’t see the connection. Nothing in that paper suggest market prices don’t incorporate all available public information. Yes, arbitrageurs know more than the public. If their principals allowed it, they’d make trades on that information, but they don’t so they can’t. There are no profit opportunities.
The intellectual history of the EMH is basically:
- Fama names the EMH
- People test EMH making assumptions about asset pricing (e.g. CAPM)
- Early tests don’t reject EMH, but later tests do
- Fama observes these rejections are of the joint test of the asset pricing model and EMH… the problem could be with the asset pricing model
- People try different pricing models and discover there aren’t any good ones (i.e. the joint test keeps getting rejected)
- The Limits of Arbitrage suggests the problem is with liquidity constraints
- People again say this is a problem with EMH
- Cue Fama…
Its funny. I’m beginning to think that EMH isn’t a testable hypothesis. Its a subsidiary hypothesis, like in cosmology the assumption that the physical constants are in fact constant, that makes it possible to test other hypothesis. Without the EMH, we can use any ad-hoc explanation to explain any particular behavior. Under this view, anomalies teach us about asset pricing models and not the rationality of markets.
True, Fama’s work depended on arbitrage having no limits, but it should be easy (as in, not easy at all) to redefine efficiency to include the liquidity constraints of arbitrageurs. Then we can say markets are constrained efficient. I have no idea what this would mean for the allocation of capital in the real economy (the paper points out that the extent of the agency problems vary by asset class but I don’t imagine asset classes are much correlated with “capital classes” or sectors of the real economy). And policy implications?