What are some theoretical explanations for deviations from the efficient market hypothesis?

Question

According (the semi-strong form of) the efficient market hypothesis, the price of an asset should reflect all publicly available information about the 'fundamental value' of that asset. The reasoning is that if there is some public information that suggests the price is below the fundamental value then people will have an incentive to act on this information and buy more of the asset. This increase in demand will cause the price of the asset to rise, closing the original price-value discrepancy.

This reasoning seems intuitively plausible, but significant price-corrections during the recent recession seem to suggest that markets had systematically priced in a manner inconsistent with underlying fundamentals. What academic theories (i.e. published or in working papers) have been advanced (and by whom) to explain why the efficient market hypothesis might fail? Is there much of an empirical consensus on the validity of the efficient market hypothesis in the literature?

Answer 1

There is a position that the efficient market hypothesis is essentially untestable. That's because purported tests of efficient markets are actually joint tests of two claims

1. Market X is efficient
2. An efficient markets looks / behaves like Y

A rejection of a market X as not behaving like Y could be because market X is efficient but it is false that efficient markets behave like Y or that X is not an efficient market. The earliest citation I know on this issue is Jensen (1978). There are numerous reasonable examples of how this can occur. Consider violations of the Consumption-CAPM. (Grossman and Shiller (1981)). A rejection of the C-CAPM can occur simply because you have failed to identify the proper stochastic discount factor or market portfolio. Or consider a superficially reasonable assumption that returns should not be predicable if markets are efficient. But if there is aggregate (non-diversifiable) risk and there are limits to the ability to move income through time (like in a Lucas-tree economy) then when aggregate output is relatively low assets will be relatively cheap even if that low output was entirely predictable.

I suspect these problems will perpetually impair consensus on the validity of the efficient market hypothesis. A related but unasked question is "Is it useful to assume that markets are efficient?" On this matter, there does seem to be a strong consensus that:

1. Prices in big, liquid markets are often very good and hard to beat
2. For models where asset prices or financial wealth are inputs to asking other questions, assuming markets are efficient is a good starting point and often good enough.

All that about untestability said, people try to test market efficiency all the time. Most of these tests fall into the family of "market anomalies". Essentially, someone comes up with an investment portfolio strategy that provides a arbitrage opportunity. Here arbitrage has a precise meaning that differs from common usage in finance. An arbitrage requires no cash outflow now or in the future and there is at least one state of the world in the present or future where it generates strictly positive value. The existence of such an anomaly seems plausible evidence that markets are not efficient. However, in practice, it isn't clear that the famous anomaly papers papers actually demonstrate this. What they more typically demonstrate is that fluctuations in some representative but flawed measure of the market portfolio do not explain all returns and therefore that one might be able to construct zero up-front cost a portfolio with no market risk exposure that nevertheless pays out in some states. But such portfolios almost always have negative outcomes as well as positive ones and so are not strictly speaking arbitrages.

An alternative explaination is that there are other risks that are priced besides the market. That's why the rightfully famous and careful Fama and French (1993) is called Common risk factors in the returns on stocks and bonds. But when we really dig into the weeds of what sorts of "risk factors" generate returns, some don't mesh to any clear source of economic risk. That firms with heavy recession exposures should command higher returns make sense. That firms should outperform in some months and not others is much harder to understand through the lens of risk and some prefer to think of as evidence of market inefficiency.

As for why the efficient market hypothesis might fail, there seem to be a few key stories often combined into "the limits of arbitrage". A few examples follow. Informed traders may lack the capital to take large enough positions so that market prices can reflect their information. Markets may not exist to hedge some risks. If risk factors A and B cannot be traded separately and you have a signal on just A it may be impossible to get risk factor A priced properly. Market participants may be risk adverse and not want concentrated exposures in unusual, under-priced risks, and so not have an incentive to make markets efficient. Transaction costs of trade may prevent the profitable exploitation of information.

Answer 2

I think this piece by John Cochrane represents a typical defense of the efficient market hypothesis from this kind of critique. Here's another. The basic summary is that market efficiency does not imply that expected returns cannot vary or that bubbles will not happen. Bubbles should be unpredictable, and variation in expected returns should be driven by risk or discount factors. Whether that's the case for the financial crisis is definitely a matter of debate.

Do people believe in efficient markets? I think that varies a lot from field to field. It might be more useful to say that efficient markets are usually the null or prior that academics in finance and economics work with when they try to understand something. That is to say if you want to use some data to try to convince them that financial markets aren't efficient, you face an uphill battle, but the struggle isn't impossible.

Answer 3

Yes, there is a simple and well accepted doctrine of why EMH fails - it is impossible, so long as arbitrageurs do not work for free, and firms only pay wages for marginal product. The argument was first formalised by Stiglitz and Grossman (1980), in the American Economic Review.

Consider that all information on the fair price of financial assets is costly to obtain, both in capital (technology) and in the disutility of human effort. To determine the impact of new information on the fair value of a bond, it may be costly to build a model of the economy, or the distribution of probabilities of the Federal Reserve's current reaction function. To determine a revision in company estimates, it is costly to research the sector outlook.

Firms pay individuals in order to find these discrepancies and use their capital to profit from them. Their pay should reflect their marginal product under perfectly competitive labour markets. Under perfectly informationally efficient markets, why will firms pay them?

Now consider a deviation from this equilibrium. If some exogenous shift in technology or market strategists' human capital causes markets to become perfectly informationally efficient, some rational firms will fire their analysts, and choose to be uninformed. As experts leave the market, that informational efficiency is lost. For this reason, EMH cannot be supported as an equilibrium.