The Adaptive Market Hypothesis

An article in today's NYT led me to Andrew Lo's 2005 paper on the Adaptive Market Hypothesis. It's a nice summary of what's wrong with the efficient market hypothesis (EMH) and how an evolutionary approach offers new insights into how markets behave. It doesn't contain any new investment strategies, but it's worth reading nevertheless.

One piece of evidence Lo offers against EMH is the following graph of the correlation of the S&P index from one month to the next. Since the EMH argues that all information is already priced into the market, any market movement must be due to new information. And since new information is essentially random, market movement must be a random walk. In that case changes in the S&P from one month to the next should be uncorrelated, i.e., a correlation coefficient of 0. Here's how it actually looks.

First-order autocorrelation coefficients for monthly returns of the S&P Composite Index using 5-year rolling windows from January 1871 to April 2003. Data Source: R. Shiller, http://aida.econ.yale.edu/~shiller/data.htm.

As Lo points out,

As a measure of market efficiency (recall that the Random Walk Hypothesis implies that returns are serially uncorrelated, hence [the auto-correlation] should be 0 in theory), [the auto-correlation] might be expected to take on larger values during the early part of the sample and become progressively smaller during recent years as the U.S. equity market becomes more efficient. However, it is apparent … that the degree of efficiency (as measured by the first-order auto-correlation) varies through time in a cyclical fashion, and there are periods in the 1950's when the market was more efficient than in the early 1990's!