Eugene Fama has written an autobiographical account of his contributions to financial economics. Without doubt the man is a giant in his field. This is what he says about the EMH.
At the end of my second year at Chicago, it came time to write a thesis, and I went to Miller with five topics. Mert always had uncanny insight about research ideas likely to succeed. He gently stomped on four of my topics, but was excited by the fifth. From my work for Harry Ernst at Tufts, I had daily data on the 30 Dow-Jones Industrial Stocks. I proposed to produce detailed evidence on (1) Mandelbrot’s hypothesis that stock returns conform to non-normal (fat-tailed) stable distributions and (2) the time-series properties of returns. There was existing work on both topics, but I promised a unifying perspective and a leap in the range of data brought to bear.
Vindicating Mandelbrot, my thesis (Fama 1965a) shows (in nauseating detail) that distributions of stock returns are fat-tailed: there are far more outliers than would be expected from normal distributions – a fact reconfirmed in subsequent market episodes, including the most recent. Given the accusations of ignorance on this score recently thrown our way in the popular media, it is worth emphasizing that academics in finance have been aware of the fat tails phenomenon in asset returns for about 50 years.
My thesis and the earlier work of others on the time-series properties of returns falls under what came to be called tests of market efficiency. I coined the terms “market efficiency” and “efficient markets,” but they do not appear in my thesis. They first appear in “Random Walks in Stock Market Prices,” paper number 16 in the series of Selected Papers of the Graduate School of Business, University of Chicago, reprinted in the Financial Analysts Journal (Fama 1965b).
From the inception of research on the time-series properties of stock returns, economists speculated about how prices and returns behave if markets work, that is, if prices fully reflect all available information. The initial theory was the random walk model. In two important papers, Samuelson (1965) and Mandelbrot (1966) show that the random walk prediction (price changes are iid) is too strong. The proposition that prices fully reflect available information implies only that prices are sub-martingales. Formally, the deviations of price changes or returns from the values required to compensate investors for time and risk-bearing have expected value equal to zero conditional on past information.
During the early years, in addition to my thesis, I wrote several papers on market efficiency (Fama 1963, 1965c, Fama and Blume 1966), now mostly forgotten. My main contribution to the theory of efficient markets is the 1970 review (Fama 1970). The paper emphasizes the joint hypothesis problem hidden in the sub-martingales of Mandelbrot (1966) and Samuelson (1965). Specifically, market efficiency can only be tested in the context of an asset pricing model that specifies equilibrium expected returns. In other words, to test whether prices fully reflect available information, we must specify how the market is trying to compensate investors when it sets prices. My cleanest statement of the theory of efficient markets is in chapter 5 of Fama (1976b), reiterated in my second review “Efficient Markets II” (Fama 1991a).
The joint hypothesis problem is obvious, but only on hindsight. For example, much of the early work on market efficiency focuses on the autocorrelations of stock returns. It was not recognized that market efficiency implies zero autocorrelation only if the expected returns that investors require to hold stocks are constant through time or at least serially uncorrelated, and both conditions are unlikely.
The joint hypothesis problem is generally acknowledged in work on market efficiency after Fama (1970), and it is understood that, as a result, market efficiency per se is not testable. The flip side of the joint hypothesis problem is less often acknowledged. Specifically, almost all asset pricing models assume asset markets are efficient, so tests of these models are joint tests of the models and market efficiency. Asset pricing and market efficiency are forever joined at the hip.
I also enjoyed this bit from the foreword.
Finance is the most successful branch of economics in terms of theory and empirical work, the interplay between the two, and the penetration of financial research into other areas of economics and real-world applications.