Efficient Markets

Historically, the origin of the audacious finance project can be traced to the much more modest idea that stock market prices follow a random walk. Initially little more than an empirical characterization of the data, the “random walk theory” developed into a powerfully evocative theoretical construct called the efficient markets hypothesis (Fama 1965, 1970, 1991), and then even further into a theory of the economy as a whole.

The efficient markets hypothesis originated as an attempt to explain a startling empiri­cal fact. If you draw a chart showing how the price of a stock (or stock index) has moved over some period of time, it will typically bear a striking resemblance to similar charts showing the movement of variables that we know to be random, such as coin flips. If you keep track of a series of coin flips, counting heads as +1 and tails as -1, the cumulative sum will go up and down in a way that is hard to distinguish from the movement of a stock price. Harry Roberts (1959), a statistician at the Graduate School of Business at the University of Chicago, convinced himself of this “random walk theory” by showing charts of stock prices and cumulative coin flips to stock market analysts, who could not tell the difference.

This new statistical way of thinking about the data focused attention on two questions: the independence of successive price changes over time, and the shape of the probability distribution of price changes at a point in time. For most economists, the independence question seemed the more important of the two, since statistical independence would imply zero expected speculative profit, which economists recognized as a characteriza­tion of equilibrium in competitive markets. Thus the random walk theory, initially only a statistical characterization of the data, came to be endowed with economic content, got renamed as the efficient markets hypothesis (EMH), and then adopted as the centrepiece of a distinctive Chicago-style approach to finance (Fama and Miller 1972).

The random walk formulation of EMH was soon shown to be not entirely correct (Samuelson 1965, Mandelbrot 1966); the correct statistical consequence of assum­ing zero expected speculative profit was a “martingale” not a random walk. What is important is not that successive increments are statistically independent, but only that current price is the conditional expectation of future price. As Samuelson stated it (under the simplifying assumption of a zero discount rate):

Notwithstanding this correction - and notwithstanding also the branding effort of Cootner’s Random Character of Stock Prices (1964) which sought to establish a dis­tinguished ancestry for this approach in the person of Louis Bachelier (1900; Davis and Etheridge 2006) - the EMH became the brand of Chicago rather than of the Massachusetts Institute of Technology (MIT), if only because Chicago wanted the brand while MIT did not. At MIT, efficiency was merely an idealization, unlikely to hold reli­ably in the real world, hence not an attractive null hypothesis and certainly not a useful maintained hypothesis (Mandelbrot 1971).

This intellectual split, right at the origin of the modern field, would have continuing reverberations throughout the subsequent decades, as fledgling finance grew to be an enormous field in its own right. One immediate consequence showed itself in dispute over the proper interpretation of statistical tests of asset pricing theories such as CAPM. Such tests were always joint tests of both the theory and the efficient markets hypoth­esis, so negative results could be met either by rejecting CAPM or by rejecting efficiency (Roll 1977). Chicago reliably chose the former, while MIT reliably chose the latter (Shiller 2000).

From this point of view, the importance of the Samuelson-Mandelbrot results was not so much to correct a technicality in the mathematical formulation of EMH, but rather to endow the formulation with a deep economic interpretation, namely, no arbitrage. Once no arbitrage was placed at the centre of finance, the new field could grow essentially unhindered by any further intellectual inheritance from economics, no utility curves or production functions, no supply and demand. So it did, for more than a decade.