Grasping Reality on TypePad, by Brad DeLong

...Unfortunately, at this point, a piece of misnomenclature occurred, and the ‘Random Walk Hypothesis’ (ie, the hypothesis that securities prices[5] are a random walk process) got renamed the ‘Efficient Markets Hypothesis’. At least at this point, it was still called a ‘Hypothesis’, which gave at least a clue that it was an empirical claim about the statistical properties of securities returns rather than a necessary truth about underlying reality--the ‘Efficient Markets Theory’ (and, god help us, ‘Theorem’[6]) was yet to come.

Although he was one of the parties to the mis-re-naming, Eugene Fama did a lot of really fundamental work in sharpening up the concept of what it might mean for securities prices to fluctuate ‘randomly’ [7]. In particular, his weak, semi-strong and strong forms of the Efficient Markets Hypothesis started the ball rolling with respect to thinking about what sort of information one should be conditioning on, when carrying out the statistical tests for a random walk. On the basis of his own research – which was very good at the time, albeit that as time and science moved on, it became apparent that the standard statistical tests of the day were pretty low in power and tended not to be very good at rejecting the hypothesis of a random walk[8] – Fama concluded that the basic answer to the question was that as far as anyone could tell, and certainly to the extent of being able to profit from them, securities prices were random and anyone charging money for the service of being able to predict them was probably lying.

This was described, rather embarrassingly, as ‘the best established proposition in social sciences’, in 1978 by Michael Jensen, who was one of a large crew of mainly American academics who kind of picked up this ball and ran with it, as we will get into discussing later.

At the start of the 1980s, though, Shiller kind of put a bomb under the EMH, by attacking it from the other side. Although as I noted above, ‘’Properly Anticipated Prices Fluctuate Randomly’ does not imply ‘Randomly Fluctuating Prices Are Properly Anticipated’, the reverse implication is valid – if prices could be demonstrated to not properly anticipate the changes in the underlying cash flows they represent claims on, then it could be demonstrated that they weren’t wholly random, as well as blowing up the larger intellectual project of ‘market efficiency’ that had been built on the foundations of the Random Walk Hypothesis.

Shiller’s 1981 paper is pretty conceptually simple[9]. Given that securities prices (particularly share prices) are meant to be based on the present value of a stream of future dividends, and given that dividend streams are not really all that volatile, why is it that share prices go up and down so much? Shiller showed that in order to believe that prices properly anticipated the risk-adjusted discounted value of future cash flows, you would have to believe something pretty implausible about the unobserved parameters (the rate of risk aversion and/or time preference) and that it was considerably more appealing to believe that the market simply and constantly got it wrong.

I like to think that, as the majority of CT readers stroke themselves with pleasure at the thought of ‘Markets are not efficient after all! I was right to do that humanities degree!’[10] and mark down Robert Shiller as A Good Thing to be counterposed to Eugene Fama who was A Bad Thing [11], a small minority of like-minded souls will be thinking ‘hang on, tell me about that ‘stock prices not random’ thing?’.

Indeed my friends. If stock prices have ‘too much’ randomness, then they overreact in both directions. Specifically, if the Shiller Cyclically Adjusted Price Earnings ratio (or price-dividends, or whatever) is historically high, then it will tend to fall and vice versa if it is low. This one works. It’s by now a really quite well-established fact about securities prices.

In fairness to Fama, his ‘round two’ contribution to this debate showed that he is a good empiricist at heart[12], as he spent most of the 80s and early 90s doing work with Kenneth French on testing what kinds of ‘anomalies’ or sources of predictability could be explained away as artifacts of the data or statistical quirks, and which were real and persistent effects that had to be taken into account. In the end, he concluded that before saying that stock prices fluctuated randomly, one had to condition on factors which might include company size, ‘value’ (book to market ratio) and even ‘momentum’ (shares which have performed strongly in recent periods do seem to have a tendency to continue to outperform which can’t fully be chalked up to statistical noise).

For reasons I don’t fully understand [13] ([13] For the longest time, the view from Chicago was that the size, value and momentum effects might be proxies for as yet unknown sources of risk, for which the associated excess returns were fair compensation. This was never disproved, but as far as I can tell the Chicago guys just got tired of getting laughed at and kind of gave up on this theory. Nowadays as I understand it, the view when pressed is that the “anomalies” are genuine features of the dataset which have to be taken into accounting when testing theories on historical data, but not really part of the underlying model, and maybe they will disappear when we have ten thousand years of equity returns to deal with. At least it’s an ethos.), Fama still thinks that a theory which allows for these factors is still worth calling a theory of ‘efficient markets’...

As Martin Wolf told me once, the British phrase "for reasons I don’t fully understand, X thinks that..." = "X is truly a nutcase and thinks that..." in American...