Risk and Return Confabulations

Jonathan's Haidt's book, The Happiness Hypothesis, spends a lot of time on the idea that the conscious mind is like the rider on an elephant, and while he moves the elephant somewhat, the elephant has a will of its own that is more powerful. A solution to this problem for the elephant rider, is that when the elephant goes somewhere you don't want to go, you just convince yourself you wanted to go there all along. This is called confabulating a rational.

I note this because I read an excellent confabulation from the recent AFA meetings in San Fran, Uncertainty about Average Profitability and Diversification Discount (by Hund, Monk and Tice). Remember that risk generates a return premium: higher risk generates higher return. Of course, risk must be specified correctly, but whatever it is, it generates high returns. So, how to explain the fact that diversified firms, those with higher profitability, lower volatility of profitability, and lower idiosyncratic return volatility, have higher-than-average returns? How is this consistent with risk and return theory?

Well, I think their answer is as follows: diversified firms have a lot of incentive (aka agency) problems. There is a lot of cross subsidization, lots of allocation of common costs, lots of finger pointing. Thus, the return premium is from when these firms basically get taken over, and become focused again. Then they don't cross subsidize value-destroying projects.

They contrast this with the model of Pastor and Veronesi, which is an even sillier model. These guys assert that firms grow geometrically, and so given the logic of pure mathematics (E[firm value]=exp((mu+s^2/2-r)*t)), the more expected volatility, the greater the expected value, the higher the price, and lower return.

Now, these models are only tenable because they were presented with a straight face using the presentation protocols of the field. They suggest a massive amount of market inefficiency, but by noting this is 'rational learning' hope nobody notices that people are not anticipating what they are going to learn. At least behavioralists like Lakonishok and Thaler say 'people overextrapolate'; here, they are trying to sell an irrationality story in a convoluted way while maintaining the veneer of rationality.

Risk used to be standard deviation. Then Beta. Then APT factors. Then Stochastic Discount Factors. Now risk is just out of the picture, and we have people learning about the future stumbling in the dark--but you don't want to upset the financial theorist's applecart too much, so it is implicitly assured that risk models still work, of course. The spooky SDF will be revealed when the messiah arrives, I guess (of course, I think risk-return theory is wrong--there is no return for risk in general, however defined).