Sunday, July 14, 2013

Altenatives to Rational Expectations

When Rational Expectations was developed in the 1960s most people thought it was a classic academic result.  Certainly mutual fund managers chuckled at the thought a monkey could outperform a skilled professional.  Further, everyone had a stupid relative, neighbor,  or co-worker, that proved people were not rational. Meanwhile, the Capital Asset Pricing Model (ie, Beta), meanwhile, was accepted as true even before data suggested it was.

Yet, we now have almost 100 years of data showing mutual fund managers do not outperform naive benchmarks (Cowles foundation (1932), Malkiel's Random Walk Down Wall Street (1973)), making the signature prediction of rational expectations surprising and true, a sign of a great theory.  Samuelson’s (1965) application of the law of iterated expectations to explain why Bachelier (1900) notices that securities prices look like a random walk, made for both elegant theory and voluminous evidence. The efficient markets theory reached its current plateau in 1970 with Fama's articulation of what efficient markets mean, but never has it been generally accepted by practitioners or non-economists, and there have always been many economists looking at exceptions to this rule.

The latest refutation of this plank of economics sounds very similar to a middle-aged economist reading this stuff:
Psychologists ... say that markets are not immune from human irrationality, whether that irrationality is due to optimism, fear, greed, or other forces.... "There's this tug-of-war between economics and psychology, and in this round, psychology wins," says Colin Camerer, the Robert Kirby Professor of Behavioral Economics at the California Institute of Technology...
What did they show? That people basically make inconsistent bets when presented with subsets vs. supersets, for example, the odds that Hillary Clinton (Democrat) will win the Presidency is less than the probability a Democrat will win the Presidency, though in aggregate surveys this doesn't always hold.

It's been long known that people over-estimate specifics based on things like the availability heuristic, as when people more easily think an introverted woman is a librarian than an accountant because it fits with what one can think about as an archetype, and this has been discussed since the 1970s (see Kahneman, Tversky and Slovic's Heuristics and Biases). This is patently illogical, and spawned the behavioralist revolution, where we have about 100 such biases.

Yet, it's one thing to say people are inconsistent, another to say liquid markets are. While a random survey of retail investors might have the equity premium at something silly like 10% or the odds of a US default at 5%, that doesn't mean derivatives are priced that way, because knowledgeable people with capital tend to arbitrage these beliefs out of the market. For example, Snowberg and Wolfers showed that horse races are illogical looking at millions of races, and this has been known since 1949 (note: the favorite has the highest expected returns). Yet, to the extent such prices reflect this inconsistency, it isn't large enough to make large amounts of money. Does this translate to anything about the general market, or tradable stocks like those in the Wilshire 3000? Almost always, no.

These findings strike me a bit like this finding listed in a eulogy of this Harvard psychologist who sounds like just an awesome mensch, where they noted that
while teaching at the University of Virginia, he devised an experiment to study secrecy and obsessions. Enlisting college students to play card games at a table, he instructed some to play footsie under the table and tell everyone they were doing so. Others had to keep their footplay secret. One result? “The subjects who made secret contact with their partner were significantly more attracted to their partners,” Dr. Wegner told the Globe in 1994, a finding that, among other things, shed light on the allure of affairs outside relationships.
Now, that is very interesting, but it doesn't really contradict any fundamental conception of humanity, and so it goes with the behavioral finance results: neat, but not profound.

The problem with the alternatives to Rational Expectations is that they generally point out that in low-stakes environments people make logical mistakes in predictable ways, but they hardly ever explain the yield curve, the equity premium, or anything else important, but rather, the poor returns to 100-1 horses or some other curiosity. I like betting on horse longshots at the track and implicitly pay a premium for that, but it probably costs less than what I spend on beer at the track, so it's really not complicated.

Now, one might think that evolutionary biology provides a fruitful alternative, in that it tries to look at predictable expectations from the standpoint of its evolutionary success. I think that's great, and indeed, think that a relative utility function makes more sense because, among other reasons, it is more evolutionarily robust.

 David Sloan Wilson, who is well-known for showing how religion is a 'rational' adaptive solution within a multi-level selection process, has organized a special issue of the Journal of Economic and Behaviour Organisation entitled ‘Evolution as a General Theoretical Framework for Economics and Public Policy’. A lot of those articles are interesting, but none really that radical.

For example, there's the article arguing economists should use multi-level selection, as opposed to the atomistic selection implicit within abstract markets. That is, in evolutionary biology, there's a big debate about where selection occurs. Before Richard Dawkins became fixated on the specter of Christian Fundamentalists, he wrote The Selfish Gene to rebut the idea that selection occurs at the level of a species, championing the ideas of George C. Williams and John Maynard Smith. These biologists noted casual speakers would talk about 'what is good for a species', but clearly gazelles don't act as a group, but as individuals, and within individuals, as cells, and then as genes.  Dawkins argued that it's selfishness at the lowest level, the gene, ergo the title. Things have gotten rather complicated since, and it's clear selection takes place at more than one level.

For example, take the Hymenoptera order, which includes ants. There's a joke about socialism: good theory, wrong species (ie, ants not humans). For these species their genes actually assist in their goupishness, and so they willing kill themselves for the group all the time, because sisters are more related to each other than a queen is to her offspring. Thus, the selection for haplodiploidy could take place at one level--the swarm--not the gene level, and this then affects the payoffs for individuals and thus genes. Selection seems to occur at the lowest level, the genes, but also cells, the organism, and among humans, in culture. Economist Herbert Gintis has argued that societies that promote pro-social norms, as in group selection, have higher survival rates than societies that do not.

I think that's all super, and think some societies might have different intrinsic levels of individualism depending on their evolutionary environment. For example, here's nice discussion about how the Western concept of marriage outside of extended families led to a unique level of individualism, something that is rather unique to Western Civilization. In contrast, in the middle east, a lot of people marry cousins, and this breeds greater groupishness because your extended family is so obviously genetically oriented, making outsiders very clear (ie, altruism towards one's clan, indifference at best towards those outside it).

Yet, it's not clear how much these insights are really new as applied to markets and theories of the firm. Many theories apply to how oligopolies and monopolies behave, and these are quite different than what one sees for perfect competition, so the idea suggested in Wilson's Special Issue, that selection occurs at the industry, firm, or worker level, is not that revolutionary. Game theory often looks at mechanism design, and the stability of coalitions, and the conditions of various equilibrium.  I get the sense they think the only behavior economics looks at is perfect competition.  It's not, and hasn't been for 100 years.

It's generally accepted that one needs the system to obey two constraints:

  1. Individuals must receive a positive return to their action
  2. Individuals must be doing their best given their information and beliefs

These are sensible restrictions, so the issue is whether behavior is rational or not, from the agent's perspective. Rational expectations merely adds the requirement that any behavior should be consistent with zero abnormal profits for the simple and sensible reason that it is very difficult to generate abnormal profits in most markets. I still think that's a good bias, because it's usually true.

1 comment:

Lauren Taylor said...

Rational expectations were forcefully argued by Lucas in the 1970s. However, 40 years have passed and there are many alternatives to rational expectations. Lucas' argument has been broadly implemented because of its simplicity. It typically yields a unique solution, but that is far from implying that is it is most realistic explanation. With the ever-increasing computer power for numerical methods, I expect rational expectations to loose ground to alternative explanations.