Steve Reads

In December of 2003 (wow, I’m shocked that it was that long ago; I had guessed it was maybe two years ago), I was in the middle of reading Richard Posner’s Frontiers of Legal Theory, and quoted his humorous line:

Behavioral economics is defined by its subject rather than by its method and its subject is merely the set of phenomena that rational-choice models (or at least the simplest of them) do not explain. It would not be surprising if many of these phenomena turned out to be unrelated to each other, just as the set of things that are not edible by man include stones, toadstools, thunderclaps, and the Pythagorean theorem.

While I still believe, as I did then, that this is a really funny quote, my reading since then has demonstrated beyond a reasonable doubt that Posner is being remarkably unfair — as in, if he’s actually the brilliant man that everyone claims he is, he was either being sloppy in his research or deliberately misleading. That book came out in 2001. Daniel Kahneman, Amos Tversky, Richard Thaler and others had long since lain the foundations for behavioral economics, which Posner tears into here; Kahneman won his Nobel a year after Posner wrote his book, and the Nobel committee isn’t noted for its speed at acknowledging brilliance.

To me the most damning critique of (Posner’s and many others’) economic orthodoxy is that the most important decisions we’re forced to make are those which must happen on extremely rapid time scales — namely, we’d have to solve them quickly to avoid dying. In order to be the sort of utility maximizers that Econ 101 teaches us that we are, we’d often need to solve tremendously complicated problems — problems that are provably hard to solve. Yet the standard economic models would have us solve these in a split second. Either the models are off, or they imply some very strong claims about how our brains are built.

Arguing the first option in sardonic detail is the job of Richard Thaler’s The Winner’s Curse, a book aimed at someone with not much training at all in economics, but with a decent intelligence and an open mind. In each of 15 brisk chapters, Thaler asks a single question and then dives in to study how behavioral economics has tried to answer it. The questions include

• Why do people who do the same jobs in different fields (say, a secretary in the mutual-fund industry and a secretary in the steel industry) get paid such different amounts? Standard economic theory predicts that they should be paid the same, or an arbitrage opportunity would develop and would quickly disappear as workers gravitated to the higher-paying industries. Over time, all wages for the same job would level out.

• People don’t save money the way that classical economics would predict; they tend to overweight present income in their lifetime savings plans. Why is that?

• If the efficient market hypothesis is correct, then the stock market already reflects all the known information about a given company. So why do companies take over other companies at a substantial premium over the market price of the stock? If the EMH is right, then those who win takeover fights will systematically overpay and will regret their purchases in the not-very-long term. This is known as the “winner’s curse,” whence the title of the book.

• Economic theory — Franco Modigliani’s, in particular — would predict that a dollar saved in a checking account is exchangeable with a dollar saved in a pension or in home equity, yet the data would say otherwise; according to Thaler, people behave more as though they divide their money into something like three buckets, with one bucket very illiquid (home mortgages), one very liquid (checking), and one whose liquidity is in between (savings). They do not treat these buckets as exchangeable.

Thaler’s arched eyebrow toward his rationalist colleagues is most evident in his description of the Modigliani model (beginning of Chapter 9):

The standard model of saving in economics, for which Franco Modigliani won a Nobel Prize, is called the life-cycle theory. It is a classic bit of economic theorizing. First, it specifies and solves an optimization problem. Then it assumes that people act as if they had solved the same problem. Here it is assumed that an individual has no interest in leaving any bequests, and values consumption equally in every period. How much should such a person consume in a given year? The answer is this: in any year, compute the present value of financial wealth, including current income, net assets, and the expected value of future income; figure out the level annuity that could be purchased with that money; then consume the amount that would be received from such an annuity. The theory is simple, elegant, and rational — qualities highly valued by economists. Unfortunately, as Courant, Grimlich, and Laitner observe, “for all its elegance and rationality, the life-cycle model has not tested out very well.”

(footnote and citation omitted)

This is not, of course, an attack on the idea of theorizing. Theories are useful. You just need the right theories. It’s also not an attack on simple theories or incorrect ones; a mistaken theory may be a step on the way to a correct one, if only because its mistakes are clearly visible and it suggests routes for future inquiry. Thaler’s point throughout The Winner’s Curse is that it’s not always clear what use current economic theories serve. As a description of how humans do behave (the “descriptive” model), they’re clearly failures. As descriptions of how humans ought to behave (the “normative” model), they’re also questionable.

There’s an interesting question lurking beneath all of this, about how one constructs scientific models in general. The oft-quoted line by Milton Friedman and L.J. Savage (from “Utility Analysis of Choices Involving Risk”) summarizes the point:

Consider the problem of predicting, before each shot, the direction of travel of a billiard ball hit by an expert billiard player. It would be possible to construct one or more mathematical formulas that would give the directions of travel that would score points and, among these, would indicate the one (or more) that would leave the balls in the best positions. The formulas might, of course, be extremely complicated, since they would necessarily take account of the location of the balls in relation to one another and to the cushions and of the complicated phenomena introduced by “english.” Nonetheless, it seems not at all unreasonable that excellent predictions would be yielded by the hypothesis that the billiard player made his shots as if he knew the formulas, could estimate accurately by eye the angles, etc., describing the location of the balls, could make lightning calculations from the formulas, and could then make the ball travel in the direction indicated by the formulas. It would in no way disprove or contradict the hypothesis, or weaken our confidence in it, if it should turn out that the billiard player had never studied any branch of mathematics and was utterly incapable of making the necessary calculations: unless he was capable in some way of reaching approximately the same result as that obtained by the formulas, he would not in fact be likely to be an expect billiard player.

The only test of a theory, say Friedman and Savage, is whether the data match the predictions. If we get good physics predictions by assuming that we’re handling point masses, so much the better; the theory is that much more manageable. Likewise if we assume that people behave as perfectly rational, selfish optimizers.

The trouble is, though, that we realize where we’re making simplifications; we must be aware that sooner or later the assumptions are going to poke through — point masses just can’t last forever, and purely rational humans will stop being a useful fiction. If we take Friedman to be making the rather modest claim that “they’re a useful fiction for now, but clearly that can’t last,” then that might be easier to swallow. As it is, he doesn’t seem that modest.

Thaler’s way of summarizing all of this is nicely smirky:

It is also interesting to note a peculiar tendency among many economic theorists. A theorist will sweat long and hard on a problem, finally achieving a new insight previously unknown to economists. The theorist then assumes that the agents in a theoretical model act as if they also understood this new insight. In assuming that the agents in the economy intuitively grasp what it took so long to work out, the theorist is either showing uncharacteristic modesty and generosity, or is guilty of ascribing too much rationality to the agents in his model.

(Internal citation omitted. I feel like quoting A Sampling Of Mathematical Folk Humor here, just because it has a similar ring:

A mathematics professor was lecturing to a class of students. As he wrote something on the board, he said to the class “Of course, this is immediately obvious.” Upon seeing the blank stares of the students, he turned back to contemplate what he had just written. He began to pace back and forth, deep in thought. After about 10 minutes, just as the silence was beginning to become uncomfortable, he brightened, turned to the class and said, “Yes, it IS obvious.”

)

The only complaint I’d register against Thaler’s work — and since it’s seemingly a complaint about an entire academic discipline, it’s likely my problem rather than Thaler’s — is that too much of the data rests on asking a) undergraduates about b) hypothetical situations. E.g., how much would you pay to delay receiving a 110-volt shock by three hours, one day, three days, one year, and ten years (p. 104). This seems like exactly the sort of question that I couldn’t be expected to answer in a survey. If I did answer it, it’s likely that my answer would bear little relation to how I’d actually behave if I were being threatened with real electrocution. From time to time Thaler’s evidence does come from the field, for instance when he studies the winner’s curse in auctions of potential oil fields, or looks at people’s refusal to spend money up front for home insulation when it would clearly pay for itself in less than a year. The trouble with data from the field, of course, is that they will be noisy and will need a lot of work to clean up; undergraduate data, by contrast, are clean even if they don’t necessarily say much. Thaler, Kahneman, Tversky et al. surely know experimental methods better than I do. All I can say is that imprecise data from the field seem far more convincing than clean data from the lab.

All in all, I couldn’t ask for a more thought-provoking book with more provocative answers than this one. Thaler’s style should appeal equally well to professional economists and to educated laymen. Highly recommended.

Daniel Kahneman’s autobiography, written on the occasion of his winning the Nobel Memorial Prize in Economics, makes for great reading. I was particularly taken by his idea of “adversarial collaboration,” which you can find maybe 90% of the way into the autobiography. The idea is that there’s something quite defective in the way that scholarly publishing happens: someone presents an idea, a critic thrashes it, and the original author responds by giving absolutely no ground to his critic. This goes back and forth for a few rounds; at no point was anyone willing to change his mind, and indeed the process encourages further entrenchment.

I can’t testify to the truth or falsehood of that claim for academic scholarship, but it’s certainly true of public debating in this country. I even noticed it, on the smallest scale, on the debate team in high school. We lost points for ceding any point to our opponent. Any debates I’ve had on the web or mailing lists have followed the same pattern. When you’re competing in front of others, any point of contention on which you compromise counts as a loss. Kahneman’s attempts to find a process that avoids this whole unproductive mess are laudable.

His description of a years-long paper-writing process smacks against what some of my college professors told me about academic publishing. They complained that publication is so competitive that you’re encouraged to put out a half-baked or small idea, rather than taking the time to let the idea blossom into a deep, complete paper. I’m curious whether academics have changed since Kahneman and Tversky worked on their foundational papers, whether disciplinary standards are different in their field, or whether Kahneman is just leaving out the intermediary steps — say, subsidiary papers that he published along the way.

Finally, it’s interesting to read Kahneman’s account of his fight with Gerd Gigerenzer. Gigerenzer’s book Simple Heuristics That Make Us Smart came highly recommended from Cosma Shalizi, and it didn’t disappoint; it’s a sustained argument, backed by lots of experimental evidence, that many mental processes which may seem initially irrational turn out to be “ecologically rational,” in the sense that they work well under the variable conditions that we find in nature. Whereas many results about optimality or economic rationality depend on rather precise conditions holding, ecologically rational ones are not sensitive to violations in the assumptions.

In particular, Simple Heuristics sets itself firmly against Kahneman and Tversky. The latter are famous for having introduced “heuristics and biases” — i.e., places where human behavior is irrational, in a way that makes even the experimental subjects themselves blush. I only have the barest connection to the debate, but Gigerenzer seems to believe that Kahneman and Tversky have given heuristics and biases a bad name, whereas he (Gigerenzer) believes they’re ecologically valuable. Kahneman, in turn, seems to believe that he’s been given a bad rap, and that heuristics and biases shouldn’t be viewed all that negatively.

What’s more interesting about the debate, it seems to me, is that it illustrates the point that Kahneman was making about adversarial collaboration; I’m pretty sure this is one of the main reasons he included the debate with Gigerenzer in his autobiography. At no point does it seem as though Gigerenzer is willing to budge in his debate with Kahneman, because his academic credibility depends on his winning a pissing contest. Kahneman is in a nice spot writing this autobiography, in that he’s just won a Nobel and can look like the high-road-taking elder statesman of psychology, and can cast Gigerenzer as an uncompromising ass. I leave it to others who have more familiarity with the debate to let me know whether this characterization is accurate.

All of the above said, I would like to make a strong recommendation for Simple Heuristics That Make Us Smart. Along with Herb Simon’s Sciences of the Artificial (another Cosma recommendation, natch), it forms a great rebuttal of economically-rational man.

(Incidentally, my friend Nina worked under Kahneman a few years ago, and told me that his preferred introduction to the subject is The Winner’s Curse, by his colleague Dick Thaler.)