Stephen Laniel’s Unspecified Bunker

If you’re interested in behavioral economics — that is, the study of how humans do behave in real economic situations, rather than how they ought to behave — this book is the locus classicus. It establishes the theoretical and experimental basis for behavioral economics; Thaler and Sunstein’s Nudge draws out its consequences for public policy.

On the topic of Thaler: I’ve loved every book and article of his that I’ve read — most notably his book The Winner's Curse and Nudge. His and Tversky’s articles are the shining lights in Choices, Values, and Frames, for opposite reasons. Thaler’s articles dig for as many examples as they can of “irrational” behavior in everyday life and public policy; they are deeply pragmatic. Tversky’s articles are theoretical and mathematical, laying a plausible conceptual foundation beneath Thaler’s and others’ work.

Follow the consequences of behavioral economics, and you end up in some philosophically and economically very interesting terrain. An example from the end of the book will capture the idea. Kahneman and Tversky have just proposed a simple model for how people remember the pleasure or pain from an experience; in particular, their data show that people remember the “peak” of the experience (most painful or most pleasurable) and the end of that experience, but very little of the rest. They’ve tested this in a squirm-worthy way, which I have to relate out of sheer fascination: patients undergoing a colonoscopy report their discomfort level every 60 seconds throughout the procedure. Under the “peak/end hypothesis,” as they call it, patients will remember the colonoscopy more favorably if the end of the procedure is less agonizing than the rest of it. So for one of the groups of patients in the experiment, doctors leave the scope in for a few minutes longer — thereby inflicting mild discomfort, but less discomfort than patients had undergone before. The results were as expected.

There’s a very important question in there for informed-consent laws. If you tell patients ahead of time, “We could leave the scope in for a while; it will almost surely improve your memory of the procedure,” almost no one will assent. This is regrettable and dangerous: if they have a less-pleasant memory of the procedure, they’ll be less likely to come back for followup exams.

So which “self” should give his consent? The self who dreads a colonoscopy that’s longer than it has to be? Or the self who looks back with regret on turning down the offer? Here’s where a bit of paternalism might do some good: if I let my doctor make some decisions for me, I’m likely to be better off than if I decide for myself. Thoughts along these lines feed into Nudge, which stakes out a position of “libertarian paternalism”: always leave people with choices, but set the default choice to what’s best for them.

That direction of thought comes at the end of Choices, Values, and Frames. To avoid falling into the peak/end trap, I should tell you about the rest of the book, too. It starts with the editors’ famous (within its field) alternative to standard economic theory. The standard theory, recall, prescribes a set of reasonable axioms for humans to follow. For instance, we’re told that if we prefer an option A to an option B, and B to C, then we should prefer A to C; this is called “transitivity.” The intuition here is that we can be made into “money pumps” if our preferences are intransitive: if you’re currently in possession of C, and someone offers you B, your preferences say that you should pay at least a little to obtain it. Now you have B. Someone offers you A, so you pay a little to obtain it. Now, since your preferences are intransitive, you can be made to pay for C again. And so one around the cycle. Other axioms in the standard toolbox include the “axiom of extensionality,” which says that identical outcomes should lead to identical preferences. Extensionality gets the severest beating in Choices, Values, and Frames. Describing a public-health intervention as “saving 200 lives” within a community of 600 people makes experimental subjects look upon it much more favorably than if it’s described as “losing 400 lives,” even though those outcomes are transparently identical. This “framing effect” can be repeated just about as often as you like, in as many contexts.

There’s a deeper point about preferences that at least a few papers in the book address: humans do not have stable preferences that they can call up with perfect accuracy whenever they’re needed. Preferences are constructed from context. One such contextual constraint on preferences is what one paper calls “extremeness aversion”. This may be connected with Thaler’s observation elsewhere that the best-selling wine on any restaurant’s menu is always the second-cheapest bottle. Likewise, if you want to sell more of a particular size of beverage, make it the “medium” size. The point is that when people walk into a café, they do not have a pre-existing set of choices in mind; their preferences depend in many ways on what they’re presented with. Here’s where one is required to invoke Ulysses and the Sirens: Ulysses at a more-rational moment knew how he would behave when he heard the Sirens’ song; he asked his men to lash him to the mast and tighten the ropes the more he asked to be let go. At a more down-to-earth level, this is why people make sure to eat something before they go to the grocery store. Knowing that your future self might be influenced into making bad choices is not compatible with standard economic theory.

The standard economic model implies that humans have a stable “utility curve” implied by their preferences; that utility curve, from the axiom of extensionality, is based on final states: I should be indifferent between any two outcomes that leave the same amount of money in my pocket. Kahneman and Tversky’s prospect theory starts with a direct attack on this axiom, based on the observation that people are loss averse. That is, they get more pain from a $1 loss than they get pleasure from a $1 gain. The ratio of pain to pleasure, estimated from various experiments, is around 2 to 1. Note that this pain and pleasure come from changes in wealth: people react differently when presented with hypothetical starting sums of $1,000 and $200 losses than they do when presented with hypothetical starting sums of $600 and $200 gains, even though the outcome is the same.

The experimental layout here is a little questionable, which is unavoidable and which behavioral economists are quick to apologize for. Much of it is based on hypothetical gambles. How much do we learn about real economic behavior if we ask people how they’d behave in an artificial situation? A few answers help. First, hypothetical gambles are just about the only way that you can compare people’s behavior under two different starting incomes with two different windfalls. Second, these experiments suggest hypotheses about real economic behavior, which may then be tested against real-world situations.

There are real-world tests within the book. The most convincing ones are, predictably, Thaler’s. Among others, he brings a behavioral-economic analysis to the question of why so few cabs are available on rainy nights. A straightforward economic answer is that the supply of taxis is geared to the typical non-rainy night; on rainy nights, more people need cabs, so supply runs out quickly. Thaler comes up with a different, intriguing answer on the basis of his own interviews with cabbies: taxi drivers put each day’s wages into a separate box, which Thaler calls “mental accounts.” They work until they reach a certain daily wage, then go home. Of course this isn’t income-maximizing; the income-maximizing approach is to work more, not less, on days when more customers need rides. Thaler’s paper shows that cabbies learn this over time: older cabbies don’t use mental accounts that much. (Mental accounts form an important part of Nudge, and serve as the behavioral response to Modigliani’s “life-cycle theory.” Thaler smirks at and pokes the life-cycle theory in Winner's Curse.)

Other papers in here are somewhat less engaging. There’s one about a supposed behavioral preference for variety, which leads to an experiment involving college students and grocery shopping. (Nearly every experiment in this book was conducted on undergrads; narrower still, the bulk of those undergrads seem to come from Tversky’s Stanford or Kahneman’s Princeton. Dear Academic Psychology Departments: please outsource some of your experiments. Love and kisses, Me.)

But then some are fascinating. Mentioned several times in Choices, Values, and Frames is the observation that longshot bets tend to be clustered at the end of the day at racetracks. The authors present a fascinating behavioral explanation, again centered on day-long mental accounts: people hate to end the day with a loss. So when the end of the day comes, and most bettors are losing, they go for broke on a horse who can pull them out of the red. This mis-allocation of bets suggests a way to beat the odds at the track: come in at the end of the day and bet on the favorite to show (that is, come in first, second, or third). An irrational amount of money will have gone to unlikely horses, and you will probably be in the black.

There are enough gems like this within the 774-page mass of Choices, Values, and Frames that the whole book is surely worth the read. If you have any interest in either the pragmatics or the theory of behavioral economics, this is certainly a book you want on your shelf.