Charlie and Vernon’s critique came right from the list of complaints discussed in chapter 6, “The Gauntlet.” First, they thought that subjects might have been confused; they would have preferred an experiment in which the subjects had an opportunity to learn. Second, they invoked a version of the invisible handwave to argue that the misbehaving observed in the Knetsch and Sinden experiment would disappear if the subjects were making choices in a market context, meaning buyers and sellers trading and prices fluctuating. Danny and I returned to Vancouver with a mission: design an experiment that would convince Plott and Smith that the endowment effect was real.
Naturally, since Jack had conducted the original experiment and was part of our fairness team, we joined forces with him on the new design. The discussion with Charlie and Vernon also led us to recognize that the endowment effect, if true, will reduce the volume of trade in a market. Those who start out with some object will tend to keep it, while those who don’t have such an object won’t be that keen to buy one. We wanted to come up with a design that could exploit this prediction.
The basic idea was to build on Jack’s original study and add a market. To make the case airtight, we wanted to show that the results were not an unintended consequence of the particular methods we employed. We decided to use one of Smith’s favorite experimental devices—induced value—to our advantage. As was mentioned in chapter 5, Vernon had used this methodology in many of his pioneering early experiments demonstrating how well markets can work. Recall that when using this method, subjects buy and sell tokens that are worthless outside the laboratory. They are each told their own personal value for a token, a value that is redeemable at the end of the experiment if the subject has a token. Seth is told that if he ends up with a token at the end of the experiment, he can sell it back to the experimenter for (say) $2.25, while Kevin is told that he can get $3.75 for a token. We used this method because we did not expect anyone to have an endowment effect for a token any more than they would have an endowment effect for a particular twenty-dollar bill.
Figure 7 illustrates how this market is supposed to work. Suppose we have twelve subjects and we have assigned induced vales to them at random, varying from 25 cents to $5.75. We then line up these subjects, with the subject given the highest induced value on the left and the one with the lowest value on the right, as shown in panel A. We then hand out six token at random to these subjects, as is illustrated in panel B. We now conduct a market by asking subjects to answer a series of simple questions. Those who own a token would get a form such as this one:
At a price of 6.00
I will sell________
I will not sell________
At a price of 5.50
I will sell________
I will not sell________
The lowest price at which a seller is willing to part with their token is called their reservation price. Someone with a valuation of $4.25 would be willing to sell at a price of $4.50 but not at $4, so her reservation price would be $4.50. Potential buyers would receive a similar form asking about their willingness to buy a token over the same range of prices. What does economic theory predict will happen? If the market works well, the six subjects who value the tokens the most, the ones on the left, will end up owning the tokens. In this example, that means that subjects 7, 8, and 11 will buy tokens from subjects 2, 5, and 6, as illustrated in panel C.
We can figure out the price that will make this market “clear,” meaning equate supply and demand, by working from the two ends of the distribution toward the middle. Subject 11 will have no trouble finding a price at which Subject 2 will give up his token, so they are bound to make a deal. The same applies to Subjects 8 and 5. But to get Subject 7 to buy a token from Subject 6, the price will have to be between their two reservation prices. Since we only allowed prices in increments of 50 cents, the market-clearing price will be $3.
FIGURE 7
Since both the values and the tokens are being handed out at random, the particular outcome will differ each time, but on average the six people with the highest valuations will have been allocated half of the tokens, and as in this example, they will have to buy three tokens to make the market clear. In other words, the predicted volume of trading is half the number of tokens distributed.
Now suppose we repeat the experiment, but this time we do it with some good such as a chocolate bar. Again we could rank the subjects from high to low based on how much they like the chocolate bar, but in this case we are not telling the subjects how much they like the good; they are determining that themselves. Now we distribute the chocolate bars at random, just as in the token experiment, and ask the same series of questions. What should happen? The theory yields exactly the same prediction. On average, half of the chocolate bars will change hands, moving from those who don’t care so much for chocolate (or are on a diet) to the chocoholics who can’t wait to start munching on one of those bars. But if there is an endowment effect, the people who are randomly assigned chocolate bars will value them more than those who don’t, and the volume of trade will be lower as a result. This is the prediction we wanted to test.
The first experiment with this design was run when I returned to Cornell in the fall of 1985. I commandeered an advanced undergraduate class in law and economics to run the experiment. In this case there were forty-four students, so there were twenty-two tokens handed out at random, and every subject was given his or her private value. Then, token owners were told that there would be a market for tokens with a price to be determined by supply and demand. Their task was to answer a series of questions, quoting different prices, e.g.:
At a price of $6.25
I will sell________
I will not sell________
At a price of $5.75
I will sell________
I will not sell________
To understand the task, subjects simply had to realize that if their private valuation was, say, $6.50, they should agree to sell at every price greater than $6.50 and refuse to sell for every price of below that amount. The lowest price at which they would be willing to sell is called the seller’s “reservation price.” Buyers also received private values and a similar form to fill out that yielded their reservation prices, i.e., the highest price at which they would be willing to buy. To be sure that everyone understood what was going on, we did this three times.
We then ran the markets right in front of the class while they watched. To do this, one simply uses the tools of supply and demand taught in any introductory economics class. Specifically, we took all the reservation prices of the sellers and ranked them from lowest to highest, and ranked the buyers’ reservation prices from highest to lowest. If the highest bid by a buyer is greater than the lowest offer by a seller, then we have at least one sale. If the second highest bid by a buyer is greater than the second lowest offer by a seller, then we have two sales, and so forth until the highest bid is less than the lowest ask. All trades happen at the same price, namely the price at which the number of tokens demanded is equal to the number supplied.
Recall that we predict about eleven trades—matching half of the twenty-two buyers with half of the twenty-two sellers—will occur. In the three trials the actual number of trades was twelve, eleven, and ten, so the market was working fine and the subjects demonstrably understood what they were being asked to do.
We were now ready for the experiment that mattered, where we would use real goods instead of the tokens. In preparation for the experiment, I went over to the campus bookstore to see what products I could buy to use in the study. I wanted something that the students might want and was not too expensive, since we had to buy twenty-two of each item. Eventually I settled on two objects: a coffee mug with the Cornell insignia and a nice ballpoint pen that came in a box. The mugs cost $6 each, and the pens were $3.98 each. In the case of the pens, the price tag was left on the box.
We began by putting a coffee mug in front of every other student. The students who got a mug were owners an
d potential sellers; the others were potential buyers. Everyone was told to inspect the mug, either their own or their neighbor’s, to ensure they all had equal information about the products. Then we conducted exactly the same market that we had used for the tokens. To allow for learning, one of Plott and Smith’s requirements, we said that we would do this four times and pick one of the trials at random to “count.” As with the tokens, economic theory predicts that the number of trades will be about eleven, but we were predicting significantly fewer trades because of the endowment effect.
Our prediction was right. On the four successive markets, the number of trades were four, one, two, and two respectively: not even close to eleven. The reason was apparent. Those who got the mugs were reluctant to sell them; the median reservation price for sellers was $5.25 in each of the four rounds. But those who did not have a mug were not eager to buy one; the median reservation price for buyers was $2.75 in one round and $2.25 in the others.
We repeated the experiment with the pens. The students who did not get a mug got a pen, so everyone had a chance to be a buyer and a seller. The students were not wild about these pens, but the results were about the same. The number of trades varied between four and five, and the ratio of selling to buying prices was again in the neighborhood of 2:1.
We ran numerous versions of these experiments to answer the complaints of various critics and journal referees, but the results always came out the same. Buyers were willing to pay about half of what sellers would demand, even with markets and learning. Again we see that losses are roughly twice as painful as gains are pleasurable, a finding that has been replicated numerous times over the years.
The endowment effect experiments show that people have a tendency to stick with what they have, at least in part because of loss aversion. Once I have that mug, I think of it as mine. Giving it up would be a loss. And the endowment effect can kick in very fast. In our experiments, the subjects had “owned” that mug for a few minutes before the trading started. Danny liked to call this the “instant endowment effect.” And while loss aversion is certainly part of the explanation for our findings, there is a related phenomenon: inertia. In physics, an object in a state of rest stays that way, unless something happens. People act the same way: they stick with what they have unless there is some good reason to switch, or perhaps despite there being a good reason to switch. Economists William Samuelson and Richard Zeckhauser have dubbed this behavior “status quo bias.”
Loss aversion and status quo bias will often work together as forces that inhibit change. Think of people who lose their jobs because a plant or a mine closes down, and in order to find work, they would have to both take up another line of work and give up the friends, family, and home to which they have become attached. Helping people get back to work can often be met with inertia. We will return to this concept later in the context of public policy. For now, let me just offer an amusing example of status quo bias.
In the years since our mugs paper was published in 1990, there have been dozens, perhaps hundreds of follow-up studies, some critical of our findings, others exploring what psychologists call the boundary conditions of the phenomenon, meaning the limits on when it will be observed and when it will not. There is one thing that nearly all these studies have in common: coffee mugs. Thousands of university insignia coffee mugs have been purchased and given away by economists and psychologists, all because at the Cornell bookstore one day, a coffee mug caught my eye. Someone that makes mugs with university insignias owes me dinner.
Near the end of my year in Vancouver, Danny made an offhand comment that was, as usual, wise. We were gossiping about some academic we both knew and Danny said: “You know, at some point people reach an age at which they can no longer be considered ‘promising.’ I think it is about the time they turn forty.” I am sure that Danny did not know my exact age, but I was thirty-nine. By the time classes resumed and I returned to Cornell, I would be forty. Damn. I had kind of enjoyed being “promising.”
V.
ENGAGING WITH THE ECONOMICS PROFESSION:
1986–94
By the time I returned to Cornell from my year in Vancouver, I had been working full time on my risky behavioral economics endeavor for eight years. And either despite or because of this endeavor, depending on whom you ask, I had managed to get tenure at Cornell and had several papers in the pipeline to be published in top journals. I was finding the project that had once looked very much like a fool’s errand as much fun as ever, and it kept a roof over my family’s head. The biggest problem was that, aside from our engagement with the experimental economics community, Amos, Danny, and I were mostly talking to one another. That state of affairs was about to change.
17
The Debate Begins
Behavioral economics got its first major public hearing shortly after I returned to Cornell from Vancouver. In October 1985, two University of Chicago Graduate School of Business professors—Robin Hogarth, a psychologist, and Mel Reder, an economist—organized a conference at the University of Chicago, home of many ardent defenders of the traditional way of doing economics. Rationalists and behavioralists were to come together and try to sort out whether there was really any reason to take psychology and behavioral economics seriously. If anyone had been laying odds on who would win this debate, the home team would have been considered the strong favorite.
The behavioral team was led by Herb Simon, Amos, and Danny, and was buttressed by Kenneth Arrow, an economic theorist who, like Paul Samuelson, deserved to win several Nobel Prizes in economics, though he had to settle for just one. The younger behavioral crowd, which included Bob Shiller, Richard Zeckhauser, and me, were given speaking roles as discussants.
The rationalists’ team was formidable, with Chicago locals serving as team captains: Robert Lucas and Merton Miller. Eugene Fama and my thesis advisor, Sherwin Rosen, were given the roles of panel moderators, but were clearly part of the Chicago-based rationalists’ side. The two-day meeting was held in a large auditorium, and every seat was taken. Thinking back on it, this conference was a highly unusual event. I don’t think I have ever been to another one quite like it.
Amos presented a new paper that he and Danny had written for the occasion. It offered some violations of economic principles that economists found especially disconcerting. One was their now famous Asian disease problem, which goes as follows:
Two groups of subjects are told that 600 people are sick from some Asian disease, and a choice has to be made between two policies. The choices offered to the first group are:
Policy A will save 200 people for sure.
Policy B offers a one-third chance to save everyone but a two-thirds chance that all 600 patients will die.
When presented with this choice, most people take the safe option A.
In the alternative version, the subjects are again given two choices:
If they go with option C, 400 will die for sure.
If they choose option D, there is a one-third chance of killing
no one and a two-thirds chance of killing everyone.
In this case, a majority preferred the risky option D.
Offhand, there does not appear to be anything remarkable about these choices, but a little arithmetic reveals that policy A is the same as C, and policy B is the same as D, so it is not logical for respondents to prefer A over B but D over C. And yet they did, and the same results were obtained with a similar problem posed to a group of physicians. Results like this clearly made the rational camp uncomfortable. Econs would certainly not misbehave so blatantly.
Danny then presented some of our work on fairness, including our Ultimatum and Dictator Game experiments. These findings were not any more popular. The economists thought that fairness was a silly concept mostly used by children who don’t get their way, and the skeptics just brushed aside our survey data. The Ultimatum Game experiments were a bit more troubling, since actual money was at stake, but of course it wasn’t all that much money, and a
ll the usual excuses could be raised.
The talk that gave me the most to think about, and the one I have gone back to read again most often, was by Kenneth Arrow. Arrow’s mind goes at light speed, and his talks tend to be highly layered fugues, with digressions inserted into digressions, sometimes accompanied by verbal footnotes to obscure scholars from previous centuries, followed by a sudden jump up two or three levels in the outline that he has in his head. While you work to digest a profound nugget disguised as a throwaway line, he has leapt back to the main argument and you are left scrambling to catch up. On this occasion, however, his talk can be summarized easily: rationality (meaning optimization) is neither necessary nor sufficient to do good economic theory.
Arrow began by dumping on the idea that rationality is necessary. “Let me dismiss a point of view that is perhaps not always articulated but seems implicit in many writings. It seems to be asserted that a theory of the economy must be based on rationality, as a matter of principle. Otherwise there can be no theory.” Arrow noted that there could be many rigorous, formal theories based on behavior that economists would not be willing to call rational. As an example, he noted that the standard theory of the consumer states that when prices change, the consumer will solve the new optimization problem and choose a new “best” set of goods and services that still satisfies the budget constraint. Yet, he noted, one could easily build a theory based on habits. When prices change, the consumer chooses the affordable bundle that is closest to what she was consuming before. Arrow could have gone even further. For example, we could have rigorous theories as bizarre as “choose the bundle with brand names in order to maximize the occurrences of the letter K.” In other words, formal models need not be rational; they don’t even have to be sensible. So we should not defend the rationality assumption on the basis that there are no alternatives.
Misbehaving: The Making of Behavioral Economics Page 17
