This story illustrates two important tools that people use to confront self-control problems. For the crew, the strategy was to remove the cues that would tempt them to do something stupid. Out of sight, out of mind. For himself, Odysseus chose a commitment strategy: he limited his own choices to prevent self-destruction. It was his version of removing the bowl of cashews. Strotz confessed to employing a commitment strategy himself, to accommodate his academic calendar year pay schedule: “I select the option of having my annual salary dispersed to me on a twelve- rather than nine-month basis, although I could use the interest!”
By the time I was thinking about self-control problems in 1978, Strotz’s paper was already more than twenty years old, and there was no one else in economics who seemed interested (though Tom Schelling would soon chime in). I turned to psychology for inspiration. Surely, I thought, there would be a vast literature in psychology on delay of gratification. Dead wrong. Although many psychologists are now interested in self-control problems, in the late 1970s that was not the case. But I did unearth two treasures.
The first was the work of Walter Mischel, which is now quite well known. Mischel, then at Stanford, was running experiments at a day care center on the school’s campus. A kid (age four or five) was asked into a room by the experimenter and given a choice between a small reward now and a larger reward a bit later. The rewards were treats such as marshmallows or Oreo cookies. The child was told that he could have one Oreo right now, or any time he wanted it, but if he could wait until the experimenter came back, he could have three Oreos. At any time the kid could ring a bell and the experimenter would return and give him the small reward.
Most of the children found this task excruciatingly difficult, but the circumstances in which the child was waiting mattered. In some versions of the experiment the treats were on a plate right in front of the kid. The sight of those Oreos had the same effect on most children that the Sirens’ tunes had on Odysseus. Waiting time averaged barely over a minute. But if the rewards were out of sight (and thus more out of mind), the average kid could hold out for eleven minutes. Children also could wait longer if they were told to think about something “fun” instead of the reward itself.
The earliest of these experiments were run in the late 1960s and early 1970s. About ten years later, as an afterthought, Mischel and his colleagues thought it would be interesting to see what had happened to the children who had been the subjects in these experiments, so they tracked down as many of the 500 or so original participants as they could find, and eventually found about a third of them who agreed to be interviewed once a decade. Somewhat surprisingly, the amount of time a kid waited in one of those experiments turned out to be a valid predictor of many important life outcomes, from SAT scores to career success to drug use. This was a particularly surprising result, because Mischel himself had done considerable research showing that so-called personality traits were not very useful in predicting behavior, even in the present, much less the future.
Mischel has priceless videos from some of the early experiments that demonstrate the difficulty kids had in exerting self-control. There is one kid I am particularly curious about. He was in the toughest setup, in which the bigger prize, three delicious Oreo cookies, was sitting right in front of him. After a brief wait, he could not stand it anymore. But rather than ring the bell, he carefully opened each cookie, licked out the yummy white filling, and then put the cookie back together, arranging the three cookies as best he could to avoid detection. In my imagination, this kid grows up to be Bernie Madoff.
The other behavioral scientist whose work captured my attention was a practicing psychiatrist named George Ainslie who was doing research in his spare time, while holding a job treating patients in a veterans’ hospital. In a paper published in 1975, which I had studied carefully during my year at Stanford, Ainslie summarized everything academics knew about self-control at the time.
I learned from Ainslie that there existed a large literature studying delay of gratification in nonhuman animals such as rats and pigeons. In a paradigm similar to Mischel’s, experimenters would give an animal a choice between a small, immediate reward and a delayed, larger reward. The animals had to press (or peck) a lever to get a reward, and, after extensive training, they would learn the length of the delays and amounts of food they could expect from pressing one lever or the other. By varying the delays and sizes of the rewards, the experimenter could estimate the animals’ time preferences, and most studies found that animals display the same discounting pattern that leads to preference reversals in humans. Animals discount hyperbolically, and have self-control problems too!†
Ainslie’s paper also provides a long discussion of various strategies for dealing with self-control problems. One course of action is commitment: removing the cashews or tying yourself to the mast. Another is to raise the cost of submitting to temptation. For example, if you want to quit smoking, you could write a large check to someone you see often with permission to cash the check if you are seen smoking. Or you can make that bet with yourself, what Ainslie calls a “private side bet.” You could say to yourself, “I won’t watch the game on television tonight until I finish [some task you are tempted to postpone].”
Armed with the insights of Strotz, Mischel, and Ainslie, I set out to create a conceptual framework to discuss these problems that economists would still recognize as being economics. The crucial theoretical question I wanted to answer was this: if I know I am going to change my mind about my preferences (I will not limit myself to a few more cashew nuts, as I intend, rather I will eat the entire bowl), when and why would I take some action to restrict my future choices?
We all have occasions on which we change our minds, but usually we do not go to extraordinary steps to prevent ourselves from deviating from the original plan. The only circumstances in which you would want to commit yourself to your planned course of action is when you have good reason to believe that if you change your preferences later, this change of preferences will be a mistake.
Removing the cashews is smart because eating the entire bowl will ruin your appetite, and you would rather not have your dinner consist entirely of cashew nuts. Likewise, a smart kid who participated in one of Mischel’s experiments would be wise to say to the experimenter, “Next time you have Oreos to give away, please do not offer me the ‘one cookie now’ option, or even mention the word Oreo. Just wait fifteen minutes and bring me my three cookies.”
At some point in pondering these questions, I came across a quote from social scientist Donald McIntosh that profoundly influenced my thinking: “The idea of self-control is paradoxical unless it is assumed that the psyche contains more than one energy system, and that these energy systems have some degree of independence from each other.” The passage is from an obscure book, The Foundations of Human Society. I do not know how I came by the quote, but it seemed to me to be obviously true. Self-control is, centrally, about conflict. And, like tango, it takes (at least) two to have a conflict. Maybe I needed a model with two selves.
As intuitively appealing as this idea was to me, any sort of two-self model had the disadvantage of being considered radical in economics but passé in psychology: not a great combination. Few economists, including me when I was getting started on this work, were aware of Adam Smith’s discussion of the battle between our passions and our impartial spectator. To most the idea just seemed wacky. Academic psychologists at that time were no longer much enamored of Freud with his id, ego, and superego, and the two-system view that is now very much in vogue had yet to emerge.‡ With much trepidation, I quietly trotted out the idea among friends. A sketch of the concept appeared in my “Toward a Positive Theory of Consumer Choice” paper, but I knew I needed something more formal, which in economics means a credible amount of math. I recruited Hersh Shefrin, a mathematical economist who was at the University of Rochester at the same time I was there, to join the effort.
Hersh was the first of many coauthors I have worked w
ith over the years. When we began talking about these questions, his chief qualifications were that he was good at math and he did not think my ideas were completely crazy. The latter was more important, since economists who are better at math than me are easy to find. In many ways Shefrin and I were polar opposites. Hersh was serious, meticulous, studious, and religious, including being a student of the Talmud, the encyclopedic compendium of ancient Jewish scholarly writing. I had none of those qualities, but we still managed to get along well. Most importantly, Hersh laughed at my jokes. We worked together in the way I had seen Amos and Danny work, through endless talking. And when it came to writing our first paper, we would talk through each sentence, just as I had watched them do. Although we began our conversations while we were colleagues at Rochester, I soon moved to Cornell and Hersh departed for sunny California at Santa Clara University, not far from Stanford. We wrote just two papers together, but Hersh got hooked on doing behavioral economics and soon formed a highly successful collaboration with Meir Statman, a colleague at Santa Clara, doing research on behavioral finance.
Our model is really based on a metaphor. We propose that at any point in time an individual consists of two selves. There is a forward-looking “planner” who has good intentions and cares about the future, and a devil-may-care “doer” who lives for the present.§ The key question for any model of this behavior was deciding how to characterize interactions between the two. One possibility would be to make the planner and doer players who interact as competitors in a game, using the branch of mathematics and economics called game theory as the core model. We rejected this idea because we did not think that the doer engages in strategic behavior; he is more of a passive creature who simply lives for the moment. He reacts to what is in front of him and consumes until sated. Instead, we chose a formulation based on the theory of organizations, namely a principal–agent model. In this choice we were undoubtedly influenced by the fact that agency theory, as it was called, was a focus of discussion at the University of Rochester Graduate School of Business while I was teaching there. Michael Jensen and the then dean of the school, William Meckling, had written a famous paper on the topic in 1976. I was not sure that they would approve of this application of their ideas, but that was part of the fun.
In a principal–agent model the principal is the boss, often the owner of a firm, and the agent is someone to whom authority is delegated. In the context of an organization, tensions arise because the agent knows some things that the principal does not, and it is too costly for the principal to monitor every action that the agent takes. The agent in these models tries to make as much money as possible while minimizing effort. In response, the firm adopts a set of rules and procedures (incentive schemes and accounting systems, for example) that are designed to minimize the costs of the conflicts of interest between the principal and the agents employed at the firm. For example, a salesperson might get paid mostly on commission, have to turn in receipts to document travel expenses, and be forbidden to fly first class.
In our intrapersonal framework, the agents are a series of short-lived doers; specifically, we assume there is a new doer each time period, say each day. The doer wants to enjoy himself and is completely selfish in that that he does not care at all about any future doers. The planner, in contrast, is completely altruistic. All she¶ cares about is the utility of the series of doers. (Think of her as a benevolent dictator.) She would like them to be collectively as happy as possible, but she has limited control over the actions of the doers, especially if a doer is aroused in any way, such as by food, sex, alcohol, or an urgent desire to go outside and goof off on a nice day.
The planner has two sets of tools she can use to influence the actions of the doers. She can either try to influence the decisions that the doers make through rewards or penalties (financial or otherwise) that still allow them discretion, or she can impose rules, such as commitment strategies, which limit the doers’ options.
Consider a simple, albeit contrived, example just to illustrate the ideas. Suppose Harry is out camping alone in a remote cabin with no means of communicating with the outside world. He was dropped off by a small plane and will be picked up again in ten days. Originally, he had plenty of food to eat (water is plentiful), but a hungry bear wandered by and walked off with every food item he had except for ten energy bars that either escaped the bear’s attention or did not suit his epicurean tastes. Since there is no way to communicate with the plane, and Harry is not good at scavenging for food, he will have to make the ten energy bars last until the plane comes back to pick him up. Of course, Harry is equipped with a planner and a doer. How will his planner deal with this problem?
Let’s assume that the planner values each doer’s consumption equally (so does not discount the consumption of distant doers relative to current ones). The doers have diminishing marginal utility for food, meaning that the first energy bar is more enjoyable than the second, and so forth, but they eat until the last bite yields no additional pleasure, and then stop. In this setup, the planner would consider the best outcome to be to eat one energy bar per day, thus giving each of the ten doers the same amount of utility.# In other words, the planner would like to impose the same kind of consumption smoothing the Econs are supposed to do if they follow the life-cycle hypothesis. To some extent, the planner is trying to get the doers to act more like Econs. If it were technologically feasible, the planner would adopt a commitment strategy that left no discretion to the doers, thus eliminating any risk of misbehaving. A cabin that came equipped with ten programmable safes, each of which can be set to open at a specific time, would be ideal.** This is the best possible outcome from the planner’s perspective.
But the cabin is unlikely to have those safes, so what can the planner do? All ten energy bars are sitting in the cupboard available to be eaten. What happens then? If the planner does not intervene, the first doer, who does not care at all about the welfare of future doers, will eat until he is full; that is, up until the point that eating one more bite of an energy bar will make him less happy. Let’s say that point comes after eating three energy bars. The second day, the doer also eats three more energy bars, as does the doer of day three. Then when the fourth day comes along, the current doer eats one energy bar for breakfast—the last of the ten left—and soon starts getting hungry. The rest of the week is no fun.
Somehow, the planner has to keep the early doers from bingeing on the energy bars in the first few days. If there are no commitment strategies available, the only other tool the planner has in our model is guilt. Through some process of indoctrination, either by the planner herself or by parents and society, doers can be made to feel bad about leaving future doers with nothing to eat. But imposing guilt is costly. In the energy bar example, the planner is unable to make the doer start feeling bad only after consuming the first energy bar. Instead, she has to make each bite of the energy bar less pleasurable.
This is illustrated in figure 6. The highest line represents the doer’s utility for eating energy bars without guilt, and the doer consumes up until the point where the utility is maximized, at three energy bars. The next highest line illustrates the case where enough guilt has been applied to get the doer to stop eating after two energy bars, and the lowest line shows the case where the doer stops after eating one. The thing to notice about this figure is that when guilt is employed, life is less pleasurable. The only way to make the doer eat fewer energy bars is to make eating them less enjoyable. Another way to think about this is that employing willpower requires effort.
FIGURE 6
This analysis suggests that if one can implement perfect rules, life will be better. The strategy of using programmable safes, each containing one energy bar, achieves much more satisfaction than the guilt-induced diet. Strotz accomplished this goal by asking his employer to pay him in twelve monthly increments, from September through August, rather than in nine, from September to May. The latter plan would yield more interest since the money comes in mor
e quickly, but he has to save up enough over the course of the academic year to ensure he has money to live on during the summer, not to mention to go on a family vacation.
Why not always use rules? One reason is that externally enforced rules may not be easily available. Even if you arrange to get a healthy dinner delivered to your home each night, ready to eat, there will be nothing stopping you from also ordering a pizza. Also, even if such rules are available, they are inflexible by design. If Professor Strotz opts for the September-to-May pay schedule, the money comes in earlier, so he might be able to take advantage of an opportunity to buy something on sale during the winter—say, a new lawnmower—that will be more expensive in the summer. But if his salary is spread out over twelve months he may not have enough slack in the budget to buy a lawnmower in winter. Of course, the flip side is that if he takes the money early he has to have the discipline to make it last through the summer.
This same principle applies in organizations. If the principal knows exactly what the agent should do in every situation, then she can create a rulebook that is never to be violated. But we have all had the frustration of dealing with a low-level agent who is working under such rules and does not have the discretion to do something that is obviously sensible but had not been anticipated, and is therefore “not allowed.”
Misbehaving: The Making of Behavioral Economics Page 12
