15
Second-best rationality
In the last two chapters I have considered the ideal of rational behavior and the frequent lapses from rationality. These lapses, however widespread and frequent, are not inevitable. If we understand our propensity to make mistakes, we can and do take precautions to make ourselves less likely to make them again, or at least limit the damage if we do. As I have said repeatedly, we want to be rational. We may think of these precautionary strategies as a form of imperfect or second-best rationality. They should be distinguished from simple learning, which occurs when the propensity simply fades away as a result of improved insight. It has been reported, for instance, that when people realize that voting is, in one sense, pointless, they are less likely to vote.1 Cognitive fallacies that are akin to optical illusions can also be overcome by learning. Just as we learn to ignore the appearance of a stick that looks broken in water, some gamblers presumably learn the hard way that the dice have no memory. I am concerned here, however, with propensities that persist over time.
To cope with our tendencies to behave irrationally, we may use either intrapsychic strategies or extrapsychic devices (precommitment). I shall first illustrate how these techniques are used to counteract hyperbolic discounting and the inconsistent behavior it generates, and then discuss their use to control emotional and addictive behavior. These various strategies are not necessarily rational, but many of them are.
Future selves as allies
An agent who is subject to hyperbolic discounting and knows it is sophisticated. Unlike the naive agent who finds himself changing his mind over and over again without understanding the underlying mechanism, the sophisticated agent both is aware of her propensity and deplores it. Anticipating future situations in which she will face the choice between an early small reward and a delayed larger reward, she would like to make herself choose the latter despite her propensity to choose the former. In some cases, she may treat her “future selves” as allies in a common effort to overcome temptations. In other cases, she may treat them as adversaries and try to limit the damage they can do to her “present self.” This language, to be sure, is metaphorical, but it will be demetaphorized.
Consider first the case in which the choice between an early small and a delayed large reward can be expected to arise over and over again. The agent can then make herself go for the delayed reward by bunching (or bundling) the choices.
Let me illustrate by an example from a time I was living up in the hills close to the university where I was teaching. Every day I took my bike to get to campus and back. The return trip involved some steep uphill climbing, so that every day I faced the temptation to get off the bike and walk rather than forcing myself to pedal. When I set out from campus I was firmly committed to staying on the bike all the way, but in the middle of the climb a seductive thought would often occur to me, “Why not walk today, and resume biking tomorrow?” Then, fortunately, a further thought occurred, “What is so special about tomorrow? If I yield to temptation today, does not that predict that I will do so again tomorrow, and the day after, and so on?” The last thought enabled me to stay on the bike.
This intrapsychic device involves a reframing of the situation. Rather than thinking about future trips home as involving a series of choices, I began to see them as a choice between two series: always biking up the hill and always walking the bike. By telling myself that my behavior on one occasion was the best predictor of my behavior on the next occasion, I set up an internal domino effect that raised the stakes and made me go for the delayed reward of improved health rather than for the early reward of relief from discomfort. Referring to Figure 6.1, it can be shown, in fact, that if we place many pairs of rewards identical to A and B on the horizontal line and then form two curves, one for the sum of the present-value curves of all the small rewards and one for the sum of the present-value curves of all the large rewards, the latter curve will lie above the former at the time the first choice has to be made, provided that the number of successive choices to be made is large enough.2 In other words, bundling the choices can make the option of always going for the larger reward preferable to always going for the smaller reward. To be sure, choosing the smaller reward today and the larger reward on all future occasions is even better, but by assumption this option is not in the opportunity set of the agent.3
Can this assumption be justified? Is my behavior today a good predictor of my behavior tomorrow? In cases that involve a genuine causal effect, this may be true. Biking today will keep my muscles strong so that I can also bike tomorrow.4 In my case, however, I relied on magical thinking rather than on causal efficacy. Just as many people vote or give to charity under the influence of the thought “If not me, who?” what kept me on the bike was the thought “If not now, when?” Or more elaborately: “There is nothing special about today. If I get off the bike, the causes that made me do it will also operate tomorrow and induce the same behavior. If I do not make an effort now, I never will.” In the absence of a genuine causal effect, however, the conclusion does not follow. If I can stay on the bike today but decide to get off, I can also stay on it tomorrow. Although false, the reasoning is compelling and, I believe, extremely widespread. It shows that we can enlist one form of irrationality (magical thinking) to combat another (hyperbolic discounting).5
To work well, such strategies may have to be framed as binary choices: always doing it or never doing it. For many people, abstention is easier than moderation. Boswell noted that “Johnson, though he could be rigidly abstemious, was not a temperate man either in eating or drinking. He could refrain, but he could not use moderately.”6 The same problem arises if instead of limiting consumption on each occasion one tries to limit the number of occasions on which one may indulge. The stratagem of laying down in advance what will count as a legitimate occasion is easily eroded. When people resolve not to drink alcohol before dinner, they may find themselves scheduling dinner earlier. The rule of drinking wine only at restaurants, never at home, may cause one to dine out more frequently.7 Kant's rule of smoking only one pipe after breakfast (Chapter 4) was not unambiguous enough to give him full protection, since as time passed he bought himself bigger and bigger pipes. Similarly, when feasible, the rule “Never do it” may be the only one that can be stably upheld. Since this policy is not feasible with regard to eating, obesity may be more recalcitrant than are addictions to private rules.
The binary choice framing can, however, induce absurdly rigid behavior. Suppose I have told myself never to suffer a single exception to the rule of brushing my teeth every night. On a given occasion I find myself without a toothbrush and decide to walk five miles in a blizzard to buy one. To sustain the decision, I tell myself that if I break the rule on that occasion, I will be on a slippery slope leading to rule violations for ever more trivial reasons, and soon there will be no rule at all and my teeth will fall out. Some people construct very elaborate systems of this kind, in which failure to follow one rule predicts failure with regard to other rules as well, thus raising the stakes even more.8 Because private rules may have these stultifying effects, they sometimes provide a remedy worse than the disease. In Freudian language (Chapter 4), the rigid impulse control exercised by the superego could do more damage than the impulses from the id.
Future selves as adversaries
Consider now the case in which the agent confronts the choice between rewards (or punishments) at one of several future dates. (Unlike the previous case, the choice is only supposed to arise once.) The agent may then adopt the intrapsychic device of responding strategically to the known propensity of “future selves” to discount the future hyperbolically. Suppose I am a “hyperbolic procrastinator” who always put things off until tomorrow, and then, when tomorrow arrives, puts them off again until the day after tomorrow. Once I understand that I am subject to this propensity, my optimal behavior changes. Suppose that I can carry out a given unpleasant task in any of three periods, and that the cost of doing so goes up with time. If I am naive, I may tell myself that I will perform the task tomorrow. If I am sophisticated, I know that tomorrow I will delay until the last period. The understanding that the cost will in fact be very high unless I perform the task right away may induce me to do exactly that.9
In this case, being sophisticated helps. In other cases, being naive may be better. Suppose you can have a reward in any one of three successive periods and that the rewards increase with time. An example might be a person who has been offered a bottle of wine that improves with time up to the third year, and then deteriorates. A naive person may form the intention to wait until the third period, and then change her mind and drink the wine in the second period. A sophisticated person will know that he is never going to wait until the third period, so that he effectively only faces the choice between the first-period reward or the second-period reward. In that choice, the early reward may win out.10 Some alcoholics report being subject to a similar kind of reasoning: “I know I am going to yield to temptation, so I might as well do it right away.” Also, naive smokers who quit for the first time may hold out longer than sophisticated smokers who have tried several times and know the odds against succeeding. Although backsliding in addiction need not be due to hyperbolic discounting, the general point is the same: if you can predict that you will deviate from your best plan, you may end up deviating even more from it or earlier than if you are unaware that you will fail.
New Year's resolutions
The theory of focal points (Chapter 18) predicts that people will try to quit a bad habit or an addiction on some salient day, such as January 1, to resist the temptation to say to oneself, “Why not wait until tomorrow?” In one study, 213 participants made an average of 1.8 New Year's resolutions. Smoking cessation (30 percent) and weight loss (38 percent) together accounted for two-thirds of the resolutions. Relationship improvement (5 percent), reduction in alcohol consumption (2 percent), and an increase in monetary savings (2 percent) were also popular resolutions. The “other” category, representing 23 percent of the primary pledges, contained a multivarious range of idiosyncratic responses, such as temper control, setting aside time for oneself, making decisions oneself, and learning to say no. Although the study reports success rates and suggests some explanatory variables, it does not say anything about the efficacy of the New Year's resolutions compared to the decision to stop on any other day.
There is some indirect evidence, however, for their efficacy: cigarette companies seem to believe in it. If most people try to quit on January 1 and, as the evidence suggests, withdrawal symptoms peak after about one month, quitters should be particularly vulnerable to cue-triggering in January and February. One might expect, therefore, that there would be a peak in cigarette advertising in those months, an expectation confirmed by an analysis of the back covers of 3,024 magazines. After considering and refuting three alternative explanations, the authors conclude that the timing of the advertisements was deliberately chosen to counter New Year's resolutions. The conclusion is certainly consistent with what we know about the culture of hypocrisy and manipulation in American cigarette companies, second only to that of the National Rifle Association.
The findings suggest a three-level model. At the first level, there is the desire of an individual to smoke or engage in other excessive (legal) behaviors. At the second level, there is the desire of an individual who engages in these behaviors, to quit them. At the third level, there is the desire of the companies who profit from the behaviors to prevent the consumers from quitting. Objectively, there is an alliance between the first and the third level. The desire of the user in a weak moment and the desire of the companies are identical. A person who wants to quit is squeezed in the middle.
Extrapsychic devices
In practice, intrapsychic devices may be less important than the precommitment devices to which I now turn. These involve affecting the external world, in ways that cannot be instantly and costlessly undone, for the purpose of making it less likely that one will choose the earlier, smaller reward in the future. Six strategies stand out: eliminating the choice of the early reward from the feasible set, imposing a penalty on the choice of the early reward, adding a premium for the choice of the delayed reward, imposing a delay between the choice and the actual delivery of the reward, avoiding cues that might trigger preference reversal, and avoiding information that might cause you to choose the immediate reward. Saving behavior can illustrate the first four options. If I begin saving for Christmas but find myself taking money out of my savings account instead of keeping it there, I may join a Christmas savings club that will not allow early withdrawals (Chapter 14). Alternatively, I may put my savings into a high-interest account that carries a penalty for early withdrawal, thus combining premium and penalty. If I want to save for my old age, I may set up a delay between the time I might make a decision to dissave and the moment the funds become available, by investing in illiquid assets rather than in stocks or bonds.
The fifth option is illustrated by the person whose craving for dessert is triggered by visual cues. The trick is to go to a restaurant where they do not wheel around the dessert trolley, so that one has to order from the menu. We may contrast this with a person who has a dessert problem because of hyperbolic discounting. For him, the best option is to go a restaurant where he has to order dessert at the beginning of the meal. Both strategies involve protecting oneself against the effects of the proximity of temptation, be it spatial or temporal.
To understand the sixth option, consider a person who is considering having unprotected sex. On the basis of her approximate knowledge about the risks of AIDS transmission, she decides to abstain. She knows that she could easily obtain more accurate information, but abstains from doing that too, on the basis of the following reasoning. If it turns out that the risk is smaller than she thought, she might then decide to have unprotected sex, even though doing so would be suboptimal both from her present point of view and from her future point of view with less accurate information. (Here, I assume that she discounts the future hyperbolically.) It is not clear how often people actually go through this chain of reasoning in a conscious manner. Strategic ignorance, such as avoiding going on the scales, is probably more common when people want to persist in harmful behavior.
People who sign up for weekly physical exercises often drop out after a week or two. To prevent this, they may (in theory at least) sign a contract with the fitness center to pay twice the normal fee up front and receive a fraction back each time they show up. People who sign up for weight loss programs may have to pay a deposit that they get back only if they lose a stipulated amount of weight, sometimes with the rider that if they fail, the deposit will be donated to the person's most disliked political cause. When I set out to give the lectures that resulted in the first edition of the present book, I precommitted myself by telling my students that I would give them a draft chapter at the end of each week. If I had failed to live up to that promise, I would have suffered the cost of their mild ridicule. If I am afraid I might cancel my appointment with the dentist when the time approaches, I can authorize him to bill me twice the regular amount if I fail to show up. In the case of wine that will improve with time, you may ask the seller to store it for you to protect yourself against premature gratification. If you are afraid that you might read the last novel by your favorite crime writer too quickly, skimming paragraphs to get to the dénouement, you might buy a book-on-tape version (and a player with no fast-forward function) that leaves you no choice but to listen to every word.
The examples in the last two paragraphs involve precommitment against two kinds of temptation. On the one hand there is procrastination, including failures to save, to seek painful treatments, to do physical exercise, or to write up a manuscript. On the other hand, there is premature gratification, such as drinking wine too early or skipping pages in a book. These temptations stem directly from hyperbolic discounting. Nothing but the sheer passage of time is involved. In a further category of cases, excessive behavior, hyperbolic discounting may interact with other, visceral motivations. These include overeating, compulsive gambling, and addictive behavior. It may be hard to know, in such cases, whether preference reversal is due to the discounting structure or to other factors. A decision to fast that is made on a full stomach may dissolve as the person again begins to feel hungry. A decision to stop smoking may be eroded by the sight of another person lighting up a cigarette. This is the phenomenon of cue dependence – cravings that are triggered by visual cues associated with the consumption of the addictive substance. A decision to stop gambling that is induced by the guilt feelings of the gambler toward her family may unravel once the emotion fades in strength (Chapter 8). It may also be hard to tell whether we are dealing with procrastination or with visceral factors. A decision to take medication regularly may be undermined by the decline of the strong emotions that caused the patient to see a doctor in the first place.
Once the agent understands that he is subject to the latter mechanisms, he may precommit himself to forestall their operation. To prevent his resolve to diet from being undermined by hunger, he may take a pill that attenuates the craving for food. More drastically, he may have his jaws wired so that he can only take in liquid sustenance. If he knows that his desire for dessert is cue dependent, he will not go to restaurants where they present a dessert trolley. Former heroin addicts will stay away from the places where they used to consume the drug. Ex-gamblers learn not to go to a casino “just to watch others play.” If the agent can predict that her anger will fade so that she will not want to punish the offender, she might carry out the punishment immediately. As noted earlier, this behavior was observed in Belgium after 1945.
In fighting addiction, the strategy of imposing costs on oneself is very common. When General de Gaulle wanted to quit smoking, he told all his friends about it, to increase the costs of backsliding. In his case, the reputation loss would have been very high. In a cocaine addiction center in Denver, doctors are offered the opportunity to write a self-incriminating letter to the State Board of Medical Examiners confessing to drug use and asking that their license be revoked. The letter is automatically mailed if the patient tests positive for cocaine use. Some former alcoholics try to stay dry by taking disulfiram, a drug that has the effect of making the user violently ill if he takes a drink.11
Self-imposed delays can also be effective in resisting cravings. To prevent myself from impulse drinking, I may store my liquor in a safe with a timing device. Alternatively, I may adopt a policy of having no liquor at home so that I have to go to a store to get it. The disulfiram technique in fact combines the imposition of costs with delays, since once you have taken the pill you have to go two days without taking it before you can drink without getting sick. The cocaine addiction center, too, combined costs with delays. It allowed people to break out of the compact by submitting a notarized declaration of withdrawal from the arrangement. There was a two-week delay. Anyone who submitted a request for withdrawal could retrieve the incriminating letter after two weeks. But if during the two weeks’ interim, the withdrawal was rescinded, then it would require another two weeks’ notice. Although many of the patients invoked the withdrawal procedure, none went two weeks without revoking the revocation.
The concern with precommitment against time inconsistency and excessive behavior is relatively recent. The classical writers on the topic focused on precommitment against passion, taken in a wide sense that also includes intoxication and psychotic states.12 In the Odyssey, Homer offered what has become the standard example of precommitment: Ulysses binding himself to the mast so that he would be unable to respond to the song of the Sirens. In On Anger, Seneca wrote, “While we are sane, while we are ourselves, let us ask help against an evil that is powerful and oft indulged by us. Those who cannot carry their wine discreetly and fear that they will be rash and insolent in their cups, instruct their friends to remove them from the feast; those who have learned that they are unreasonable when they are sick, give orders that in times of illness they are not to be obeyed.” In Mme de Lafayette's novel La Princesse de Clèves, the princess flees the court to avoid the temptation of responding to the overtures of the Duc de Nemours; even later, when her husband is dead and she is free to remarry, she stays away. “Knowing how circumstances affect the wisest resolutions, she was unwilling to run the risk of seeing her own altered, or of returning to the place where lived the man she had loved.” In Stendhal's novel Lucien Leuwen, Mme de Chasteller takes care to see Lucien only in the company of a chaperone, to make it prohibitively costly to give in to her love for him.
These strategies are quite common. When people burn their bridges, it may be for strategic reasons (Chapter 18), but probably more often to prevent themselves or others from giving in to fear. I may stay away from the office party because I know from past experience that I am likely to have a drink or several, and that because of its disinhibitory effect alcohol will induce aggressive or amorous behavior that I will later regret. Alternatively, I may decide to take my spouse along, to raise the cost of such behavior. Merely resolving not to drink (or not to get emotional if I do) is less likely to be effective, given “the power of the situation” (Chapter 12). Similarly, controlling anger by the intrapsychic device of counting to ten before talking back or lashing out presupposes a detachment that tends to be lacking in the heat of the moment. A general advice of self-help is in fact to “break the chain early” rather than to rely on self-control in the face of temptation or provocation. As Mark Twain said, “It is easier to stay out than get out.” For an extreme case, I refer to a New York Times headline (April 5, 1996): “Texas Agrees to Surgery for a Molester: Soon to Leave Prison, Man Wants Castration to Curb His Sex Urge.”
Delay strategies might seem to hold out the best promise for dealing with emotion-based irrationality. Since emotions tend to have a short half-life, any obstacle to the immediate execution of an action tendency could be an effective remedy. As I note later, public authorities do indeed count on this feature of emotion when they require people to wait before making certain important decisions. It is rare, however, to observe people imposing delays on themselves for the purpose of counteracting passion. The requisite technologies may simply be lacking. One example, however, is the “covenant marriage” offered by three American states (Arkansas, Arizona, and Louisiana), an optional form of marriage that is harder to enter and harder to leave than the regular marriage. Typically, a couple who have entered a covenant marriage can be granted a divorce only after two years of separation, as compared to a normal waiting time of six months. The small minority (less than 1 percent of marrying couples) who use this option presumably do so to signal their commitment to each other and to protect themselves against short-lived passions and temptations.
Precommitment often involves the help of other individuals, organizations, or public authorities. These need, however, to be independent from the agent issuing the precommitment instructions, since otherwise he might revoke them. To fight his opium addiction, Samuel Coleridge hired a man to oppose by force his entrance into any druggist's shop. When the man tried to restrain him, Coleridge said, “Oh, nonsense. An emergency, a shocking emergency has arisen – quite unlooked for. No matter what I told you in times long past. That which I now tell you, is – that, if you don't remove that arm of yours from the doorway of this most respectable druggist, I shall have a good ground of action against you for assault and battery.” Similarly, Mao Zedong gave orders that any orders he might issue after taking sleeping pills were to be ignored. When, after having taken the pills, he ordered his aide to send an invitation to the American table tennis team to visit China (the beginning of Chinese-American relations) and the aide asked him, “Do your words count?” he answered, “Yes, they do. Do it quickly. Otherwise there won't be time.”
Organizations are more reliable tools for self-binding. The cocaine clinic in Denver and the Christmas clubs offer self-binding options that the individuals could not have come up with on their own and that were deliberately designed to help them to overcome their problems13 and to prevent them from rescinding their instructions. In Norway, the Law of Psychic Health Protection allows individuals to commit themselves voluntarily but irreversibly to a three-week treatment in a psychiatric institution. It seems, however, that the system does not work, because doctors have the right but not the duty to retain individuals once they are in the clinic. To make it effective, patients would have to be allowed to sue their hospital if, at their request, it released them prematurely.
In 1996, the state of Missouri began a self-exclusion program for compulsive gamblers. Anyone who signs up for a self-exclusion list is banned for life from entering any of Missouri's casino riverboats. If she tries to ignore the ban and gamble on one of Missouri's riverboats anyway, she is to be removed from the boat, and “the licensee shall cooperate with the commission agent in reporting the incident to the proper prosecuting authority and request charges be filed … for criminal trespassing, a class B misdemeanor.” Self-excluded gamblers are to be denied any winnings if they somehow manage to go aboard a riverboat, gamble, and win.
The state can also take a more active role, by imposing delays on abortion, gun purchase, or divorce (and marriage!) and by allowing consumers a three-day or week-long cooling-off period during which they can cancel purchases made in a moment of enthusiasm. To illustrate, consider the 238,292 purchasers of handguns in California in 1991. In the year following the purchase, 21.9 percent of the deaths among the purchasers were suicides by firearm, as against 0.9 percent in the general population. The temporal profile of the suicides among the purchasers illustrates strikingly the short half-life of emotions (Chapter 8): after the obligatory waiting time of fifteen days, the number of suicides in the first week was double that of the fourth, the number in the first month five times that of the twelfth, and the number in the first year six times that of the sixth. Had there been no waiting time, there would presumably have been more suicides; with a longer wait, fewer.
Softer methods are also used. At Gardermoen airport in Norway, tobacco products are no longer on open display in the tax-free shop, but shuttled away in a separate room, presumably to prevent cue-triggered cravings in smokers who are trying to quit.14 In Norwegian bars, you cannot order a double whisky (8 cl), but nothing prevents you from ordering two singles in succession.15 When Mayor Bloomberg tried to impose a similar limitation on the size of soda bottles, he failed.
Sometimes, political constitutions are understood as precommitment devices, or a form of collective self-paternalism. John Potter Stockton, writing in 1871, said that “constitutions are chains with which men bind themselves in their sane moments that they may not die by a suicidal hand in the day of their frenzy.” Another common metaphor is that constitutions are ties imposed by Peter when sober on Peter when drunk. Bicameralism is often cited as an example of political precommitment: by having all legislation pass through two houses, one creates time for impulsive passions to cool down and reason (or interest!) to regain the upper hand. In Chapter 25, I make some skeptical comments on this argument. Imposing delays on constitutional amendments has been justified by the same argument. If precommitment is understood as self-binding, however, the extension from the individual to the collective case, and from the intragenerational to the intergenerational case, is quite dubious. Rather than a community's binding itself, we find majorities binding minorities and the present generation binding the future. Moreover, since constitutions are typically written in turbulent times, framers or founders are often themselves in the grip of passion. Being “drunk,” they may not see the need to take precautions against drunkenness. Thus on September 7, 1789, when the French Assemblée Constituante was debating whether to write unicameralism or bicameralism into the constitution, the deputy Adrien Duquesnoy wrote the following entry into his journal: “If one can be allowed to make a probability assessment, it seems clear that the majority of the assembly will never vote for the two chambers. This outcome may have great disadvantages, but the situation is such, and the minds are so exalted, that no other is possible; perhaps it will be possible to make a change in a few years. One will come to understand that a unique assembly, in a nation as extremely impetuous as ours, can produce the most terrible effects.”
Bibliographical note
In this chapter I draw on my book Agir contre soi (Paris: Odile Jacob, 2007). The intrapsychic device of bundling or bunching the options has been extensively discussed by G. Ainslie, notably in Picoeconomics (Cambridge University Press, 1992). The information-avoidance strategy was proposed by J. Carrillo and I. Mariotti, “Strategic ignorance as a self-disciplining device,” Review of Economic Studies 67 (2000), 529–44, and confirmed experimentally by T. Brown, R. Croson, and T. Eckel, “Intra- and inter-personal strategic ignorance: a test of Carrillo and Mariotti” (accessible at http://stiet.cms.si.umich.edu/node/809). The discussion of decision myopia draws on O.-J. Skog, “Hyperbolic discounting, willpower, and addiction,” in J. Elster (ed.), Addiction: Entries and Exits (New York: Russell Sage Foundation, 1999). Strategic responses by sophisticated individuals who are aware of their propensity to discount the future hyperbolically are discussed by T. O'Donoghue and M. Rabin, “Doing it now or later,” American Economic Review 89 (1999), 103–24. The study of New Year's resolutions is J. Norcross, A. Ratzin, and D. Payne, “Ringing in the new year: the change processes and reported outcomes of resolutions,” Addictive Behaviors 14 (1989), 205–12, and the study of the countermeasures by the cigarette companies is M. Basil, D. Basil, and C. Schooler, “Cigarette advertising to counter New Year's resolutions,” Journal of Health Communication 5 (2000), 161–74. The idea of precommitment or self-binding to cope with one's irrational propensities is discussed in T. Schelling, “Egonomics, or the art of self-management,” American Economic Review: Papers and Proceedings 68 (1978), 290–4, and in several of his later publications. I have discussed it in Ulysses and the Sirens, rev. edn (Cambridge University Press, 1984); in Ulysses Unbound (Cambridge University Press, 2000); and in “Don't burn your bridge before you come to it: ambiguities and complexities of precommitment,” Texas Law Review 81 (2003), 1751–88. A book-length treatment of the failure to take prescribed medications is G. Reach, Pourquoi se soigne-t-on? (Paris: Éditions de Bord de l'Eau, 2005). The story about Coleridge is found in Thomas de Quincy, Confessions of an Opium Eater (London: Penguin, 1968), p. 145. The story about Mao is found in J. Chang and J. Halliday, Mao: The Unknown Story (New York: Knopf, 2005), 580–1. The increase in the sales of alcohol in Sweden is documented in O.-J. Skog, “An experimental study of a change from over-the-counter to self-service sales of alcoholic beverages in monopoly outlets,” Journal of Studies on Alcohol 61 (2000), 95–100. The data on handguns and suicides in California are from G. Wintermute et al., “Mortality among recent purchasers of handguns,” New England Journal of Medicine 341 (1999), 1583–9. Whereas in Ulysses and the Sirens I was enthusiastic about the idea of constitutions as precommitment devices, I recanted in Ulysses Unbound.
1 Students of economics, in particular, seem to behave in this way.
2 For an illustration, suppose that the present value of 1 unit of utility at time t in the future is 1/(1 + t) and that the agent is twice exposed to the choice between a small reward of 3 and a large of 10. The small rewards become available at times 0 and 6, the large ones at times 3 and 9. At time 0, the present value of the first large reward is 10/(1 + 3) = 2.5, which is less than the present (instantaneous) value of the small reward. If the choice is made on the basis of this comparison only, the small reward will be chosen. The same choice, for identical reasons, will be made at time 6. As the sum of the present values of the two small rewards is 3 + 3/(1 + 6) ≅ 3.43 and the sum of the present values of the two large rewards is 10/(1 + 3) + 10/(1 + 9) = 3.5, bunching will make the agent prefer the two large rewards.
3 If the agent suffers from “decision myopia” (Chapter 14), the bunching may not work. Suppose, namely, that the agent bundles his choices well ahead of the time when the first small reward becomes available. At the time of bunching, the present value of the stream of large rewards is greater than that of the stream of small rewards, and the agent firmly intends to wait for the first large reward. As he moves closer in time to the moment when the first reward becomes available, this intention may or may not survive. This preference reversal is not due to hyperbolic discounting per se, but to decision myopia.
4 For an analogy, suppose that by voting I could influence many others, who would otherwise have abstained, to cast a vote as well. Note that on this assumption, no magical thinking is involved, only a causal multiplier effect.
5 In voting, the effect of magical thinking is to help us overcome the propensity to socially harmful rational behavior rather than to counteract irrationality.
6 According to Montaigne, the problem is quite general: “It is perhaps easier to do without women altogether than duly and scrupulously to restrict yourself to the company of your wife: a man has more means of living an unworried life in poverty than in duly controlled abundance; behavior governed by reason is more thorny than abstinence.” He had a generous conception of reason: “I now defend myself against temperance as I used to do against voluptuousness … Wisdom has its excesses and has no less need of moderation than folly.”
7 Chief Justice John Marshall used a more transparent device to get around the rule that the court would indulge in drinking wine only when it was raining. Looking out of the window on a sunny day, he would say that “our jurisdiction extends over so large a territory that the doctrine of chances makes it certain that it must be raining somewhere.”
8 They may believe, erroneously (Chapter 12), that cross-situational consistency will induce cross-situational triggering of breakdowns.
9 For an illustration, suppose that the present value of 1 unit of utility at time t in the future is 1/(1 + t) and that I will suffer increasingly by delaying my visit to the dentist: if I go today I suffer a pain of –2.75, tomorrow it will be –5, and the day after –9. From today's perspective, the present values are, respectively, –2.75, –5/(1 + 1) = –2.5, and –9/(1 + 2) = –3. Hence it might seem that the optimal choice from today's perspective is to postpone the visit until tomorrow. However, being sophisticated I know (today) that tomorrow the present value of going tomorrow will be –5 and that of going the day after –9/(1 + 1) = –4.5, inducing a preference to wait until the day after. Since today, however, I prefer going today to going the day after tomorrow, I go today.
10 For an illustration, suppose that the present value of 1 unit of utility at time t in the future is 1/(1 + t) and that I can benefit more and more by delaying my consumption of a bottle of wine: if I drink it this year I derive a pleasure of 2.75, next year it will be 5, and the year after 9. From today's perspective, the present values are, respectively, 2.75, 5/(1 + 1) = 2.5 and 9/(1 + 2) = 3. Hence it might seem that the optimal choice from the first year's perspective is to drink the wine the year after next. However, being sophisticated I know (this year) that next year the present value of drinking it next year will be 5 and that of drinking it the year after 9/(1 + 1) = 4.5, inducing a preference to drink it the next year. Since in the first year, however, I prefer drinking it the first year to drinking it next year, I drink it today.
11 The most common form is by oral intake, which works by causing the consumption of alcohol to make you sick. The (physically inefficacious but psychologically efficacious) implantation under the skin is less common.
12 The phenomenon, briefly mentioned in the text, of precommitting oneself while being in the grip of passion and fearful that it will abate is much less common.
13 Safes with timing devices, by contrast, were not made to help people fight their drinking problems.
14 The importance of cue-triggering was observed in Sweden, where sales of alcohol went up by 10 percent after a change from over-the-counter to self-service sales.
15 When he was a political commentator, my father argued, unsuccessfully, that the government should ban advertising that offers products at the price of $499.95 and similar prices slightly below a round number, because they exploit consumer irrationality.