Procrastination, Deadlines, and Performance: Self-Control by Precommitment PDF

PSYCHOLOGICAL SCIENCE Research Article PROCRASTINATION, DEADLINES, AND PERFORMANCE: Self-Control by Precommitment...

PSYCHOLOGICAL SCIENCE Research Article PROCRASTINATION, DEADLINES, AND PERFORMANCE: Self-Control by Precommitment Dan Ariely1 and Klaus Wertenbroch2 1 Massachusetts Institute of Technology and 2INSEAD, Fontainebleau, France Abstract—Procrastination is all too familiar to most people. People One of the causes for the apparent changes in preferences over delay writing up their research (so we hear!), repeatedly declare they time is changes in the saliency of the costs and benefits of the activity will start their diets tomorrow, or postpone until next week doing odd in question (Akerlof, 1991). For example, well in advance of actually jobs around the house. Yet people also sometimes attempt to control taking on the responsibility of writing a book, the benefits of com- their procrastination by setting deadlines for themselves. In this arti- pleting such a task loom large, and the costs seem small. Conse- cle, we pose three questions: (a) Are people willing to self-impose quently, authors take on such tasks. But as the deadline draws closer, meaningful (i.e., costly) deadlines to overcome procrastination? (b) the saliency of the costs and benefits changes. Authors become in- Are self-imposed deadlines effective in improving task performance? creasingly aware of the costs (the time needed for completing the (c) When self-imposing deadlines, do people set them optimally, for task), while the benefits become increasingly less clear. maximum performance enhancement? A set of studies examined these Although such time-inconsistent preferences may form serious issues experimentally, showing that the answer is “yes” to the first two obstacles to following a planned course of action, they can be over- questions, and “no” to the third. People have self-control problems, come. In addition to exercising willpower to resist temptation (Hoch they recognize them, and they try to control them by self-imposing & Loewenstein, 1991; Muraven & Baumeister, 2000), people can costly deadlines. These deadlines help people control procrastination, bind, or precommit, their own behavior (Prelec, 1989; Schelling, but they are not as effective as some externally imposed deadlines in 1992; Strotz, 1956; Thaler & Shefrin, 1981; Wertenbroch, 1998). For improving task performance. example, people who want to diet, but recognize that crème brûlée will tempt them to deviate from their plan, can preempt temptation by going to a restaurant with a less tempting menu. A wealth of anec- Good resolutions are useless attempts to interfere with scientific laws. Their or- dotes describes examples of binding behaviors, including frequenting igin is pure vanity. Their result is absolutely nil. health retreats where some food types are not available, saving in —Oscar Wilde, The Picture of Dorian Gray non-interest-bearing Christmas clubs, or buying small packages of cigarettes in order to reduce consumption (Wertenbroch, 1998). An Self-control problems arise when preferences are inconsistent extreme example was provided by Schelling (1992), who described across time or context (e.g., Ainslie, 1975; Loewenstein, 1996). For drug addicts sending self-incriminating letters to be held in trust (and example, before going to a restaurant dieters may choose not to have mailed to the person they fear the most will find out about their addic- crème brûlée, but when the time comes to have dessert they may give tion) in the event of a relapse into drug use. What characterizes bind- in to the temptation and order it after all, only to regret having eaten it ing behavior is the voluntary imposition of constraints (that are costly after the meal is over. The issue is not whether having crème brûlée is to overcome) on one’s future choices in a strategic attempt to resist right or wrong, but that ordering it is inconsistent with the decision future temptations. makers’ preferences both before and after the event. One way to think Although time-inconsistent preferences and self-control have about these issues is that individuals have a set of preferences, X, at been the subject of much theoretical analysis in psychology and eco- some point in time (or under a certain set of environmental condi- nomics (Ainslie, 1975; Bargh & Gollwitzer, 1994; Hoch & Loewen- tions) and a different set of preferences, Y, at some other point in stein, 1991; Muraven & Baumeister, 2000; O’Donoghue & Rabin, time. In the case of the crème brûlée, dieters may prefer not to con- 1999, 2000; Prelec, 1989; Strotz, 1956; Thaler & Shefrin, 1981; sume it (Y) before going to the restaurant, prefer to eat it (X) when Tversky & Shafir, 1992), controlled empirical evidence of self-con- ordering dessert and consuming it at the restaurant, and prefer not to trol strategies is scarce. The few studies that have looked at self-con- have eaten it after the meal is over (Y). This type of systematic pref- trol show that people do attempt to impose costly restrictions on erence reversal is often described by hyperbolic time discounting themselves. In the domain of consumer choice, Wertenbroch (1998) (e.g., Ainslie, 1975; Kirby, 1997; Laibson, 1997), under which imme- showed with experimental and field data that people are willing to diately available rewards have a disproportionate effect on prefer- forgo quantity discounts on goods that they may be tempted to over- ences relative to more delayed rewards, causing a time-inconsistent consume, effectively paying a “self-control premium” to implement taste for immediate gratification. Crème brûlée poses but a minor a precommitment strategy of rationing their own consumption of self-control problem. Examples of more important self-control prob- such “vices.” Similarly, Read, Loewenstein, and Kalyanaraman (1999) lems include not exercising enough, scratching a rash, nail biting, asked participants to pick three rental movies either simultaneously smoking, engaging in unsafe sex, abusing drugs, overspending, pro- (for later consumption) or sequentially (for more immediate consump- crastination, and so forth. tion). Their results showed that participants used the simultaneous choices to precommit to watching more “high-brow” (as opposed to more tempting “low-brow”) movies. In the domain of medical testing, Trope and Fishbach (2000) allowed participants to set the magnitude Address correspondence to Dan Ariely, Massachusetts Institute of Technol- of self-imposed penalties for failing to undergo small, unpleasant ogy, 38 Memorial Dr., E56-329, Cambridge, MA 02142; e-mail: [email protected]. medical procedures. Their results showed that participants used these VOL. 13, NO. 3, MAY 2002 Copyright © 2002 American Psychological Society 219  PSYCHOLOGICAL SCIENCE Procrastination, Deadlines, and Performance penalties strategically as precommitment devices, setting higher pen- ticipated in the class via interactive video. The two sections of the alties for more aversive procedures. course (which, based on records provided by the executive-education What remains unclear from the studies that have documented such program, did not differ in overall academic performance) were each self-control behavior is the extent to which attempts to impose restric- assigned to a different condition (so there was no random assignment tions on oneself are successful. The work we report here examined self- of individuals to treatments but rather a random assignment of sec- control empirically, with a focus on procrastination. In particular, we tions to treatments). were interested in the effectiveness of setting potentially costly dead- lines as a way to overcome procrastination. To address this issue, we Procedure looked at tasks on which performance could be evaluated objectively. Using performance measures, we could test not only whether people use During the first lecture, the instructor went over the syllabus, self-imposed deadlines as precommitment mechanisms, but also which included instructions for the study. One part of the course re- whether or not these mechanisms improve performance. We asked three quirements was to write three short papers. Students in the no-choice questions regarding procrastination, self-control, and performance: section (48 students) were given fixed, evenly spaced deadlines for the papers (a paper at the end of each third of the course). Students in the Do people self-impose costly deadlines on tasks in which procrasti- free-choice section (51 students) were given detailed instructions nation may impede performance? about setting their own deadlines (as in the pilot studies). These in- Are people correct in imposing deadlines on themselves? In other structions indicated that each student was free to choose the dates by words, are self-imposed deadlines effective in improving task per- which he or she wanted to hand in the short papers. Four external con- formance? straints were set regarding the dates: First, students had to hand in Do people set their deadlines optimally, for maximum performance their papers no later than the last lecture; second, students had to an- enhancement? nounce the deadlines for submission prior to the second lecture; third, the dates were final and could not be changed; and fourth, the dates PILOT STUDIES were binding, such that each day of delay beyond the deadline would The two pilot studies took place within the context of a semester- cause a 1% penalty in the paper’s overall grade. Finally, it was ex- long course (14 weeks) at the Massachusetts Institute of Technology plained clearly that there were no grade advantages for early submis- (MIT). Participants were students in the class, and as part of their sions because the instructor would not provide grades or feedback on course requirement had to write either three short papers (Pilot Study the assignments before the end of the course. Explaining to the stu- 1) or one short paper (Pilot Study 2). The instructor explained that dents that there would be no feedback before the end of the course was each student was free to choose the dates by which he or she commit- important because it eliminated incentives for students to hand in pa- ted to hand in the short papers, but that the deadlines had to be an- pers early in order to get feedback that they could use to improve sub- nounced in advance and were binding. sequent papers. Each of the deadlines was scored by taking its distance (number of In fact, the external incentives for the students in the free-choice days) from the last day of class. Thus, a score of zero implies a planned section encouraged submission of all three papers on the last possi- submission on the last day of class (as would be predicted in the ab- ble day. By setting their deadlines as late as possible, the students sence of self-control problems). Any other response indicates a more would have the most time to work on the papers, the highest flexibil- severe deadline than necessary. In the first pilot study, the mean dead- ity in arranging their workload, and the opportunity to learn the most line across all three papers was 21.2 days before the end of the course, about the topic before submitting the papers. Students also had an and significantly earlier than the last possible deadline, t(83)
8.05, p  incentive to set submission dates late because the penalty would be.001. The mean deadline was 32.8 days before the end of the course for applied only to late submissions and not to early ones. Finally, stu- the first paper, t(27)
5.72, p .001; 20.4 days before the end for the dents who wanted to submit assignments early could privately plan second paper, t(27)
5.04, p .001; and 10.4 days before the end for to do so without precommitting to the instructor. Of course, such pri- the third paper, t(27)
4.45, p .001. These results show that the stu- vate deadlines might be less psychologically meaningful than the dents set themselves deadlines well before the last day of class. deadlines they set with the instructor, and hence more pliant and less To rule out the possibility that students self-impose deadlines be- effective. cause of a preference for distributing events evenly over time (Loe- wenstein & Prelec, 1993), in Pilot Study 2 we gave the students a Results and Discussion single task. The mean self-imposed deadline in this case was 41.59 days before the end of the course, t(21)
15.44, p .001, suggesting First, we examined the declared deadlines for each of the three pa- that setting early deadlines is strategic, and not an outcome of a desire pers. Again, each deadline was scored by taking its distance (number to space tasks evenly. of days) from the last day of class, so that a score of zero indicates a planned submission on the last day of class (perfectly normative). Other responses indicate the severity of the deadlines the students im- STUDY 1: THE FREE-CHOICE/NO-CHOICE STUDY posed on themselves. The mean deadlines were significantly earlier than the last possible deadline—41.78 days before the end of the Method course for the first paper, t(44)
8.41, p .001; 26.07 days before Participants the end for the second paper, t(44)
8.10, p .001; and 9.84 days before the end for the third paper, t(44)
4.97, p .001. Figure 1 Study 1 took place during a semester-long executive-education shows that only 43 deadlines (32%) were set for the final week of course at MIT. Participants were 99 professionals, most of whom par- class. The majority of the deadlines were set prior to the last lecture, 220 VOL. 13, NO. 3, MAY 2002  PSYCHOLOGICAL SCIENCE Dan Ariely and Klaus Wertenbroch Fig. 1. Frequency distribution of the declared deadlines in Study 1 as a function of the week of class (Week 1 is the first week, and Week 14 the last week), plotted separately for the three papers. and in fact, only 12 students (27%) chose to submit all three papers on self-control problems, and they use deadlines to overcome these prob- the last day of class.1 lems, but do not set these deadlines optimally, greater flexibility might These results indicate that people are willing to self-impose dead- lead to lower grades. In sum, flexibility, compared with evenly spaced lines to overcome procrastination, even when these deadlines are deadlines, should lead to lower grades only if people have self-control costly (our first question). The students could have chosen less bind- problems yet do not set their own deadlines optimally.2 The results ing private deadlines, but instead chose deadlines that involved more supported the third prediction. The grades in the no-choice section commitment and greater potential cost (a grade penalty for being late). (M
88.76) were higher than the grades in the free-choice section (M
It seems that they were willing to take the risk of losing grade points 85.67), t(97)
3.03, p
.003. to apply the self-control mechanism of precommitment. In addition to having a direct effect on performance, deadlines can Next, we compared the grades in the two sections to see if flexibil- have a secondary effect on other aspects of performance that also re- ity in setting deadlines caused higher or lower grades compared with quire the investment of time as a resource. A natural candidate for externally imposed, evenly spaced deadlines. There were three possi- this measure is the students’ performance on a final project that was ble predictions: (a) If students do not have self-control problems, due on the last day of class. Grades for the final project showed the greater flexibility should lead to higher grades. (b) If students do have same effect: Scores were lower in the free-choice section (M
77) self-control problems, and if they both use deadlines to overcome than in the no-choice section (M
86), t(95)
4.15, p .001, sug- these problems and set these deadlines optimally, greater flexibility gesting that students with late self-imposed deadlines for the three fo- should allow them to achieve higher grades. (c) If students do have 2. We use the term optimally relative to participants’ performance under the 1. Because of missing data, the percentages do not fit with the total number evenly spaced deadlines in the no-choice section. If performance can be im- of students in the class. proved, it is suboptimal by definition. VOL. 13, NO. 3, MAY 2002 221  PSYCHOLOGICAL SCIENCE Procrastination, Deadlines, and Performance cal tasks might not have had sufficient time to dedicate to the final be contingent on the quality of the proofreading, with 10¢ paid per project. correctly detected error and a $1 penalty for each day of delay. A total Although the students were instructed about the penalties associ- of 60 students participated in the study, randomly assigned to the three ated with missing the deadlines, it is possible that students in the free- experimental conditions. choice section, compared with those in the no-choice section, treated these deadlines as less binding because they were self-imposed.3 To Procedure demonstrate that the better performance in the no-choice section was caused by the timing of the deadlines and not by the perceived force of We chose a task that people cared about but one whose outcome the externally imposed deadlines, we compared the performance of was not central to their lives (in contrast to the course grades in the the students in the no-choice section with the performance of those previous studies). We also wanted a task for which performance scores students in the free-choice section who chose evenly spaced (or almost would be more objective and for which we could pay participants ac- evenly spaced) dates for submission. This comparison isolates the ef- cordingly. We therefore designed a proofreading task in which we de- fect of deadline type (self vs. external) on performance. If these two liberately planted spelling and grammatical mistakes. We used a groups with similarly spaced deadlines differed in their performance, postmodern text generator4 to create text that was grammatically cor- the overall difference between the sections could be attributed to the rect but not meaningful, as shown by the following sample: nature of the deadlines (self vs. external). However, if students who spaced their deadlines evenly showed similar performance regardless “Sexual identity is intrinsically impossible,” says Foucault; however, according of the nature of the deadline, the overall difference between the sec- to de Selby, it is not so much sexual identity that is intrinsically impossible, but rather the dialectic, and some would say the stasis, of sexual identity. Thus, tions was likely due to the timing of the deadlines. The results showed D’Erlette holds that we have to choose between premodern dialectic theory that the performance difference between the two sections decreased and subcultural feminism imputing the role of the observer as poet. dramatically and became nonsignificant when only those students who had evenly spaced deadlines were included in the analysis (effect size We created three such texts with a length of about 10 pages each, and reduced by 59%). This comparison suggests that the overall effect of inserted in each of them a total of 100 grammatical and spelling er- self-imposing deadlines was due primarily to the timing of the dead- rors. lines, not just a weaker perceived potency of self-imposed deadlines. Study 2 included three different conditions. In each condition, we clearly explained to the participants that their payoffs would depend STUDY 2: THE PROOFREADING STUDY on how many errors they detected and on the time of submission of each proofread text. Participants were told that submitting their tasks The combined results of the pilot studies and Study 1 suggest that early was permitted (without increasing their compensation), but that decision makers who face situations in which they can self-impose delay in submission would result in a penalty of $1 for each day of de- deadlines recognize two conflicting forces. On the one hand, they real- lay. In the evenly-spaced-deadlines condition, participants had to sub- ize the value of binding themselves to overcome procrastination; on mit one of the three texts every 7 days; in the end-deadline condition, the other hand, they understand the normative reasons to set the dead- they had to submit all three texts at the end of 3 weeks (21 days); and lines as late as possible. We propose that decision makers combine in the self-imposed-deadlines condition, they had to choose their own these two perspectives and come up with deadlines whose timing is deadline for each of the three texts within the 3-week window (as in suboptimal (as shown in Study 1) but better than delaying all dead- the previous studies). lines to the last possible day. Thus, we hypothesize that performance under self-imposed deadlines is lower than performance under exter- nally imposed, evenly spaced deadlines but higher than performance Results and Discussion under maximally delayed deadlines (when all tasks are due simulta- neously at the end of the period). To examine this hypothesis, we now First, we determined whether the self-imposed-deadlines condition focus on our second and third questions: whether self-imposed dead- replicated the results of the previous studies. The results showed that lines improve performance and, if so, whether people know how to set participants in this condition chose to space out their proofreading deadlines for maximum performance enhancement. Study 2 was de- tasks, F(2, 38)
63.28, p .001, thus showing a preference for self- signed to examine these questions in a controlled experimental setup, imposing costly deadlines. providing a more sensitive test of the effect of deadlines on perfor- We analyzed three aspects of performance across the different con- mance than Study 1 did, coupled with a more objective performance ditions: number of errors detected, delays in submissions, and earn- measure. ings (see Fig. 2). All differences were statistically significant (all ps .01) in the expected direction. As predicted, the number of errors cor- rectly detected was highest in the evenly-spaced-deadlines condition, Method followed by the self-imposed-deadlines condition, with the lowest per- Participants formance in the end-deadline condition. Results were similar for par- ticipants’ delays in submitting their proofreading work (in this case, In MIT’s newspaper and on bulletin boards, we placed an ad look- shorter delays resulted in higher payoffs). Participants’ earnings re- ing for “native English speakers to help us proofread papers by other flected a combination of error detection and delay and thus show the students to evaluate writing skills.” We also noted that payment would same pattern of results. 3. Evidence against this argument is that all students (in both sections) 4. The text-generating engine is on the Web at http://www.elsewhere.org/ handed in their papers on or before the deadlines. cgi-bin/postmodern/. 222 VOL. 13, NO. 3, MAY 2002  PSYCHOLOGICAL SCIENCE Dan Ariely and Klaus Wertenbroch Fig. 2. Mean errors detected (a), delays in submissions (b), and earnings (c) in Study 2, compared across the three conditions (error bars are based on standard errors). Delays are measured in days, earnings in dollars. Next, we examined the same measures focusing on the participants ticipants in the self-imposed-deadlines condition, and finally by par- in the self-imposed-deadlines condition who had spaced their tasks ticipants in the end-deadline condition. evenly, or approximately evenly (n
10). Mirroring the results of Study In addition, we asked participants to estimate how much time they 1, the differences between the evenly-spaced-deadlines condition and had spent on each of the three texts. The time estimates revealed a mirror the (“sophisticated”) self-imposed-deadlines condition decreased dra- image of the subjective evaluations, F(2, 57)
45.76, p .001, indicat- matically and became nonsignificant for all dependent measures: delay ing that increased time spent on the task caused the evaluation to be more in submissions (effect size reduced by 55%), errors detected (effect negative. Participants in the evenly-spaced-deadlines condition indicated size reduced by 79%), and earnings (effect size reduced by 55%). This they spent the most time on the task (M
84 min), participants in the reduction in effect sizes provides additional evidence that a central self-imposed-deadlines condition spent an intermediate amount of time cause of the lower performance in the self-imposed-deadlines condi- on the task (M
69.9 min), and participants in the end-deadline condi- tion compared with the evenly-spaced-deadlines condition was subop- tion spent the least time on the task (M
50.8 min). Taken together, the timal spacing of the tasks. results show that when deadline constraints increased, performance im- Finally, we asked participants to evaluate their overall experience proved, time spent on the task increased, and enjoyment of the task de- on five attributes: how much they liked the task, how interesting it creased (because of enhanced recognition of the true low quality of the was, how good the quality of the writing was, how good the grammat- texts). The effectiveness of the constraints themselves depended on the ical quality was, and how effectively the text communicated the ideas type of constraint—self-imposed deadlines improved performance, but contained in it. Responses to all questions were on a 100-point scale, not to the same degree as evenly spaced deadlines. on which higher numbers represented higher quality ratings. An analy- sis of the average subjective evaluation across the five questions re- vealed a pattern that was the opposite of the performance results, F(2, GENERAL DISCUSSION 57)
17.06, p .001. Participants in the evenly-spaced-deadlines The studies presented here show that people sometimes impose condition liked the task the least (M
22.1), followed by the partici- deadlines on themselves, even when missing these deadlines leads to pants in the self-imposed-deadlines condition (M
28.12), followed penalties. In a world without self-control problems, such behavior by participants in the end-deadline condition, who liked the task the would seem nonnormative. A rational decision maker with time-consis- most, or disliked it the least (M
37.9). These results are not surpris- tent preferences would not impose constraints on his or her choices. But ing, as the texts were meaningless and the tasks were boring, if not an- if people impulsively procrastinate, and if they also are aware of their noying. We suggest that the pattern would have been reversed if the procrastination problems (e.g., Bénabou & Tirole, in press; O’Dono- task had been inherently enjoyable; participants in the evenly-spaced- ghue & Rabin, 1999), self-imposing costly deadlines can be strategic deadlines condition would have enjoyed it the most, followed by par- and reasonable. Study 1 demonstrated that self-imposed deadlines do VOL. 13, NO. 3, MAY 2002 223  PSYCHOLOGICAL SCIENCE Procrastination, Deadlines, and Performance not enhance performance as much as externally imposed, evenly spaced Akerlof, G.A. (1991). Procrastination and obedience. American Economic Review, 81, 1–19. Bargh, J.A., & Gollwitzer, P.M. (1994). Integrative views of motivation, cognition and deadlines. The results from Study 2 show that performance under self- emotion. In W.D. Spalding (Ed.), The Nebraska Symposium on Motivation: Vol. 41. imposed deadlines is lower than performance under evenly spaced dead- Integrative views of motivation, cognition and emotion (pp. 71–124). Lincoln: Uni- lines, but higher than performance under maximally delayed deadlines. versity of Nebraska Press. Bénabou, R., & Tirole, J. (in press). Self-knowledge and self-regulation: An economic ap- We can now return to the three questions posed earlier. (a) Do people proach. In I. Brocas & J.D. Carrillo (Eds.), Collected essays in psychology and eco- self-impose costly deadlines to overcome procrastination? (b) Are self- nomics. Oxford, England: Oxford University Press. imposed deadlines effective in improving task performance? (c) Do Hoch, S.J., & Loewenstein, G.F. (1991). Time-inconsistent preferences and consumer self- control. Journal of Consumer Research, 17, 492–507. people set self-imposed deadlines optimally? The answer to the first two Kirby, K.N. (1997). Bidding on the future: Evidence against normative discounting of de- questions is “yes,” and the answer to the last question is “no.” Our find- layed rewards. Journal of Experimental Psychology: General, 126, 54–70. Laibson, D. (1997). Golden eggs and hyperbolic discounting. Quarterly Journal of Eco- ings demonstrate that people understand the value of binding them- nomics, 62, 443–477. selves to overcome procrastination, even in the face of strong normative Loewenstein, G.F. (1996). Out of control: Visceral influences on behavior. Organizational reasons for setting deadlines as late as possible. Our participants showed Behavior and Human Decision Processes, 65, 272–292. Loewenstein, G.F., & Prelec, D. (1993). Preferences for sequences of outcomes. Psycho- some sophistication in their understanding of their own procrastination logical Review, 100, 91–108. problems, but many did not set their deadlines to bind themselves opti- Muraven, M., & Baumeister, R.F. (2000). Self-regulation and depletion of limited resources: mally. Whether our evidence of such “imperfect” sophistication (or Does self-control resemble a muscle? Psychological Bulletin, 126, 247–259. O’Donoghue, T., & Rabin, M. (1999). Doing it now or later. American Economic Review, “partial naiveté”) reflects biased self-perception, cognitive limitations in 89, 103–124. calibrating deadlines, or a deliberate mixed strategy of balancing flexi- O’Donoghue, T., & Rabin, M. (2000). The economics of immediate gratification. Journal of Behavioral Decision Making, 13, 233–250. bility and self-control is a question for future research. What is clear Prelec, D. (1989). Decreasing impatience: Definition and consequences. Unpublished manu- from our empirical evidence is that procrastination is a real behavioral script, Massachusetts Institute of Technology, Cambridge, MA. problem, that people strategically try to curb it by using costly self-im- Read, D., Loewenstein, G.F., & Kalyanaraman, S. (1999). Mixing virtue and vice: The combined effects of hyperbolic discounting and diversification. Journal of Behav- posed deadlines, and that self-imposed deadlines are not always as ef- ioral Decision Making, 12, 257–273. fective as some external deadlines in boosting task performance. Schelling, T. (1992). Self-command: A new discipline. In J. Elster & G.F. Loewenstein (Eds.), Choice over time (pp. 167–176). New York: Russell Sage Foundation. Strotz, R.H. (1956). Myopia and inconsistency in dynamic utility maximization. Review of Acknowledgments—We thank Jose Fernando Camoes Mendonca Oliveira Economic Studies, 23, 165–180. Silva for his humor and advice, and the editor and two anonymous review- Thaler, R.H., & Shefrin, H.M. (1981). An economic theory of self-control. Journal of Po- litical Economy, 89, 392–406. ers for their helpful comments. Trope, Y., & Fishbach, A. (2000). Counteractive self-control in overcoming temptation. Journal of Personality and Social Psychology, 79, 493–506. Tversky, A., & Shafir, E. (1992). Choice under conflict: The dynamics of deferred deci- sion. Psychological Science, 3, 358–361. REFERENCES Wertenbroch, K. (1998). Consumption self-control by rationing purchase quantities of vir- tue and vice. Marketing Science, 17, 317–337. Ainslie, G. (1975). Specious reward: A behavioral theory of impulsiveness and impulse control. Psychological Bulletin, 82, 463–496. (RECEIVED 1/2/01; REVISION ACCEPTED 6/13/01) 224 VOL. 13, NO. 3, MAY 2002