American Economic Association - CMU
American Economic AssociationNegative Time PreferenceAuthor(s): George Loewenstein and Drazen PrelecSource: The American Economic Review, Vol. 81, No. 2, Papers and Proceedings of theHundred and Third Annual Meeting of the American Economic Association (May, 1991), pp.347-352Published by: American Economic AssociationStable URL: http://www.jstor.org/stable/2006883 .Accessed: 22/07/2011 12:37Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at ms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at erCode aea. .Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@jstor.org.American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to TheAmerican Economic Review.http://www.jstor.org
Negative Time PreferenceBy GEORGE LOEWENSTEIN AND DRAZEN PRELEC*reached at the end. Thus a declining seriesof consumption levels ought to be preferable to an increasing series, holding totalconsumption constant.In this paper, we present a short selectionof findings (reported more fully in our 1990paper) that sharply contradict the normativesequencing rule just described. To most persons, a deteriorating series of utility levels isa rather close approximation to the leastattractive of all possible patterns, regardlessof the nature of events that are being ordered. As a secondary violation of the discounted utility model, the preferences ofmany people are not additive. Such additivity violations often reflect a concern forspreading utility levels evenly over time thatis not attributable to diminishing marginalutility within periods.The man who lives within his incomeis naturally contented with his situation, which, by continual, though smallaccumulations, is growing better andbetter every day. He is enabled gradually to relax, both in the rigour of hisparsimony and in the severity of hisapplication; and he feels with doublesatisfaction this gradual increase ofease and enjoyment, from having feltbefore the hardship which attendedthe want of them.Adam SmithThe Theoryof Moral SentimentsPlanning for the future invariably requires one to choose among alternative sequences of outcomes. Even simple shortterm scheduling decisions about work, play,chores, vacations, etc., involve choosing between sequences, because events that takeup time cannot be rescheduled withoutchanging the timing of other activities.Most economic analyses of preferencesbetween temporally spaced sequences relyon the discounted utility model, along withthe assumption of positive time preferenceand diminishing marginal utility.1 Barringany preferential interactions across differenttime periods, the predictions of this modelfor determining the optimal sequencing of agiven set of events are simple: Place thebest event at the start, then proceed indescending order until the worst event isI. Sequences vs. Simple OutcomesSeveral recent studies have documentedan apparently negative rate of time preference for choices among outcome sequences.Loewenstein and N. Sicherman (1991) foundthat a majority of museum visitors preferredincreasing wage profiles over those that areflat or decline over time (holding total valueconstant). Pointing out that the flat anddeclining wage profile could produce adominating consumption stream through asuitable savings program, did not have muchimpact on preference. C. Varey and D.Kahneman (1990) found that subjectsstrongly preferred brief sequences of decreasing discomfort, even at the cost of experiencing overall greater discomfort, whileW. T. Ross and I. Simonson (1990) showedthat people prefer sequences that end on agood note.Preference for improvement is an overdetermined phenomenon, driven in partby anticipatory savoring and dread (Loewenstein, 1987), and in part by loss aver-*Department of Social and Decision Sciences,Carnegie Mellon University, Pittsburgh, PA 15215, andHarvard Graduate School of Business Administration,Boston, MA 02136, respectively.1The applications of this model have been remarkably diverse, ranging from savings behavior, educational investment, labor supply, bargaining, to criminal behavior and addiction. In recent years, however,its descriptive validity has come under challenge (seeour forthcoming article, and Loewenstein and RichardThaler, 1990).347
348AEA PAPERS AND PROCEEDINGSsion (A. Tversky and Kahneman, 1990) andhabit forming changes in the adaptationlevel (J. Duesenberry, 1949). Savoring anddread contribute to the preference for improvement because, for gains, improving sequences allow decision makers to savor thebest until the end of the sequence. Withlosses, getting the worst outcomes over withquickly eliminates dread. Adaptation andloss aversion induce preference for improvement because, over time, people tendto assimilate to ongoing stimuli and to evaluate new stimuli relative to their assimilation level. Thus, changes in, rather thanlevels of, consumption are the carriers ofvalue. Improving sequences afford a continual series of positive departures (gains) fromone's adaptation level; declining sequencesprovide a series of relative losses.Savoring and dread apply to single outcomes as well as to sequences, but assimilation and loss aversion apply only to sequences. The fact that two motives operatefor sequences but only one for simple outcomes suggests that the tendency to deferdesirable outcomes will be stronger whenthose outcomes are embedded in sequences.Such a pattern is illustrated by a surveyconducted with undergraduates at HarvardUniversity, who were asked the followingthree questions:1. Which would you prefer if both werefree?A. Dinner at a fancy French restaurant (86%)B. Dinner at a local Greek restaurant (14%)For those who prefer French:2. Which would you prefer?C. Dinner at the French restauranton Friday in 1 month (80%)D. Dinner at the French restauranton Friday in 2 months (20%)3. Which would you prefer?E. Dinner at the French restauranton Friday in 1 month and dinnerat the Greek restaurant on Friday in 2 months (43%)F. Dinner at the Greek restauranton Friday in 1 month and dinnerat the French restaurant on Friday in 2 months (57%)MAY1991We anticipated that more people would delay the fancy French dinner when it wascombined in a sequence with the Greekdinner than when it was expressed as asingle outcome prospect. This is indeed whathappened. Of the 86 percent of subjectswho preferred the fancy French dinner, 80percent preferred a more immediate dinner(option C) over a more delayed dinner (option D). However, when the French dinnerwas composed into a sequence with theGreek dinner, a slight majority (57 percent)preferred to have the better dinner comelater. Even with single-outcome events thereis some motivation to defer the French dinner-witness the 20 percent of subjects whoopted for the longer delay. However, thistendency is stronger for sequences than forindividual items.A similar pattern is observed when "Dinner at home" is substituted for the Greekdinner. Since most people eat dinner athome on most nights anyway, embeddingthe French dinner in an explicit binary sequence does not introduce any real modification of the problem, but the subject isreminded that the choice is "really" between complete sequences. Like other framing effects, such reminders cause preferences to shift, in this case in favor of theimproving sequence.II. WhatIs a Sequence?If impatience and the desire for improvement are simultaneously present within asingle individual, what determines the prepotent motive on a given occasion? A reasonable conjecture is that the desire forimprovement depends on the "integrity" ofthe sequence-theextent to which theevents that comprise it are of a similar type,are regularly spaced, and are not stretchedtoo far apart.The following example illustrates how theintegrity of sequence can be reduced in apredictable way. Visitors to a science museum were asked to choose between alternative scheduling of two visits to a city wherethe respondent once lived, one to be spentwith "an irritating, abrasive, aunt who is ahorrendous cook," the other with "former
INTERTEMPORALCHOICEVOL. 81 NO. 2work associates whom you like a lot." Eachsubject made three choices, as describedbelow. Response frequencies are reportedin untfriends(10%)(90%)C.D.Thisweekendfriendsaunt26 weeksfrom nowauntfriends(48%)(52%)E.F.26 weeksfrom nowfriendsaunt27 weeksfrom nowauntfriends(17%)(83%)Observe that it is not possible to interpretthe three modal choice patterns as the result of a single time preference function,denominated in absolute time. Responses tothe first and third questions suggest a negative rate of time preference over a 1-weekinterval, irrespective of whether the week isclose or far. By interpolation, one may conclude that the aunt would be scheduled inthe earlier of any consecutive 2 weekends.Yet, for many subjects, the negative timepreference over adjacent weekends does notcompound into a negative time preferencefor the complete 6-month interval, as shownby responses to the second question.Applied uniformly to all choices, negativetime preference would require harsh reductions in present consumption in favor of thefuture. The fact that one does not observesuch sacrifices, even given the additionalinducements of a positive interest rate, isnormally taken as evidence for positive timepreference (M. Olson and M. J. Bailey,1981).We propose a different interpretation:Negative time preference is applied selectively, to those events that are seen as partof a meaningful sequence, having a welldefined starting and ending point. As theprevious example shows, a pair of adjacentweekends defines a minimal but coherent349sequence, while a pair of weekends separated by 6 months does not.The saliency of particular intervals is notan objective matter, but depends rather onperceptual framing. By deliberately manipulating the subjective frame, it is possible toinduce normatively peculiar patterns ofchoices, as the following example shows.Subjects were asked to indicate their preferred times to eat two free dinners at therestaurant of their choice. One group wasgiven no time constraint concerning whenthey could eat the dinners. A second, constrained group was told that the dinnersmust be consumed within the next 2 years.We anticipated that constrained subjectswould prefer to delay the dinners more thanunconstrained subjects, because the introduction of an explicit planning interval reminds the subjects that, by selecting a particular pair of dates, they are also choosingnot to consume the meals on all of the otherdates. As predicted, the mean preferreddelay for the first dinner was 3.3 weeks forthe unconstrained group and 7.7 for theconstrained. Mean preferred delay times forthe second dinner were 13.1 and 31.1 weeks.These results are inconsistent with theaxiom of revealed preference. The imposition of a time constraint on an initiallyunconstrained population should only affectthe responses of that fraction of the population whose preferred delays are longer thanpermitted by the constraint. The populationaverages should, therefore, be longer in theunconstrained condition.III. Nonadditive PreferencesAside from a preference for improvement, choices between sequences also reveal a sensitivity to certain global or"ggestalt"properties, having to do with howevenly the good and bad outcomes are arranged over the total time interval. Consider the following problem presented to 37Yale University undergraduates (fromLoewenstein, 1987). Subjects were first givena choice between A and B, then between Cand D. Percentages choosing each of theoptions are presented in the right-hand column.
AEA PAPERS AND PROCEEDINGS350AlternativeWeekend 1Weekend 2ABFancy FrenchEat at homeEat at homeFancy FrenchEat at homeEat at home1684CDFancy FrenchEat at homeEat at homeFancy FrenchFancy LobsterFancy Lobster5446Weekend 3ChoicesIn the first problem, the majority of subjectspreferred to postpone the fancy dinner toweekend 2, in keeping with the improvement principle. However, in the secondproblem, the insertion of the common lobster dinner in weekend 3 caused preferenceto shift slightly in favor of having the Frenchdinner right away. This is a violation ofadditive separability, which implies that anindividual preferring sequence A over Bshould continue to prefer A over B if anyelements shared by the two sequences arealtered in the same way.Intertemporal additivity has never beenviewed as normatively compelling, sincethere are many situations in which it isreasonable for consumption at one point intime to influence the marginal utility ofconsumption at another. Models of the"habit formation" type (Duesenberry; R. A.Pollack, 1970; G. M. Constantinedes, 1990)assume that instantaneous utility dependsnegatively on past consumption. Othermodels incorporate the rate of consumptionchange into the utility function (R. Frank,1989), or a preference/aversion for utilityvariation between adjacent periods, as in I.Gilboa's elegant formulation (1989).Can the separability violation describedabove be reasonably attributed to some sortof loss aversion, following an adaptation toa reference point? An aversion to utilityreductions from one period to the nextwould seem to work in favor of alternativeD over C, while leaving preferences between A and B unchanged (to a first approximation). Hence it would predict theopposite violation pattern to the one actually observed.Furthermore, we have observed the samepreference pattern when common "Eat athome" weekends are inserted between theoriginal weekends 1 and 2, and 2 and 3, inall four alternatives (thereby creating 5weekend sequences). Because the neutralMAY1991filler weekends should reduce adaptation,and altogether eliminate differential interactions between adjacent periods, theyshould attenuate the separability violations.The results with the 5-weekend version ofthe problem were essentially equivalent,however: Only 11 percent of respondentsopted for option A over B, while 49 percentpreferred C over D.The problem here is that habit formation/loss aversion models do not well capture the global properties that people findattractive in sequences. The relative advantage of sequences B and C is due to the factthat they both "cover" the 3-week intervalbetter than their respective alternatives. Inother words, they interleave the good andindifferent events in a more nearly uniformmanner.We have developed a theoretical modelthat measures both the degree of improvement and spreading of any sequence interms of cumulative utility sequences (seeour 1990 paper). The model defines improvement over time as the sum of deviations of the cumulative sequence that wouldbe obtained by spreading total utility evenlyover time from the cumulative utility streamof the sequence being evaluated. Evennessof spread is represented by the sum of theabsolute value of these deviations.The derivation of our notions of improvement and uniformity is depicted in Table 1,using the last illustrative example on theassumption that the "Eat at home" eventhas utility zero, and the "Fancy French"and "Fancy Lobster" events have utility one.Note that a simple preference for improvement would lead to a preference for B overA and D over C, while discounting alonewould produce the opposite pattern. Thedesire for spreading outcomes over time,however, designated in the row markedSpread (lower numbers signify more evenspreading of outcomes), can explain thepreference for B over A and for C over D.In several longer surveys (see our 1990paper), we have systematically mapped outpreferences over multiple period sequences.The judgments of the average person couldbe briefly described as follows: There is astrong liking for improving sequences, mod-
VOL. 81 NO. 2INTERTEMPORALCHOICETABLE 1-DERIVATIONOF IMPROVEMENTAND SPREAD 0.66,1.33,2-.33, .33,0.66,1.33,2.66, .33,0-11001.66.661eratedby a penaltyfor deviationfromglobaluniformness,and a small premiumfor sequences that startwell.IV. ConclusionPreviouspsychologicalworkon time preference has focused almost entirely on thetradeoff that arises when two outcomes ofdifferentdates and differentvalues are compared. The tacit premise was that suchjudgmentswill reveal an individual's"raw"time preference,from which one can thensynthesize preferences over more complexobjects-retirement plants, intertemporalincome profiles,and such. This view we feelis fundamentallyincorrect:As soon as anintertemporaltradeoff is embedded in thecontext of two alternativesequencesof outcomes, the psychological perspective, or"frame"shifts,and individualsbecomemorefarsighted,usuallywishing to postpone thebetter outcome to the end. The same person who prefersa good dinnersoonerratherthan later, if given a choice between twoexplicitlyformulatedsequences,one consisting of a good dinner followed by an indifferent one, the other of the indifferentdinner followed by the good one, may wellprefer the latter alternative.Sequences ofoutcomes that decline in value are greatlydisliked, indicatinga negative rate of timepreference.A byproduct of the sequence frame isthat subjectswho are given a time interval,within which to schedule some enjoyableactivity, may schedule it later on average351than people who are given no time frameatall. Apparently,as soon as the relevantinterval is specified, a person becomes concerned with shiftingthe good events out tothe end. This result has implicationsfor lifecycle choices; for example, it suggests thepossibility that some individuals wouldchoose an earlier retirementin the absenceof a mandatedretirementpoint.The sensitivityof time preference to thesequence "frame" casts new light on theoften-repeatedchargethat certaingroupsofpeople (consumers,managers,membersof aparticularnation or culture)have an excessively steep rate of time preference.Such aclaim is a psychologicallyimprecise definition of the problem,at best. The differencesthat do prevail should instead perhaps betraced to different styles of mental bookkeeping, which will alone produce differentdegrees of impatienceeven with a commonunderlyingrate of time preference.Any operation, custom, or habit that causes thestream of purposeful activity to fragmentinto a series of isolatedchoices, each involving a simple intertemporal tradeoff, andeach unrelatedto a largerplan, encouragesimpatientchoices. Whereas the integralsequence frame,by fusingevents into a coherent sequence, promotesconcernfor the future, thereby creating an appearance ofnegativetime preference.REFERENCESConstantinides,G. M., "Habit Formation: AResolution of the Equity Premium Puzzle," Journal of Political Economy, June1990, 98, 519-43.Duesenberry,J., Income, Saving, and theTheory of Consumer Behavior, Cambridge: Harvard University Press, 1949.Frank, R., "Frames of Reference and theQuality of Life," American Economic Review Proceedings, May 19
American Economic Association Negative Time Preference Author(s): George Loewenstein and Drazen Prelec Source: The American Economic Review, Vol. 81, No. 2, Papers and Proceedings of the Hundred and Third Annual Meeting of the American Economic Association (May, 1991), pp. 347-352 Published by: American Economic Association
CMU-CS QTR-120 CMU-CS-13-119 School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Carnegie Mellon University, Qatar campus. The author can be reached at afahim@qatar.cmu.edu or amtibaa@cmu.edu or kharras@cs.cmu.edu. Abstract It is common practice for mobile devices to o oad computationally heavy tasks o to a cloud, which has
American Economic Association U.S. Economic Growth since 1870: One Big Wave? Author(s): Robert J. Gordon Reviewed work(s): Source: The American Economic Review, Vol. 89, No. 2, Papers and Proceedings of the One Hundred Eleventh Annual Meeting of the American Economic Association (May, 1999), pp. 123-128 Published by: American Economic Association
American Economic Association The Invisible Hand and the Grabbing Hand Author(s): Timothy Frye and Andrei Shleifer Source: The American Economic Review, Vol. 87, No. 2, Papers and Proceedings of the Hundred and Fourth Annual Meeting of the American Economic Association (May, 1997), pp. 354-358 Published by: American Economic Association
American Economic Association The Peculiar Economics of Bureaucracy Author(s): William A. Niskanen Source: The American Economic Review, Vol. 58, No. 2, Papers and Proceedings of the Eightieth Annual Meeting of the American Economic Association (May, 1968), pp. 293-305 Published by: American Economic Association
American Economic Association Why Doesn't Capital Flow from Rich to Poor Countries? Author(s): Robert E. Lucas, Jr. Source: The American Economic Review, Vol. 80, No. 2, Papers and Proceedings of the Hundred and Second Annual Meeting of the American Economic Association (May, 1990), pp. 92-96 Published by: American Economic Association
Your path to CMU-Q 2 Why study at CMU-Q? 3 Academics at a glance 4 Biological Sciences 5 Business Administration 6 Computational Biology 7 Computer Science 8 Information Systems 9 Where do our graduates work? 10 Clubs and activities 11 What can you do at CMU-Q? 12 Five degree programs. A world of opportunities. Your university education starts .
Recovering Temporally Rewiring Networks: A Model-based Approach Fan Guo fanguo@cs.cmu.edu Steve Hanneke shanneke@cs.cmu.edu Wenjie Fu wenjief@cs.cmu.edu Eric P. Xing epxing@cs.cmu.edu School of Computer Science, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213 USA
Reading: Literature Second Grade Key Ideas and Details Range of Reading and Level of Text Complexity Craft and Structure Integration of Knowledge and Ideas. Indicator Date Taught Date Retaught Date Reviewed Date Assessed Date Re-Assessed CCSS.ELA-LITERACY.RI.2.1 I can ask and answer who, what, where, when, why and how questions to show that I understand nonfiction. CCSS.ELA-LITERACY.RI.2.2 I .
