Choice Over Time -- Chapter 1

8/3/2019 Choice Over Time -- Chapter 1

1/16

Preface

Daily life requires frequent trade-offs between costs and benefits thatoccur at different points in time. To sleep late or rise early, munchsnacks or eat a healthy lunch, buy the snazzy sports car or the reliablesedan, get a job or go to college, risk pregnancy or use a contracep-tive-these and many other decisions require us to weigh conse-quences that are distributed over time. How they are resolved is thecentral focus of this book. In it, we explore the psychological pro-cesses underlying intertemporal choice, propose formal models ofthese processes, and apply the models to various social problems.

Choice O ver Time is the product of a group that met annually forfour years to discuss intertemporal choice. Brought together underthe auspices of the Russell Sage Foundation's program on behavioraleconomics, scholars from diverse disciplines presented their own per-spectives, listened skeptically as others did the same, discussed em-pirical findings, and debated interpretations of those findings. Be-cause these discussions revolved around the axis of behavioraleconomics, much of the book that emerged from them is devoted to acritique of the current economic approach to modeling intertemporalchoice. The dominant model of intertemporal choice employed byeconomists-the discounted utility model (DU)-provides an im-plicit (and sometimes explicit) benchmark against which many of theauthors contrast their contributions. The reader who is unfamiliarwith DUwill find a brief description of the model in Chapter 1, and anin-depth analysis of its assumptions and implications in Chapter 5.

ix


2/16

x Preface

Despite their extensive use of DU, economists do not have a strongintellectual commitment to the model. This ambivalence stems fromtwo sources. First, few economists versed in the details of intertem-poral choice are comfortable with DU's behavioral assumptions. In-deed, Samuelson, who first proposed the model in 1938, commentedthat several features of the model were "completely arbitrary" (seeChapter 1). Koopmans (1960) who proposed a set of axioms consis-tent with the model, enumerated serious objections to all of the im-portant ones. Thus, DU is flawed both as a descriptive and normativemodel.

Second, and perhaps more importantly, the conventional dis-counting model does a very poor job of predicting or explaining be-havior. When scrutinized through the lens of DU, intertemporalchoices observed in practice appear hopelessly inconsistent. Exam-ples of behavior at the extremes of high, low, and even negative timediscounting are easy to identify, not only across different persons(which the model can accommodate), but also in the behavior ofa single person. For example, when purchasing energy-consumingappliances or insulating their houses, people display extreme tempo-ral myopia, underinvesting in efficiency to a degree that implies dis-count rates in the hundreds of percents. At the same time, however,many of these people visit the dentist regularly, eat nutritious butbland breakfast foods, and manage to finance comfortable retire-ments. These behaviors are difficult to reconcile with an economicperspective which assumes that each person has a single discountrate representing his or her attitude toward the future.

The same apparent inconsistency can be observed at a societallevel. We construct shoddy highways, refuse to switch to the metricsystem, even type on a keyboard that was designed to slow the typistdown to a rate that 19th-century typewriters could accommodate. Butwe also invest in basic research whose payoff is remote and eschewnuclear waste disposal sites because they may cause problems centu-ries hence. While the mechanisms that drive discounting behavior atthe individual and societal levels are different, both domains displaygreat variance in apparent concern for the future.

The elegance of the economic approach inheres in its simplicity; itsmain drawback is its low congruence with reality. The psychologicalperspectives presented here, on the other hand, offer not one, butseveral models, each of which captures a critical dimension of realhuman behavior. It is our hope that bringing these contributionstogether in one volume will help to spawn a synthesis that incorpo-rates the advantages of each approach.

The book consists of five parts. The first examines historical per-


3/16

Preface xi

spectives on intertemporal choice, focusing specifically on the causesof time discounting and the relationship between social organizationand individual time discounting. The second part offers a variety ofcontemporary perspectives on intertemporal choice by authors whospan a broad spectrum of disciplines and subdisciplines: psychiatry,behavioral psychology, social psychology, and behavioral decisiontheory. The third part examines several multiple-self models thatapply game-theoretic reasoning to the analysis of intra personal con-flict and self-control. The fourth part focuses on internalities in inter-temporal choice-situations in which actions at one point in timeaffect utility or tastes at other points. The fifth part applies insightsfrom the previous parts to a variety of real world problems: savingsbehavior, addiction, and wage-setting practices.

Part One: Historical OverviewIn Part One, two chapters provide historical overviews of differentaspects of intertemporal choice. Chapter 1, "The Fall and Rise ofPsychological Explanations in the Economics of IntertemporalChoice" by Loewenstein, examines the evolution of economists' an-swers to the fundamental question of why people discount thefuture.

One can draw a parallel between 17th century speculation overman's natural state-his nature in the "state of nature" -and theruminations of 19th-century economists about man's baseline or"natural" time discounting. One school of thought, exemplified bythe contributions of W.S. [evons, insisted that the natural humantendency is to ignore the future. The other, first propounded byN.W. Senior, assumed that equal treatment of present and futurewas natural. These contrary assumptions led to two very differentexplanations of time discounting.To explain why people took account of the future at all, the "myo-pia" school pointed to the existence of emotions experienced in thepresent but associated with the anticipation of future pleasures andpains. Such emotions were thought to give the future an immediateimpact, overcoming what would otherwise be a total preoccupationwith the present. In this view, discounting occurred due to the rela-tive weakness of such emotions and from an inability to imagine thefuture in vivid and accurate detail. Those 19th-century economistswho assumed that human beings weighted present and futureequally, on the other hand, argued that the temptation of immediateconsumption distorted people's natural inclinations toward such abalanced treatment of the present and future.


4/16

xii Preface

Far from being just of antiquarian interest, the question of whethertime discounting derives from the temptations of the moment, orfrom the inadequate motivational force of the future, continues toguide current investigations. In considering emotions associated withanticipation, several contributions in Part Four of this book revivethe "myopic" perspective. Contributions in Part Two, in contrast,continue the effort to show how the lure of immediate consumptioncan produce extremes of shortsightedness. Part Three, on self-control, considers the tactics that people use to resist such tempta-tions.

In "Intertemporal Choice and Political Thought," Elster positionsintertemporal choice in the social arena. Anticipating the discussionof self-control tactics contained in the third part of the book, Elstershows that provisions which make it difficult for democracies toamend their constitutions act as a political self-control device. In thesame way that individuals learn rules such as "count to ten beforeresponding," such measures protect legislatures from overreactingto momentary passions. Elster also provides an introduction to thecontributions of Tocqueville on the question of how civic cultureand individual time preference influence one another. He identifiestwo mechanisms through which political institutions influence indi-vidual time discounting. The "spill-over" effect refers to situations inwhich farsighted public institutions exert a demonstrational or learn-ing effect, inducing a comparable perspective at the individual level.The "displacement effect," on the other hand, refers to situations inwhich the government fulfills needs relating to the future, leavingindividuals free to pursue more short-term goals. De Tocqueville'sview was not, however, a simple one in which government influ-enced individual time preferences. Instead, as Elster documents, heenvisioned a complex interaction between the civic, religious, andfamilial spheres of activity.

Part Two: General PerspectivesIn Part Two, four contributions offer an overview of contemporarypsychological research on intertemporal choice. The first two of thesefocus specifically on the shape of the discount function, which speci-fies the relationship between the time at which an object will bereceived, and its present subjective value. Both "Hyperbolic Dis-counting" by Ainslie and Haslam, and "Irrationality, Impulsiveness,and Selfishness as Discount Reversal Effects," by Rachlin and Raineriexamine the evidence for and implications of the observation that


5/16

Preface xiii

time discounting tends to follow a hyperbolic rather than exponentialform. However, as discussed below, their perspectives differ on sev-eral critical points.The discounted utility model assumes exponential discounting ata constant rate. This implies that a given time delay leads to thesame degree of time discounting regardless of when it occurs. Underexponential discounting, a one-day delay has the same significanceif it means deferring an outcome from today until tomorrow, or fromone year from today to a year and a day from today. Thus, for exam-ple, a person who is indifferent to the choice between one appletoday and two apples tomorrow should be indifferent to the choicebetween one apple in one year and two apples in a year and a day.'

As first noted by the economist Strotz (1955), exponential discountfunctions have the desirable property of implying that behavior willbe consistent over time. For example, an exponential discounter whoprefers an expensive French dinner in twelve months to a nice Chi-nese dinner in eleven, should also prefer the French dinner after sixmonths have passed (and the Chinese dinner is five months awayand the French dinner six), and should continue to prefer the Frenchdinner even when the actual date for the Chinese dinner arrives.Nonexponential discounters, in contrast, are subject to a phenome-non Strotz called "time inconsistency"; their preferences change sys-tematically with the passage of time.

The hyperbolic discount curves discussed in chapters 3 through 5produce a particular form of time inconsistency. When faced with achoice between an inferior early option and a superior later option,hyperbolic discounters will tend to prefer the later, superior, optionwhen both are remote, but switch to a preference for the earlier,inferior, option as both approach in time. A hyperbolic discounterwho preferred the French restaurant over the Chinese when the for-mer was delayed by six months and the latter by five, would be likelyto opt for the Chinese dinner when it became immediately available;under hyperbolic discounting, the Chinese dinner becomes dispro-portionately tempting when it is imminent but the French dinner isstill a long month away.

Time inconsistent behavior is different from a high rate of timediscounting per se. High time discounting signifies that outcomesexperienced earlier are much more highly valued than outcomes ex-perienced later. Time inconsistency, on the other hand, refers tochanges in time discounting as a function of delay. A high rate of

1Example from Thaler (1981).


6/16

xiv Preface

time discounting is possible without inconsistency, and vice versa.Indeed, as Elster and Loewenstein show in Chapter 9, examples oftime inconsistency associated with negative time discounting are notuncommon-as illustrated by the behavior of misers who plan tosplurge in the future, but inevitably fail to carry out the plan whenthe appointed time arrives.Given the frequency with which hyperbolic discounting arises, itmay be worthwhile to provide some historical background on theevolution of the idea. In the early 19th century, before the discountedutility model established exponential discounting as the hallmark ofrationality, the notion of exponential discounting had no special sta-tus, and its significance for the consistency of behavior was not ap-preciated. Nonetheless, many 19th-century economists displayed anawareness of the nonconstancy of discounting. [evens, for example,noted that:

An event which is to happen a year hence affects us on the averageabout as much one day as another; but an event of importance, whichis to take place three days hence will probably affect us on each of theintervening days more acutely than the last. (1871:34)

Bohrn-Bawerk commented that:It is not true that the individual who undervalues a future utility thatwill mature in one year by 5% will necessarily undervalue one due intwo years by 10% and one that is only three months off by exactly1.25%.... We [may) look for a very marked difference in valuationas between momentary and nonmomentary satisfactions, but on theother hand very little or no difference at all as between moderatelydistant and remotely distant gratifications. (1889:272)

In fact, much earlier, Spinoza remarked on this phenomenon:We can only distinctly imagine distance of time, like that of space, upto a certain limit, that is just as those things which are beyond twohundred paces from us, or, whose distance from the place where weare exceeds that which we can distinctly imagine, we are wont to imag-ine equally distant from us and as if they were in the same plane, soalso those objects whose time of existing we imagine to be distant fromthe present by a longer interval than that which we are accustomed toimagine, we imagine all to be equally distant from the present, andrefer them all as it were to one moment of time. (1677:145)The first empirical studies to demonstrate nonexponential dis-counting were conducted with animals, beginning with a study by


7/16

Preface xv

Chung and Herrnstein (1964). In Chung and Herrnstein's procedure,pigeons pecked two buttons, each of which occasionally producedfood according to a prearranged schedule. A "successful" peck didnot produce food immediately, but only after a brief delay, lastingon the order of several seconds. Chung and Herrnstein found thatpigeons divided their choices between the two keys in inverse pro-portion to the ratio of delays imposed by the keys. In other words,if the delay on one key was twice as long as the delay on the other,the pigeon would peck twice as much on the latter alternative, allother things being equal. On the basis of this result, they concludedthat the subjective value of a delayed reward is inversely related tothe duration of the delay. The "biological" discount function, in otherwords, is the simple reciprocal function: f(t) = lit.

More recent animal studies, notably Mazur (1987) have slightlyqualified this finding. Mazur presented pigeons with two delayedrewards of different magnitudes, and explored the combination ofdelay durations that made the animals indifferent between the twoalternatives. Exponential discounting predicts that these matchingdelays should differ by a constant amount, while reciprocal dis-counting predicts that the delays should be in constant proportion.Mazur found that the relationship between the two delays was linearand located in between the two extremes of constant proportionalityand constant difference. Such a relationship is consistent with thehyperbolic discount function, f(t) = 1/(1 + kt), which had alreadybeen suggested by Herrnstein (1981). Indeed the most general dis-count function consistent with the observed linear indifference rela-tions between delays for small and large rewards is the power hyper-bola, f(t) = (1 + at) - b/a (see Chapter 5).

Richard Thaler (1981) was the first to document the effect withdata from human subjects. His study also produced the first evidenceof other choice patterns incompatible with DU, including higher dis-counting of gains than losses and higher discounting of small out-comes than large. Subsequently, many experiments (e.g., Benzion,Rapoport, and Yagil, 1989), including some using real money out-comes (e.g., Holcomb and Nelson, 1989), have replicated the findingof nonexponential time discounting in humans. (For a review of thesefindings see Chapter 5.)

Ainslie and Haslam, in the first chapter of this section, use hyper-bolic discounting to explain a broad range of human behaviors-fromthe mundane (e.g., semi-voluntary scratching an itch) to the momen-tous (e.g., addictions). In their view, the hyperbolic discounting andthe reversals of preference which it produces (e.g., from the Frenchto the Chinese dinner) are a powerful motive force underlying human


8/16

xvi Preface

behavior. They believe that the lure of immediate self-gratification atevery moment establishes competing interests within the individualthat struggle between themselves for dominance (see also their Chap-ter 8). Most controversially, they argue that pain can be explained interms of hyperbolic discounting-that the experience of pain resultsfrom a surrender to a momentary temptation.

Rachlin and Raineri, while also endorsing hyperbolic discounting,disagree with Ainslie and Haslam on several fundamental issues.First, hyperbolic discounting assumes far less prominence in theiroverall view of human behavior. They see time discounting as onlyone of a broad class of subjective phenomena that follow a hyperbolicpattern. These other phenomena include the discounting of uncertainrewards according to the probability of receiving them, which tendsto overweight certain outcomes relative to uncertain outcomes in thesame way that hyperbolic discounting overweights immediate out-comes. Forgetting also tends to assume a hyperbolic form-a regular-ity known as Jost's law in which recently assimilated information ismore likely to be forgotten than information assimilated in the distantpast. Similarly, the loudness of a sound, intensity of a light source,size of an object, and other stimulus dimensions vary hyperbolicallyas a function of the distance between the observer and the source. Allof these phenomena can produce reversals similar to the preferentialreversals observed in intertemporal choice. For example, when faraway from a city, one can generally gauge accurately the relativeheights of buildings. But when in its midst, buildings that are neartend to look taller than those which are far.

Rachlin and Raineri view hyperbolic discounting not only as onlyone of many related phenomena, but also see the reversals of prefer-ences it implies as relatively rare. As a supportive analogy they pointto the phenomenon of size constancy in perception. Even though theimage formed on the retina of the eye grows hyperbolically as anobject approaches, given appropriate visual cues, we are usually ableto compensate and estimate the object's size without bias. In thesame way, they argue, people can naturally overcome the effect ofhyperbolic discounting and behave in a time-consistent manner.

Taking the analogy one step further, they contend, just as sizeconstancy can be made to appear or disappear through appropriateprovision of cues, seemingly minor situational variations can lead thedecision-maker either to extreme myopia or farsightedness. Mischel,Shoda, and Rodriguez, for example, show that simply having some-one wait in the presence of the immediate and deferred objects cansignificantly affect willingness to delay (see Chapter 6).

Ainslie and Haslam disagree on all of these points. They view


9/16

Preface xvii

time discounting as quite distinct from other subjective phenomenaassuming a hyperbolic shape. And they disagree that human beingshave the same effortless ability to overcome time inconsistency thatthey have in compensating for source distance when judging size,loudness, or intensity. Indeed, for these authors the lack of such anability is a key feature, which distinguishes time discounting fromthese other effects.

The difference between Ainslie and Haslam, on the one hand,and Rachlin and Raineri, on the other, recalls the dispute among19th-century economists about whether complete myopia or equaltreatment of present and future was the norm, deviations from whichneeded to be explained. Ainslie and Haslam view time inconsistencyas a pervasive problem only to be overcome through intricate andenergetic efforts at self-control. For Rachlin and Raineri, however, itis an aberrant behavior that people display only when normal cuesare absent. The specifics have changed, but debate continues to re-volve around the question of the natural form of time discounting.

Chapter 5, "Anomalies in Intertemporal Choice: Evidence and anInterpretation," by Loewenstein and Prelec, criticizes the discountedutility model. They enumerate a series of intertemporal choice anom-alies-common patterns of behavior that are inconsistent with DU-that are analogous to the well-known expected utility anomalies indecision-making under uncertainty. The intertemporal choice anoma-lies include nonconstant discounting, as discussed in the prior twochapters, and a variety of other phenomena such as the fact thatlosses tend to be devalued less rapidly than gains, that large lossesand gains are devalued less rapidly than small ones, and that peopledisplay asymmetric preferences for speed-up and delay of consump-tion. To account for these anomalies, Loewenstein and Prelec developa mathematical model of intertemporal choice which, like the contri-butions in the prior two chapters, incorporates hyperbolic dis-counting, but which also makes a variety of assumptions about theform of the utility function.

The last chapter in this section, "Delay of Gratification in Chil-dren," by Mischel, Shoda, and Rodriguez, reviews three decadesof research on delay of gratification by Mischel and his colleagues.Mischel's research has gone through several transitions during thisperiod. His earliest work focused on delay-choice-that is on chil-dren's initial choices between rewards to be obtained at differentpoints in time. In the mid-1960s, however, Mischel became interestedin a different aspect of intertemporal choice, namely the implementa-tion of the initial choice, which he termed "delay of gratification."The initial decision to eschew an immediate reward in favor of a


10/16

xviii Preface

greater reward in the future exposes the individual to the temptationof bailing out and opting for the inferior immediate reward duringthe waiting period. Mischel and his co-authors examine a variety ofsituational and cognitive factors that influence the likelihood of sucha bail-out occurring.

The finding that subtle changes in situational variables can pro-duce striking changes in modal behavior led Mischel to reevaluatethe significance of situation as a determinant of behavior. In his 1968book Perso na lity a nd A ssessm en t, he argued that situation shapes be-havior more than personality does-or at least more than was com-monly believed. However, Mischel's most recent work takes aslightly different perspective. Drawing on recent research with trou-bled children at a summer camp, Mischel now contends that individ-uals differ in terms of measurable and fairly stable competencies thatinteract with situational variables to produce behavior. In a follow-upof the subjects from his early delay of gratification experiments, hehas discovered strong correlations between delay of gratification inpreschoolers and their adult characteristics and achievements. How-ever, these correlations occur only within specific experimental condi-tions-when the rewards children waited for were exposed. Thisevidence, of temporal consistency in certain situations but not inothers, support his current interactionist perspective.

Part Three: Self-ControlThe hyperbolic discounting models presented in this book imply thatbehavior will be inconsistent over time in the absence of efforts atself-control. Similarly, the difficulties of executing a decision to delay(as documented by Mischel), particularly when the delayed rewardis exposed to view, induce a form of time inconsistency-subjectswho initially chose to wait for a delayed reward bailout during thewaiting period.

In daily life, however, consistency appears to be the rule andinconsistency the exception. How is such consistency achieved?Rachlin and Raineri suggest that consistency is learned through day-to-day experience much in the same way that we learn to compensatefor distance when judging the size of a remote object. Mischel andhis co-authors point to a wide range of cognitive tactics, such as dis-traction and cognitive restructuring, that people employ to avoid time-inconsistent bail-outs. Elster and Loewenstein in Chapter 9 show thatpleasure and pain derived from memory and anticipation can miti-gate myopia, in some cases actually leading to negative time dis-


11/16

Preface xix

counting-e.g., deferring desired outcomes or getting unpleasantoutcomes over with quickly. Herrnstein and Prelec, in Chapter 13,argue that social norms (e.g., eat only three meals a day; don't drinkin the morning) provide additional defenses against the lure of temp-tation. Shefrin and Thaler, in Chapter 12, examine a wide range ofindividual and social mechanisms that promote self-control.

The most influential explanation for time-consistency in the pres-ence of myopic discount curves, however, relies on the notion ofmultiple selves, and conflicts between those selves. Such multiple-self models give expression to the introspective experience of internalconflict ("the devil made me do it") that philosophers and writers ofall historical periods have commented upon. Whereas the theoreticalmodels reviewed in the second section have their intellectual rootsin neoclassical economics, the multiple choice models presented inthis section have a closer affinity to game theory.

Schelling, whose chapter "Self-Command: A New Discipline"opens this section, applies insights from his well-known work oninterpersonal bargaining to bargaining within the self. In his view,self-control can be seen as an intimate struggle between two selves,one farsighted, the other shortsighted. Schelling depicts the strate-gies and tactics employed by the farsighted self to constrain the be-havior of the myopic self. These are very similar to those that mightbe used by one individual in bargaining with another. In the sameway that a union leader might commit himself to attaining a particularwage demand by staking his reputation on achieving that wage, indi-viduals can stake their reputations or, through side-bets with otherpeople, stake material possessions on their own avoidance of temp-tation.

Ainslie and Haslam's Chapter 8 "Self-Control," building on aframework first articulated in a highly influential earlier paper (Ain-slie, 1975),models self-regulation as a repeated two-person prisoners'dilemma, with one person representing the long-term, and the otherthe short-term, perspective. The authors point out that conflicts arerarely one-shot; internal struggles such as those involved in dietingor waking up in the morning tend to repeat themselves over time.And, just as early interactions between two individuals can influencelater ones by creating precedents and establishing reputations, thesame is true in the intrapersonal context. Any lapse by the self canundermine the individual's belief that future temptations can beavoided. Both the reluctance to destroy reputational capital, and thetendency to perceive a breach as only the first of a long string oflapses (bunching), promote the long-term perspective. Ainslie andHaslam's perspective differs from Schelling's mainly in that the latter


12/16

xx Preface

focuses predominantly on tactics that involve third parties (e.g., giv-ing one's car keys to the party's host), while the former are morecentrally interested in "intrapsychic" tactics in which the self playsthe policing role.

Stimulated by these authors' work, many other multiple-self mod-els have been proposed to deal with the subjective experience ofinternal conflict. In the "planner-doer" model, presented by Shefrinand Thaler in Chapter 12, intrapersonal conflict results from the inter-action of a single farsighted superego-like planner with a series ofid-like myopic doers. The planner represents the notion that at alltimes an individual has some sense of his or her long-term best inter-est; the doers reflect the temptations of the immediate moment thatinevitably conflict with the individual's long-term interests. The plan-ner-doer model, however, is tied more closely to principal-agent the-ory in economics than to game theory, as are the two contributionsjust discussed.

Part Four: InternalitiesChoice at anyone time has consequences for other points in time.Just as empathy causes people to share another person's pleasure orpain, anticipation and memory extend the hedonic impact of an epi-sode across time. The authors in this section examine a variety ofbehavioral implications that are rooted in these simple observations.

In "Utility from Memory and Anticipation," Elster and Loewen-stein enumerate the various ways that emotions arising from mem-ory and anticipation influence utility, and discuss the implication ofsuch effects for behavior. They distinguish between two types ofemotional influences that can operate either through memory or an-ticipation: that is, consumption effects, whereby utility experiencedin the past or future positively influences current utility, and contrasteffects in which the relationship is negative. For example, a deliciousdinner at a four-star restaurant last weekend could enhance currentwell-being by providing pleasurable memories (a consumption ef-fect), or detract by providing an unfavorable contrast to tonight'sfast-food repast. An individual who includes such effects in his orher calculations of utility, even if a discounted utility maximizer, willdisplay anomalous patterns of preference, such as preferring wageprofiles that improve over time over declining profiles of comparableundiscounted value (see Chapter 15).

In "Melioration," Herrnstein and Prelec examine suboptimal pat-


13/16

Preface xxi

terns of choice that occur when tastes result from a long series ofchoices, each of which has a small individual impact. In these situa-tions, the choices of an optimizing decision-maker would take intoaccount the incremental effect of current choice on satisfaction fromsubsequent choices. But the authors present experimental resultsshowing that, in a conceptually simple situation, people ignore indi-rect effects and, as a result, fail to maximize their well-being. Theyargue that the failure to take account of taste internalities underliesa large number of suboptimalities commonly observed in choice.Among these are underinvestment in skill-intensive consumption ac-tivities (such as music appreciation), and overconsumption of activi-ties, such as eating, which have negative indirect effects. Their laterchapter, in the following section, develops these insights into a theo-retical model of addiction.

Although the arguments of Elster and Loewenstein, on the onehand, and Herrnstein and Prelec, on the other, might appear contra-dictory, the conflict is largely illusory. Herrnstein and Prelec focuson the impact of the present on the future, arguing that this impactis routinely ignored or severely underweighted in decision-making.Elster and Loewenstein focus primarily on the impact of the futureon the present, arguing that this influence is quite profound.

In Chapter 11, "The Role of Moral Sentiments in the Theory ofIntertemporal Choice," Robert Frank examines the social functionthat emotions play in overcoming social dilemmas. In his book Pas-sions W ithin R eason, Frank argues that emotions such as anger andgratitude benefit society by causing people to reciprocate defectionand cooperation in situations in which narrow self-interest wouldsuggest the opposite course. Anticipation of retaliation and reciprocalcooperation alters the prisoners' dilemma in a manner that promotescooperation. According to Frank, emotions can affect cooperationwhile other mechanisms cannot because emotions are experienced asimmediately as the temptation to defect. In contrast to, for example,the contemplation of the potential future benefits of cooperation,emotions circumvent the steep discounting of even slightly delayedrewards, as implied by hyperbolic discounting.

Part Five: Applications and ExtensionsThe final section offers a small sampling of how theoretical insightsinto intertemporal choice can illuminate real world phenomena. She-frin and Thaler's chapter "Mental Accounting, Saving, and Self-


14/16

xxii Preface

Control" applies a self-control perspective to savings behavior. Inmuch the same way that Chapter 5 challenged the validity of thediscounted utility model, this chapter criticizes the life-cycle modelthat dominates contemporary economic analysis of savings behavior.Using a model based on Thaler's influential work on mental account-ing, the authors show that the central predictions of the life-cyclemodel are refuted, and that actual behavior typically deviates fromthe predictions of the life-cycle model in a manner predicted by theirbehavioral life-cycle model.

The next two chapters take radically different approaches to theproblem of addiction. Herrnstein and Prelec in Chapter 13, "A The-ory of Addiction," present a theory of addiction based on the conceptof the "primrose path" -the idea that addiction is a trap that peopleenter slowly and unknowingly and then find it difficult to extricatethemselves. They dismiss the notion that people can predict andrespond optimally to the changes in taste which accompany addic-tion, a perspective they dismissively label "rational self-medication."Such a model is precisely what Becker, Grossman, and Murphy offerin their chapter "Rational Addiction and the Effect of Price on Con-sumption." In their model, addiction has a long-term negative im-pact on well-being, but overall, the addict is better off than he orshe would have been in the absence of the addiction.

Given the major difference in starting assumptions between thetwo chapters, their theoretical frameworks are remarkably similar.This affinity is most strikingly evident in the resemblance betweenBecker et al.:s Figure 1, and Herrnstein and Prelec's Figure 2, eachof which contains two stable and one unstable equilibrium point lo-cated at approximately equivalent positions.

In the final chapter of the book, "Frames of Reference and theIntertemporal Wage Profile," Robert Frank applies the idea that pastconsumption sets a standard against which current consumption isevaluated (the contrast effect from Chapter 9) to the analysis of wageprofiles. As already noted, the contrast effect can induce a preferencefor sequences of utility that improve over time, even among thosewho discount the future. If people found it easy to implement savingplans, the creation of such an improving living standard would sim-ply be a matter of optimal saving. But, as Thaler and Shefrin note inChapter 12, saving is anything but effortless. As a result, such apattern of improving consumption, while desired in the abstract, maybe difficult to achieve. Frank argues that institutions that satisfy thedesire for improvement have evolved over time-that is, seniority-pegged wage increases which exceed the rates of increase explicableon purely economic grounds.


15/16

Preface xxiii

Concluding CommentsWith ever greater insistence, American social scientists are beingcalled upon to explain and offer remedies for a broad range of societalproblems that seem to reflect suboptimal time discounting. Theseinclude paradigmatic economic variables that must appear on thedebit side of the U.S. balance sheet: a low savings rate, low rates ofcorporate investment, and ballooning private and public debt. Theyalso encompass social ills such as abysmal educational attainments,AIDS, environmental degradation, high rates of infant mortality,crime, and teenage pregnancy. Whether in the role of consumer,manager, voter, student, criminal, or parent, many Americans exhibitbehavior that reflects a heavy weighting of the present relative tothe future.

The economic approach to intertemporal choice provides few use-ful insights into these problems. It assumes that each individual hasa rate of time discounting that he or she applies to all domains ofbehavior, economic and otherwise. But the research reviewed in thisbook suggests a much more complicated picture. While individualdiscount rates, to the extent that they exist, may be high, humanspossess remarkable capabilities to overcome these rates. Social andpolitical organizations, elaborate strategies and tactics of self-control,social norms, and emotions arising from memory and anticipation allmitigate the otherwise pernicious effects of hyperbolic discounting.

Simplistic solutions that ignore psychology will inevitably founder.Successful policy must be based on a more nuanced understandingof the psychological mechanisms underlying time preference. Wehope that C hoice O ver Tim e will provide a significant step in the devel-opment of such an understanding.

References

Ainslie, G. "Specious Reward: A Behavioral Theory of Impulsiveness andImpulse Control," Psycholog ical Bu lle tin , 82 (1975): 463-509.

Benzion, U., A. Rapoport, and J . Vagi!. "Discount Rates Inferred from Deci-sions: An Experimental Study." M ana gement Science, 35 (March, 1989):270-284.

Bohm-Bawerk, E.v. Cap ita l a nd In terest. South Holland, IL: Libertarian Press(1889), 1970.

Jevons, W.S. Theory of P olitica l E conom y. London: Macmillan, 1871.Koopmans, T.e. "Stationary Ordinal Utility and Impatience," Econometrica,28 (1960): 287-309.


16/16

xxiv Preface

Mazur, J.E. "An Adjustment Procedure for Studying Delayed Reinforce-ment," Chap. 2 in M.L. Commons, J.E. Mazur, J.A. Nevins, and H. Rach-lin (eds.). Q uantitative Analysis of Behavior: The E ffect of D elay and of Interven-ing Events on Reinforcement Value. Hillsdale, NJ: Erlbaum, 1987.Mischel, W. P erson ality a nd A ssessm en t. New York: Wiley, 1968.

Rae, J. The Sociolog ica l Theory of Capital (reprint of original 1834 edition).London: Macmillan, 1905.Samuelson, P. "A Note on Measurement of Utility," R eview of Economic Stud-ies 4 (1937): pp. 155-161.Senior, N.W. An Outline of the Science of Political Economy. London: Clowesand Sons, 1836.

Spinoza, Ethics. London: J.M. Dent & Sons Ltd., 1989.Thaler, R., "Some Empirical Evidence on Dynamic Inconsistency." EconomicsLetters, 8 (1981): 201-207.

Choice Over Time -- Chapter 1

Documents