Routing

Economists have long argued that govern-ment intervention makes most sense in situa-tions that involve externalities. Externalities arecosts or benefits that spill over onto third parties.When individuals bear the full costs and receivethe full benefits of their own actions, the justifi-cation for government involvement is muchweaker. But a new generation of economists con-tends that paternalistic intervention can be justi-fied to correct problems of self-control. If peopledon’t fully consider the costs their choicesimpose on their own future selves, the theorygoes, those choices impose within-person exter-nalities dubbed “internalities.” The internalitiesapproach provides a novel argument in favor ofpaternalistic government policies such as sintaxes (including fat taxes), marketing restric-

tions, mandatory savings plans, and so on.The theory of internalities is explicitly modeled

on the theory of externalities. However, the formerstands about where the latter stood in 1960, justprior to Ronald Coase’s seminal work on the sub-ject. Exposing internality theory to Coaseaninsights reveals serious flaws. Specifically, internal-ity theory in its current form unjustifiably “takessides” when it chooses to favor some personalinterests over others. Furthermore, it ignores thepossibility of within-person bargaining and otherprivate solutions to self-control problems. Finally,it gives insufficient attention to the possibility ofgovernment failure. Taking those objections intoaccount severely damages the case for paternalisticgovernment intervention to address problems ofself-control.

Against the New PaternalismInternalities and the Economics of Self-Control

by Glen Whitman

_____________________________________________________________________________________________________

Glen Whitman is associate professor of economics at California State University, Northridge.

Executive Summary

No. 563 February 22, 2006

Introduction: How Is a Twinkie like

a Smokestack?How is a Twinkie like a smokestack? Set

aside the vaguely similar shape and thinkabout the harms they create. The harm froma Twinkie falls primarily on the consumer, inthe form of worse health and larger hips.Because the consumer bears most of the costof eating Twinkies, he is well positioned todecide whether their perceived benefits out-weigh their harms. A smokestack, by con-trast, affects more than the factory and theconsumers of its products—specifically, itharms everyone in the surrounding commu-nity who has to breathe the air. The firm’smanagement might not, in its pursuit ofprofit, take into account all of the costs asso-ciated with its polluting activity.

The difference in who is harmed—thedecisionmaker versus someone else—mightseem an important distinction. And indeed,economists have generally held that the casefor government intervention is strongestwhen third parties are involved—that is, whenthere are externalities. Externalities are costs orbenefits of an activity that spill over ontopeople not involved in the activity (typicalexamples include people who breathe pollut-ed air or neighbors disturbed by loud music).Self-regarding activities, on the other hand,can safely be engaged in by free individuals.But a growing literature in economics, as wellas the popular press, argues that Twinkiesand smokestacks are more similar than theyappear. Although your choice to eat Twinkiesor smoke cigarettes or skip exercising todaydoesn’t generally harm anyone else, it doesharm your future self. If we think of a personas consisting of multiple selves—the presentself who wishes to indulge in transient plea-sures versus the future self who wishes to behealthy—then arguably the present self ’schoices can force externalities on the futureself. Those within-person externalities havebeen dubbed “internalities.” And just as wemight impose a pollution tax on a factory tocontrol the externality problem, we might

impose a sin tax on items like cigarettes, alco-hol, and fatty foods to control the internalityproblem.

The concept of internalities, although notyet a part of mainstream economics, is gain-ing attention. It is one among many noveleconomic models recently deployed by a newgeneration of paternalists. Paternalistic argu-ments advocate forcing or manipulatingindividuals to change their behavior for theirown good, as distinct from the good of others.At one time paternalists argued that adults,like children, don’t really know what’s bestfor them. Some preferences, they argued,such as those for unhealthy food or casualsex, are just wrong. But such arguments holdlittle sway in a free society, where most peoplebelieve they should be able to pursue theirown values and preferences even if othersdon’t share them. So the “new” paternalistshave wisely chosen not to question people’spreferences directly; instead, they argue thatinternalities (and other sources of error indecisionmaking) can lead people to makedecisions that are unwise even according to theirown values and preferences.

In short, the old paternalism said, “Weknow what’s best for you, and we’ll make youdo it.” The new paternalism says, “You knowwhat’s best for you, and we’ll make you do it.”1

The internalities approach is clever. Eventhe staunchest skeptics of government inter-vention will usually concede that governmentis needed to prevent people from harmingeach other. By treating the individual as a mul-tiplicity of selves, the new paternalism invitespolicy analysts to import the theory of exter-nalities into the realm of individual choice.

Of course, thinking of a person as havingmultiple selves is a controversial philosophicalposition, and we might be tempted to reject itoutright. We could object that “multipleselves” is, at best, just a metaphor. But here, Iwish to make an immanent critique. I will takethe idea of multiple selves as given—for argu-ment’s sake—and then argue that the analysisof internalities is seriously incomplete.2 In“translating” the concept of externalities, inter-nality theorists have drawn on economic theo-

2

Because the consumer bearsmost of the cost

of eatingTwinkies, he is

well positioned todecide whethertheir perceived

benefits outweightheir harms.

ry that is at least 40 years out of date. In hisfamed 1960 article “The Problem of SocialCost,” Nobel Prize–winning economist RonaldCoase started a revolution in the economicanalysis of externalities.3 His approach showedthat externalities need not always produce inef-ficient outcomes, because both institutions(such as private property rights) and marketexchanges can deal with them. The applicationof Coase’s ideas to internality theory casts seri-ous doubt on the new paternalism.

Internalities without Coase

Prior to the Coasean revolution, externalitytheory followed the analysis of A. C. Pigou.4

The Pigovian argument is straightforward:

• In the presence of externalities (some-times called “spillover” effects), privatedecisionmakers do not face all the costsof their choices.

• As a result, they will do too much of theactivity in question. Factories will pro-duce additional goods even when theadded revenue doesn’t cover the addedcost, for instance, because part of the costis foisted onto the air-breathing public.That is inefficient.

• An appropriately chosen tax on theactivity in question can correct the inef-ficiency. If the factory has to pay a taxfor each unit of output, or for eachcubic foot of soot it pumps into the air,it will adjust its activity accordingly.

Internality theorists take those three claimsand apply them, with little change, to choiceswithin the individual. They say that individualshave a systematic tendency to pay too littleattention to costs and benefits of decisions totheir future selves. As a result, they will engagein excessive amounts of certain activities, suchas smoking, eating, and drinking. Taxes onthose activities will reduce them, thus makingpeople better off. To illustrate:

• Jonathan Gruber and Botond Koszegi

contend that cigarette smoking producesnegative internalities, argue that govern-ment policies “should depend not onlyon the externalities that smokers imposeon others but also on the ‘internalities’that smokers impose on themselves,”and calculate that “there are sizable opti-mal ‘internality’ taxes on the order of $1per pack or more.”5

• Ted O’Donoghue and Matthew Rabindiscuss “optimal sin taxes” designed tocorrect self-control problems, using thespecific example of overeating.6 In theirnot-yet-published extension of that arti-cle, they note that “since people withself-control problems impose negativeexternalities on their future selves—dubbed ‘negative internalities’ . . . therole that sin taxes play in our analysis ismuch like a Pigovian tax to correct nega-tive externalities.”7

• Other authors, such as Colin Camerer etal.8 and Richard Thaler and Cass Sun-stein,9 have offered more tentative poli-cy prescriptions, preferring to weakenthe prima facie case against paternal-ism. Recent research in behavioral eco-nomics, they say, “potentially broadensthe scope of situations in which pater-nalistic policies could usefully be devel-oped.”10 Still, they invoke the term“internalities” along with its hefty mar-ket-failure baggage: “When consumersmake errors, it is as if they are imposingexternalities on themselves because thedecisions they make . . . do not accurate-ly reflect the benefits they derive.”11

All of these authors use old-fashionedPigovian externality arguments, with no refer-ence to Coase or his insights. Yet Coase threwa monkey wrench into Pigou’s works long ago.

The Reciprocal Nature of Internalities

The question is commonly thought of asone in which A inflicts harm on B and

3

By treating theindividual as amultiplicity ofselves, the newpaternalisminvites policyanalysts toimport the theoryof externalitiesinto the realm ofindividual choice.

what has to be decided is, How should werestrain A? But this is wrong. We are deal-ing with a problem of a reciprocal nature.To avoid harm to B would be to inflictharm on A. The real question that has tobe decided is, Should A be allowed toharm B or should B be allowed to harmA? The problem is to avoid the more seri-ous harm.12

Imagine that the individual consists ofmore than one self. One common approachsays that the individual contains many differ-ent selves, one present-oriented self for eachmoment in time and a single future-orientedself.13 Another approach imagines just twoselves, a single present-oriented self and a sin-gle future-oriented self. When it eases expla-nation, I will use the simpler model.

Coase’s first major insight concerningexternalities was that they are reciprocal innature. In other words, an externality is notsimply a harm imposed by one person onanother. Rather, an externality arises becausetwo (or more) people have conflicting inter-ests, and championing the interests of oneparty means denying the interests of the other.

Take, for example, the case of a residentialneighborhood near an airport. The incomingand outgoing flights create a noise nuisancefor neighboring homes; this is a negativeexternality. Allowing the airport to continueoperation causes harm to the residents. But italso works the other way around: restrictingflights to give residents more peace and quietdoes harm to the airport (and its customers).This point becomes especially clear when wenote that the residents often move into thearea around an airport voluntarily, after theairport is already in operation. Apparently,such residents think the benefits of livingnear the airport compensate for the noise.

But Coase’s argument does not depend onwho was there first. The point is simply thatharm is a two-way street. The existence of anexternality tells us that a tradeoff existsbetween some people’s interests and others’. Itdoes not tell us how the tradeoff should bemade. In some cases, it might make sense to

shut down an airport (or restrict its flights)because the cost of the noise exceeds the bene-fit. In other cases, it will be better to allow theairport to operate unimpeded and let the resi-dents adjust—by moving elsewhere, sound-proofing their homes, or putting up with thenoise in exchange for lower housing costs.

Internality theorists observe that theshort-run self can take actions, such as smok-ing or overeating, that will harm the interestsof the long-run self. They predict the short-run self will therefore engage in too much ofthe harmful activities at the expense of thelong-run self. This approach is fundamental-ly Pigovian. It regards one actor (the short-run self) as the sole cause of the harm, andthe other actor or actors (the long-run self orall future short-run selves) as the passive vic-tims of that harm. As Coase observed in thequotation above, that analysis is one-sided.True, allowing the present self to smoke orovereat means harming the future self. But,by the same token, preventing smoking orovereating on behalf of the future self meansharming the present self.

To take the notion of multiple selves seri-ously, the analyst must consider both sets ofinterests or preferences. We should not simplyassume that the long-run self’s interests some-how supersede those of the short-run self, anymore than we should assume the residents’interests supersede those of the airport. Thus,adopting policies solely on grounds that theyadvance the interests of the long-run selfwould be inappropriate.

As Coase observes, the real problem is toavoid the more serious harm. But nothingabout the situation, certainly not the factthat the short-run self may impose harm onthe long-run self, shows that the long-runharm exceeds the short-run harm. Thisbecomes apparent if we consider that thelong-run self can harm the short-run self byadopting self-control devices—such as flush-ing cigarettes down the toilet, refusing toallow ice cream in the house, checking into aclinic, and so on. The future long-run selfmay also impose the cost of guilt on the pre-sent self.14 Such actions help the long-run

4

The applicationof Coase’s ideas

to internality theory casts

serious doubt on the new

paternalism.

self at the expense of the short-run self. Giventhe reciprocal nature of the problem, andhaving no further information, we could justas easily conclude that the long-run selfimposes internalities on the short-run selfthat require correction. Perhaps we shouldtax weight-loss clinics.

Going a step further, we could observe thatfuture-oriented selves sometimes induce behav-iors that, at least to outside observers, appearexcessive. In contrast to the obese and the prof-ligate, whose short-run selves constantly trumptheir long-run selves, we might point to themisers, workaholics, and anorexics for whomthe reverse appears to be true.15 Even among“normal” individuals, studies show that exces-sive self-control efforts can lead to undercon-sumption of desirable things.16 Perhaps weshould subsidize Krispy Kreme.

Or, following Coase, we could recognizethat harm goes both ways. The existence ofan interactive effect does not, in itself, tell usthat an inefficiency exists; nor does it tell uswhether the inefficiency, if any, results fromtoo much or too little of an activity.

Time Inconsistency and Multiple Selves

Where does this notion of “multiple selves”come from? You’re just one person—right? Itmight seem that way, but internality theoristssay your actions betray you. People often makechoices that reflect internal conflict, or ten-sion between different sets of preferences.Specifically, people display time inconsistency,which (in simple terms) means a conflict be-tween your preferences tomorrow and yourpreferences today.

Suppose, for example, that you’re offered achoice between $100 to be received 100 daysfrom now, or $110 to be received 101 daysfrom now. Many people will choose the largeramount of money. But now take that samechoice and move it forward 100 days, so thatyou’re choosing between $100 today and $110tomorrow. Given that choice, many willchoose the smaller amount—including people

who chose the larger amount for the more dis-tant choice.17 This phenomenon, known aspreference reversal, can lead to inconsistentbehavior, such as making promises for thefuture (“I promise to start my diet Monday”)and then breaking them when the date arrives(“No, I guess I’ll start my diet next Monday”).And lest it seem that those are just idle promis-es, people will even limit their own future optionsto make promise breaking more difficult—forinstance, by emptying their pantries of tempt-ing snacks.

Time inconsistency, according to internali-ty theorists, shows the existence of competinginterests within the self—or, more dramatical-ly, competing selves. And this is not just a run-of-the-mill tradeoff, like the tradeoff betweenwatching TV and going to the movies. Rather,time inconsistency means that the rate oftradeoff itself changes systematically overtime. It reveals a kind of schizophrenia in theindividual, albeit a schizophrenia present evenin the most normal people. The impatientshort-run self places more weight on immedi-ate gratification, whereas the long-run selfplaces greater weight on delayed gratification.And then, the argument goes, the short-runself takes advantage of its control of the bodyto foist harms on the helpless long-run self.

Although time inconsistency does reveal akind of internal conflict, it tells us nothingabout how to resolve the conflict. Look againat the choice between $100 one day and $110a day later. If someone chooses the smalleramount when the choice is near but the larg-er amount when the choice is distant, wecould “correct” him by manipulating him tochoose the larger amount always. That wouldmake his choices consistent—and this is, inessence, what the internality theorists thinkwe should do. But we could also “correct”him by making him choose the smalleramount always. That, too, would make hischoices consistent. So there is more than oneway to “fix” a time inconsistency, and inter-nality theory tells us nothing about which fixto use.18 Once again, we observe the recipro-cal nature of internalities: assisting one set ofpreferences means harming the other.

5

We should notsimply assumethat the long-runself’s interestssomehow supersede thoseof the short-runself.

The Least-Cost AvoiderPrinciple

Suppose that those who suffer thedamage could avoid it by moving toother locations or by taking variousprecautions which would cost them[less than the polluter’s avoidancecost]. Then there would be a gain in thevalue of production . . . if the factorycontinued to emit its smoke and thosenow in the district moved elsewhere ormade other adjustments to avoid thedamage.19

Typically, there exists more than onemeans of averting a harm. The object is toinduce action by the least-cost avoider of harm.

Consider a classic externality story: acement factory spews dust on the residencesthat surround it.20 A simple Pigovian analysissays the factory creates harm to the residentsand, therefore, ought to be taxed for the dustit creates. A Coasean analysis points out thatthe dust nuisance might be avoided orreduced in more than one way. The factorycould shut down or reduce its production.Alternatively, the residents could move away,not move there in the first place, or act tomitigate the dust’s impact (by not hangingtheir washing outdoors, for instance). Whichcourse of action ought to be taken? Nothingin the description of the situation gives usthe answer. In some situations, it would becheaper for the factory to reduce its output(or shut down) than for the residents to move(or reduce their exposure in some other way).In other situations, it would be cheaper forthe residents to change their behavior. In thelatter situation, where the residents are theleast-cost avoiders, a tax on the factory wouldnot improve the situation. The tax wouldtend to reduce the factory’s production, eventhough the value of the lost productionwould exceed the cost to residents of avertingthe same harm.21

Analogously, the harm resulting from aninternality might be avoided in more than one

way. The short-run self could reduce its Twinkieconsumption, eat a Twinkie Lite instead, or haveit with a Diet Coke instead of a Coke. Altern-atively, the long-run self could adopt measuresdesigned to reduce the Twinkie’s future effects.It could, for instance, commit to exercising moreoften (or more vigorously) by joining a gym ormaking agreements with workout partners. Orthe long-run self might resign itself to takingheart medications. Which route is most efficientdepends on the subjective cost of the differentoptions. If the future-oriented self were the least-cost avoider, a Twinkie tax would not improvematters. It would induce the present self to eatfewer Twinkies, even though the future selfcould have avoided or reduced the harm at alower cost.

Returning to the cement factory example,there is a third outcome that might, depend-ing on the parameters, prove efficient: doingnothing. If the value of the factory’s output(which would be lost if the factory shut down)is greater than the damage done, and anyavoidance measures by the factory or resi-dents would impose costs greater than thedamage avoided, then it makes sense to createdust with no avoidance measures at all.Analogously, if the value of the Twinkie to theshort-run self is greater than the damagedone to the long-run self, and avoidance mea-sures by either self involve costs higher thanthe damage they avoid, it makes sense to eatthe Twinkies without countermeasures.

Property Rights andExchange

It is always possible to modify by trans-actions on the market the initial legaldelimitation of rights. And, of course, ifsuch market transactions are costless,such a rearrangement of rights willalways take place if it would lead to anincrease in the value of production.22

Assuming the harm exceeds the avoidancecost, what can induce the least-cost avoiderto take the appropriate action? Coase sug-

6

In contrast to theobese and the

profligate, whoseshort-run selves

constantly trumptheir long-run

selves, we mightpoint to the

misers, workaholics, and

anorexics.

gests the possibility of transactions betweenthe parties. If the cement factory is the least-cost avoider, but the factory is allowed by lawto pollute, the residents can pay the factoryto shut down or cut back production. Morebroadly, Coase’s point is that people can findcreative ways to negotiate with each other toexploit opportunities for gain. Could a simi-lar solution apply to internalities?

I contend that the answer is yes. Just as dif-ferent people can make deals, different selveswithin a person can make deals. I will call suchdeals intrapersonal bargains. Those bargains canbe struck in at least three different ways.

Collusion for Mutual GainThis sort of intrapersonal bargaining, first

explained in detail by psychologist GeorgeAinslie,23 relies on a specialized form of cooper-ation among a succession of present selves.Ainslie observes that even your impatient pres-ent self cares to some degree about your futureselves—just not enough to make the preferencesof present and future selves perfectly consistent.Your present self would actually like to seegreater self-control, because that would benefitall future selves, which the present self caresabout, too. The self-control problem arisesbecause the present self would like to make aspecial exception for itself. In choosing betweenstarting a diet today and starting a diet tomor-row, the present self prefers the latter. But thesame goes for tomorrow’s present self oncetomorrow arrives, so the diet will be postponeduntil the day after tomorrow. And so on, withthe result that the diet never begins.

But what if the present self thought that ifthe diet didn’t begin today, it never would? Orwhat if the diet had already begun, and thepresent self knew breaking the diet would setin motion a series of exceptions that wouldeventually destroy the diet? In that case, theconsequences of breaking the diet, or failing tostart one, could be sufficiently great that thepresent self would choose to “be good.” Andthe same goes for tomorrow’s self facing thesame choice.

This means there’s room for a cooperativedeal to be struck by all the selves (both the pres-

ent self and all the future selves that will even-tually arrive in the present). They all agree tolimit their indulgences. Each self gives upsome transient pleasures in return for therestraint exercised by all the others, and on netthey are all better off. The agreement isenforced by each self’s desire not to destroythe agreement.

The key to Ainslie’s intrapersonal bargain-ing solution is that the cooperative agreementeffectively confronts each present self with a“package deal.” If the present self weighed thebenefit of overeating just this once against thesmall future cost of overeating just this once, itwould choose to indulge. But the agreementmakes it impossible to overeat just once;overeating triggers more overeating. The rele-vant cost is therefore the cost of overeatingrepeatedly, which is large enough to make thepresent self abstain.

The agreement just described may soundodd, because we are not accustomed to think-ing of ourselves as multiple selves. But in fact,the process is quite common. In enforcing ourresolutions, we are loath to make exceptionsfor fear that they will set a precedent for ourown future actions. Successful dieters oftenadopt rigid personal rules to govern their eat-ing. People trying to quit smoking often do so“cold turkey,” because they fear smoking onecigarette will lead to smoking another, andanother, until the resolution to quit has beendefeated. Wage earners will save a certainamount of money each month, and they willstrongly resist reducing the amount—even forjust one month—lest they get in the habit ofspending more every month.

Personal rules help define the amount ofrestraint expected of the selves. Bright-line rules,in particular, are valuable as precedents becausethey can clearly indicate when a present self haschosen to defect from the agreement. Ainsliesees “rationalizations, blind spots, and circum-scribed lapse districts” as exceptions to the rulesthat can “defeat your resolutions.”24 He attrib-utes the problem of backsliding to the tempta-tion to make exceptions.25 The downside of thecollusive solution is that it often results inexcessive rigidity. Nonetheless, many people

7

The self-controlproblem arisesbecause the present selfwould like tomake a specialexception foritself.

voluntarily adhere to rigid personal rules, pre-sumably because they think the benefits of self-control outweigh the costs.

Establishing Property BoundariesA key element of the intrapersonal bargain-

ing model just described is that only presentselves make decisions. Future selves’ interestsmatter only because the present self happensto care about them (although maybe not asmuch as it should). But that’s not the only wayto look at the problem. As an alternative, sup-pose that the present self does not have exclu-sive decisionmaking rights. Instead of repre-senting a temporal locus of control, the presentself represents certain interests with a moreimmediate payoff, while the future self repre-sents interests with a more distant payoff. Thetwo selves exercise joint decision rights overthe person. This approach treats the body as akind of common asset, over which the selvesseek to exert control.

In this situation, war is one possible out-come.26 Each self seeks to advance its owninterests while sabotaging the other. The pres-ent self searches for chances to overindulge infood, drink, sex, spending, and so forth. Thefuture self finds ways to limit the present self’spleasure—by ridding the household of snacks,throwing cigarettes away, or signing up forautomatic savings plan contributions. Thefuture self may also spoil the present self’spleasure by creating guilt or by imposing“oversight” and planning on activities the funof which derives from their spontaneity.

War is costly to both parties. The presentself consumes with attenuated pleasure. Thefuture self ’s expenditures on enforcementdiminish the gains from satisfying its more dis-tant interests. As a result, each self prefers anegotiated outcome. The bargain takes theform of a redistribution of property rights:instead of both selves exercising control at thesame time, each self cedes some control overcertain kinds of decisions in exchange forexclusive control over others.

Commonplace experience affirms that dif-ferent interests tend to operate in different cir-cumstances. Individuals adopt rules of self-

control, such as, “I will smoke only in social sit-uations,” “I will not drink alone,” “I will not eatafter midnight,” “I can ignore my diet while onvacation.” Obviously, such personal rules pro-scribe behavior, and so they are typically inter-preted as tools of one’s long-run interests. Yetthe rules are as notable for what they allow asfor what they prohibit. Within specified zones,they enable the individual to “let loose” andenjoy life’s pleasures without guilt and over-sight.

An even better example of an intrapersonalbargain that both constrains and enables is theestablishment of separate budgets or accountsfor particular activities, such as when a gamblercreates a personal gambling fund. Although thefund limits total losses from gambling, it alsoenables the gambler to gamble freely withoutworrying about the effects of (sufficientlysmall) losses on other kinds of consumption.This represents a mutually beneficial exchangebetween the present and future selves.

Research confirms that people use mentalaccounts as a means of establishing bound-aries.27 Heath and Soll show that people dividetheir total resources into “separate mentalaccounts (e.g., entertainment or householdexpenses) and then track expenses against thebudgets.”28 Wertenbroch observes that peopleration their consumption of both “virtue” and“vice.”29 Kivetz and Simonson provide what islikely the best evidence that separate mentalaccounts enable as well as limit consumption:people will deliberately precommit to indul-gence by (for instance) choosing luxuries overnecessities or cash as lottery prizes.30

Here, as in the collusion model discussedabove, personal rules assist in the enforcementof an intrapersonal bargain. Unlike that model,the “exceptions” in this model are an integralpart of the bargain itself; they are the presentself ’s compensation. The collusion modeldemonstrates that a certain amount of intra-personal altruism—the fact that your selvescare about each other—can assist in creatingrules of self-control. The urge to make excep-tions to the rules constitutes a threat. Theproperty rights model, however, shows thatsome cooperation can occur even if the selves

8

The downside of the collusive

solution is that itoften results in

excessive rigidity.

don’t care about each other. In this case, theexceptions don’t necessarily threaten the agree-ment; on the contrary, the exceptions allow theagreement to happen in the first place.

Mutually Beneficial ExchangeIf the present and future selves value goods

or activities on more than one dimension,then we can imagine yet another kind ofintrapersonal bargain. For this kind of bar-gaining to work, one’s selves must possesssome form of nonjoint control over the per-son—either because they have established itthrough a prior agreement as outlined aboveor because they possess such control inherent-ly. Suppose, for instance, there are two dimen-sions of choice: money (present versus futureconsumption) and food (present indulgenceversus future health). And suppose initiallythe future self has greater control over finan-cial decisions, whereas the present self hasgreater control over eating decisions. Thefuture self could offer the present self a deal:don’t eat that fried chicken, and buy a CDinstead. The present self exchanges eatingpleasure for listening pleasure. The future selfexchanges money (the price of the CD plusinterest) for health.

Again, ample evidence supports the idea ofintrapersonal exchange. Kivetz and Simonsonfind that people are most likely to choose luxu-ry rewards for frequent-use programs whenthey have exerted more effort to obtain therewards,31 and they are also more likely tochoose luxury rewards when the necessaryefforts were related to work rather than plea-sure—such as using frequent flier miles for plea-sure travel if they were earned via business trav-el. People also engage in self-gifting to rewardthemselves for virtuous behavior.32 Such giftsoften perform an “exchange” function by act-ing as “self-contracts in which the reciprocityfor the gift is also personal effort and achieve-ment.”33 Studies have demonstrated the effica-cy of self-imposed reward schemes in motivat-ing greater effort and performance.34

The three modes of bargaining I havedescribed—collusion, establishment of prop-erty rights, and exchange—need not be mutu-

ally exclusive, of course. Just as both altruismand self-interest operate between persons, theyalso both operate within persons. A limiteddegree of intrapersonal altruism could allowfor collusive agreements among the selves,while still allowing room for “détente” agree-ments to avoid costly wars between competingpresent and future interests. Further opportu-nities for gain can be exploited via exchangesbetween the selves.

What Could PossiblyGo Wrong?

In order to carry out a market transac-tion, it is necessary to discover who it isthat one wishes to deal with, to informpeople that one wishes to deal and onwhat terms, to conduct negotiationsleading up to a bargain, to draw up thecontract, to undertake the inspectionneeded to make sure that the terms ofthe contract are being observed, and soon.35

Given the multiple possibilities for bar-gaining among one’s selves, what mightobstruct an efficient and workable outcome?Coase’s answer to that question was high trans-action costs. Transaction costs are the costs(monetary and otherwise) of coordinating theparties to a bargain, negotiating terms, andenforcing the agreement that results.

Transaction costs arise from varioussources, but in this context, the most prob-lematic is contract enforcement. Agreementsbetween individuals can be made legally bind-ing by means of explicit contracts enforced bythe state legal system. But most intrapersonalagreements must be enforced internally, asthe legal apparatus is not usually available.This does not rule out intrapersonal bargain-ing entirely, but it does mean bargainingselves must depend on mechanisms that aretypically less reliable: repeated dealings andreputation. A virtue of the collusion model isthat it explicitly incorporates the problem ofenforcement, with the solution depending on

9

There are at leastthree strong reasons to beskeptical of governmentinterventionsdesigned to fixinternality problems.

each self’s interest in sustaining cooperationin the future. The other forms of bargainingdiscussed earlier might be enforced by similarmeans. If one self persistently violates theterms of its agreements, it signals to otherselves its lack of reliability, thus reducing theirwillingness to make future deals with the vio-lator. The potential violator, realizing this,has reason not to act opportunistically.

The viability of repeated dealings and repu-tation as modes of contract enforcementdepends on the open-ended character of the sit-uation. When the cessation of interactionbecomes imminent, “end-game” behaviors canlead to the breakdown of bargaining solutionsin both interpersonal and intrapersonal con-texts. We might, therefore, expect less self-con-straint on the part of people whose lives arecoming to an end (though a rational unifiedself would also engage in greater indulgenceunder the same circumstances). Also, agree-ments require adequate policing. John Ameriks,Andrew Caplin, and John Leahy identify “mon-itoring abilities” as one of the skills that enablehouseholds to rein in excessive spending to savemore money.36 Presumably, someone with bet-ter monitoring skills can monitor internalagreements at a lower cost.

Although legal enforcement is usually un-available, other forms of external enforcement doexist. Ainslie refers to such means as “extrapsy-chic commitments,”37 a category that includesjoining Alcoholics Anonymous and WeightWatchers to enlist the support of other people oradvertising one’s resolutions to friends and fam-ily. Precommitments can also help to enforcecontracts by making deviation impossible or verycostly. Such commitments include deadlines38

and automatic savings plan deductions,39 as wellas the tactics mentioned earlier, such as banningfatty foods from the household.

Transaction costs can also arise from theparties’ lack of information about each other. Abargainer may, for instance, hold out for a larg-er share of the gains from trade simply becausehe thinks the other party values the transactionmore than he actually does. In the intraperson-al context, such a problem is less likely.Although it is conceivable that selves may lack

perfect knowledge of each other,40 such knowl-edge will still be markedly greater than that pos-sessed by different people in an interpersonalcontext. Hiding or falsification of informationcannot be accomplished as easily, given thatboth selves have access to the same mind.

Bargaining Failure versusGovernment Failure

There is, of course, a further alterna-tive, which is to do nothing about theproblem at all. And given that the costsinvolved in solving the problem by reg-ulations issued by the governmentaladministrative machine will often beheavy . . . it will no doubt be common-ly the case that the gain which wouldcome from regulating the actionswhich give rise to the harmful effectswill be less than the costs involved ingovernmental regulation.41

Given the difficulty of internal contractenforcement, it stands to reason that people donot always succeed in exercising self-control.Some individuals fail at finding effective intra-personal bargains; they tend to overindulge.Others find imperfect solutions that result insome self-control but not enough. And yetother individuals find solutions that are tooeffective, resulting in excessive self-control andunderindulgence. Does it follow that some formof paternalist intervention would correct thoseproblems?

The Coasean perspective argues otherwise.There are at least three strong reasons to beskeptical of government interventions designedto fix internality problems. First, even thoughsome individuals fail to exercise self-control,others succeed. That means internalities are, tosome degree, already addressed through intra-personal bargains. Government interventionscould thwart or supersede such bargains. Inaddition, new bargains would be struck in a dif-ferent regulatory environment, so we have toask whether the new bargains would be prefer-able to the old ones.

10

The policy that effectively corrects some

people’s problems will failto correct, or may

even exacerbate,others’ problems.

Second, interventions have problems oftheir own. Just as it is incomplete to arguethat a market failure alone justifies econom-ic regulation, it is incomplete to argue that afailure of individual choice justifies paternal-ist regulation. In both cases, the possibility ofgovernment failure must be taken intoaccount. Governments often lack the infor-mation, the incentives, or both to make wiseregulatory decisions.

Third, regulations usually have a “one-size-fits-all” quality, inasmuch as they affectall citizens (though to differing degrees). Butpeople are heterogeneous, meaning the policythat effectively corrects some people’s prob-lems will fail to correct, or may even exacer-bate, others’ problems.

To make these arguments more concrete,we need to consider specific proposals. Here,I will focus on the most obvious and com-monly suggested proposal for controllinginternalities: the fat tax. More generally, theanalysis here will apply to any “sin” taxdesigned to induce individuals to make bet-ter personal health decisions.

Coase + Pigou = Trouble:The Interaction of Taxesand Private Bargaining

Suppose, for simplicity, that the presentself makes all decisions about eating, andthat the present self cares only about itself.The present self ’s choice to eat Twinkies cre-ates benefits for itself and imposes costs onfuture selves. Given these assumptions, anaïve policy analyst would predict that thepresent self would keep eating Twinkies aslong as doing so created any benefits—even ifthe future costs were very large and the bene-fits very small. So it might seem like a goodidea to tax the present self ’s eating choices.The tax should be set equal to the futurecosts of eating Twinkies, so that the presentself will take exactly those costs into account.

But this analysis is incomplete, because itignores the possibility of intrapersonal bar-gaining. Suppose, for instance, that there are

no transaction costs, meaning the selves canreach and enforce internal bargains with lit-tle difficulty. In this case, an optimal out-come would occur without any tax. Any timethe present self ’s choice to eat something“bad” would incur greater costs (to the futureself) than benefits (to the present self), thereis room for a trade. The future self can offersome compensation to the present self, per-haps by offering a reward for abstaining.Since the cost of eating poorly is greater thanthe benefit, any reward between the twowould work: the future self would willinglyoffer the reward, and the present self wouldwillingly accept.

In this situation, no tax is necessary. Butimagine that a tax is imposed anyway, in themistaken belief that internal bargaining doesnot occur. Assume that the tax is paid entirelyout of the present self’s budget. This tax willactually result in too little consumption. Supposethe present self could eat a Twinkie, and the pre-sent benefit is great enough to justify the futurecost. The present self should eat. But now the taxdiminishes the perceived benefit of eating. Thefuture self will be able to offer a reward to thepresent self that will induce it to eat this time,whereas without the tax no reward offered bythe future self would be large enough. For exam-ple, if we measure value in dollar terms, onemore Twinkie might be worth $5 to the presentself, while causing $4 worth of future damage.This is a Twinkie worth eating, and without thetax it will be eaten. (The future self would notwillingly offer more than $4 worth of compen-sation, and thus the present self would notagree.) But suppose a tax of $1.50 is levied oneach Twinkie. Now the present self expects a netbenefit of only $3.50, so it will accept a rewardoffered by the future self not to eat.

This analysis assumes that the tax is paidfrom the present self’s budget. But that maynot be true. If the present self does not care(enough) about the future self, why not simplygo into debt to pay sin taxes? Incurring debtpasses the tax on to the future self. In this case,the future self would perceive an even larger illeffect from the present self’s consumption—first the reduction in health and second the

11

Any one-size-fits-all policy will necessarily beefficient for onlya fraction of thepublic at best.

reduction in budget. Given the greater cost,the future self would willingly offer largerrewards to induce the present self to reduce itsconsumption. And once again, the result willbe too little consumption. Take the same fig-ures as above: a Twinkie that produces $5 pres-ent benefit and $4 future cost. If the future selfalso expects to experience a $1.50 increase indebt, then it will offer as much as $5.50 for thepresent self’s cooperation—and the presentself will accept the offer. The Twinkie does notget consumed, even though it should. In addi-tion, the future self ends up making largerpayments to the present self, thus ironicallyreducing the future self’s welfare relative towhen there was no tax.

I’ve assumed so far that transaction costsare zero. At the opposite extreme, supposetransaction costs are prohibitively high, sothat no intrapersonal bargains are made.Here, the case for a fat tax is stronger. The taxforces the present self to consider the cost tothe future self, when otherwise that costwould not have been considered. If the taxcomes entirely from the present self ’s budget,then (as before) the tax effects a welfare shiftfrom the present to the future self. This is theideal situation for the fat-tax advocate.

On the other hand, if the present self canoffload the tax to the future self by going intodebt or depleting savings, then the tax has noimpact on the present self’s consumption.The tax is experienced by the future self as anincrease in the cost of present consumption—but by supposition, high transaction costs pre-vent the future self from making a viablereward offer to the present self to induce it toreduce its consumption. In addition, the taxrevenue is lost to the future self. The future selfactually ends up worse off, unless the tax rev-enues are rebated or spent in a way that bene-fits only the future self.

In reality, transaction costs are neither zeronor prohibitively high. Some internal bargainswill be made, others not. The outcome willexhibit elements of both situations justdescribed. The key insight is that some,though probably not all, of the present self’sfuture costs will already have been internalized

through intrapersonal bargains. Any tax thatfails to account for this process, or to accountfor it fully, will be too large and thus result inunderconsumption by some people. In addi-tion, if the present self can shift taxes to thefuture self, the policy will tend to diminish thefuture self’s welfare.

Unraveling IntrapersonalBargains

In the tax analysis above, I treated the bar-gains between the present and future self asthough they were struck in a precise manner,corresponding to exact quantities of con-sumption. But bargains often take the form ofpersonal rules that divide up or reallocate deci-sionmaking power. Instead of specifying thenumber of fat calories the present self mayconsume, the rule might specify times andplaces at which the present self may freely con-sume fat and other circumstances in which itmay not.

The effects of a fat tax on idiosyncratic bar-gains of this kind are more difficult to parse.In the short run, existing personal rules willlikely persist. Especially when transactioncosts are high, bargainers have an interest inmaintaining existing agreements to econo-mize on such costs and avoid a breakdown inthe relationship, even if those agreements areno longer optimal. The present self may con-tinue eating fatty foods only on weekends andvacations, for example. This could occur eventhough it would make sense, given the tax, toshrink the set of allowed indulgence zones. Ifthe present arrangement is already optimal ornearly so, such persistence could be desirable.If the individual had not succeeded in reach-ing internal bargains for self-control, then thepersistence of old rules would be undesirable,though the tax would not aggravate the situa-tion (except by reducing the selves’ income).

Eventually, however, people will renegoti-ate their internal bargains. To minimize thetax’s impact, they will find it worthwhile toreduce their level of consumption. They willtry to find a new set of personal rules that

12

If internalitytheory is to be

taken seriously, it should

incorporateat least some of

the lessonslearned in the

last half centuryfrom research on

externalities.

approximate the desired level of consump-tion, which may be difficult to do (perhapsindulgence is allowed only every other week-end or only on Sundays). In any case, whatev-er new rules appear might or might notimprove overall welfare. To the extent thatthe tax falls on only the present self ’s incomeand transaction costs prevent the negotiationof efficient personal rules in the absence ofthe tax, the tax will tend to induce better per-sonal rules. But if transaction costs are lowenough that the selves eventually tend toarrive at near-efficient rules absent the tax,the tax will tend to reduce consumptionbelow the optimal level. And if transactioncosts are high while the present self canoffload the tax to the future, the tax willreduce the future self ’s income while failingto reduce the present self ’s consumption.

Information and Incentives

Given the difficulties outlined above, would-be paternalist regulators face a daunting task.Even without Coasean considerations, optimaltaxation of internality-producing behaviorwould be no simple task. The optimal tax wouldbe equal to the marginal cost of the behavior tofuture selves. To calculate this amount, regula-tors would first need to find the “true” rate oftradeoff between present and future satisfac-tion. The problem, as noted earlier, is that the“true” rate of time-discounting is a phantom. Topick one rate of tradeoff over another is to privi-lege one set of subjective preferences over anoth-er, without any basis for doing so.

But suppose the regulators somehowfound the “right” rate of tradeoff between thepresent and the future. Even then, theywould need to discern the degree to whichpeople have already dealt with their internal-ities through intrapersonal bargaining. Suchinformation will not be readily available. Thephenomenon of time inconsistency has beenidentified primarily under laboratory condi-tions, in which test subjects are presentedwith stylized choice situations (e.g., “Wouldyou prefer $100 now or $110 tomorrow?

Would you prefer $100 a year from now or$110 in a year and a day?”). The rates of trade-off over time revealed by such experimentswill not necessarily, or even likely, approxi-mate the rates of tradeoff used by people inreal-world situations. The actual devices peo-ple use to define and enforce intrapersonalbargains, and thus to induce more future-ori-ented behavior, most often involve personalrules based on circumstances (e.g., “Am I in abar right now? Am I on vacation?”) that donot appear in the laboratory setting.

Moreover, even if regulators could discernboth the “right” rate of time tradeoffs and theactual rates of tradeoff implicit in people’sbehavior, they would still face the unenviabletask of estimating the degree to which subse-quent choice by the regulated people willundermine their policies’ intended results.Since people may change their choice processin response to policy changes—for example, byaltering the terms of their internal agree-ments—it follows that realized rates of trade-off will be endogenous to the policy choice.

Furthermore, people are heterogeneous—in the size of their initial internality problems,the magnitude of their internal transactioncosts, and the type of personal rules availableand attractive to them. Any one-size-fits-allpolicy will necessarily be efficient for only afraction of the public at best. Others will beunaffected or affected adversely by beingmanipulated into suboptimal consumption,or affected too little because the policy doesn’tgo far enough. Any attempt to improve thepolicy’s effectiveness vis-à-vis the latter groupwill have undesired and often unexpected con-sequences for the other groups.

And with all of the informational difficul-ties, we have not even begun to ask whetherregulators will have the appropriate incentivesto find the correct answers and implementthem.

Conclusion

Does the theory of internalities justifygovernment intervening in people’s lives “for

13

There’s no validreason to assume,when there is aninconsistencybetween presentand future interests, that thelatter must trumpthe former.

their own good”? The new paternalists clear-ly think so. But their argument is extraordi-narily weak.

The theory of internalities is explicitly mod-eled on the theory of externalities. If internalitytheory is to be taken seriously, it should incor-porate at least some of the lessons learned in thelast half century from research on externalities.But that hasn’t happened yet. As it stands, thecase for paternalism based on internality theorysuffers from several major flaws.

First, the new paternalism blithely assumesthat, when your present self can impose costson your future self, the outcome is necessarilybad. But preventing harm to the future selfmight involve even greater harm to the presentself. There’s no valid reason to assume, whenthere is an inconsistency between present andfuture interests, that the latter must trump theformer.

Second, the new paternalism ignores thefact that harms can be avoided in multipleways. Restricting present behavior is one wayto reduce future harms, but that doesn’tmake it the best way. The future self might becapable of mitigating the harm at lower costby other means.

Third, the new paternalism neglects thepossibility of internal bargains and privatesolutions. All of us face self-control problemsfrom time to time. But we also find ways tosolve, or at least mitigate, those problems. Wemake deals with ourselves. We reward our-selves for good behavior and punish ourselvesfor bad. We make promises and resolutions,and we advertise them to our friends and fam-ilies. We make commitments to change ourown behavior. Internality theorists point tothese behaviors as evidence that the internali-ty problem exists. But they are actually evidenceof the internality problem being solved, at leastto some degree.

People are not perfect, so we should notexpect real people’s actions to mimic those ofperfectly rational and perfectly consistentbeings. Mistakes will occur; self-control prob-lems will persist. But paternalist solutionswill solve them no better than personal solu-tions. What is really at stake is how self-con-

trol problems will be addressed—through pri-vate, voluntary means or through the force ofgovernment.

The new paternalists would have us believethat benevolent government can—through taxes,subsidies, restrictions on the availability of prod-ucts, and so on—make us happier according toour own preferences. But even if we place little orno value on freedom of choice for its own sake,the paternalists’ recommendations simply don’tfollow. Public officials lack the information andincentives necessary to craft paternalist policiesthat will help the people who most need help,while not harming those who don’t need thehelp or who need help of a different kind.Individuals, on the other hand, have every reasonto understand their own needs and find suitablemeans of solving their own problems.

Notes1. This is a simplification, since some new pater-nalists also draw on arguments about individuals’lack of information or poor information-process-ing skills. Here, I will set aside those related butdistinct arguments to focus on the problem ofintrapersonal conflict and choice.

2. “I say only that people act as if there were twoselves alternately in command . . . the ways inwhich people cope, or try to cope, with loss ofcommand within or over themselves are muchlike the ways that one exercises command over asecond individual.” Thomas Schelling, “Ethics,Law, and the Exercise of Self-Command,” inChoice and Consequence: Perspectives of an ErrantEconomist, ed. Thomas Schelling (Cambridge, MA:Harvard University Press, 1984), p. 84.

3. Ronald H. Coase, “The Problem of Social Cost,”Journal of Economic Literature (October 1960).

4. A. C. Pigou, The Economics of Welfare, 4th ed.(London: Macmillan, 1932).

5. Jonathan Gruber and Botond Koszegi, “IsAddiction ‘Rational’? Theory and Evidence,” QuarterlyJournal of Economics 116 (2001): 1261–94.

6. Ted O’Donoghue and Matthew Rabin, “Study-ing Optimal Paternalism, Illustrated by a Modelof Sin Taxes,” American Economic Association Papers& Proceedings 93 (2003): 186–91.

7. Ted O’Donoghue and Mattew Rabin, “Optimal

14

Individuals haveevery reason to

understand theirown needs and

find suitablemeans of solving

their own problems.

Sin Taxes,” unpublished manuscript, Universityof California at Berkeley, 2003, p. 2, n. 3.

8. Colin Camerer et al., “Regulation for Conservatives:Behavioral Economics and the Case for ‘AsymmetricPaternalism,’” University of Pennsylvania Law Review151 (2003): 1211–54.

9. Richard H. Thaler and Cass R. Sunstein, “Libertar-ian Paternalism,” AEA Papers and Proceedings 93(2003): 175–79.

10. Camerer et al., p. 1214.

11. Ibid., p. 1221.

12. Coase, p. 2.

13. See, for instance, Richard H. Thaler and H. M.Shefrin, “An Economic Theory of Self-Control,”Journal of Political Economy 89 (1981): 392–406. Seealso Jon Elster, “Weakness of the Will and theFree-Rider Problem,” Economics and Philosophy 1(1985): 231–65.

14. Ran Kivetz and Itmar Simonson, “Self-Controlfor the Righteous: Toward a Theory of Precommit-ment to Indulgence” Journal of Consumer Research 29(2002): 199–217; and Dana N. Lascu, “ConsumerGuilt: Examining the Potential of a New MarketingConstruct,” in Advances in Consumer Research 18(1991): 290–95. These scholars, among others, doc-ument the importance of guilt as a motivator indecisionmaking.

15. See, for example, Tyler Cowen, “Self-Constraintversus Self-Liberation,” Ethics 101 (1993): 360–73;and George Ainslie, Breakdown of Will (New York:Cambridge University Press, 2001), p. 115.

16. See Chip Heath and Jack B. Soll, “MentalBudgeting and Consumer Decisions,” Journal ofConsumer Research 23 (1996): 40–52.

17. Mathematically, time inconsistency followsfrom a quasihyperbolic utility function such asthis one:

T

Ut(ut + ut+1, . . . , uT) = ut + bnd5uss=t+1

The discount factor b, which is less than one, rep-resents the agent’s degree of present bias. Since bdoes not apply to the present period (time t), thetradeoff between any two periods will depend onwhether one of those periods is the present. As aresult, the tradeoff changes over time instead ofbeing consistent. If b equals one, the agent is per-fectly consistent.

18. The two “corrections” can be seen using the

quasihyperbolic utility function above. One cor-rection involves setting b equal to one; this favorsfuture selves. The other correction involves apply-ing b to every period instead of just the present(in effect, lowering d to db); this favors the pres-ent self. With either correction, the person’s be-havior will become consistent over time.

19. Coase, p. 41.

20. See, for example, Boomer v. Atlantic Cement Co.,26 N.Y.2d 219, 257 N.E.2d 870; 309 N.Y.S.2d 312(NY Court of Appeals 1970).

21. I ignore, for the sake of simplicity, the possi-bility of the factory paying the residents to moveaway in expectation of a reduced tax burden.

22. Coase, p. 15.

23. George Ainslie, Picoeconomics: The StrategicInteraction of Successive Motivational States within thePerson (Cambridge: Cambridge University Press,1992).

24. Ibid., p. 189.

25. George Ainslie, “How Do People Choosebetween Local and Global Bookkeeping?” Behavioraland Brain Sciences 19 (1996): 574–75.

26. Consider Thomas Schelling’s notion of limit-ed war, which George Ainslie describes as the rela-tionship of “bargaining agents who have someincompatible goals but also some goals in com-mon.” George Ainslie, “A Research-Based Theoryof Addictive Motivation,” Law and Philosophy 19(2000): 100.

27. See Richard H. Thaler, “Mental Accounting andConsumer Choice,” Marketing Science 4 (1985):199–214.

28. Heath and Soll, p. 40.

29. Klaus Wertenbroch, “Consumption Self-Controlby Rationing Purchase Quantities of Virtues andVice,” Marketing Science 77 (1998): 317–37.

30. Kivetz and Simonson, “Self-Control for theRighteous.”

31. Ran Kivetz and Itamar Simonson, “Earningthe Right to Indulge: Effort as a Determinant ofCustomer Preferences toward Frequency ProgramRewards,” Journal of Marketing Research 39 (2002):155–70.

32. David Glen Mick and Michelle DeMoss, “Self-Gifts: Phenomenological Insights from FourContexts,” Journal of Consumer Research 17 (1990):322–32; David Glen Mick, “Giving Gifts to Our-

15

selves: A Greimassian Analysis Leading to TestablePropositions,” in Marketing and Semiotics: SelectedPapers from the Copenhagen Symposium, ed. HanneHartvig Larsen, David Glen Mick, and ChristianAlsted (Copenhagen: Hendelshojslolens, 1991), pp.142–59; and David Glen Mick, “Self-Gifts,” in GiftGiving: A Research Anthology, ed. Cele Otnes andRichard F. Beltramini (Bowling Green, OH: BowlingGreen State University 1996), pp. 99–120.

33. Mick and Demoss, p. 326.

34. See, for example, Albert Bandura and BernardPerloff, “Relative Efficacy of Self-Monitored andExternally Imposed Reinforcement Systems,” Jour-nal of Personality and Social Psychology 7 (1967):111–16; and Albert Bandura and Dale H Schunk,“Cultivating Competence, Self-Efficacy, and IntrinsicInterest through Proximal Self-Motivation,” Journalof Personality and Social Psychology 41 (1981): 586–98.

35. Coase, p. 15.

36. John Ameriks, Andrew Caplin, and John J.Leahy, “Wealth Accumulation and the Propensityto Plan,” Quarterly Journal of Economics 118 (2003):1007–47.

37. George Ainslie, Breakdown of Will, pp. 74–76.

38. Dan Ariely and Klaus Wertenbroch, “Procrastina-tion, Deadlines and Performance: Self-Control by Pre-commitment,” Psychological Science 13 (2002): 219–24.

39. Richard H. Thaler and Shlomo Benartzi, “SaveMore Tomorrow ™: Using Behavioral Economicsto Increase Employee Savings,” Journal of PoliticalEconomy 112 (2004): S164–S182.

40. See, e.g., Roland Bénabou and Jean Tirole,“Will Power and Personal Rules,” Journal of PoliticalEconomy 112 (2004): 848–86, for a model of self-control that incorporates imperfect recall.

41. Coase, p. 18.