The Myth That Promisees Prefer Supracompensatory Remedies: An

The Myth that Promisees Prefer SupracompensatoryRemedies: An Analysis of Contracting for Damage Measures

Alan Schwartzt

Courts will not enforce liquidateddamage clauses when a stipulatedsum exceeds (i) the harmthat the promiseecould reasonablyexpect to suffer from breachor (ii) the actualharm that breach turned out to cause.' Courtstradition- ally have not awardedpunitive damages "for a breachof contractunless the conduct constitutingthe breachis also a tort for which punitive damagesare recoverable."2Courts also will not grant specific performance"if damages would be adequateto protectthe expectationinterest of the injuredparty," nor will courts enforce contractsthat accordpromisees a right to specific relief.3 These threerules sharethe goal of limitinga promisee'srecovery to his lost expectation.The liquidateddamage rule preventsthe promiseefrom contracting for a supracompensatoryremedy, and the punitivedamage rule preventscourts from awardingsuch a remedy.The specific performancerule achievesthe law's goal indirectly.A promiseewho has a rightto specific performancecan compel the promisor to perform even when the promisor'sloss from performance would exceed the promisee'sgain. A promisorcan purchaseher freedom,but sophisticatedpromisees sometimes will demandmore thantheir expectation as the price.Permitting specific performanceonly whendamages could not protect the expectation interest limits the ability of promisees to obtain supracom- pensatorypayments by threateningto seek specific relief. The ban on "specific performanceclauses" prevents a promiseefrom obtaining by contractthe power that the general specific performancerule aims to abolish.

t William K. TownsendProfessor of Law, Yale Law School; Professor,Yale School of Organization & Management. Earlier versions of this article were presented at the 1989 Canadian Conference of CommercialLaw Professorsand a Symposiumon ContractTheory at Tel Aviv Law School held in March 1990. Peter Cramtonmade very helpful suggestions. Jules Coleman, RichardCraswell, Henry Hansmann, JonHanson, Jason Johnston, Avery Katz, andMichael Trebilcock made perceptive comments on priordrafts. The authoris an Associate Reporterto the AmericanLaw InstituteProject: Compensation and Liabilityfor Productand Process Injuries.The thoughtsexpressed here are the author's,reflecting neitherthe views of other Reporterson the Project nor the views of the ALI. 1. RESTATEMENT(SECOND) OF CONTRACTS? 356 (1981); U.C.C. ? 2-718(1) (1989). 2. RESTATEMENT(SECOND) OF CONTRACTS ? 355 (1981); U.C.C. ? 1-106(1) (1989) (by inference). 3. RESTATEMENT(SECOND) OF CONTRACTS ? 359(1) (1981); U.C.C. ? 2-716(1) (1989). 369 370 The Yale Law Journal [Vol. 100: 369

These three rules rest on a normativepremise and on a positive premise. The normativepremise holds thatsupracompensatory remedies are undesirable. The positive premiseholds thatpromisees prefer supracompensatory remedies, and so must be preventedfrom getting them.The normativepremise is truebut the positive premise is false. It is shown below that promisees do not want contractualdamage measures that would grant more than their lost expectation. Several legal implicationsfollow from this showing. First, the initial, or "ex ante," branchof the liquidateddamage rule is unnecessary.Courts do not have to preventpromisees from obtainingpenalty clauses if promiseesdo not wantpenalty clauses. The ex ante rule is not merely unnecessary:judicial review produces mischief. Courts sometimes mistake compensatorydamage measures for penalties,and so have foundthat particular liquidateddamage clauses would inevitablyovercompensate promisees when those clauses only protectedthe expectation.Thus, the ex ante branchof the liquidateddamage rule should be abandoned. The "expost" branch of the liquidateddamage rule, which bans clausesthat overcompensatein fact, seemsjustifiable at first glance. An ex ante reasonable estimate of the damages that the promisee will later incur may exceed the promise's actualloss. The normativepremise that supracompensatory remedies are undesirablethen apparentlyjustifies the ex post branchof the rule, because it implies that courtsshould not enforceliquidated damage clauses thatexceed actualdamages. This view is unpersuasive:party estimates may err,but courts generallydo not review contractsto ensurethat performance under a contract's termsyields the consequencesthat the partiesexpected it to have. The general absence of judicial review follows from two premises:even with the benefit of hindsight,courts seldom could do betterfor the partiesthan the partiescan do for themselves; and the willingness of courts to attemptto rescue parties from baddeals reducesthe parties'incentive to write good contractsoriginally. These premisessupport foregoing judicial review of the liquidateddamage term just as they supportforegoing review of othercontract terms. Thus the ex post branchof the liquidateddamage rule should be repealed as well. Some courts have relaxed the traditionalprohibition against awarding punitive damages in ordinarycontract actions. The traditionalrule should be restored. A state-suppliedright to sue for punitive damages is similar to a contract-suppliedright to sue for a penalty; in both cases, a disappointed promisee would sue on the right if it existed. Showing that promiseesprefer the right not to exist implies both that a promiseewould reject a contractthat containeda penalty clause and would vote againsta punitivedamages regime if he could. This promisee preferenceshould control.4

4. Punitivedamage awardsare sometimesjustified on retributiveand deterrencegrounds. Retributive concerns are outside this Article's scope; they also seem irrelevantto the commercialbreaches that are its subject. Punitive damages will increase deterrencewhen the law is underenforced.For example, if some tort victims will not sue, injurersface a suboptimalincentive to behaveproperly. Letting plaintiff victims 1990] SupracompensatoryRemedies 371

Finally, parties should be permitted to contract for specific relief. The currentban on such contractscan be justified on groundssimilar to those that underlie the liquidateddamage rule. The ex ante aspect of this justification holds that a right to specific performancefunctions similarly to a right to sue on a penalty clause: both rights permitpromisees to compel inefficient perfor- mances. If promisees want this power, then they should be preventedfrom contractingfor specific relief, just as they are preventedfrom contractingfor penalties. But promisees do not want this power. Rather, promisees prefer specific performanceprimarily when their expectation cannot be monetized;in this circumstance,a promiseeneither could prove damages nor create a liquidated damageclause. Because specific relief is efficientrelative to no relief, courts should enforce specific performancecontracts. If courts were to do this, then arguablythey should police these contractsspecially, just as they police liquidated damage clauses specially. Again, the ex post justification for judicial review holds that the parties' initial belief that specific performancewas necessary may turn out to be mistaken,but the promisee might then use his contractright to specific relief to extorta supracompensatorypayment from the promisor.As is shown below, promiseesdo not use the specific performance remedyto exploit today,and would be unlikelyto use the remedyfor exploitation were they able to contractfreely for it. Hence, courts should not police specific performance contracts specially.' If specific relief clauses become enforceable,the currentspecific performancerule would not be the best default (what would be is a difficult question). PartI shows thatpromisees prefer compensatory remedies. Part II explores the normativeimplications of this preference.Part III defends the use of certain assumptionsthat underliethe argument.Part IV considersone importantand sometimes unrealistic assumption in detail. This assumption holds that promisees always sue to protect their expectation interest; in particular, promisees always detect breachesand never are deterredfrom suing by high legal costs. The former aspect of the assumptionoften is plausible because breachusually is easy to detect;the latteraspect is more questionablebecause legal costs do seem to deter some suits. Part IV makes two claims. First, promisees prefer not to solve the "collection cost problem" with penalty clauses. Second, the state should not respondto the inefficiencies that sometimes follow from underenforcementof the law with punitive damageawards. Rather, the state should encouragepromisees to sue by reducing collection costs. PartV considersthe recenttrend to awardpunitive damages for contract

recover punitive damages responds to this underenforcementconcern. Part IV below considers whether underenforcementjustifies punitive damagesin contractcontexts. The claim made in the text above is that punitive damages should not be awardedto vindicate the promisee'sinterest in realizing gains from trade. 5. Courtsdo and should police all contractsto ensure thatparties do not violate the duty of good faith in performance.An early and perceptiveargument that the partiesshould be allowed to contractfor specific performanceis in Kronman,Specific Performance,45 U. CHI.L. REV. 351 (1978). 372 The Yale Law Journal [Vol. 100: 369 breacheswhich disadvantageindividuals. This trendis particularlymanifest in two contexts, the wrongfuldenial of insurancebenefits to individualinsured6 and dismissals from employment.7Awarding punitive damagesin the former case is undesirable;awarding them in the latter case may be wise only when an unregulatedright to dismiss would create an externality,such as reducing a firm's incentive to obey the law.

I. CONTRACTINGFOR REMEDIES

Part I shows that promiseeswould reject supracompensatoryremedies in three contexts that cover much of the contractingground: (i) when parties function in competitive markets;(ii) when parties bargain (in competitive markets, there is no bargaining)and are well informed about the relevant economic parameters;(iii) when partiesbargain but are imperfectlyinformed. In the first two cases, the promisee'spreference for compensatoryrelief follows fromthe familiarresult that the expectationinterest remedy is efficient respecting the breachdecision. The expectationremedy is efficient respectingbreach because it induces the partiesto performwhen performancewould maximize theirjoint gains, and to breachotherwise.8 In cases (i) and (ii), the promise's preference for expectation damages follows from the remedy's efficiency because the promisee's share of the gains from trade is exogenously determined; hence, the promise cannot maneuverto obtain a larger share of a smallerpie. The promiseethus wants the pie to be as large as possible. Since a penalty clause would shrinkthe pie-that is, reduce the parties' joint gains from contracting-the promisee would reject it. In the thirdcase, strategicbehavior is possible.A standardresult in bargaining theory is thatuninformed parties sometimes rationally sacrifice efficiency gains in the course of attemptingto maximize their own shares. Thus, it is a separatequestion whether promisees would seek inefficient damagemeasures

6. Courts that award punitive damages for the breachof an insurancecontract usually describe the breachas a tort-the violation of the implied covenant of good faith and fair dealing. See, e.g., Egan v. Mutual of Omaha Ins. Co., 24 Cal. 3d 809, 169 Cal. Rptr. 691, 620 P.2d 141 (1979); Anderson v. ContinentalIns. Co., 85 Wis. 2d 675, 271 N.W.2d 368 (1978). These are best regardedas contractactions because the companies' conduct was a breach of contractand because similarly situated parties could respondin their contractsto the legal rule. Some courts do let promisees sue in contractitself. See, e.g., Linscott v. Rainier Nat'l Life Ins. Co., 100 Idaho 854, 606 P.2d 958 (1980); Kirk v. Safeco Ins. Co., 28 Ohio Misc. 44, 273 N.E.2d 919 (1970). 7. These courts also say that the action is in tort. See Carterv. CatamoreCo., 571 F. Supp. 94 (N.D. fll. 1983); Dare v. MontanaPetroleum Mktg. Co., 212 Mont. 274, 687 P.2d 1015 (1984); K Mart Corp. v. Ponsock, 103 Nev. 39, 732 P.2d 1364 (1987). 8. That the expectation measureis efficient respecting breachwas first proved formally by Shavell. See Shavell, Damage Measuresfor Breach of Contract, 11 BELL J. ECON.466 (1980). Sam Rea inferred from this result that "the parties to a contractare unlikely to agree ex ante to damages that exceed the expected loss." Rea, EfficiencyImplications of Penalties and LiquidatedDamages, 13 J. LEGALSTUD. 147, 159 (1984). This Article confirms and extends Rea's inferenceby showing that promisees would not seek penalty clauses in a variety of contractingcontexts. The Article also drawsdifferent normative implications than Rea does from the promisee'spreference. See inf-a notes 24-32 and accompanyingtext. 1990] SupracompensatoryRemedies 373 when bargainingunder conditions of imperfectinformation. Part I.C. shows that promisees never would seek penalty clauses in this context but may seek damage measuresthat inefficiently undercompensate.The normativeimplications of this latter preferenceare not pursuedhere.

A. The CompetitiveCase

The conclusionthat a promiseefunctioning in a competitivemarket would not purchasea contractualremedy in excess of his best estimate of the harm that breachwould impose rests on the following assumptions:(1) the promisee is not risk preferring;(2) the promisorand promiseeshare the same estimate of the breachprobability; (3) the promisor'scosts are determinedexogenously9 (this assumption implies that the promisee cannot affect the probabilityof performanceexcept throughhis choice of a contractdamage measure); (4) the promisee does not engage in post-contractualreliance;10 (5) the promisee always detects breach11and will sue if breachoccurs; (6) renegotiationafter the contractis made is costly;12 (7) if the promisorbreaches, a court cannot accuratelydetermine what the promisee'sgain from performancewould have been." It is helpful to begin by recreatingthe result that a contractrequiring the promisorto pay the expectationinterest remedy on breachwould maximizethe

9. In the analysis below, the promisorestimates performance cost when the contractis made;her actual costs are determinedby the state of the worldex post. The promisor'scosts will be higherthan herestimate if input prices unexpectedlyrise. The promisoris assumedto have no control over input prices. Breachin all the cases considered here is preferableto the promisorwhen her costs come to exceed the price by a sufficient amount. 10. This assumptionis made for convenience and does not affect the analysis. If the contractspecifies a particularsum as the promisee's damages-the situation considered here-then the promisee's legal expectation would not be a function of his postcontractreliance expenditures.Consequently, the promisee will engage in optimal reliance. The question is whetherthe sum that the contractspecifies will be penal or not. This is taken up next. The effect of the legal damage rules on the parties' reliance incentives is thoughtfully reviewed by Craswell and Rogerson. See Craswell, Performance,Reliance and One-Sided Information,18 J. LEGALSTUD. 365 (1989); Rogerson,EfficientReliance and Damage MeasuresforBreach of Contract, 15 RANDJ. ECON.39 (1984). 11. The assumptionthat breachalways is detected is plausiblein the cases consideredhere, in which breachentails the failure to supply promisedgoods or services, or to pay money when it is due, or to retain someone in employment.Undetected breaches seem primarilyto occur in certainprincipal-agent contexts, as where an employee covertly shirks.An optimalprincipal-agent contract could have a penal elementwhen one possible outputcorrelates perfectly with the shirkinginput. Such "shifting supportschemes" are not discussedhere. See E. RASMUSEN,GAMEs AND INFORMATION: AN INTRODUCTION TOGAME THEORY 148-50 (1989). Undetected breachesare briefly consideredinfra at note 67. 12. If renegotiationis costly, the parties have an incentive to write an efficient initial contract.Since an efficient damage measureexists-the expectation interest-renegotiation will not be discussed. 13. Goetz and Scott were the first to show that promisees seek liquidateddamage clauses when the expectation interest would be difficult to prove. See Goetz & Scott, LiquidatedDamages, Penalties, and the Just CompensationPrinciple: Some Notes on an EnforcementModel and a Theoryof EfficientBreach, 77 COLUM.L. REV.554 (1977). The seven assumptionslisted above apply to all the cases consideredhere unless it is otherwise stated. 374 The Yale Law Journal [Vol. 100: 369 parties'joint gains from contracting.'4If the contractis performed,the promisor, who is assumed to be a seller, would earn the price less her costs; the promiseewould earn the value he places on performanceless the price. The sum of these gains is the surplusfrom contracting.To calculate this surplus, a little notationis helpful.Let p be the product'sprice, c be the promisor'scost of producingor purchasingthe productand v be the promisee'sgross gain from performance. Then the contracting surplus is p - c (promisor gain) + v - p (promiseegain) = v - c (promiseegain less promisorcost). Whenthe gain from performancewould exceed the cost, the partiesshould perform; when the gain would be below the cost, the parties should not perform.In this latter case, breach would generate a positive gain; the promisorwould save her cost c, which exceeds the promisee'slost value v. Hence, the applicabledamage rule should induce performancewhen v > c but not otherwise. The expectationinterest remedy awardsthe promisee his expected gross gain less the price (v - p). Any damage rule bites only when the promisor's costs exceed the price (otherwisethe promisorwill perform).Under the expectation remedy,the promisorwill performwhen her loss from doing so (c - p) is less than her loss from paying damages(v - p), or when v > c; the promisor will breachwhen her performancecost would exceed the damagepayment, or when v < c. The expectationmeasure thus inducesperformance when performance would generatepositive gains, and breachotherwise. Next define a contractdamage measure as a termthat requires the promisor to pay k times the expectationinterest to the promisee upon breach,where 0 < k < 0. If k = 1, the damage measure would award the promisee his expectation; if k > 1, the damage measure would award a penal sum (k is then referredto as the "penal multiplier").When the contractcontains a damage measure,the promisorwill performwhen her cost of performanceis less than k times the expectationremedy. If k exceeds one, the promisorthus may be induced to perform although performancecost would exceed the value of performanceto the promisee.Similarly, if k is less than one, the promisormay breachtoo often. Thus, only the expectationmeasure maximizes the parties' joint gains from contracting. Promisees functioning in competitive markets would not purchase supracompensatoryremedies (contract damage measures with k exceedingone). A competitive marketis defined by the "free entry"condition. There is free entry if a new firm could enter a marketat no cost disadvantagerelative to incumbentfirms. Whenfree entry obtains,price in equilibriumequals average cost (includinga competitivereturn on the sellers' investment).If firms in the market were earning profits, new firms would enter until the profits were

14. When the parties' comparativeadvantages at reducingor insuringagainst breachare considered, some partiesfunctioning in competitivemarkets will preferdamage rules thataward less thanthe expectation interest.This idea is interestinglydiscussed by Epstein.See Epstein,Beyond Foreseeability: Consequential Damages in the Law of Contract, 18 J. LEGALSTUD 105 (1989). 1990] SupracompensatoryRemedies 375 competed away. That firms earn zero profits implies that buyers-the promisees-receive the entire surplus from contracting. Consequently,the promiseeswant the surplusto be as large as possible. The expectationinterest measure maximizes the contractingsurplus. Thus, promisees would contract for a damage measurethat just equals their expectation. It may be helpful to say a little more concerningwhy a promisee would prefer the expectation measureto a chance at a large penalty if the promisor breaches.Let pc be the price associatedwith a compensatorydamage measure dc, where dC = - pc, and let pP be the price associated with a penal damage measure dP = k(v - Vp), where k > 1. As said above, the promisor wants to performwhen price exceeds cost. Hence, a contractuallyinduced performance always is at a loss to the promisor.The greater are the damages that the promisor must pay on breach, the more often will the promisorchoose to perform(at a loss) ratherthan pay. Thus, the cost of the contractto the promisor is increasingin the damagemeasure. Since price equalscost in the competitive case, the price associated with the damage rule dP exceeds the price associated with the damage rule dc; that is, pP > pc. Next ask whetherthe promiseewould prefer a penaltyclause. Thereare two cases to consider.In the first, the promisorperforms. The promisee'snet gain underthe compensatorydamage measure dc is his gross gain minus the price (v - pc) while his net gain under the penal measure dPwould be the same gross gain minus the "penal price" (v - pP). Since the penal price pP exceeds the compensatoryprice pC,the promisee'snet gain would be lower underdP. The promisee always does worse under the penal measure when the promisor performs because the contractprice is increasingin k, the penal multiplier, while the promisee's gross gain from performance,v, is fixed. In the second case, the promisor breaches. The promisee then does better with a penalty clause because his recovery is greater.The loss to the promiseein the former case necessarilyexceeds his gain in the latter.Recall thatthe parties'-here the promisee's-surplus is maximizedunder the expectationdamage measure. This implies that the surplusmust be smallerunder any otherdamage measure; and the surpluscould be smallerwith the penal measureonly if the promisee'sloss from promisorperformance would exceed his gain from promisorbreach. 376 The Yale Law Journal [Vol. 100: 369

This conclusionis illustratedwith a simplepicture that plots the promise's expected gain from the contractas a function of the penal multiplierk:15

Figure I

Expected Gain

K K = I In Figure 1, the promisee'sexpected gain, whichis plottedon the verticalaxis, is maximized at k = 1 (damagesare compensatory)and becomes constantat the value for k that is high enough always to induce the promisorto perform. Hence, in the competitive case a promisee would not knowingly contractfor damage measuresthat would overcompensatehim relative to the expectation measure.

B. The Perfect InformationBargaining Case

Whenthere is heterogeneityin the marketfor goods or services,a promisor- seller has bargainingpower; she is supplyingsomething that is (at least slightly) unique.The promiseealso has bargainingpower because the demandfor unique items is relativelylimited. The bargainingproblem has been extensivelystudied for cases when each bargainerknows the other'spayoffs from agreementand disagreement.This study shows that the parties divide the gains from trade accordingto theirrespective discount rates and disagreementpayoffs. Respect- ing the former factor, suppose that the promisor'sdiscount rate exceeds the promisee's discount rate. Then the failure to agree promptlyimposes higher opportunitycosts on the promisorthan on the promisee;the promisor'shigher discountrate implies thatshe attachesrelatively less weight to futuregains than to presentgains. Consequently,the promisoris less willing than the promisee to reject a low offer now in the hope of getting a higherone later.The more patientplayer-the partywith the lower discountrate-thus has morebargain- ing power; he can afford to wait longer to get his price. Respecting the disagreementfactor, suppose that there is an exogenous probabilitythat the parties'bargaining process will cease-a selling seasonwill

15. FigureI is a copy of a computergraph of the promisee'sexpected gain as a function of the penal multiplierk. The graphwas obtainedby solving the promisee'smaximization problem mathematically and putting sample values of the variablesinto the solution. The math is not set out here because the intuition seems clear enough. The graph identifies the "global max"; the promisee's gain declines below the gain undercompensatory damages when the multiplierdrops below one because then the promisorbreaches too frequently. 1990] SupracompensatoryRemedies 377 end, for example. Were bargaining,to terminate,each party would receive its "disagreementpayoff," the value of the party's next best option. These disagreementpayoffs often differ. Then the party with the higher disagreement payoff has more bargainingpower. Because she is relativelyless prejudicedby the failure to agree, she can hold out for a better deal in the relationship."6 Now considerthe promisee'spreferences respecting the contract'sdamage measure.The promisee's share of the gains from trade is exogenously determined; it is a function of the parties' discount rates and disagreementpayoffs-their respectivebargaining power. A promiseewho cannotaffect the size of his share will want to maximize the size of the pie. This he can partly do becausethe gains from tradeare a function,inter alia, of the contract'sdamage measure, which the parties choose. As shown above, the damage measure influences the gains throughits effect on the promisor'sdecision whetherto perform. Consequently,the promiseewould bargainfor the expectationmeasure. Only it maximizes the contractingsurplus-the pie's size. Experimental evidence is consistent with the conclusion that promisees would seek the expectationmeasure. This evidence shows that partiesbargain to the efficient outcome when they know what that outcome is."7

C. The ImperfectInformation Bargaining Case

Parties that bargainoften do not know their adversaries'payoffs or other relevant information about them. This leads parties to engage in strategic behavior."8For example, a buyer will attemptto persuadethe seller that the buyer has a low valuationfor the object of sale because a seller who believes this will accept a lower price. When bargainingitself is costless to the parties but they have positive discountrates, strategicbehavior will delay agreement unduly;when continuingto bargainis costly, the parties sometimes may not agree althoughpositive gains from tradeexist."9 It is shown here that another inefficiency would not arise: promisees in imperfect informationbargaining environmentswould reject supracompensatoryremedies.

16. This intuitivestory respectingthe sourcesof bargainingpower is formalizedby Sutton.See Sutton, Non-CooperativeBargaining Theory:An Introduction,53 REV.ECON. STUD. 709 (1986). 17. See Harrison & Mckee, ExperimentalEvaluation of the Coase Theorem,28 J.L. & ECON.653 (1985); Hoffman & Spitzer,Experimental Effects of the Coase Theoremwith Large Bargaining Groups, 15 J. LEGALSTUD. 149 (1986). 18. For a recent, illuminatingdiscussion of strategic bargaining,see Johnston,Strategic Bargaining and the Economic Theoryof ContractDefault Rules, 100 YALEL.J. (forthcoming 1990). 19. By now, there is a large literaturerespecting bargaining.For recent discussions, see Cramton, Dynamic Bargainingwith TransactionCosts, 37 MGMT.So. (forthcoming 1991); J. Keenan & F. Wilson, Bargaining with Private Information(forthcoming 1991). 378 The Yale Law Journal [Vol. 100: 369

1. An IntuitiveStory

A seller/promisorwants to tradean object to a buyer/promisee.There is a distributionof buyer "types,"which means that potential buyersdiffer in the valuation they attach to the object. A buyer's type is just his valuation.The seller knows the distributionof buyertypes but does not know the type-the particularvaluation-of the personwith whomshe is bargaining.The sellerwill offer a contractwith two terms, a price and a liquidateddamage clause. The clause specifies a gross buyer valuation-v in the analysis above-for the object. If the seller breaches,she must pay the specified sum to the buyer as damages. The seller prefers to sell to the buyer with the highest valuation because this buyer is willing to pay the highest price. Accordingly,the seller will initially proposea contractwith a liquidateddamage clause that specifies the highestbuyer valuation (recall that the sellerknows the distributionof buyer valuations) and a correspondinglyhigh price. The price is increasingin the damage measure;that is, a contractwith a high liquidateddamage clause has a higher price than a contractwith a low liquidateddamage clause (because, as Part I.A. showed, higher damagemeasures impose higher costs on sellers, and a seller will not propose a price that is below cost). The buyercan accept the initial high price, high damagemeasure contract or reject it (buyershave positive discountrates and so are motivatedto agree ratherthan reject because rejectiondelays and thus partly dissipatesthe gains from trade).If the buyerdoes reject,the seller will proposea new contractwith a lower price and a liquidateddamage clause that correspondsto the second highest buyervaluation. This process continuesuntil the buyerfinally accepts a contract.Because the seller does not know the buyer's actual type, a buyer can accept a contractwith a liquidateddamage clause that is above his true valuation for the object of sale, equal to his true valuationor below his true valuation.The damagemeasure is penal if it is abovethe buyer'strue valuation; then the buyer will be overcompensatedin the event of breach.The question is what kind of contractthe buyer will accept. A buyerwill not accept a contractwith a damagemeasure that is above his true valuation,but may accepta contractwith a damagemeasure that is below it. Respecting the former conclusion, a buyer cannot fool the seller to the buyer'sadvantage by acceptinga contractwith a damagemeasure that exceeds his true valuation.The buyeris disadvantagedby this strategywhen the seller performs,because the buyerpays the price of a personwith a highervaluation than his own. The buyer is overcompensatedin the event of breach-he receives damagesthat exceed his truevaluation-but he has paid the (appropri- ately high) price for this opportunity.Thus, the buyergains nothingby misrep- resenting his type-by accepting a contractwith a penal damage measure. In contrast,a buyer may fool the seller to his advantageby pretendingto have a lower valuationthan he actuallyhas. Such a buyergains when the seller 1990] SupracompensatoryRemedies 379 performs, because he pays the price of a person with a lower valuationthan his own; the buyer loses when the seller breaches, because he then is undercompensated.The gain will exceed the loss when the seller has considerable bargainingpower. To see why, realize that the buyer can make positive profits in two ways: (i) by realizing the differencebetween his true valuation and the appropriateprice or (ii) by fooling the seller into charginga price that is below the price that the buyer's true valuation would warrant.When the seller has considerablebargaining power, she is able to appropriatemuch of the gain from tradeby charginga price that is close to the buyer's "revealed valuation"-the valuationthat the contract'sdamage measure specifies. In this event, the buyer'sstrategy of earningprofits by realizingthe differencebetween his true valuation and the price is relatively unattractive;that difference is small. The high valuing buyerdoes betterby "revealing"a falsely low valuation and paying the correspondinglylow price. Hence, buyerssometimes may accept contractswith undercompensatoryliquidated damage clauses. In sum, when the partiesare uninformedbargainers, promisees still will not accept contractswith penal damagemeasures, but may agree to contractswith undercompensatorydamage measures. The analysisthat generates this conclusion is somewhatartificial; in it, the seller does all the proposingwhile in real life the parties often make proposalsto each other.The logic thatgenerates the conclusion seems general,however. It is difficult to see how a promiseein an imperfectinformation environment could maximize utility by pretendingto a valuationfor the object of sale that is higherthan his own. Finally,a contract with an undercompensatorydamage measureis inefficient because the seller will breachtoo often. This inefficiency is anothercost that imperfectinformation imposes.'

2. An AnalyticalStory

20. The analysis above assumesthat both partiesknow the breachprobability. When promisees/buyers do not know it, there is a disputein the literatureas to whetherreliable promisors would offer penal damage measuresto signal that they are unlikely to breach.Less reliablepromisors could(not mimic these signals because they would have to pay off too frequently.For argumentsthat promisorswould offer penalties in some asymmetricinformation environments, see R. COOTER& T. ULEN,LAW AND ECONOMICS 295 (1988); Kornhauser,An Introductionto the EconomicAnalysis of ContractRemedies, 57 U. COLO.L. REV.683, 720-21 (1986). For an argumentagainst, see Rea, supranote 8. This disputeis irrelevanthere. The liquidated damage rule attempts to protect promisors from promisees. If a sophisticatedpromisor wants to send a penalty signal, only paternalisticconsiderations would justify a court in preventingher from doing so. 21. This section puts the analysis above in formal terms. Personsfamiliar with bargainingtheory will recognize that the text sets out a screening model, in which the uninformedparty-the seller-"screens" buyer types by proposing contracts.The novelty here is that the screening is done by proposingcontract clauses-the contract's damage measure-along with prices. Readers who find the intuitive story clear enough and who are uninterestedin the details can skip to the summarysection and Part II. An interesting screeningmodel that uses contractterms but is set in a marketrather than a bargainingcontext is Matthews & Moore,Monopoly Provision of Qualityand Warranties:An Explorationin the Theoryof Multidimensional Screening, 55 ECONOMETRICA441 (1987). 380 The Yale Law Journal [Vol. 100: 369

Two partieswant to tradean object.The seller does not know the valuation of the buyer with whom she deals but knows the distributionof buyer types. For simplicity,this distributionis assumedto have two members:v e {Ivl,vh, where 0 < v < Vh. The buyerknows the distributionof seller costs but ex post will not know the seller's cost realization.The seller can proposea contractthat is the pair {p,d), where p is the price and d is the damagemeasure. There are two damagemeasures (reflecting the two valuations):dh = vh and d, = v,. There also is a high and a low price: Ph > p,. Four contractsare possible but two would not exist in equilibrium.The seller will not proposethe contract{pl,dh} because if the buyer actuallyhas a high valuation,the seller believes that she can do better than the low price p,; thus the seller would not begin with the contract{pl,dh). The buyerwill not acceptthe contract{ph,d,) because he would lose money; knowing this, the seller will not propose that contract.Hence, attention can be restrictedto the contracts {pl,dl} and {ph,dh}. To keep the problem interesting, assume that the seller earns greater profits under the contract {p,,d,} than underthe seller's next best alternative. There are two relevant questions. First, would penalty contractsexist in equilibrium?In this story,a penaltycontract is the pair Iph,dh}when accepted by the low valuingbuyer. Second, would undercompensatory damage measures exist in equilibrium?The contract{ p,,d, I is undercompensatorywhen accepted by the high valuing buyer.The seller begins by proposingthe contract{ph,dh}; underit, she earns the maximumprofit in performancestates because she is paid the highestprice, and she is appropriatelycompensated for riskingthe high penalty.If this contractis rejected,she proposesthe contract{ pl,dl}. The two questionscan be approachedby determiningwhen (if ever) a buyerwould do better by pretendingto be a type differentthan his own. To pursue this issue, the contractprices must first be determined.The seller's price is the sum of threeelements: (i) the expected cost of performance when she does perform;(ii) the expected cost of the damage paymentwhen she breaches;and (iii) the shareof the expected surplusfrom contractingthat the seller's bargainingpower permitsher to command.The bargainingpower parameteris x where 0 < x < 1. When x = 1, the seller gets all the gains from trade.For convenience,the seller's costs are assumedto be uniformlydistribut- ed on the unit interval.22First considerthe contract {ph,dhl.The seller takes the damage measureto equal the buyer's valuation.Then the price is:

1) Ph=f cdc+(l-Vh)vh+xf (vhc)dc

22. This means that all seller cost realizationsare equally probableand that these (and the buyer's valuations)are normalizedto lie between zero and one. The argumentbelow holds for the normaldistribu- tion and any other standarddistribution. The uniformdistribution is the easiest to work with. 1990] SupracompensatoryRemedies 381

The solutions to this and for the similarly derived price p, are 2 2 2) V Vh V Ph 2 +X( 2

2 2 3) VI, VI 2 2

The first question is whethera buyer with a low valuationwill accept the contract {ph,dh}. His expected gain from doing this is

4) E(Bh ) =vh(v,-ph) +( lVh)(VhaPh)

His expected gain from rejectingthe contract{ph,dh} and acceptingthe appropriate contract {pl,dl) is

5) E(B)=vl-p

Therefore,the low valuing buyer will accept the appropriatecontract23 when 6) E(B,)>E(B,)1 h

Substitutingthe values for the prices into the inequalityand letting x = 0 (the seller has no bargainingpower), the inequalitysimplifies to

7) (2+V)>

This is satisfied for all values of v, and Vh.As x increases, the two prices increasebut the buyer's valuationis unchanged.Consequently, expression (4) and expression (5) become smaller in the same proportion.Therefore, the inequality in (6) is satisfied for all values of x that exceed zero as well; low valuing buyers would not accept the contract {ph,dh}.No buyer, that is, will pretendto have a valuation higher than his own. There will be a separatingequilibrium if the high valuing buyer would accept the contract{Phdh} that the low valuing buyerwill reject. An example shows that a high valuing buyer would reject this contract and accept the

23. The first three terms in equation (1) are the three elements of the seller's price that are described in text. In these equations, the probabilities of performance and breach can be expressed as Vh and (1 - Vh) respectively. This is because the seller's possible costs and the buyer's valuations are assumed to lie on the unit interval. The seller in this case believes that she is dealing with the high valuing buyer. Hence, the probability that the seller will perform is the probability that her costs will equal or be below the sum vh (for she must pay the damage measure vh on breach). No probabilities appear in equation (5) because when the low valuing buyer chooses the appropriatecontract, his gain is the same in every state; he receives his valuation less the price either throughperformance or throughreceiving expectation damages of v, - pi. 382 The Yale Law Journal [Vol. 100: 369 contract{pldl} when the seller has enoughbargaining power. Using the solutions above, let Vh = .8, v1= .5 and x = .4. Then the high valuing buyerhas an expected gain of .19 underthe contract {Ph,dh} and an expected gain of .20 underthe contract {p1,d1}.When the seller has enough bargainingpower, the equilibriumwill be pooling: all buyer types reject the contract {ph,dh}and accept the contract {p1,d1}. The intuition underlyingthese results is as given above: the low valuing buyer never can fool the seller to his advantageby pretendingto a valuation higherthan his own while the high valuing buyercan fool the seller in performance states. When the seller has enough bargainingpower, the latter buyer type does better by fooling the seller than by accepting the contractthat is appropriateto his type. Thus buyerswill reject contractswith penal damage measuresbut sometimeswill acceptcontracts with undercompensatorydamage measures.The lattercontracts are inefficient;when the damagemeasure is too low, the seller breachestoo often. Nevertheless, the low "pooling contract" {p1,d1}would exist in equilibrium:the seller earns greaterprofits underthe pooling contractthan she would earnwithout a deal;and acceptingthis contract is a best responsefor every buyertype when the seller has sufficientbargaining power.24Contracts sometimes are inefficientin imperfectinformation environments. Underpresent understanding, contracts will not containpenal damage measures.

D. Summary

Promiseeswould not contractfor supracompensatoryremedies when they act in competitive marketsor bargainunder conditions of full or asymmetric information.In the former two cases, promisees cannot engage in strategic behavior,but ratherearn an exogenously determinedshare of the gains from trade.Hence, promiseesprefer the contractualdamage measure that maximizes these gains. The optimal measure restricts the promisee to his expectation

24. This result should be viewed as tentative. The analysis does not specify the seller's beliefs respectingthe probabilitythat the buyer who rejects the contract{ph,dh) has a high or a low valuation,but these beliefs may matter.To see why, suppose the seller believes that such a buyer has a high valuation with probability.9. Then, when the buyerrejects, the seller's best responsemay be to offer {ph,dh) again. Since the buyer's discountrate is positive, if the seller pursuesthe strategy of offering {ph,dh)again, the high valuing buyer may do better accepting it initially ratherthan waiting. The low valuing buyer never would accept {ph,dh). Hence, particularseller beliefs respectingthe probabilitythat each type exists in the relevantpopulation together with positive discountrates could cause the pooling equilibriumto disappear. This possibility is not pursuedformally because it is enough to show here that the low valuing buyernever would accept the contract {ph,dh). The message space is very restrictedin bargainingmodels, the conventionbeing to permitthe parties to communicatetheir types only by their willingness to delay agreement.See, e.g., J. Kennan& R. Wilson, Theoriesof BargainingDelays (StanfordCenter on Conflict and NegotiationWorking Paper No. 12, 1990). Whetherparties also can communicatewith the contractclauses they are willing to propose or accept has received relatively little attention.This Article argues that the parties can communicatewith the damage measure,and such communicationwould not producepenalties. More sophisticated treatments of the subject obviously would be helpful. 1990] SupracompensatoryRemedies 383 interest.In the last case, promiseescan engage in strategicbehavior but nevertheless do not prefersupracompensatory remedies. This is because a promisee would have to pay for this preference,and so would gain nothingby pursuing it. The three situationsanalyzed here cover much of the contractingground.25 Thus the law regulating liquidated damage clauses, punitive damages and specific performance should be evaluated in light of the recognition that promisees prefer compensatoryremedies.

II. LEGALIMPLICATIONS

A. LiquidatedDamage Clauses

The ex ante branchof the liquidateddamage rule directsa courtto put itself in the position of the partieswhen the contractis made, and to ask whetherthe contract's stipulated sum exceeds a reasonableestimate of the loss that the promisee could suffer from breach. This judicial review is justified on the ground that promisees preferpenalty clauses and so must be preventedfrom obtainingthem. The ex post branchof the liquidateddamage rule directsa court to strike a liquidateddamage clause that would overcompensatethe promisee. This judicial review is justified on the ground that some ex ante reasonable estimates of the promisee'sharm will turnout to be too high. PartII.A begins with the ex ante rule. Part I implies that a penal sum would be stipulatedonly if the contracting process was unfair-for example, the promisorwas uninformed-or if the parties made an erroneousprediction respecting the harm that breachwould

25. This note briefly discusses unincludedcases. A seller may be uninformedabout buyer valuations but functionin a marketrather than a bargainingenvironment. It can be shown thatpenalty contracts would not emerge in such markets;again, a penalty contractwould specify a liquidateddamage sum that exceeds the promisee'svaluation. The seller would offer buyersa menuof contracts;different buyer types sometimes will choose the pooling contractbut no buyerwill select a penalty contract.The reason for this preference is identical to that set out above. This line of argumentwas suggested to me by RichardCraswell. A recent analysis shows that a monopolistseller may extracta penalty from buyersin orderto deter entry by other sellers; since the new entrantmust compensatea buyerfor breachinghis contractwith the monopolist,entry costs are increasingin the size of the contractpenalty for breach.Aghion & Bolton, Contractsas a Barrier to Entry,77 AM.ECON. REV. 388 (1987). Thereis no evidence thatsuch contractsare used. Also, the promisor/buyerin this analysis is made no worse off by the penalty; he will breach only when a new entrant's price is low enough to permit the buyerto pay the monopoly seller the penalty and still be at least as well off as if the buyer had purchasedunder the original contract.Since the liquidateddamage rule seeks to protectpromisors from having to pay large penalties on breach,and in this model it is actually the entrant rather than the buyer/promisorwho would pay the penalty, the model is not relevant here. Rather,the antitrustlaws should be used to ban entry deterringcontractual practices. Finally, Avery Katz suggests that promisees with marketpower may use penalty clauses to price discriminate;promisors whose demandis high may be segregatedby their willingness to acceptpenalty sanctions.The empirical significance of this insight is unclear.Also, if the price discriminationis inefficient, again it probablyis betterdealt with under the antitrustlaws. There do not seem to be other importantsituations in which promisees would prefer supracompensatoryremedies. 384 The Yale Law Journal [Vol. 100: 369 cause. The unconscionabilitydoctrine applies in the formercase.26 The genus of which the latter case is a species is generallyunregulated: except for the liquidateddamage clause itself, courts do not ask whetherthe parties' agree- mentrested on predictionsthat were objectivelyreasonable given the evidence that the parties had before them when they signed the contract.This restraint follows from the courts' correct belief that they are not as good at drafting contractsas businesspeople are. The liquidateddamage rule, however,permits enforcementonly of damage predictionsthat are reasonableex ante and so directs courts to evaluate the parties' predictions.Courts should be expected to do this job badly. Thereis evidencethat this pessimisticexpectation is plausible.Consider two federal appeals court cases that invalidatedwidely used liquidateddamage clauses that did not overcompensate.In ChandlerLeasing Division v. Florida- VanderbiltDevelopment Corp.,27 an equipmentlease requiredthe lessee to pay on default(a) fifteen percentof the equipment'scost to the lessor plus (b) the unpaidrent minus (c) the net proceedsof resale of the leased equipment.The court held that requiringthe lessee to pay fifteen percent of the equipment's cost was "anunenforceable attempt to contractfor a penaltyin excess of actual damages."In the court's view, the lessor was entitled only to the unpaidrent plus any damageto the leased propertyand retakingexpenses; these damages "were certainly capable of accurateestimation." The court added that if the lessee breached"in the lattermonths of the lease,"there would be little unpaid rentyet the lessee would have to pay the full fifteen percentof the equipment's original cost. That would be "unconscionable."28Since enforcementof the fifteen percentclause would necessarilyovercompensate the lessor, the clause was unreasonablein expectation. This decision is incorrectbecause fifteen percentof the equipment'scost almostcertainly reflected the expectedvalue of the lessor'sreversionary interest in the leased property-the probablevalue the propertywould have on the expirationof the lease. To prohibitthe lessor from collecting the value of the reversionfrom the defaultinglessee wouldproduce undercompensation. To see why, consider a simple example. A lease requiresthe promisorto make ten monthlypayments of $10 each ($100 in total), after which the promisormust

26. See, e.g., U.C.C. ? 2-302 (1989). As is well known,a contractclause will be found unconscionable if it is both procedurallyand substantivelydefective. A supracompensatorydamage measure may be substantively unconscionable.Part I implies that a promisee might contract for such a measure if the contractingprocess was procedurallyunconscionable. For example, a promisorwho could not read the contractwould not exact the full price for a penalty clause; in that event, the clause could producepositive profits for the promisee. The unconscionabilitydoctrine is sufficient to deal with such cases. Hence, the liquidateddamage rule should be analyzed on the assumptionthat the parties' contractingprocess was procedurallyfair. 27. 464 F.2d 267 (5th Cir. 1972). This and the next case apparentlyinvolved marketsrather than bargainingenvironments. 28. Id. at 270-71. The courtdid not mean thatthe contractwas unconscionableunder the unconsciona- bility doctrine, which it made no effort to apply.Rather, the court used the word "unconscionable"to mean supracompensatory. 1990] Supracompensatory Remedies 385 return the leased property, which then is expected to be worth $15 (the reversion). The promisee/lessor's expectation interest thus sums to $115 (rent plus reversion). Let the promisor make five payments and breach. The promisee resells the property and sues. The lease, suppose, authorizes the promisee to recover $50 (the unpaid rent) plus $15 (the reversion or fifteen percent clause), less the resale proceeds. Thus, the promisee receives a total sum of $115: $50 in voluntary payments plus $65 by resale and by action. The $115 sum, recall, is the promise's expectation interest so the contract ensures him full compensation, not overcompensation. The court in Chandler Leasing struck the fifteen percent clause. In the illustration here, to do that would permit the promisee to recover only $100, the promised rental payments. This undercompensates.29The fifteen percent clause thus was not a penalty but rather a necessary part of the lessor's compensatory damages. Also, because the type of liquidated damage clause involved in the case is widely used in equipment leases, the court created a precedent that, if followed, would seriously affect the leasing industry.30 The second example is equally troubling.31There the promise, A, was to process materials that the promisor, B, was to ship to it. The contract required B to ship a minimum quantity or to pay A the full processing price for material that B did not ship, up to the specified quantity. B shipped less than the minimum quantity and claimed, in the lawsuit, that the minimum quantity term created a penalty. Judge Posner agreed, holding that the clause overcompensated the promisee A on an expected basis. Because A would incur no variable processing costs for unshipped material, yet would be paid as if it had done the work, the contract would permit A to do better on breach than it would do on performance (when it would get the full price less its processing cost). This analysis was erroneous. The minimum quantity clause was functionally equivalent to a take-or-pay clause. These clauses require buyers/promisors to pay sellers/promisees the full price on a specified minimum quantity. If the promisor breaches by taking less than the stated minimum quantity, the promis-

29. The court was concerned that the promisormight breach toward the end of the lease and that enforcementof the 15%clause would then create a penalty.This is incorrectbecause the timing of breach is irrelevant.In the example in text, suppose that the promisormade $99 in paymentsand then breached. The contractwould permit the promiseeto recover$16 (unpaidrent plus reversion)while the court would permit him to recover $1. This is too little because the promisee'sexpectation interestsums to $115. 30. The Comment to ? 2A-504 of the proposednew U.C.C. Article 2A dealing with lease contracts describes clauses such as the one used in this case as "common in leasing practice."A similar case to ChandlerLeasing is Lee Oldsmobile,Inc. v. Kaiden, 32 Md. App. 556, 363 A.2d 270 (Md. Ct. Spec. App. 1976) in which a retail automobiledealer's contractauthorized it to retainthe downpaymentas liquidated damages. Such clauses are common in retail contexts. The court struckthe clause "becauseat the time the contractwas made, . . . any damages which would result from a possible futurebreach .. . would be easily ascertainable."These damages would be the difference between the contractand marketor resale prices. The dealer, however, apparentlywas a lost volume seller, whose typical suit would be for profits. These are hardto prove. The liquidateddamage clause probablyrepresented a reasonableestimate of those profits and thus should have been sustained. 31. The case is Lake River Corp. v. CarborundumCo., 769 F.2d 1284 (7th Cir. 1985). 386 The Yale Law Journal [Vol. 100: 369 ee can recover the price although it saved variable cost by not having to producethe goods. Take-or-payclauses authorizepromisees to recoverthe full price on a minimumquantity rather than the contractmarket differential on the larger amount that the parties expect will be shipped because the contract marketdifferential can be difficult to prove in court.The formulafor creating the minimumquantity is chosenso thatthe specific performancedamages under the contractequal the contractmarket differential for the larger amountthe parties expected to trade. Thus damages under take-or-payclauses are not penal.32Hence, a judge as expert in economics and as friendly to freedomof contractas JudgePosner created a precedentthat, if generalized,would outlaw an efficientpractice in severalindustries (take-or-pay clauses are used in natural gas, coal, and electricity contracts).33 These examples suggest that there may be more to fear from judicial oversightthan from poor partypredictions. Thus the branchof the liquidated damage rule that authorizesex ante judicial review should be repealed.34 The ex post branchof the liquidateddamage rule authorizesa courtto strike a stipulatedsum that turnsout to be above the promisee'sloss. This aspect of the rule may seem justifiable because a remedy that is supracompensatoryin

32. For discussion of take-or-payclauses, see Crocker& Masten,Efficient Adaptation in Long-Term Contracts:Take or Pay Provisionsfor Natural Gas, 75 AM. ECON.REv. 1083 (1985); Masten,Minimum Bill Contracts:Theory and Policy, 37 J. INDUS.ECON. 85 (1988). 33. JudgePosner recognized that the contractin this case resembleda take-or-payclause but believed that the take-or-payclause in such industriesas naturalgas could be compensatoryonly because suppliers there had a very large ratio of fixed to variablecost; the clause must overcompensate,he held, where the promiseeincurred "only a fractionof its costs beforeperformance began." This is incorrect:the clausenever overcompensateson an expected basis. A take-or-payclause may be ex post inefficient when demand collapses; then the efficient tradingquantity may be less than the contractminimum. To find the efficient tradingquantity, a court would have to know the relevantdemand and cost curves. The discussion of the ex post branch of the liquidateddamage rule shows that courts should not attemptto reconstructsuch economic variables. See text this page. Judge Posner's difficulty probablystemmed from the fact that he had made no study of liquidateddamage clauses generallyand neitherparty mentioned take-or-pay clauses during the case. That courts and lawyers seldom are industryexperts is the reason why courts generally do not review contractclauses. Courtsoften strike liquidateddamage clauses if the promisorcan breachin severalways butis required to pay the same stipulatedsum for any breach.Such clauses are consideredpenal becausethey necessarily overcompensatewhen the defaultis small. This reasoningerroneously assumes that it is costless to specify the expected loss for every type of breachthat could occur.Creating stipulated sums actuallyis costly, so parties sometimes specify sums that reflect mean or minimumexpected losses. Contractsthat do this will not overcompensatepromisees on average.An example of judicial misunderstandinghere is Stock Shop, Inc. v. Bozell & Jacobs,Inc., 481 N.Y.S.2d 269 (N.Y. Sup. Ct. 1984). The plaintiff shippeda large number of originalcolor transparenciesto defendantfor possible use by one of defendant'sclients in an advertising campaign. Specifying the value of each photographin a large batch would be expensive so the contract recited that defendant"agrees . . . that the reasonableminimum value of . . . [each individuallost] ... transparencyshall be no less than fifteen hundred($1,500) dollars."Defendant lost several photographs. The court did not dispute plaintiff's claim, in the suit for stipulateddamages, that its liquidateddamage clause was standardin the industry.Nevertheless, the clause was held to be invalid in expectationbecause the "$1,500 per photographfigures may bear no relationshipto the actual value of a photograph." 34. Rea argues that a liquidateddamage clause that is penal in expectation probablyreflects either unconscionabilityor some form of mistakeand so should be stricken.See Rea, supra note 8. This argument is unpersuasivefor the reasonsgiven in text; that is, courts should apply the unconscionabilitydoctrine if there is reason to believe that a proceduraldefect existed, and, absent such a defect, should not review contractsto correct mistakes. 1990] SupracompensatoryRemedies 387 the event could induce an inefficient performance.Courts, however, cannot easily know when a clause is supracompensatory.Promisees contract for stipulatedsums when it would be expensive or impossible for them to prove their actualloss. In addition,the existence of ex postjudicial review sometimes forces a promiseeto attemptto prove thatloss in orderto defeatthe promisor's claim that the stipulatedsum is penal in application.Because this is hardto do, a promisor'sability to force substantivereview encouragespromisors to breach when performancewould be efficient. Since promiseesstipulate sums just in orderto avoid this danger,substantive ex post review reduces the efficacy of liquidateddamage clauses. And since these clauses are no more likely than other contractterms to malfunction,ex post judicial review is no morejustifi- able here than elsewhere. Therefore,the branchof the liquidateddamage rule that authorizesthis review should be repealed also.35

B. Specific Performance

Contractsfor specific performanceare unenforceable,but courts award specific relief when damagesare inadequate.The ban on specific performance contractsis best justified in the same (unpersuasive)way as the ban on penalty clauses. There is an ex ante aspect. The specific performanceremedy can function in the same way as a penal damagemeasure; a promisorwho fails to performspecifically is subjectto severesanctions. Hence, promisees could seek specific performancecontracts in orderlater to induce promisorsto perform

35. Section 2A-504(1) of the proposed new U.C.C. Article dealing with leases eliminates ex post judicial review of liquidateddamage clauses in lease contracts.The Commentdoes not explain why, but good reasons for the reform exist. Commentatorsoccasionally claim thatliquidated damage clauses createexternalities while expectation damages do not. See Chung, On the Social Optimality of Liquidated Damage Clauses (unpublished manuscripton file with author);Rubin, Unenforceable Contracts: Penalty Clauses and SpecificPerformance, 10 J. LEGALSTUD. 237 (1981). It apparentlyfollows that courtsshould awardthe expectationin preference to enforcing a liquidateddamage clause. This conclusion assumes that courts know what the expectation is. In contrast,promisees commonly use liquidateddamage clauses when the rules that damages must be foreseeable and provablewith reasonablecertainty would otherwisepreclude recovery of the expectation. Since compensatorydamages are efficient relativeto no damages,courts should enforce liquidateddamage clauses though they may create externalities.Also, the scope of the externalityargument is unclear.The interestingChung paper, supra, for example, arguesthat the promisorand promiseewill agree to a penalty clause in orderto improvethe bargainingpower of the promisorin an ex post negotiationwith a laterbuyer, who is expected to appearin the interval between the time the contractis made and the time when it is to be performed.Such penalty clauses sometimes could preventhigher valuing buyers from purchasing,but the circumstancesunder which the argumentholds seem limited. ProfessorDaniel Friedmannrecently criticized efficient breachtheory. See Friedmann,The Efficient Breach Fallacy, 18 J. LEGALSTUD. 1 (1989). According to Professor Friedmann,giving the promisee a propertyright in a promisedperformance is more consistent with our society's general normativeviews than, and would not increase transactioncosts relative to, a rule that permittedpromisors to breachon the paymentof compensation.Professor Friedmann does not claim thatsupracompensatory damages are efficient nor does his argumentsupport general judicial review of liquidateddamage clauses. He concludes that "partiesin a contractualsetting should be left free to define the ambitof theirrights, and it is open to them to stipulatethat the promisorwill be allowed to terminatethe contractsubject to paymentof damages."Id. at 23. 388 The Yale Law Journal [Vol. 100: 369 more frequentlythan their costs and the promisees'valuations would justify, or to requirepromisors to purchasetheir freedom by makingsupracompensatory payments.It was shown above that promiseesprefer contracts that deny them these powers. Consequently,the "ex ante"justification for banningcontracts for specific relief is incorrect. If courts were to enforce these contracts,there remainsthe questionwhat should be done when the contractis silent respecting remedies. Promisees generally prefer damages when the market offers good substitutes for the promisor'sperformance. An orderfor specific relief may take years to get; in the meanwhile,the promiseecould not carry on his affairs.A promiseewould ratherpurchase a substitutepromptly and sue later for monetaryrelief.36 In contrastto this general promiseepreference, specific performanceis optimal relativeto damagesin two importantcases. In the first,the expectationmeasure is not monetizable.Hence, the partiescould neitherrely on courts to awardit nor createliquidated damage clauses. Specific performanceis efficient relative to no remedy.In the second case, renegotiationis cheap.Then specific performance can be efficient relative to damages.37Promisees also may preferspe- cific performancewhen they anticipatedifficulty in collecting money. The parties'preferences respecting specific relief thus may be too context-dependent to supportthe creationof any generaldefault. On the otherhand, some rule is necessary.Perhaps the best solutionis to make damagesthe default,in consequence of the general promisee preferencefor substitutionalrelief, but to enforce specific relief clauses.38 Therealso is an "ex post"aspect to the ban on specific performancecontracts.A promiseewith a contractright to specific performancemay laterlearn thatthe marketprovides good substitutesbut forego these in favorof threaten- ing specific performanceto exploit the promisor.Restricting specific performance to cases when the legal remedyis inadequatedeters this behavior.The concernto preventexploitation cannot justify the ban on specific performance contractsfor two reasons. First, promiseeswould not seek these contractsin orderto exploit. Promisorsknow when substitutesare convenientlyobtainable ex post. Hence, a requestfor a specific performanceclause whenit likely would be unnecessarymay excite suspicion(and a highprice). Thus exploitation could occur only when a promiseewho contractedfor the right to specific relief later discoveredthat such relief was unnecessarybut chose to demandit to extort

36. This argumentwas first made in Schwartz,The Casefor Specific Performance,89 YALE L.J. 271 (1979). 37. See Rogerson,supra note 10 (when renegotiationis costless, buyerengages in ex post relianceand buyer has bargainingpower, specific performanceremedy induces more optimal reliance than damage remedy). 38. An implicationof this solution is repeal of the rule that a promisee/buyeralways can get specific performanceof a promise to sell realty. There seems no evidence that such promiseescommonly prefer specific relief to damages. To the contrary,agreements to sell real propertyfrequently contain liquidated damage clauses. 1990] SupracompensatoryRemedies 389 a large payment. This apparentlywould seldom happen. Second, bad faith promiseesseldom could make credibleextortion threats. As said above, when substitutesexist, promiseescommonly do betterto buy them. Also, the promisor could purchasea substituteand supply it to satisfy her contractualduty. Hence, a promisee who threatensa specific performancesuit when a thick marketexists ex post is unlikely to be believed. The argumentthat promisees would not use the right to obtain specific performanceto exploit promisorsgenerates two predictions.First, requests for specific performancewill be as uncommon in jurisdictions where specific performanceis freely available as they are here. If promisees seek specific performanceonly when substitutesare unavailable,and do not seek to exploit, then the incidence of specific performancerequests should be invariantto the legal regime, so long as marketconditions are roughly the same across these regimes.This predictionis consistentwith impressionisticevidence that specific performanceis seldom sought in Europeanjurisdictions where the remedy is easy to get.39The second predictionis that promisees who litigate specific performancecases to judgmentalmost always win. Americandoctrine holds that the promisee is entitled to specific performancewhen the remedy at law is inadequate.If promiseesseek specific performancewhen marketsubstitutes clearly do not exist, ratherthan to extort promisors,then promiseeplaintiffs obviously would lack an adequatelegal remedyand so should always prevail. This predictionwas confirmedby a recentsurvey of many cases, which found that specific relief almost never is denied on the groundthat plaintiff had an adequatelegal remedy.' Promiseeshave no greaterincentive to exploit when the remedyis madeavailable by contractthan when it is madeavailable by law. In sum, specific relief clauses as well as liquidateddamage clauses should be freely enforceable.Promisees would not seek eithertype of clause to exploit promisors,nor is the dangerof ex post exploitationserious. Just as courtsapply the standarddamage rules when the contractdoes not contain a liquidated damageclause, they should applythese rules if the contractis silent respecting specific performance.The damageremedy usually is best, butthe partiesshould have discretion to substitute other remedies. If courts enforced clauses for specific relief while letting damagesbe the default,the currentrules regulating the grantingof specific relief are unnecessary.Regulation is superfluouswhen contractwill do.

39. Schwartz,supra note 36, at 277 & n.24. 40. Laycock,The Death of the IrreparableInjury Rule, 103 HARv.L. REV. 688 (1990). Makingspecific performanceclauses enforceableprobably would expand use of the remedy;parties could define damage adequacyfor themselves ratherthan rely on the courts' relatively restricteddefinition. 390 The Yale Law Journal [Vol. 100: 369

C. Punitive Damages

Promisees prefer there not be a punitive damages remedy for the same reasonthat they rejectpenal contractmeasures. The likelihoodthat the promisee would sue on eitherremedy, if it existed, wouldinduce the promisorto perform too frequently.Promisors charge extra if they have to performin loss states. These higherprices reducethe promisees'expected gains from tradebelow the gains they would make underthe expectationmeasure. Because both punitive damagesand penal contractmeasures have this effect, promiseeswant neither remedy. Thus the traditionalrule against awarding punitive damages for contractbreach should be retained.4" Punitivedamages also shouldnot be awardedfor the tortof inducingbreach of contractbecause imposing a penalty on the thirdparty has the same effect as imposingthe penaltyon the promisor.To see how, supposethat a thirdparty is consideringwhether to bid for the promisor'sperformance. The thirdparty would reducethe size of her bid by the value of the sanctionthat the promisee could impose on her. If the promisee could only sue the third party for the promisee's expectation, the third party would induce breach whenever she valued the promisor'sperformance more highly than the promiseedid. If the promiseecould sue for punitivedamages, the thirdparty would reduceher bid by the amountof the penalty.When the penaltyis high enough,the thirdparty would not attemptto induce breachalthough her valuationis highest. In these cases, the promisorwould performalthough the (opportunity)cost of performance-the thirdparty's best offer-exceeded the value of performanceto the promisee.Promisees ex ante prefernot to have damagerules thatgenerate this result.42Hence, only compensatorydamages should be awardedfor the tortof

41. An example of the cases that are disapprovedhere is HibschmanPontiac, Inc. v. Batchelor,266 Ind. 310, 362 N.E.2d 845 (1977). In Hibschman, compensatorydamages for breach of a repair and replacementwarranty on a new car were $1,500. The IndianaSupreme Court held that while the evidence would supporta verdict that "appellantmerely attemptedto fulfill its contractand to do no more than that contractrequired," the facts also would supportan inference"that Hibschman Pontiac acted tortiouslyand in willful disregardof the rights of Batchelor"by not makingrepairs and by trying"to convince Batchelor that the problems were not with the car, but ratherwith Batchelor."The jury awardedpunitive damages of $15,000, ten times actual losses. The supremecourt reducedthe penal multiplierto five and affirmed. In a more recent example, accordingto the WallStreet Journal, plaintiff sued E.F. Hutton& Co. because his brokerpursued a more risky tradingstrategy than plaintiff believed he had authorized.Plaintiff lost $27,000 which thejury awardedhim; the jury also awarded$800,000 in punitivedamages, a penal multiplier of almost thirty.See Wall St. J., Sept. 26, 1990, ? B, at 7, col. 1. 42. This analysisrests on the assumptionthat promisees prefer to profitfrom performance, not speculate on the laterappearance of parties with valuationshigher than their own. If promiseesplan to searchex post for otherparties who have valuationshigher than their own and proposethese parties to the promisor,then they must be compensatedfor this activity. Such promiseescould contractfor "penal"damage measures (or prefer a punitive damage award);the incrementof damages above the expectationactually would be compensationfor incurring search costs. See Haddock, McChesney & Spiegel, An OrdinaryEconomic Rationalefor ExtraordinaryLegal Sanctions,78 CALIF.L. REV.1, 34-36 (1990). This possibility is ruled out by the assumptionhere that promiseeswant only performance.The possibility also is consistentwith this Article's argument:Haddock, et al. show that the promiseewould contractto get his expectationplus a competitivereturn for the service of finding a highervaluing promise, not thatthe promiseewould prefer a supracompensatoryremedy when he providedno se-rch services. As an empiricalmatter, promisees seem 1990] Supracompensatory Remedies 391 inducing breach. The Pennzoil case shows that this reform would have impor- tant consequences.43 This analysis of the punitive damages remedy rests on the assumptions that the expectation interest is fully compensatory and that promisees always will sue to obtain it. Given these assumptions, punitive damages are inappropriate to vindicate the promisee's interest in realizing gains from trade. Both assumptions are questionable, however. Respecting the former, punitive damages sometimes are awarded to compensate the promisee for dignitary losses that the expectation remedy traditionally excluded." This is a weak rationale for punitive damages. The law generally refuses to compensate for emotional harm because the damages are difficult to foresee and to prove with reasonable certainty. If the unforseeability and uncertainty objections are too weak to bar the award of damages for dignitary losses, then these losses should be included in the promisee's expectation. If the objections are telling, then compensation for dignitary losses cannot be justified by calling the compensation "penal." Thus the argument against awarding punitive damages for breach of contract is unaffected by the recognition that some losses are intangible. The assumption that promisees always sue is more troublesome. Part IV considers this assumption in detail.45

III. ASSUMPTIONSOF THE ARGUMENT

The analysis in Part I rested on several assumptions. Some of these, such as that the parties do not engage in postcontractual reliance, have been discussed above. Four assumptions have not been considered. Part III discusses three of them. The fourth, that promisees always sue, is analyzed in Part IV.

to provide these services infrequently. 43. The Pennzoil case involved a suit for inducingthe breachof a mergeragreement. The Texasappeals court affirmeda judgmentagainst the inducerof $10 billion, of which $3 billion were punitive.See Texaco, Inc. v. Pennzoil Co., 729 S.W.2d 768 (Tex. Ct. App. 1987). A general criticism of the tort of inducing breach of contract is in Perlman,Interference with Contractand Other Economic Expectancies:A Clash of Tortand ContractDoctrine, 49 U. CHI.L. REV.61, 78-89 (1982). For recent defenses of the tort see BeVier,Reconsidering Inducement, 76 VA. L. REV.877 (1990); Epstein,Inducement of Breachof Contract as a Problem of OstensibleOwnership, 16 J. LEGALSTUD. 1 (1987). The argumentin text holds only that punitive damages should not be awardedagainst the tortfeasor. 44. See Chapman& Trebilcock,Punitive Damages. Divergencein Search of a Rationale, 40 ALA. L. REV.741 (1989). The Restatement(Second) now providesthat damages "for emotional disturbance" should be granted when "the breachis of such a kind that serious emotional disturbancewas a particularlylikely result." RESTATEMENT(SECOND) OF CONTRACTS ? 353 (1981). 45. The analysis here of the normativeimplications of a promiseepreference for compensatoryrelief can be extended to any remedy that may be thoughtto overcompensate,such as restitutionor the cost of completion rule. For example, a breachingpromisor is permittedto receive restitutiondamages. RESTATE- MENT(SECOND) OF CONTRACTS? 374(1) (1981). This rule is justifiable because the contraryrule would put the promisee in a betterposition than performancewould have done; that is, a legal rule which denied restitutionwould function as a penal damage measure.Since the parties preferthere not to be penalties, they would contractfor the rule awardingrestitution if the law did not supply it. Other extensions of the analysis here are omitted for brevity.A general discussion of possibly supracompensatoryremedies is in Craswell, ContractRemedies and the Theoryof Efficient Breach, 61 S. CAL.L. REV.629 (1988). 392 The Yale Law Journal [Vol. 100: 369

A. Promisees Do Not Prefer Risk

The analysis assumes that promisees do not preferrisk. A promisee that likes risk may want to treathis contractas a lottery ticket that pays his valuation minusthe price uponperformance and a multipleof this sum uponbreach. The price increasethat this preferenceinduces is the cost of the lottery ticket. Thereseems not to be evidencethat parties treat commercial contracts as lottery tickets.Rather, these partiesmake contractsto obtainthe gains from engaging in contractingbehavior. Hence, the assumptionthat the parties do not prefer risk seems innocuous.46

B. Promisees CannotAffect the Odds of Breach

Part I also assumedthat the promiseecannot affect the likelihoodthat the promisor will breach (except through the promisee's choice of a damage measure).In the analysis above, the promisor'sbreach decision is a function of her costs, which are assumed to be exogenously determined.Hence, the assumptionthat the promiseecannot affect the breachprobability is impliedby the assumptionthat costs are exogenous.Suppose instead that a promisee,after the contractis made,could increasethe promisor'scosts such thatthe promisor would preferbreach. Such a promiseemay seek a penal damagemeasure if his expected gain from inducingbreach exceeded the reductionin his gain from the contract'shigher price.47 The ex ante branchof the liquidateddamage rule permits a court to strike a clause that is so motivated.Also, a liquidatedsum that is compensatoryin the expectationmay turnout to be supracompensatory in the event. A promisee then may attempt to induce breach to collect the penalty.48The ex post branch of the liquidateddamage rule eliminates the incentiveto inducebreach because it holds thatsupracompensatory clauses are unenforceable.Thus the conclusionsreached above would have to be modified if promisees commonly could, and also would, substantially increase a promisor'scosts after the contractwere made. The assumptionthat promiseescannot affect the odds of breachis often true. For example, a buyer seldom can affect the failure rate on the seller's

46. When the parties are risk averse, their preferencefor contractremedies is influenced by their concern for risk sharing. See, e.g., Polinsky,Risk Sharing ThroughBreach of ContractRemedies, 12 J. LEGALSTUD. 427 (1983). Liquidateddamage clauses arecommonly sought by commercialpromisees. These promisees are not risk averse if they are maximizingprofits, because then their utility functions are linear in income. The analysis above goes throughif promisees are risk neutralor risk averse. 47. A promisee could profit from this strategyonly if the promisordid not anticipateit and raise the price to reflect the possibility that breachcould be caused by the promisoras well as by exogenous events. 48. Clarkson,Miller, and Muris argue that this is a serious danger.See Clarkson,Miller & Muris, LiquidatedDamages v. Penalties: Sense or Nonsense, 1978 WIs. L. REV.351. The text next argues that the danger ordinarilyis remote. In unusual cases when such strategic behavior is feasible, the parties seemingly avoid it by using third-partybonding mechanisms.See Kroeber,An AlternativeMechanism to Assure ContractualReliability, 12 J. LEGALSTUD. 333 (1983). 1990] Supracompensatory Remedies 393 assembly line. When the promisee could affect the odds of performance, the question is whether he would bargain for a penalty rule in anticipation of doing so; if not, the assumption of promisee powerlessness does not affect the conclusions reached above. Promisees would not bargain for penalty clauses in the hope of later causing contract breaches, nor would promisees often exploit promisors when stipulated sums turn out to be penal, because in both situations the costs of inducing breach seem higher than the gains. A promisee would incur three categories of cost. First, the contract price is higher when the damage rule is penal. Second, inducing breach can create reputational losses. Third, the promisee could himself be liable for damages. Respecting the third cost, inducing breach to collect liquidated damages is a violation of the legally implied covenant of good faith. A promisee who is found to have violated this covenant cannot collect liquidated damages and also is liable for the promisor's lost expectation. The gains from successfully inducing breach are unlikely to exceed these three categories of cost. The first cost is incurred for sure (because the price is higher when the contract's damage measure is higher); the last two costs are increasing in the possible gains. To see why this is so, realize that when the liquidated damages that the promisee would collect are large, the promisor is determined to perform. The promisee would have to make strenuous efforts to induce a determined promisor to breach. The more strenuous are these efforts, the more likely is it that third parties will recognize that the efforts had been made. Attempting to induce breach when the penalty is high thus disadvantages the promisee in two ways: it increases the promisee's reputational loss and it eases the promisor's task of proving in court that the promisee violated his duty of good faith. Therefore, the strategy of attempting to induce breach is least likely to pay off when it is most desirable to pursue-when the liquidated damages would be large. On the other hand, a promisee would not seek a small liquidated damage clause with the object of inducing its breach, nor would he engage in ex post strategic behavior just to collect slight damages. Thus the assumption that promisees cannot affect the likelihood of breach also is effec- tively true.

C. Both Parties Share the Same Breach ProbabilityEstimate

This assumption's relevance to the argument should first be explained. The promisee wants a damage measure that equals the value he expects performance to have. The price that the promisee will pay for such a damage measure is this value times the probability he assigns to breach. A mistake in predicting the breach probability does not imply a mistake in predicting the value of performance unless these variables are correlated. When they are not, a promisee mistaken about probabilities will be willing to pay too much for a compensatory damage clause (if he overestimates) or unwilling to pay the correct amount (if 394 The Yale Law Journal [Vol. 100: 369 he underestimates),but will not buy the wrong sum. Hence, relaxing the assumptionthat promisees know the breachprobability can affect the argument only when the two variablesof breachprobability and performancevalue are correlated.49 The possibility that this correlationexists does not supportthe case for judicial review of liquidateddamage clauses for threereasons. First, promisee mistakesmay generatefew penalties.Promisees in the correlatedcase will not seek penalties when they underestimatebreach probabilities. Promisees who overestimatewill seek penalties only sometimes, because overestimatesmay not only cause promiseesto overestimatethe value thatperformance will have but also may cause promiseesto underestimatevalue. Respecting the former possibility,suppose that the likely cause of a breachis a foreign embargo.If the embargooccurs but the promisornevertheless delivers, the productwill be worth a lot. A promiseewho overestimatesthe likelihoodof an embargothus may also overestimatethe value thatperformance would havewere an embargo to occur.Such a promiseewill (inadvertently)contract for a penalty.Respecting the latterpossibility, let the likely cause of breachbe shrinkingdemand for the product (the promisorwill get so few orders that she will do better not to produce).A promiseewho overestimatesthe likelihoodof a fall in demandwill (inadvertently)purchase an undercompensatorydamage measure, because when demand falls the productis worth little to him. In sum, overestimatesin the breachprobability will not systematicallyinduce promisees to purchasepenalty clauses while underestimatesin this probabilitycannot generate penalties. Thus promisee mistakes may producefew penalty clauses in the correlatedcase. Second, there is no reasonto think thatpromisee errors are systematically high or low. If promisee estimates of the breach probability are unbi- ased-partly random but correct on average-judicial intervention could improve matterson averageonly if courts could shrink the error term. This would happenif courts acting ex post could create damage measuresthat are materiallycloser to the measuresthat the true economic variablesimply than the parties could create ex ante.50This is unlikely. As argued in Part II.A, judicialreview of liquidateddamage clauses apparently has not been successful. Third, the breach probability and promisee value variables seem uncorrelatedmuch of the time. The factorsthat may induce a producerof ball bearings to breach, for example, seem uncorrelatedwith the value of ball bearingsto each memberof the heterogeneousset of ball bearingbuyers. When these variablesare uncorrelated,promisees will not contractfor penalties.And

49. When the parties' estimates of the breach probabilitydiffer, there is an incentive to speculate respecting the damage measure.This outcome is ruled out here by the assumptionthat the parties prefer not to speculate with liquidateddamage clauses. That assumptionseems plausible.Hence, Part Ll.C. omits discussion of speculation. 50. The adverb"materially" is addedto the sentenceabove becauselitigation is expensive. Thus minor improvementsin the accuracyof liquidateddamage clauses could not justify systematicjudicial review. 1990] Supracompensatory Remedies 395 for these three reasons, the likelihood that promisees sometimes mistake the breach probability should not affect the argument here.5'

IV. POSITIVECOLLECTION COSTS AND PUNITIVEDAMAGES

Promisees sometimes do not sue because they are uninformed about their rights or because litigation is too expensive. The former difficulty can be ameliorated by disclosure.52 The latter difficulty is discussed here. The policy issues that "rational apathy" raises have generated a large literature. To make sense of the limited aspect of the subject analyzed below, three connected issues should be noted. The first concerns the relation between the private and public spheres. Sophisticated promisees know that litigation costs are an obstacle to enforcing legal rights; their contracts sometimes deal with these costs. The state, prima facie, should provide unsophisticated promisees with the protections that sophisticated promisees obtain. If these parties reject certain solutions, such as penalties and the awarding of counsel fees to a successful plaintiff, should the state provide the rejected solutions to ordinary people? The second issue concerns whether there are economies of scale to promise keeping. When no economies of scale exist, positive collection costs are unlikely to cause inefficiency. The promisee will take the cost of enforcing the contract into account when he makes it. Regulation may be justified in this case only if the state could reduce collection costs more cheaply than private parties could.53 In the scale economy case, it is uneconomical for a promisor to keep her promise to just one or a few promisees; rather, she will keep her promise to a large set of promisees or none. Here positive collection costs can cause inefficiency. The threat that a given promisee will sue cannot itself induce the promisor to perform; rather, a promisee who sues can only guarantee himself compensation. This is because when not every promisee will sue, the promisor will breach if the costs of keeping her promise are below the gains to promisees as a group but above the costs of compensating the promisees that litigate. What kind of regulation would best ameliorate this inefficiency?

51. Mistakes in the breach probability should not influence a promisee preference for specific performancebecause that preferenceis largely a function of whethera monetaryremedy can be devised, not what the probableloss from breachwill be. 52. Federallaw and such statutesas the UniformConsumer Credit Code often requirefirms to advise consumersof their legal rights in credit and sales contracts. 53. This sentence is put tentativelybecause the questionwhen the state shouldencourage or discourage litigation in tort contexts is complex. A concise review of the issues is in Rose-Ackerman& Geistfeld,The Divergence Between Private and Social Incentives to Sue: A Commenton Shavell, Menell, and Kaplow, 16 J. LEGALSTuD. 483 (1987). Also, the incentive to sue is a function not only of litigation costs but also of the variance in expected returns.See Cornell, The Incentiveto Sue: An OptionPricing Approach,19 J. LEGALSTUD. 173 (1990). Thus focusing only on one variable can be misleading. Part IV abstractsfrom these considerations:it asks whetherpunitive damages is an appropriatelegal response when it is known (here assumed) that underenforcementof the law in consequence of high litigation costs would generate suboptimalpromise keeping on the part of firms. A complete treatmentof the relation between efficiency and incentives to sue would consider the contributionsof the scholars cited here (and others). 396 The Yale Law Journal [Vol. 100: 369

The thirdissue concernssolutions. There are two ways to inducepromisees to sue, by reducingtheir costs or by increasingtheir gains. Punitive damages increase the gains from suit. Should the state solve the "underenforcement problem"by reducingpromisee costs or by increasingpromisee gains? This Partargues that sophisticated private parties prefer cost reductionwhen economiesof scale exist and when they do not. Hence,the claim thatpromisees would not contractfor penaltiesis unaffectedwhen the assumptionthat collection costs are zero is relaxed. Part IV also arguesthat the state should prefer the cost reduction solution too when it assists unsophisticatedpromisees or respondsto underenforcementinefficiencies. This second conclusiondoes not imply that punitive damages never should be awardedfor behaviorthat is associated with a breach of contract.The analysis above showed that penal measuresare inappropriate to vindicatethe promisee'sinterest in realizinggains from trade, given that promisees sue to protect their expectation;and it is arguedbelow thatpenal measuresare an inappropriatemeans of inducingsuit. Promisorssometimes may behavemaliciously and despicablyenough to justify a retributivesanction. This behaviorseems rarein connectionwith the typical contract breach, but there is a plausible case for punitive damages when it occurs. Analyzing that case is beyond the scope of the analysis here.54

A. Private Contractsand Collection Costs

Promisees could reduce collection costs in three ways: (1) contractfor a penal damage measure out of which collection costs will be recovered;(2) requirethe promisorto pay these costs if the promisoris held liable for breach; (3) reduce disputeresolution costs cooperatively.The first responsecollapses into the second in the private context. A promisee does not want a damage measurethat would awardmore than the sum of his lost expectationplus legal costs (for the reasons given above); and a damagemeasure that would award less will not solve the promisee's problem. Hence, a promisee who wants money would not purchasea penal damagemeasure but ratherwould bargain to receive legal and related costs.

54. The AmericanLaw InstituteProject on Productand ProcessInjuries is likely to rejectthe punitive damages sanctionto cure the underenforcementproblem, if successfulplaintiffs can recoverlegal fees, but retain the sanction to deal with cases of outrageousproducer behavior. See ALI, COMPENSATIONAND LIABILITYFOR PRODUCT AND PROCESS INJURIEs, FINAL REPORT, PRELIMINARY DRAFT No. 3, vol. 2, ch. 2 (Oct. 1990). A thoughtfulargument for applyingpunitive sanctions to such conductis providedby Grady. See Grady,Punitive Damages and SubjectiveStates of Mind:A Positive EconomicTheory, 40 ALA.L. REV. 1197 (1989). There also seems to be a case for punitive damages in thin markets, when a party takes another'sentitlement to propertyor bodily integrity.This is becauserepairing the injuredparty's loss may be insufficient to deter such takings;rather, the takermust be "madewhole" too, by being forced back to its status quo ante utility level. A punitive sanction sometimes is necessary to achieve this. See Haddock, McChesney& Spiegel, supra note 42. This is an interestingargument but seems unrelatedto the concerns discussed in this part of the article. Part IV considers cases in which the parties operate in thick markets and just breachor perform. 1990] SupracompensatoryRemedies 397

Commercialparties, however, commonly reject this solution.This probably is because each party is the most efficient bearerof the risk that it will incur collection or defense costs. The magnitudeof these costs is difficult to predict at contractingtime; hence, the typical contractwould have to award a party "reasonable"attorneys' fees on breach. This arrangementwould cause the moral hazardproblems associated with any cost plus contract.Counsel would have an incentive to devote excessive resourcesto the prosecutionor defense of any case it may win becausein such cases the party'slitigation costs are less than the expected final burden;there is a positive probabilitythat legal costs can be imposed on the other side. Also, the cost of an attorneys'fee clause is hardto calculateex ante. Thus, each partyprobably would chargeher contract partnertoo much for bearing the moral hazardrisk. In addition,a party that requestsa legal fee clause may be takento signal that it would rathersue than work disputesout privately.This is a bad signal to send. Consistentwith these analyses, legal fee clauses seldom are observed in commercialcontracts although courts will enforce them.5 The thirdresponse, to reduce collection costs jointly, seems more popular. Meansto do this include:making contract clauses clear, which reduces the costs of litigating whether there was a breach and thus reduces the likelihood of breach itself;56 stipulatingdamages, which eliminates the costs of having to prove them; using performancebonds, which have relatively low collection costs; using collateral that can be privately repossessed; and incorporating arbitrationclauses. All of these methods of reducingdispute resolutioncosts are observed in practice. When contractorsare unsophisticated-consumersor unorganizedemploy- ees-they seldom will contractfor these cost-minimizingoptions. The statecan help with the same set of responseslisted above. If there are no economies of scale to promise keeping, the penalty solution collapses into the legal fee solutionhere as well. Unsophisticatedpromisees also do not wantpenal damage measuresthat awardsums above or below their costs of suit. States sometimes do awardlegal fees to successfulplaintiffs,57 and the Magnuson-MossAct and

55. Consumercontracts sometimes requirethe breachingconsumer to pay the promisee firm's legal fees. The analysis here implies that the legality of these contractsshould be rethought. 56. An interestingexample of the effect that clarity has on breachis given by the coal industry,where contractindices in the early 1980's generatedprices that were too high as measuredby currentspot market prices but buyersseldom breached.A knowledgeableanalyst claimed that there was little breach"because the terms and conditions of long term coal contractsare typically fairly explicit and the obligationsof the parties quite clear." Joskow, The Performanceof Long Term Contracts: Further Evidence From Coal Markets, 21 RANDJ. ECON.251 (1990). 57. While no state has adopted across-the-boardfee-shifting in commercialdisputes, a numberhave legislated fee-shifting in specific areas. The majordispute so affected concerns warranties,especially the automobile warranty.See ALA. CODE? 8-20-8 (1984); CAL.CIV. CODE ? 1794(d) (West 1985 & Supp. 1990); COLO.REV. STAT. ? 42-12-103(3) (1990); MD. COM.LAW CODE ANN. ? 14-1502(l)(1) (1990); N.Y. GEN.Bus. LAW? 742 (McKinney 1984 & Supp. 1990). New York also shifts fees to winning plaintiffs in consumer credit disputes. See N.Y. GEN.Bus. LAW? 458(i) (McKinney 1984 & Supp. 1990). 398 The Yale Law Journal [Vol. 100: 369

Truthin LendingLaw also let winning plaintiffs collect legal fees."8Legisla- tors may believe that the moral hazard risk is lower when individuals are plaintiffs. Finally, the state can reduce collection costs by encouragingthe mediationand arbitration of certaindisputes. These solutions also are becoming more common.59 These reformsreduce promiseelitigation costs, but the questionremains whetherthe state should increasepromisee litigation gains instead or as well. Before discussingthis, it is helpful to analyze the economies of scale case and private responses to it.

B. Private Contractingand Underenforcement

The promisor,when decidingwhether to perform,will comparethe cost of compliance with the expected cost of a damagejudgment.60 When damages are undercompensatoryor the promiseewill sue with probabilityless thanone, a promisormay not keep her promisealthough the cost of complianceis below the promisee'sloss. Sophisticatedpromisees respond to the first cause of this inefficiencyby purchasingcompensatory damage measures; they respondto the second by the cost reducingmethods just discussed.When there are economies of scale to promisekeeping, these methodscannot ensure that the promisorwill do as she agreed;they can only ensurethat the sophisticatedpromisee will be compensated. Economies of scale exist when a promisormust keep her promiseto many promisees.For example, suppose that the promisoragrees to sell a productwith a particularquality feature; it is uneconomicalto install this featureon just one item. Then the promisor,when decidingwhether to perform,will comparethe cost of installingthe featureon every item with the expected cost of promisee lawsuits. When many promisees will not sue, the costs of compliance may exceed the costs of compensatingthe promisees who litigate. Hence, when economies of scale exist, a promiseecan assurehimself of compensation,but cannot assurehimself of performance,so long as the applicabledamage measure provides only for compensation.Would such a promisee contractfor a penalty?

58. The American Law Institute Committee on Product and Process Injuries will probably also recommendone way fee-shiftingfor injuredindividuals. ALI, COMPENSATION ANDLIABILITY FOR PRODUCT ANDPROCESS INJURIES, FINAL REPORT, supra note 54, at chap. 23. 59. Twentystates have adoptedthe Magnuson-Mossapproach in connectionwith warrantiesof motor and farm vehicles. E.g., ILL.ANN. STAT.ch. 121-?h,para. 1204 (Smith-Hurd1960 & Supp. 1990). The Magnuson-MossAct specifies that if a manufactureradopts a disputeresolution procedure complying with FTC guidelines, then a consumerwith a warrantydispute cannot go to court until she has exhaustedthe arbitrationprocedure. Magnuson-Moss Warranty-FTC Improvement Act ?110, 15 U.S.C. ?23 10(a)(3)(1988). The FTC guidelines requirethat the arbitrationpanel be well fundedand independentof the manufacturer, that it charge no fee, and that it keep written records. 16 C.F.R. ?? 703.1-8 (1990). 60. For convenience, the text assumes that breachcreates no reputationallosses. 1990] SupracompensatoryRemedies 399

To understandhow a privatepenal sanctionwould work,suppose that there are n promiseeseach of whom is injuredwith probabilityq and suffersthe loss v." Then the expected total loss that all of the promisees would suffer from breachis the sum of the probabilityof harmtimes each promisee'slost value, n or ) qv. Let the promisor's cost of complying with a promise to produce a certainquality level be z per unit, and assumethat this cost is less thanthe cost of breach (l z < l qv). The promisorthus should comply with her contracts. Promiseessue upon breachwith probabilityI` where0 < I < 1. The promisor's true expected cost of breachthen is E J`qv,and she will comply when E z < E J`qv.Let F be low enough so that the promisorwould ratherbreach and pay damages of v to any promisee who sues than keep her promise. She could be made to comply if each promiseebargained for a damagemeasure that would pay him his lost value dividedby the portionof promiseeswho will sue, or v/F. Then the promisor's expected cost from breach would exceed the cost of n n n compliance (E Tq * v/F = E qv > E z) and she will keep her promise.62Be- cause the portion of promisees who sue now is assumedto be less than one, the damage measurev/F' exceeds the promisee'slost value v. Promisees would not contract for this penal damage measure for two reasons.First, a promiseewho buys a contractwith a penal sanctionis provid- ing a public good, because the promiseepays the higherprice associatedwith the sanction but would make no gain. As said above, a promiseecould assure himself of full compensationby securinga compensatorydamage measure and a reduction in his costs of collection. Even if promisees particularlywant performance, each of them may reason that if he buys the compensatory measurev at a lower price than the promisorwould charge for supplyingthe penal measurev/F, enough other promisees will purchasethe penal measure to induce the promisorto supplythe promisedquality. If so, the promiseewho buys only the damage measure v gets quality partly for free. Because each promisee has an incentive to freeloadin this way, the penal damage measure likely will not be seen. The second reason why promisees probably would not purchasepenal sanctions is that agreementswith them are not equilibriumcontracts. This is because F, the probabilitywith which promiseessue, is endogenous;it is partly a function of the damage measureitself. To see why this creates a problem, assume that each promisee knows F and let pPbe the price if the contract's damage measureD generatesa paymentof v/r and pc be the price if D = v; pP>pcbecause the price increaseswhen the promisor'sexpected damagepay-

61. Here v is the value that the promisee attaches to performance,just as in the analysis of Part I. 62. Commentatorshave observed that the underenforcementproblem can be cured by "grossingup" compensatorydamages-i.e., dividing the promisee's actual loss by the probabilityof suit. See, e.g., R. COOTERAND T. ULEN,supra note 20, at 391-96; Chapman& Trebilcock,supra note 44, at 818-19; R. Cooter,Punitive Damages For Deterrence:When and How Much?(University of CaliforniaSchool of Law at Berkeley Programin Law and Economics WorkingPaper No. 89-4, 1989). 400 The Yale Law Journal [Vol. 100: 369 ment increases.Promisors will comply with promiseswhen the proportionof active promisees-those that sue-is high enough. The idea of a compliance thresholdcan be capturedby letting F* be the proportionof active promisees that is high enough to induce the promisorto comply; the promisorperforms when 1 ? F* and breacheswhen F < F*. The optimal strategyfor a promiseewho wants performanceapparently is to purchasethe damagemeasure D = v/ITif F < P* at price pP;and to purchase the damage measure D = v if F ? F* at price pc. In the former case, an insufficient number of promisees sue to induce compliance so the promisorwill breachunless she faces the penal sanction v/I; thus our illustrativepromisee should be willing to buy this more expensive sanction. In the latter case, enough active promiseesexist so the promisorcan be inducedto comply with her promise if the illustrative promisee only buys the contractualdamage measure D = v; thus the promisee should be willing to buy only this less expensive measure. This promisee strategywould not be pursuedin equilibrium.To see why, first considerthe case when an insufficientnumber of promiseessue (F < F*). Then the strategyrequires a promiseeto purchasethe expensive penal damage measureD = v/F. The proportionof promiseesthat sues is partly a function of the damage measureitself; more promiseeswill sue when they can collect v/ITthan when they can collect only v. Suppose next that enough promisees obtainthe penal measureD = v/ITto cause the proportionof promiseesthat sue to rise above r*. Then the posited strategyrequires a promiseeto shift to the less expensive damagemeasure D = v. But if promiseesdo this, the proportion of promiseesthat sues may again fall below r*; in that event, promiseesmust return to the penal measure D = v/F. This process has no naturalstopping point-that is, it is not an equilibrium.The posited promisee strategy thus probablywould be too difficult to use in practice.Promisees could not condition theircontract choices on the proportionof litigatingpromisees-r-as the strategyrequires, because F would be cycling up and down in consequenceof the actions that promiseeswould be taking in accordancewith the strategy. This cycling problemis curablein theory if the parties play mixed strategies. To do this would requirepromisees to obtain the damage measureD = v/ITin a certainpercent of theirdeals-say a percent-and the damagemeasure D = v in (1 - x) percent of their deals (O < a < 1); and it would require promisorsto comply with theirpromises a certainpercent of the time-say 0 percent-and to breach them (1 - 0) percent of the time. Then an equilibrium exists in which the promiseeswould have no reason to vary the proportionof the cases in which they pick one or the other damage measure, given the noncomplianceprobability that the promisorschoose; and the promisorswould have no reason to vary this probabilitygiven the promisees' actions. In this equilibrium,promisees sometimes would bargain for penal clauses and promisors sometimes would keep their promises. 1990] SupracompensatoryRemedies 401

This possibilityseems moreinteresting theoretically than practically. Parties seldom are observedto use mixed strategiesin life, apparentlybecause it seems irrationalto people to shift arbitrarilybetween courses of action-i.e., to buy one damage rule and then for no apparentreason to switch and buy the other. Also, mixed strategies are complex to devise and play.63If the possibility of a mixed strategyequilibrium is rejected,the difficultyof contractingfor a penal damage measureremains.' In sum, two reasonssuggest that sophisticated promisees would not contract for a penal damage measureof the form v/ITin the economies of scale case. First, to purchasethe penal measureis to providea public good. Second, the strategyof conditioningthe magnitudeof the damagemeasure on the current proportionof promisees that litigate would not be pursuedin any plausible equilibrium.Because blatantlypenal clauses are unenforceable,this argument cannot be tested by observingactual contracts; the law ratherthan the reasons discussed here may account for the lack of penal contractresponses to the promisees'compliance concern. Nevertheless, the reasonsseem plausible.They suggest that were penal clauses lawful, promiseesstill would not bargainfor penal damage measures. Rather, promisees would respond to the underenforcementconcern in the economies of scale case as they do in the usual case. An externalbenefit of the cooperativemethods described above to reduce dispute resolutioncosts is that these methods increase the probability of contractenforcement generally and thus amelioratethe underenforcement concern.

C. The Public Response to Underenforcement

When there are no economies of scale to promisekeeping, there is no first orderefficiency case for intervention.Promisees commonly sue when the costs of suit are less than the gains, and pay less for performanceswhen the utility

63. Both criticisms are made in A. Rubenstein,Comments on the Interpretationof Game Theory (London School of Economics Discussion PaperNo. TE/88/181, 1988), who adds "thatthe use of mixed strategies is particularlyproblematic in any situationwhere the execution of mixed strategiesis costly in terms of devoting attention or time." Id at 10. When there is a cost, "a player strictly prefers to use any of the pure strategies which appearin the supportof the mixed strategy."Id. A pure strategyequilibrium is possible in the circumstancesabove if there is exogenous heterogeneityamong the promisees. Should some promisees never sue regardlessof the size of the legal damages, then 17may be fixed over relevant ranges of the parametersand the other promiseescould condition on it. There seems no reason to suppose that a nontrivialportion of promiseesis unaffectedby the damagemeasure. The analysis in the text assumes that the promisor supplies performance in "lumpy" units-she makes all products more durable or none-rather than in continuousincrements. The assumptionis made for convenience and becauseit often is true. The conclusion that a pure strategypenal equilibriumdoes not exist also holds if continuouscost curves are assumed. 64. The observationthat the enforcementprobability is endogenous-more people sue when penalties increase-has led some commentatorsto question whether antitrust damages should be trebled. See Easterbrook,Detrebling AntitrustDamages, 28 J.L. & ECON.445 (1985); Hovenkamp,Treble Damages Reform,33 ANTITRUSTBULL. 233 (1988). 402 The Yale Law Journal [Vol. 100: 369 calculus would go the other way.65The state apparentlyshould intervenein this case only when it can reducelitigation costs more cheaplythan the parties could. In the consumer case, this comparativeadvantage sometimes exists. When there are economies of scale, there is an efficiency case for regulation; promisorsmay fail to keep promises even when the aggregate gains from promisekeeping exceed the costs. The state also could respondto this case by reducing promisee litigation costs. Should it instead (or in addition)permit promisees who sue to recover the penal sanction v/I? Two considerations suggest that punitive damages should not be used. First, a public decisionmakerseldom would know the enforcementerror-the proportionof promiseeswho will not sue-especially as this proportion is partlycontext dependent.66Second, the equilibriumproblem exists here too. The parties' ability to respondto penal damagerules by alteringpurchase contractsmakes questionablethe strategyof using these rules to increasethe suit probability.67 To see why, suppose that the state requirescontracts to contain the penal damagemeasure v/r and F is caused to rise above 1*. Then promiseeswould prefer a different damage measure but the promisorsare prohibited from supplying it. The parties thus will be induced to contract out indirectly. Promisorswill supply contractsthat differ from the "old contract,"but that

65. This assumes that promiseesknow their legal rights or that the state has respondedto imperfect informationby requiringdisclosure. Awarding punitive damages apparently could not cureunderenforcement resultingfrom ignoranceof legal rights;promisees would be uninformedabout their right to recoverpunitive damages. 66. Cooter suggests that courts sometimes could infer the degree of underenforcementby comparing the ratio of the marginaldecisionmaker's precaution costs to the social harm that breachcould cause. See R. Cooter, supra note 62, at 9-10. For example, if the costs of precautionare $100 and the harm that precautionwould avoid is $200, the ratio of the two is ?h.A firm would be on the marginbetween taking precautionsand not only if its expected liability for not doing so equalled the precautioncost. In the example, given that the harm from noncomplianceis $200 the expected liability could equal the $100 precautioncost only if the enforcementerror (r in the text above) is 50%. This is the same value as the ratio betweenprecaution costs and breachcosts. Thus the enforcementerror can be inferredfrom this ratio. Cooter's method for finding the enforcementerror is imaginative,but may not help in real cases. The method requires a jury to recover ex post what would have been a correct ex ante promisorestimate of expectedharm (including nonpecuniary harm) as well as the promisor'sprecaution costs at the margin.Juries seldomcould find these facts accuratelyenough to enablethe enforcementerror to be inferredwith sufficient confidence. 67. An interestinganalysis of positive litigation costs when parties do not bargainis in Polinsky & Rubenfeld,The WelfareImplications of Costly Litigationfor the Level of Liability, 17 J. LEGALSTUD. 151 (1988). Some commentatorssuggest thatpunitive damages are an appropriatecontract law sanctionbecause breachescan go undetected.See Farber,Reassessing the EconomicEfficiency of CompensatoryDamages for Breach of Contract, 66 VA. L. REV. 1443 (1980): Sebert, Punitive and NonpecuniaryDamages in Actions Based Upon Contract:Toward Achieving the Objectiveof Full Compensation,33 UCLA L. REV. 1565 (1985). Undetectedbreaches also createan underenforcementconcern. This possibilityis not discussed here. Undetectedbreaches could exist in productor service marketsonly in connection with what in the economics of informationare called credencegoods. Somethingis a credencegood if the purchaserwould have difficulty knowing whether he needs it or not or. if he knows that he needs it, whether it was appropriatelysupplied. Examples include a subset of automobilerepairs and medical operations.Though undetected breaches in connection with credence goods are a concern, respondingto them with penal sanctionsencounters the problemsdiscussed here. For example. a publicdecisionmaker is unlikelyto know the percentageof undetectedbreaches, and so could not use a penal sanction of the form v/r. 1990] SupracompensatoryRemedies 403 promiseesprefer only becausethe statehas amendedthe old contractby adding a compulsorypenal damage measure.When rI> J7*,however, the promisees would preferthe old contractwith D = v. As a concreteexample, understrict products liability promisors may respond to a compulsory penal damage measureby making saferproducts than promiseeswant, becausethe promisor is liable to all victims and they can collect supracompensatorydamages. A contractthat is identical to the old contractand that generatesexcess safety would not exist in equilibrium.The partieswould prefer a contractthat supplies less productvariety or a largerdownpayment, etc., to restorethem to the utility level they would have been on had the punitive measurenot been imposed. Solving the underenforcementproblem by requiringa penal measure of the form v/I will producean overenforcementproblem that itself leads to contracting inefficiencies.The questionis whethera betterstrategy exists thanrequiring promisorsalways to supply the penal damage measureD = v/F. The state could use anothertype of penal measure.Three exist. Undereach of them the jury is asked whetherpunitive damagesshould be awardedgiven the applicablelegal standard.If the answeris yes, then (1) the jury has discretion over the sum; or (2) the jury must awarddamages of kv, where k > 1 and is set by statute(i.e., damagesare some multipleof actuallosses);68 or (3) the jury has discretion to award punitive damages up to a statutorilyspecified sum.69These penal measuresdo not respondto the underenforcementconcern that purportedlyjustifies the resortto punitivedamages, because none of them make the size of the sanctiona functionof the extent of underenforcement.In addition,the first penal measurecreates considerable uncertainty for promisors. Thus it seems betterall in all to respondto the likelihoodthat some promisees will not sue by reducingthe costs of suit ratherthan by increasingthe gains.70

V. TWO APPLICATIONS

A. The Insurance Cases

Courts sometimes awardpunitive damages when individualinsureds are deniedbenefits. In some of these cases, the insurancecompany promisor denied benefits because it made a mistake. In other cases, the company apparently acted undera plan. Courtsreject punitive damageswhen they believe that the

68. See, e.g., CAL.CIV. CODE ? 1794 (West 1985) (punitivedamages in warrantydisputes limited to twice compensatorydamages); CONN. GEN. STAT. ANN. ? 52-240b (West Supp. 1989) (punitive damages limited to twice compensatorydamages). The ALI Committee on Productand Process Injuriesprobably will recommend the specified multiplier solution for cases in which it believes punitive damages are appropriate. 69. See, e.g., VA. CODEANN. ? 8.01-38.1 (Supp. 1990) (punitive damages limited to $350,000). 70. In additionto these considerations,any damagemultiplier should be low when thereis a probability of legal error.See Calfee & Craswell, Deterrence and UncertainLegal Standards,2 J.L. ECON.& ORG. 279, 292-95 (1986). 404 The Yale Law Journal [Vol. 100: 369 company made an honest mistake' but authorize them when the company behavedbadly. The courts may be grantingpunitive damages too frequently. Courts award punitive damages for three reasons: to vindicate the promisee'sinterest in makinggains from trade;to inducepromisees to sue; and to punish despicablepromisor behavior. The first two reasons cannotjustify the awardof punitive damages.In some of the insurancecases, the promisor's behavior seems to have reflected the incompetenceof low level employees ratherthan actualmalice. Becauseincompetence is regrettablebut not despicable, punitive damages perhaps are awardedtoo frequentlyin the insurance context.

B. The EmploymentCases

Wrongfuldismissal cases come in threekinds. First, the dismissalis without good cause; a foremantakes an unjustifieddislike to an employee. In this case, if the dismissal is a breachof contract,the employee loses the expectationof receiving futurewages. Punitive damagesshould not be awardedin addition to this expectation;the ability to claim them would reduce the value of the employment contractjust as it reduces the value of other contracts.In the second case, the dismissalviolates a publicpolicy. For example, the employee was dischargedfor cooperatingwith a publicagency investigating the company or its customer.73The question whetherpunitive damages should be awarded in this case is complex. On the one hand,awarding them encourages employees to help enforce the law. On the otherhand, since the expected cost of punitive judgmentsis reflectedin wages, employeesas a groupare draftedin the service of the state. The appropriateresolution of this conflict is beyond the scope of this Article. In the thirdcase, the employeeis dismissedunder humiliating circumstanc- es, causing her to incur a noncommercial loss.74 Punitive damages would be inappropriatehere because there are no economies of scale to good firm behavior.The promisor'sability to dischargea particularpromisee employee with dignity is independentof its abilityto dischargeany otheremployee with

71. E.g., Garule v. Illinois Mut. Life and CasualtyCo., 152 Ariz. 600, 734 P.2d 85 (1987); California Shoppers,Inc. v. Royal Globe Ins. Co., 175 Cal. App. 3d 1, 56, 221 Cal. Rptr. 171, 201 (1985). 72. Courts affirm large punitive damage awards in these cases apparentlybecause they believe that legal fees deter insuredsfrom suing on moderateclaims, and that the companies, in consequence,are too reluctantto pay these claims. The punitive sanction is thought to encouragethe companies to pay. The companies' reluctanceto comply with their contractsis unfortunate(if it exists), but awardinglegal fees is a better solution for the reasons given. 73. E.g., Palmateerv. InternationalHarvester Co., 85 111.2d 124, 421 N.E.2d 876 (1981) (where employee was dischargedin retaliationfor supplyinginformation to local law enforcementofficials, action is in tort and punitive damages are appropriate). 74. E.g., Agis v. HowardJohnson Co., 371 Mass. 140, 355 N.E.2d 315 (1976); Harless v. First Nat'l Bank, 169 W. Va. 673, 289 S.E.2d 692 (1982). 1990] Supracompensatory Remedies 405 dignity. This third case may involve unsophisticated promisees, however. Thus here the state may wish to reduce collection costs for individual employees.75

VI. CONCLUSION

Contract Law had been consistent in its treatment of liquidated damage clauses, punitive damages, and specific performance. The courts refused to enforce contractually specified sums that appeared to be penalties; refused to permit disappointed promisees to recover penal damages; and refused to enforce contract clauses that required specific performance. Recently, this consistency has been breaking down; the liquidated damage and specific performance rules are unchanged, but there has been greater leniency in granting punitive damages to certain classes of promisees, such as insureds whose claims were wrongfully denied. This Article urges courts to be consistent on a deeper level. It shows that promisees do not want contracts that would award them greater compensation than the expectation provides. Thus courts should enforce all liquidated damage and specific performance clauses; there is no persuasive justification for the current practice of treating these contract terms specially. Promisees that reject contractual penalties also prefer not to receive punitive damages for contract breach; the rules authorizing penal awards are just implied terms in the promisees' contracts. Consequently, punitive damages should not be granted to vindicate a promisee's interest in making gains from trade. The analysis that generates these conclusions rests on several assumptions, of which the most controversial is that promisees always sue to protect their expectation. Because promisees do not always sue, there is a case for legal intervention. The state can increase the frequency of suit by reducing litigation costs or by increasing litigation gains. The latter is done by awarding punitive damages to disappointed promisees. Private parties prefer the cost reduction solution and so, it is argued, should the state. Hence, punitive damages should not be awarded to encourage parties to vindicate their contractual rights. There remains a retributive case for punitive damages when promisors behave in a despicable fashion. Such behavior constitutes a tort and so is beyond the scope of the analysis here. This Article's argument rests on the premise that because the rules criticized here presuppose a promisee preference for supracompensatory remedies, showing that that presupposition is mistaken is a sufficient justification for reform. The rules, however, seem also to be the product of a "court centered-

75. For example, California requiresthe losing defendantin a dispute involving the nonpaymentof wages or benefits to pay the plaintiff's legal fees. CAL.LABOR CODE ? 218.5 (West 1985). Punitivedamages are granted in medical malpracticecases, where patient and doctor are in a contractualrelationship. The analysis above applies to these cases, except that the retributivejustification for a punitive sanction may exist more frequentlyin the malpracticecontext because the defendantis an individualperson. See Nelson v. Gaunt, 125 Cal. App. 3d 623 (1951) (punitivedamages awarded against doctor who injectedpatient with illegal substance after having been previously arrestedfor prescribingthe substance). 406 The Yale Law Journal [Vol. 100: 369 ness" thatis at ContractLaw's core. Courtcenteredness, loosely defined,holds that courts should ensurethat performance under a fair readingof a contract's terms would not contradictthe parties' "deeper"intentions and fairness. Court centerednessproduced the famous and still influential opinion in Jacob & Youngs v. Kent.76As is well known, the contract there required plaintiff contractorto install "Reading"pipe in the defendant'shome, but the contractorinstalled a substantialamount of "Cohoes"pipe. The marketvalue of defendant's home seemed trivially affected by the substitution, but defendant'sarchitect refused to certify that the work was completedproperly. The defendantthen refused to make the final payment.Judge Cardozorelied on "considerationspartly of justice and partlyof presumableintention" to hold thatwhen the value diminutionwas small and the remedialcost large,the home owner was limited to the value diminution.This result seemed precludedby the contract,which explicitly dealt with remedies.It recited that work "which is defective or which is not fully in accordancewith . . . the specifications,in every respect,will be rejectedand is to be immediatelytorn down, removed . . . or replaced in accordance with the ... specifications, whenever discov- ered . . . ." However, "The owner shall have the option at all times to allow the defective or improperwork to stand and to receive from the Contractora sum of money equivalentto the differencein value of the work as performed and as herein specified."77The contract thus gave defendantthe power to choose between the cost of completionand diminutionin value damagemeasures, but the court reservedthis power for itself. A similarreluctance to yield controlover remediesis reflected in the rules criticized here. Liquidateddamages law permits sophisticated commercial parties to agree only to those monetarydamage measures that courtswill find "reasonable."Specific performancelaw reservesto courtsthe power to decide when specific or substitutionalrelief is appropriate.The recent expansionof punitive damage awardsreflects court centerednessof a differentkind. The underenforcementof substantiverules can be amelioratedby reducing the parties' litigation costs or increasing the litigation gains for disappointed promisees. The parties and the legislaturescan best reduce litigation costs. Courtscan increase the gains, and have actively done so. Courtcenteredness in ContractLaw extends beyond remedialissues. An example is the courts' tendency to restrict party efforts to contractout of statutoryrules that were explicitly meant to be defaults.78 Courtcenteredness as well as a mistakenview of the economicsof contract thus seems to explain the existence of the rules analyzed above. This Article arguesthat a betterunderstanding of the economicsshows thatthe commitment

76. 230 N.Y 239, 129 N.E. 889 (1921). 77. The quoted clause is set out in R. Scorr & D. LESLIE,CONTRACT LAW AND THEORY 78 (1988). 78. This tendency is analyzed in Goetz & Scott, The Limitsof ExpandedChoice: An Analysis of the InteractionsBetween Express and Implied ContractTerms, 73 CALIF.L. REv. 261 (1985). 1990] SupracompensatoryRemedies 407 to courtcenteredness as appliedto contractremedies is misplaced.Another way to put this conclusionis thatthe parties will choose appropriateremedies when left to their own devices. Similararguments have been made respectingother aspects of contract.The questionnow, it seems, is whetherdefenders of court centerednesscan show how theircommitment to it is reconcilablewith Contract Law's parallel commitmentto party autonomy.Absent this showing, one of these foundationalcommitments must be abandoned.