2y ago

164 Views

2 Downloads

1.25 MB

26 Pages

Transcription

JLEO,V9 N2 399The Role of Precedentsin RepeatedLitigationYeon-KooCheUniversityof Wisconsin, MadisonJong Goo YiShearman & Sterling1. IntroductionOne of the most distinctive characteristicsof the common law system is thatthe judiciary is an important"source of law."1While judges in the civil lawtraditionmostly interpretthe statutorylaw, common law judges can make thelaw by creating precedents (stare decisis).2 Once established, a precedentrequiresjudges in subsequent cases to follow the precedent when factualsituations are not significantlydifferent.In this article, we consider precedentsin the context of repeatedlitigationwhere a series of claimantsbring suits against a single defendanton relatedissues. Repeated litigation of this type is quite common and increasinglysignificant. In many productliability, securities fraud, antitrust,and environmental cases, a defendant, typically a large corporation, is sued over anextended period of time by numerousplaintiffs allegedly hurt by the defenBoth authors are grateful to Tim Bresnahan, lan Gale, Mitch Polinsky, Larry Samuelson,Steven Shavell, Ed Steinmueller,Joe Stiglitz, an anonymousreferee, and seminarparticipantsatKorea University and StanfordUniversityfor helpful comments. Both authors'researchon thisarticle was supportedby the John M. Olin Programin Law and Economics at StanfordLawSchool. The firstauthoralso acknowledgesfinancialsupportfrom the Centerfor EconomicPolicyResearch at StanfordUniversity.1. The U.S., England, Ireland,Canada,Australia,and New Zealandare in the common lawtradition,while most of WesternEurope, Centraland South America, Africa, and Asia are in thecivil law tradition.2. For more details on the distinction between the common law system and the civil lawsystem, see Merryman(1985).? 1993 by Oxford University Press. All rights reserved. 8756-6222/93/ 5.00

400 The Joural of Law,Economics,& Organization,V9 N2dant's unlawful conduct or defective products. In the productliability area,for instance, thousandsof claims have been filed againstproducersof asbestos, cigarettes, motor vehicles, or pharmaceuticals.Examples include theDalkon Shield, Benedectin, and Agent Orange.3Precedentscan have distinctive strategicimplicationsfor litigatingpartiesin this kind of situation.Once acase goes to trial, a precedentwill be set and have a lasting effect on successive trial outcomes. For a repeat player, then, going to trial means not onlyfacing a particularcourtdecision, but also setting a good or bad precedentforfuturecases. Recognizing this, the playerswill altertheirstrategiesin pretrialbargaining,based on theirexpectationof the precedent.Ourobjective here isto identify the ways in which precedentsaffect litigating parties' bargainingstrategies and thereby their settlementbehavior.We apply the two-periodlitigationmodel developed in Che and Yi (1990),where two plaintiffs, one in each period, file claims againsta single defendantover related issues. To do a meaningfulanalysis on settlement, we adopt anasymmetric informationmodel where the defendant does not have perfectinformationon each plaintiff's damage.We focus on two forms of decision inertiacreatedby precedents:(i) correlated decisions and (ii) correlateddamages. The term "correlateddecisions"refers to a situationwhere a decision favoringa partyin one case (positively)affects the party's chance to win in a later case. This feature of precedentsfollows from the fact that common law judges must apply the holdings of aprior case to subsequentfactually similar cases.4 Sometimes, correlationofdecisions can result from a court's applying"collateralestoppel doctrine"(or"issue exclusion"),which prohibitspartiesfrom relitigatingthe common issueonce ruled on by previous courts.5To a lesser degree, we postulatethat"correlateddamages"can be regardedas anothercharacteristicof precedentsin the repeatedlitigationcontext. Although courts are not legally bound by the previousjudgment awards, theirjudgmentsare likely to be correlatedover similarcases as they apply commonrules in assessing damages. Then, a court'sjudgmentin one case can serve as3. One notable example is Johns-ManvilleCorporation,which was a majormanufacturerofasbestosproducts.In 1982, it filed for bankruptcyunderthe pressureof the overwhelmingnumberof claims. A series of private antitrustlitigations against IBM in the late 1970s after the U.S.initial action provides anotherexample. For details, see Dungworth(1988).governmment's4. To follow the strictlegal interpretation,we must distinguishbetween the actualholding ofthe case and the ratio decidendi. The formerdenotes a precise point at issue while the latteris agenerally applicablerule of law on which the court says the holding is based. To be precise, thecourts must follow the ratio decidendi of the previouscase in reachinga particularholding in asimilarcase (Llewellyn, 1989). Therefore,the correlationbetween courtdecisions is not perfect.Also, similar as it may appear,each case involves somewhatdifferentfactualsituations. Therefore, a judge presidingover a subsequentcase has some discretionto distinguishthe case from theprevious one and fashion its holdings accordingly.5. Certainconditions must be satisfiedbefore the doctrineapplies:(i) the issue of the secondperiod must be identical with the issue decided in the prior proceeding; (ii) the issue wasnecessarily decided in the prior proceeding; (iii) the litigant had a full and fair opportunitytolitigate the issue in the prior proceeding. For details, see Friedenthal,Kane, and Miller (1985).

in RepeatedTheRoleof PrecedentsUtigation401a form of a precedent(or a referencepoint) for a repeatplayerin figuringoutthe judgment patternof courts for futurecases.Often these two types of correlationexist simultaneouslyin some areas ofrepeatedlitigation, such as asbestoscases. In this article, however,for analytical clarity, we examine the two features of precedents separately.We firstinvestigatehow correlateddecisions affect litigants'incentivesfor settlement.We show that there is some criticalthresholdlevel in the defendant'swinningprobabilityagainst the firstplaintiffsuch thatwhen his winning probabilityishigher (lower) thanthis thresholdlevel, the equilibriumsettlementratewill behigher(lower) than it would be when there is no precedent.This can be easilyexplained. When the defendantis less likely to win at trial against the firstplaintiff (relative to some thresholdwinning probability),he has an incentiveto make a larger settlement offer relative to the situation where there is noprecedent. This is because if the case goes to trial in the first period, thedefendant (with a small winning probability)will face the possibility of anunfavorableprecedentbeing set, which will hurthim in the second period. Toavoid this unwelcome consequence, the defendantwill make a higher settlement offer in the first period. This will increase the first period's settlementrate. Conversely, if the defendant'swinning probabilityis high in the firstperiod, he is more likely to risk trial, and more so as the precedentialvalueincreases.Two interestingimplicationscan be obtainedfrom this observation.First,when the partiescan influence a trial outcome throughlitigation efforts, theywill front-load their efforts to the initial (precedent-setting)stage, becausesetting a favorableprecedentis more effective than fighting against an unfavorableone alreadyset. This might explain why landmarkcases often lead tointensive campaigns by those most likely to be affected by the resultingprecedents.Second, the settlement behavior under correlateddecisions indicates thatthe defendantis more aggressive againsta plaintiffwith a low winning probability than against the one with a high winning probability.The defendant'sdifferingresponses to plaintiffsbased on their likelihood of prevailingat trialmay imply that correlateddecisions discourage nuisance suits (just as theEuropeanfee system does).Finally, when damagesare correlatedacross periods, we findthatthe initialtrial gives the defendant a valuable learning opportunity,enabling him tomake an offer better tailoredto the subsequentplaintiff. This learningopportunity may induce the defendantto risk trial more often than otherwise.Therehave been some legal studies on the collateralestoppeldoctrine,6butmost of them are concernedabout the fairness issue arising from nonmutualapplication of the doctrine. A couple of recent articles (Spurr, 1991; Note,1992) have formally analyzed the implicationsof the doctrineon the welfareof disputing parties. In particular,Note (1992) identifiedthe possibility that6. Early contributions include Currie (1957) and Polasky (1954). More recent ones areGeorge (1980), Ratliff (1988), and Schroeder(1982).

402The Journalof LavwEconomics,& Organization,V9 N2defendantsfacing a sequence of plaintiffsare extortedthroughsettlementbythe initial plaintiffswhen the collateralestoppel doctrineis appliedunfavorably against the defendants. Our approachdiffers from these previousones inthat we attemptto draw implicationson the settlementincentives, intertemporal allocation of litigation efforts, and nuisance suits.This article is organizedas follows. Section 2 brieflyreviews a one-periodlitigation problem. Section 3 introducesrepeated litigation with correlateddecisions and examines parties' settlementbehavior,and Section 4 exploresfurtherimplicationsof correlateddecisions. In Section 5, we providea numerical example to illustrateour findings. Finally, Section 6 examines correlateddamages.2. A Review of One-Period LitigationIn this section, we review a one-periodproblem. This review provides someuseful resultsthat areused in our analysison precedents.Also, the one-periodproblem later serves as a point of contrast, because, absent precedents, thetwo-period model is equivalent to a replication of two one-period games.The partiesare risk neutral,and the Americanfee system underwhich eachpartybearshis own litigationcosts is adopted.The following notationis usedthroughout:w plaintiff's damage, distributedover [0,w,]F(w) a twice differentiabledistributionfunction of w with a positivedensityf.s the defendant'ssettlementoffer.c the plaintiff's litigation costs.d the defendant'slitigation costs.p the plaintiff's winning probability.A litigation cycle consists of three stages. In the first stage, the plaintiffisprivately informed about w, while the defendantknows only its probabilitydistribution,F(w). In the second stage, the defendantmakes a settlementoffers on a take-it-or-leave-itbasis. In the last stage, the game ends if the plaintiffaccepts the offer, but the case goes to trial if the plaintiffrejects it. This oneperiod problem is similar to Bebchuk (1984) in that pretrialbargaininginvolves an uninformedparty making an offer to an informedparty.We use backwardinductionto solve for each party'soptimaldecision rule.7In the last stage, the plaintiffwould accept an offer if and only if it is no lessthan what she expects to gain from trial. That is, an optimal decision rule forplaintiff, [P], with damage w is[P]Accept s( 0),if and only if (iff) s - pw - c.87. The equilibriumconcept we use is that of Perfect Bayesian Equilibrium.8. We restrictattentionto nonnegativeoffers because the plaintiffcan always refuse a negative offer and eithergo to trial or dropthe case (when the expected net trialrecoveryis negative).

in RepeatedLitigation403TheRoleof Precedentss cdenote a threshold-typeplaintiff, who is indifferentbeLet w(s) Ptween settling and going to trial at the given s. In the second stage, thedefendant makes a settlement offer s to minimize the expected loss. Sincethere is a one-to-one correnspendencebetween s and the thresholdtype w(s),without loss of generality, the defendant can be regarded as picking thethresholdtype w w(s). It is convenient to work with this thresholdlevelratherthan the settlementoffer, since the equilibriumvalue of the thresholdlevel will provide an intuitive measureof the likelihood of settlement. (Thesettlement likelihood will be simply F(w), the probabilitythat the plaintiffhas lower judgmentthan the thresholdtype.)9 Let s(w) w,p- c denote theminimumoffer needed to induce the thresholdtype to settle. Then, the defendant's decision rule, [D], can be expressed as[D]L*(p) L(L(p) v c/pi ps(w)F(W) d[l - F(w)]I wdF(w).The defendant'sexpected loss consists of three elements:the expected settlement payment, the expected litigation costs, and the expected judgment attrial.Let v*(p) denote a solution to [D]. Then, the plaintiff'sexpected utility inthe first stage is given byU*(p)L*(p) - [1 - F(w*(p))][c d],which is the expected loss L(p) less the expected litigation costs borne byboth parties.The first-ordercondition for an interioroptimumof [D] is given by- c dH(w*) where H F-is the inverse hazardrate. Throughoutthis article, in consider-ing the defendant's optimization problems, we restrict attentionto interiorsolutions.109. A settlementoffer can be a misleading indicatorfor the settlementprobability.When theplaintiff's winning probabilityincreases, in equilibriumthe defendantmay increasehis offer, andyet the settlementrate can go down, since the plaintiffnow demands more.10. This is just to simplify our exposition and does not change the qualitativeresults of thisarticle. As can be easily checked, incorporatingcomer solutionsjust changes strictinequalitiestoweak inequalitiesin all the subsequentpropositions.

404 The Joural of Law,Economics,& Organization,V9 N2This first-orderconditionis similarto the one obtainedin Bebchuk(1984) andThe defendantoptimallybalancesbenefitsand costshas a similarinterpretation:of increasingthe settlementoffer. An incrementalincrease in the settlementoffer, say by ds, increasesthe defendant'sloss for the settlementrange [whichaccountsfor the whole inframarginaltypes whose measureis F(w(s))]. ButF(w(s)) and enit also increases the settlement rate byf(w(s)) dsdspables the defendantto save the litigation costs c d at the margin. At theoptimum, these marginalbenefits must equal the marginalcosts. This firstordercondition is necessary for the offer to be optimal. It is also sufficientifthe inverse hazardrate is strictlyincreasing.12 Roughly speaking, this impliesthat the marginalcost is increasingin the settlementoffer. This assumptionislabeled as [A1] and used throughoutthe article.The following lemma reportssome importantcomparativestaticresultsthatwe will refer to in later sections. All the proofs, includingthat of Lemma 1,are relegated to the Appendix.Lemma1. Suppose [D] has a unique solution for p, then (i) wi* is strictlydecreasingin p; (ii) L* is strictly increasingin p.The second statementimplies thatthe defendant'swelfare is reducedwhenthe trialprospectbecomes less favorable.The argumentfor the firststatementis less trivial. It implies that settlement is less likely the more likely is theplaintiff to win the trial. Intuitively, with higher p, the defendant finds itharderto get his offer acceptedbecause the plaintiffdemandsmore to forgo atrial option (which has become more favorable). Therefore, at the samethresholdlevel w the marginalbenefits from increasings are reduced, whileits marginalcosts remainthe same. Thus, the equilibriumsettlementrategoesdown.133. Settlement Behavior under Correlated DecisionsIn this section, we extend the one-periodmodel to study the effect of correlated decisions. In this extension, the litigationcycle describedin the previoussection is repeatedtwice. To model correlateddecisions, we assume that thesecond plaintiff's winning probabilitydepends on the outcome of the firstcase: Her winning probabilityincreases (decreases) when the first plaintiff11. The reasonthatthe defendantsaves the whole litigationcost (notjust his portiond) throughsettlementis because, being the firstmover in the bargaininggame, he extractsall the surplustheplaintiffreceives from settlement.Althoughthis particularresultappearsmodel-specific, it is notessential to the centralpoints we make in this paper.12. This condition roughly says that the density functionft() does not grow too fast. It issatisfied with most of the well-known distributionfunctions, includingthe uniform,exponential,and normaldistributions.13. Whether the defendantwould want to raise his settlement offer as p increases dependson the elasticity e, of H with respectto w. The equilibriumsettlementoffer will increasein p iffeCHS1.

inRepeatedTheRoleof PrecedentsLitigation405wins (loses), while it remainsunchangedwhen the firstcase is settled. Let pl,P2 denote the first and second plaintiff's winning probabilities,respectively.Formally,then, P2 is determinedin the following way:Ps P,P2jPW P SE,PL P - Edif the first case is settled ("S");if the first plaintiff wins trial ("W");if the first plaintiff loses trial ("L"),where p and e are such that 0 p - Ed P Ep 1. From now on, allthe other variableswill be similarly subscriptizedby 1, 2, S, W, and L. Herethe parametersEpand ed measurethe precedentialeffect of the firsttrialon thesecond case. Specifically, ep denotes an added winning probabilityfor thesecond plaintiff when the first plaintiff wins trial, which may result from anapplicationof the so-called rine, the defendantis prohibitedfromrelitigatingan issue decided againsthim in an earliertrial. Similarly,Ed can measurethe likelihoodof the application of the "defensive collateral estoppel doctrine," which prohibits thesecond plaintifffrom relitigatingan issue decided againstthe firstplaintiff.Inmost of this article, we take the precedentialeffect to be "mutual"(i.e., p,Ed 0) and, when simplicity serves exposition, "symmetric"(i.e., Ep Ed ).However, as a thought experiment-and, more importantly,as reflection ofcurrent debates14-we also consider extreme cases where the precedentialeffect is "nonmutual"(i.e., Ed O,Ep 0; or Ed O,Ep 0).To analyze the effects of correlateddecisions, we use backwardinductionby consideringthe second-periodproblemfirst. In the second period, partiesface the same problems(and thus adoptthe same decision rules, [P] and [D])as in the one-period game, except for the fact that they now inheritdifferenttrialprospects (winning probabilities)dependingon the first-periodoutcome.Therefore, the defendant'soptimal offer in the second periodis characterizedas the following first-orderconditions:C dH(Wi) c dPiwhere i W, L, or S. Let Li denote the defendant's second-period lossfollowing the first-periodoutcome i for i W, L, S. Then, we obtain thefollowing result by applying Lemma 1.Corollary1. If the precedentialeffect is mutual(Ep,ed 0), then (i) L WS v\W; and (ii) LL Ls Lv. On the other hand, if there is no prodefendantprecedentialvalue (Ed 0), then WL vS and LL Ls, whileif there is no pro-plaintiffprecedentialvalue (Ep 0), then w's w andLs L,.14. For discussion of currentdebates on the mutualityof collateralestoppel doctrine, see theconcluding remarksin Section 7.

406 The Journalof Law,Economics,& Organization,V9 N2Note that the second plaintiff reacts to a given settlementoffer differently.In the case of the mutualprecedentialeffect, she is less willing to settle whenthe firstplaintiff wins trial (i.e., demandsmore to settle) and more willing tosettle when the first plaintiff loses. Her winning probability,however, willremainunaffectedif the firstcase is settled. Consequently,the settlementrateis the highest, lowest, and in between in the second periodif the firstplaintiffloses, wins, and settles, respectively.Let Lr and UT denote the defendant'sexpected second-periodloss and thesecond plaintiff's expected utility when the first case is tried. Then, Lr PiLw (1 - Pl)LL, and UT PIUW (1 - P)UL, where Li and Ui aredefined in the same way as L* and U* with p replacedby pi.Now, we are in a position to discuss the first-periodproblem.Since the firstplaintiff is a short-runplayer, her optimal decision is the same as in the oneperiod game (i.e., [P]). But the defendantis not myopic. He takes accountofthe long-termconsequences when determiningthe first-periodsettlementoffer. Formally,the defendant'sproblem is representedby[Dl]min LI(1,pI1) [s(w1) Ls]F(,1)x [Lr d] p [1 - F(w)]wdF(w).This ex ante loss functionhas the second-periodloss termsLs andLq addedtothe one-period loss function in [D]; the defendant, when making an offer,takes its second-periodconsequences into consideration.The first-ordercondition for [D1] isH( d L -LPiComparedwith the first-orderconditionfor [D], the only differencehere is theadditionalterm, LT-- Ls, in the right-handside. This termrepresentsthe costof setting a precedent.When it is positive, the precedentis unfavorableto thedefendant; when negative, it is favorable. The cost of setting a precedentdepends on the first-periodwinning probability.It follows from Corollary1and the definitionof Lr that, when the precedentialeffect is mutual,precedentis unfavorable(favorable)if P2 is sufficientlyhigh (low).Proposition 1. When the precedentialeffect is mutual (p,ed 0), thereexists a thresholdprobabilityp1E (O, 1) such thatthe first-periodequilibriumsettlement offer and the settlementrate are higher with correlateddecisionsthan without, if and only if the first plaintiff's winning probabilityis higherthan the threshold level, p,. With nonmutual offensive collateral estoppel

TheRoleof PrecedentsinRepeatedUitigation407(ep 0, ed 0), 1I 0, while PI 1 with nonmutualdefensive collateralestoppel (ep O,d 0).The intuition behind the proposition is transparentfrom the above argument. Since with the mutual precedentialeffect the first plaintiff's winningprobabilityp, coincides with the likelihood of an unfavorableprecedentbeingset for the defendant,ifp1 is sufficientlyhigh (low), the defendantis willing toraise (lower) his offer to avoid (face) trial. If the precedentis nonmutual,thispropositionpredicts a unilateralresult. Under the offensive collateralestoppel, any precedentis unfavorableto the defendant(LT Ls), so the defendanttries to avoid trial regardlessof his winning probabilityrelative to the oneperiod problem. By the same token, underthe nonmutualdefensive collateralestoppel, the defendantis eager to go to court to establish a favorableprecedent for the second period regardlessof Pl.We next examine how changes in the magnitude of precedential effectinfluencethe parties'settlementincentives. To this end, we againfocus on thecost of precedentsetting, Lr - Ls. Since the increase in pro-plaintiffprecedencial value, ep, makes a pro-plaintiffprecedent more unfavorableto thedefendant without affecting a pro-defendantprecedent, the cost, Lr - Ls,increases with Ep, while the opposite is true for an increase in the prodefendantprecedentialvalue Ed.The following propositionfollows from thisobservation.Proposition 2. (i) The er are increasing(decreasing)in e, (Ed).Suppose the precedentialeffect issymmetric(i.e., Ep ed e); then (ii) the first-periodequilibriumsettlementrate and offer are increasing(decreasing)in the simultaneousincreasein Ep Ed e if P1 is sufficiently high (low); (iii) P 1 for all e 0; and (iv) Pincreases with e.The second result is not obvious because it refers to the simultaneousincrease in pro-plaintiffand pro-defendantprecedentialvalues. The intuitionbehindthe resultcan be explainedas follows. Whenpi is sufficientlyhigh (thecase with sufficientlylow pi is analogous),the defendantputs more weight tothe pro-plaintiffprecedentialvalue, Ep,even if both ep and Edincreaseby thesame magnitude,since the formeris more likely to be realized. Cases (iii) and(iv) concern the behavior of the threshold probability.That the thresholdprobabilityis greaterthan one-half may be surprisinggiven the symmetryofthe precedentialeffect. The implied bias towardtrialis due to the fact that thedefendantcan exploit the option value associated with trial.15Depending on15. It is useful to interpretLr as a lotteryof takingLw 2L*(p e) andL1r L*(p - e) withprobabilityp, and (1 - pi), respectively.Whenp, i, Lr representsan actuariallyfair lotteryfora certain loss expected from settlement (LS L*(p)). Since L* is concave in p, the defendantprefers the actuariallyfair lottery of losses to a certain loss.

408The Journalof Law.Economics,& Organization,V9 N2the firsttrialoutcome, the defendantwill face differentprospectsto respondinthe second period. This creates the option value for the defendant:Withvariable winning probabilities, he can tailor his settlement offer on eachcontingency and can do better than when faced with a fixed winning probability. The intuitionfor the last result is similar.4. Further Implications of Correlated Decisions4.1 Front-loading of Litigation EffortsWhen an issue before the court raises broadpublic interests, there are oftenintensive efforts on both sides in the initial precedent-settingstage. This isespecially true in the so-called landmarkcases, which have profoundimpacton futurecases. For example, Roe v. Wade,the 1973 SupremeCourtdecisionthat legalized abortion,and Brown v. Board of Education, the 1954 SupremeCourtdecision that found segregationin public schools unconstitutional,aretwo such frequentlycited cases. As we have witnessedrecently,the controversy over a woman's constitutionalright to abortionhas drawnnationalattention, and people on both sides of the issue have staged aggressive campaignsto win the court's favor. The underlyingforce that drives these efforts is theconcern over precedents.To explore this issue within our framework,let us now suppose that eachpartyof litigationcan influencethe winningprobabilityby undertakingeffortsin the two periods. By examininghow effortin each periodaffects the welfareof each party, we can study the relative effectiveness of effort in one periodover effort in anotherperiod. First, to measure the defendant'sfirst-periodincentive for litigation effort, we compute the total derivativeof the indirectutility function in [Dl] with respect to pi. This shows the extent to which agiven increase in the winning probabilityimproves the defendant'swelfare.dL, aLwd- F(w))(LdLpp (1- LL) wdF F(wl)WI,Reducingp, has both long-termand short-termeffects. The firsttermof theright-handside representsthe long-termeffect, and the remainingtwo termsrepresentthe short-termeffect. The short-termeffect capturesa directbenefitthe defendantreceives by improving his winning chance in the first period.The long-term effect measures the expected benefit of setting a favorableprecedent.Similarly, the defendant'ssecond-periodincentive for litigation effort canbe measuredbydL (p)dpwdF F(i)i,dependingon the first-periodoutcome i W, L, and S. Notice thatthe longtermeffect is absentin the second periodsince the defendanthas no preceden-

TheRoleof PrecedentsinRepeatedUtigation409tial effect to worry about. If these short-termeffect terms are not muchdifferent, the first-periodincentive will be greater than the second-periodincentive, due to the precedential effect. Thus, we have the followingproposition.Proposition 3. If either the litigation costs c d are sufficiently low, ordLp) for i dLl(p)dL1(p)W, L, andH'(-) K for sufficiently large K, then dLp) dpdpIS; that is, the first-periodeffort is more effective than the second-periodeffort. In addition, if H'(-) K for sufficientlylargeK, the relativeeffectiveness of the first-periodeffort over the second-periodeffort, measuredby theirdd(p(p) increasesas the precedential(/marginalrate of substitution,value, ep or ed, increases.Underthe identifiedcircumstances,this propositionconfirmsa familiarnotion thatit is moreimportantto set a favorableprecedentthanto fightagainsttheunfavorableone alreadyset. That an increasedprecedentialeffect makes thefirst-periodeffort relatively more importantis also intuitive. The immediateimplicationof this propositionis that the defendantis likely to front-loadhiseffort to the initial precedent-settingstage. This will be indeed the case if thedefendant faces a standard symmetric, increasing disutility functionD(-Pl,-P2), associated with his winning probabilityin each period. If, in 0)addition, the two period efforts are substitutes (i.e.,(-p ))o(-P2)(-p\/a(-P1the front-loadingeffect is likely to be acute, since the defendantwill try todivert this effort away from the second period into the first period.16Several remarks are in order. First, the first condition has an intuitiveexplanation. In our model, the precedentialeffect arises only throughtrial.Since low litigation costs make trial more likely, this implies that setting agood precedentbecomes more importantwhen the litigationcosts are smaller.Second, the hypotheses of the propositionare sufficient conditions.'7 Thus,the results may hold even when the conditions are not met. (We conjecturethis to be the case for a wide range of parametervalues.) Finally, the resultofthe propositiondoes not depend on whetherthe precedentialeffect is mutualor not. This is because the long-term incentive for the first-periodeffortdepends only on Lw - LL, the second-periodwelfare differencein winningand losing.Given the way our model is structured,the front-loadingeffect does notapply to the plaintiffs. They are essentially myopic and lack the long-term16. The substitutes case is most relevant when the defendant faces a budget constraint infinancinghis litigation efforts. Under a fixed budget, for example, the first-periodeffort can beincreased only at the expense of the second-periodeffort.17. The conditionregardingH(-) is not intuitive. Its purposeis to ensurethatan indirecteffectthroughthe change in the settlementrates does not dominate the direct effect.

&Organization,V9N2410 TheJournalof Lavw,Economics,effort incentives. However, throughmutuallybeneficial contracting,the firstplaintiffcan be inducedto internalizethe ex

poral allocation of litigation efforts, and nuisance suits. This article is organized as follows. Section 2 briefly reviews a one-period litigation problem. Section 3 introduces repeated litigation with correlated decisions and examines parties' settlement behavior, and Section 4 explores further implications of correlated decisions.

Related Documents: