W

La

b

a

ARRA

JKK

KLFPC

1

lrcprsbp

cmc

ced

n

R(G

0d

International Review of Law and Economics 32 (2012) 233– 241

Contents lists available at SciVerse ScienceDirect

International Review of Law and Economics

hy not adopt a loser-pays-all rule in criminal litigation?

uciana Echazua, Nuno Garoupab,∗

Clarkson University, School of Business, Potsdam, NY 13699, United StatesUniversity of Illinois College of Law, 504 East Pennsylvania Avenue, Champaign, IL 61820, United States

r t i c l e i n f o

rticle history:eceived 11 January 2011eceived in revised form 17 January 2012ccepted 18 January 2012

EL classification:1

a b s t r a c t

In this paper we consider the potential effects that the application of a loser-pays-all rule may have oncriminal litigation, including the decision to prosecute, criminal deterrence and legal error. We find thatthe effects of fee shifting on deterrence and on miscarriage of justice go in opposite directions. We alsolook at the effects of this rule on the rate of settlements (plea-bargaining) and when one party is wealth-constrained. We apply the insights of our model to current policy discussions such as the use of RICOproceedings and the financing of enforcement authorities in the United States.

4

eywords:oser-pays-all ruleee shiftingrosecutor

© 2012 Elsevier Inc. All rights reserved.

dant to pay for the costs of prosecution for an amount consideredreasonable and appropriate.4 In the case of a guilty plea, the courtshould adjust the order to take into account that the expenses of the

riminal litigation

. Introduction

Many common law legal systems use loser-pays-all rules in civilitigation. The economic literature refers to the English or the Britishule when discussing the situation where the party that loses a civilase pays for the legal fees and the costs incurred by the prevailingarty. Most civil law jurisdictions also use some form of the Englishule (in some, the fees paid to the lawyers are excluded from thehifting rule or only a percentage of the total costs can be reim-ursed). The opposing example is the American rule where eacharty pays their own costs.

The economic literature on the topic is by now extensive.1 Theontroversial issues include the extent to which shifting costs pro-ote settlement, enhance civil litigation, favor more meritorious

laims, or decrease the number of nuisance lawsuits.However, loser-pays-all rules are generally not used in criminal

ases. Usually public prosecution bears the cost of its prosecutorialfforts and defendants bear the cost of their defense. A criminalefendant brought to trial and found not guilty is typically not

∗ Corresponding author.E-mail addresses: Lechazu@clarkson.edu (L. Echazu), ngaroupa@law.illinois.edu,

garoupa@law.uiuc.edu (N. Garoupa).1 See, among others, Shavell (1982), Bebchuk (1984), Katz (1987), Cooter andubinfeld (1989), Hause (1989), Snyder and Hughes (1990), Polinsky and Rubinfeld1996, 1998), van Wijck and van Velthoven (2000), Hylton (2002), Spier (2007),aroupa (2009b), and Katz and Sanchirico (2010).

144-8188/$ – see front matter © 2012 Elsevier Inc. All rights reserved.oi:10.1016/j.irle.2012.01.004

compensated for the defense costs incurred.2 Likewise, prosecu-torial expenses are not compensated by guilty defendants.

There is some evidence that the loser-pays-all rule can beapplied to criminal litigation. In Britain, for instance, the recoveryof costs is regulated by the Prosecution of Offences Act 1985. Sec-tion 16 allows a court order to compensate for defense costs (out ofgovernment funds) with a strong presumption that such order willbe awarded to favor a defendant who has been acquitted (unlessthe defense is funded by the government itself).3 Section 17 regu-lates the recovery of costs for private prosecutors and Section 18for public prosecutors. The court can order the convicted defen-

2 Generally speaking, a defendant who cannot afford a lawyer has a right to oneprovided by the state. In many jurisdictions, the prosecutor is subject to a rule ofmandatory disclosure of evidence. These two rules might be regarded as forms ofshifting costs from defendants to prosecutors. However, they are not conditionedon the defendant being acquitted.

3 We follow Hungerford-Welch (2004), at chapter 13, here. In fact, unless thedefendant brought suspicion on himself or the defense is publicly funded, it isexpected that the defense costs are borne by the government when a defendantis acquitted. The fact that the prosecution has acted properly or that the defen-dant was acquitted on a technicality does not allow the court to refuse a cost orderfavorable to the defendant.

4 Following Hungerford-Welch (2004), the amount should reflect the ability ofthe defendant to pay as well as the fine. If the defendant cannot pay both, the finehas precedence.

2 iew of

pJc

staa

raghai

bffd

asdwtrdsh

dtalitgfasadom

ma

i

AvTfipefa§2

34 L. Echazu, N. Garoupa / International Rev

rosecution have been reduced. Finally, Section 17 of the Access toustice Act 1999 introduces the possibility of a recovery of defenseosts order against defendants funded by the government.5

Notwithstanding the statutory developments, the application ofuch orders is much less common in criminal litigation and subjecto the behavior of the parties and the will of the court. There islways the possibility of a claim in torts to sue for compensationnd payment of legal expenses, but these claims are occasional.6

In the United States, the Hyde Amendment (1997) permitsecovery of attorneys’ fees by criminal defendants who have beencquitted following a (willful and frivolous) indictment by theovernment.7 At the same time, criminally culpable defendants canave fees levied against them in their convictions, so long as there is

statutory basis for it.8 A loser-pays-all rule is occasionally appliedn the context of corporate criminal litigation.9

The use of a loser-pays-all rule in criminal litigation should note confused with compensation for wrongful conviction. In oneorm or another, all legal systems potentially compensate wrong-ully convicted people. However, such compensation has little too with the legal costs incurred by the accused during trial.10

In this paper, we develop an economic model to consider thepplication of a loser-pays-all rule in criminal litigation. We con-ider the effects of this hypothetical rule primarily on criminaleterrence and on the possibility of legal error. In the basic model,e investigate the extent to which shifting costs from the loser

o the winner provides additional deterrence and reduces miscar-iage of justice. Under realistic conditions, shifting legal costs helpseterrence and therefore the English rule should prevail. At theame time, fee shifting increases the frequency of legal error andence the American rule is more appropriate from that viewpoint.

Our paper also explicitly addresses the situation where defen-ants do not have resources to reimburse the prosecution in ordero assess the effects of an asymmetric application of a loser-pays-ll rule. One practical reason why a loser-pays-all rule in criminalitigation might not be frequently considered and applied is that,n a significant proportion of the cases, both sides are funded byaxpayers. Prosecutors are public by nature and operate with a bud-et financed by taxes. Many defendants in criminal litigation use aorm of public funding (legal aid, public defender). In this context,

loser-pays-all rule would just imply a transfer of resources acrosstate bureaucracies. Consequently, it is possible that a loser-pays-ll rule in criminal litigation is only politically feasible when theefendant is wealthy. In the context of corporate criminal litigation

r wealthy defendants, there is no immediate reason to under-ine the application of a symmetric loser-pays-all rule. In fact, we

5 As discussed by Hungerford-Welch (2004), the order could be made up to theaximum of the full cost borne by the government. These orders have been used to

void abuses by the defendant.6 See a summary of remedies in civil action for wrongful or mistaken prosecution

n Sanders and Young (2007), at chapter 12.7 See Departments of Commerce, Justice, and State, the Judiciary, and Relatedgencies Appropriations Act, 1998 sec. 617, 18 U.S.C.A. sec. 3006A note; Masterson. U.S., 200 F.Supp.2d 94 (2002); Morgan v. Perry, 142 F.3d 670, C.A.3 (Pa. 1998).he Hyde Amendment provides in relevant part: during fiscal year 1998, and in anyscal year thereafter, the court, in any criminal case. . . may award to a prevailingarty, other than the United States, a reasonable attorney’s fee and other litigationxpenses, where the court finds that the position of the United States was vexatious,rivolous, or in bad faith, unless the court finds that special circumstances make suchn award unjust. Pub. L. 105-199,111 Stat. 2440, 2519 (1997) (reprinted in 18 U.S.C.3006A, historical and statutory notes).8 See McCray v. State, 665 So.2d 384, (Fla. 1st DCA 1995); Wright v. State, 654 So.2d

52 (Fla. 1st DCA 1995).9 We could not find any quantitative analysis of the Hyde Amendment.

10 For an economic analysis, see Fon and Schaefer (2007).

Law and Economics 32 (2012) 233– 241

conclude that a loser-pays-all rule applied only to wealthy defen-dants helps deterrence.11

The paper focuses on deterrence (i.e. crime rates) and legal error(i.e. miscarriage of justice) unlike the literature on civil litigation.The latter is usually more concerned with the impact of loser-pays-all rules on the incentive to litigate and settle (rather than theunderlying rate of wrongdoings or the impact on false positives).

As already stated, criminal and civil litigation are not the same.Our model reflects these differences and, in that context, departsfrom the previous literature that only looks at civil litigation.12 Civillitigation usually uses contingency fees which are not allowed incriminal litigation. In criminal litigation the objective is twofold:to increase deterrence, on one hand, and to decrease the numberof wrongful convictions, on the other hand. Given that it is impor-tant to avoid punishing innocent defendants, the standard of proofis higher under criminal litigation. Public enforcement prevails incriminal law. As a consequence, the fine paid by the defendant isnot the gain for the prosecutor. The fine is also not the equivalentto compensatory damages in civil litigation. Furthermore, publicenforcement is usually financed by public agencies with limitedbudgets and highly regulated or constrained by guidelines and pro-cedural rules. The wealth constraints for the defendant are moreserious in criminal litigation. Finally, plea-bargaining is not exactlythe equivalent of out-of-court settlements in civil litigation becausethe prosecutor is not directly benefited from imposing a higher fine.

Some of the features we derive in our model show similaritieswith the civil litigation literature. For example, as with the civillitigation literature, we find that fee-shifting will change the incen-tives for settlements and alter the incentives of defendants to takeprecaution (in our context, to comply with the law). The direction ofthese changes is frequently ambiguous and depends on the param-eters of the model. The parameters that determine the results aredifferent in our model given the criminal context.

We do not think our model is a mere thought experiment.Although cost-shifting is uncommon in criminal litigation, thereare significant applications such as the use of RICO proceedingsand the financing of enforcement authorities in the United States.

The basic model is developed in Section 2. A more detailed dis-cussion of our results in terms of policy implications is presentedin Section 3. Section 4 provides concluding remarks.

2. Basic model

As in the usual Polinsky and Shavell (2000) framework, we startby assuming that each risk-neutral individual chooses whether ornot to commit an offense, for example, an environmental violationor financial fraud. The offender’s gain from committing the offenseis b, which is distributed across the population according to a prob-ability density function g(b) and a cumulative distribution functionG(b), where b ∈ [0, ∞). The fraction of individuals deriving benefitsless than b from the crime is G(b). The size of the population is nor-malized to one. Each offense generates a social damage given by h.The net social harm from the crime is h − b (we take the usual viewthat the illegal gain is a social gain), where h< ∞.

The decision of whether or not to commit an offense generates asignal � that can be observed by the prosecutor (such as evidence tobe used in a criminal trial), and can be regarded as the prosecutor’s

expectation of obtaining a conviction. The signal � is itself a randomvariable distributed according to a probability density function rj(�)and a cumulative distribution function Rj(�), where � ∈ [0, 1] and

11 Notwithstanding, we acknowledge that our model does away with other com-plications on the application of criminal corporate liability already identified in theeconomic literature, hence, tempering our results with caution.

12 See survey by Katz and Sanchirico (2010).

iew of

tocoTtgpp

tpvtc

picetp

(

((

(

(

Fpg

2

pssa

apeHe

fiGa

L. Echazu, N. Garoupa / International Rev

he subscript j stands for a (actual offenders or guilty) or i (deterredffenders or innocent). The interpretation of the signal � is that zeroorresponds to no evidence (no probability of conviction) whereasne means overwhelming evidence (probability one of conviction).hen, Rj(�) represents the probability of acquittal conditional onhe individual’s type. For each � , we pose that Ra ≤ Ri, that is, theuilty are expected to generate a stronger signal than innocents. Aerfectly correlated signal would be Ri(� = 0) = 1 and Ra(� < 1) = 0. Aure noisy signal would be Ri = Ra for all � .

An offender will be prosecuted if the observed signal � is greaterhan a critical threshold �* chosen by the prosecutor. The costs ofrosecution are m for the prosecutor and x for the offender. If con-icted, an offender pays a monetary penalty f assumed to be costlesso collect. Let us assume there is a maximal fine F for financial oronstitutional reasons.13

The loser (offender or prosecutor) pays his own expenditurelus a fraction � of the other party’s expenditures. There are three

mportant cases to note: (i) the loser-pays-all or English rule inriminal litigation (� = 1); (ii) non-recoverable expenses, whereach party pays his or her cost, or American rule in criminal litiga-ion (� = 0); (iii) partial recovery of litigation costs by the prevailingarty (0 < � < 1).

The sequence of the game is the following:

1) The government announces a fine f and a rule of allocation ofcosts �;

2) Potential criminals decide whether or not to commit an offense;3) A signal � concerning the culpability of a potential offender is

observed by the prosecutor.4) The prosecutor and the defendant engage in criminal litigation

if and only if � ≥ �*, that is, if the signal is above some criticalthreshold. The costs are m and x respectively.

5) If the defendant is convicted, he is ordered to pay a fine f andreimburse the prosecutor for a fraction � of her costs. If thedefendant is acquitted, the prosecutor must reimburse him fora fraction � of his costs.

We need to solve the game backwards for subgame perfection.irst, we look at the decision to prosecute. Second, we assess theotential offender’s choice. Finally, we consider the policy of theovernment.

.1. The prosecutor

There is a vast economic literature on the utility function of therosecutor.14 For example, it is unclear if prosecutors maximize theeverity of sanctions or the number of convictions. Also controver-ial in the economic literature is the extent to which prosecutorsre benevolent or merely opportunistic concerning convictions.15

We assume a simple objective function. The prosecutor cares

bout the costs of prosecution because they are borne by therosecutorial budget (the prosecutor bears the actual costs of pros-cution as she cares about the budget of the prosecutorial office).owever, the fine is revenue for the government, not for the pros-cutor. Let u be the opportunity cost borne by the prosecutor if the

13 An extension that explicitly includes a judgment-proof problem caused by thene and expenditure in legal defense would replicate results similar to Gravelle andaroupa (2002). In our model, we assume defendants have two deep pockets thatllow them to spend on defense and later pay a fine.14 See discussion by Garoupa and Rizzolli (2011) and references therein.15 On the objective function of the prosecutor, see Garoupa (2009a).

Law and Economics 32 (2012) 233– 241 235

case is dropped.16 The prosecutor takes a particular case as long asthe expected cost is less than the opportunity cost u:

V = �(1 − �)m + (1 − �)[m + �x] ≤ u

As a consequence, the critical threshold for the prosecutor canbe easily derived:

�∗ = m + �x − u

�(m + x)

As long as � ≥ �*, the defendant will be prosecuted. The thresh-old is the ratio of the difference between the cost of losing a case,m + �x, and the opportunity cost for dropping a case u to the netbenefit from winning a case (and this is given by the differencebetween the payoffs from winning a case and losing it), �(m + x).Intuitively, if the signal � concerning the culpability of the defen-dant is sufficiently strong, prosecution takes place. It is easy to seethat a higher opportunity cost for dropping a case will make theprosecutor take on more cases, while a higher cost in the instanceof losing will make her take fewer cases. Also, as the net benefitfrom victory increases, the prosecutor will also be willing to takeon more cases.

Quite importantly, the threshold varies with �. An increase in� (shifting costs to the loser) has a positive impact on the criticalthreshold if and only if u > m. Therefore shifting costs to the losermakes prosecution less likely when the opportunity cost from drop-ping the case is significantly high. Otherwise, shifting costs to theloser makes prosecution more likely.

We can understand this result by starting with the Amer-ican rule. In this specific case, �* goes to plus infinite orminus infinite, depending on u > m (always prosecutes) or u ≤ m(never prosecutes). With respect to the English rule, we have�* = (m + x − u)/(m + x) which is strictly in between plus and minusinfinite. Therefore, when we move from the American rule to theEnglish rule, we can have more prosecution (if u ≤ m) or less pros-ecution (if u > m).

We can summarize the result in the following way:

Proposition 1. Shifting costs to the loser increases the critical thresh-old for the prosecutor if the opportunity cost from dropping a caseis higher than the cost of prosecution (i.e. if u > m, �∗

� > 0); but itdecreases the critical threshold if the converse is true (i.e. if u < m,�∗

� < 0). In cases where these two costs are of equal magnitude, shiftingcosts has no effect on the threshold (i.e. if u = m, �∗

� = 0).

The American rule induces more prosecution when the oppor-tunity cost borne by the prosecutor is very significant. The Englishrule is associated with more prosecution when the opportunity costfrom dropping a case is not very significant. The intuition regardingthe American rule is pretty straightforward. Under the Americanrule, the decision to prosecute is given by V = m < u, and clearly theprosecutor will pursue a case whenever the cost she bears for pros-ecuting is lower than the cost of dropping the case, even if theprobability of winning a case is virtually zero.

The case of the English rule may at first glance seem some-what counterintuitive. One may wonder why a prosecutor woulddrop more cases with increases in � if the cost for dropping a

16 We find this simple model appealing given the clear disagreement in the litera-ture about what prosecutors actually care about (convictions, crime rates, degree ofculpability, salary, promotion, etc.). Mathematically, the parameter u can be adjustedto any of the different alternatives. There are two important differences between aprosecutor in this model and a plaintiff in civil litigation: the prosecutor does notget the fine (more generally, the prosecutor cares only about winning but not byhow much) and the prosecutor operates under a resource constraint (measured byan opportunity cost given by u). If the prosecutor only cares about a fraction of theactual costs of prosecution, the opportunity cost u should be adjusted appropriately.

2 iew of

csVcs�

bwuwtotahfSs

2

t

w

filiccfg

b

ato

tc(rtttzt

l

The American rule diminishes deterrence whereas the Englishrule promotes deterrence by shifting costs to the loser who is likelyto be the defendant given the decision of prosecuting only when

36 L. Echazu, N. Garoupa / International Rev

ase is higher than the cost of prosecuting the case. The deci-ion of the prosecutor under the English rule is determined by

= (1 − �)(m + x) < u. There are two different effects. If the prose-utor loses, she pays an additional x, but she only pays m + x ifhe loses; if she wins she pays nothing. A closer examination of* reflects these two different effects.

The costs from losing the case increase (the numerator effect),ut the difference between winning and losing a case increases asell (the denominator effect). Now, the question is which one goesp by more and how that is related to u. Let us look at the casehere the opportunity cost for dropping a case is rather high. In

hose cases, the prosecutor will already be taking several cases, inther words, �* is closer to zero. Inevitably shifting costs increaseshe burden for the prosecutor since the probability of bearing thesedditional costs is significant, that is, 1 − �* is high. Therefore, �*

as to go up. Consider now the opposite case. The opportunity costor dropping a case is minimal. The threshold �* is closer to one.hifting costs actually reduces the burden of the prosecutor becausehe does not expect to lose. Therefore, �* should go down.17

.2. The offenders

Each potential offender decides to commit a crime if and only ifhe payoff for being a criminal is higher than that for being innocent:

b − Ca ≥ 0 − Ci

here

Ca =∫ 1

�∗[�(f + x + �m) + (1 − �)(1 − �)x]dRa(�)

Ci =∫ 1

�∗[�(f + x + �m) + (1 − �)(1 − �)x]dRi(�)

On the left-hand-side of the inequality, the expected net gainrom committing an offense is the illegal gain obtained by the crim-nal minus the expected fine minus the expected costs of criminalitigation borne by a guilty defendant. On the right-hand-side of thenequality, we have the (normalized to zero) expected gain from notommitting an offense minus the expected fine minus the expectedosts of criminal litigation borne by an innocent defendant. There-ore a potential offender commits a crime if and only if the illegalain offsets the cost differential:

≥ Ca − Ci =∫ 1

�∗[�(f + x + �m) + (1 − �)(1 − �)x][ra(�) − ri(�)]d�

= z(f, �)

As the literature suggests, false positives (that is, prosecutionnd conviction of innocent individuals) decrease deterrence sincehe expected cost is more severe when there are no errors in termsf prosecuting innocents.18

17 This contrast between the American and the English rule can potentially impacthe incidence of false positives. When u > m the American rule leads to sure prose-ution, whereas under the English rule leads to a higher threshold for prosecutionthus, prosecuting fewer cases). In this situation, the English rule could potentiallyeduce the incidence of wrongful convictions. On the other hand, when u < m underhe American rule the prosecutor will never file a suit, while under the English rulehe threshold decreases (thus, there will be more cases prosecuted). In this case,he American rule dramatically decreases the number of wrongful convictions toero, while the English rule increases the potential for wrongful convictions. We arehankful to the editor for pointing this out.18 See Garoupa (1997) and Polinsky and Shavell (2000). On the context of civilitigation, see Rubinfeld and Sappington (1987).

Law and Economics 32 (2012) 233– 241

The critical value of deterrence z(.) varies with �. In particularnotice that:

z∗� =

∫ 1

�∗[�m − (1 − �)x][ra(�) − ri(�)]d� + [�∗(f + x + �m)

+ (1 − �∗)(1 − �)x][ri(�∗) − ra(�∗)]�∗�

The first term reflects the difference in litigation costs when anindividual is guilty and innocent. It depends on the relative magni-tude of those costs, m and x, and the probability of being convicted.While the sign of the expression depends on the parameters of themodel, we can easily suppose it is more realistic to expect a posi-tive sign for two reasons. First, � has to be higher than the thresholdand therefore closer to one than to zero. Second, given the condi-tional distribution of the signals, an individual is more likely tobe prosecuted when guilty than when innocent. This intuition canonly be reversed if the litigation costs borne by the defendant xare overwhelmingly larger when compared to those borne by theprosecutor m, an extremely unrealistic scenario.19

The final term measures the difference in probabilities when thelikelihood of prosecution changes due to cost shifting. Since we arelooking at continuous probability density functions when � is at thethreshold (hence, with a zero probability density), we can considerthe final term a second-order effect and assume it is arbitrarily closeto zero (in fact, with a zero probability density, the term is zero).

We can also show that deterrence increases unambiguouslywith the fine:

z∗f

=∫ 1

�∗�[ra(�) − ri(�)]d� > 0

given the assumptions about the probability density functions ofthe signal when an individual is guilty and when an individualis innocent. It basically says that the probability of paying a fineis higher when an individual is guilty than when an individual isinnocent.

In conclusion, we have shown the following proposition:

Proposition 2. Assume that defense expenses are not significantlylarger than prosecutorial expenses and that the difference in proba-bilities of conviction is arbitrarily small (i.e. ri(�*) − ra(�*) ≈ 0). Underthese conditions, shifting costs to the loser increases deterrence (i.e.z� > 0). Similarly, increasing fines enhances deterrence (i.e. zf > 0).20

evidence is above a strategic threshold.

19 In particular,

∫ 1

�∗[�m − (1 − �)x][ra(�) − ri(�)]d� > 0

given the assumptions about the probability density functions of the signal when anindividual is guilty and when an individual is innocent and the relationship betweenm and x.

20 We can argue that, under realistic conditions, deterrence increases with cost-shifting. However, in our analysis, we will also consider in a footnote the less realisticpossibility that deterrence does not increase with cost-shifting and derive morecomplete results.

iew of

2

Tcojw

W

toddii

w

ibmqrm

t(fstfii

Plboeflm�

ae

oPs

t

co

L. Echazu, N. Garoupa / International Rev

.3. Policy objectives

The first policy objective we consider is minimization of costs.21

his setup reflects the social value of deterrence. However, in theontext of criminal litigation, we also want to focus on the impactn false positives, namely a possible social cost of miscarriage ofustice that plays an important role in shaping criminal law. That

ill be the second policy objective we analyze.Enforcement costs plus social damage is given by:

= (x + m)[1 − Ri(�∗)]G(z(f, �))

+(x + m)[1 − Ra(�∗)][1 − G(z(f, �))] + h[1 − G(z(f, �))]

The first term refers to the costs borne by the prosecutor andhe defendant when an innocent is prosecuted whereas the sec-nd reform refers to the costs borne by the prosecutor and theefendant when a guilty is prosecuted. The last term reflects socialamage. The social objective is to choose � and f in order to min-

mize the above expression, assuming the second-order conditions satisfied. The first-order derivatives are:

Wf = Wzzf < 0W� = Wzz� − (x + m){ri(�∗)G(z(f, �)) + ra(�∗)[1 − G(z(f, �))]}�∗

�

here

Wz = {(x + m)[Ra(�∗) − Ri(�∗)] − h}g(z) < 0

Let us start with the optimal fine. The first term in Wf is Wz. Its necessarily negative and measures the cost imposed by criminalehavior (social damage and additional costs of prosecution). It isultiplied by a positive term zf (fine increases deterrence). Conse-

uently, the optimal fine is necessarily maximal, that is, f = F. Thisesult reflects the standard maximal fine result in the law enforce-ent literature (Becker, 1968; Polinsky & Shavell, 2000).We can now focus on the allocation of litigation costs. The first

erm in W� is negative under the assumptions of Proposition 2it reflects the fact that shifting costs increases deterrence). Theollowing term measures the impact of shifting costs on the deci-ion of whether or not to prosecute. When �∗

� > 0, we have W� < 0,herefore the English rule is necessarily optimal. When �∗

� < 0, therst and second terms push in opposing directions. Therefore, an

nterior solution is generally optimal.22

We can now write the following proposition:

roposition 3. Assume that defense expenses are not significantlyarger than prosecutorial expenses and that the difference in proba-ilities of conviction is arbitrarily small (i.e. ri(�*) − ra(�*) ≈ 0). If thepportunity cost from dropping a case is higher than the cost of pros-cution (i.e. u > m), the optimal fine is maximal and the optimal ruleor cost shifting is the English rule (i.e. � = 1). If the opportunity cost isower than the cost of prosecution (i.e. u ≤ m), the optimal fine is maxi-

al and the optimal rule of cost shifting yields an interior solution (i.e. is between zero and one).23

The immediate observation is that the American rule is gener-lly not optimal. The English rule is optimal when it simultaneouslynhances deterrence while reducing expected costs from criminal

21 In the law enforcement literature, social welfare is usually defined to be the sumf illegal gains minus social damage minus enforcement costs, as in the standardolinsky and Shavell (2000) framework. Minimization of enforcement costs andocial damage has been shown to produce equivalent results.22 As noted, the two effects push in opposite directions. However, if the seconderm dominates, we may end up with an American rule as optimal.23 Notice that under a continuous probability density function, it should be thease that ri(� *) and ra(� *) are zero. If so, W� is strictly negative under the conditionsf Proposition 2 and the English rule is always optimal.

Law and Economics 32 (2012) 233– 241 237

litigation (due to the negative impact on the likelihood of prosecu-tion). The intuition is that fewer individuals are prosecuted but, ifconvicted, they pay more because they bear prosecutorial costs. Inthis context, the English rule delivers more deterrence at a lowercost.

An interior solution is more appropriate when shifting costsincreases expected costs from criminal litigation (by enhancing thelikelihood of prosecution) which have to be traded off against gainsin deterrence. The American rule can only emerge as optimal whenthe first effect completely dominates the second effect.24

In the context of criminal law, miscarriage of justice could bean important aspect to consider more closely. The proportion ofinnocents wrongly prosecuted is given by:

M = (1 − Ri(�∗))G(z(f, �))

Suppose the government uses shifting rules to minimize theproportion of innocents wrongfully prosecuted (while presum-ably using the fine to achieve efficient deterrence). The first-orderderivative is given by:

M� = −ri(�∗)G(z(f, �))�∗� + g(z(f, �))[1 − Ri(�∗)]z�

The result depends on z� and �∗� . Under the conditions of

Proposition 2, shifting costs in criminal litigation increases deter-rence. As a consequence, shifting costs also increases the pool ofindividuals subject to miscarriages of justice. When �∗

� is negative,we know that M� is positive, therefore, the optimal � is necessarilyzero. When �∗

� is positive, the two terms have opposite signs andan interior solution is likely to be optimal. In the extreme case thatthe first effect dominates, the optimal � is one.

We have proved the following proposition:

Proposition 4. Assume that defense expenses are not significantlylarger than prosecutorial expenses and that the difference in proba-bilities of conviction is arbitrarily small (i.e. ri(�*) − ra(�*) ≈ 0). If theopportunity cost for dropping a case is relatively smaller than the costof prosecution (i.e. u ≤ m), the American rule is adequate (i.e. � = 0).When the converse is true, the optimal cost shifting rule yields aninterior solution (i.e. � is between zero and one).25

In terms of miscarriage of justice, the English rule is generallynot adequate. When shifting costs increases the likelihood of prose-cution, the American rule is more appropriate. When shifting costsdecreases the likelihood of prosecution, an interior solution seemssuitable.

Clearly M indicates that there is a tradeoff between enhancingdeterrence and reducing the miscarriage of justice. Under the con-ditions of Proposition 4, the American rule decreases deterrence,therefore decreasing the pool of individuals subject to miscarriagesof justice. If, at the same time, the American rule also reduces theconditional probability of being convicted given that one is inno-cent, then shifting costs can never be optimal. However, if it hasthe opposite effect, that is, the American rule increases the con-ditional probability of being convicted given that one is innocent,then shifting some costs is adequate. The English rule can only beoptimal if such effect completely dominates the deterrence effect(note that the English rule increases deterrence, increasing the pool

of individuals subject to miscarriages of justice).

Comparing Propositions 3 and 4, the conclusion is that the Amer-ican rule dominates the English rule for the purpose of avoiding

24 Outside of the conditions specified in Proposition 2, it could be that shifting costsdecrease deterrence, a less realistic possibility. However, in this context, the resultsof Proposition 3 have to be reversed.

25 Notice that under a continuous probability density function, it should be the casethat ri(� *) is zero. If so, M� is strictly positive under the conditions of Proposition 2and the American rule is always optimal.

2 iew of

mc

msmddgo

eti

3

htm

3

itpbtcoesn

bim(o�(g

fbttstiUHl

p

cstc

ec

38 L. Echazu, N. Garoupa / International Rev

iscarriage of justice while the English rule dominates the Ameri-an rule for the purpose of minimizing social costs.

Our results illustrate the difficult choice between deterrence andiscarriage of justice. Addressing optimal deterrence requires one

et of policies (maximal fine and English rule) that will create tooany miscarriages of justice. Minimizing legal error demands a

ifferent policy (American rule) that is inconsistent with optimaleterrence. One possibility is to achieve a balance between bothoals. Such balance is likely to beg for a cost-shifting policy shortf the English rule.26

Another perspective is to understand the overwhelming pref-rence for the American rule in criminal procedure as evidencehat avoiding miscarriage of justice is paramount while other policynstruments (sanctions) can be used for deterrence purposes.

. Discussion

Our basic model indicates that shifting criminal litigation costsas an impact on deterrence as well as on miscarriage of justice. Inhis section, we discuss our results in light of recent legal develop-

ents and other possible more realistic extensions.

.1. The decision to engage in plea-bargaining

A cost-shifting policy may be desirable if, in addition to increas-ng deterrence, it encourages settlements in criminal litigation,hus, saving on the costs of a trial.27 In general, the possibility oflea-bargaining exists if the expected cost from conviction borney the offender (the willingness to pay for the offense) is greaterhan the expected benefit from conviction enjoyed by the prose-utor (the willingness to accept for the prosecutor). In a situationf symmetric information, plea-bargaining fails if there are inher-nt costs to negotiation. As a consequence, if the range of possibleettlements increases, it is more likely that the inherent costs ofegotiation will not undermine plea-bargaining.

In the context of this model, the possibility of a settlementetween prosecutor and offender depends on the extent to which

t saves costs to both sides. The prosecutor accepts a settle-ent as long as the costs are lower than V, given that � ≥ �*

that is, there is enough evidence to prosecute the offender). Theffender accepts a settlement if it less than the expected trial costs,(f + x + �m) + (1 − �)(1 − �)x, for the guilty and the innocent alike

since, independent of culpability, the evidence is strong enough toenerate prosecution).

Shifting costs changes the expected costs for the prosecutor andor the offender. The first derivative of the expected trial costs borney the offender with respect to � is �m − (1 − �)x. The first deriva-ive of the expected trial costs borne by the prosecutor with respecto � is the exact symmetric, (1 − �)x − �m. This is not surprisingince shifting costs is a zero-sum game in our basic model. Underhe conditions of Proposition 2, the first derivative for the offenders positive while the first derivative for the prosecutor is negative.

nder different conditions, the opposite result can be obtained.owever, we can easily see that there is always one player who

oses and one player who wins with cost-shifting.28

26 An explicit formulation of the problem would follow the objective functionroposed by Miceli (1990).27 See Polinsky and Shavell (2000) and Garoupa (1997) for a more general dis-ussion. See Bebchuk (1984) for the effect of a cost-shifting rule on out-of-courtettlements in civil litigation. Shifting costs could reduce settlement in civil litiga-ion if there is significant asymmetric information or divergence of beliefs about thease.28 Except in the unrealistic case that � = x/(x + m). Under this condition, thexpected trial costs borne by offender and by prosecutor do not change with aost-shifting rule.

Law and Economics 32 (2012) 233– 241

The effect of cost-shifting is necessarily ambiguous on the like-lihood of settlement. The player who bears more costs with theEnglish rule (the offender under the conditions of Proposition 2)is more willing to settle; the player who bears less costs with theEnglish rule (the prosecutor under the conditions of Proposition 2)is less willing to settle.

The model cannot provide a definite answer to the questionof whether or not shifting costs increases the likelihood of plea-bargaining. Notwithstanding, two observations are in order. First,it seems more likely that shifting costs does not increase the like-lihood of plea-bargaining under the conditions of Proposition 2;in particular the prosecutor who initiates the process of plea-bargaining is less keen on pursuing such possibility. Second, underthe basic model where cost-shifting is a zero-sum game and thereare no transaction costs, it is highly possible that moving from theAmerican to the English rule has no impact on the likelihood ofplea-bargaining.

Plea-bargaining reduces costs of trial. Therefore our results indi-cate that cost-shifting rules are unlikely to produce major savingsin trial costs.29 At the same time, plea-bargaining could have animpact on screening the guilty from innocent. The expected costfrom criminal litigation is higher for the guilty than the inno-cent, on average, given the probability density functions ra(�) andri(�). Under the conditions of Proposition 2, a cost-shifting rulehurts defendants. Individuals with a significantly large � have highexpected costs anyway and would be willing to settle with or with-out cost-shifting. Individuals with � just above �* might be willingto settle under the English rule only. The first group is likely to bedominated by the guilty. Innocents are more likely to be in the sec-ond group. Under these conditions, a cost-shifting rule reduces thescreening power of plea-bargaining since more individuals wantto settle, not just those with a significantly large � (who are morelikely to be the guilty).

3.2. One side cost-shifting rules

In this section we discuss the basic model when only one partyhas enough resources to compensate the other. We take an extremeassumption, that is, the wealth-constrained party cannot reim-burse the litigation costs borne by the other party. A more generalapproach would be to consider the party being constrained reim-bursing the other side up to the constraint. However, the resultswe derive are robust and highlight the role of each rule. The gen-eral conclusion depends on how constrained a party is and shouldreflect a combination between the two extremes we consider,unconstrained or constrained to the point of not being able to reim-burse the other party.

Poor defendants: So far we have developed the model assumingeach and every potential criminal has assets to actually be able topay for court costs. Suppose now that defendants cannot afford toreimburse legal costs. Evidently the loser-pays-all rule only appliesto the prosecution.

Let us start by focusing on the expected payoff for the prosecu-tor:

V = �m + (1 − �)[m + �x] = m + (1 − �)�x ≤ u

Cost shifting hurts the prosecutor unambiguously. Thereforefewer individuals will be prosecuted under the English rule. Whenthere is a wealth-binding liquidity constraint for the defendant,

29 Unless we assume a model with asymmetric information and both sides arehighly pessimistic about trial outcomes. If so, by shifting litigation costs, both sidesexpect to pay more and are more willing to settle. This result is in line with theliterature on civil litigation.

iew of

sa

rTETda

ape

waab

t

Trpd

barew

atj

daA

i

it

cG

clnm

L. Echazu, N. Garoupa / International Rev

hifting costs hurts the prosecutor since she is the only one actuallyffected by this rule.

Let us turn our attention to the criminals:

z(f, �) =∫ 1

�∗[�(f + x) + (1 − �)(1 − �)x][ra(�) − ri(�)]d�

Not surprisingly, the English rule unambiguously hurts deter-ence by subsidizing defendants in case they do not get convicted.here are more criminals and fewer are prosecuted. Therefore, thenglish rule is inappropriate from the viewpoint of deterrence.he effect on deterrence now is unambiguously negative since theefendant benefits from shifting costs. When defendants are poor,n American rule should prevail for sake of deterrence.30

As for miscarriage of justice, there are fewer innocents and fewerre prosecuted. As a consequence, a cost-shifting rule that hurts therosecutor reduces the likelihood of miscarriages of justice at thexpense of more crime.

Wealthy or corporate defendants: A different situation existshen the defendant is wealthy and the prosecution, in turn, faces

limited budget.31 In this specific case, the loser-pays-all rule onlypplies to the defendant since the prosecutor cannot afford to reim-urse legal costs.

Let us start by considering the expected payoff for the prosecu-or:

V = �(1 − �)m + (1 − �)m ≤ u

Cost-shifting now benefits the prosecutor unambiguously.herefore more individuals will be prosecuted under the Englishule. When there is a wealth-binding liquidity constraint for therosecutor, shifting costs helps the prosecutor since she is subsi-ized by this rule.

Let us turn our attention to the criminals:

z(f, �) =∫ 1

�∗[�(f + x + �m) + (1 − �)x][ra(�) − ri(�)]d�

As expected, the English rule unambiguously helps deterrencey punishing wealthy defendants in case they get convicted. Therere fewer criminals and more are prosecuted. Therefore, the Englishule is extremely helpful from the viewpoint of deterrence. Theffect on deterrence now is unambiguously positive since theealthy defendant is hurt by shifting costs.32

As to miscarriage of justice, there are more innocents and morere prosecuted. As a consequence, a cost-shifting rule that hurtshe wealthy defendant augments the likelihood of miscarriage ofustice in the context of enhanced deterrence.

This extension further shows the existing trade-off between

eterrence and legal error. If reducing legal error is paramount, theppropriate policy seems to be English rule for poor defendants andmerican rule for wealthy defendants.

30 Note that in the case where the criminal uses public defense the expected fine

s reduced to z =∫ 1

�∗ �f [ra(�) − ri(�)]d� and expected cost borne by the prosecutor

s V = m. In that case there is no point in implementing either rule. Furthermore, ifhe defendant is unable to even pay the fine, z = 0 and there is no deterrence.31 It is in this light that we consider corporate defendants. An economic theory oforporate criminal liability is beyond the scope of this article and can be found inaroupa (2000) and Garrett (2011).

32 The result is subject to the standard caveat that additional corporate liabilityould generate perverse effects by reducing the company’s incentive to enforce theaw against potentially liable managers. If enhancing prosecution of corporations isot adequate due to external costs that are not included in the model, the conclusionust be reversed.

Law and Economics 32 (2012) 233– 241 239

3.3. Costly sanctions and transfers

In the tradition of Becker (1968), we have assumed that fines andreimbursement of legal costs are costless to enforce. The extensionof the model to costly sanctions (such as imprisonment) and tocostly enforcement of a loser-pays-all rule is not difficult.

Suppose the reimbursement of legal costs is costless and sanc-tions are costly. Then, sanctions should be lower than when theyare costless to enforce, so the socially optimal fine should be aslow as possible in order to avoid the maximal fine restriction. Theallocation of legal costs, however, should not be affected if reim-bursements of them are costless.

The conclusion is quite different in the opposite scenario. Withcostless fines but costly reimbursement of legal costs, the Englishrule is more problematic. In the situations where we concluded anEnglish rule would be efficient or more appropriate, now the resultneeds to be reconsidered. Presumably, some shifting of legal rulesis still efficient but not to the full extent of the English rule. At thesame time, fines should go up to keep deterrence unchanged unlessthey are restricted by the maximal fine constraint.

There is another point concerning sanctions that requires someadditional consideration. According to our model, the loser-pays-allrule would further enhance deterrence under realistic conditions.It is possible that courts could be tempted to reduce the imposedsanction, or raise the standard of proof, to offset the perceived addi-tional burden from paying for the costs of the prosecution. Sucheffect should be taken into account when determining optimal lawenforcement. If juries are more likely to acquit or judges are morelikely to impose reduced sanctions when, following conviction, thedefendant is required to cover the prosecution’s costs of trial, thencost-shifting is likely to be inefficient.

In fact, if juries react to cost-shifting rules, one could argue that� , which is the probability of conviction in our model, should beadjusted. The prosecutor may anticipate the reaction of the juryand choose �* accordingly. If the prosecutor believes the jury maybe more inclined to acquit under the English rule, she might be lesswilling to prosecute (since her expected costs are augmented byfee-shifting) and, consequently, �* goes up. In this context, cost-shifting could have a detrimental effect on deterrence.

3.4. Asymmetric information

The basic model operates under symmetric and perfect informa-tion, except in the context of false positives. However, asymmetricinformation has played an important role in the discussion ofloser-pays-all rules in civil litigation. More precisely, if, due toasymmetric information both sides are overly optimistic abouttheir chances of prevailing, the English rule could promote morelitigation (since each side expects the other side to pay for legalcosts) and fewer settlements. At the same time, an English rulecould reduce nuisance litigation since the strategic plaintiff knowsthe chances of actually winning are low and therefore might beliable for all legal costs.

We argue that similar results can be derived for criminal liti-gation. An optimistic prosecutor will be more willing to prosecuteunder an English rule than under an American rule since she antic-ipates the defendant will pay for legal costs (provided he canreimburse them). Plea-bargaining is less likely to be used (keepingin mind that even with symmetric information, we have observedthat the English rule could hurt plea-bargaining under some con-ditions). Both consequences should presumably help deterrence.More prosecution and less plea-bargaining increase expected pun-

ishment for potential offenders. Therefore, under asymmetricinformation that makes the prosecutor more optimistic, the Englishrule helps deterrence. The opposite reasoning applies if asymmet-ric information makes the prosecutor more pessimistic. An English

2 iew of

rda

3

fblwd

stidtpors

3

oa

aEEtaoe

altompab

ttlsgr

4

hccgl

q

40 L. Echazu, N. Garoupa / International Rev

ule also hurts deterrence if asymmetric information makes theefendant more optimistic (therefore, thinking he can be acquittednd escape legal costs).

.5. Risk aversion

Our model follows the standard Polinsky and Shavell (2000)ramework with risk neutral individuals. Consider now the possi-ility of risk aversion. It is clear that the English rule makes criminal

itigation riskier than the American rule. Inevitably, all rest equal,hen potential offenders are risk averse, an English rule enhanceseterrence.

Regarding the rate of settlements, considering risk aversion addsome asymmetry to the payoffs. If the defendant is risk averse, buthe prosecutor is risk neutral, then the English rule may actuallyncrease the willingness to settle for the defendant more than it mayecrease the willingness to settle for the prosecutor. In that case,he English rule will increase the rate of settlements. Even if therosecutor is risk averse, depending on the degree of risk aversionf the defendant compared to that of the prosecutor, the Englishule could still increase the rate of settlements if the defendant isufficiently more risk averse than the prosecutor.

.6. Policy applications

The results of our model can also be understood in the contextf ongoing policy discussions not directly framed in the context ofn application of the English rule.

Under RICO33 proceedings, the enforcement authorities arellowed to seize and keep the assets owned by the defendant.ssentially we can see this situation as a one-sided version of thenglish rule where the enforcement authorities can shift part ofheir legal costs to the defendant and where the seized assets oper-te as reimbursement. We have already seen that the consequencef such a provision is likely to be more deterrence and more legalrror (false positives).34

A similar interpretation can be given to statutory provisions thatllow prisoners to work while serving their sentence for salariesower than the market salary. Effectively, in these circumstances,he government has shifted legal and enforcement costs to the pris-ners. This is another version of a partial cost-shifting policy. Ourodel shows that such statutory provisions are likely to incentivize

rosecution (if they expand effectively prosecutorial resources) andllow for more severe (costly) sanctions since the costs are borney the defendant after conviction.

Another important policy application is the discussion overhe financing of enforcement authorities (police and prosecutors)hrough a public budget or self-funding mechanisms (includingoans, funding opportunities or law enforcement grants). Underelf-funding enforcement authorities, any potential deterrenceain from an English rule must be balanced against the burden ofaising funds.

. Concluding remarks

Loser-pays-all rules are generally not used in criminal cases. Weave considered the effects of this hypothetical rule primarily onriminal deterrence and on legal error. The overall desirability of a

ost-shifting rule should thus be evaluated comparing the marginalains obtainable on deterrence with the marginal costs induced byegal error.

33 The Racketeer Influenced and Corrupt Organizations Act enacted in 1970.34 Subject to the caveat that the model does not explain other possible conse-uences such as the perverse incentive for the prosecution to overperform.

Law and Economics 32 (2012) 233– 241

The effect of a loser-pays-all rule on deterrence is positive underrealistic conditions. We also concluded that it would be inappropri-ate when the defendant is poor, but could be better when we thinkof wealthy and corporate defendants. More generally, balancing theadvantages in terms of deterrence against the additional costs inlegal error leads us to conclude that, in general, the English rule aswell as the American rule are unlikely to be efficient (alternatively,an English rule with caps in order to reduce legal error is probablythe efficient approach).

The model could not provide a definitive answer to the questionof whether or not shifting costs increases the likelihood of plea-bargaining. Notwithstanding that, it seems more likely that shiftingcosts does not increase the likelihood of plea-bargaining, unless weassume that criminal defendants and prosecutors are significantlypessimistic. The impact of fee-shifting on the quality of screeningby plea-bargaining seems negative if we take the view that it islikely to less severely shape the incentives of the guilty than of theinnocent.

We have used RICO proceedings and the financing of enforce-ment authorities as two examples of how the insights of our modelcan be useful in current legal policy discussion.

Acknowledgments

The authors would like to thank two anonymous referees, theeditor Avery Katz, and Pinaki Bose, Brandon Garret, Rich Hynes, JonKlick, Tom Miles, Francesco Parisi, Mitch Polinsky, Andrew Sandersand the participants at the UVA Conference on Criminal Law andEconomics. Roya H. Samarghandi has provided excellent researchassistantship. The usual disclaimers apply.

References

Bebchuk, L. A. (1984). Litigation and Settlement Under Imperfect Information. RANDJournal of Economics, 15, 404–415.

Becker, G. S. (1968). Crime and Punishment: An economic approach. Journal of Polit-ical Economy, 76, 169–217.

Cooter, R. D., & Rubinfeld, D. L. (1989). Economic analysis of legal disputes and theirresolution. Journal of Economic Literature, 27, 1067–1097.

Fon, V., & Schaefer, H.-B. (2007). State liability of wrongful conviction: Incentiveeffects on crime levels. Journal of Theoretical Economics, 163, 269–284.

Garoupa, N. (1997). The theory of optimal law enforcement. Journal of EconomicSurveys, 11, 267–295.

Garoupa, N. (2000). Corporate criminal law and organization incentives: A manage-rial perspective. Managerial and Decisions Economics, 21, 243–252.

Garoupa, N. (2009a). Some reflections on the economics of prosecutors: Mandatoryv. selective prosecution. International Review of Law and Economics, 29, 25–28.

Garoupa, N. (2009b). The English rule with payments upfront. Portuguese EconomicJournal, 8, 177–181.

Garoupa, N., & Rizzolli, M. (2011). The Brady Rule May Hurt the Innocent. AmericanLaw and Economics Review, 13, 168–200.

Garrett, B. (2011). Globalized corporate prosecutions. Virginia Law Review, 97,1775–1876.

Gravelle, H., & Garoupa, N. (2002). Optimal deterrence with legal defense expendi-ture. Economic Inquiry, 40, 366–379.

Hause, J. C. (1989). Indemnity, settlement, and litigation, or I’ll Be Suing You. Journalof Legal Studies, 28, 157–179.

Hungerford-Welch, P. (2004). Criminal litigation and sentencing. Cavendish Publish-ing.

Hylton, K. (2002). An asymmetric model of litigation. International Review of Law andEconomics, 22, 153–175.

Katz, A. W. (1987). Measuring the demand for litigation: Is the English rule reallycheaper? Journal of Law, Economics and Organization, 3, 143–176.

Katz, A. W., & Sanchirico, C. W. (2010). Fee Shifting in Litigation: Survey and Assess-ment. Institute for Law and Economics, University of Pennsylvania Law School.Research Paper 10–30.

Miceli, T. (1990). Optimal prosecution of defendants whose guilt is uncertain. Journalof Law, Economics and Organization, 6, 189–201.

Polinsky, A. M., & Shavell, S. (2000). The economic theory of public enforcement oflaw. Journal of Economic Literature, 38, 45–76.

Polinsky, A. M., & Rubinfeld, D. L. (1996). Optimal awards and penalties when theprobability of prevailing varies among plaintiffs. RAND Journal of Economics, 27,269–280.

Polinsky, A. M., & Rubinfeld, D. L. (1998). Does the English rule discourage low-probability-of-prevailing plaintiffs? Journal of Legal Studies, 27, 519–535.

iew of

R

SS

L. Echazu, N. Garoupa / International Rev

ubinfeld, D. L., & Sappington, D. E. M. (1987). Efficient awards and stan-dards of proof in judicial proceedings. RAND Journal of Economics, 18,

308–315.

anders, A., & Young, R. (2007). Criminal justice. Oxford University Press.havell, S. (1982). Suit, settlement, and trial: A theoretical analysis under alter-

native methods of the allocation of legal costs. Journal of Legal Studies, 11,55–81.

Law and Economics 32 (2012) 233– 241 241

Snyder, A., & Hughes, J. W. (1990). The English rule for allocating legal costs: Evidenceconfronts theory. Journal of Law, Economics and Organization, 6, 345–380.

Spier, K. E. (2007). Litigation. In A. M. Polinsky, & S. Shavell (Eds.), Handbook of lawand economics. North Holland.

van Wijck, P., & van Velthoven, B. (2000). An economic analysis of the Americanand the continental rule for allocating legal costs. European Journal of Law andEconomics, 9, 115–125.