Content downloaded/printed from HeinOnline

Fri Dec 20 07:11:38 2019

Citations:

Bluebook 20th ed.

John P. Gould, The Economics of Legal Conflicts, 2 J. Legal Stud. 279 (1973).

ALWD 6th ed.

John P. Gould, The Economics of Legal Conflicts, 2 J. Legal Stud. 279 (1973).

APA 6th ed.

Gould, J. P. (1973). The economics of legal conflicts. Journal of Legal Studies 2(2),
279-300.

Chicago 7th ed.

John P. Gould, "The Economics of Legal Conflicts," Journal of Legal Studies 2, no. 2
(June 1973): 279-300

McGill Guide 9th ed.

John P Gould, "The Economics of Legal Conflicts" (1973) 2:2 J Leg Stud 279.

MLA 8th ed.

Gould, John P. "The Economics of Legal Conflicts." Journal of Legal Studies , vol. 2,
no. 2, June 1973, p. 279-300. HeinOnline.

OSCOLA 4th ed.

John P Gould, 'The Economics of Legal Conflicts' (1973) 2 J Legal Stud 279

-- Your use of this HeinOnline PDF indicates your acceptance of HeinOnline's Terms and
Conditions of the license agreement available at
https://heinonline.org/HOL/License
-- The search text of this PDF is generated from uncorrected OCR text.
-- To obtain permission to use this article beyond the scope of your license, please use:
Copyright Information

Use QR Code reader to send PDF to your smartphone or tablet device

 THE ECONOMICS OF LEGAL CONFLICTS
JOHN P. GOULD*

I. INTRODUCTION'

IN neoclassical economic theory, utility functions and indifference curves

provide a mechanism for the analysis of how consumers and other decision
makers make choices among alternative goods or courses of action. The
neoclassical model is a theory of behavior under certainty, however, and does
not provide much help in the analysis of choices involving uncertain outcomes.
Fortunately, the gap was filled by the introduction of the expected utility
hypothesis (or theorem) by von Neumann and Morgenstern in their path-
breaking work, Theory of Games and Economic Behavior.2 The expected
utility hypothesis can be described simply as follows. Suppose an individual
with wealth W derives utility from that wealth according to a function U(W).
U(W) is the amount of utility derived from wealth W and is assumed to
increase as W increases. Now suppose that the individual is faced with the
following choices: (A) a certain wealth of $9,000; (B) a wealth of $10,000
with probability p and of $8,000 with probability 1 - p. According to the ex-
pected utility hypothesis the individual will prefer alternative A if
U(9,000) > pU(10,000) + (1 - p) U(8,000)
whereas B will be preferred to A if the inequality is reversed and indifference
* Associate Professor of Business Economics, Graduate School of Business, University of
Chicago. An earlier version of this paper, entitled On the Economics of Going to Court,
was distributed in September 1968 as a report from the Center for Mathematical Studies
in Business and Economics. Since then, a more general model, which applies to a broader
class of problems and which appears to provide more insight as to the structure of the
court problem, has been developed, and it is the focus of this paper. Some of the results
of the earlier paper have appeared in the published works of other authors who were at
work concurrently on this and related problems. The relationship of this paper to these
other papers will be taken up in the text. I gratefully acknowledge the comments of
members of the Industrial Organization Workshop at the University of Chicago on an
earlier draft of this paper. This research has been supported in part by the National
Science Foundation.
I This part and Part II describe the expected utility hypothesis and contain expository
material for the benefit of readers who are not familiar with these techniques. Those
readers who are familiar with this material may wish only to skim these parts of the
paper.
2 John von Neumann & Oskar Morgenstern, Theory of Games and Economic Behavior
(1944). For another valuable discussion of this theory see Milton Friedman & L. J.
Savage, The Utility Analysis of Choices Involving Risk, 56 J. Pol. Econ. 279 (1948).
 THE JOURNAL OF LEGAL STUDIES

U (W)

u~w

W
FIGvUE 1

obtains when the inequality is replaced with equality. Notice that the expected
value of alternative B is 10,000p + 8,000(1 - p). An individual is said to be
risk averse if he always prefers the certain payment of the expected value of
a risky choice to the risky choice itself.
U(pW 1 + (1 - p)W 2 ) > pU(W 1 ) ± (1 - p)U(W 2 ). (1)
Inequality (1) is the definition of a concave function. Hence, if U(W) is
concave, as in Figure 1, the individual must be risk averse, and vice versa.
It also follows from the inequality that a risk-averse individual is willing to
pay something for certainty. Thus, if (1) holds we can find values of W that
are less than the expected income pW1 + (1 - p)W 2 and that will be pre-
ferred to the risky alternative. If Wo is such that indifference is obtained, i.e.,
U(Wo) -pVJ(W 1 ) + (1 -p) U(W 2 ),
then a risk averse individual is willing to "pay" as much as
pWI + (1 - p)W 2 - Wo

to avoid the risky alternative in the sense that he is willing to accept this
much less than the expected value of the gamble to avoid the gamble.
The fact that risk-averse individuals are willing to pay something to avoid
risk suggests an intuitively appealing proposition: when two risk-averse
individuals become involved in a conflict that has an uncertain outcome for
 THE ECONOMICS OF LEGAL CONFLICTS

each, they can both gain by eliminating uncertainty and settling the conflict
with a riskless transfer of wealth. To examine this proposition and related
issues in greater detail, this paper will develop a general framework for analyz-
ing the problem of trading among individuals in the face of uncertainty.
One obvious application of such a model is to the question of the disposition
of lawsuits. It has been observed that a large proportion-perhaps 95 per cent
-of all lawsuits are settled out of court. One aim of the paper is to show
why such behavior is to be expected and to provide some hypotheses about
what causes five per cent of the cases to go to court. A closely related question
is the phenomenon of negotiated sentences in criminal cases. William Landes
in his study of criminal courts uses a model analogous to the one discussed
8
here to explain why 80 per cent or more of criminal cases do not go to trial.
Similarly, Richard Posner, in his study of administrative agencies, applies a
4
related model to explain how such agencies allocate resources among cases.
Other applications, to labor-management negotiations and to problems of the
allocation of social cost, are possible.

II. THE BASIC MODEL

The familiar advantages of indifference-curve analysis in the neoclassical
"riskless" case can be extended to the expected-utility hypothesis. Suppose
that an individual is faced with a risky situation in which two states of the
world can obtain. State 1 occurs with probability p and state 2 with prob-
ability (1 - p). Given these probabilities the individual's expected utility
will be a function of the wealth he will have in each of the two states. In
other words, for given probabilities, his expected utility is a function of W,
and W 2, his wealth in state 1 and 2 respectively. Thus
V(Wl, W 2 ) = pU(W 1 ) ± (1 - p) U(W 2 ). 5 (2)
Given (2), it is straightforward to construct indifference curves where an
indifference curve is defined to be all the combinations of W, and W 2 such
that expected utility remains a constant (i.e., such that V(W 1 , W 2 ) - k for
fixed k). If U(') is increasing and concave (i.e., if U'(W) > 0 and U"(W)
< 0), then typical indifference curves would look like those shown in Figure 2.
The line labeled ko can be interpreted as all combinations of W, and W 2 such
that expected utility is constant and equal to ko. The line labeled ki (where

3William M. Landes, An Economic Analysis of the Courts, 14 J. Law & Econ. 61

(1971).
4 Richard A. Posner, The Behavior of Administrative Agencies, 1 J. Leg. Studies
305 (1972).
5 The specification could be made more general by making utility dependent on the
state as well as on wealth. We will comment on the implications of this generalization
later.
 THE JOURNAL OF LEGAL STUDIES

wealth in State 1

ko

Wealth in State 2
FIGURE 2

kz is greater than ko) has a similar interpretation. It can be shown without

much difficulty that the indifference curves such as ko and k, will have the
indicated convex shape when the individual is risk averse and that the slope
of an indifference curve at the coordinates W1 and W 2 will be
dW1 ( p) U'(W 2 )
dW 2 p U'(W1 ) ' (3)
dU

where U'(W) - d- Equation (3) tells us by how much W 1 will have

to be increased to compensate just enough for a reduction in W2 to leave

expected utility unchanged. Equation (3) tells us, for example, that if W2 is
larger than Wi, then a given decrease in W2 will require a smaller compensat-
ing increase in W, than would be the case if W2 were less than W1 (because
U'(W 2 ) < U'(W1 ) when W 2 > W 2 ). This simply means that when there
are relatively more resources in state 2, an additional dollar is valued more in
state 1 than in state 2. This is a logical implication of the risk aversion of the
individual.
Similarly, the larger p is, the smaller is the compensating increase in W 1 for
a given decrease in W 2 (because (1 - p)/p decreases as p increases). This is
reasonable since when p is larger the individual is more certain of obtaining
U(W).
 THE ECONOMICS OF LEGAL CONFLICTS

The Analysis of Two Individuals

The trading behavior of two individuals in the presence of uncertainty is

analyzed in Figure 3. If we start at the corner designated 01 and look in a
northeast direction into the box, we see indifference curves (such as I,) for
individual 1, just as in Figure 2. The origin for individual 2 is at 02, and
looking southwest from this corner we see his indifference curves (such as 12).
In other words, the diagram corresponding to Figure 2 for individual 2 has
simply been inverted in Figure 3. Accordingly, the distance 0 1A is the wealth
that individual 1 will have in state 2 and the distance BA is the wealth
individual 2 will have in state 2. In state 1, individual 1 will have wealth 0 1C
and individual 2 will have wealth DC. Thus the whole distance O1B(= 0 2D)
is the total wealth of both individuals in state 2, whereas the distance 0 1 D
(= 02B) is the total wealth of both individuals in state 1.
It is easy to show that the initial allocation of wealth (the point (C, A)
where I, and 12 intersect) can be improved upon in the sense that one or both
individuals can be made better off. Suppose individual 2 agrees to pay indi-
vidual 1 an amount CC' if state 1 obtains and individual 1 agrees to pay 2 an
amount AA' if state 2 obtains. This change leaves individual 2 on the same
indifference curve, but individual 1 is now on the indifference curve I',, which
represents greater expected utility since it is above and to the right of I,.

D State 2
I I 11
II,
12

StatelI , State 1
Cc I '

I I 12

01 A' A B
State 2
FIGURE 3
 THE JOURNAL OF LEGAL STUDIES

Clearly any trade to a point inside the wedge-shaped area between I and 12
will be an improvement over the initial position for both individuals.
The only cases where individuals cannot both benefit from such trades are
those like point (C', A') where their indifference curves are tangent. This is
because a southwest move from such a tangency will help individual 2 but
hurt individual 1, whereas a northeast move has the opposite effect.
These results can be expressed in terms of the concept of compensating
wealth changes discussed earlier. If two individuals have different compensat-
ing changes for a given distribution of wealth in the two states (i.e., different
values for equation (3)), then there exists a trade such that both can be
made better off. For example, if individual 1 is willing to trade $1 in state 1
for $1 in state 2 and individual 2 is willing to trade $2 in state 2 for $1 in
state 1, then individual 2 can offer individual 1 say $1.25 of state 2 wealth
for $1 of state 1 wealth. Both are better off after this trade. Individual 1 gets
$1.25 in state 2 for giving up $1 in state 1 even though he would be fully
compensated by a payment of only $1 in state 2-the $.25 is a gain to him. On
the other hand, individual 2 would be willing to pay as much as $2 in state 2
for the additional $1 in state 1 but has only paid $1.25 for the state 1 dollar-
the difference of $.75 is a gain to him. (This discussion assumes, of course,
that the trading costs do not exceed the potential gains.)
The conditions for gains from trading can thus be expressed in terms of
equation (3). Let PA and PB be the personal assessment of the probability of
state 1, by individuals A and B, respectively; let UA( ) and UB( ) be their
utility functions; let W*Aj and W*Bj be their respective wealth levels in
state 1 and let W*A2 and W*B2 be their respective wealth levels in state 2.
Then, assuming trading costs are zero, there will be a gain from trading unless
(1 - PA) UA(W*A2) (I - pB) U'B(W*B2)
PA U'(W*A1) PB U'B(W*Bl)

where the left-hand side of (4) is equation (3) for individual A and the right-
hand side is equation (3) for individual B.
Equation (4) is a necessary condition for Pareto optimality. In what follows
we assume that, if individuals can find a trade that makes both better off,
they will indeed trade (unless, of course, the costs of trading are greater than
the benefits).

III. AN APPLICATION TO THE ECONOMICS OF GOING TO COURT

The model of the last section can be applied in a variety of situations in-
volving risky conflicts between individuals. We begin with a simple conflict
involving a lawsuit.
 THE ECONOMICS OF LEGAL CONFLICTS

Economics of Going to Court

Consider the following elementary problem: Individual A suffers an accident
on B's property and decides to sue for $X. When will A and B find it in their
interest to settle out of court? Let WA and WB be the post-accident wealth of
A and B, respectively, and let PA and PB be their respective estimates of the
probability of A's winning the law suit. 6 If CA and QB are the costs of bring-
ing the case to court, then A's expected utility is
PA UA(WA + X - CA) + (1 - PA) UA(WA - CA)
and B's expected utility is
PB UB(WB - X - CB) + (I - pB) UB(WB - CB).
From the last section we know that this situation (litigation) will be a point
of Pareto optimality only if
-(I - PA) U'A(WA- CA) 1 - PB U'B(WB - CB)
PA U'A(WA+ X - CA) PB U'(WB - X - CB)
(5)
When risk aversion obtains, equation (5) can hold only if PA > PB, that is,
only if A's assessment of his probability of winning is greater than B's assess-
ment of that probability. 7 The reason for this is easy to see. When both
parties are risk averse,
U'A(WA - CA)
U'A(WA + X -aCA)
and
UB(WB -CB)
U' BXC)<
UYB (WB - X - CB) 1 (Sb)

because U"A and U"B are negative. To further interpret (5a) and (5b), sup-
pose the disputants agree on the probabilities (i.e., suppose PA = PB). Pareto
optimality cannot obtain in this case because
-U'A(WA - CA) U'B (WB- CB)
U'A(WA+ X - CA) U'(W - X - CD)
6 We assume here that these probabilities are not affected by the resources devoted to
legal services by either individual and that the costs of going to court are independent
of the outcome. This simplifying assumption will be relaxed later in the paper.
7 It is important to emphasize here that the condition PA > PB does not assure the
equality in (5). Given the ratios of the marginal utilities for each party and PA (or PB),
the other probability is completely determined if equality is to hold as in (5). Hence,
s
i
PA > pb necessary but not sufficient for (5).
 THE JOURNAL OF LEGAL STUDIES

Inequality (6) says that A is willing to trade a state 1 dollar (state 1 being the
state where he wins in court) for fewer state 2 dollars than is B. Thai is, state
2 dollars are relatively more valuable to A than to B. Referring back to (5),
this means that if equality is to obtain, A must perceive the probability of
state 1 to be greater than B does because this gives more relative weight to
state 1 dollars for A and less relative weight to state 2 dollars for B.
If the two parties have the same assessment of the probabilities, then
inequality (6) must hold and both the parties will be better off by a settle-
ment out of court. Indeed a settlement is the only way to obtain Pareto
optimality in this case. To see why, first note that a side payment of some
amount S from B to A will give A the same wealth in both states and
similarly for B. Thus, the ratio of marginal utilities will be unity for both
parties, and this is a point of Pareto optimality. The argument is completed
by the following proposition which shows that there can be no points of Pareto
optimality involving different payments for two states.

Proposition:
If PA - pB, then a payment of S (negative or positive) from B to A in
state 1 and a payment S 2 (negative or positive) from B to A in state 2 is not
Pareto optimal if SL =-S2 and the parties are risk averse.

Proof:
Let p be the agreed upon probability; then because of risk aversion

pU(WA ±S1-CA) -(1 -p)UA(WA+S2- CA) <,UA(WA±- CA)

and

pUB(Wn - S1 - CB) + (1 - p)UB(WB - S2 - CI) - UB(WB - 9- CB)

where

S=pS + (1 - p)S2.

The inequalities show that a payment of S from B to A is preferrable to both,

and hence, if SI = S2 the "risky" deal is not Pareto optimal.
We have established that if the costs of making an agreement out of
court are not excessive, parties agreeing on the probability of the outcome will
make an out-of-court settlement. In fact, if the costs of making an out-of-
court settlement are lower than the costs of going to court, both parties further
gain by the net saving in court costs. It is worth emphasizing, however, that
risk aversion is enough to derive these conclusions so that out-of-court settle-
ments might occur even if court costs were zero.
 THE ECONOMICS OF LEGAL CONFLICTS

Some Qualifications
Before going on to examine what happens when parties disagree on the
probabilities, some further comments on the out-of-court settlement are in
order.
1. The "all or nothing" nature of the example is not essential. In the ap-
pendix, it is shown that both parties will benefit from settlement when the
amount of the award is uncertain and when countersuits are allowed as long
as all probabilities are agreed upon.
2. It was assumed that the probabilities did not depend on the amount in
suit or the resources spent by the parties on legal services. When these depen-
dencies are introduced the problem takes on the characteristics of a two-person
game, as in the classical duopoly model. Let PA and pB be functions of the
amount of the suit X and the amount of resources spent on legal services, RA
and RB, by A and B respectively. Then the probabilities may be written
pA(RA, Rn, X) and pB(RA, RB, X). A key difficulty with this specification is
that the optimal expenditure of resources on legal services for each defendant
will depend on the behavior (or assumed behavior) of the other defendant.
In the absence of special behavioral assumptions, there is no way to assure
that the implied cycle of actions and reaction will, in fact, converge to an
equilibrium solution. 8 An interesting aspect of this problem is that the implied
uncertainty that A has about B's case (and vice versa) may put off an out-
of-court settlement until the case gets near to trial or until the evidence is
actually presented in court. Once each party has a clear understanding of the
other party's case, differences in probability estimates presumably narrow and
may vanish. This may explain in part why many cases remain unresolved while
waiting for a court appearance but then get resolved in an out-of-court
settlement shortly before or during the trial.
3. The optimality of the out-of-court settlement holds quite generally for
risk-averse individuals who agree on the probability of the outcome. Wealth
differences and differences in utility functions do not affect this conclusion
and neither does the value of the probability. Nothing has been said, how-
ever, about the amount of the settlement. The Pareto optimal point is not
unique, and the particular settlement will depend on the relative bargaining
skills of the parties. The agreed-upon probability is clearly of importance to
their bargaining strengths. If the plaintiff is very likely to win, the settlement

8 To resolve this problem, William M. Landes supra note 3, and Richard A. Posner,
supra note 4, assume that one of the parties does not react to the resources spent on
legal services by the other. In the situation discussed here, however, the risk aversion
and limited wealth of the disputants may be assumed to give an upper limit on the legal
services expenditures of each and this may be enough to establish the existence of an
equilibrium solution. For a discussion of similar issues in the context of duopoly models
see Kalman Cohen & Richard Cyert, Theory of the Firm ch. 12 (1965).
 THE JOURNAL OF LEGAL STUDIES

will be close to the amount of the suit; if he is not likely to win, the settle-
ment will be small. Incidentally, this helps to explain why the plaintiff will not
choose an arbitrarily large amount X for the suit in the hope of increasing his
settlement. Presumably, the larger is X for any given loss, the smaller the
probability that the court will decide in his favor. This result depends,
however, on the "all-or-nothing" nature of the present model; in actual
practice the court can decide on a payment other than X, in which case X
may be chosen arbitrarily. The multiple-outcome case is dealt with in the
appendix.
The Effect of Insurance
In this analysis, it has been assumed that the individuals deal with each
other either directly or through legal counsel and assume all costs and risks
themselves. It is no doubt more realistic to assume, at least in personal injury
and property damage cases, that the dispute will involve at least one company
insuring the respective parties. However, this does not alter the results in any
significant manner.
To see why, consider the situation in which A sues B for $X. If B is in-
sured, he turns the case over to his insurance company, which then deals
directly with A. If the insurance company maximizes expected profits, it
deals with A in a manner similar to the way in which two individuals will
handle the dispute and the case is likely to be settled out of court. 9
A more interesting possibility is that A carries private insurance which will
cover his losses no matter what the court decides. In such a situation, it might
seem that A will always take the case to court, since, while he has some chance
of gain by doing so, he cannot lose because his personal insurance covers his
losses in any circumstance. This outcome does not appear reasonable, however,
as the analysis of the following possibilities indicates: (1) A can collect a
payment directly from his insurance company and also sue B. Since our
analysis holds no matter what A's wealth level is, the payment from the
insurance company will not affect the way in which he deals with B, and the
case is not likely to come to court. (2) The insurance company guarantees
payment only if it is allowed to handle the litigation. In this case, the in-
surance company will negotiate with B out of court to get some payment
from him which will reduce its out-of-pocket costs to A.

Differences in Probability Assessments

In the discussion of equation (5), it was noted that equality can obtain and
the parties may go to court if probability assessments differ. Equality in (5)

9 Some writers assume that firms, like individuals, maximize expected utility rather than
profits. I do not find much appeal in this assumption, but it does not affect this con-
clusion if the company is risk averse.
 THE ECONOMICS OF LEGAL CONFLICTS

requires the probabilities to differ in a very specific way, however, and there
is no reason to suppose the actual disagreement yields the necessary "balanc-
ing" probabilities. What happens, then, when differences in probabilities can-
not be eliminated and equality does not obtain in (5)?
Several cases can be distinguished, but perhaps the most interesting is where
A's estimate of his chance of winning is lower than B's estimate of that
probability. In this case, pA < PB and

I - PA U'A(WA - CA) i - PB U'B(WB - CB)

PA U'A(WA + X - CA) PB U'B(WB- X -- CB)
Geometrically, condition (7) is represented by the point C in Figure 4. At this
point the slope of A's indifference curve is greater in absolute value than
the slope of B's indifference curve. As is intuitively obvious from the state-
ment of the problem, there are out-of-court settlements that are preferred

State I
A Wins
0

w +X-C A -

WA + -sC A -
WA+ ACA E

WA-CA WA+-SB-CA

State 2
A Loses
FtGURa 4
 THE JOURNAL OF LEGAL STUDIES

by both parties. The set of such settlements is represented in Figure 4 by the

segment of the 450 diagonal line that lies in the wedge-shaped area described
by the indifference curves A and B. It is perhaps somewhat less obvious that
none of the out-of-court settlements is Pareto optimal. 10 To see why, consider
the settlement indicated by the point D along the 450 line. At this point (in
fact, at any point on the 45 ° line), inequality (7) is

-PA> I-P] (7a)

PA Pu
because the disputants have different probability estimates. Clearly, any
point in the wedge-shaped area between A' and B' represents a further im-
provement over this out of court settlement. When (7a) holds, the set of
Pareto optimal points is the line 0, which is below the 450 line. This means,
in effect, that the disputants will go to court and at the same time make a
"bet" on the outcome of the trial. A more appealing interpretation of this
finding is that a "probabilistic" settlement is made wherein A agrees to
reduce his claim against B in the event that A wins in court in exchange for
an agreement that B will pay A something if A loses in court.1 ' In the diagram,
SA represents the reduction of A's claim and SB the payment from B to A.
A curious consequence of this disagreement in probability estimates is that
A will get more by losing the case than by winning, the difference representing
the amount by which the "bet" exceeds the certain settlement out of court.
Obviously, the disputants are paying the court costs in order to engage in this
bet. If these court costs are high enough, they may cancel the gains from
betting, in which event an out-of-court settlement will be more attractive.
The fact that A can gain more by losing than by winning introduces "moral
hazard" into the problem. If A can influence the outcome of the trial by
presenting a stronger or weaker case, then the larger SB, the more A will be
motivated to lose the case. If B recognizes this, he may adjust his estimate of
A's winning (i.e., pu) downward and this will tend to push the solution to
one that is closer to an out-of-court settlement. This element of moral hazard
also suggests, in the case of the criminal justice system analyzed by Landes,
that when the prosecutor is relatively pessimistic about winning the case, he
will be more inclined to take a bribe in exchange for presenting a weak case
against the defendant. In this situation, the payment of the bribe (perhaps

10 This corrects an error in an earlier draft of this paper where it is implied that
since the out-of-court settlement is better than the initial allocation, it will occur. William
M. Landes, supra note 3, also makes this error. In a full analysis, one cannot assume,
as Landes and I did, that necessary conditions for settling out of court are sufficient.
"1Subsequent to deriving this result, I was interested to learn that lawyers sometimes
arrange pretrial agreements which make the amount of the claim dependent on the vote
of the jury; a unanimous jury will result in a higher payment than a split jury.
 THE ECONOMICS OF LEGAL CONFLICTS

in the form of a contribution to a political campaign fund) would depend on

the case's being dismissed or the defendant's being found not guilty.
This analysis deals with only one of the possibilities that can obtain when
the disputants disagree about the probabilities of the outcome of a trial. In
all such cases, however, the disputants will not be able to attain a Pareto
optimum with an out-of-court settlement and the case will go to court unless
the court costs are higher than the disputants are willing to pay to engage in
an optimizing "bet."

Risk Preference
The assumption of risk aversion has been critical to the analysis of this
problem. What happens when risk preference obtains for one or both parties?
To answer this question, we refer to Figures 5 and 6. In Figure 5, both

BA

State 1
A Wins
AA

State 2
A Loses
FxUo' S
 THE -JOURNAL OF LEGAL STUDIES

State 1
A Wins

C
A
B
450

State 2
A Loses
FIoa 6

parties are risk preferrers (globally) and the indifference curves are concave
instead of convex. (This concavity can be shown by examining the sign of
the derivative of (3) with respect to W 2.) The point of tangency indicated
by C is not Pareto optimal in this case because we can shift A's utility to a
higher level by moving along B (B's indifference curve). This movement can
continue until the point D, where further increases in A's utility would de-
crease B's utility. Hence points of Pareto optimality such as D will have the
characteristic that one of the parties will lose all of his wealth in one of the
states. In Figure 5, B loses all of his wealth when A wins. This situation will
obtain even if the disputants agree on the probabilities, because they prefer
gambles. However, if they agree on the probabilities, they would be better
off not going to court and making the bet on the outcome of a suitably chosen
 THE ECONOMICS OF LEGAL CONFLICTS

random number generator, because the court costs could then be avoided. 12
If A is risk averse and B is a risk preferrer (globally), then the solution
will depend on the relative curvature (the second derivatives) of the indiffer-
ence curves of A and B. In Figure 6, A, who is risk averse, has sufficient
curvature in his indifference curve to make the out of court settlement at C,
a point of Pareto optimality, even though B is a risk preferrer. If the labels
were switched (curve A being labeled B and vice versa), then C would not
be a point of Pareto optimality and the case might go to court.

IV. SoM OTHER APPLICATIONS

It is easy to see that the above analysis could be applied to many situations
where the conflicting parties are in a position to affect the riskiness of the
outcome. In this part, we examine some of these applications.

The Behavior of Administrative Agencies

Richard A. Posner has applied a model similar to the one discussed here
in the analysis of how the FTC and similar administrative agencies allocate
their budget among cases.' s He finds that the tendency of agencies to devote
a "disproportionate" amount of their budgets to relatively minor cases is
consistent with rational, utility maximizing behavior despite frequent criticism
to the contrary.
The settling of disputed tax claims by the Internal Revenue Service is
another place where one would except the model to apply. The model sug-
gests that most disputes between taxpayers and the IRS would be settled out
of court and that, in the out-of-court settlements, the IRS would settle for
less than the amount of the alleged claim. Because of a recent lawsuit in
Seattle, the IRS has been forced to release some internal operating statistics
that bear on this question. These data show that the IRS settles out of court
more than 90 per cent of the disputed cases. Table 1 indicates the percentages
of the initial claim at which the settlement was made. Observe that the
percentage of the settlement decreases as the amount of the alleged claim
increases. This is consistent with the additional resources that the taxpayer
can be expected to devote to the case as the potential loss to him increases.
It is also consistent with Posner's finding that the agency will devote more

12 This analysis suggests that global risk preference is not a good description of the
disputants' utility function because risk preferrers would gain from such gambles ir-
respective of law suits. If risk preference is local, the parties would bargain to a position
of local risk aversion. Local risk preference around the 45* line could work against out
of court settlements.
13
Richard A. Posner, supra note 4.
 THE JOURNAL OF LEGAL STUDIES

TABLE 1
PERCENTAGE OF AGENT-ALLEGED EXTRA TAX DUE THAT APPELLATE DIVISION OF IRS
SETLED FOR IN CASES NOT GOING To COURT

Alleged Extra Tax 1971 1972

$1-$ 999 68% 67%
1,000- 9,999 63 61
10,000- 49,999 50 53
50,000- 99,999 44 44
100,000- 499,999 38 39
500,000- 999,999 37 35
$1 million and over 24 34
Source: Internal Revenue Service, internal documents reported in the Wall Street Journal, February 5,
1973, at 25, cols. 3 and 4.

resources to smaller cases when it is maximizing expected gains from a fixed

budget.

Labor-Management Relations
The economics of bargaining between labor unions and managers of firms
involves many complex issues, but certain aspects of compulsory arbitration
can be analyzed in terms of the model of this paper.' 4 One interesting pro-
posal on the handling of "national emergency" labor-management bargaining
(i.e., railroad and longshoremen strikes) is to have the disputants each enter
a "final" offer to the arbitrator, who would then choose one or the other of
them for the final settlement.1 5 An appealing aspect of this "forced choice"
method of dealing with critical labor negotiations is that it would strongly
motivate a prearbitration settlement as long as the disputing parties agreed
on the probabilities of the arbitrator's final choice. In order to achieve this
objective, it would be desirable to have the arbitrator make clear the relevant
probabilities before the final choice is made, and then to make his choice by
using an appropriate random number generator. The prearbitration settlement
would then be obtained by allowing the parties a period of "final" negotiation
between the time the probabilities were announced and the final choice was
made. The probabilities could be determined on the basis of the reasonable-
ness of the offers to avoid capricious offers and possibly inequitable settle-
ments. The objective of getting a prearbitration settlement per se, however,

14 For more detail on the theory of bargaining, see John G. Cross, The Economics of
Bargaining (1969).
15 The Nixon Administration actually proposed such legislation. See S. 560, 92d Cong.,
1st Seas. § 219 (1971). S. 560 and similar bills were withdrawn late in 1972 because of
strong opposition from the AFL-CIO and the Teamsters Union. See National Emergency
Disputes, 1971-72, Hearings before the Subcomm. on Labor of the Senate Comm. on
Labor and Public Welfare, 92d Cong., 1st and 2d Sess., pt. 1, at 11, for the text of
S. 560 § 219.
 THE ECONOMICS OF LEGAL CONFLICTS

could be achieved with any probabilities as long as these are known to the
disputants in advance of the "drawing."

The Problem of Social Cost

In an important article, Ronald H. Coase discussed the problem of social
cost and showed that, if the pricing system works without cost, the allocation
of resources will be that which maximizes the value of production irrespective
of where liability is placed. In Coase's words,
It is necessary to know whether the damaging business is liable or not for the
damage caused since without the establishment of this initial delimitation of rights
there can be no market transactions to transfer and recombine them. But the
ultimate result (which maximizes the value of production) is independent of the
legal position if the pricing system is assumed to work without cost.' 0

It is interesting to note the parallels between the model of this paper and
Coase's result. Suppose that individual A engages in a production activity
that has value k to him, and that this activity affects individual B's welfare
because of the noise it produces. The noise could be avoided by the intro-
duction of additional noise-muffling equipment with market price V. We
assume that V is less than k and also less than the monetary value of the
welfare loss to B. If B threatens to take A to court in an effort to stop him
from making the noise, the expected utility to A and B will be
pUA(WA -k) + (1 - p) UA(WA - k- V)
and
P UB(WB - V) + (1 - p) UB(WB)
respectively, where (1 - p) is the agreed upon probability of the court's
assigning liability to A. These expected utilities reflect the assumption that B
will pay for the muffling equipment if A is not found liable. This situation is
strictly analogous to the model discussed in Part III of this paper: provided
that the parties agree on p (and are risk averse), they will settle the dispute
between themselves. The settlement will give them utilities
UA(WA + k - SA)
and
UB(WB - SB)

respectively, where SB - V - SA. In other words, the out-of-court settlement

17
will divide the costs of muffling the noise between them.
16 Ronald H. Coase, The Problem of Social Cost, 3 J. Law & Econ. 1, 8 (1960).
17 Because B will always find it desirable to have the noise muffler installed, he will
pay the difference between its cost and A's contribution.
 THE JOURNAL OF LEGAL STUDIES

This result parallels the Coase theorem in that, no matter how the liability
is assigned (even if they split the cost between themselves), the production
activity continues and the noise-muffling equipment is installed. The value of
production is maximized because k - V > 0. The example shows, however,
that the optimal solution can be obtained without the actual assignment of a
property right by the court. What is critical is agreement on the relative
probability of the court's potential action. The analysis thus provides an
economic rationale for, and illustrates the economic significance of, the con-
cept of legal precedent. If the courts acted arbitrarily and without precedent,
an enormous number of property right disputes would have to be resolved in
court, and this would be a costly and serious impediment to the efficient
organization of economic activity. The concept of legal precedent is in effect
a means to provide stationarity over time to the probabilities and hence to
increase the opportunities for out of court agreements.' 8
Low court costs may therefore actually be harmful to efficient economic
organization, since cases in which there are small disagreements about the
probabilities of the outcomes are less likely to be settled out of court than if
court costs are high. Thus a combination of high court costs and clear-cut
legal precedents can operate to offset the transactions costs that may prevent
the price system from working smoothly.

V. SUMMARY AND CONCLUSIONS

This paper has developed a fairly general framework for dealing with the
economics of risky conflicts. The model was used to examine how individuals
engaged in civil suits will behave and it was found that a critical component
in the motivation of such individuals to settle out of court is agreement on
the probabilities of the court's action. The actual probabilities affect the terms
of the settlement but do not affect the optimality of an out of court settle-
ment. This conclusion holds for any risk-averse utility function and for any
wealth level of the individuals. 19 If differences in probability estimates exist,
then the individuals will go to court and make what is in effect a "side bet"
on the outcome. However, costs of going to court may outweigh the utility
of the bet, and the out-of-court settlement will then be preferred.
The analytical techniques used here have been applied in related areas such
as the criminal courts (in a paper by William M. Landes) and the behavior
of administrative agencies (in a paper by Richard A. Posner). Other applica-
18 Stationarity is the critical aspect of legal precedent in this context because the actual
probabilities are irrelevant to the final use of resources aside from the effect of wealth
transfers (Coase theorem).
19 If the utility function of the individual is state-dependent (i.e., if he gets utility from
going to court, per se) he may, of course, go to court even if there is agreement on the
probabilities.
 THE ECONOMICS OF LEGAL CONFLICTS

tions, such as the problem of labor-management disputes, also can be viewed

in terms of the model discussed here. The model also has close parallels to
the problem of social cost which was analyzed by Professor Coase. An inter-
esting aspect of the model discussed in this paper is that it provides a way
for the Coase theorem to apply before the court assigns liability for damages
in a specific instance. If the disputants in a "social cost" problem agree on
the probabilities of the court's action, the economic effect is equivalent to the
actual assignment of property rights.

APPENDIX
A. SOME COMPARATIVE STATICS

The condition for Pareto optimality that must obtain when the case goes to
court (equation (5)) can be used to derive some useful relationships among the
parameters of the problem. Let
U'A(WA - CA)
U'A(WA + X - CA)

and

U'B(WB - CB)
h =
U'n(WB - X - CB)

We have seen that hA - h1 > 0 which means that PA > PB if going to court is
Pareto optimal. Clearly, if hA increases relative to hB, then PA must increase rela-
tive to PB if (5) is to continue to hold. To simplify the analysis of relative changes
inPA and PB, we will suppose that pB is fixed, and hence the relative change is
determined by the change in PA- We consider the effect on PA of the following
changes in parameters of the model:
(a) An increase in X. The derivatives of hA and hB with respect to X are
0hA
2
-= -U'A(WA - CA) [U'A(WA + X - CA)] U"A(WA+X- CA) >0

ahB = U'B(WB - CB) [U'B(WB - X - CB))]-2 U"(W] - X - CB) < 0.

ax
ahA ahB
Hence, X - - > 0 and PA must increase relative to PB if (5) is to hold. In
other words, the more optimistic A is relative to B about his chances of winning,
the larger must be the amount of the suit, X, if the Pareto optimal solution given
by (5) is to hold.
(b) An increase in WA or WB. The sign of the derivative of hA with respect to
WA is greater than, equal to, or less than zero, according as
 THE JOURNAL OF LEGAL STUDIES

-U"A(WA + X - CA) > -U"A(WA - CA)

U'A(WA + X - CA) < U'A(WA - CA)

The ratio - is Pratt's measure of risk aversion. 20 Hence, if risk aversion is

U,
increasing, h is positive and PA must increase if (5) is to hold. This also says

that as wealth increases, a plaintiff with decreasing risk aversion will require a less
optimistic relative assessment of his chances of winning if (5) is to hold.
The sign of the derivative of ha with respect to W] is greater than, equal to, or
less than zero according as:
-- B(WB - X - Cn) > -U"B(WB - CB)

UB(W n - X - CB) < U'B(WB - CB)

Thus, when B's risk aversion is decreasing, is positive and the difference
between hA and h B decreases as hB increases. To maintain (5) when WB increases
and B has decreasing risk aversion, PA must get smaller relative to PB.
(c) Increases in CA and CB. The sign of the derivative of hA with respect to CA
is greater than, equal to, or less than zero according as
-U"A(WA - CA) > -U"A(WA + X - CA)
U'A(WA - CA) < U'A(WA + X - CA)

The sign of the derivative of hR with respect to CB is greater than, equal to, or
less than zero according as
-U"B(WB - CB)> -U"B(WB - X - CB)

U'B(WR - CB) < U'B(WB - X - CB)

Thus, if the parties have decreasing relative risk aversion hA- hB increases with
an increase in CA or CB. An increase in court costs for either party requires that
PA increase relative to PB if (5) is to continue to hold.
(d) An increase in aversion to risk. If g is a concave function then VA = g(UA)
is a transformation of UA that exhibits more risk aversion than UA. Clearly,
- V'A(WA - CA) g'[UA(WA - CA)] UA(WA - CA)
V'A(WA + X - CA) g'[UA(WA + X - CA)] U'A(WA + X - CA)

_ g'[UA(WA - CA)]
g'[UA(WA + X - CA)] hA.
Since g is concave, the coefficient of hA in the last expression is larger than one.
Hence, h* > hA and the transformation requires that PA increase relative to PB if
'A A
2
OJohn W. Pratt, Risk Aversion in the Small and in the Large, 32 Econometrica
122 (1964).
 THE ECONOMICS OF LEGAL CONFLICTS

(5) is to continue to hold. Using the same analysis, it can be shown that when B
becomes more averse to risk. PA must increase relative to PB if (5) is to hold.
These results are comparative static in nature. They do not imply that the in-
dicated changes in PA relative to PB will necessarily occur, but simply identify
what relative changes are needed to maintain equality in expression (5). Some
caution must be used in interpreting these results, however. Suppose that WA
increases but PA and pB remain unchanged. Equation (5) no longer holds, but this
does not mean the case will be settled out of court since we have seen that a
court settlement will always be involved when PA # pI so long as the court costs
(CA and CB) are not excessive. What must happen in such a situation, then, is
that other variables will be changed to find a new Pareto optimum. These changes
might involve bargaining over X in exchange for "side bets" as discussed above in
the section on differences in probability assessments.

B. THE MULTIPLE-OUTCOME MODEL

In the body of the paper it was assumed that the court made a single decision
either for or against the plaintiff and awarded the exact amount of the suit if the
plaintiff won. Moreover, we ignored the possibility of a countersuit by the defendant
and also assumed that the court costs were independent of the outcome of the case.
Here we extend the model by relaxing these assumptions.
If U(.) is a concave function defined on the real numbers, then by Jensen's
inequality,
n ti

for n 2,Z-l'i-1, and Vi' 0.

i-i

For i = A, B, let j = 1,..., n denote each of the n possible outcomes of taking

the case to court; then
Xj is the payment of B to A (which may be negative if A files a countersuit)
when outcome j prevails;
C1 is the cost incurred by i when outcome j occurs; and
&'i is i's assessment of the probability the outcome j will occur.
Using this notation,
n

E oJn UB(WB - Xj - CjB)

J=1

is the expected return to B if the case goes to court and

Z J
A
UA(WA + Xj - CI
A)

J-i
 THE JOURNAL OF LEGAL STUDIES

is the expected return to A if the case goes to court. Following the expected-utility
hypothesis, there exist wealth levels ZA and ZB such that individual i would prefer
any certain wealth level in excess of Z, to the uncertain prospect of going to court.
Hence, B will be willing to pay an amount S to A to avoid going to court as long
as WB - ZB > S. Similarly, A will accept a payment S rather than go to court if
S > ZP - WP. Hence, a necessary condition to settle out of court is
WB + WA > ZA + ZB. (Al)
By Jensen's inequality and monotonicity,
n

ZB Z
E OjB (WR - Xi - Cj')

and
n

ZA J _OA (WA + X 5 - CJA).

j=1
Thus, a sufficient condition for the inequality (Al) is
n n n

WB +WA> W + WA- E (6 1B - E8A) X- E EB CjB - 1:E)A CJA

j=1 j=1 j=1

or, rewriting,

B A
(0
B
-- OEA) Xj + L
E j B Cj + OjA C > 0. (A2)
j=1 5=1 j=l

If the utility functions are strictly concave, then the inequality in (A2) need not
be strict. Hence, we have
Proposition: When O B = OA for all j and when CB, CA 1 0 for all j, parties with
strictly concave utility functions will gain from settling out of court.