** JOINT TORTFEASORS**

Lewis A. Kornhauser

*Alfred and Gail Engelberg Professor of Law*

*New York** University School** of Law*

Richard L. Revesz

*Professor of Law*

*New York** University School** of Law*

© Copyright 1999 Lewis A. Kornhauser and Richard L. Revesz

** **

**Abstract**

This chapter compares the properties of joint and several liability with those of

non-joint liability. It considers three criteria: deterrence, settlement inducing

properties and fairness. The analysis is performed for both full and limited

solvency.

The central conclusion is that neither rule dominates the other. With respect

to deterrence, the relative desirability of the two rules depends on the levels of

solvency of the defendants. In contrast, with respect to settlements and fairness,

the comparison turns on the correlation of the plaintiff’s probabilities of succes

against the defendants.

*JEL classification: *K1, K2, K4

*Keywords: *Joint and Several Liability, Settlement, Joint Tortfeasors, Hazardous

Waste Regulation

** **

**1. Introduction**

The law and economics analysis of joint tortfeasors focuses on the comparison

between joint and several liability and non-joint (several only) liability. Part A

provides a brief background of the legal regimes. Parts B, C and D compare,

respectively, the deterrence, settlement effects and fairness of the two rules.

** **

**A. Legal Regimes**

** **

**2. Legal Regimes**

The choice between joint and several liability and non-joint liability arises in

situations in which the plaintiff’s injury results from the actions of multiple

parties. Under joint and several liability, if the plaintiff litigates against many

defendants and prevails against only one, he can recover his full damages from

that defendant; if the plaintiff prevails against all defendants but some are

insolvent, he can recover his full damages from the solvent defendants; and if

the plaintiff prevails against all defendants and all are solvent, he can

nonetheless choose to recover his full judgment from any defendant or to a

recover a portion from each. In contrast, under non-joint liability, the plaintiff

can recover from a losing defendant only the share of the damages attributable

to that defendant.

For joint and several liability, the legal regime needs to be specified further.

As shown in Kornhauser and Revesz (1993), the various choices presented

below can affect the economic analysis of the consequences of joint and several

liability.

First, a right of contribution permits a defendant that has paid a

disproportionately large share of the plaintiff’s damages as a result of the

application of joint and several liability to obtain compensation from a

defendant that has paid a disproportionately small share of these damages.

Absent a right of contribution, such reallocation is not possible. Second

contribution shares are usually determined either pro rata (equal division

among the defendants) or by reference to comparative fault.

Third, the question of an appropriate set-off rule arises when the plaintiff

settles with one defendant and litigates against the other. Under the pro tanto

set-off rule, the plaintiff’s claim against the non-settling defendant is reduced

by the amount of the settlement. In contrast, under the apportioned share set-off

rule (sometimes referred to as a proportional set-off rule), the plaintiff’s claim

against the non-settling defendant is reduced by the share of the liability

attributable to the settling defendant.

Fourth, under the pro-tanto set-off rule, when one defendant settles and the

other litigates and ultimately loses, the question arises whether the settling

defendant is protected from contribution actions. Fifth, the legal regime must

also specify whether settling defendants are entitled to bring contribution

actions against defendants who settled for less than their share of the liability.

Sixth, under the pro-tanto set-off, if the plaintiff enters into an inadequately

low settlement with one defendant, the other defendant is responsible for the

shortfall if he litigates and loses. To protect the interests of non-settling

defendants, courts sometimes require ‘good faith’ hearings on the adequacy of

settlements.

Seventh, if the plaintiff joins all the joint tortfeasors in a single suit, his

claims against all of them will be adjudicated in the same proceeding. If the

plaintiff chooses not to join all the tortfeasors as defendants, the question arises

whether a named defendant can join another tortfeasor as a third-party

defendant. Otherwise, the named defendant would have to file a separate action

for contribution after the adjudication of his liability to the plaintiff.

** **

**B. Deterrence**

** **

**3. Deterrence: Several Remarks**

We compare here the deterrence effects of joint and several liability and

non-joint liability, when coupled with both rules of negligence and strict

liability. We perform the comparison first for cases in which the defendants are

fully solvent (Kornhauser and Revesz, 1989) and then consider the effects of

limited solvency (Kornhauser and Revesz, 1990).

We develop our argument by reference to a model in which two firms, Row

and Column, dump hazardous wastes at a single landfill. The actors benefit

from this dumping because the wastes are the byproduct of profitable economic

activity. At some time in the future, these wastes may leak into the environment

and cause serious damage; we think of this damage as the cost of cleaning up

the landfill and the surrounding area affected by the release. We take the

damage function to be convex (the additional damage caused by one unit of

waste increases with increasing amounts of waste in the landfill).

The expected damage of a release is a ‘social’ loss because it does not fall

directly on the dumpers absent a legal provision shifting the liability to them.

Instead, it falls on the victim that would have legal responsibility for the

cleanup, or, alternatively, that would suffer the consequences if the problem

were left unattended. Under our model, each dumper chooses the amount of

waste that it will dump.

The socially desirable amount of waste is that which maximizes the social

objective function: the sum of the benefits derived by the actors minus the social

loss. An economically rational firm, however, does not make this decision

based on the social objective function. Instead, it seeks to maximize its private

objective function: the benefit that she derives from the activity that leads to the

production of the waste minus whatever share of the social loss the legal regime

allocates to her.

We model a joint and several liability regime with contribution shares

determined by reference to the amount of waste dumped. (Other rules are

considered in Landes and Posner, 1980; Kornhauser and Revesz, 1989;

Tietenberg, 1989 and Wright, 1988, pp. 1169-1179.) We assume that a

plaintiff, say for example the government, sues both defendants in the same

proceeding and we exclude the possibility of settlement (the deterrence effects

of joint and several liability when settlement is possible are analyzed in Kahan,

1996 and Spier, 1994).

** **

**4. Full Solvency: Negligence**

We assume in the case of negligence that the standard of care will be chosen

at the level that maximizes social welfare; departures from the social optimum

in setting the standard of care are considered in Kornhauser and Revesz (1989,

pp. 862-870). For expositional convenience, we assume that negligent actors

are liable only for the losses that would have been prevented through due care

(in this example, for the additional losses that result if a firm dumps more than

the socially optimal amount, rather than the socially optimal amount). We show

in Kornhauser and Revesz (1989) that essentially the same results hold if

negligent actors are responsible for the full losses (even ones that would have

occurred with due care).

Under these circumstances, joint and several liability will produce the

socially optimal result. If one of the actors, say Row, is non-negligent, it would

not be rational for Column to be negligent. If this actor were contemplating

dumping more than the standard of care, she would face liability for the full

increase in the resulting damage. If the standard of care is set at the social

optimum, the increased benefits that this actor would obtain through negligent

conduct would be less than the increase in the damage for which she would be

liable. Thus, assuming that one of the actors is non-negligent, the remaining

actor will be non-negligent as well. This argument shows that it is a Nash

equilibrium for each actor to meet her standard of care.

We now show that this efficient Nash equilibrium is unique. Consider

whether it would be rational for both actors to be negligent. These actors will,

jointly, face liability equal to the full increase in the resulting damage. If

negligent action on the part of these actors were preferable to non-negligent

action for each of them, then the total social welfare would exceed that

attainable when all actors meet the standard of care which, once again, is not

possible if the standard of care is set at the social optimum. Thus, regardless of

how the increased damage was allocated between the defendants, at least one

of them would have to pay more than the increased benefit that it obtained by

acting negligently. An equilibrium in which both actors are negligent is

therefore not possible.

The analysis is different for a non-joint liability rule, under which a

negligent defendant would not be liable for the share of the damage attributable

to the non-negligent defendant. Instead, the negligent defendant would be liable

for an amount proportional to waste that it had dumped. Assume that Row is

non-negligent and that Column is contemplating dumping more than the

standard of care. Column would then pay only a fraction of the increase in

damage. Under this apportionment rule, the remainder of the increase would

be attributable to Row and would be unrecoverable by the plaintiff as a result

of Row’s lack of negligence. Thus, in this situation, non-joint liability leads to

under-deterrence.

** **

**5. Full Solvency: Strict Liability**

The analysis is different for strict liability. Under strict liability, as long as both

actors are fully solvent, there is no difference between joint and several liability

and non-joint liability. Strict liability ensures that the victim is compensated for

the full damage and thus the question whether the victim will have to bear the

share of the damage caused by the actions of non-negligent defendants does not

arise.

Assume that Row is dumping the optimal amount of waste (the amount that

would have met the standard of care if a rule of negligence had been in effect)

and that Column is contemplating whether to dump more than this amount.

Such a decision on the part of the Column would, of course, increase the

damage to the victim. Column would, in turn, be liable for a larger share of the

damage, as he would pay in proportion to the amount of waste that he dumped.

As long as the damage function is convex, however, the increase in Column’s

liability is less than the increase in the social loss. Thus, Column’s decision to

dump more than the socially optimal amount has the effect of increasing Row’s

liability as well. As a result of this externality, strict liability leads to

under-deterrence, regardless of whether it is coupled with joint and several

liability or non-joint liability.

Miceli and Segerson (1991) consider a modification of the strict liability

rule that does in fact lead to efficiency both in terms of the level of care adopted

and of entry into the activity. Under their formulation, each actor is responsible

for the marginal damage that it causes. This rule, coupled with the assumption

of convex costs, implies that the total payments from two defendants would

exceed the plaintiff’s actual damages.

** **

**6. Limited Solvency**

Here, each defendant is defined not only by her benefit function (the rate at

which her generation of waste is transformed into net benefits) but also by a

fixed solvency, which represents the actor’s available amount of assets to offset

her share of the social loss. Under this formulation of the problem, the actors

cannot shed their solvencies over time. We present here the analysis for strict

liability, which makes it possible to explain the basic intuitions. The

comparison of joint and several liability and non-joint liability under

negligence when the actors have limited solvency is presented in Kornhauser

and Revesz (1990).

Consider a situation under which Row’s solvency is zero and Column’s

solvency is infinite, and that both firms are otherwise identical. The liability

rule thus transmits no deterrence incentive to Row. Row will therefore dump

up to the point at which any additional benefit (in terms of reduced costs of

production) from additional dumping becomes zero. This amount, which we

call , is greater than, the amount that Row would have dumped if both

defendants had been infinitely solvent. Note that, as a result of the

underdeterrence caused by strict liability, discussed above, is in turn larger

than *x****, the socially optimal amount of dumping by Row.

Under joint and several liability, because Row has no solvency, Column will

be responsible for the whole liability and will dump an amount *a *(smaller than

*x****), which is the optimal amount of dumping by Column conditional on Row

being insolvent. The equilibrium is thus (, *a*). If Column is not infinitely

solvent, there are two possible equilibria: ( *a*), if Column’s solvency is

greater than a critical solvency which we call *s**j *or (, ), if Column’s

solvency is lower.

In contrast, under non-joint liability, Column is not responsible for the

whole liability, but only for its proportional share. If Column has infinite

solvency, it will dump *b*, an amount larger than *a*, though smaller than *x****.

Here, too, there are two possible equilibria if Column is not infinitely solvent:

(, *b*), if Column’s solvency is greater than a critical solvency which we call

*s**nj *or (, ), if Column’s solvency is lower. Because for any level that it

dumps Column faces less liability under a rule of non-joint liability, over a

larger range of solvencies it chooses to act as if it were infinitely solvent rather

than wholly insolvent. Thus, *s**nj *is smaller than *s**j*. Table 1 summarizes the

relevant equilibria.

** **

** **

In region C in Table 1, joint and several liability is therefore preferable to

non-joint liability. From a social welfare perspective, an equilibrium at (, *a*)

is preferable to an equilibrium at (, *b*). When one actor is generating , joint

and several liability makes the other actor see the full social cost of its actions,

whereas non-joint liability does not. Thus, *a *is the optimal response by Column

to Row’s choice of .

In region B, however, the reverse is true. Joint and several liability induces

Column to act in the same manner that it would if it were wholly insolvent,

dumping *x**H*, whereas non-joint liability induces Column to act in the same

manner that it would if it were infinitely solvent, dumping *b*. Thus, in this

region, non-joint liability has better social welfare properties. (Of course, in

region A, both rules have the same properties.)

This discussion illustrates that, when solvency is limited, neither rule

dominates the other. (The same is true under negligence (Kornhauser and

Revesz, 1990.) The intuition behind this result is that Row’s insolvency creates

a ‘domino’ effect, leading Column, under certain circumstances, to act as if it

were insolvent as well. Because under joint and several liability Column is

responsible for a greater proportion of the total harm, the range under which

this ‘domino’ effect occurs is greater. In a model in which an actor’s probability

of insolvency is independent of the other actor’s solvency (or probability of

insolvency), a ‘domino’ effect is not possible and the results are different

(Watts, 1998).

** **

**C. Settlements**

** **

**7. Settlement: Basic Framework**

The basic framework for the analysis of the impact of joint and several liability

on settlements is set forth in Kornhauser and Revesz (1994a), which deals with

fully solvent defendants, and Kornhauser and Revesz (1994b), which deals with

potentially insolvent defendants. The discussion here proceeds by reference to

a numerical example, as in Kornhauser and Revesz (1993, 1995), which serves

to illustrate in a straightforward manner the game-theoretic interactions

generated by the competing rules.

We model the following rule of joint and several liability. First, there is a

right of contribution among defendants found jointly and severally liable.

Second, in contribution actions, the relevant shares are determined by reference

to the amount of waste dumped. Third, following a settlement, the plaintiff’s

claim against the nonsettling defendants is reduced by the amount of the

settlement (a pro tanto set-off rule); the effects of different formulations of the

apportioned share set-off rule are analyzed in Kornhauser and Revesz (1993,

p. 465-469) and Klerman (1996). Fourth, a settling defendant is protected from

any contribution actions. Fifth, a settling defendant can bring contribution

actions against non-settling defendants. Sixth, there is no detailed judicial

supervision of the substantive adequacy of settlements. Seventh, the claims

involving the joint tortfeasors are litigated together in a single proceeding.

Kornhauser and Revesz (1993) show that the results derived here are robust to

many changes in the legal regime governing joint and several liability.

To perform the comparison between joint and several liability on the one

hand and non-joint liability on the other, we consider a situation in which the

plaintiff has a claim of $100 against two defendants, Row and Column, each

equally at fault. All the parties are risk neutral. We assume initially that the

defendants are sufficiently solvent that they can satisfy the plaintiff’s judgment.

Later, we consider the effects of limited solvency.

The probability that the plaintiff will prevail against each defendant is 50

percent. All the parties have accurate information about this value and the costs

of litigation are zero. As shown in Kornhauser and Revesz (1994a), the results

derived here hold even if the two defendants were not equally at fault, if the

plaintiff’s probability of success were not 50 percent, and if litigation costs are

not zero.

With respect to the relationship between the plaintiff’s probabilities of

success against the two defendants, we consider two polar situations. In the

first, these probabilities are independent. Thus, the plaintiff’s probability of

success against one defendant is 50 percent regardless of whether the plaintiff

has prevailed against, lost to, or settled with, with the other defendant.

In the second case, the probabilities are perfectly correlated. Thus, if the

plaintiff litigates against both defendants, it either prevails against both (with

a probability of 50 percent) or loses to both (also with a probability of 50

percent).

The parties may either litigate or settle the claim. Settlement negotiations

have the following structure. The plaintiff makes settlement offers to the two

defendants. Row and Column decide simultaneously whether to accept these

offers. (The effects of different offer structures are examined in Donohue, 1994;

the effects of ‘Mary Carter’ agreements between the plaintiff and a subgroup

of defendants is analyzed in Bernstein and Klerman, 1995). We assume that

defendants’ costs of coordinating their actions are sufficiently high that they act

non-cooperatively. The plaintiff then litigates against the non-settling

defendants, if any. We adopt the convention that, if a party is indifferent

between settlement and litigation, it settles.

The central conclusion of our analysis is that the comparison of the

settlement inducing properties of joint and several liability and non-joint

liability depends critically on the correlation of the plaintiff’s probabilities of

success. When these probabilities of success are independent, joint and several

liability unambiguously discourages settlements, relative to non-joint liability.

When, in contrast, these probabilities are perfectly correlated, joint and several

liability has a more complex effect: it encourages settlement when the litigation

costs are low, but may discourage settlements when these costs are high

(Kornhauser and Revesz, 1994a). Earlier analyses had focused, implicitly, only

on perfectly correlated probabilities (Easterbrook, Landes, and Posner, 1980;

Polinsky and Shavell, 1981).

A recent experimental study of auditors’ liability considers a more

complicated correlation structure under which the probabilities are perfectly

correlated if the manager is not liable (because under the securities’ laws the

auditor then cannot be liable) but independent if the manager is liable (Dopuch,

Ingberman, and King, 1997).

** **

**8. Non-Joint Liability**

The analysis of the choice between settlement and litigation under non-joint

liability is straightforward. The plaintiff’s expected recovery from litigation is

$50: she has a 50 percent probability of obtaining $50 from each defendant;

each defendant’s expected loss is therefore $25. Absent litigation costs, the

plaintiff and the defendants are indifferent between litigation and settlement.

For any level of litigation costs, settlement becomes preferable. For example,

if each party’s litigation costs were $5, the plaintiff’s expected recovery from

litigation would be only $20 and each defendant’s expected loss would be $30.

The plaintiff and each defendant would prefer any settlement between $20 and

$30 to litigation.

The result that under non-joint liability the parties are indifferent between

settlement and litigation in the absence of litigation costs and prefer to settle for

any level of litigation costs does not change if the defendants have limited

solvency. Say, for example, that Row’s solvency is only $20. Then, in the

absence of litigation costs, the plaintiff and Row are indifferent between

litigation and a settlement for the plaintiff’s expected recovery of $10 (a 50

percent probability of recovering Row’s solvency of $20). For any level of

litigation costs, the parties prefer to settle. Thus, while limited solvency affects

the expected value of the plaintiff’s claim as well as amount at which the case

would settle, it does not affect the choice between settlement and litigation.

** **

**9. Joint and Several Liability**

* *

*Independent Probabilities*

As a consequence of joint and several liability, the plaintiff recovers his full

damages not only if he prevails against both defendants but also if he prevails

against one and loses to the other. When the plaintiff’s probabilities of success

against the two defendants are independent, each of four different scenarios

carries a probability of 25 percent: that the plaintiff prevails against both

defendants, that the plaintiff prevails against Row and loses to Column, that the

plaintiff prevails against Column and loses to Row, and that the plaintiff loses

to both defendants. In the first three cases, carrying an aggregate probability of

75 percent, the plaintiff recovers his full damages of $100. Thus, his expected

recovery from litigating with both defendants is $75. In turn, each defendant’s

expected loss is $37.50. We proceed by analyzing a situation in which litigation

costs are zero.

A risk-neutral plaintiff will not accept a settlement with both defendants

that yields less than $75, but would find acceptable an aggregate settlement for

$75 or more. What would happen if the plaintiff made settlement offers to the

two defendants for $37.50 each, so that its aggregate recovery was equal to the

expected recovery of litigating against both defendants? If one defendant, say

Row, accepted the offer, would the other defendant accept it as well? Column

would accept the settlement only if his expected loss from litigation is at least

$37.50. Under the pro tanto set-off rule, Column’s exposure in the event of

litigation is reduced to $62.50: the plaintiff’s damages of $100 minus Row’s

settlement of $37.50. But Column faces only a 50 percent probability of losing

the litigation. Thus, in light of Row’s settlement, its expected loss from

litigation is only $31.25.

It therefore follows that if the plaintiff were to make offers of $37.50 to each

defendant, at least one of them would reject the offer. The plaintiff’s expected

recovery would then be $68.75 (Row’s settlement of $37.50 plus an expected

recovery of $31.25 from litigating against Column). This amount is lower than

the plaintiff’s expected recovery from litigating against both defendants. Thus,

the plaintiff would never make offers of $37.50 to each defendant. Similar logic

establishes that no other pair of offers would give the plaintiff an expected

recovery of at least $75 and yet be acceptable to the two defendants. Also, there

is no scenario under which the plaintiff would receive an expected recovery of

at least $75 by settling with one defendant and litigating against the other.

This phenomenon has two sources (1) the surplus that the plaintiff obtains

from litigation as a result of joint and several liability when its probabilities of

success against the defendants are independent, and (2) the benefit that a

non-settling defendant receives from the set-off created by the plaintiff’s

settlement with the other defendant.

If the plaintiff were litigating against only one defendant rather than two,his

expected recovery from litigation would be $50 rather than $75: he would have

a 50 percent probability of recovering from that defendant its full damages of

$100. Similarly, as we have indicated, if the plaintiff were litigating against

two defendants under non-joint liability, his expected recovery would also be

$50: he has a 50 percent probability of recovering $50 from each of the

defendants. Finally, if the plaintiff were litigating against two defendants under

joint and several liability but his probabilities of success against the defendants

were perfectly correlated, he would also have an expected recovery of only $50

(a 50 percent probability of recovering his full damages if he prevails against

both defendants).

As a result of the surplus that the plaintiff obtains from litigating under

joint and several liability when the probabilities of prevailing are independent,

the plaintiff will not accept from one defendant a settlement that is too low even

if he intends to litigate against the other. Say, for example, that the plaintiff

accepted a settlement of $0 from Row and litigated against Column. His

expected recovery would then be only $50 (a 50 percent probability of

recovering $100); the settlement with Row would have reduced his expected

recovery by $25. If the plaintiff accepted a settlement of $10 from Row, his

expected recovery from litigating with Column would be $45 (a 50 percent

probability of recovering $90), for a total expected recovery of $55; the loss

from the low settlement with Row would be $20.

So as not to lose his surplus, the plaintiff would thus have to demand a

sufficiently high settlement from Row. But a settlement that is sufficiently

desirable for the plaintiff to accept confers a benefit upon Column. If, for

example, the plaintiff were to settle with Row for $25, Column’s expected loss

from litigation would be $37.50 - the same expected loss as if Row litigated.

Any higher settlement with Row reduces Column’s expected loss. We have

already shown that a settlement with Row for $37.50 reduces Column’s

expected loss from $37.50 to $31.25, giving him a benefit of $6.25. In order to

recover $75, the plaintiff would have to obtain from Row a settlement of $50

(which would leave an expected recovery from Column of $25 and confer upon

Column a benefit of $12.50). Row, however, would not agree to such a

settlement because, given that Column litigates, he is better off litigating as

well and facing an expected loss of only $37.50.

We have thus illustrated why the plaintiff cannot capture the full benefit of

Row’s settlement if his probabilities of success are independent. Part of this

settlement confers an external benefit upon Column. It is this externality that

stands in the way of settlement. Indeed, the only way that the plaintiff can

obtain the full benefit of a defendant’s payment is by litigating, because if he

settles, part of the benefit accrues to the other defendant, reducing the

plaintiff’s expected recovery from litigation.

The role of joint and several liability in discouraging settlements is not

limited to the case in which litigation costs are zero. The externality described

above also impairs the possibility of settlement when litigation when litigation

costs are positive but lower than a particular threshold.

* *

*Perfectly Correlated Probabilities*

The problem changes considerably when the plaintiff’s probabilities of success

against both defendants are perfectly correlated. If the plaintiff litigates against

both defendants, he either prevails against both (with a probability of 50

percent) or loses against both (also with a probability of 50 percent). His

expected recovery from litigation is $50 rather than $75; each defendant’s

expected loss is then $25.

In the case of perfectly correlated probabilities, the plaintiff will settle with

both defendants. It is easy to see that the plaintiff will settle with at least one

of the defendants. Say that the plaintiff settles with Row for $10. He faces a 50

percent probability of recovering $90 from Column, and his total expected

recovery is $55 - $5 higher than his recovery from litigating against both

defendants. The effect of this settlement is to give the plaintiff $10 with

certainty, but reduce his expected recovery from litigation by $5. As a result,

settlement with one defendant and litigation against the other is always more

attractive to the plaintiff than litigation against both defendants. Unlike the

case of non-joint liability, where the parties are indifferent between settlement

and litigation when litigation costs are zero, here there is a positive surplus that

the plaintiff and a defendant can divide if a settlement takes place.

It is also easy to show that, for the example that we are analyzing, the

plaintiff in fact settles with both defendants, for $25 and $37.50, respectively.

Given that Row settles for $25, Column’s expected loss through litigation is

$37.50 (a 50 percent probability of paying the plaintiff’s damages of $100

minus Row’s settlement of $25), and would therefore accept a settlement for

that amount. Moreover, given that Column settles for $37.50, Row’s expected

loss through litigation is $31.25 (a 50 percent probability of paying the

plaintiff’s damages of $100 minus Column’s settlement of $37.50), and would

therefore prefer to settle for $25. The same argument establishes that the

plaintiff would be no better off settling with one defendant and litigating

against the other.

We show elsewhere that, for perfectly correlated probabilities, the plaintiff

settles with both defendants if their shares of the liability are sufficiently

similar, and settles with one defendant--the one with the larger share of the

liability - and litigates against the other if the defendant’s shares of the liability

are sufficiently different (Kornhauser and Revesz, 1994a).

* *

*The Effects of Limited Solvency*

As indicated above, under non-joint liability, the limited solvency of the

defendants does not affect the choice between settlement and litigation. The

situation is different under joint and several liability. We consider first how

limited solvency would affect the choice between settlement and litigation if the

plaintiff’s probabilities of success are independent. If one of the defendants, say

Row, has limited solvency, the plaintiff nonetheless litigates against both

defendants if this solvency is above a threshold. For example, if Row’s solvency

is $80 and the plaintiff litigates against both defendants, his expected recovery

is $37.50 from Column but only $32.50 from Row (with a probability of 25

percent, the plaintiff prevails against both defendants and recovers $50 from

Row, and, also with a probability of 25 percent, the plaintiff prevails only

against Row and recovers Row’s solvency of $80 rather than its full damages

of $100). In contrast, if the plaintiff settles with Column for $37.50, Row’s

expected loss from litigation, and consequently the maximum settlement that

it would offer, would be only $31.25 (a 50 percent probability of paying the

plaintiff’s damages of $100 minus Column’s settlement of $37.50).

When Row’s solvency is sufficiently low, however, the plaintiff settles with

both defendants. Consider the case in which Row’s solvency is $40. If the

plaintiff litigates against both defendants his expected recovery is $60 (with a

probability of 25 percent, he prevails only against Column and recovers $100;

with a probability of 25 percent, he prevails against both and recovers $40 from

Row and $60 from Column; and with a probability of 25 percent, he prevails

only against Row and recovers $40). In turn, Row’s expected loss is $20 and

Column’s expected loss is $40.

If the plaintiff offered Row a settlement of $20, his expected recovery from

Column is $40 (a 50 percent probability of recovering his damages of $100

minus Row’s settlement of $20), and Column would be willing to settle for this

amount. In turn, if the plaintiff offered Column a settlement of $40, his

expected recovery from Row is $20 (a 50 percent probability of recovering his

solvency of $40), and Row would be willing to settle for this amount. Thus, as

in the case of non-joint liability, when the solvency of one of the defendants is

sufficiently low and litigation costs are zero, the parties are indifferent between

settling and litigating.

In summary, the result that joint and several liability discourages

settlements when the plaintiff’s probabilities of success are independent holds

over a range of solvencies. A similar analysis (Kornhauser and Revesz, 1994b)

establishes that, when the plaintiff’s probabilities of success are perfectly

correlated, joint and several liability promotes settlements over a range of

solvencies. For solvencies below a given threshold, however, joint and several

liability has the same settling- inducing properties as non-joint liability. The

relevant results are summarized in Table 2.

** **

** **

** **

**D. Fairness**

** **

**10. Fairness: Several Remarks**

The comparison of the relative fairness of joint and several liability and

non-joint liability raises four principal issues (Kornhauser and Revesz, 1995).

Three of these issues arise when the defendants are fully solvent: (1) the size

of the plaintiff’s expected recovery when she litigates against the defendants;

(2) the division of the plaintiff’s recovery among litigating defendants; and (3)

the effects of settlements. A fourth issue arises when the defendants have

limited solvency: the division of the burden of insolvency between the plaintiff

and the solvent defendant (Wright, 1992). A question relevant to all four issues

is whether one should assess fairness *ex ante *(in terms of the parties’ expected

payments) or *ex post *(in terms of the actual payments in particular cases). We

largely confine our remarks here to ex ante assessments.

** **

**11. Size of the Plaintiff’s Recovery**

First, as indicated in Part C, except when the plaintiff’s probabilities of success

against the defendants are perfectly correlated, joint and several liability leads

to a higher expected recovery than non-joint liability. Recall the example in

which the plaintiff’s damages are $100 and her probabilities of success against

each of the defendants are 50 percent, and the defendants are equally at fault

and fully solvent. The plaintiff’s expected recovery is $50 under non-joint

liability, $50 under joint and several liability when the plaintiff’s probabilities

of success are perfectly correlated, and $75 under joint and several liability

when the plaintiff’s probabilities of success are independent. (In the range

between independence and perfect correlation, the plaintiff’s recovery is

between $50 and $75.)

Thus, except when the plaintiff’s probabilities of success are perfectly

correlated, an effect of joint and several liability is to transfer resources from

the defendants to the plaintiff. The fairness consequence of this transfer

depends upon why the plaintiff’s probability of success against each of the

defendants is only 50 percent. It could be that the defendants are in fact liable

but that the plaintiff has difficulty in proving their liability. In this case, joint

and several liability is attractive on fairness grounds because it brings a

defendant’s expected liability closer into line with the harm that it caused.

Alternatively, it could be that there is true uncertainty about whether the

defendants are liable, and that this uncertainty is captured by the 50 percent

probability. Then, joint and several liability is undesirable because it increases

a defendant’s expected liability beyond the level of the harm the defendant

caused.

** **

**12. Division of the Plaintiff’s Recovery**

The second issue concerns the allocation of expected liability among litigating

defendants. From this perspective, joint and several liability performs badly: it

places a disproportionate burden on the defendant with the smaller share of the

liability, except when the plaintiff’s probabilities of success are perfectly

correlated. Consider an example in which, instead of being equally at fault,

Row and Column are 25 percent and 75 percent at fault, respectively; the

plaintiff’s probabilities of prevailing against each of the defendants remains at

50 percent and these probabilities are independent. There are then four possible

scenarios, each carrying a probability of 25 percent:

1. the plaintiff prevails against both defendants and collects $25 from Row and

$75 from Column;

2. the plaintiff prevails against Row and loses to Column, and collects $100

from Row;

3. the plaintiff loses to Row and prevails against Column, and collects $100

from Column; and

4. the plaintiff loses to both defendants and does not recover anything.

Thus, Row pays $25 with probability 25 percent and $100 with probability 25

percent; her expected liability is then $31.25. In turn, Column pays $75 with

probability 25 percent and $100 with probability 25 percent, and her expected

liability is $42.75. Thus, while Row’s contribution to the harm is only one-third

that of Column’s, her expected liability is about three-quarters that of

Column’s.

The preceding example shows that this disproportionate effect stems

exclusively from the fact that under joint and several liability the plaintiff might

prevail against the defendant with the lower responsibility for the harm but lose

against the other defendant, and that the defendant with the lower responsibility

is then required to pay the plaintiff’s full damages. In contrast, under non-joint

liability (and under joint and several liability when the plaintiff’s probabilities

of success are perfectly correlated), each defendant’s expected liability is

proportional to its responsibility for the harm.

** **

**13. The Effects of Settlements**

The possibility of settlements introduces a third fairness issue, also by placing

a disproportionate burden on the defendant with the smaller share of the

liability. Indeed, for the legal regime analyzed in Part C, which employs a pro

tanto set-off rule, each defendant settles for the same amount, even when their

shares of the harm are different. Consider the example in which the litigation

costs are sufficiently high that they induce the parties to settle, and in which the

plaintiff makes take-it-or-leave-it offers to the defendants.

The largest settlement that Row will accept, *S**R*, conditional on Column

settling for *S**C *(which is less than the plaintiff’s damages *D*) is given by

where *p *is the plaintiff’s probability of success against each defendant, *t *is each

defendant’s litigation costs, and *D *are the plaintiff’s damages. Similarly, the

largest settlement that Row will accept, *S**C*, conditional on Column settling for

*S**R *(which is less than the plaintiff’s damages *D*) is given by

Thus,

As a result, when litigation costs are sufficiently high that the parties settle

despite the independence of the plaintiff’s probabilities of success, the plaintiff

extracts from each defendant an equal settlement, regardless of the differences

in the defendants’ shares of the harm.

In contrast, recall that under non-joint liability, each defendant’s expected

liability is proportional to its responsibility for the harm. The plaintiff, if she

made take-it-or-leave-it offers, could extract from each defendant in settlement

this amount plus the defendant’s litigation costs. If each defendant’s litigation

costs are independent of their share of the liability, the defendant with the

smaller share will pay a disproportionate amount, but it will be less

disproportionate than what she would have paid under joint and several

liability.

** **

**14. Division of the Burden of Insolvency**

The fourth fairness issue arises if one of the defendants has limited solvency.

Our assessment of fairness here is neither fully *ex ante *nor fully *ex post*. A fully

*ex ante *perspective would consider the likelihood that each defendant would

become insolvent; instead our discussion assumes that one defendant is already

insolvent. On the other hand, our discussion is not fully *ex post *because we

assess fairness in terms of expected litigation (and settlement) outcomes.

We have studied elsewhere how the shortfall caused by the limited solvency

of one defendant is allocated between the plaintiff and the remaining solvent

defendant under joint and several liability (Kornhauser and Revesz, 1994b).

That study revealed that, over a broad range of solvencies, the plaintiff bears

the full shortfall, and it is never the case that the full shortfall is borne by the

solvent defendant. This conclusion challenges the accepted wisdom that, under

joint and several liability, the burden of one defendant’s insolvency falls

exclusively on its co-defendants (Sugarman, 1992).

The reason for the entrenchment of this erroneous view may be that judges

and commentators implicitly consider only the situation in which the plaintiff’s

probabilities of success are perfectly correlated and the plaintiff litigates against

both defendants. Then, any shortfall caused by one defendant’s limited solvency

is borne by the other defendant. If, however, the correlation of the probabilities

is less than perfect, the plaintiff’s expected recovery is reduced because it might

prevail only against the defendant with limited solvency. Moreover, the focus

on litigation overlooks the fact that settlement might occur.

Thus, under joint and several liability the shortfall caused by one

defendant’s limited solvency is generally shared between the solvent defendant

and the plaintiff. In contrast, as shown in Part C, under non-joint liability, the

full shortfall is borne by the plaintiff.

In summary, joint and several liability performs worse in terms of fairly

allocating liability among defendants but does not necessarily perform worse

in terms of fairly allocating liability between the plaintiff on the one hand, and

the defendants on the other.

** **

**15. Conclusions**

In sum, from the perspectives of inducing deterrence and inducing settlements,

and promoting fairness, there is no dominant relationship between joint and

several liability and non-joint liability. From a deterrence perspective, the

comparison between the two rules turns on the levels of solvency of the

defendants. In contrast, from settlement and fairness perspectives, the

comparison turns on the correlation of the plaintiff’s probabilities of success

against the defendants.