GOLDBERG, Circuit Judge:
This case raises the interesting and knotty question of what credit, if any, should be given a nonsettling section 1983 defendant because of a pretrial settlement between the plaintiff and a codefendant joint tortfeasor. The issue is complicated in this case by the facts that the settling tortfeasor was also being sued for other torts and that the amount of the settlement was not segregated by claim. We hold that the dual policies of section 1983, compensation and deterrence, allow credit for a settlement by a joint tortfeasor in proportion to the amount of the injury caused by that tortfeasor; thus, if a settling tortfeasor is found to have caused one-third of the plaintiff’s damage, a nonsettling tortfeasor is entitled a credit of one-third of the amount of damages found by the jury, regardless the amount of the settlement. Because the district court applied a different rule, we reverse and remand.
I. INTRODUCTION
A. Facts
The facts of this case are relatively simple and not particularly relevant to any of the legal issues involved. Plaintiff Willie Dobson, who worked for an auto parts company in Arkansas, came to Houston for a convention of automobile parts distributors. During the course of his stay, he and several fellow conventioneers went into a Denny’s restaurant at about 7:00 p.m., where they had dinner and a few drinks. Later in the evening most of group returned to their adjacent hotel, but Dobson and another man stayed discussing business. Later in the evening the two had another meal and Dobson’s companion left. Dobson retired to the men’s restroom.
The assistant manager became worried by the length of time Dobson was spending in the men’s room. Fearful that Dobson might be trying to avoid his fiscal duty to Denny’s, the assistant manager asked some Houston policemen then present in Denny’s to check up on Dobson. The policemen, went into the men’s room and apprehended Dobson. He was arrested and put in jail for the night.
The next morning Dobson was released from jail. He returned to his hotel and then proceeded to the hospital. At various times between Dobson’s arrest in the Denny’s men’s room and his release the next morning Dobson had been severely beaten by the Houston police. These injuries caused his hospitalization in Houston, subsequent medical treatment, and various other consequential damages.
B. Procedural History
On July 7,1977, Dobson filed suit against Camden, two other police officers, the City of Houston, the mayor, and the police chief under 42 U.S.C. § 1983 (1976). The essence of the suit was that policeman Camden and the two other officers had used excessive force in arresting Dobson, violating his constitutional rights. On January 11, 1978, Dobson amended his complaint to include Denny’s, Inc. as a codefendant. The claims against Denny’s included pendant state claims for false arrest and malicious prosecution as well as a claim under section 1983 that Denny’s had conspired with the offi[761]*761cers to violate Dobson’s constitutional rights.
Denny’s settled with Dobson for $30,000 prior to trial, and the case proceeded to trial against the remaining defendants. Before the case was submitted to the jury, the trial court granted a directed verdict in favor of the police chief and the mayor. The court submitted issues to the jury on liability of the city and the three officers and on nine different elements of damages flowing from the use of excessive force. No issues were submitted regarding Denny’s liability or damages for false arrest or malicious prosecution. The jury found Camden alone liable.
The jury apparently decided to award Dobson $25,000 one way or another, but it was not sure how to allocate the $25,000 among the nine elements of damages submitted to it. Taking a shortcut, the jury simply awarded one-ninth of the $25,000 under each of the nine elements of damages. The jury specifically declined to award punitive damages. The jury’s division was more numerically accurate than legally meritorious; there was no evidence in the record to support some of the items for which they awarded damages. On appropriate motion by defendant Camden, the court reduced those damages by $3,864.04, with a resulting verdict of $21,135.96.
Camden also moved for a reduction in the amount of the judgment on the verdict based on Dobson’s settlement with Denny’s. The judge ruled that a pro tanto — dollar-for-dollar — reduction would violate the policies of section 1983. Instead, the court adopted a kind of pro rata credit based on Rose v. Associated Anesthesiologists, 501 F.2d 806, 808 (D.C.Cir.1974). This pro rata concept did not “depend on a mechanical count of number of defendants,” id. at 808 n. 5, but rather grouped the defendants by theory of liability. The court placed the defendants in four groups. Because Denny’s, as the only private party conspirator, comprised one of the four groups and had settled, the judge reduced the verdict by twenty-five percent to reflect the settlement, leaving $15,851.97. Dobson was also awarded $6,340.79 in attorney’s fees based on an evidentiary hearing.
C. Issues on Appeal
Camden on appeal argues that Dobson’s settlement with Denny’s should have resulted in a dollar-for-dollar reduction in the amount awarded by the jury, leaving Dob-son with a take-nothing judgment. Second, in regard to attorney’s fees, Camden claims that the moral victory of a take-nothing judgment does not make Dobson a prevailing party entitled to attorney’s fees. Alternatively, Camden claims that the attorney’s fees award should also be subject to a dollar-for-dollar reduction based on the Denny’s settlement.
In response to Camden’s arguments Dob-son makes an argument that, if correct, would mean that no setoff was justified. Dobson argues that the claim against Denny’s for malicious prosecution, false arrest, and conspiracy, was separate from the claim against Camden for excessive force. Thus, no setoff could be given for the settlement. Alternatively, Dobson argues that the trial court’s roughly equitable reduction was required by the policies of section 1983. Dobson did not cross-appeal, however, so even if we agree with his argument, we cannot increase the judgment in his favor.
D. Plan of Attack
One of the initial questions we face in deciding this case is determining what source of substantive law we shall use. In making this decision we are guided by 42 U.S.C. § 1988 (1976), which states:
The jurisdiction in civil and criminal matters conferred on the district courts by the provisions of this Title ... for the protection of all persons in the United States in their civil rights, and for their vindication, shall be exercised and enforced in conformity with the laws of the United States, so far as such laws are suitable to carry the same into effect; but in all cases where they are not adapted to the object, or are deficient in the provisions necessary to furnish suitable remedies and punish offenses against law, [762]*762the common law as modified and changed by the constitution and statutes of the State wherein the court having jurisdiction of such civil or criminal cause is held; so far as the same is not inconsistent with the Constitution and laws of the United States, shall be extended to and govern the said courts in the trial and disposition of the cause ....
This statutory scheme establishes a three part test for finding substantive law. First, if federal law is neither deficient nor inapplicable, it will apply. Second, if federal law does not apply, state law does apply, unless, third, state law would be inconsistent with the Constitution and laws of the United States.
As mentioned earlier, the instant case is complicated by the fact that the settling tortfeasor was being sued for both joint and separate torts. The conspiracy claim against Denny’s made Denny’s potentially liable for the same elements of damages as were found against Camden. Thus, we cannot take the simple solution suggested by Dobson and hold that the settlement with Denny’s was entirely separate from the claims against Camden. Accordingly, we must address the complicated issues regarding settlements. For analytical simplicity, we shall first consider the problem in the context of pure joint tortfeasors. Once that question has been resolved we may then reintroduce the complexity of a settlement for both joint and separate torts. The issue before us, the effect of a settlement on a nonsettling tortfeasor, is merely the tip of the iceberg. It is impractical to consider the effect of a settlement without also considering the problem of contribution and, indeed, the very nature of joint liability. Accordingly, we must become voyageurs, embarking on the titanic journey of circumnavigating tort law.
One final preliminary matter is that of vocabulary. There are at least four schemes for measuring the effects of payment by one tortfeasor on the rights of a joint tortfeasor. The traditional common law approach is to do nothing. The principal behind the “no contribution” rule was that the courts should not assist a wrongdoer in spreading the consequences of his or her wrongful acts among other culpable parties. See, e.g., Merryweather v, Nixon, 101 Eng.Rep. 1337 (1799); W. Prosser, Handbook of the Law of Torts 307 (4th ed. 1971). A second approach is to allow pro tanto credit. Under this scheme, a settlement with a tortfeasor would produce a dollar-for-dollar reduction in an award against a nonsettling tortfeasor. See, e.g., Unif. Contribution Among Tortfeasors Act § 4 (1939). A third approach, the pro rata approach, apportions the economic consequences of a judgment equally among the joint tortfeasors. See, e.g., Unif. Contribution Among Tortfeasors Act § 4 (1955); see generally Note, Settlement in Joint Tort Cases, 18 Stan.L.Rev. 486 (1966) [hereinafter cited as Stanford Note]. This might better have been called a per capita approach, but in this opinion we will retain the more common term of “pro rata.” Finally, under the proportional approach, each tortfeasor is responsible for a share of the damages in proportion to that tortfeasor’s share of the fault. See Comment, Comparative Negligence, Multiple Parties, and Settlements, 65 Calif.L.Rev. 1264 (1977) [hereinafter cited as California Comment]. Thus, a tortfeasor found seventy-five percent at fault would be responsible for payment of seventy-five percent of the damages awarded.
With this vocabulary in place we shall proceed first, to see if federal law is deficient regarding settlements, second, to examine the state law applicable to this question, and third, to determine whether the result provided by state law is consistent with federal law. We find that federal law is deficient in this area and that application of state law would result in a pro tanto reduction in the judgment against the non-settling tortfeasor. This result, however, is inconsistent with the policies underlying section 1983. As a result, we must fashion a rule that is consistent with these policies. This approach leads us to adopt a rule of proportional reduction. Finally, we find that reintroducing the complexity of separate as well as joint torts has no effect on the rule of proportional reduction.
[763]*763II. SECTION 1988 — STATE LAW AND FEDERAL POLICIES
Initially we note that federal law is deficient on the effect of a plaintiff’s settlement with a joint tortfeasor on a judgment against a nonsettling tortfeasor. In Miller v. Apartments and Homes of N.J., Inc., 646 F.2d 101 (3d Cir.1981), the court found that federal law was not deficient on this subject. The Third Circuit, however, relied upon the power of federal courts to generate common law rules to fill out the interstices of federal statutory schemes. We do not follow that line of reasoning. Indeed, section 1988 would make no sense if Miller were correct on that point. Section 1988 assumes as a trigger the presence of a federal statutory scheme. If that same statutory scheme carried with it the power to generate common law rules, section 1988’s reference to a federal law deficiency would make no sense. Instead, we follow Johnson v. Rogers, 621 F.2d 300 (8th Cir.1980), in holding that federal law is deficient on the effect of a settlement with a joint tortfeasor on the liability of a nonsettling tortfeasor. Consequently, we must turn to an examination of the state law governing this issue.
A. State Law Regarding Settlements
Section 1988 requires that we apply state law to a section 1983 action where federal law is deficient, unless that state law conflicts with other federal law or policies. Having found that federal law is deficient on the effect of a settlement on the liability of a nonsettling tortfeasor, we must now examine state law. This case comes to us from the Southern District of Texas, so we must examine Texas law.
When a joint tortfeasor settles before trial, Texas applies a “one satisfaction” rule.1
“It is a rule of general acceptation that an injured party is entitled to but one satisfaction for the injuries sustained by him .... There being but one injury, there can, in justice, be but one satisfaction for that injury ....’’ Bradshaw v. Baylor University, 126 Tex. 99, 84 S.W.2d 703, 705 (1935). Accord T.L. James & Co. v. Statham, 558 S.W.2d 865, 868 (Tex.1977); McMillen v. Klingensmith, 467 S.W.2d 193, 196 (Tex.1971); Palestine Contractors, Inc. v. Perkins, 386 S.W.2d 764, 766 (Tex.1964). Application of this rule would result in a pro tanto credit for a settlement, which would ensure precisely one satisfaction for a wrong. In this case, assuming Denny’s settled for the same injury for which Camden was found liable, the Texas rule would result in a take nothing judgment. Denny’s settled for $30,000, the jury and judge found damages of $21,135.96, so Dobson has been fully compensated. Our next task is to determine whether the Texas pro tanto reduction rule conflicts with federal law.
B. Federal Law and Policies Regarding Damages
“In resolving questions of inconsistency between state and federal law raised under § 1988, courts must look not only at partic[764]*764ular federal statutes and constitutional provisions but also at the policies expressed in [them].” Robertson v. Wegmann, 436 U.S. 584, 590, 98 S.Ct. 1991, 1995, 56 L.Ed.2d 554 (1978) (quoting Sullivan v. Little Hunting Park, Inc., 396 U.S. 229, 240, 90 S.Ct. 400, 406, 24 L.Ed.2d 386 (1969)). Dobson’s action is under section 1983, so we must examine the policies of section 1983. “The policies underlying § 1983 include compensation of persons injured by deprivation of federal rights and prevention of abuses of power by those acting under color of state law.” Id. 436 U.S. at 590-91, 98 S.Ct. at 1995-96. See also City of Newport v. Fact Concerts, Inc., 453 U.S. 247, 101 S.Ct. 2748, 2759-60, 69 L.Ed.2d 616; Owen v. City of Independence, 445 U.S. 622, 651, 100 S.Ct. 1398, 1415, 63 L.Ed.2d 673 (1980); Carey v. Piphus, 435 U.S. 247, 256-57, 98 S.Ct. 1042, 1048-49, 55 L.Ed.2d 252 (1978). Thus, section 1983 embodies two policies: compensation and deterrence. These two goals have different theoretical underpinnings, which suggest different perspectives on section 1983 actions. First we will discuss compensation, then we will discuss deterrence.
1. Section 1983, Fairness, and Compensation. — Compensation, the first goal of section 1983, is easily understood, at least on a superficial level. The initial point of compensation is to make the plaintiff whole for damages caused by a violation of federal rights under color of state law. The jury need merely look back at the harm suffered by the plaintiff and place a dollar value on those injuries. When the plaintiff has received that dollar amount, the goal of compensation has been satisfied.
The principle of compensation carries with it two additional characteristics. First, it is concerned with fairness, and second, it is backward-looking. The reason for compensating a plaintiff under this theory, the reason Congress and the courts went to the effort of providing a remedy, is that it seems wrong for someone to suffer harm from a lawless action. That person deserves, in some moral sense, to be made whole. See, e.g., F. Harper & F. James, The Law of Torts 743 (1956); Epstein, A Theory of Strict Liability, 2 J. Legal Stud. 151 (1973); Fletcher, Fairness and Utility in Tort Theory, 85 Harv.L.Rev. 537 (1972). Because deterrence also relies in part on a measure of compensation, see infra Part I.B.2, we shall refer to the theory underlying section 1983’s goal of compensation as the “fairness theory.”
It is this same moral sense that suggests that on some occasions compensation alone is not enough. Some wrong actions under color of state law may be so egregious that our sense of justice is not satisfied by compensation alone. So, for example, if a deprivation of constitutional rights is prompted by malice, a jury might find that punitive damages were warranted. See City of Newport v. Fact Concerts, Inc., 453 U.S. 247, 101 S.Ct. 2748, 2759-60, 69 L.Ed.2d 616 (1981). Restatement (Second) of Torts § 908 (1979); W. Prosser, Handbook of the Law of Torts 9-10 (4th ed. 1971).
The second characteristic of this fairness theory is that it is backward looking. Acts that may or may not occur are hypothetical and not yet subject to a consideration of fairness. The inquiries into compensation and punitive damages, based on fairness, of necessity involve looking backward to determine what has happened.
This discussion provides a brief look at the characteristics of one of section 1983’s goals. It is a moral inquiry, involving concepts of fairness. It is also a retrospective inquiry, involving a rectification of past évents. This backward looking moral theory provides a rationale for asking the jury questions on both compensatory and punitive damages.
2. Section 1983 and Deterrence. — The second goal of section 1983 is deterrence. Again, on a superficial level the policy of deterrence is easily understood. Society does not want persons to have their federal rights violated under color of state law. If people considering doing such things know that they will have to pay money if they do [765]*765violate someone’s federal rights, they will be less likely to do so.2
Of course, a simple mechanism for deterring violations such as this would be to amend section 1983 to provide that violators will be drawn and quartered. This seems like a very powerful deterrent and might substantially reduce violations of federal rights under color of state law. Aside from problems relating to fairness, however, this solution also poses problems in the deterrence framework. A powerful deterrent such as drawing and quartering offenders might also deter worthwhile conduct — police officers might decide never to arrest anyone because the risks to the police officer would be too great. So, the deterrence rationale calls for neither too much nor too little deterrence; we need to find the right amount.
The “science” of economics can provide us with some insight as to how to reach the correct level of deterrence. Each activity produces costs and benefits. An action is appropriate, from this economic point of view, when the benefits outweigh the costs. If the person contemplating an action will reap the benefits but will not pay the costs, we have no assurance that the socially correct decision will be made. The solution to this problem is to force the decisionmaker to absorb the costs as well as the benefits of a given action. This concept is known as “cost internalization.” See, e.g., G. Calabresi, The Cost of Accidents 68-129 (1970).
Cost internalization provides us with a mechanism for reaching the correct level of deterrence for official misconduct. If people acting under color of state law know that they will bear the consequences of their actions, they will be deterred from violating a person’s federal rights, but will not be over-deterred. The “correct” level of deterrence will be established. The costs of a violation of a person’s federal rights are the damages suffered by that person as a consequence of the violation. Thus, under a deterrence theory, one of the questions put to the jury would involve the damage suffered by the plaintiff.
On some occasions, however, the actions of the violator may generate costs to people other than the plaintiff. See Webster v. City of Houston, 689 F.2d 1220, 1237-38 (5th Cir.1982). (Goldberg, J., concurring). For example, if the actions of a violator were particularly egregious, those actions might produce trauma within the society at large. Merely assessing the officer with the amount of damage suffered by the plaintiff will not adequately internalize all of the societal costs of the officer’s actions. Accordingly, to deter official misconduct adequately the jury should be asked in an appropriate case to levy punitive damages against the offender. See City of Newport, supra, 101 S.Ct. at 2760; Owen v. City of Independence, supra, 445 U.S. at 651, 100 S.Ct. at 1415; Carlson v. Green, 446 U.S. 14, 22, 100 S.Ct. 1468, 1473, 64 L.Ed.2d 15 (1980), Carey v. Piphus, supra, 435 U.S. at 257 n. 11, 98 S.Ct. at 1049 n. 11.
This discussion, though brief, serves to sketch the outlines of the theory underlying the deterrence goal of section 1983. That theory, in contrast to the fairness theory discussed earlier, is forward looking — it seeks to govern future activity. It is somewhat agnostic as to the moral elements involved, except as they provide a guide to what kind of • future activities should be promoted or deterred. It sets the appropriate level of deterrence by internalizing to the violator the costs of violating federal rights. Interestingly, this theory provides a separate and distinct justification for asking the jury the same two questions regarding compensatory and punitive damages that we asked under the fairness rationale.
8. Consistency of State Law with the Policies of Section 1988. — Section 1983, as just discussed, embodies two policies: compensation of plaintiffs and deterrence of wrongdoers. The Texas pro tanto rule is based upon one policy, the “one satisfaction rule.” See, e.g., Bradshaw v. Baylor Uni[766]*766versity, 126 Tex. 99, 84 S.W.2d 703 (1935). The Texas rule is designed to ensure compensation and nothing more. “[A]n injured party is entitled to but one satisfaction for the injuries sustained by him .... ” Id. 84 S.W.2d at 705. This rule rather clearly supports the section 1983 policy of compensation. But being designed to promote solely a policy of compensation, the Texas rule does not perform as well in the context of deterrence.
The basic problem of the Texas rule in deterring wrongdoers is quite evident from the facts of this case. If the Texas rule were applied to this case, Camden would pay nothing because Dobson has been (presumably) fully compensated by Denny’s. Section 1983 deters wrongdoing by imposing damages upon the wrongdoer. See supra Part II.B.2. A rule that removes the burden of damages from the wrongdoer certainly conflicts with the policy of deterrence. We must always be conscious that one of the functions of section 1983, if not the primary function, is therapeutic, seeking to eradicate the disease of violations of federal rights under color of state law. The one satisfaction rule, as this case well demonstrates, would tend to eliminate this constitutional medication. Accordingly, we hold that application of the Texas one satisfaction rule to this case would contravene the deterrence policy of section 1983, and thus is impermissible under section 1988.
C. Through the Looking Glass
We now have a problem. Federal law is deficient concerning the effect of settlements on section 1983 judgments, and state law conflicts with federal policies, so we may not apply state law. At this point section 1988 runs out of gas. The logical alternative, however, is to create federal common law to fill the void. It is interesting that, finally, we reach the same point that the court in Miller, supra, used as its starting point. We, too, resort to federal common law for a rule of decision, but only after following the route prescribed by section 1988. We emphasize the contrast between our position and that of the Third Circuit in Miller. We decide this case under federal common law not out of a belief that we have the general power to generate rules of decision, see, e.g., Texas Industries, Inc. v. Radcliff Materials, Inc., 451 U.S. 630, 101 S.Ct. 2061, 68 L.Ed.2d 500 (1981) (no common law power to create rule of contribution in antitrust), but rather, we apply federal common law pursuant to section 1988’s implicit grant of authority to do so when federal law is deficient and state law would conflict with federal law. See, e.g., Shaw v. Garrison, 545 F.2d 980 (5th Cir.1977), rev’d on other grounds, 436 U.S. 584, 98 S.Ct. 1991, 56 L.Ed.2d 554 (1978).
III. SECTION 1983 AND JOINT TORTFEASORS
At this point we have climbed through the rabbithole of section 1988 and are in the Wonderland of federal common law. Because federal law is deficient and state law contravenes the policies of section 1983, we must generate a rule of decision as best we can. As mentioned above, we find it impossible to discuss intelligently the issue of the effect of a settlement on a non-settling tortfeasor without detouring into the related areas of the purpose and nature of joint liability.
First, having examined the goals of section 1983, we shall ponder what treatment these goals suggest for joint tortfeasors. Second, we shall investigate the problem of a settling joint tortfeasor in light of section 1983’s goals of compensation and deterrence.
A. Section 1983 and Joint Tortfeasors
Now that we have a basic understanding of the dual goals of section 1983 and the theoretical concepts behind them, we can introduce the first level of complexity— joint tortfeasors. Again, we will first discuss the effect of multiple joint tortfeasors on the fairness, or compensation, goal of section 1983, and then on the deterrence goal.
1. Joint Tortfeasors and Fairness. —Fairness considerations have produced different theories for the treatment of joint [767]*767tortfeasors. The first treatment was holding joint tortfeasors jointly and severally liable for the entire consequences of their wrong. This treatment resulted from focusing on the need to compensate the plaintiff — by making each tortfeasor liable for the full amount, the risk of one tortfeasor being insolvent was placed on the joint tortfeasors rather than on the plaintiff. This treatment was coupled with two other doctrines. First, courts would not help one tortfeasor who had paid the entire judgment collect from another tortfeasor who had paid nothing — there was no right of contribution. See, e.g., Union Stockyards Co. v. Chicago, Burlington & Quincy R. Co., 196 U.S. 217, 25 S.Ct. 226, 49 L.Ed. 453 (1905); W. Prosser, Handbook of the Law of Torts § 50, at 305-07 (4th ed. 1971); Leflar, Contribution and Indemnity Between Tortfeasors, 81 U.Pa.L.Rev. 130, 130-34 (1932). Second,'if the plaintiff was not totally free from contributory negligence, the plaintiff would collect nothing. See Fleming, Foreword: Comparative Negligence at Last — By Judicial Choice, 64 Calif.L.Rev. 239, 240 (1976) [hereinafter cited as Fleming Foreword].
The intrinsic unfairness of these harsh common law rules has been steadily eroding. See, e.g., Gregory, Contribution Among Joint Tortfeasors: A Defense, 54 Harv.L.Rev. 1170 (1941); Leflar, Contribution and Indemnity Between Tortfeasors, 81 U.Pa.L.Rev. 130 (1932). The primary hydraulic force has been a realization that the world is not solely black and white, that shades of gray exist. Furthermore, not all bad consequences have a single, distinct cause. In the case of joint tortfeasors or a contributorily negligent plaintiff, the harm suffered by the plaintiff was caused by two or more parties. The fairness rationale places financial responsibility on a tortfeasor out of a moral feeling that the tortfeasor should pay for the damage he, she, or it caused. Accordingly, if a tortfeasor was
only a partial cause of an injury, that tortfeasor should pay for only a part of the damage. See California Comment, 65 Calif. L.Rev. at 1267-74.
This, in brief, is the fairness rationale for comparative fault. If an injury is caused by two or more parties the damages should be apportioned among the parties in proportion to their fault. Thus, under this fairness rationale a jury should be asked to apportion liability among all contributing parties, including the plaintiff when appropriate. This places on a plaintiff the risk that a tortfeasor may be insolvent.3 However, that risk is always on the plaintiff in the case of a single tortfeasor; furthermore, we see no reason for placing the financial burden of damage caused by an insolvent tortfeasor on a joint tortfeasor, who has been found by the jury not to have caused that damage, rather than on the plaintiff, who also has been found not to have caused that damage.
2. Joint Tortfeasors and Deterrence. —The key principle in deterrence is the economic idea of internalizing the costs of an action to the party that made the decision to act. The theory does not alter in the case of joint tortfeasors. All we need do is change the singular to plural — the costs of an action should be internalized to the parties that caused them. It further follows that each party should internalize the costs to the extent that that party caused them.
Under this deterrence idea of comparative causation the jury should apportion the cause of damage among all the potential candidates. Each party would then internalize the costs properly allocated to that party’s actions and the correct level of deterrence would be achieved. Again, under both the fairness and deterrence theories the same questions would be put to the jury-
[768]*768
B. Settlements and Joint Liability
The second complication, after introducing multiple tortfeasors, is the complication of a settling tortfeasor. Before inquiring into the problem of a settling joint tortfeasor we will first examine the issue of a settling single tortfeasor. Then we will look at a settling joint tortfeasor under both fairness and deterrence theories.
I The Settling Tortfeasor. — Most lawsuits are settled. Fleming Foreword, 64 Calif .L.Rev. at 247. A settlement generally represents a contractual agreement between the settling parties that a dollar amount passing between them is a satisfactory resolution of the contingent liabilities between the parties. The philosophy of settlements does not encompass, save in infrequent cases, elements of logic or fairness. The dollar figure is not a precise mathematically derived figure; rather it represents a melange of factors including the parties’ estimates of damages, the probability of recovery, costs of litigation, and preferences of the parties for avoiding uncertainty. See California Comment, 64 Calif.L.Rev. at 1275-82.
Most settlements are aleatory exercises. They are rarely perfectly prognosticated and they evolve almost as a literary “who done it.” When two parties settle in a bipolar lawsuit, no one would consider objecting to the settlement on the ground that the dollar figure reached unfairly overcompensates the plaintiff. What would have happened had the parties gone to trial is nothing more than a “might have been.” The “might have beens” were considered in reaching the settlement; the parties reached a bargain and they must live with it.
2. Settlements and Joint Tortfeasors. —The various techniques of handling settlement credits — pro tanto, pro rata, proportional — all have varying virtues and problems. The different techniques, however, all rest on different conceptions of joint liability. Under a theory of joint and several liability, the appropriate treatment would seem to be to provide a pro tanto credit. A plaintiff is entitled to one satisfaction, and the pro tanto credit ensures that.
The problems with a pro tanto credit arise from the fact that it provides no rational basis for allocating damages among the joint tortfeasors. See California Comment, 65 Calif.L.Rev. at 1273-74. Certainly, nothing prevents a collusive, low settlement with a less solvent tortfeasor, who might even be largely responsible for the harm, in exchange for that tortfeasor’s assistance in prosecuting a claim against the less responsible, more solvent joint tortfeasors. See California Comment, 65 Calif. L.Rev. at 1280-81; Stanford Note, 18 Stan. L.Rev. at 490-91. This problem has been recognized in those proposals allowing pro tanto credits, and the response is usually to require that settlements be in good faith. See, e.g., Unif. Contribution Among Tortfeasors Act § 4 (1955).
What we have found so far, however, is that joint and several liability does not mesh well with section 1983’s goals of fairness and deterrence. Under both fairness and deterrence theories we found that fault or causation should be apportioned among the parties and each party should be responsible for a proportional share of the damages. In essence, this scheme of comparative fault/comparative causation separates joint torts into multiple single torts. A settlement by one tortfeasor thus should have no effect on the liability of another tortfeasor- — each nonsettling tortfeasor will still be responsible for a proportional share of the damages.4
IV. PROBLEMS WITH COMPARATIVE FAULT/COMPARATIVE CAUSATION
A. Overcompensation
Camden argues strenuously that anything other than a pro tanto credit would result [769]*769in overcompensation for Dobson and would be unfair. Ignoring for the moment the added complexity of joint and separate torts in this case, we are unpersuaded that this claimed overcompensation introduces a problem in our comparative fault/comparative causation scheme. We will address this point, first, under a fairness theory and, second, under a deterrence theory.
1. Overcompensation and Fairness. —Camden’s overcompensation argument sounds most strongly in chords of fairness. Camden argues that it simply is not fair that Dobson receive a double recovery. Even under a fairness rationale, however, we see nothing wrong with a proportional credit for settlement.
First, in invoking compensation Camden neglects to consider carefully the fairness rationale underlying it. The fairness of a tortfeasor paying an injured party rests not solely on the fact that the injured party deserves compensation, but also on the fact that the tortfeasor deserves to pay.5 If another tortfeasor has already compensated the plaintiff that payment obviates one reason for transferring money between the parties, the injured party receiving money, but the fairness involved in the tortfeasor paying still remains. As the New Jersey Supreme Court stated:
If there is enrichment, it is not at the defendant’s expense. Defendant does not seek to make [the settling defendant] whole but rather to profit from the injustice that [the settling defendant] supposedly experienced. If the voluntary agreement between plaintiff and [the settling defendant] were thought so to offend public policy as to require redress, the remedy would run to [the settling defendant] rather than to a stranger to the bargain.
Theobald v. Angelos, 44 N.J. 228, 208 A.2d 129 (1965). Alternately stated,
The one-compensation rule, grounded in unjust enrichment, is not to be applied in such a way as to generate unjust enrichment to the only litigating defendant. ... It would be unjust enrichment ... to give the only defendant who was eventually found liable ... a full pro tanto credit for the full amount paid by the others.
Rose v. Associated Anesthesiologists, 501 F.2d 806, 809 (D.C.Cir.1974).
A second fairness argument suggesting that overcompensation does not make a proportional credit unfair flows from the fact that comparative fault actually separates a joint tort into independent multiple torts. There is no reason that a settlement with one tortfeasor should affect the judgment against another tortfeasor. If, in fact, the settlement was more generous to the plaintiff than a trial "with that joint tortfeasor would have been, all that indicates is that the plaintiff got a good deal.6 There certainly is no reason why a nonsettling tortfeasor should claim the benefit of the plaintiff’s bargain with the settling tortfeasors. Fairness demands that the benefits of the bargain remain where the bargaining parties place them, and that nonsettling tortfeasors pay damages in proportion to their fault. See Larson, A Problem in Contribution: The Tortfeasor with an Individual Defense Against the Injured Party, 1940 Wis.L.Rev. 467, 486-81; California Comment, supra, 65 Calif.L.Rev. at 1278-79.
2. Overcompensation and Deterrence. —The argument against overcompensation is even weaker under a deterrence theory. Deterrence is not at all concerned with compensating the injured party. Deterrence, rather, is concerned with making the tortfeasors pay. The fact that the money goes [770]*770to the plaintiff is not really relevant.7 “[F]rom the deterrence standpoint, it is irrelevant to whom damages are paid, so long as someone redressed the violation.” Illinois Brick Co. v. Illinois, 431 U.S. 720, 760, 97 S.Ct. 2061, 2028, 52 L.Ed.2d 707 (Brennan, J., dissenting). In order for deterrence to function properly, each tortfeasor must pay for the proportional share of damage caused. Whether this payment results in over, under, or exact compensation is beside the point.
Another way of understanding this is to realize that deterrence is concerned with establishing a rule to shape future conduct. A potential tortfeasor’s actions will probably not be shaped by considerations of whether the injured party will be compensated nearly as much as they will be shaped by considerations of whether the tortfeasor will have to pay. Accordingly, the issue of whether another joint tortfeasor has settled with the plaintiff just is not relevant at all to the deterrence rationale for requiring each nonsettling tortfeasor to pay a proportional share of damages.8
B. Procedural Problems in This Case
Finally we must deal with two procedural problems presented on the facts of this case. First, the settlement in this case was for multiple torts, not all of which were joint. Denny’s was sued for false arrest, malicious prosecution, and conspiracy to violate Dobson’s federal rights. The only joint aspect of these allegations was the conspiracy claim, and we have no clear way of allocating the settlement among the joint and separate claims. Second, we do not have any jury findings as to comparative fault/comparative causation.
These problems are linked in one regard. With a jury finding on comparative fault we would not need to concern ourselves with identifying which part of the settlement was for which claim. We would merely assign to Camden his share of the damages and be done with it. Again, under a system of comparative fault/comparative negligence the existence of a settlement is irrelevant to the liability of a nonsettling tortfeasor.
That principle is of little comfort to us now because we have no jury finding on comparative fault. Our question at this point is how to handle this procedural default. In future cases we would have no problem in holding that the party desiring a reduction in judgment on account of a settlement has the burden of proving its entitlement to that reduction by showing another party shares responsibility for the injuries and submitting jury issues on relative fault/causation. In this case, however, we are reluctant to charge a party with the consequences of default of a procedure that did not even exist prior to this appeal. Accordingly, we must remand for a determination of relative liability.
CONCLUSION
This deceptively simple case regarding the effect of a settlement on a nonsettling joint tortfeasor has taken us on quite a journey.. Section 1988 led us from a deficiency in federal law, past state law conflicting with federal policy into the netherworld of federal common law. Once there we found that the dual policies of section 1983, fairness and deterrence, led us by different paths of reasoning to the same result. Fairness and deterrence both call for the same inquiries into compensatory [771]*771and punitive damages. Furthermore, fairness and deterrence both call for apportioning joint liability in proportion to fault and causation. Having come that far it is a simple matter to realize that we have separated joint torts into independent multiple torts. Accordingly, a settlement by one tortfeasor has no effect on the liability of a nonsettling tortfeasor.
This simple theory we have used to unify the field of joint liability is difficult to apply here because it differs from the unspoken premises used in the court below. The record developed below does not provide either the trial court or this Court a factual basis to apportion damages fairly with anything approaching the limits of Einsteinian accuracy. We are left with post hoc hypotheses and a relatively greater quantum of uncertainty than Heisenberg requires. If the rule regarding comparative liability we announce today had been in effect below, we would simply state that the party claiming a reduction in judgment has the burden of proving that another party, though not joined as a defendant, shares the blame for having caused the injury. Absent proof one way or the other, we would allow no reduction. Because we are imposing a new rule on Texas section 1983 litigants, however, we must remand to give Camden an opportunity to prove that he does not bear sole responsibility for Dob-son’s injuries.9
This case is a confection of many legal ingredients, of which some are complicated, some are intersecting, and others are philosophical. We hope we have examined all the recipes available to us and trust that our conclusion is neither half-baked nor overcooked, but a palatable blend of the policies of section 1983. If this ease does not prove anything else, it demonstrates that the law cannot be static. The law, as it must, evolves in its uncertainties, functions pragmatically and rationally responds to perform its societal function even as our conception of that function changes.
REVERSED AND REMANDED.