dissenting.

I respectfully dissent with regard to the attorney’s fees award. A lawyer’s fee of $139,788.25 for getting a client $17,500 is obscene. The district court properly determined that the amount was inappropriate under controlling law and cut the fee to $20,000. We should affirm that discretionary determination.

The majority purports to limit Farrar v. Hobby, 506 U.S. 103 , 113 S.Ct. 566 , 121 L.Ed.2d 494 (1992), to nominal damages cases. Farrar applies to all 42 U.S.C. § 1988 cases. We have already established that interpretation of Farrar , and it is inescapable in the context of the line of Supreme Court decisions of which Farrar is a part. The district court applied Farrar correctly and exercised discretion appropriately:

That exercise of discretion was in accord with controlling law.

Farrar considered whether a civil rights plaintiff who receives only a nominal damages award is a prevailing party. The Fifth Circuit had held that plaintiffs who win only $1 are not prevailing parties, so attorneys’ fees cannot be awarded under Section 1988. Farrar, 506 U.S. at 107-09 , 113 S.Ct. at 571 . The Supreme Court affirmed the denial of fees, but on a different theory. The Court held that such a plaintiff is a prevailing party, but that attorneys’ fees awarded to prevailing parties under § 1988 must reflect the “degree of success obtained.” Id., at 114, 113 S.Ct. at 574 .

*366 I am mystified by the majority’s insistence upon interpreting only Justice O’Connor’s concurring opinion in Farrar , and neither quoting from nor applying the reasoning of the opinion of the Court. Justice Thomas wrote the opinion of the Court in Farrar , joined by Chief Justice Rehnquist, Justice O’Connor, Justice Sealia, and Justice Kennedy. Five is a majority. Justice O’Connor expressly wrote “I join in the Court’s opinion.” Id., at 116 , 113 S.Ct. at 575 . Justice Thomas’s opinion said that “the only reasonable fee is usually no fee at all.” Id. (emphasis added). Justice O’Connor did not suggest that the opinion should be limited to nominal damages cases. Rather, she took the position that “no fees can be awarded” where the plaintiffs victory is “purely technical or de minimis,” but not all nominal damages cases are mere technical victories. Id., at 117, 120-22 , 113 S.Ct. at 576, 578 (emphasis added).

The reasoning of Farrar compels the conclusion that it applies generally to section 1988 cases, not just to nominal damages cases. Farrar holds that “the most critical factor in determining the reasonableness of a fee award is the degree of success obtained.” Id., at 114 , 113 S.Ct. at 574 (quotations omitted). Farrar applied the rule in Hensley v. Eckerkart, 461 U.S. 424 , 103 S.Ct. 1933 , 76 L.Ed.2d 40 (1983), that using the product of hours times rate as the award may be excessive in cases of “partial or limited success.” Farrar, 506 U.S. at 114 , 113 S.Ct. at 574 . Hensley had held, in an injunctive relief (not damages) case, that the “lodestar” fee may be excessive in cases of partial or limited success, and in such cases, “the district court should award only that amount of fees that is reasonable in relation to the results obtained.” Hensley, 461 U.S. at 440 , 103 S.Ct. at 1943 . Farrar made it clear that this rule applied to damages cases as well as injunc-tive relief cases, and that in a nominal damages case, the fee “reasonable in relation to the results obtained” was usually nothing.

The Supreme Court in City of Riverside v. Rivera, 477 U.S. 561 , 106 S.Ct. 2686 , 91 L.Ed.2d 466 (1986) affirmed an award of attorneys’ fees exceeding the amount of damages recovered by the plaintiff. A plurality, but not a majority, of the Court rejected the proposition that section 1988 fees must necessarily be no higher than the damages award, or no higher than the one third contingent fee commonly used in personal injury litigation. Id. at 574 , 106 S.Ct. at 2694 . The plurality conceded that “the rates charged in private representations may afford relevant comparisons.” Id. at 578 n. 9, 106 S.Ct. at 2696 n. 9. The plurality decision rejected the proposition that fee awards “should necessarily be proportionate to the amount of damages a civil rights plaintiff actually recovers,” on the theory that “[ujnlike most private tort litigants, a civil rights plaintiff seeks to vindicate important civil and constitutional rights that cannot be valued solely in monetary terms.” Id. at 574 , 106 S.Ct. at 2694 . The plurality considered that a strict rule limiting fee awards to monetary damages fails to reflect the important social benefits secured by a successful civil rights plaintiff. Id.

Justice Powell, the fifth vote to affirm in Riverside, “join[ed] only the Court’s judgment.” Id. at 581 , 106 S.Ct. at 2697 . He did not join in the plurality opinion. He thought that the “fee award seems unreasonable,” and that it would be a “rare case in which an award of private damages can be said to benefit the public interest to an extent that would justify the disproportionality between damages and fees reflected in this case.” Id. at 586 n. 3, 106 S.Ct. at 2700 n. 3. He affirmed only because the detailed findings supporting the award were not clearly erroneous, that is, because the standard of review did not permit him. to substitute his judgment for that of the district judge.

Putting these three cases together, the law is plain: while section 1988 fees are not necessarily held to a ceiling of the damages award or the one third contingent fee common in personal injury cases, the fee award has to be reasonable in relation to the results obtained, including societal benefits, and a limited success requires that a fee be limited by reasonableness in relation to the results obtained, even where the “lodestar” amount is much higher.

The majority opinion errs by requiring the district court to begin by making a lodestar calculation, and then to assess whether any *367 adjustment should be made to it. Farrar says just the opposite. The “critical factor” of “degree of success” trumps the old Kerr v. Screen Extras Guild, Inc., 526 F.2d 67 (9th Cir.1975), twelve factors approach to adjusting the “lodestar” of hourly rate times hours spent. Farrar teaches that the district judge neither has to mention the twelve factors, nor even multiply hours times rate. “Having considered the amount and nature of damages awarded, the court may lawfully award low fees or no fees without reciting the 12 factors bearing on reasonableness or multiplying the number of hours reasonably expended ... by a reasonable hourly rate.” Farrar, 506 U.S. at 117 , 113 S.Ct. at 575 (emphasis added) (citation and quotations omitted). Thus, the district judge need not start by multiplying hours spent times hourly rate, and then adjust it by the old twelve factors and extent of success, in a limited success case. 1 Instead, the judge can start by considering “the amount and nature of damages awarded.”

Our own decisions leave no room for the majority to adopt its different analysis of Farrar today. We have already rejected the majority’s limitation of Farrar to nominal damages cases, and we can reject our earlier decisions only by en banc review. United States v. Camper, 66 F.3d 229, 232 (9th Cir.1995).

In Corder v. Brown, 25 F.3d 833 (9th Cir.1994), where the damages award was $24,006 (not nominal), we held that under Farrar , full attorney’s fees based on multiplication of hours times rate would be unreasonable. We required the district court on remand to recalculate the fee award giving primary consideration to the fact that “the amount of damages awarded as compared to the amount of damages sought is minimal,” for which we cited Farrar. Corder, 25 F.3d at 837 . Thus we applied Farrar to a case other than a nominal damages case.

In Harris v. Marhoefer, 24 F.3d 16 (9th Cir.1994), we followed Carder in another case of a moderate (not nominal) award, $25,000, and required reduction of the lodestar amount because of limited success.

We carefully considered how Farrar applied to a $34,000 award (not nominal) in McGinnis v. Kentucky Fried Chicken, 51 F.3d 805 (9th Cir.1994). We construed Far-rar to mean that “it is an abuse of discretion for the district court to award attorney’s fees without considering the relationship between the extent of success and the amount of the fee award.” Id. at 810 . Because the district court expressly refused to relate the extent of success to the amount of the fee award and refused to reduce the attorney’s fee award “so that it is commensurate with the extent of the plaintiffs success,” id. at 810 , we reversed. In McGinnis , plaintiffs attorney had obtained a disappointingly low compensatory damages award, considering Mr. McGinnis’s suffering, but a high punitive damages award. The punitive damages award had to be disallowed as a matter of law, so the plaintiffs success was limited. We noted that “no reasonable person would pay lawyers $148,000 [the fee sought] to win $34,000 [the damages obtained].” Id. at 810 . Although the attorneys’ fees award could exceed what a reasonable individual would pay a lawyer for the private benefit, because of the public benefit associated with the plaintiffs victory, the public benefit “is not infinite,” and “[w]hat the lawyers do for their actual client is an important measure of ‘extent of success.’” Id. We held that the district court “must reduce the attorneys’ fees award so that it is commensurate with the extent of the plaintiffs success.” Id.

After Carder, Harris, and McGinnis , it is impossible, without rehearing en banc, for us to construe Farrar to apply only to nominal damages cases. 2 Even if our own circuit *368 authority did not bind us to a broader reading, the Supreme Court’s reasoning in the majority opinion in Farrar , and in Hensley , would leave no room for the majority’s interpretation. “Where the relief sought and obtained is limited to money, the terms ‘extent of success’ and ‘level of success’ are euphemistic ways of referring to money.” McGinnis, 51 F.3d at 810 . In a limited success case, the district court may begin by considering “the amount and nature of damages awarded,” and award fees commensurate with the extent of success “without ... multiplying the number of hours reasonably expended ... by a reasonable hourly rate.” Farrar, 506 U.S. at 115 , 113 S.Ct. at 575 .

The majority also errs in substituting its judgment for the district court’s, regarding the extent of benefits Mr. Morales’s attorneys conferred on persons other than their client by their limited victory. It is firmly established that the district court has discretion over how much to award for attorney’s fees, and our review authority is limited to determining whether there was abuse of discretion. Hensley v. Eckerhart, 461 U.S. 424, 437 , 103 S.Ct. 1933, 1941 , 76 L.Ed.2d 40 (1983); Franceschi v. Schwartz, 57 F.3d 828, 830 (9th Cir.1995); Oviatt v. Pearce, 954 F.2d 1470 , 1481 (9th Cir.1992); Jordan v. Multnomah County, 815 F.2d 1258, 1261 (9th Cir.1987); Hardin v. White Mountain Apache Tribe, 779 F.2d 476, 480 (9th Cir.1985). Indeed, the reason Justice Powell gave for his fifth vote to affirm our decision in Riverside was the limited scope of appellate review. Riverside, 477 U.S. at 581-86 , 106 S.Ct. at 2697-2700 .

Mr. Morales’s attorney had made a formal demand for $250,000, and urged the jury to award $150,000 to $250,000 in compensatory damages, plus $500,000 in punitive damages. The jury decided that the police officer “unlawfully arrested or detained plaintiff as a result of negligent performance of his duties.” But it also decided that Mr. Morales himself was negligent and that his own negligence was also a proximate cause of his injury. They attributed 45% of the negligence to Mr. Morales and 55% to the police officer, and found total damages suffered by Mr. Morales of $17,500. Mr. Morales’s attorney wanted $139,783.25 in fees for getting his client $17,500 (the court decided that $17,500, not 55% of $17,500, was the appropriate award).

Mr. Morales won damages of $17,500, so a one-third contingent fee to would have been $5,833. That is the private market measure of the fee commensurate with the extent of success obtained for Mr. Morales. Cases like Mr. Morales’s are rarely handled on an hourly basis, because the clients cannot afford the risk of paying more than they win, and the lawyers cannot afford the risk of nonpayment. The district judge thought a fee of $20,000 took proper account of the reasonable time spent and rates chargeable, the extent of success for Mr. Morales, and the benefit conferred on the public. That is an extra $14,167 for what the lawyers did for the public, on top of the $5,833 for what they did for Mr. Morales. The judge noted that the $17,500 award “constitutes a warning to law enforcement officers not to treat civilians unconstitutionally.” Only that benefit could justify awarding more to the lawyer than his client obtained from his services. As we said in McGinnis , “[t]he ‘private attorney general’ theory lets the attorneys recover more than the benefit to their client would make reasonable, because they also confer benefits on others throughout society by winning a civil rights claim.” McGinnis, 51 F.3d at 810 .

There are several things wrong with an excessive award of attorneys’ fees in a case such as this. One is over-deterrence. The police, having been called by a private guard who had spotted a weapon, had taken a loaded M-l carbine from a van. The arresting officer knew there was a law against *369 carrying a loaded weapon in a vehicle. The van was registered to a “Robert Garcia,” and the police had an outstanding arrest warrant for a “Roberto Garcia” for murder. Mr. Morales had asked a security guard about the seizure of the weapon, at the request of another individual whose English was limited, but the police officer had no reason to suppose that Mr. Morales’s inquiry had been on behalf of someone else. When asked for his identification, Mr. Morales, who had obviously been drinking, did not produce it, so the policeman could not ascertain that he was Mr. Morales and not Roberto Garcia, the murder suspect. The jury did not think the arrest was egregious, outrageous police misconduct. They thought the fault was about 55% to 45%, and that no punitive damages were appropriate. Just how much do we want to deter police from arresting a man in this situation? Though he was not guilty of anything, the policeman had reason to think that Mr. Morales was at least a violator of the law against possession of a loaded gun in a vehicle, and perhaps was wanted for murder. Do we want a police officer to say “better a man suspected of murder who has been drinking should carry a loaded firearm in his vehicle, than that I should risk making an unlawful arrest”? That does not make sense. We should deter police from acting recklessly, but lawyers confer harm, not benefit, on society if the police are deterred by immense awards from touching a situation like this with a ten foot pole.

Second, excessive attorneys’ fees awards in section 1983 litigation unduly divert legal resources from other uses. People injured by negligent drivers and on the job, people who need wills drawn, people who need tenants evicted or landlords restrained from unlawful evictions, people who need their construction contracts reviewed and bank loans secured so they can start work, have to compete for legal services with Mr. Morales, who can offer a lawyer $139,000 to sue a policeman for arresting him erroneously and holding him for two hours. Congress chose to give lawyers more of an incentive than the usual contingent fee out of the damages recovery to take section 1983 cases, but neither the words nor the purpose of section 1988 justify a $139,000 jackpot in a $17,500 case.

Third, excessive attorneys’ fees awards in civil rights cases harm the victims of civil rights violations. That sounds paradoxical, but is not surprising to anyone with experience in personal injury work. The wronged victim of the civil rights violation is likely to want, or if physically injured, need, money, as compensation for the wrongs done to him. But the cases are hard to win, and probably most wrongful arrest cases taken to trial lose on liability, or result in low or nominal awards. So the safest, fastest and most efficient way for the victim of the wrong to obtain money is likely to be settlement. But no municipality is going to settle a $17,500 case for $157,000, with the lawyer to get the lion’s share of the money. In any one case, it is unlikely that the plaintiffs lawyer will “ring the bell,” as personal injury lawyers put it. The settlement value to the city is likely to be the risk of a bad loss multiplied by its size, perhaps 10% of $150,000 to $200,-000. Also, a defense lawyer would be concerned about the ethics of a settlement where the plaintiffs lawyer, not the wronged plaintiff, gets the lion’s share of the money. It would look like a bribe to the lawyer to dump the case. The plaintiffs lawyer would have a very great financial incentive to talk his client out of settling, and to take the ease to trial. All the fees he might hope for from ten small settlements are unlikely to add up to one jackpot fee, so even if he wins a very small percentage of cases at trial, the lawyer will do better to try them all. If the fee is too high relative to the client’s recovery, then the lawyer’s economic interest conflicts with his client’s, because the client has only one case but the lawyer has many. The majority’s approach today turns the ease into a dollar slot machine for the lawyer. Most lawyers are honorable people, and the lawyers representing plaintiffs in civil rights cases are often highly idealistic, but economic incentives work in the long run on the best people. A lot of people wronged by the police are going to be wronged again by being talked into exchanging small settlements for big trial defeats, under the attorneys’ fees rule the majority adopts today.

Mr. Morales’s attorneys conferred a moderate benefit on their client, and are entitled *370 to a moderate fee commensurate with that benefit, increased by a moderate extent to reflect the benefit conferred on the rest of society. I do not understand why the majority characterizes a $20,000 attorney’s fee as “low,” in a $17,500 ease. Three or four fees a year this big, and a large number of smaller fees, are the typical path to grossing six figures and netting in the high five figures for a successful lawyer. A successful lawyer does not, however, spend 938 hours, equivalent to working full time on nothing else for half a year or more, on a $17,500 case. The district judge exercised discretion precisely in accord with controlling law. The district judge expressly considered the lodestar fee request made, determined that the fee would be excessive when considered relative to the results achieved, and awarded a fee commensurate with those results, increased by the benefit conferred on persons other than the lawyer’s client. A $20,000 fee to a lawyer for getting $17,500 for his client in this circumstance is defensible. A $139,783.25 fee to a lawyer for getting $17,500 for his client is an incentive for the lawyer to serve, not his client, but himself.

. The majority mistakenly relies on Jordan v. Multnomah County, 815 F.2d 1258, 1262 (9th Cir.1987), for the proposition the lodestar determination is the "predomina[nt] element” of a fee award. We decided Jordan in 1987, but the Supreme Court decided Farrar in 1992, so Farrar controls. Farrar holds that " 'the most critical factor' in determining the reasonableness of a fee award 'is the degree of success obtained.' " Farrar, 506 U.S. at 114 , 113 S.Ct. at 574 .

. We say in Stivers v. Pierce, 71 F.3d 732, 753 (9th Cir.1995), that “Farrar holds that a prevailing party may be denied fees where he obtains only a narrow, technical victory such as nominal damages.” Stivers was a § 1983 case for an *368 injunction to grant a license, and damages. The Board yielded in response to the lawsuit, and agreed to grant licenses to Stivers, so the plaintiff won substantial relief, even though he lost what was left of the case on summary judgment. We therefore reversed the denial of attorneys’ fees. Our purported limitation of Farrar showed why the lack of a substantial damages award did not bar an attorney's fees award. Stivers is consistent with McGinnis and the other cases discussed above. Where a plaintiff wins substantial relief in response to the lawsuit, he is generally entitled to an attorney’s fee award, even where the monetary portion of the relief is absent or nominal. Farrar, 506 U.S. at 111-12 , 113 S.Ct. at 573 ; Stivers, 71 F.3d at 751 .