After judge Orval Anderson of the Lake County Court was indicted, the Supreme Court of Indiana appointed James J. Krajewski in February 1985 to fill the position temporarily. Under Indiana law, Krajewski had the authority to appoint any necessary public defenders. The chief public defender had been indicted, but Krajewski asked Steven A. Kurowski and David H. Nicholls, the assistant public defenders, to stay on. Judge Anderson resigned in June 1985 following his conviction, see United States v. Anderson, 798 F.2d 919 (7th Cir.1986), and the Governor of Indiana gave Krajewski the regular appointment. Krajewski praised the work of Kurowski and Nicholls to date and increased their salaries.

Six months later he fired them. Kurowski and Nicholls, like the judge who hired them, are Democrats. Nicholls is a member of the Central Committee of Lake County's Democratic Party; Kurowski is a Vice-Committeeman of his precinct. They campaigned for Judge Anderson's retention in 1976, 1980, and 1984; Kurowski represented him in the criminal prosecution. Krajewski, like the Governor who appointed him, is a Republican. Although Krajewski appointed as chief public defender a friend who generally votes Democratic (but is not active in politics), he filled the assistants' jobs with Republicans. The assistants responded with this suit under 42 U.S.C. Sec. 1983, contending that the use of political criteria in selecting public defenders violates the first amendment, applied to the states through the fourteenth.

The parties consented to final disposition by a magistrate, 28 U.S.C. Sec. 636(c). The magistrate granted partial summary judgment to the plaintiffs on issues concerning official immunity and held a trial on the question whether Judge Krajewski fired the plaintiffs because of their politics. He concluded that Krajewski had done this and awarded compensatory damages aggregating $59,075 and punitive damages of $1,000 per plaintiff; he also ordered the plaintiffs reinstated as assistant public defenders. On motion of the plaintiffs, the magistrate awarded $13,905 as attorneys' fees under 42 U.S.C. Sec. 1988. Krajewski's appeal contests every aspect of the decision except the award of punitive damages, which he apparently concedes are appropriate if he is liable at all.

* Public defenders represent criminal defendants, and their loyalty as advocates runs to their clients, not to their employer. Although paid by the state, in their role as counsel they are not even "state actors". Polk County v. Dodson, 454 U.S. 312, 102 S. Ct. 445, 70 L. Ed. 2d 509 (1981). The Supreme Court said in Branti v. Finkel, 445 U.S. 507, 100 S. Ct. 1287, 63 L. Ed. 2d 574 (1980), that political beliefs and affiliations are not permissible criteria for firing public defenders, precisely because a public defender is the advocate of the accused rather than a maker or implementer of political decisions. Branti held that a chief public defender may not fire assistant public defenders on political grounds. Krajewski argues that things are different when a judge fires an assistant public defender, but this misunderstands Branti. The Court's decision turned on the characteristics of the assistant public defender's job, not on the office of the person doing the firing.

Branti dealt with public defenders in New York. Krajewski tells us that public defenders in Indiana are different. Indiana Trial Rule 63(E) permits a trial judge to appoint a "judge pro tempore " to serve while he is away. The judge pro tempore exercises the full powers of the office, conducting trials, sentencing defendants, and so on. See Survance v. State, 465 N.E.2d 1076 (Ind.1984). Krajewski's predecessor frequently appointed public defenders to sit as judges pro tempore in his absence. Krajewski continued this practice with respect to Kurowski, although there is no evidence that he appointed Nicholls to sit as judge pro tempore.

Judge Krajewski maintains that the duties of the office of public defender in Indiana therefore include functions in addition to those considered in Branti, and that because the judge pro tempore is a policy-maker for the State of Indiana, political criteria may be applied to the job of public defender. The magistrate disagreed on the ground that judges pro tempore are not the instruments of the judge who appoints them; because judges pro tempore are free agents, the magistrate thought, and do not carry out the regular judge's policy, it is inappropriate for the regular judge to employ political criteria in selecting a judge pro tempore.

The magistrate's approach is unsatisfactory because it assumes that Branti and its progenitor, Elrod v. Burns, 427 U.S. 347, 96 S. Ct. 2673, 49 L. Ed. 2d 547 (1976), permit an appointing officer to consider the appointee's political views only when the appointee carries out the appointing official's own "policy". If this is so then, for example, the governor could not consider a would-be judge's politics when deciding whom to appoint (because the judge is independent of the governor once in office), and the President could not consider the views of a prospective appointee to the Federal Trade Commission when making that selection. Neither Elrod nor Branti makes anything turn on the relation between the job in question and the implementation of the appointing officer's policies.

A governmental officer holding the power of appointment may make any decision he pleases, unless the Constitution bars the way. The pertinent provision--the guarantee of free speech contained in the first amendment--disables the government from firing a person based on his speech and political beliefs (and thus penalizing the employee for holding or advocating those beliefs) unless the beliefs are relevant to the job in question. They will be relevant, the Court said in Branti and Elrod, when the office involves making on the state's behalf the sort of decisions about which there are political debates. That is to say, the first amendment does not remove political beliefs from politics; it would undermine the democratic process to hold that the winners at the polls may not employ those committed to implementing their political agenda.

So the right question is whether there may be genuine debate about how best to carry out the duties of the office in question, and a corresponding need for an employee committed to the objectives of the reigning faction. Shondel v. McDermott, 775 F.2d 859, 864 (7th Cir.1985); Lindahl v. Bartolomei, 618 F. Supp. 981, 987-88 (N.D.Ind.1985). When the officeholder wields the final authority of the government, the answer to that question almost always will be yes. It will frequently be yes even lower within the bureaucracy. See, e.g., Tomczak v. City of Chicago, 765 F.2d 633 (7th Cir.1985), holding that the number two person in a bureaucracy devoted to delivering water could be fired on political grounds, because the operation of municipal enterprises is embroiled in politics--because there may be genuine debate both about how to allocate resources between waterworks and snow removal, and about how best to deliver water--the "ins" need the loyalty of the senior managers in order to implement their programs.

A judge both makes and implements governmental policy. A judge may be suspicious of the police or sympathetic to them, stern or lenient in sentencing, and political debates rage about such questions. In most states judges are elected, implying that the office has a political component. Holders of the appointing authority may seek to ensure that judges agree with them on important jurisprudential questions. The Governor of Indiana was entitled to consider Krajewski's views about the role of judges--or even simply Krajewski's political affiliation--when making the appointment, just as the voters may consider these factors without violating the first amendment when deciding whether to retain Judge Krajewski in office. (We put aside all debate about whether recourse to politics in selecting judges is good or bad; we are concerned only with the constraints the first amendment imposes on the way the State of Indiana prefers to organize its government.) What is true about the office of judge is true about the office of judge pro tempore. It is the same job; only the tenure of the officeholder differs. The regular judge holds office for four years; the judge pro tempore holds office at the pleasure of the regular judge. Since the disposition under Elrod and Branti depends on the duties of the office rather than the tenure of the incumbent, it follows that political beliefs may be the basis for the appointment of a judge pro tempore. If Judge Krajewski wants a judge pro tempore committed to "law and order" or to sympathy for criminal defendants, the first amendment is no obstacle.

Accordingly, we must decide whether the plaintiffs' job included duties as judge pro tempore, for if it did then Judge Krajewski may apply political criteria. And we must ask what the powers of the actual office were, not what an "ideal" public defender does. Tomczak, 765 F.2d at 640. Because the magistrate granted partial summary judgment for Kurowski and Nicholls on the question whether Judge Krajewski could take their politics into account, any disputed question of material fact requires a remand for retrial.

There is no material dispute, however. Indiana Trial Rule 63 makes it pellucid that judicial service is not part of a public defender's duties in Indiana--one can be a public defender but not a judge pro tempore, or a judge pro tempore but not a public defender. The only qualification is that the judge pro tempore "shall be an attorney in good standing at the bar of the Supreme Court of this state." Rule 63(C). Krajewski concedes that he has appointed as judge pro tempore attorneys who are not public defenders. After he fired Kurowski and Nicholls, he asked them to sit in for him while he was on vacation the next month, the most vivid demonstration that public defender and judge pro tempore are not the same job. More, Nicholls testified without contradiction that he could not recall serving as judge pro tempore after Judge Krajewski's appointment in February 1985, although he was an assistant public defender for another eleven months. Judge Krajewski may have turned frequently to the attorneys closest to hand, the public defenders. This does not make judicial service part of the public defender's office, any more than it would make such service part of the job of a lawyer in private practice who happened to be a confidant of the judge. We doubt that the Attorney General of Indiana--who represents Judge Krajewski in this court--would argue that Democrats may be excluded from the Bar of Indiana because judges may choose to appoint only Republicans as judge pro tempore. There is no greater reason why a political test may be imposed for the job of assistant public defender.

After hearing the evidence at trial, the magistrate had to choose between competing stories. The plaintiffs insisted that Krajewski fired them to make room for Republicans, in order to maintain the support of the Republican Party in his campaign for retention in office. Judge Krajewski insisted that he fired Kurowski and Nicholls because they had been hired by his convicted predecessor and had worked closely with Lee J. Christakis, the convicted former chief public defender. Since Kurowski and Nicholls had the taint of their associations about them, Krajewski maintained, he removed them in order to improve the public standing of the court. Whether Kurowski and Nicholls were saints or sinners, Judge Krajewski contended, the court as a whole was better off without them. Since a public official may fire subordinates for a good, bad, or indifferent reason--any but an unconstitutional reason--the actual moral standards of Kurowski and Nicholls are irrelevant. It was enough that he had lost confidence in them. See Livas v. Petka, 711 F.2d 798 (7th Cir.1983).

This presented the magistrate with the need to decide who to believe. If Judge Krajewski in fact thought Kurowski and Nicholls a blot on the court's escutcheon, he could fire them, whether they were or not. If Krajewski just thought Kurowski and Nicholls two obstacles to patronage for the Republican Party, he could not fire them. The magistrate took the latter view. That finding is not clearly erroneous. See Anderson v. Bessemer City, 470 U.S. 564, 105 S. Ct. 1504, 84 L. Ed. 2d 518 (1985).

Judge Krajewski's conduct undercuts his contention that he came to office with a broom and swept out the clutter he encountered. He retained Kurowski and Nicholls for eleven months and never inquired of the prosecutor or any other source whether there were grounds to suspect the two of misconduct. When Krajewski was given the permanent appointment in June 1985, he told Kurowski and Nicholls they had done good work so far, and he raised their salaries. He warned them at the same time that as a Republican appointee he was under pressure for their jobs, and that they must understand "how the game is played". Judge Krajewski testified at trial that he was indeed under such pressure. When he told the plaintiffs in December that they would be replaced effective January 1, 1986, Krajewski repeated the comment that he was under pressure from the Republican Party. He gave Kurowski and Nicholls no other reason for their discharge and allowed that he could probably be sued for what he was doing. He then asked whether either would be willing to serve as judge pro tempore the next month. This is not the path trod by one intent on turning out rascals.

A. Krajewski contends that as a judge he is immune from liability in damages for actions taken in a judicial capacity. Although the parties do not discuss our jurisdiction to consider this question, we must consider the subject anyway. Judge Krajewski moved before trial for summary judgment on the grounds of absolute and qualified immunity. The magistrate denied Krajewski's motion and granted partial summary judgment to plaintiffs on the issue. This decision was immediately appealable. Mitchell v. Forsyth, 472 U.S. 511, 105 S. Ct. 2806, 86 L. Ed. 2d 411 (1985); Scott v. Lacy, 811 F.2d 1153 (7th Cir.1987). Krajewski did not appeal; instead he persevered and presented the immunity question as a ground for reversal after losing on the merits. At least one court of appeals has held that a public official must appeal as soon as Mitchell permits, and if he waits he is barred from raising the question later--at least on a subsequent interlocutory appeal. Kennedy v. City of Cleveland, 797 F.2d 297, 304-05 (6th Cir.1986). No court has addressed the question whether the rationale of Kennedy, by extension, forbids raising the question on appeal from a final judgment.

Mitchell allows an immediate appeal on the theory that the rejection of a claim of absolute immunity is a "collateral order" disposing of rights separate from the merits, rights lost if appeal does not come until after trial. See Cohen v. Beneficial Industrial Loan Corp., 337 U.S. 541, 69 S. Ct. 1221, 93 L. Ed. 1528 (1949); Gulfstream Aerospace Corp. v. Mayacamas Corp., --- U.S. ----, 108 S. Ct. 1133, 1136-38, 99 L. Ed. 2d 296 (1988). Absolute immunity establishes both a right not to pay damages and a "right not to be tried"--that is, it grants freedom from the emotional travail, commitment of time, and legal expense involved in going to trial. Immunity doctrines lose some of their force if these benefits are withheld until after trial. Trial is not "irreparable injury", see Paine Webber Inc. v. Farnam, 843 F.2d 1050, 1051-52 (7th Cir.1988), (collecting authority), but it may so defeat the purpose of this privilege that an immediate appeal is an essential ingredient of the right itself. See also Abney v. United States, 431 U.S. 651, 97 S. Ct. 2034, 52 L. Ed. 2d 651 (1977). The right to appeal forthwith may be combined with the principle that a final judgment must be appealed at once or not at all, e.g., Exchange National Bank v. Daniels, 763 F.2d 286, 291-92 (7th Cir.1985), to produce the conclusion that a public official who foregoes an appeal under Mitchell forfeits arguments based on immunity.

This neat little argument encounters a fundamental problem: only final judgments need be appealed. Until a judgment is rendered "final" by entry of a separate document under FED.R.CIV.P. 58, no one need appeal. United States v. Indrelunas, 411 U.S. 216, 93 S. Ct. 1562, 36 L. Ed. 2d 202 (1973). An appeal from an interlocutory order is not one from a final judgment, even if by virtue of Cohen it is from a final decision. Interlocutory orders therefore may be stored up and raised at the end of the case, see Daniels, 763 F.2d at 290; In re Kilgus, 811 F.2d 1112, 1116-17 (7th Cir.1987). Other interlocutory decisions, even those producing interlocutory "judgments" (such as the denial of a preliminary injunction) receive the same treatment. See United States v. Clark, 445 U.S. 23, 25-26 n. 2, 100 S. Ct. 895, 898-99 n. 2, 63 L. Ed. 2d 171 (1980). This makes sense. The privilege to take an interlocutory appeal exists for the appellant's protection. Such appeals come at great cost to the judicial system because they may prolong litigation and require appellate courts to cope with each case more than once. Most interlocutory appeals end in affirmance (thus entail wasted motion), because district judges dispose correctly of the vast majority of motions. If the aggrieved party is content to swallow his losses and proceed with the case--if, for example, Judge Krajewski was content to surrender any "right not to be tried" he may have possessed, so long as he did not have to pay damages at the end--no interest of either the judicial system or the adverse party is served by treating the whole subject as forfeit. That would simply induce public officials to file more interlocutory appeals. We therefore hold that a public official may raise questions of immunity on appeal from a final judgment, even though he bypassed an opportunity to take an interlocutory appeal.

B. Before the magistrate, and in his opening brief here, Krajewski relied principally on Forrester v. White, 792 F.2d 647 (7th Cir.1986), which held that a judge is entitled to absolute immunity when hiring and firing probation officers. Shortly after Krajewski filed his opening brief, the Supreme Court reversed that decision, Forrester v. White, --- U.S. ----, 108 S. Ct. 538, 98 L. Ed. 2d 555 (1988), holding that judges act in an "administrative capacity" when hiring and firing staff--even staff intimately connected with the judicial office--and therefore possess only the qualified immunity that applies to other public officials so acting.

One would think that Forrester disposes of any claim of absolute immunity. Krajewski stubbornly clings to it, however, saying in his reply brief that

[w]hile [Forrester ] does indeed hold that a judge is not absolutely immune for terminating the employment of a probation officer, that does not answer the question of whether that judge is immune as to the decisions to hire and fire public defenders. That case should be limited to its facts and to cases where the person whose job is in question can only function through advice to the judge.

Litigants should take adverse decisions more gracefully. Forrester concludes that absolute immunity does not protect the judge, even when the employee in question "can only function through advice to the judge." It would make no sense to hold the judge liable in a case such as Forrester but to excuse the judge in a case where the employee is farther removed from judicial duties. Forrester holds that the extent of immunity depends on the capacity in which the judge acts; judicial decisions are immune absolutely, while administrative decisions must be evaluated under principles of qualified immunity. Judge Krajewski played an administrative role when firing Kurowski and Nicholls.

C. Qualified immunity is the next question. Krajewski contends that he could not reasonably have been expected to know when he fired Kurowski and Nicholls that his conduct was unconstitutional. Since Branti was decided in 1980, five years before Krajewski fired them, this is not a powerful submission. To the extent it has any substance, it rests on the proposition, which we have already rejected, that judges have greater freedom to fire assistant public defenders than do chief public defenders (the defendant in Branti ). The proposition is no more persuasive now than it was a few pages ago.

True, Branti did not involve a judge as a defendant, but earlier cases need not be identical to the one at hand. The question is whether "the contours of the right [are] sufficiently clear that a reasonable official would understand that what he is doing violates the right." Anderson v. Creighton, --- U.S. ----, 107 S. Ct. 3034, 3039, 97 L. Ed. 2d 523 (1987). Branti gave substance to the generalities of the first amendment. A reasonable person in Judge Krajewski's position would have understood in December 1985 that the Constitution forbids firing an assistant public defender on the basis of politics. The magistrate concluded that Krajewski actually did so understand--which is why the magistrate awarded punitive damages.

D. Krajewski was gambling that absolute immunity would protect him from the consequences of a clear constitutional wrong. The bet did not pay off. The Supreme Court's reversal of our opinion in Forrester is, however, the basis of a last gasp. Krajewski contends that even if Branti clearly established the constitutional right in question, a judge acting in December 1985 could not reasonably have predicted that he would be held liable for such an act. Since principles of immunity have been clarified so recently, the argument concludes, Krajewski is entitled to avoid liability.

This is equivalent to the argument that Forrester should not be applied retroactively, despite the presumption that a court applies the law in force at the time of its decision. Bradley v. Richmond School Board, 416 U.S. 696, 94 S. Ct. 2006, 40 L. Ed. 2d 476 (1974). It is an application of the positivism of Justice Holmes, of his view that "law" is nothing more than a prediction of what judges will do. Oliver Wendell Holmes, The Path of the Law, 10 Harv.L.Rev. 457, 460-61 (1897), reprinted in Collected Legal Papers 173 (1920). The argument runs: In December 1985 a reasonable person might have predicted that a judge would be immune from damages for firing a public defender; therefore "the law" in 1985 had not clearly established that judges could not fire public defenders; that law has changed, but officials are not obliged to anticipate changes in the law.

A case such as this illustrates that there may be "law" without a judicial remedy. Holmes's law-as-prediction approach focuses on what judges do and is useful for that limited purpose. Holmes had a "bad man" theory of the law and wanted to be able to tell that "bad man" the potential penalties, for these penalties determine the degree of compliance. The Constitution creates rules that apply to all governmental actors, however, no matter what judges will do to implement them. In some cases no one will have standing to sue. United States v. Richardson, 418 U.S. 166, 94 S. Ct. 2940, 41 L. Ed. 2d 678 (1974). There is nonetheless a rule of law, which all parts of the government must observe. The Guarantee Clause did not lose its meaning when the Supreme Court declared controversies under it non-justiciable, see Luther v. Borden, 48 U.S. (7 How.) 1, 43, 12 L. Ed. 581 (1849). The decision meant only that the rules would be enforced by officials other than judges. Cf. Deborah Jones Merritt, The Guarantee Clause and State Autonomy: Federalism for a Third Century, 88 Colum.L.Rev. 1 (1988). Every public actor must conform to the governing rule, no matter how slight the penalty for disobedience. Rules without judicial sanctions are rules just the same. The legislature could impeach and remove Judge Krajewski from office for violating the Constitution, or the Supreme Court of Indiana impose intra-branch discipline, whether or not federal judges are prepared to extract damages.

The principle that law may exist independently of the anticipated sanction has more general application. Changes in the probability of suit and the availability of defenses will affect the sanction but do not necessarily change "the law". A would-be robber might think to himself that the expected penalty for theft is slight, because the local prosecutor has been focusing on drugs. If a new prosecutor cracks down on street crime, so that the expected penalty (the sentence on conviction times the probability of successful prosecution) rises dramatically after the crime has been committed, the robber has no legitimate beef. He has simply lost a gamble. Cf. Prater v. United States Parole Commission, 802 F.2d 948, 953 (7th Cir.1986) (en banc), remarking that even "[s]ettled expectations regarding the vigor of enforcement are unreasonable"--not because factually unsupported, but because the social costs of recognizing them as legitimate would be too high. If the legislature increases the length of the statute of limitations after tortious acts have been committed, the tortfeasor has no legitimate complaint. Chase Securities Corp. v. Donaldson, 325 U.S. 304, 65 S. Ct. 1137, 89 L. Ed. 1628 (1945). Consider too what must be a frequent occurrence: a court of appeals says that A, B, and C are the elements of bank robbery (or some other crime), and a year later the Supreme Court reverses, concluding that A and B are sufficient in themselves. During the interim, was it "the law" that someone who did A and B had robbed a bank? Perhaps not in the sense that the appellate court would penalize the conduct immediately; certainly yes in the sense that the acts met the statutory definition of a crime, and a penalty ultimately might be imposed.

So here. If Judge Krajewski thought that judicial immunity would allow him to get away with a violation of the Constitution, this expectation--factually reasonable though it may have been--is not one the judicial system is prepared to recognize as legitimate. Gamblers must be prepared to pay up if they lose; the gamble is not itself a source of immunity. Branti sets the standards to which public officials must conform; Forrester holds that judges should be treated like other officials to the extent they play administrative roles. Judges are supposed to look over their shoulders when making administrative decisions, just as other public officials must; that is the rationale of Forrester, which treated absolute immunity as a harsh medicament, to be administered sparingly.

Judge Krajewski might respond that to change an immunity rule retroactively would dissipate part of the benefit of immunity. Absolute immunity is designed not only to prevent erroneous decisions and spare judges the costs of discovery and trial, but also to promote fearless action. A judge concerned about the possibility of legal change would act as if at risk, even when the rule then in force was nominally "absolute" immunity. To preserve the benefits of truly absolute immunity, the argument would conclude, courts must enforce whatever rule prevailed at the time of the acts. This argument loses its punch, however, if we assume--as judges of inferior courts must--that the rule applied by the Supreme Court today is superior to the rule in force earlier. The possibility of change will induce judges and other public officials to behave more like they would have behaved under the "correct" rule. For example, if absolute immunity were the rule in 1985, and a judge feared that it would be replaced by qualified immunity, he would take some extra care. But we now know that this is a good thing. Similarly, if the rule in 1985 had been qualified immunity, only to be replaced by absolute immunity, a judge with his eye on the development of the law would have taken a little less precaution in 1985 than the prevailing rule seemed to require; and we would know from the adoption of the new rule that this, too, was a move in the right direction. Anticipation of change in rules of liability always induces improvements, if the "new" rules are better than the "old".

For what it is worth, we doubt that a reasonable judge would have believed in December 1985 that he had absolute immunity when firing public defenders. Six years earlier the Supreme Court had held that Members of Congress, whose general immunity is as absolute as that of judges, may be liable for employment discrimination. Davis v. Passman, 442 U.S. 228, 99 S. Ct. 2264, 60 L. Ed. 2d 846 (1979). Simultaneously with Forrester, this court released an opinion holding that judges are not absolutely immune when firing court reporters. McMillan v. Svetanoff, 793 F.2d 149 (7th Cir.1986). The opinions of both panels recognized that judges ordinarily are not absolutely immune when making administrative decisions; the immunity covers only "judicial" acts. Forrester portrayed its holding as a narrow exception to the rule, one depending on the interaction between judge and probation officer in fulfilling judicial tasks. Even so, one judge dissented, and the Supreme Court unanimously reversed. Public defenders are advocates, not judicial officers; they are farther removed from core judicial functions than are the court reporters dealt with in McMillan. The magistrate denied Krajewski's motion for summary judgment before the Supreme Court's decision in Forrester. See also McDonald v. Krajewski, 649 F. Supp. 370, 373-76 (N.D.Ind.1986) (a pre-Forrester decision rejecting Judge Krajewski's immunity defenses in the case of a politically-motivated discharge of a clerk-secretary in Judge Krajewski's court). The Supreme Court's opinion in Forrester did not strip from Judge Krajewski an immunity he previously enjoyed.

Kurowski and Nicholls recovered about $14,000 for their attorneys' fees under Sec. 1988. This is less than 25% of the monetary recovery, and of course plaintiffs won reinstatement to boot. The request--the magistrate gave plaintiffs everything they wanted--was modest as requests go in civil rights litigation. See City of Riverside v. Rivera, 477 U.S. 561, 106 S. Ct. 2686, 91 L. Ed. 2d 466 (1986) (award of about $245,000 in a case with damages under $34,000). It was also less than the contingent-fee benchmark that we have said is appropriate, when such a fee provides adequate compensation. Kirchoff v. Flynn, 786 F.2d 320 (7th Cir.1986). Appellate review of an award is deferential, see City of Riverside, and the magistrate did not abuse his discretion in giving plaintiffs 100% of a modest request. All litigation should be conducted so frugally!

Krajewski challenges the attorneys' hourly rate ($100 per hour for the two lawyers, one of whom is a professor of constitutional law at Valparaiso University), the fact that two lawyers worked on the case, counsel's dedication of a few hours' research to ideas that did not pan out (one of which was not even raised at trial), and--apparently the crowning blow in defendant's mind--the submission of a bill for 1.6 hours' time to prepare the request for fees.

The bill in this case is an impressively low figure for taking a case through discovery, several rounds of motions for summary judgment, and trial. The request in Kirchoff for less work (ending in a lower recovery) was $50,000, and we held an award of only $10,000 to be an abuse of discretion. In Frantz v. United States Powerlifting Federation, 836 F.2d 1063 (7th Cir.1987), counsel asked for $44,000 for less work than this case required; in In re Central Ice Cream Co., 836 F.2d 1068 (7th Cir.1987), on statement of fees, 841 F.2d 732 (7th Cir.1988), counsel requested about $65,000 for monitoring a case while it was not being briefed, and we awarded $20,000 for this task.

It is fair to judge a pudding by the eating, and fair to judge litigating decisions by their results. The use of two (or more) lawyers, which solvent clients commonly pay for because they believe extra help beneficial, may well reduce the total expenditures by taking advantage of the division of labor. The results in this case suggest that this benefit has been realized. The scholar did the research, the litigator the litigating. Awards of fees under Sec. 1988 are supposed to give counsel the market rate for their time, Blum v. Stenson, 465 U.S. 886, 892-95, 104 S. Ct. 1541, 1545-47, 79 L. Ed. 2d 891 (1984), and the kinds of billing and staffing arrangements struck when clients must pay their own lawyers are the best benchmarks of the arrangements (and bills) appropriate under Sec. 1988. See Bohen v. City of East Chicago, 666 F. Supp. 154 (N.D.Ind.1987). The use of two lawyers was appropriate here, as it was in Bohen--a case in which counsel spent more time (and were awarded more than $30,000) to achieve lesser results. The hourly rate of $100 is appropriate in the Northern District of Indiana; the magistrate has experience with the billing rates in that market, and we are not inclined to disturb his judgment.

That the lawyers spent some time in blind alleys is irrelevant; this is inevitable, and the hourly rate reflects the fact that not all time is equally productive. It would be possible, but silly, to bill $150 for productive time and nothing for other time; the standard hourly rate reflects the standard productivity of time. Hensley v. Eckerhart, 461 U.S. 424, 433-35, 103 S. Ct. 1933, 1939-40, 76 L. Ed. 2d 40 (1983), says that the court should subtract time for losing claims, but a losing argument in support of a successful claim for relief is fully compensable time. Lenard v. Argento, 808 F.2d 1242 (7th Cir.1987). This is still another aspect of the point that the hourly rate includes hours with different degrees of productivity. The defendant need only compensate plaintiff for fees to the extent plaintiff succeeds; losing claims are not compensable; but this suit was 100% successful. As for Krajewski's bete noire, the request for $160 to prepare the bill: counsel are entitled to full compensation, including the time spent pursuing requests for fees. Muscare v. Quinn, 680 F.2d 42 (7th Cir.1982). This request is the smallest we have ever seen for this function, and we commend plaintiffs' counsel for their parsimony.

Counsel are entitled to full compensation under Sec. 1988 for productive work. The work in this case was completely successful. The plaintiffs' appellate briefs and argument (for which plaintiffs are entitled to further compensation under Sec. 1988) were first-rate. Niggling objections such as those raised by Krajewski simply increase the ultimate bill and waste judicial time. We trust that the Attorney General of Indiana will in future abjure the buckshot approach to appellate litigation.

AFFIRMED.