Section 803(d) of the Prison Litigation Reform Act, codified at 42 U.S.C. § 1997e(d), sets both absolute and relative limits on attorneys’ fee shifting. The district court held these limits unconstitutional because they disadvantage prisoners compared with other plaintiffs, whose recoveries under 42 U.S.C. § 1988(b) in constitutional-tort litigation are not subject to any statutory maximum. Johnson v. Daley, 117 F.Supp.2d 889 (W.D.Wis.2000). Every court of appeals that has considered this question has held, to the contrary, that § 1997e(d) is within Congress’ authority. See Boivin v. Black, 225 F.3d 36 (1st Cir.2000); Hadix v. Johnson, 230 F.3d 840 (6th Cir.2000); Walker v. Bain, 257 F.3d 660 (6th Cir.2001); Foulk v. Charrier, 262 F.3d 687 (8th Cir.2001); Madrid v. Gomez, 190 F.3d 990 (9th Cir.1999); Jackson v. State Board of Pardons & Paroles, 331 F.3d 790 (11th Cir.2003). Accord, Collins v. Algarin, 1998 WL 10234 (E.D.Pa. Jan. 9, 1998) 1998 U.S. Dist. Lexis 83, affirmed by an equally divided court under the name Collins v. Montgomery County Board of Prison Inspectors, 176 F.3d 679, 686 (3d Cir.1999) (en banc). Like these other circuits, we hold that § 1997e(d) is rationally related to valid objectives and hence is within the legislative power, whether or not it is wise.
I
Section 1997e(d) provides:
(1)In any action brought by a prisoner who is confined to any jail, prison, or other correctional facility, in which attorney’s fees are authorized under [42 U.S.C. § 1988], such fees shall not be awarded, except to the extent that—
(A) the fee was directly and reasonably incurred in proving an actual violation of the plaintiffs rights protected by a statute pursuant to which a fee may be awarded ...; and
(B)(i) the amount of the fee is proportionately related to the court ordered relief for the violation; or (ii) the fee was directly and reasonably incurred in enforcing the relief ordered for the violation.
(2) Whenever a monetary judgment is awarded in an action described in paragraph (1), a portion of the judgment (not to exceed 25 percent) shall be applied to satisfy the amount of attorney’s fees awarded against the defendant. If the award of attorney’s fees is not greater than 150 percent of the judgment, the excess shall be paid by the defendant.
(3) No award of attorney’s fees in an action described in paragraph' (1) shall be based on an hourly rate greater than 150 percent of the hourly rate established under [18 U.S.C. § 3006A] for payment of court-appointed counsel.
(4) Nothing in this subsection shall prohibit a prisoner from entering into an agreement to pay an attorney’s fee in an amount greater than the amount authorized under this subsection, if the fee is paid by the individual rather than by the defendant pursuant to [§ 1988].
Subsections (1) and (2) establish relative limits: fees must be “proportionately related to the court ordered relief’ and, when monetary relief is awarded, the fees attributable to that relief cannot exceed 150% of the damages. Subsection (3). establishes an absolute limit at 150% of the hourly rate for defense counsel under the Criminal Justice Act, times the number of hours reasonably devoted to the litigation. Because the CJA rate (set by the Judicial Conference of the United States) currently is $90 per hour, the maximum that the defendant may be directed to underwrite *584is $135 per hour.† The total amount that an attorney may receive, however, is greater, not only because the attorney is entitled to 25% of the judgment under subsection (2) but also because the client is free under subsection (4) to agree by contract to pay more — out of the recovery or out of other assets.
This case shows how the statute works. Cedric Johnson sued George Daley, the medical director of the Bureau of Correctional Health Services for the Wisconsin Department of Corrections, under 42 U.S.C. § 1983, contending that Daley subjected him to cruel and unusual punishment by waiting three years before certifying that Johnson, whose alcoholism had damaged his liver, was eligible for a transplant at public expense. Johnson contended that Daley had been deliberately indifferent to his serious medical need. See Farmer v. Brennan, 511 U.S. 825, 114 S.Ct. 1970, 128 L.Ed.2d 811 (1994); Estelle v. Gamble, 429 U.S. 97, 97 S.Ct. 285, 50 L.Ed.2d 251 (1976). Johnson was put on the eligibility list in June 1999 and appears to have suffered no long-term injury from the delay. Nonetheless, a jury agreed with Johnson that he should have been made eligible sooner and awarded him $10,000 in compensatory damages, plus $30,000 in punitive damages. Attorneys from Foley & Lardner and Heller Ehrman White & McAuliffe represented Johnson at the district judge’s request; Johnson did not enter into an agreement with counsel under § 1997e(d)(4), so compensation depends entirely on the application of subsections (1) through (3). Counsel asked the judge to direct Daley to pay $92,997.20 in attorneys’ fees. This request exceeds both relative and absolute máximums: the relative cap under subsection (2) is $60,000 in fees (150% of the judgment), and application of the absolute cap in subsection (3) produces a lower award of $36,451.50. (This figure comes from Johnson’s lawyers. Daley and the United States have not questioned its accuracy, nor have we plumbed the details of its calculation.) Counsel sought compensation for 525.1 hours of work — much of it by paralegals with rates under $135 per hour, but some time by partners who contended that their market rate is as high as $325 per hour. Daley did not deny that this legal time had been reasonably devoted to the case, though he did dispute the hourly rates.
Under § 1997e(d) counsel could receive a maximum of $46,451.50 for legal services — $10,000 from the award plus $36,451.50 extra from Daley. As we read *585subsection (2), attorneys’ compensation comes first from the damages, as in ordinary tort litigation, and only if 25% of the award is inadequate to compensate counsel fully may defendant be ordered to pay more under § 1988. Cf. Gisbrecht v. Barnhart, 585 U.S. 789, 122 S.Ct. 1817, 152 L.Ed.2d 996 (2002) (using this approach in Social Security cases, where any excess would be provided by the Equal Access to Justice Act, 28 U.S.C. § 2412(d), rather than § 1988). The district court proceeded in a different way, first calculating counsel’s entitlement and then determining how much of this should be satisfied from the damages. After notifying the United States, which intervened to defend the constitutionality of the statute, the district court declined to enforce subsections (2) and (3). Subsection (l)(B)(i), which provides that fees must be “proportionately related” to the violation, is in the district court’s view an appropriate cap — because the judge gets to determine how high a “proportion” to use. The court concluded that attorneys’ fees should be set at $80,000, or 200% of the judgment, under § 1988, writing that this level is not disproportionate to the damages. Next the judge held that Johnson should contribute only $200 of this out of the judgment, leaving him with $89,800 while Daley has been ordered to pay a total of $128,578.81: $10,000 in compensatory damages, $30,000 in punitive damages, $79,800 in attorneys’ fees, and $8,778.81 in costs. The record does not disclose how much, if any, of this tab will be picked up by the State of Wisconsin as Daley’s employer. Both Daley and the United States have appealed; the appeal is limited to the amount by which the attorneys’ fees exceed the maximum allowed by § 1997e(d)(2) and (3). Counsel have not cross-appealed to seek a greater portion of the damages awarded to Johnson.
II
A
The district court held that § 1997e(d)(2) and (3) are incompatible with the due process clause of the fifth amendment, which since Bolling v. Sharpe, 347 U.S. 497, 74 S.Ct. 693, 98 L.Ed. 884 (1954), has been deemed to include an equal-protection principle. See, e.g., Vance v. Bradley, 440 U.S. 93, 99 S.Ct. 939, 59 L.Ed.2d 171 (1979). Yet the Prison Litigation Reform Act (PLRA from now on) does not rest on any of the powers granted by Article I of the Constitution; its genesis is § 5 of the, fourteenth amendment, which says that “Congress shall have power to enforce, by appropriate legislation, the provisions of this article.” Legislation under the power granted by § 5 is not necessarily subject to limitations on the original grants of national power; this is why Fitzpatrick v. Bitzer, 427 U.S. 445, 96 S.Ct. 2666, 49 L.Ed.2d 614 (1976), holds that Congress may use its § 5 power to subject states to suit in federal court, despite the contrary text of the eleventh amendment. So the question to ask is whether § 1997e(d) “enforces” the fourteenth amendment, which points us to the equal protection clause in § 1 of that amendment without any need to detour through the fifth amendment — for a law at odds with the fourteenth amendment’s substantive provisions cannot be one to “enforce” them. See, e.g., City of Boerne v. Flores, 521 U.S. 507, 117 S.Ct. 2157, 138 L.Ed.2d 624 (1997); University of Alabama v. Garrett, 531 U.S. 356, 121 S.Ct. 955, 148 L.Ed.2d 866 (2001).
Legislation that does not burden a suspect class or affect fundamental rights satisfies the equal-protection requirement if the legislature could think the rule rationally related to any legitimate goal of government. Prisoners are not a *586suspect class; conviction of crime justifies the imposition of many burdens. See, e.g., Connecticut Dep’t of Public Safety v. Doe, 538 U.S. 1 (2003) (public identification as a felon); Hudson v. United States, 522 U.S. 93, 118 S.Ct. 488, 139 L.Ed.2d 450 (1997) (occupational debarment). Nor is there a fundamental right to have one’s adversary, or the public treasury, defray all or part of the cost of litigation. That is why we held the three-strikes provision in the PLRA, 28 U.S.C. § 1915(g), compatible with the Constitution. See Lewis v. Sullivan, 279 F.3d 526 (7th Cir.2002). Although prisoners enjoy a fundamental right of access to the courts, see Lewis v. Casey, 518 U.S. 343, 116 S.Ct. 2174, 135 L.Ed.2d 606 (1996), there is no right of subsidized access. See United States v. Kras, 409 U.S. 434, 93 S.Ct. 631, 34 L.Ed.2d 626 (1973). The right to publish a newspaper does not imply a right to governmental funding, nor does a right to read books imply entitlement to a public library that circulates books without charge. A woman’s right to choose whether to have an abortion does not imply a right to have the government cover the medical costs. See Maher v. Roe, 432 U.S. 464, 97 S.Ct. 2376, 53 L.Ed.2d 484 (1977). A right to education does not imply a right to free transportation to school. See Kadrmas v. Dickinson Public Schools, 487 U.S. 450, 108 S.Ct. 2481, 101 L.Ed.2d 399 (1988). A right to petition for redress of grievances does not imply a right to free writing paper and stamps. And a right to seek redress in court does not imply entitlement to have someone else pay for your lawyer. The Supreme Court made this pellucid, for prisoners in particular, when holding in Murray v. Giarratano, 492 U.S. 1, 109 S.Ct. 2765, 106 L.Ed.2d 1 (1989), that a right to wage a collateral attack on one’s conviction does not entail a right to counsel at public expense, even in a capital case.
Legislation singling out prisoners accordingly is analysed under the rational-basis standard — as the plaintiffs concede, and as the Supreme Court held in McDonald v. Board of Election Commissioners, 394 U.S. 802, 89 S.Ct. 1404, 22 L.Ed.2d 739 (1969). See also Zehner v. Trigg, 133 F.3d 459 (7th Cir.1997) (applying the rational-basis standard to 42 U.S.C. § 1997e(e), a part of the PLRA that requires prisoners to show physical injury as a condition to recovery, and holding that the statute is valid under that standard). Cf. Marshall v. United States, 414 U.S. 417, 94 S.Ct. 700, 38 L.Ed.2d 618 (1974). Chief Justice Warren explained in McDonald how the rational-basis standard works:
The distinctions drawn by a challenged statute must bear some rational relationship to a legitimate state end and will be set aside as violative of the Equal Protection Clause only if based on reasons totally unrelated to the pursuit of that goal. Legislatures are presumed to have acted constitutionally even if source materials normally resorted to for ascertaining their grounds for action are otherwise silent, and their statutory classifications will be set aside only if no grounds can be conceived to justify them. See McGowan v. Maryland, 366 U.S. 420, 81 S.Ct. 1101, 6 L.Ed.2d 393 (1961); Kotch v. Board of River Port Pilot Commissioners, 330 U.S. 552, 67 S.Ct. 910, 91 L.Ed. 1093 (1947); Lindsley v. Natural Carbonic Gas Co., 220 U.S. 61, 31 S.Ct. 337, 55 L.Ed. 369 (1911). With this much discretion, a legislature traditionally has been allowed to take reform “one step at a time, addressing itself to the phase of the problem which seems most acute to the legislative mind,” Williamson v. Lee Optical of Oklahoma, Inc., 348 U.S. 483, 489, 75 S.Ct. 461, 99 L.Ed. 563 (1955); *587and a legislature need not run the risk of losing an entire remedial scheme simply because it failed, through inadvertence or otherwise, to cover every evil that might conceivably have been attacked. See Ozan Lumber Co. v. Union County National Bank, 207 U.S. 251, 28 S.Ct. 89, 52 L.Ed. 195 (1907).
394 U.S. at 809, 89 S.Ct. 1404. Under this standard, a legislative decision “is not subject to courtroom fact-finding and may be based on rational speculation unsupported by evidence or empirical data.” FCC v. Beach Communications, Inc., 508 U.S. 307, 315, 113 S.Ct. 2096, 124 L.Ed.2d 211 (1993). See also, e.g., Nordlinger v. Hahn, 505 U.S. 1, 15, 112 S.Ct. 2326, 120 L.Ed.2d 1 (1992); Minnesota v. Clover Leaf Creamery Co., 449 U.S. 456, 463-65, 101 S.Ct. 715, 66 L.Ed.2d 659 (1981); Vance v. Bradley, 440 U.S. at 111, 99 S.Ct. 939. “[R]ational-basis review in equal protection analysis ‘is not a license for courts to judge the wisdom, fairness, or logic of legislative choices.’” Heller v. Doe, 509 U.S. 312, 319, 113 S.Ct. 2637, 125 L.Ed.2d 257 (1993) (citations omitted). See also, e.g., New Orleans v. Dukes, 427 U.S. 297, 96 S.Ct. 2513, 49 L.Ed.2d 511 (1976); Massachusetts Board of Retirement v. Murgia, 427 U.S. 307, 96 S.Ct. 2562, 49 L.Ed.2d 520 (1976); University of Alabama v. Garrett, 531 U.S. at 366-67, 121 S.Ct. 955 (holding that Cleburne v. Cleburne Living Center, 473 U.S. 432, 105 S.Ct. 3249, 87 L.Ed.2d 313 (1985), on which the district court relied in this case, does not establish a sliding-scale approach).
B
Whether the line drawn in the PLRA between prisoners’ suits and free persons’ suits is rational may depend on answering the question: “equal with respect to what ”? The district court compared § 1997e(d) with § 1988, which entitles prevailing parties in constitutional cases to recover reasonable attorneys’ fees. Section 1988 has been treated as asymmetric — that is, prevailing plaintiffs recover their legal expenses but prevailing defendants do not. See Christiansburg Garment Co. v. EEOC, 434 U.S. 412, 98 S.Ct. 694, 54 L.Ed.2d 648 (1978) (defendants recover legal expenses only if the suit is frivolous — in which event they would be entitled to recompense even without regard to a statute). A plaintiff achieves “prevailing party” status by recovering any judgment, even for nominal damages. Compare Farrar v. Hobby, 506 U.S. 103, 113 S.Ct. 566, 121 L.Ed.2d 494 (1992), with Buckhannon Board & Care Home, Inc. v. West Virginia Dep’t of Health & Human Resources, 532 U.S. 598, 121 S.Ct. 1835, 149 L.Ed.2d 855 (2001). Although fees must be “reasonable”, so that de minimis or Pyrrhic victories do not support fee-shifting, see Maul v. Constan, 23 F.3d 143, 145, 147 (7th Cir.1994), the flexible reasonableness standard permits a court to award legal fees that substantially exceed the damages, see Riverside v. Rivera, 477 U.S. 561, 106 S.Ct. 2686, 91 L.Ed.2d 466 (1986), even though no solvent private litigant in tort or contract litigation would agree to pay counsel more than the anticipated recovery (or the value of equitable relief). So § 1988 is exceptionally pro-plaintiff. The district judge asked why, if medicine this strong is needed to promote constitutional claims by free persons against state actors, prisoners do not need at least as much assistance.
This approach assumes, however, that prisoners and free persons have similar constitutional claims. As Johnson’s case shows, that is not altogether true. Free persons are not constitutionally entitled to liver transplants or other costly medical care at public expense. States have extra obligations toward prisoners and must provide care appropriate to their serious med*588ical needs because imprisonment takes away their ability to fend for themselves. See DeShaney v. Winnebago County Department of Social Services, 489 U.S. 189, 109 S.Ct. 998, 103 L.Ed.2d 249 (1989). A complaint along the lines of Johnson’s, if made by a free person, usually would take the form of a contention that a physician committed the tort of malpractice. And the prevailing party in tort litigation must bear 100% of his own attorneys’ fees; that’s the American Rule. See Alyeska Pipeline Service Co. v. Wilderness Society, 421 U.S. 240, 95 S.Ct. 1612, 44 L.Ed.2d 141 (1975). Now it is true that a free person may recover for negligence, while a prisoner must show intentional misconduct (where “deliberate indifference” to his needs counts as intentional); the free person therefore is more likely to prevail in any given situation. But the free person still must cover his own legal costs, even if he shows gross negligence or wilful misconduct by the physician. Such proof may lead to awards of punitive damages (as in Johnson’s case) but will not require the defendant to reimburse the plaintiff for the cost of legal representation. If we compare prisoners under § 1997e(d) with free persons who received bad medical care, perhaps the question should be why prisoners receive legal services at defendants’ expense, and free persons do not.
Ordinary tort litigation is not the only option. Suppose Johnson had been free and unable to pay for a liver transplant. He might have requested medical care under Medicaid — but if the responsible agency had delayed making him eligible, there would not have been any option to litigate. The agency’s decision about applications for individual benefits is the end of the line under the Medicaid program. See Heckler v. Ringer, 466 U.S. 602, 104 S.Ct. 2013, 80 L.Ed.2d 622 (1984). Such a free person is decidedly worse off than a prisoner. Or consider how Johnson would have been treated as a veteran seeking medical care in a Veterans Administration hospital. A decision by the hospital’s staff deferring his eligibility for a liver transplant would have been subject to review by an administrative tribunal but not the Article III courts. Compare 38 U.S.C. § 7252 with 38 U.S.C. § 7292(d)(2). And, until recently, veterans were prohibited from paying attorneys more than $10 out of their own resources to seek a favorable result in the administrative process. The Supreme Court held in Walters v. National Ass’n of Radiation Survivors, 473 U.S. 305, 105 S.Ct. 3180, 87 L.Ed.2d 220 (1985), that this cap was constitutional because Congress was entitled to rely on the bar (and veterans’ organizations) to donate free legal services. Once again prisoners are decidedly better off under the PLRA.
If a veteran suffers on account of substandard medical care at a VA hospital, the legal remedy is under the Federal Tort Claims Act. This statute forbids punitive damages. 28 U.S.C. § 2674 ¶ 1. It caps attorneys’ fees at 25% of any judgment (20% if the case is settled). 28 U.S.C. § 2678. Those legal fees are deducted from the plaintiffs recovery; the United States does not cover any part of a prevailing plaintiffs legal expenses. A prisoner can offer his lawyer at defendant’s expense as much as 150% of the recovery (plus a quarter of the recovery itself, and any additional amount provided by contract), while a veteran can offer counsel no more than 25% of a recovery that never includes punitive damages. Moreover, the plaintiff is forbidden to supplement these fees by private contract with his lawyer. The Westfall Act, 28 U.S.C. § 2679(d), makes this remedy against the United States exclusive; any effort to evade the cap by suing federal employees will be defeated by replacing them with the United States as the sole defendant. See generally Gu*589tierrez de Martinez v. Lamagno, 515 U.S. 417, 115 S.Ct. 2227, 132 L.Ed.2d 375 (1995). A litigant who can show that the United States took in the litigation itself a position that was not substantially justified may recover some attorneys’ fees under the Equal Access to Justice Act, 28 U.S.C. § 2412(d), but even the most egregious pre-litigation conduct by federal agents or employees does not permit an award of attorneys’ fees. Litigants who qualify for an award under the EAJA encounter a statutory cap: “attorney fees shall not be awarded in excess of $125 per hour unless the court determines that an increase in the cost of living or a special factor, such as the limited availability of qualified attorneys for the proceedings involved, justifies a higher fee” (§ 2412(d)(2)(A)(ii)). By contrast, the PLRA permits the court to award fees up to $135 per hour, to be paid by the defendant, without any need to prove that the defendant’s conduct in the litigation was not “substantially justified.”
Counsel readily agree to represent veterans and other injured persons in suits under the FTCA, even though these suits are subject to caps more stringent than those confronting prisoners under the PLRA. Congress rationally could conclude in light of this experience that prisoners, like veterans, will be able to enjoy the benefit of counsel, when they have good claims (a vital qualifier).
It is difficult to ascribe the caps in the PLRA to irrational antipathy, when prisoners fare better under the PLRA than do veterans and other free persons who must bear their own legal expenses under the American Rule and the FTCA. Caps on attorneys’ fees, far from being unique to prisoners (as the district judge supposed), are common in litigation against governments and their employees. For example, persons claiming benefits under the Social Security program must pay their own lawyers (unless the EAJA applies because the government’s litigating position was unreasonable); and it is unlawful for any claimant to agree to pay counsel more than 25% of past-due benefits as compensation for legal services rendered in obtaining those benefits. 42 U.S.C. § 406(b) (discussed in Gisbrecht, supra). School children and their parents, seeking a better education under the Individuals with Disabilities Education Act, may not recover more than $1,300 in attorneys’ fees (at rates of more than $50 per hour) in suits within the District of Columbia. D.C.Code. § 11-2604. That limit (which has since been raised) was sustained as rational, even though Congress allows greater awards outside the District. See Calloway v. District of Columbia, 216 F.3d 1, 4-5 (D.C.Cir.2000). In a world where veterans, employees who have paid employment taxes for a half century, and school children must pay for their own lawyers, or accept reimbursement at low rates, it is difficult to treat as irrational a statute limiting to $135 per hour, or 150% of the recovery, the amount of legal fees that courts may require defendants to pay in prisoners’ cases.
Prisoners may think of $135 per hour as munificent, compared with the compensation of their lawyers in pretrial and trial proceedings. Solvent defendants must pay all of their legal expenses and are not reimbursed if they prevail. For insolvent defendants, the Criminal Justice Act, 18 U.S.C. § 3006A, in conjunction with decisions of the Judicial Conference, caps at $90 per hour how much the public fisc will pay for counsel — yet criminal defendants have a right to counsel far stronger than any entitlement to fee-shifting in civil suits. See Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963). The hourly cap under the CJA combines with a per-case cap of $5,200 in a felony prosecution. Although courts are *590authorized to approve compensation exceeding $5,200 in exceptional cases, few criminal trials are so exceptional that defense counsel take home $36,000, the PLRA cap for Johnson. If this had been a criminal prosecution, and Johnson had been a defendant rather than the plaintiff, his lawyers would have been paid considerably less than $86,000. If $5,200 is enough to secure a competent criminal defense— and it is — how can the PLRA cap of $36,000 be unconstitutionally low for a civil plaintiff? The stakes of the criminal prosecution for the defendant are considerably greater than the stakes in most post-imprisonment civil suits.
The kind of post-imprisonment litigation most important to inmates is the collateral attack — on the conviction itself or on decisions affecting good-time credits (and thus the computation of the release date). Yet, when seeking release, a prisoner is entitled to no legal assistance at all. If the prisoner obtains legal assistance and prevails, with the court declaring that the incarceration was unconstitutional from the outset, the state government will not be required to pay a single penny of counsel’s fees. Courts and defendants alike rely on — and receive — legal services donated by the bar or supplied by legal assistance bureaus (defender programs, legal aid clinics at law schools, programs underwritten by the Legal Services Corporation, or programs funded by IOLTA proceeds, see Brown v. Legal Foundation of Washington, — U.S. -, 123 S.Ct. 1406, 155 L.Ed.2d 376 (2003)). Only in capital cases does any federal statute provide for post-conviction counsel, and even then the rate is capped at $125 per hour, see 21 U.S.C. § 848(q)(10)(A), or $10 less than the current maximum under the PLRA.
If asked why $0 for most collateral attacks and prosecutions of solvent persons, $90 for criminal defense under the CJA if the defendant is insolvent, $125 under the EAJA and collateral attacks in capital cases, $135 under the PLRA, and 25% of back benefits under the Social Security system, we might be unable to give a convincing answer. Caps have an arbitrary quality; different majorities in the legislature at different times have different willingness to expend public funds (or write checks that must be paid by state actors such as Dr. Daley, or the states themselves as indemnitors). Some of these caps depend on an interaction between legislation and decisions of the Judicial Conference, and these different bodies may have different objectives. Some Congresses favor more litigation and adjust fee schedules to promote it; other Congresses pay more attention to the costs of adding cases to the docket (including the costs to defendants and to the federal system when Congress is adopting rules for state actors) and adjust rules to make litigation less attractive. There is no one right answer to the question how much litigation there should be, and who should pay for that litigation.
We are conscious that the numbers we have given are not directly comparable. Criminal defense counsel receives $90 per hour win or lose; a prisoner’s lawyer receives as much as $135 per hour, or a Social Security claimant’s lawyer 25% of the back benefits, only in the event of victory, as with a tort lawyer on contingent fee; an award at $125 per hour under the EAJA depends on both prevailing in the litigation and showing that the government’s position was not substantially justified. So the actuarial value of an hour devoted to the case by criminal defense counsel may exceed the value of an hour under the PLRA. But the value of an hour under the PLRA exceeds that of an hour under the EAJA — and any fee-shifting system offers more to counsel than does *591the no-shifting approach on post-conviction challenges to the fact of incarceration.
The United States Code and its implementing rules are the work of thousands of different actors over scores of years; consistency is not possible. But what this means is that all of these different systems could be thought rational solutions to the question “how much may plaintiffs be allowed to spend for legal services, how much of that must be paid for by the losing side, and how much of the cost of litigation will be covered by the public fisc?” Litigation produces benefits (and sometimes costs) for third parties; it is to this extent a public good, and determining how much of a public good to supply (and at whose cost) is an intractable problem. The American Rule is a rational approach; the British loser-pays rule is a rational approach; asymmetric fee-shifting in § 1988 is a rational approach; asymmetric fee shifting plus compensation for the risk of loss in order to induce counsel to be indifferent between paying clients and chancy constitutional claims would be rational (and is used in common-fund cases, though not under statutes such as § 1988, see Burlington v. Dague, 505 U.S. 557, 112 S.Ct. 2638, 120 L.Ed.2d 449 (1992)); and fee caps such as the FTCA, the EAJA, and the PLRA also represent rational approaches. The observation that prisoners receive less under the PLRA than under § 1988 no more shows that the PLRA is irrational, than the fact that defendants pay more under § 1988 than under the PLRA (or the FTCA, or the EAJA, or the American Rule) shows that § 1988 is itself irrational. These are simply different legislative solutions to an enduring problem; in a democracy, each of these options is open to the people’s representatives.
C
Even if § 1988 is the right benchmark for the question “compared to what?”, the rational-basis standard permits Congress to 'distinguish prisoners from free persons when deciding how much the defendant must subsidize a prevailing plaintiff. The PLRA draws many distinctions between prisoners and other persons covered by civil rights statutes. We held in Zehner that Congress rationally could distinguish prisoners from free persons for the purpose of suits seeking compensation for mental distress. We held in Lewis that Congress rationally applied the three-strikes rule to prisoners but not free persons. See also Lucien v. DeTella, 141 F.3d 773 (7th Cir.1998) (PLRA’s fee-collection and prepayment system is constitutional). The Supreme Court has twice interpreted and enforced the PLRA’s rule, 42 U.S.C. § 1997e(a), thát prisoners (and only prisoners) must exhaust administrative remedies before filing suit under 42 U.S.C. § 1983. Porter v. Nussle, 534 U.S. 516, 122 S.Ct. 983, 152 L.Ed.2d 12 (2002); Booth v. Churner, 532 U.S. 731, 121 S.Ct. 1819, 149 L.Ed.2d 958 (2001). Neither decision hints that there is anything problematic about treating prisoners differently. After we held unconstitutional the portion of the PLRA making equitable relief harder to get (and harder to keep) in prison-reform suits than in litigation about other institutions, see French v. Duckworth, 178 F.3d 437 (7th Cir.1999), the Supreme Court reversed and held that the PLRA must be applied as written. Miller v. French, 530 U.S. 327, 120 S.Ct. 2246, 147 L.Ed.2d 326 (2000). None of the Justices suggested that there is any constitutional problem in distinguishing prisoners’ suits from free persons’ suits. And in Martin v. Hadix, 527 U.S. 343, 119 S.Ct. 1998, 144 L.Ed.2d 347 (1999), the Supreme Court considered one issue about the scope of § 1997e(d) itself, again without any of its members suggesting that the statute has a constitutional flaw.
*592Congress could conclude that prisoners differ from free persons in ways relevant to litigation. A rational legislature would be entitled to believe the following:
• Prisoners have time on their hands. Unlike free persons, who must skip work or shun family to visit a law library and draft legal documents, prisoners have ample leisure. Persons with low opportunity cost of time substitute their own efforts for purchased commodities, such as legal services, and demand more of those things (such as litigation) that can be had for the investment of time alone. As a result, prisoners file many more federal suits per person than do free persons. (Prisoners, who account for less than 1% of the population, file more than 20% of all civil actions in the federal courts. See Administrative Office of the U.S. Courts, Judicial Business of the United States Courts 2002 Table C-2. In the year preceding the PLRA’s enactment, prisoners filed federal suits about 35 times as frequently as noninmates. See Margo Schlanger, Inmate Litigation, 116 Harv. L.Rev. 1555,1575 (2003).)
• Prisoners receive paper and postage; they have access to legal materials, see Bounds v. Smith, 430 U.S. 817, 97 S.Ct. 1491, 52 L.Ed.2d 72 (1977); inmate writ-writers provide assistance that, for free persons, would be deemed the unauthorized practice of law. These materials and services make it easier for prisoners to litigate.
• Many prisoners have a burning desire to turn the tables on the guards and other prison personnel, discomfiting if not hurting or humiliating them. This non-economic incentive to litigate produces additional suits, which as grudge matches are particularly hard to resolve.
• For some prisoners, litigation is recreation. Although most free persons shun litigation, because they have many better ways to amuse themselves, prisoners may see a trip to court as a vacation.
• These and other circumstances, including defendants’ desire not to attract additional nuisance suits, make prisoners’ suits unusually hard to settle, so that they impose on the judicial system (and thus on other litigants) a burden disproportionate to their numbers and intrinsic difficulty. (Professor Schlanger found that 6% of prisoners’ civil suits are settled, compared with 28% of noninmates’ civil-rights litigation and 50% of ordinary tort litigation. 116 Harv. L.Rev. at 1598. She also determined that the costs federal courts and prisons incur in handling prisoners’ suits top $175 million annually, far exceeding prisoners’ damages recoveries. Id. at 1622-26.)
• Prisoners are less honest than free persons and thus more likely to tell tall tales of victimization. The convictions that put them in prison establish their proclivity to violate the law when they see a personal advantage in doing so. (This is a premise of Fed.R.Evid. 609, which permits use of prior convictions to impeach the veracity of testimony.) The pot o’ gold at the end of litigation is a lure that induces dishonest claims.
• At the same time, however, prisoners are less amenable to sanctions for making false claims. Many are destitute; none earns wages subject to garnishment. The threat of prosecution for perjury holds little terror for a person already in prison, as a perjury sentence would be deferred until the expiration of existing terms; for those serving life sentences without prospect *593of parole, no further penal sanction is possible.
It is not our part to determine which of these things is true; it is enough to say that a legislature could think them true without taking leave of its senses. They make it apt to ask why, if prisoners file so many suits even though they receive legal assistance only 4% of the time, they require (or deserve) a subsidy to provide more legal assistance at defendants’ expense. The 4% figure for prisoners who have counsel comes from the Bureau of Justice Statistics and is reported in Roger A. Hanson & Henry W.K. Daley, Challenging the Conditions of Prisons and Jails: A Report on Section 1983 Litigation 21-22 (1995). Some prisoners have counsel from the outset; others, including Johnson, benefit from the district court’s assistance in recruiting counsel. Even after the PLRA’s adoption, about 4.4% of prisoners’ suits are prosecuted with the benefit of counsel. See Schlanger, 116 Harv. L.Rev. at 1609. Many of these lawyers — who are not appointed or otherwise conscripted but serve voluntarily, see Mallard v. United States District Court, 490 U.S. 296, 109 S.Ct. 1814, 104 L.Ed.2d 318 (1989) — are willing to assist for slight compensation, but a legislature rationally may think that the supply of lawyers prepared to serve on judicial request is affected by the compensation available if the plaintiff prevails. See Stewart J. Schwab & Theodore Eisen-berg, Explaining Constitutional Tort Litigation: The Influence of the Attorney Fees Statute and the Government as Defendant, 73 Cornell L.Rev. 719 (1988).
As the district judge perceived matters, the respects in which prisoners differ from free persons affect only frivolous and small stakes litigation. Prisoners file a superabundance of frivolous suits, the judge allowed, but none of these ends in an award of attorneys’ fees under § 1988 and so none is affected by § 1997e(d). Prisoners also file a profusion of small-stakes claims over $10 losses, such as the hobby kit in Parratt v. Taylor, 451 U.S. 527, 101 S.Ct. 1908, 68 L.Ed.2d 420 (1981), that any free person would let slide by rather than pay the $150 filing fee and devote costly time to litigation. Professor Schlanger found that prisoners lose some 93% of all civil suits they file. About 6% are settled: these are the only numerically significant victories, as defendants win 9/10 of the 3% of cases that make it to trial. 116 Harv. L.Rev. at 1598. The median recovery when a prisoner prevails at trial is $1,000. Id. at 1602-03. But the district judge was sure that pestiferous suits could be deterred by limiting attorneys’ fees to what is “reasonable” in light of the ends in view (including deterrence of similar petty misconduct by guards), without any need to curtail awards of legal fees in more substantial litigation.
Let us start with the second of these propositions — that the reasonableness requirement in § 1988, coupled with decisions such as Farrar v. Hobby, supra, and, e.g., Cole v. Wodziak, 169 F.3d 486 (7th Cir.1999), and provisions in the PLRA making it harder for prisoners to litigate without paying the filing fee, suffice to discourage suits over trivial harms (or at least .to make prisoners no more likely to file such suits than are free persons). This assumes that Congress is entitled to use only one tool to achieve a given objective. Why would that be? Legislatures often, and legitimately, select multiple devices. A “reasonableness” requirement varies, like the length of the chancellor’s foot, from judge to judge. Some are more generous than others with defendants’ money, and the district judge’s substantial discretion, see Hensley v. Eckerhart, 461 U.S. 424, 437-38, 103 S.Ct. 1933, 76 L.Ed.2d 40 (1983), which implies deferential appellate review of fee awards, ensures inconsisten*594cy. Congress rationally could conclude— on the ground of equitable treatment alone — that a quantitative rule (such as “fees can’t exceed 150% of the judgment”) should supplement a qualitative one (such as “fees can’t exceed what is reasonable”).
As for the contention that § 1997e(d) does not affect frivolous suits (because none is eligible for fees under § 1988): a legislature rationally may conclude that, because prisoners litigate even frivolous claims to excess, they do not need any extra incentive to litigate meritorious claims.
The district court’s view that the PLRA does not affect prisoners’ decisions depends on two assumptions, neither of which it justified. The first (which the dissenting opinion likewise indulges) is that each suit is known from the outset to be either frivolous or meritorious. The second is that no prisoner makes rational calculations of gains and losses from suit. Yet Congress is not bound to use these assumptions.
Take the first: that a suit’s merit (or lack thereof) is known to court and counsel. Why is this necessarily true? Some suits entail legal uncertainty, others factual uncertainty — and a legislature could conclude that prisoners’ proclivity for deceit makes it hard for outsiders to tell which end is up. District judges have a hard time telling the two apart: prisoners lose half of all cases in which the judge deems the claim sufficiently meritorious to recruit counsel, and others are settled for sums that may reflect the defendants’ litigation costs rather than a high probability that the prisoner would prevail. See Schwab & Eisenberg, supra, 73 Cornell L.Rev. at 773-74. If district judges cannot reliably separate strong from weak claims, neither can counsel; and if the suit’s strength is hard to determine at the outset, then a law reducing fees in successful suits also affects the filing of weak claims. An example makes the point.
Suppose a prisoner’s complaint alleges that a guard set upon and beat him without provocation, and the prisoner files an affidavit to that effect. The guard responds that the supposed battery never occurred and adds that the prisoner’s injury was inflicted by his cellmate in a gambling dispute. Is this suit frivolous or meritorious? If the prisoner is lying, it is frivolous; if the guard is lying, the suit is meritorious. The judge cannot be sure who is telling the truth and is not authorized to resolve the case short of trial. Nor could a lawyer representing the prisoner be sure who is honest; deceitful prisoners have no reason to be candid with counsel and may be able to recruit fellow inmates as witnesses (for other prisoners may want to make life miserable for the guard, and like the plaintiff are hard to sanction for perjury).
Suppose that 9 out of 10 prisoners making such claims are lying, while 1 of 10 guards is lying. (This assumption tracks Professor Schlanger’s finding that prisoners lose 90% of suits that go to trial.) Suppose further that, if the jury finds for the plaintiff, it will award $50,000 in damages and, but for the PLRA, the judge would award $150,000 in legal fees (if a lawyer takes the case). Finally, suppose that with the assistance of counsel the plaintiff would prevail before a jury 20% of the time, while unaided the plaintiff will win only 10% of these cases. Before the PLRA’s enactment, this case on filing is worth $10,000 (= $50,000 x 0.2) to the prisoner, who will win 20% of the time with legal assistance (a form of regression to the mean, since most errors will come from the 90% of claims in which the prisoner is deceitful); the expected legal award is worth $30,000 to counsel ($150,-000 x 0.2), which induces counsel to assist; *595and the expected judgment from the guard’s perspective is $40,000 (($50,000 + $150,000) x 0.2). With the PLRA in force, the expected legal fee falls to $15,000 (assuming that the 150% cap is the effective limit), and counsel may be unwilling to take the case. Then the prisoner’s expected recovery falls to $5,000 ($50,000 x 0.1), and the guard’s anticipated judgment is the same (for there will be no legal fees). If it turns out that $15,000 is enough to recruit counsel, then the prisoner’s expected recovery rises to $10,000 (just as before the PLRA), but the guard’s outlay falls to $20,000 — which, considering the 90% chance that the prisoner is lying, still is too high but is more appropriate than the pre-PLRA anticipated loss of $40,000.
By reducing the prisoner’s expected recovery from $10,000 to $5,000 in this class of cases, the statute discourages suits that are frivolous when viewed objectively — or so Congress rationally could conclude. The decision is made ex ante, and at this stage the PLRA has a direct effect on counsel’s willingness to represent those prisoners making frivolous claims. See Boivin, 225 F.3d at 45. By reducing the defendant’s anticipated cost of litigation from $40,000 to $20,000 for this set of claims even when counsel takes the prisoner’s case, the PLRA brings the outcome closer to the real loss entailed- — or so, again, Congress rationally could conclude.
We come to the district court’s second assumption: that prisoners just don’t think this way. The judge was certain that prisoners who are deluded, or determined to lie in court, will do so no matter what, and that prisoners determined to file frivolous suits are incorrigible. The PLRA then will fail; prisoners won’t desist from filing in the class of suits we have described (or any other). Yet a legislature could believe that, even if the law has no effect on plaintiffs, its effect from defendants’ perspective (reducing the fees that must be paid on top of damages) is a public benefit. Moreover, a legislature rationally could conclude that some prisoners will respond to the adjustments. Marginal effects may differ from average effects. Even if 80% of prisoners are insensible to changes in the litigation process, different behavior by the other 20% would be welcome. A reduction in prisoners’ suits by 20% would reduce the total caseload of the federal courts by about 5%. The actual reduction between 1995, the year before the PLRA’s enactment, and 2001 is even greater. In 1995 prisoners filed 39,008 federal civil-rights suits, or 24.6 suits per 1,000 inmates. In 2001 they filed 22,206 such suits, at a rate of 11.4 per 1,000 inmates. See Schlanger, 116 Harv. L.Rev. at 1583. While the number of prisoners rose, the number of suits dropped dramatically. (Note that this is a decline before application of two other changes made by the PLRA: screening under 28 U.S.C. § 1915A and dismissals under § 1997e(a) for failure to exhaust administrative remedies.)
That the PLRA has led prisoners to cut by half their propensity to sue shows that they do respond to incentives. Litigation is never free, even to a prisoner, if only because it diverts time from exercise and watching television. Congress rationally could believe that the size of the expected recovery influences some prisoners’ decisions about how to allocate their time. Although the amount of the effect attributable to § 1997e(d) is hard to calculate, its direction is knowable. A rational legislature could conclude that a small reduction in weak, trivial, or bogus suits is worth achieving even at some potential cost to prisoners’ ability to prevail in the less common meritorious suit. See, e.g., National Paint & Coatings Ass’n v. Chicago, 45 F.3d 1124 (7th Cir.1995) (discussing how the difference between marginal and infra-*596marginal behavior affects analysis under the rational-basis standard).
D
Even if all of this is wrong, Congress rationally could suspect that fee awards under § 1988 are excessive as a rule, and that prisoners’ suits are an appropriate place to explore the results of a cutback. The ability to take one step at a time, to alter the rules for one subset (to see what happens) without changing the rules for everyone, is one of the most important legislative powers protected by the rational-basis standard. And it would be entirely rational to conclude that fees under § 1988 are in need of recalibration. Although a plurality of the Court in Riverside rejected the contention that attorneys’ fees must be proportional to damages, see 477 U.S. at 573-81, 106 S.Ct. 2686 (opinion of Brennan, J., joined by Marshall, Black-mun & Stevens, JJ.), four other Justices concluded that § 1988 as it stands should be understood to limit awards against defendants to the amount that a solvent plaintiff would be willing to pay his own lawyer, a sum generally less than the prospective recovery. See 477 U.S. at 588-96, 106 S.Ct. 2686 (Rehnquist, J., dissenting, joined by Burger, C.J., and White & O’Connor, JJ.). Justice Powell, who wrote the dispositive opinion, concluded that the award — about $245,000 in fees for legal work that led to $33,350 in damages — was unjust to the defendants (why should these defendants be made to pay for legal services whose principal effect is to improve deterrence with respect to other would-be violators?) but could not be set aside given deferential appellate review and statements in the legislative history implying that some Members of Congress thought a hard cap unwarranted.
Members of Congress who agreed with Justice Powell about fairness to defendants (or for that matter with Justice Rehnquist about the appropriate rule for all civil-rights cases) rationally could decide to alter the approach of § 1988 (as Justice Brennan interpreted it) for a subset of all cases to which it applies. For reasons we have covered, prisoners are a sensible subset with which to begin, and § 1997e(d) is a modest step. The district court’s award requires Daley to pay more than 12 times the jury’s estimate of actual damages; even with the PLRA cap in full effect, Daley must pay about 8 times actual damages, which is still a hefty multiplier. (A larger multiplier than the double-damages approach that, according to State Farm Mutual Automobile Insurance Co. v. Campbell, — U.S. -, —, 123 S.Ct. 1513, 1526, 155 L.Ed.2d 585 (2003), may be the constitutional limit for awards based on emotional distress.) Experience under the PLRA may lead to more general modifications — or, if too many good claims no longer can be vindicated, this experience may lead Congress to lift or adjust the cap for prisoners. The Constitution permits this sort of tinkering; it does not put the legislature to an all-or-none choice at the outset, before data can be collected.
The step-at-a-time corollary to the rational-basis standard tolerates the sort of inconsistency to which all systems of majority voting are prone. See Kenneth J. Arrow, Social Choice and Individual Values (2d ed.1963). Suppose one-third of all legislators believe with Justice Brennan that no limit should be applied to awards of attorneys’ fees in any constitutional suit; that one-third believe with Justice Rehnquist that awards in all constitutional suits should be limited to the amount (almost always less than the damages) that a solvent private litigant would be willing to pay for legal assistance; and that one-third believe that the plaintiffs with the lowest opportunity costs of time (which is *597to say, prisoners) should be required to substitute their own time for lawyers’ time to some extent. Two-thirds of these legislators would agree with the proposition: “Prisoners and free persons should be treated alike with respect to recovering attorneys’ fees from defendants.” But though they would agree with an equal-treatment rule, they would not agree on the content of that rule. As a result, a proposal to reduce fees in suits by prisoners would carry by a two-thirds majority, though a proposal to reduce fees across the board would fail. If that form of inconsistency can be called “irrational” and condemned as unconstitutional, then it is democracy itself that the Constitution forbids — because, as Arrow proved, inconsistency is an unavoidable consequence of decision by majority rule. The rational-basis approach tolerates this sort of legislative inconsistency by asking, not what legislators (or judges) actually believe, but whether it is possible for a sensible person to believe that the law does -something useful. People could, and many do, believe that § 1997e(d) does something useful.
E
We have not yet mentioned two mainstays of Johnson’s argument and the district court’s holding: Rinaldi v. Yeager, 384 U.S. 305, 86 S.Ct. 1497, 16 L.Ed.2d 577 (1966), and Lindsey v. Normet, 405 U.S. 56, 92 S.Ct. 862, 31 L.Ed.2d 36 (1972). Neither case supports the conclusion that § 1997e(d) is invalid.
Rinaldi dealt with a law requiring prisoners — but not criminal defendants whose sentences had been suspended, or those fined but not sentenced to prison — to repay the cost of any transcripts prepared for use on direct appeal. The state apparently failed to offer any defense of this distinction (at least, the Justices did not discuss any proffered justification), and the majority held it to be irrational because inexplicable and unreasoned. Wisconsin and the United States have defended the line drawn by § 1997e(d); as Part II.C demonstrates it is not an “unreasoned distinction” (384 U.S. at 310, 86 S.Ct. 1497); and that is enough to show that Rinaldi does not undermine the PLRA. But there is more.
Rinaldi is among a series of decisions in the 1950s and 1960s that clear away what the Justices deemed to be obstacles to appeals by indigent criminal suspects. See, e.g., Griffin v. Illinois, 351 U.S. 12, 76 S.Ct. 585, 100 L.Ed. 891 (1956); Douglas v. California, 372 U.S. 353, 83 S.Ct. 814, 9 L.Ed.2d 811 (1963). These and other similar decisions were revisited in Ross v. Moffitt, 417 U.S, 600, 94 S.Ct. 2437, 41 L.Ed.2d 341 (1974), which concluded that Griffin and its successors rest more comfortably on the due process clause — because they determine what process states must afford to those accused of crime— than on the equal protection clause. Ever since Ross it has been understood that cases such as Rinaldi protect only the rights of defendants. This means principally criminal defendants, though occasionally the principle has some application to defendants in civil litigation of exceptional import, such as those that may terminate parental rights concerning their children. See M.L.B. v. S.L.J., 519 U.S. 102, 117 S.Ct. 555, 136 L.Ed.2d 473 (1996). Neither Rinaldi nor any of the other cases in its sequence ever has been used to hold that the defendant must subsidize plaintiffs’ civil litigation. That is why we held in Lewis that Congress.may require civil plaintiffs to pay their own way.
As for Lindsey: the Court held that, consistent with the equal protection clause, a state may distinguish eviction suits from other litigation about property and may require tenants to post bonds for the rent *598accruing pending decision. 405 U.S. at 69-74, 92 S.Ct. 862. Indigent litigants unable to post bonds for accrued rent may lose the litigation summarily, the Court held, as far as the Constitution is concerned. These aspects of Lindsey strongly support the kind of distinctions drawn in the PLRA. The Court added in Lindsey that requiring the tenant to post a bond (to be forfeited in the event of affirmance) for double all accrued and impending rentals is irrational; double-or-nothing is a game of chance, not a mode of civilized litigation. Id. at 74-79, 92 S.Ct. 862. What Lindsey establishes is that forcing one party (in Lindsey, the tenant-defendant) to pay extra (double rent) to the adversary as a condition of receiving a decision on the merits violates the Constitution. If that holding has any resonance here, it casts a shadow on § 1988 itself, for, as a condition of receiving a decision on the merits, Daley had to wager not only his own legal expenses but also Johnson’s. We do not think that fee-shifting laws are problematic; they have been sustained many times, see In re Mann, 311 F.3d 788, 790 (7th Cir.2002) (collecting cases); but nothing in Lindsey implies that fee-shifting ever is required.
Unlike the double-or-nothing statute in Lindsey, the PLRA does not compel the defendant to pay extra (compared with § 1988) as a condition of receiving a decision on the merits. The district court turned Lindsey on its head, treating it as a requirement that one side subsidize the other by enough to ensure that litigation occurs. Nothing of the kind can be found in Lindsey (which, recall, held that a single-bond requirement is valid even for impecunious tenants) or in any other decision of which we are aware. If the American Rule is constitutional, which it is, there can be no doubt about the validity of the PLRA, which does not impose a “litigation tax” on prisoners but simply reduces the extent to which defendants must underwrite prisoners’ suits.
The judgment of the district court is reversed, and the case is remanded for an award of attorneys’ fees that complies with § 1997e(d).
Section 3006A, the Criminal Justice Act, authorizes payment at $60 per hour for work in court and $40 per hour for other work. It permits the Judicial Conference to raise the cap to the greater of $75 per hour or an amount calculated with respect to cost-of-living increases awarded to federal employees. In September 2000 the Judicial Conference authorized use of the $75 rate for all work nationwide and determined that the inflation-adjusted rate would be $113 per hour, but that appropriated funds did not permit compensation at more than $75. The 2002 and 2003 appropriations acts for the judiciary provide funds sufficient to pay appointed counsel $90 per hour, and it is the policy of the Judicial Conference that all work performed after May 1, 2002, should be compensated at that level. Johnson’s case came to trial in the district court before this increase, so the CJA maximum at the time was $75, and the PLRA maximum therefore was $112.50 per hour. It is possible that some of the legal work performed on Johnson’s behalf is affected by the earlier $60 and $40 máxi-mums, which remained in effect in scattered districts. To facilitate exposition, we use throughout the opinion the $90 CJA funded rate, which implies a maximum of $135 per hour under the PLRA. By employing this figure, we do not imply any view on the question whether it is the right one, or whether instead $169.50 (150% of $113) is today's cap. Compare Webb v. Ada County, 285 F.3d 829, 838-39 (9th Cir.2002), with Hernandez v. Kalinowski, 146 F.3d 196, 201 (3d Cir.1998).