with whom COFFEY and MANION, Circuit Judges, join, dissenting.
“[T]he decisions of the Administrator on any question of law or fact under any law administered by the Veterans’ Administration providing benefits for veterans ... shall be final and conclusive and no other official or any court of the United States shall have power or jurisdiction to review any such decision by an action in the nature of mandamus or otherwise.” 38 U.S. C. § 211(a). This is a bold assertion of the sovereign immunity of the United States. Decisions of the Administrator are “final and conclusive”. The “decision”, not the legal issues that might influence it, is off limits. No wonder the Supreme Court has described this statute as the model of language that Congress uses when it wants to “bar judicial review altogether”. Lindahl v. OPM, 470 U.S. 768, 779-80 & n. 13, 105 S.Ct. 1620, 1627 & n. 13, 84 L.Ed.2d 674 (1985).
I
Our case lies at the core of § 211(a). Marozsan does not contend that any statute governing the award of benefits is unconstitutional or that any of the VA’s regulations is illegal. Contrast Johnson v. Robison, 415 U.S. 361, 366-74, 94 S.Ct. 1160, 1165-69, 39 L.Ed.2d 389 (1974) (constitutional challenge to statutory eligibility criteria); Traynor v. Turnage, — U.S. -, *1486108 S.Ct. 1372, 1378-80, 99 L.Ed.2d 618 (1988) (contention that a regulation of general applicability is inconsistent with a subsequent statute). Marozsan wants judicial review of the decision to rate him 20% rather than 100% disabled. His complaint asks the court to “restore the Plaintiff to his proper disability, retroactive to the date of disability” and to give him $5 million as damages. Marozsan does not seek prospective relief for a class; he wants money for himself. Marozsan’s current theory about why the court should review his case is that the procedures the VA used violated the Due Process Clause of the fifth amendment. Section 211(a) does not distinguish among reasons for believing the VA’s decision to be erroneous; it protects the “decision” from review. Other courts say that § 211(a) bars review of individual veterans’ cases even when the veteran contends that the VA violated the Constitution in the course of resolving his claim. E.g., Higgins v. Kelley, 824 F.2d 690 (8th Cir.1987); Pappanikoloaou v. Administrator of Veterans Administration, 762 F.2d 8 (2d Cir.1985); Rosen v. Walters, 719 F.2d 1422, 1423 (9th Cir.1983); Anderson v. Veterans Administration, 559 F.2d 935 (5th Cir.1977); Ross v. United States, 462 F.2d 618 (9th Cir.1972); de Rodulfa v. United States, 461 F.2d 1240, 1257-58 (D.C.Cir.1972); Milliken v. Gleason, 332 F.2d 122 (1st Cir.1964). Devine v. Cleland, 616 F.2d 1080, 1083-85 (9th Cir.1980), disagrees, but Devine conflicts with Ninth Circuit cases before (Ross) and since (Rosen). We stand alone among the circuits in declining to give § 211(a) its ordinary meaning.1
The history of § 211(a) shows that it has the scope its language portends. Versions of this statute date back to 1933. Johnson, 415 U.S. at 371-73, 94 S.Ct. at 1167-68, traces the history. The earliest version quickly was invoked in Lynch v. United States, 292 U.S. 571, 587, 54 S.Ct. 840, 847, 78 L.Ed. 1434 (1934), in which Justice Bran-déis described the statute as a rule that
concerns ... grants to veterans and their dependents — pensions, compensation allowances and special privileges, all of which are gratuities. The purpose of the section appears to have been to remove the possibility of judicial relief in that class of cases even under the special circumstances suggested in Crouch v. United States, 266 U.S. 180 [45 S.Ct. 71, 69 L.Ed. 233]; Silberschein v. United States, 266 U.S. 221 [45 S.Ct. 69, 69 L.Ed. 256]; United States v. Williams, 278 U.S. 255 [49 S.Ct. 97, 73 L.Ed. 314]; Smith v. United States, 57 F.(2d) 998 [(4th Cir.1932) ].
The “special circumstances” suggested in Crouch, Silberschein, and Williams—three cases that held the courts without jurisdiction to review administrative decisions concerning benefits — were that review might be possible when the decision was wholly arbitrary or without colorable support; Lynch seems to say that even in these “special circumstances” there could be no review of veterans’ benefit cases.
A modification in the statute some years after Lynch led the D.C. Circuit to assert a limited power of review. Congress was unwilling to accept even a small erosion of administrative finality, so in 1970 it passed an amendment giving § 211(a) its current language. The House Report, H.R.Rep. No. 91-1166, 91st Cong., 2d Sess. 10 (1970), U.S.Code Cong. & Admin.News 1970, p. 3723, identifies the three decisions to be rejected by the amendment. One of these, Thompson v. Gleason, 317 F.2d 901 (D.C.Cir.1962), was a constitutional challenge to a decision to deny benefits. Congress certainly wanted to cut off any possibility of judicial review of individual veteran’s benefits decisions, and it was aware — to the extent institutions can be “aware” — that § 211(a) would block constitutional challenges. It is still so aware. Since 1970 many Members of Congress have favored judicial review of the VA’s decisions. Bills to provide for review have cleared committee, e.g., S.Rep. No. 97-466, 97th Cong., 2d Sess. 37-49 (1982), but have not become law. One bill that has attracted substantial support in the 100th Congress would create review of constitutional and strictly *1487“legal” questions, while preserving administrative finality on factual ones. No one participating in this ongoing debate believes that § 211(a) already provides for judicial review of constitutional claims in individual eases.
Congress enacted § 211(a) for two reasons:
(1) to insure that veterans’ benefits claims will not burden the courts and the Veterans’ Administration with expensive and time-consuming litigation, and (2) to insure that the technical and complex determinations and applications of Veterans’ Administration policy connected with veterans’ benefits decisions will be adequately and uniformly made.
Johnson, 415 U.S. at 370, 94 S.Ct. at 1167 (footnote omitted). See also Traynor, 108 S.Ct. at 1379 (adhering to this assessment);
Rose v. Rose, — U.S. -, 107 S.Ct. 2029, 2035, 95 L.Ed.2d 599 (1987) (identifying the same purposes). The statute is part of a system of benefits decision-making designed to operate outside the adversarial process. Walters v. National Ass’n of Radiation Survivors, 473 U.S. 305, 105 S.Ct. 3180, 87 L.Ed.2d 220 (1985). It is hard to have a non-adversarial administrative procedure if the parties’ next stop is a court. Everyone will want to build the record for review.
Johnson and Traynor concluded that litigation about the constitutionality of the statutory criteria used to award benefits, and the legality of the VA’s regulations, would not prevent § 211(a) from achieving any of its purposes. The Court drew a line between reviewing cases and reviewing rules. The line the Court adopted in Johnson and Traynor accords not only with the functions of § 211(a) but also with its language. The statute forbids review of a “decision of the Administrator”, and it is natural to read this as a decision in a case. Hiring employees, leasing office space, and similar choices come to mind as “decisions” outside the scope of § 211(a), if for no other reason than that they are not decisions concerning “benefits for veterans” within the scope of the statute.2
The line between reviewing decisions in cases and reviewing rules is familiar in governmental law. Substantive criteria, whether in a statute or in a regulation, may be reviewed in litigation seeking prospective relief. Such suits are authorized by 5 U.S.C. § 702 (waiving sovereign immunity in injunctive cases), and are common against the states, despite the eleventh amendment. Edelman v. Jordan, 415 U.S. 651, 94 S.Ct. 1347, 39 L.Ed.2d 662 (1974). Suits seeking money for a particular claimant are something else again, not permissible against the states because of the eleventh amendment, or against the federal government because of sovereign immunity. Because “[n]o money shall be drawn from the Treasury, but in Consequence of Appropriations made by Law” (Art. I, § 9, cl. 7), only an unambiguous waiver permits monetary awards. United States v. King, 395 U.S. 1, 4, 89 S.Ct. 1501, 1502, 23 L.Ed.2d 52 (1969); United States v. Testan, 424 U.S. 392, 399-403, 96 S.Ct. 948, 953-55, 47 L.Ed.2d 114 (1976); Army & Air Force Exchange Service v. Sheehan, 456 U.S. 728, 738-41, 102 S.Ct. 2118, 2124-25, 72 L.Ed.2d 520 (1982). Section 211(a) is an unambiguous non-waiver. The line drawn in Johnson and Traynor respects the nature of § 211(a) as an assertion of sovereign immunity, limiting judicial review to the kind of prospective relief that courts may award against the government.
II
Perhaps it seems that the majority and I are discussing different cases. The court insists that Marozsan has asked for class-wide, prospective relief, not for a review of his disability rating. It affirms the district court’s conclusion that Marozsan may not attack the disposition of his particular case. Majority op. at 1471 n. 3. If Marozsan were *1488waging a campaign along the lines of Johnson and Traynor, § 211(a) would have little to do with the case. But I fear that my colleagues are as inventive with Marozsan’s complaint as they are with § 211(a).
The complaint starts by describing the handling of Marozsan’s case. It grumbles, for example, about “the VA’s Foreign asi-atic general practitioner (M.D.), who could not even grammically write a concise medical report with out assistance from the Plaintiff”. (Here, and throughout, I omit [sic]s.) The complaint is concerned only about Marozsan’s treatment. It ends by asking for a review of the disability rating, a restoration “to his proper disability, retroactive to the date of the disability”, and $5 million. Not a peep about class-wide practices, and no request for injunctive relief. (The amended complaint, filed by counsel, deleted the request for an injunction that had been in the original complaint, filed by Marozsan personally.) When the defendants moved to dismiss under § 211(a), Marozsan did not reply that he was challenging practices of general applicability. He instead said that § 211(a) is unconstitutional, a violation of the equal protection component of the Due Process Clause of the fifth amendment. The district court rejected that contention.
Marozsan’s briefs on the original argument and the reargument say two things: that § 211(a) is unconstitutional, which if true entitles him to relief on his particular claim, and that the court en banc should follow Winslow v. Walters, 815 F.2d 1114, 1116-18 (7th Cir.1987), and Mathes v. Hornbarger, 821 F.2d 439, 441-42 (7th Cir.1987). These cases hold that despite § 211(a) a veteran may obtain review in his individual case, provided his legal theory rests on the Constitution. Again he did not ask for prospective relief (except to the extent' of holding § 211(a) unconstitutional), and at oral argument he disclaimed such a request.
His opening brief explains why he believes § 211(a) violates the equal protection component of the Due Process Clause. It lists claims cut off by § 211(a), and it contends that these are sufficiently serious that the statute must be unconstitutional. Here is his recitation:
(1) There is no requirement for the VA to confront Veteran’s evidence and as a result, decisions may be rendered without addressing any of the Veteran’s contentions, and when the Veteran continues to protest this he is accused of repetitiveness.
(2) Decisions are rendered at stages of the VA process without affording the Veteran the opportunity for a hearing. [Citation omitted.]
(3) A conflict of interest in that the same rating personnel who originally denied the claim may be hearing the appeal. A further conflict exists between the various service organizations and the VA with the crux being the use of VA facilities by service organizations, with space provided free....
(4) Rating personnel making the decision are often not present for the hearing.
(5) Informed discussion with VA personnel is denied, leaving the Veteran, who does not understand the rules, at a loss to pursue the case.
(6) Expert opinions in behalf of Veteran’s are disregarded in favor of the opinion of the general practitioners who are without medical specialization, and no attempt is made to reconcile differences of medical opinion.
(7) The VA acts as the investigative arm, judge, jury and attorney for both sides, using their subpoena power only in their behalf and not for the Veteran.
(8) Although the Veteran is denied representation by an attorney, attorneys for the VA dominate every stage.
(9) The imposition of a quota system as demonstrated by the percentages of successful appeals which varies less than 1% over the years.
It is impossible to figure out which of these Marozsan thinks happened to him, which happens frequently to others and is unconstitutional, and which illustrates why (in his view) § 211(a) is undesirable.
None of these, however, is stated as a ground of injunctive relief against the VA. *1489Each is a reason, according to Marozsan, why § 211(a) should be held unconstitutional, so that veterans may litigate case by case. None of his assertions, moreover, finds support in the record. The defendants moved to dismiss, and the district court converted the motion to one for summary judgment. To avoid summary judgment Marozsan had to put in evidence, or at least file an affidavit under Fed.R.Civ.P. 56(f) to explain why evidence was unavailable. He did neither. My search for support in the record for any of his assertions was fruitless. They are, so far as this case is concerned, nothing but hot air.
We have to take Marozsan’s case the way he and the record frame it: as a challenge to the denial of his claim for benefits, expanded to an attack on the constitutionality of § 211(a). There is, nothing else to consider — not only because nothing else was raised, but also because any rumblings in his appellate brief are unsupported by the record and so cannot stave off summary judgment. My colleagues say that § 211(a) is proof against an equal protection attack (maj. op. at 1471 n. 3). That ought to end things; it is not our business to resurrect a case so that the court may address contentions that the litigants never advanced.
At all events, the majority is trying to have things both ways. One would have thought that the conclusion (maj. op. at 1471 n. 3) barring review of Marozsan’s own claim would require us to overrule Winslow and Mathes; instead the court reaffirms them (maj. op. at 1472-73, 1479). Those cases hold that a veteran may obtain review of his individual case, provided he contends that the VA committed constitutional error in the course of making a decision. The court writes that those cases “reflect the proper construction and application of § 211(a)”. If that is so, the hoo-ha about Mar-ozsan's claim being class-wide and prospective throws sand in the reader’s eyes. What the court holds today is that review may be had, case by case, despite § 211(a) — and also may be had on a prospective basis in Maroz-san’s ease.
Under Winslow and Mathes every case may be reviewed to see whether the VA violated the Constitution in the course of reaching a decision. To say that a “veteran may obtain review, not of his individual claim determination, but of unconstitutional methods employed by the V.A. in arriving at that benefits decision” (maj. op. at 1473 n. 10) is to say that the veteran may obtain review in his individual case. Review of “methods employed” by an agency to make a decision is review of the individual case. If not, it is an advisory opinion. Such review is what the Administrative Procedure Act provides, what the federal courts provide when resolving collateral attacks on criminal convictions. Because any principle of administrative law may be recast in the language of “due process”, every benefits decision is reviewable, if only to determine whether the procedural error should be called “constitutional”. And a court may not evaluate claims of procedural error in the abstract. It must get down to the nitty-gritty, looking at the whole proceedings, both to determine whether the gaffe is an “error” and to evaluate whether the error is harmless. In other words, the kind of review the majority approves will in Johnson’s words “burden the courts and the Veterans’ Administration with expensive and time-consuming litigation” and require many judges to examine the “technical and complex determinations” of veterans’ benefits law that are supposed to be “uniformly made” by a single Administrator. Hundreds of district judges cannot match one Administrator’s (or Board’s) decisions for “uniformity”. Homemakers North Shore, Inc. v. Bowen, 832 F.2d 408, 412 (7th Cir.1987).
Dissenters often cry “Wolf!”, but this prediction is based on the federal courts’ daily experience. I speak of collateral review of criminal convictions. Such review generally is limited to claims of constitutional error, just as review of veterans’ benefit decisions is restricted to constitutional questions under the majority’s approach. Federal courts engage in collateral review of a significant portion of all state and federal criminal convictions, sometimes more than once per ease. The *1490ability to state any procedural question in constitutional terms gets these cases over the transom, and the court must review the record to satisfy itself that no error properly denominated “constitutional” occurred. This task frequently requires extensive study. Jackson v. Virginia, 443 U.S. 307, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979), also requires the court to review the sufficiency of the evidence. Although review under Jackson depends on the reasonable-doubt standard inapplicable outside criminal law, Jones v. Thieret, 846 F.2d 457, 460-62 (7th Cir.1988), the Due Process Clause often demands at least a smidgeon of evidence, once the government sets up the sort of substantive rules that create a liberty or property interest. Superintendent of Walpole v. Hill, 472 U.S. 445, 453-55, 105 S.Ct. 2768, 2772-73, 86 L.Ed.2d 356 (1985); cf. University of Michigan v. Ewing, 474 U.S. 214, 106 S.Ct. 507, 88 L.Ed.2d 523 (1985). Every evidentiary question has a constitutional dimension under Hill, and although most such claims are dead losers, each must be reviewed. Yet we know from Lynch that § 211(a) is designed to put federal courts out of the business of reviewing the sufficiency of evidence behind veterans’ benefits decisions.
Constitutional “questions” are everywhere — it is easy to pose questions and occasionally is possible to characterize a mistaken view of the law or the facts as “arbitrary”, an “abuse of power”, hence unconstitutional. Compare Cole v. Young, 817 F.2d 412, 423-26 (7th Cir.1987), and Argento v. Village of Melrose Park, 838 F.2d 1483, 1491-92 (7th Cir.1988), with Jones v. Thieret and Archie v. City of Racine, 847 F.2d 1211 (7th Cir.1988) (en banc). Only an unimaginative lawyer would be unable to find a constitutional “question” in most proceedings before the VA. If veterans take their cases to court at the same rate social security claimants do, today’s decision will produce some 13,-000 civil actions and 1,000 appeals per year.3 Judicial review to sift constitutional wheat from statutory chaff may be even more time consuming than review under the Administrative Procedure Act, for a court will need to grapple with a series of anterior questions that are missing in “straight” administrative review. We need only think of the forfeiture and exhaustion rules in criminal cases, e.g., Wainwright v. Sykes, 433 U.S. 72, 97 S.Ct. 2497, 53 L.Ed.2d 594 (1977), and the nagging questions of characterization, e.g., Donnelly v. DeCristoforo, 416 U.S. 637, 94 S.Ct. 1868, 40 L.Ed.2d 431 (1974), to see what lies ahead. The process of line-drawing may consume more energy than the resolution of the substantive claims.
Ill
What is the justification for engaging in review in the teeth of the language, history, and functions of § 211(a)? The best line of argument is that § 211(a) is limited to statutory and evidentiary challenges because it addresses only decisions “on any question of law or fact under any law administered by the Veterans’ Administration providing benefits” (emphasis added). A constitutional question does not arise “under” a law providing for benefits. So, the argument goes, § 211(a) does not block review of constitutional questions.
The approach has some force; it is derived from the Court’s analysis in Johnson. But it is unsatisfactory. The statute bars review of “decisions”, not of the legal issues the VA must resolve to reach ded-*1491sions. One may intelligibly say that a suit seeking an injunction against a statute or regulation “arises under” the Constitution (Johnson) or Rehabilitation Act (Traynor) because in those eases the Constitution or statute was the foundation of the claim for relief. The plaintiff filed an independent suit; the claim did not depend on a veterans’ law. The only thing to be construed in order to adjudicate the case was the non-veterans’ rule.
That a constitutional issue comes up in the course of adjudicating a claim for benefits does not mean that the claim itself “arises under” the Constitution. Someone who wants benefits necessarily makes a claim under the veterans’ laws, and § 211(a) therefore applies. Under the well-pleaded complaint rule, the claim arises under the rule of law that gives it vitality. Gully v. First National Bank, 299 U.S. 109, 57 S.Ct. 96, 81 L.Ed. 70 (1936); Franchise Tax Board v. Construction Laborers Vacation Trust, 463 U.S. 1, 27-28, 103 S.Ct. 2841, 2855-56, 77 L.Ed.2d 420 (1983). Without a claim under the veterans’ laws, Marozsan has nothing, no matter what constitutional arguments he may make; without a claim under the veterans’ laws Mar-ozsan does not even have standing. Cases concerning the federal question jurisdiction show that a claim does not “arise under” federal law for purposes of 28 U.S.C. § 1331 just because a court needs to consider a federal issue to reach a decision. The arising-under jurisdiction depends on “what necessarily appears in the plaintiff’s statement of his own claim ..., unaided by anything alleged in anticipation or avoidance of defenses which it is thought the defendant may interpose.” Franchise Tax Board, 463 U.S. at 10, 103 S.Ct. at 2846; see also, e.g., Christianson v. Colt Industries Operating Corp., — U.S. -, 108 S.Ct. 2166, 100 L.Ed.2d 811 (1988). “The well-pleaded complaint rule ... focuses on claims, not theories”. Id. at -, 108 S.Ct. at 2176; see also Franchise Tax Board, 463 U.S. at 26 & n. 29, 103 S.Ct. at 2855 & n. 29. Marozsan’s claim is for benefits, even if his theory about why he gets them has something to do with the Constitution. We know what issues Marozsan raised in his complaint: he wanted review of the decision denying his claim for benefits. That he anticipated the VA’s invocation of § 211(a) by nodding lightly to constitutional arguments does not give him a constitutional claim for arising-under purposes.
The Court has not accepted Justice Holmes’s position that the claim invariably arises under “the law that creates the cause of action”, American Well Works Co. v. Layne & Bowler Co., 241 U.S. 257, 260, 36 S.Ct. 585, 586, 60 L.Ed. 987 (1916), but this is at least the starting point and in most cases the finishing point as well. E.g., Merrell Dow Pharmaceuticals Inc. v. Thompson, 478 U.S. 804, 808, 106 S.Ct. 3229, 3232, 92 L.Ed.2d 650 (1986). For example, a claim based on state law becomes a “federal” claim only when federal law so occupies the field that it supplies the only possible remedy, or when the federal right is so integral to the state claim that one cannot think of them separately. E.g., Caterpillar Inc. v. Williams, — U.S. -, 107 S.Ct. 2425, 2429-30, 96 L.Ed.2d 318 (1987). That a court inevitably will consider an issue of federal law to reach a final decision does not mean that the claim “arises under” that law. E.g., Luckett v. Delpark, Inc., 270 U.S. 496, 46 S.Ct. 397, 70 L.Ed. 703 (1926); cf. Christianson.
This way of defining what kind of claim “arises under” a given rule governs jurisdiction-withdrawing statutes with an “arising-under” , component, just as it governs jurisdiction-granting statutes. Heckler v. Ringer, 466 U.S. 602, 615-16, 104 S.Ct. 2013, 2021-22, 80 L.Ed.2d 622 (1984); Weinberger v. Salfi, 422 U.S. 749, 760-64, 95 S.Ct. 2457, 2464-66, 45 L.Ed.2d 522 (1975). Salfi dealt with a portion of the Social Security Act providing that claims arising under that statute could be reviewed in one way only. The plaintiff attempted to get review in a different way on the argument that his claim really arose under the Constitution, because he was making a constitutional challenge to the way the program was administered. The Court acknowledged that in a sense such a claim arises under the Constitution but concluded that it also arises under the Social *1492Security Act and held that it was therefore barred. The suit, though invoking the Constitution, arose under the statute because the statute was an essential ingredient of the claim for benefits. Franchise Tax Board and Christianson, decided since Salfi and Ringer, strengthen their holdings. The same principles should apply to § 211(a). The veteran’s claim for benefits arises under the veterans’ laws. This claim is what gets the administrative procedure going. The “decision” of the administrator therefore occurs under those laws. The complaint challenges that decision, and the anticipation of a defense based on § 211(a) does not change the basis of the complaint. An objection to the procedure used during any given case, even to the Administrator’s regular ways of taking evidence and issuing decisions, is an issue arising in the course of a proceeding under the statute. Section 211(a) therefore covers it. A choice by the VA to act in a certain way, even to dispense with a hearing, is a “decision”, and a constitutional objection to that decision is one based on “law”. The Administrator could follow a different course if he were persuaded of the validity of the legal argument. Although the Administrator has disclaimed the power to review the constitutionality of Acts of Congress, he has not disclaimed the power to think about the constitutionality of his own decisions implementing the statutes — how could the Administrator promise not to change his own mind? The VA has full authority to entertain, and resolve, the kind of constitutional arguments Marozsan presses.4
What reason could Congress have had for treating a procedural contention as reviewable just because articulated under the Constitution? There is no functional reason; I have already remarked that constitutional, statutory, and common law procedural contentions are just three different ways of expressing a demand for “fair” procedure — which here means procedure reasonably calculated to produce an accurate result. Marozsan’s contentions could be founded on the Administrative Procedure Act more naturally than on the Constitution, yet as claims under the APA they are securely barred by § 211(a) — even though § 211(a) does not mention the APA (or the common law) any more than it mentions the Constitution. If we read § 211(a) to reject judicial determination of APA claims, the analysis of APA claims decked out in constitutional garb must be the same.
Marozsan’s constitutional contentions add up to the belief that the VA’s proce*1493dures are not sufficiently careful, or adversarial, to produce accurate resolutions of disputed factual issues. They are in the mold of Mathews v. Eldridge, 424 U.S. 319, 96 S.Ct. 893, 47 L.Ed.2d 18 (1976). Yet on what account would constitutional arguments concerning the procedures used to achieve accuracy be reviewable, when the substantive decisions themselves are not? We know for sure — from Lynch and the debates in 1970 — that Congress meant to cut off review of claims that decisions are inaccurate. Lynch tells us that even the no-evidence case is not reviewable, and what could be less accurate than a decision without a scintilla of evidence behind it? A clear review-preclusion statute must be enforced even when it cuts off constitutional arguments. Swain v. Pressley, 430 U.S. 372, 377-78, 97 S.Ct. 1224, 1227-28, 51 L.Ed.2d 411 (1977). Since Congress did not want the VA’s decisions to be reviewed for accurate implementation of the statutory criteria, why attribute to Congress a decision to provide judicial review of contentions that the procedures used in reaching those decisions did not ensure a sufficient degree of accuracy?
The political process is the guarantor of sufficient accuracy over the run of cases. To say that Congress meant to leave the decision in individual cases to the agency is not to say that the decision has been left to solitary, unaccountable bureaucrats. The staff at the VA is less solitary, and more accountable, than is the federal judiciary. Hearing officers receive evidence and make decisions; dissatisfied applicants may appeal to the Board of Veterans Appeals. 38 C.F.R. § 3.103. State judges with the benefit of less tenure regularly hear and decide cases fairly; the public servants at the VA can do likewise. Savage v. CIA, 826 F.2d 561, 563-64 (7th Cir.1987). Before Congress made the Court of Claims an Article III tribunal, final decisions concerning claims against the government almost invariably were made outside the Article III system. Cf. Glidden Co. v. Zdanok, 370 U.S. 530, 82 S.Ct. 1459, 8 L.Ed.2d 671 (1962). A veteran dissatisfied with the handling of his claim appeals not to the courts but to veterans’ organizations (which have exercised considerable influence in shaping the current system) and to the successors in office of those who made the decision. The VA does not use res judicata, 38 C.F.R. § 3.105, and Marozsan has received numerous dispositions through the years. That the VA rated him 20% disabled effective March 1980, for an injury supposedly incurred in 1949, says something for the agency’s patience and open-mindedness. Congress also monitors the VA closely, at the behest of organized veterans’ groups. Judges are not the only or the best friends of applicants for benefits.
IV
The majority offers several additional reasons for giving a stingy reading to § 211(a). One is the presumption in favor of judicial review of the constitutionality of agency action. The Supreme Court has said many times that there is such a presumption. E.g., Califano v. Sanders, 430 U.S. 99, 109, 97 S.Ct. 980, 986, 51 L.Ed.2d 192 (1977) (constitutional contention reviewable unless there is clear and convincing evidence of a contrary legislative decision). The most recent, and most emphatic, of these cases is Webster v. Doe, — U.S. -, 108 S.Ct. 2047, 100 L.Ed.2d 632 (1988), which holds that a district court may review the constitutionality of the CIA’s dismissal of an employee even though a statute, 50 U.S.C. § 403(c), gives the Director of Central Intelligence authority to dismiss any employee whenever “in his discretion” he “shall deem such termination necessary or advisable”.
Presumptions in favor of review are not, however, bits of “evidence” that stand unless overcome by stronger evidence. This presumption is a rule of statutory construction, which leaves in place all of the usual tools of that art. United States v. Fausto, — U.S. -, 108 S.Ct. 668, 675-76, 98 L.Ed.2d 830 (1988). Cases such as Webster deal with statutes giving broad discretion to the Executive Branch but not regulating *1494the relations between the Executive and Judicial Branches. Nothing on the statute books said that the CIA’s decision was un-reviewable. Webster and like cases hold that given the grant of federal-question jurisdiction by 28 U.S.C. § 1331, and its application to the United States under 5 U.S.C. § 702, when Congress does not say that review of constitutional issues is precluded, only the clearest evidence of intent closes the doors. One may question why intent divorced from language would ever do the trick, given multiple grants of jurisdiction to award prospective relief. When Congress passes a law withdrawing the authorization of § 1331 and § 702, that law is to be enforced. Department of the Navy v. Egan, — U.S. -, 108 S.Ct. 818, 823-27, 98 L.Ed.2d 918 (1988); Briscoe v. Bell, 432 U.S. 404, 409-10, 97 S.Ct. 2428, 2431, 53 L.Ed.2d 439 (1977). Cf. Block v. Community Nutrition Institute, 467 U.S. 340, 104 S.Ct. 2450, 81 L.Ed.2d 270 (1984); Southern Ry. v. Seaboard Allied Milling Corp., 442 U.S. 444, 99 S.Ct. 2388, 60 L.Ed.2d 1017 (1979); Morris v. Gressette, 432 U.S. 491, 97 S.Ct. 2411, 53 L.Ed.2d 506 (1977) (three cases inferring preclusion from statutory structure). It is to be enforced even when, as in Swain v. Pressley, the subject foreclosed is a constitutional argument. Now the “intent” shoe is on the other foot. Given language barring review, it takes a showing of intent to authorize it. Section 211(a) is a law barring review; Lynch and Lindahl show that the Supreme Court recognizes it as one. How could Congress have been clearer, except by adding “and we really mean it!” at the close?
The notion of a presumption of judicial review brews trouble in litigation, for it collides with another line of cases — the one saying that waivers of sovereign immunity are to be construed narrowly. In addition to King and Testan, see, e.g., Block v. North Dakota, 461 U.S. 273, 287, 103 S.Ct. 1811, 1819, 75 L.Ed.2d 840 (1983); United States v. Sherwood, 312 U.S. 584, 590, 61 S.Ct. 767, 771, 85 L.Ed. 1058 (1941). These lines of cases do not cross-cite. Many of the cases employing a presumption of review involved claims to benefits — the very circumstance that calls for strict construction of the waiver. A formal distinction is that only the substantive terms of the waiver are strictly construed, while procedural attributes (such as judicial review to ensure compliance with substance) are generously read. The distinction does not harmonize the cases, however, because several in the “strict” line dealt with the availability of judicial review. Block v. North Dakota, for example, declines judicial review by construing a statute of limitations in the sovereign’s favor. And the distinction does not convince because it does not explain why substantive disputes should be resolved with a thumb on the government’s side of the scales and procedural disputes with the other thumb on the other side.
A more plausible distinction is that the presumption of review comes into play when a court is asked to issue prospective relief, while the strict construction approach takes over when the court is asked to compel the payment of money. Then the line is drawn where the eleventh amendment and correlative principles of governmental immunity draw it: prospective relief is permitted against the sovereign once the fiction of Ex parte Young, 209 U.S. 123, 28 S.Ct. 441, 52 L.Ed. 714 (1908), is indulged, even though awards of damages are not. United States v. Mitchell, 463 U.S. 206, 212-19, 103 S.Ct. 2961, 2965-69, 77 L.Ed.2d 580 (1983); United States v. Mottaz, 476 U.S. 834, 841, 106 S.Ct. 2224, 2229, 90 L.Ed.2d 841 (1986). When courts have the constitutional power to award relief, there is a presumption that Congress authorized the award; when the power to award relief depends on an affirmative authorization, this waiver is read more cautiously. If this is the right line, then Johnson and Traynor were on one side (the prospective-injunction side), while Maroz-san’s claim, along with Winslow and Mathes, are on the other (the retrospective-money side). Webster v. Doe, the most potent of the pro-review cases, is on the Johnson and Traynor side of the line. Plaintiff Doe sought only prospective relief: reinstatement but not back pay, — U.S. at -, 108 S.Ct. at 2051.5 Marozsan *1495wants money; even if we construe Maroz-san’s claim as looking forward, what he wants prospectively is cash, only. So although judges are apt to get lost hacking through this jungle of inconsistent precedents, a bold assertion of sovereign immunity in a money case still ought to be respected.
Another of the majority’s contentions is that we ought to construe § 211(a) to avoid holding it unconstitutional, perhaps even to avoid addressing constitutional questions about it. See Edward J. DeBartolo Corp. v. Florida Gulf Coast Building & Construction Trades Council, — U.S. -, 108 S.Ct. 1392, 1397, 99 L.Ed.2d 645 (1988); The Charming Betsy, 6 U.S. (2 Cranch) 64, *118, 2 L.Ed. 208 (1804). Construction to avoid unconstitutionality or a serious question, see NLRB v. Catholic Bishop of Chicago, 440 U.S. 490, 500-01, 99 S.Ct. 1313, 1318-19, 59 L.Ed.2d 533 (1979), must be distinguished from revising statutes to avoid any questions at all. What with the proliferation of constitutional “questions”, courts could do anything they pleased. Henry J. Friendly, Benchmarks 210-11 (1967). This canon properly may be invoked only when a substantial problem coincides with the possibility of fairly reading the statute to avoid that problem. “This canon of construction does not give a court the prerogative to ignore the legislative will in order to avoid constitutional adjudication.” CFTC v. Schor, 478 U.S. 833, 106 S.Ct. 3245, 3252, 92 L.Ed.2d 675 (1986) (collecting cases); see also Swain, 430 U.S. at 378 n. 11, 97 S.Ct. at 1228 n. 11. Construing statutes to avoid all constitutional questions treats the penumbra around the Constitution as if it has independent force, and thereby denies effect to real laws on the basis of insubstantial “concerns”. See Boxhorn’s Big Muskego Gun Club, Inc. v. Electrical Workers, 798 F.2d 1016, 1021 (7th Cir.1986).
Is there a serious constitutional problem in § 211(a)? No; there is not even a humorous one. When Congress creates a right to recover money from the United States, it may “provide an administrative remedy and make it exclusive.” Dismuke v. United States, 297 U.S. 167, 172, 56 S.Ct. 400, 403, 80 L.Ed. 561 (1936). Accord, United States v. Babcock, 250 U.S. 328, 331, 39 S.Ct. 464, 465, 63 L.Ed. 1011 (1919). It does so as a condition on the waiver of sovereign immunity, for the power to stand on immunity implies the power to waive immunity with conditions. E.g., Lehman v. Nakshian, 453 U.S. 156, 101 S.Ct. 2698, 69 L.Ed.2d 548 (1981) (the power of Congress to withhold any remedy against the Treasury permits it to grant a remedy without a jury trial). All § 211(a) does is make the administrative remedy exclusive. Until recently there was doubt whether Congress could allow Article III tribunals to resolve claims against the Treasury. Five Justices held in Glidden (on two incompatible theories) that Congress may. It is quite a leap to transmute this grudging permission into constitutional compulsion to provide review in the Article III courts.
Perhaps one could say, as two judges did in dicta in Bartlett v. Bowen, 816 F.2d 695, 703-11, rehearing en banc vacated, 824 *1496F.2d 1240 (D.C.Cir.1987), that sovereign immunity no longer permits the United States to close the courts to requests for funds from the Treasury; in that event the foundation for cases such as Dismuke would be gone, and there could be a substantial question about § 211(a). The ruminations in Bartlett are unpersuasive, however, and a majority of the D.C. Circuit has rejected them, see 816 F.2d at 712-28 (Bork, J., dissenting, and discussing cases on, and scholarly treatments of, the subject), 824 F.2d at 1248-49 (Bork, Starr, Buckley, Williams & D.H. Ginsburg, JJ., dissenting, calling the dicta “plainly wrong”), 824 F.2d at 1247 (Silberman, J., concurring in the order vacating rehearing but “inclined to favor the dissent's analysis” on the jurisdictional question). It is familiar in the jurisprudence of the eleventh amendment that the constitutional footing of a claim does not authorize judicial review or open state treasuries, e.g., Kentucky v. Graham, 473 U.S. 159, 169-70, 105 S.Ct. 3099, 3107, 87 L.Ed.2d 114 (1985); Cory v. White, 457 U.S. 85, 90, 102 S.Ct. 2325, 2328, 72 L.Ed.2d 694 (1982); Alabama v. Pugh, 438 U.S. 781, 98 S.Ct. 3057, 57 L.Ed.2d 1114 (1978). Congress’ power to assert or waive sovereign immunity is no less than the states’ immunity under the eleventh amendment. See Maricopa County v. Valley National Bank, 318 U.S. 357, 362, 63 S.Ct. 587, 589, 87 L.Ed. 834 (1943).
My colleagues’ assertion that “[sjince the Administrator lacks sovereign authority to contravene the Constitution, he cannot assert sovereign immunity from liability for such acts” (maj. op. at 1477) misunderstands sovereign immunity. They might say with equal force that because no one has “authority to contravene the Constitution” (or law) there is no such thing as sovereign immunity. That doctrine nonetheless exits. It is not a privilege to violate the law, any more than judicial immunity is such a privilege. No person or institution has “authority to contravene” the law. Sovereign immunity has to do with the allocation of powers among the branches of the government. Congress alone has the authority to decide when to permit suit against the United States as an entity. Graham, Cory, Pugh, and a host of other cases show that governmental immunity defeats a claim for money as a remedy, even when the plaintiff has shown a violation of the Constitution. That is why in Bivens v. Six Federal Agents, 403 U.S. 388, 91 S.Ct. 1999, 29 L.Ed.2d 619 (1971), the plaintiff seeking damages on account of unconstitutional action by an agent of the federal government had to sue the person who acted unconstitutionally: sovereign immunity barred access to the Treasury. Ours is a suit seeking money from the United States in its own name, which is what sovereign immunity is all about. Officers of the government are personally liable for their wrongs, for they are not the sovereign, but no such officer is a party to this case.
The many subtle questions that attend legislative efforts to close the Article III courts to suits seeking to protect natural liberty and private property from governmental interference — or to close courts selectively in ways reflecting hostility to certain constitutional rights — are irrelevant to the validity of § 211(a). See Gerald Gunther, Congressional Power to Curtail Federal Court Jurisdiction: An Opinionated Guide to the Ongoing Debate, 36 Stan.L. Rev. 895 (1984). Marbury v. Madison, 5 U.S. (1 Crunch) 137, *177-80, 2 L.Ed. 60 (1803), heldt that because a court with jurisdiction must decide, it is entitled to resolve a constitutional question logically entailed in reaching a conclusion. Judicial review came from the authority to decide. It is not yet true that the authority to decide may be inferred from the institution of judicial review, for that institution presupposed the rightful authority to decide. So it is not true that whenever someone presents a constitutional issue, the court must have power to decide. Quotations (maj. op. at 1477) from cases such as Crowell v. Benson, 285 U.S. 22, 56-57, 52 S.Ct. 285, 294-95, 76 L.Ed. 598 (1932), miss this point. The statute at issue in Crowell authorized a federal agency to transfer money from private person A to private person B on finding that A had committed a civil wrong. Congress passed another statute authorizing judicial review of the administrative *1497decision. Crowell, which dealt with the rights of the person called on to pay, had nothing to do with suits against the United States seeking welfare payments, or with the power of courts to engage in review in the teeth of statutes denying them jurisdiction. No case has held that Congress must allow review of claims against the Treasury; Dismuke and many other cases that my colleagues choose not to discuss hold that it need not.6
The Supreme Court has never held a door-closing statute unconstitutional, and it has seen many, starting with the denial of all federal-question jurisdiction to the federal courts in the first Judiciary Act. It has enforced those it encountered. Ex parte McCardle, 74 U.S. (7 Wall.) 506, 19 L.Ed. 264 (1868); and see the string of modern cases such as Fausto (no review of certain disputes arising out of governmental employment); NLRB v. United Food Workers, — U.S. -, 108 S.Ct. 413, 98 L.Ed.2d 429 (1987) (no review of decisions of the Board’s General Counsel not to issue, or to dismiss in settlement, unfair labor practice complaints); and United States v. Erika, Inc., 456 U.S. 201, 102 S.Ct. 1650, 72 L.Ed.2d 12 (1982) (no review of certain decisions under the Medicare statute). Cases such as Erika show that even a request for a remand to the administrative agency requires statutory authorization and may be blocked by a statute precluding review. Interpreting Maroz-san’s suit as one requesting a remand rather than money does not give us the statutory authority necessary to grant that relief.7 A remand is just a step on the way to money, and statutes precluding judicial review of agency action invariably are treated as forbidding adjudication followed by remand to the agency. Cf. Florida v. Long, — U.S. -, 108 S.Ct. 2354, 101 L.Ed.2d 206 (1988) (treating an increase in pension payments for future months as a “retroactive” award of money because the payments are attributable to earlier work — a conclusion equally applicable to veterans’ disability awards).
*1498Marozsan contends that § 211(a) violates the equal protection component of the Due Process Clause, which is ludicrous. The majority gives it the back of the hand, maj. op. at 1471 n. 3. Veterans are hardly a disfavored class; instead they receive benefits and preferences in governmental employment. E.g., Personnel Administrator of Massachusetts v. Feeney, 442 U.S. 256, 99 S.Ct. 2282, 60 L.Ed.2d 870 (1979). Veterans have not been singled out compared with, say, persons whose security clearances are revoked, see Egan, or persons alleging unfair labor practices, see United Food Workers, or chemical companies that believe that their formulae have been appropriated, see Thomas v. Union Carbide Agricultural Products Co., 473 U.S. 568, 582-84, 105 S.Ct. 3325, 3333-34, 87 L.Ed.2d 409 (1985), or railroad workers who believe that their employers have disregarded collective bargaining agreements, see Union Pacific R.R. v. Sheehan, 439 U.S. 89, 94, 99 S.Ct. 399, 402, 58 L.Ed.2d 354 (1978). Section 211(a) is rationally related to the purposes of ensuring that the veterans’ benefits program operates outside the adversarial system (see National Ass’n of Radiation Survivors) and ensuring that complex questions get consistent treatment from a unitary Administrator. It serves these functions well — too well, Marozsan tells us — which dissipates any equal protection challenge. There is no plausible constitutional challenge to § 211(a), and no reason to “construe” the statute to sidestep a phantom attack.
One more part of my colleagues’ approach calls for comment. The majority uses a familiar technique that starts with an unpalatable hypothetical, observes that there must be some way to deal with that, and concludes that given the power to reach the necessary result in the hypothetical situation, there is no principled ground to stop short of the result about which there is greater doubt. See Paul Gewirtz, The Jurisprudence of Hypotheticals, 32 J. Legal Education 120 (1982). Applied to § 211(a) the method proceeds: (1) racial and religious discrimination are obnoxious; (2) the Nation has decided that judicial review is necessary to wipe out these invidious forms of discrimination; (3) surely Congress would allow judicial review of claims that the Administrator of the VA was granting benefits to white or Christian veterans but denying them to identically-situated black or Jewish veterans; (4) there is no evidence in the legislative debates that Congress wanted to permit racial discrimination, which a denial of judicial review would do; yet (5) nothing in § 211(a) would permit judicial review of claims based on racial or religious discrimination but deny it for other claims; and because (by # 3) there must be review of claims of invidious discrimination, (6) the legislation necessarily permits review of other constitutional claims.
I do not doubt these premises; I doubt only the conclusion. This is too syllogistic, and the “life of the law has not been logic: it has been experience.” Oliver Wendell Holmes, Jr., The Common Law 1 (1881). See also Lydon v. Justices of Boston Municipal Court, 698 F.2d 1, 10 (1st Cir.1982) (Campbell, J., dissenting), reversed, 466 U.S. 294, 313, 104 S.Ct. 1805, 1815, 80 L.Ed.2d 311 (1984). The technique starts by putting a hypothetical question to the minds of deceased legislators who never thought about it while they were living — let alone while they were in Congress assembled. These ghostly legislators always give the answer the questioner prefers; they are in no position to do otherwise. Their “answer” becomes the basis for insisting that the statute they actually wrote be construed consistently with the views we have put in their mouths — for they did not deny that they wanted the statute construed consistently with the answer they did not give! Yet the fact that they did not answer a question that was not asked of them does • not grant us the authority to disregard the answer they gave to the question that was asked. Rodriguez v. United States, 480 U.S. 522, 107 S.Ct. 1391, 1393, 94 L.Ed.2d 533 (1987).
The terror of extreme hypotheticals produces much bad law. Just as the answer to the claim that “the power to tax involves the power to destroy” (and the conclusion that there can be no power to tax) is “not
*1499... while this Court sits”, Panhandle Oil Co. v. Mississippi ex rel. Knox, 277 U.S. 218, 223, 48 S.Ct. 451, 453, 72 L.Ed. 857 (1928) (Holmes, J., dissenting), overruled by Alabama v. King & Boozer, 314 U.S. 1, 62 S.Ct. 43, 86 L.Ed. 3 (1941), so the generic answer to the extreme hypothetical is “we’ll cross that bridge when we come to it”. If Congress actually were to do something that from today’s perspective seems “extreme” maybe it would have a good reason; the flow of events may make things more sensible when done than they appear from a distance. Legislation that in 1905 appeared to a majority of Justices to be bizarre and destructive appears to a majority today to be benign, perhaps inevitable. Only after understanding an “extreme” proposal are we in a position to assess its consequences and craft the proper response. Taking anticipated reactions to an extreme result — one divorced from plausible justifications — and using this as the basis of reconstructing a law on the principle of parity of treatment, disregards the genius of common law development. We must start from the cases and laws at hand and understand them as best we can. The stuff of daily litigation must be resolved under existing statutes; fear of the future, of what’s at the bottom of a long, slippery slope, is not a good reason for today’s decision. See Frederick Schauer, Slippery Slopes, 99 Harv.L.Rev. 361, 368-77 (1985).
Consider the application of the extreme hypothetical to § 211(a). My colleagues are convinced that Congress would not close the courts to claims of racial discrimination. Detestable though discrimination is, Congress rationally might conclude that judicial review of claims of discrimination would come at too great a purchase. Many claims of discrimination take the form: “My case was sufficient under the governing law, yet I was denied benefits; therefore my race must have been held against me”, or “My claim is as good as that of someone else, yet his request was granted and mine was denied; therefore my race must account for the difference.” This line of reasoning, common in litigation about age, sex, and race discrimination in employment, would break down all barriers against review. If claims agents of the YA engage in racial, religious, or political discrimination, the Board of Veterans’ Appeals can correct the situation. The federal courts do not receive complaints that adjudicatory tribunals are themselves discriminators; the vast jurisprudence arising out of social security disability claims (a parallel to veterans’ claims) does not reveal any substantiated charge of discrimination by the administrative law judges and the Appeals Council of the Department of Health and Human Services. Congress could well conclude that the ratio of actual discrimination by professional adjudicators to efforts to obtain back-door review of the merits is too small to authorize a search by another system of tribunals. Judges, of all people, should recognize that this conclusion may be reached by conscientious persons. Judges have absolute immunity, including immunity from suits seeking damages on account of racial discrimination, because of a belief that the costs of such litigation would swamp the good it could do in correcting wrongs. A thoughtful legislature also could come out the other way, providing review of claims of racial (and other) discrimination but not of “the merits” of benefits decisions.
A belief that Congress is likely to do this, if asked, is not a sufficient reason to say that it has already done so, and because it has neglected to distinguish racial from other constitutional contentions, every constitutional argument is reviewable today. This confuses a prediction with a fact. Section 211(a) was passed in 1933 and last amended in 1970. The first Congress of the Franklin Roosevelt Administration was not Searching for ways to expand the power of the courts, and any proposal to protect civil rights would have produced an endless filibuster. Not until 1972, in an amendment to Title VII of the Civil Rights Act of 1964, did Congress first expose the United States to suits based on claims of racial discrimination. Not until 1976, in an amendment to 5 U.S.C. § 702, did Congress first permit litigation against the United States, as a general matter, *1500when subjects other than money are at issue. It is revisionism to read earlier statutes as if they had been enacted by recent Congresses. The clash in this case is not between sensible statutory construction and wooden “plain meaning” construction, as the concurrence would have it; no one believes that words have meanings apart from linguistic (and other) contexts. The clash is between understanding and enforcing a 1933 law on its own terms and the impulse to follow “strong modern trend[s]”, (Posner, J., concurring op. at 1484).
If we must choose between applying the statute we have and imagining how today’s Congress would handle the same problem, we must take the former path. “[UJnen-acted approvals, beliefs, and desires are not laws.” Puerto Rico Department of Consumer Affairs v. Isla Petroleum Corp., - U.S. -, 108 S.Ct. 1350, 1354, 99 L.Ed.2d 582 (1988). If we must choose between applying the statute we have and imagining how the enacting Congress would have handled the questions we pose for it (Posner, J., concurring op. at 1483), again the path is clear. The case “must be measured against the ... statute Congress passed, not against the hypothetical statute it is most ‘reasonable to believe’ Congress would have passed had it considered the question”. Lukhard v. Reed, 481 U.S. 368, 107 S.Ct. 1807, 1812-13 n. 3, 95 L.Ed.2d 328 (1987) (plurality opinion).8 If we must wonder whether in 1933 Congress “would have agreed to extinguish all constitutional remedies against the Veterans’ Administration that veterans might otherwise possess, or might acquire by virtue of subsequent enactments” (Posner, J., concurring op. at 1483), the answer is that they had none against the United States as an entity in 1933, and nothing in § 211(a) inhibits Congress from doing in “subsequent enactments” whatever it pleases. The concurrence thinks it implausible to conclude that § 211(a) “forever preclude[s]” (Posner, J., concurring op. at 1483) remedies — which would indeed be implausible, but which none of the contending interpretations implies. Congress may amend or supersede § 211(a) tomorrow, yet later laws such as 5 U.S.C. § 702 leave existing bars alone. Our question is whether that statute authorizes relief against the United States: for if it does not, sovereign immunity bars what Maroz-san wants. My colleagues must believe that courts are responsible for keeping laws “up to date” even if that means altering them; the Supreme Court believes that “a statute ‘is not an empty vessel into which [a] Court is free to pour a vintage that ... better suits present-day tastes.’ ... Considerations of this kind are for Congress, not the courts.” National Broiler Marketing Ass’n v. United States, 436 U.S. 816, 827, 98 S.Ct. 2122, 2130, 56 L.Ed.2d 728 (1978).
In trying to produce consistency between the treatment of racial claims to which my colleagues believe Congress “would have agreed” had it been asked, and the actual treatment of other claims, the court may create a real inconsistency. Nothing in § 211(a) suggests that constitutional claims are to be treated differently from claims based on statutes other than veterans’ benefits legislation. The Administrative Procedure Act is one such statute. If consistency is our aim, must we not hold that all contentions based on the APA are reviewable? The majority opinion seems to balk, treating constitutional claims as special; the concurring opinion bites the bullet for most laws, concluding: “section 211(a) does not foreclose judicial review of actions by the Veterans’ Administration that are alleged to violate laws other than the Veterans’ Benefits Act itself” (Posner, J., concurring op. at 1481-82). The APA is such a law, but the concurrence shies away. Congress — the one sitting in 1933, the one sitting in 1970, the one sitting today — would be surprised to learn that standard APA review is available in veterans’ benefits cases, just because my colleagues are confident how it *1501would answer questions about racial discrimination. Stopping short, however, entails the same sort of line-drawing that the court otherwise abjures. The right lesson to draw from this cascade of inferences is that the method of instilling meaning into a statute based on lines Congress did not draw, as a result of things it did not consider, is fundamentally flawed.
V
It still may be worth recognizing what this case does not involve. It does not involve a demand for prospective relief of the sort that is granted despite governmental immunity. It does not involve a “claim” arising under the Constitution; it involves only a constitutional issue arising in the course of adjudicating a claim arising under the veterans’ laws. It has long been recognized that litigants have no general right to resolution in federal court of constitutional issues arising in other forums. E.g., Johnson v. Mississippi, 421 U.S. 213, 95 S.Ct. 1591, 44 L.Ed.2d 121 (1975) (collecting cases). It does not involve a contention that the VA used racial, religious, political, or other criteria. It does not involve a claim for damages against the persons said to have violated the Constitution. Compare Bivens with Bush v. Lucas, 462 U.S. 367, 103 S.Ct. 2404, 76 L.Ed.2d 648 (1983), and Schweiker v. Chilicky, — U.S. -, 108 S.Ct. 2460, 101 L.Ed.2d 370 (1988). It is a suit seeking money against the United States in its own name, a claim traditionally barred unless a statute unambiguously authorizes the suit.
The submission my colleagues find most appealing is Marozsan’s assertion that the YA uses a “quota system”. What he apparently means is that the Board of Veterans’ Appeals reverses approximately the same percentage of decisions each year.9 This is true of federal appellate courts too. The table shows the percentage of reversals by the Board, and by the United States courts of appeals in civil cases, in the same years:10
Reversal Rate
U.S.
Year BVA Courts
1975 30.7 20.7
1976 29.5 21.9
1977 28.3 16.9
1978 25.9 19.7
1979 26.3 18.8
1980 26.8 19.0
1981 26.7 19.3
1982 28.2 18.4
1983 29.1 18.4
1984 28.7 18.6
1985 29.6 18.3
1986 31.6 20.0
1987 — 17.5
*1502Both institutions have stable rates, though the Board’s is not so flat as Marozsan believes. The stability is no surprise, for the rate of reversal depends on litigants’ choices. Because taking an appeal is costly, litigants do so only if they believe that the law and facts offer them a decent chance of success. Unless the costs of appealing change, the rate of success for a given kind of case should be stable. See George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. Legal Studies 1 (1984). Stability no more suggests that the Board of Veterans’ Appeals uses a “quota” than that courts do. Even if it did, what would judicial review entail? Determining whether the stability in the rate of reversal comes from equilibrium forces or from inclination to damn the facts and get the cases out would require the most painstaking case-by-case inquiry. This “quota” contention, if taken seriously, would demand the examination of the files even of the veterans who do not seek judicial review. The Board of Veterans’ Appeals decides on the merits between 35,000 and 45,000 cases per year, about double the output of all of the U.S. courts of appeals put together, so the inquiry could take a little time. A process farther removed from the language and functions of § 211(a) is impossible to imagine.
This is not to praise § 211(a). Arguments may be made for and against review. Judicial review by the more than 700 district judges would sacrifice consistency and technical expertise, but it might improve the implementation of the rule of law at the VA. Many people are dissatisfied with the way the VA handles claims, which may be why proposals to provide for judicial review have currency. A conscientious Congress might strike the balance either way. We know, however, how Congress has struck the balance between 1933 and today. Wherever the fringe of § 211(a) may be, its core is the principle that there shall be no case-by-case review to ensure accurate implementation of the statutory rules. It is exactly case-by-case review to ensure accuracy that the court today requires.
“Without jurisdiction the court cannot proceed at all in any cause. Jurisdiction is power to declare the law, and when it ceases to exist, the only function remaining to the court is: that of announcing the fact and dismissing the cause.” Ex parte McCardle, 74 U.S. (7 Wall) at 514. The federal courts lack jurisdiction of Marozsan’s case. Courts must scrupulously observe the limits on their own jurisdiction. Judges have only power granted, not whatever powers they think best. If I must choose between reading even this simple statute as license to follow a “modern trend” (Posner, J., concurring op. at 1484) of decisions under laws post-dating § 211(a) — a process that makes one assertion of judicial power the foundation for the next — and reading this statute as a choice binding on me until amended by its author, I choose the latter without hesitation or regret.
. Rescigno v. Walters, 773 F.2d 47 (3d Cir.1985), may support Devine, but more likely holds only that § 211(a) does not bar a challenge to the validity of § 211(a) itself.
. The specter that a literal construction of § 211(a) nullifies Title VII of the Civil Rights Act of 1964, extended to the government in 1972, see concurring op. at 1482, is therefore insubstantial. The introductory clause to § 211(a) — "except as provided in sections 775, 784, and as to matters arising under chapter 37 of this title” — permits review of some additional matters. Sections 775 and 784 deal with veterans’ life insurance, chapter 37 with home loans.
. The Appeals Council of the Department of Health and Human Services heard 44,621 administrative appeals in fiscal year 1986. Social Security Administration, Executive Handbook of Selected Data 32 (May 1987). During fiscal year 1987, the claimants took 13,322 of these to the district courts under 42 U.S.C. § 405(g). (This assumes a lag between administrative closure and the institution of the civil action.) Almost 1,000 appeals in social security disability cases are filed yearly. Director of the Administrative Office of the United States Courts, Annual Report 1987 Table B-7 (1988) (982 appeals classified as “U.S. Defendant — Social Security Laws” in fiscal 1987). Social security disability cases thus make up some 6% of the district courts' cases and 3% of the appellate courts’. The Board of Veterans' Appeals hears almost exactly as many cases every year as does the Appeals Council. E.g., Veterans Administration, Annual Report 1984 p. 125 (1985) (61,328 appeals filed in fiscal 1984; after settlement of many of these, the Board issued 44,064 decisions).
. The majority’s contrary statements, e.g. maj. op. at 1472 (because "the V.A. disclaims authority to consider constitutional claims" the veteran “has ... no forum at all in which to raise his due process claim”), rest on a confusion of challenges to statutes with challenges to practices. The VA told the Supreme Court in Johnson that it lacked the authority to disregard statutes on constitutional grounds. I doubt that this is correct; the Constitution is supreme for administrative as well as judicial personnel, and nothing about the constitutional hierarchy implies that only judges have the power to place the Constitution above mere law. Every governmental official has the duty to do this. The power of judicial review comes from the hierarchy or rules, with the Constitution superior to law; that same hierarchy applies to every other governmental actor, and each takes an oath of obedience to the Constitution. See Rex E. Lee, The Provinces of Constitutional Interpretation, 61 Tulane L. Rev. 1009, 1012-14 (1987) (collecting presidential views on the subject). But see Lear Siegler, Inc. v. Lehman, 842 F.2d 1102, 1119-26, (9th Cir.1988), and Continental Air Lines, Inc. v. Department of Transportation, 843 F.2d 1444, 1455 (D.C.Cir.1988), both saying that the Executive Branch is forbidden by the Constitution from giving the Constitution precedence over statutes. To state the position is to refute it. Each case relies on the requirement in Article II § 3 that the President "take Care that the Laws be faithfully executed”, but “Laws” for this purpose must include the Constitution. Cf. National Wildlife Federation v. ICC, 850 F.2d 694 (D.C.Cir.1988) (remanding a case so that the agency may consider a constitutional objection to its decision). Let this pass. Assume for the moment that the VA takes a contrary view of its powers and duties. The VA’s view in Johnson was that when Congress says “jump” it jumps. Nothing about the VA’s belief that statutes must be followed implies that when promulgating or revising 38 C.F.R. § 3.103 (establishing rules for holding hearings), the Administrator does not consider constitutional questions; indeed, that regulation is captioned "Due process — procedural and appellate rights ...”. If the Board uses a "quota” in deciding cases, or exalts the medical views of general practitioners over those of specialists, someone at the VA has decided to do this and he, she, or a successor in office can reverse course.
. Webster v. Doe does not make anything turn on this, but that is not surprising in light of the *1495absence of a claim for money, and the waiver of sovereign immunity in 5 U.S.C. § 702 for claims not seeking damages. I concede, however, that another aspect of Webster is both at odds with the approach taken here and not so easy to explain. Webster relied, - U.S. at -, 108 S.Ct. at 2054, on Johnson v. Robison for the proposition that "where Congress intends to preclude review of constitutional claims its intent to do so must be clear." The Court did not remark on the potential difference between reviewing the constitutionality of enabling legislation (as in Johnson) and reviewing day-today decisions (as in Webster). That difference was one Johnson itself stressed in order to demonstrate that the Court could review the constitutionality of a law because it was not reviewing a decision of the Administrator within the meaning of § 211(a). And, the Court added in Johnson, review of the law itself would not create a risk of embroiling the courts in the tens of thousands of disputes under veterans’ benefits acts, the very thing § 211(a) is designed to prevent. Although Webster did not refer to those limiting aspects of Johnson, neither did it repudiate them. The parties in Webster did not join issue on the scope and effect of Johnson; the Solicitor General’s briefs do not mention that case. Instead of reading tea leaves, an inferior court should stick to Johnson’s stated rationale until told that it is no longer authoritative. *1497151, 29 S.Ct. 458, 53 L.Ed. 742 (1909) (no review of constitutional claim to money concerning the government's handling of liquor), are just a smattering of the other cases on point. Perhaps these cases need revision; the majority’s opinion oozes hostility to sovereign immunity; but until the Supreme Court tells us to throw that doctrine away we ought to implement it.
. For example, Mottaz invoked sovereign immunity to prohibit adjudication of a suit that included a claim that the government had violated the Due Process Clause; Lynch, 292 U.S. at 581, 54 S.Ct. at 844, remarked that in a suit seeking recovery from the Treasury "[t]he rule that the United States may not be sued without its consent is all-embracing” and "applies alike to causes of action arising under acts of Congress ... and to those arising from some violation of rights conferred upon the citizen by the Constitution" (citing Schillinger v. United States, 155 U.S. 163, 166, 15 S.Ct. 85, 86, 39 L.Ed. 108 (1894)); Maricopa County used sovereign immunity to eject from federal court a contention that the grant of certain powers to the Reconstruction Finance Corp. violated the Due Process Clause because it retroactively burdened property, remarking, 318 U.S. at 362, 63 S.Ct. at 589: "No such suit may be maintained without the consent of the United States”; see also California v. Arizona, 440 U.S. 59, 63, 99 S.Ct. 919, 922, 59 L.Ed.2d 144 (1979) (“It is clear, of course, that Congress could refuse to waive the Nation’s sovereign immunity in all cases or only in some cases but in all courts. Either action would bind this Court even in the exercise of its original jurisdiction” — a jurisdiction granted by the Constitution and therefore not subject to the legislative power to make "exceptions”). Morrison v. Work, 266 U.S. 481, 45 S.Ct. 149, 69 L.Ed. 394 (1925) (no review of constitutional claims concerning management of Indians’ property); and Murray v. Wilson Distilling Co., 213 U.S.
. The concurring opinion says that courts may remand benefits cases to agencies. A remand is a judicial order running against the government as an entity and therefore requires authorization. The Administrative Procedure Act provides that authorization for many cases, even when the claimant wants money — provided the money is not “damages”, see Bowen v. Massachusetts, — U.S. -, 108 S.Ct. 2722, 101 L.Ed.2d 749 (1988). Section 702 of the APA, the statute interpreted in that case, waives immunity, and the question was "how far?" Section 211(a) asserts immunity; the more general rules established by the Administrative Procedure Act do not govern. See 5 U.S.C. § 701(a)(1), providing that the judicial-review sections of the APA (including § 702) do not apply when “statutes preclude judicial review”. Sections 702(1) and (2) reinforce this; each subsection states that § 702 does not permit review that is forbidden by some other law. So we are back to § 211(a), for if it applies, § 702 (like the rest of the APA) does not, and we are missing statutory authority to grant relief against the United States as an entity.
. The plurality in Lukhard was Chief Justice Rehnquist and Justices White, Stevens, and Sca-Iia. Chief Justice Rehnquist and Justices Black-mun, O'Connor and Scalia expressed the same position even more strongly in K Mart Corp. v. Cartier, Inc., — U.S. -, 108 S.Ct. 1811, 1834-35, 100 L.Ed.2d 313 (1988) (concurring and dissenting opinion). The view therefore has the support of a majority of the Court. (Justice Kennedy has not addressed this subject.)
. There is another sense of "quota system” to which Marozsan may refer. Until recently, see New York Times, June 10, 1988, § 1, p. 8, col. 6 (reporting the abolition of this system), the VA gave bonuses to those among the 62 members of the Board of Veterans’ Appeals who disposed of the most cases per year. The line separating bonus from no bonus may be called a "quota”. Perhaps the pressure to dispose of more cases leads the members to do less well with each. This could cut in favor of the veteran as readily as against. At all events, the crush of business is attributable to legislative choices. Congress determines how much money is available to hire appellate adjudicators, which fixes how many cases each will have. Cf. Heckler v. Day, 467 U.S. 104, 104 S.Ct. 2249, 81 L.Ed.2d 88 (1984). It is not clear why financial incentives to clear up a backlog of cases — incentives that might lead to inaccurate decisions but also may induce the adjudicators to spend the extra time necessary to keep up with their work — are a constitutional problem.
. The data come from the annual reports of the Administrative Office of the U.S. Courts and of the Veterans Administration and are for the fiscal or statistical years ending in the calendar year in question. The VA has three categories of dispositions: claims allowed, denied, and remanded (or "closed"). I combine the "allowed” and "remanded” figures to produce the reversal rate. In most years there were slightly more remands than allowances. The reversal rate for the courts of appeals is the column "reversed or denied” in the Administrative Office’s annual report for the year. After 1984 the Administrative Office stopped reporting an aggregate "civil” rate and gave the rates for sub-categories of civil cases. For 1985-87 I give a reversal rate derived by dividing the total civil reversals (all reversals listed in Table B-5, less the number of criminal reversals) by the total number of terminations on the merits (less criminal decisions on the merits). Data for the VA in 1987 are not yet available.