dissenting:
When the Supreme Court decided Metro Broadcasting in 1990, the opinion was criticized vigorously for its indulgent approval of an affirmative action program adopted by the FCC and approved by Congress. “The Court abandons [its] traditional safeguard against discrimination for a lower standard of review, and in practice applies a standard like that applicable to routine legislation,” Justice O’Connor wrote in dissent. Metro Broadcasting, Inc. v. FCC, 497 U.S. 547, 110 S.Ct. 2997, 3029, 111 L.Ed.2d 445 (1990). Justice Kennedy, for his part, compared the decision’s “relaxed standard of review” to Plessy v. Ferguson, 163 U.S. 537, 16 S.Ct. 1138, 41 L.Ed. 256 (1896). Id. 110 S.Ct. at 3044.
The Supreme Court, of course, may now decide to overrule Metro Broadcasting and to require strict scrutiny of Congress’s affirmative action policies, as the four dissenters urged. But as appellate judges, our duty is to follow Supreme Court precedents, not to anticipate them. And it strikes me as impossible to reconcile the Supreme Court’s decision in Metro Broadcasting with my colleagues’ decision today.
The Supreme Court held that Congress may require the FCC to make certain broadcasting licenses available only to minorities; yet this Court reaches the surprising conclusion that Congress may not require the FCC to adopt a far milder and far less discriminatory preference program for women. Although they pay lip service to Justice Brennan’s majority opinion, my colleagues apply in practice the more exacting scrutiny of Justice O’Connor’s dissent.
My colleagues point to only one difference between this case and Metro', the statistics in one report showed a greater correlation between minority ownership and programming diversity than between female ownership and programming diversity. This poses an interesting question for social scientists, but not for judges. The study that my colleagues invoke to strike down the preference explicitly concluded that there is a link between female ownership and programming diversity. The study showed that stations owned by women are 20% more likely than stations owned by men to broadcast “women’s programming,” and about 30% more likely than stations owned by non-minorities to broadcast “minority programming.” My colleagues do not share with us the text or history of the Equal Protection clause that tells them that a 20% or 30% increase is unconstitutional, while a larger increase is not. And even though the Supreme Court in Metro cited empirical studies offered by amicus groups rather than by the FCC or Congress, my colleagues ignore similar studies that provide more of the statistics they require.
In striking down the preference policy, my colleagues have done precisely what the Supreme Court forbids them to do: they have rejected Congress’s conclusion that more female owners of broadcast stations will lead to more diverse programming, even though Metro says repeatedly that courts should defer to Congress’s conclusions about the link between ownership and programming, as long as the conclusions reflect reasoned analysis rather than archaic stereotypes. Metro’s holding is consistent with a long line of gender cases; and this, after all, is a case about gender, not race. Applying the Supreme Court’s test, I think Congress clearly used reasoned analysis when it concluded that increasing the number of women who own and manage television and radio stations will increase programming diversity. And it would be hard to conclude that the gender preference reflects outdated stereotypes, since the FCC and Congress designed it to reduce outdated stereotypes. My colleagues are free to question the wisdom of the gender preference, but the Supreme Court forbids them (for now) from striking the policy down.
I cannot join them, and respectfully dissent.
A. The Metro Case
Mr. Lamprecht was understandably nervous when the Supreme Court agreed to hear the Metro case. He was so convinced, *405in fact, that his pending challenge to the FCC’s gender preference would stand or fall on the Supreme Court’s decision about the FCC’s racial preferences, that he filed an amicus brief in Metro confessing his fears:
The preferential treatment policy employed by the Commission to advance women is a component of its more comprehensive preference policy toward minority applicants. This Court’s ruling with respect to the constitutionality of the FCC’s minority preference policy in the case sub judice will necessarily affect, and may well prove determinative of, amicus’ s pending challenge to the Commission’s female preference scheme.
Brief of Amicus Curiae Jerome Thomas Lamprecht In Support of Petitioner at 2, Metro Broadcasting, 497 U.S. 547, 110 S.Ct. 2997, 111 L.Ed.2d 445 (1991) (emphasis added).
Mr. Lamprecht had good reason to be apprehensive. The FCC’s gender preference is far milder than its race preferences, and if the Supreme Court upheld the latter, it seemed unlikely that lower courts could strike down the former. The race policies reviewed in Metro include two separate programs — a “distress sale” policy that makes certain broadcast licenses available to minorities alone; and a preference policy that makes race a “plus factor” to be weighed with other factors in comparative hearings open to minorities and non-minorities. Metro, 110 S.Ct. at 3005. The gender policy is far less severe. There is nothing remotely like a set-aside — no “distress sale” for women that excludes men from consideration. There is only a preference that makes gender a “plus factor” to be weighed against other factors; and the Supreme Court has consistently approved affirmative action programs in which race or sex are one of many factors considered in open competition. See, e.g., Johnson v. Transportation Agency, 480 U.S. 616, 641-42, 107 S.Ct. 1442, 1456-57, 94 L.Ed.2d 615 (1987); id. at 649, 656, 107 S.Ct. at 1461, 1464 (O’Connor, J., concurring).
The “plus factor” awarded for gender, furthermore, is less valuable than the “plus factor” awarded for race. According to the FCC, the “merit for female ownership and participation is warranted upon essentially the same basis as the merit given for black ownership, but ... it is a merit of lesser significance.” Mid-Florida Television Corp., 69 F.C.C.2d 607, 652 (Rev.Bd.1978), set aside on other grounds, 87 F.C.C.2d 203 (1981) (emphasis added) (awarding a “slight” merit to female applicant while granting “substantial” merit to black applicant). Since the FCC’s gender preference has far less bite than its racial preferences, Mr. Lamprecht was right to worry that a favorable decision in the Metro case would doom his own.
The Metro decision, when it came, fulfilled Mr. Lamprecht’s worst fears. A majority of the Court reaffirmed a position that had previously won only a plurality: that “benign, race-conscious measures mandated by Congress” should be reviewed under intermediate rather than strict scrutiny. 110 S.Ct. at 3008-09. The opinion repeatedly stressed the importance of deferring to Congress’s conclusions about the connection between station ownership and programming diversity, as long as the conclusions were based on reasoned analysis rather than archaic stereotypes. In light of the fact that the Supreme Court has traditionally treated gender classifications much more permissively than racial classifications, it now seemed indisputable that Metro’s approval of the FCC’s racial preferences would compel appellate courts to uphold, the FCC’s gender preference.
My colleagues, however, have reached the unexpected conclusion that Metro “compels us to strike down the program.” Ante at 391 (emphasis added). Their analysis is easily summarized. They agree that the program is subject to intermediate scrutiny — that it must be upheld if it is “substantially related” to an “important governmental objective.” They concede that encouraging diverse programming is an “important” goal for the government to pursue. But they conclude that the gender preference program is not “substantially related” to the goal of broadcast diversity because they are not convinced by some of the statistical evidence Congress had before it. In my view, they turn Metro on its head.
*406Justice Brennan’s opinion stresses the importance of deference to Congress so often the point is hard to miss. “It is of overriding significance in these cases that the FCC’s ... ownership programs have been specifically approved — indeed mandated — by Congress.” Metro, 110 S.Ct. at 3008 (emphasis added). Once again: “The FCC’s conclusion that there is an empirical nexus between minority ownership and broadcasting diversity is a product of its expertise, and we accord its judgement deference. ” Id. at 3011 (emphasis added). And finally: “[B]oth Congress and the Commission have concluded that the minority ownership programs are critical means of promoting broadcast diversity. We must give great weight to their joint determination.” Id. at 3016 (emphasis added).
Here is the central distinction (which my colleagues quote selectively, ante at 391-392):
Although we do not “ ‘defer’ to the judgment of the Congress and the Commission on a constitutional question," ... we must pay close attention to the expertise of the Commission and the factfind-ing of Congress when analyzing the nexus between minority ownership and programming diversity. With respect to this “complex” empirical question, we are required to give “great weight to the decisions of Congress and the experience of the Commission.”
Id. at 3011 (emphasis added) (citations omitted). The holding of Metro, in other words, is that although courts should not defer to Congress on constitutional questions, we should defer — or give “great weight” — to Congress on empirical questions. I see no difference between “great weight” and “deference,” which the Supreme Court and this Court have consistently treated as synonyms. See, e.g., Rostker v. Goldberg, 453 U.S. 57, 64, 101 S.Ct. 2646, 2651, 69 L.Ed.2d 478 (1981); Commonwealth Edison Co. v. United States Dep’t of Energy, 877 F.2d 1042, 1045 (D.C.Cir.1989).
This is the central point of Metro, and my colleagues do not acknowledge it. There is not even a pretense of deference to Congress anywhere in their opinion. On the contrary, they refuse to defer to Congress because they themselves are not convinced by one report of the Congressional Research Service. But Metro does not say that courts may reject Congress’s factual conclusions if they themselves are not convinced by the statistical evidence. Nor does it say what my colleagues say it says: “Any ‘predictive judgements’ concerning group behavior and the differences in behavior among different groups must at the very least be sustained by meaningful evidence.” Ante at 393. Metro says that courts must defer to Congress’s judgement that there is a link between ownership and broadcast diversity as long as “the policies are ... a product of ‘ “analysis” ’ rather than a ‘ “stereotyped reaction” ’ based on ‘ “[hjabit.” ’ ” Metro, 110 S.Ct. at 3018 (citation omitted).
B. The Gender Cases
This conclusion was not new with Metro. In a long line of gender cases, the Court has made clear that the underlying purpose of the legal test that requires a gender classification to be “substantially related” to “important objectives” is to make sure that the government is not making distinctions between men and women based on “archaic and stereotypic notions.” Mississippi University for Women v. Hogan, 458 U.S. 718, 725, 102 S.Ct. 3331, 3336, 73 L.Ed.2d 1090 (1982). As Justice O’Connor wrote for the Court:
The purpose of requiring that close relationship [between means and ends] is to assure that the validity of a classification is determined through reasoned analysis rather than through the mechanical application of traditional, often inaccurate assumptions about the proper role of men and women.
Id. at 725-26, 102 S.Ct. at 3337. In subsequent cases, the Court has reaffirmed the “firmly established principle” that the focus of the “substantial relationship” test is to insure that gender classifications are based on reason rather than stereotypes. See Heckler v. Mathews, 465 U.S. 728, 744, 750, 104 S.Ct. 1387, 1398, 1401, 79 L.Ed.2d 646 (1984). My colleagues suggest that intermediate scrutiny in the gender cases is *407about statistics, not stereotypes. But the cases themselves say the opposite.
The Supreme Court, for example, has upheld logical distinctions between men and women, even when the empirical evidence was open to question. In Michael M. v. Superior Court, 450 U.S. 464, 101 S.Ct. 1200, 67 L.Ed.2d 437 (1981), the Court upheld a state statute punishing men, but not women, who have sexual relations with minors. The Court found that the law did not “res[t] on ‘the baggage of sexual stereotypes,’ ” id. at 476, 101 S.Ct. at 1208 (cita: tion omitted), but conceded that the substantial relationship of the law to the goal of reducing teenage pregnancy was “at best an opaque one,” id. at 474 n. 10, 101 S.Ct. at 1207 n. 10. Rather than requiring the state to produce statistical evidence to support the nexus, however, the Court said: “[w]here such differing speculations as to the effect of a statute are plausible,” courts should defer to those “ ‘armed ... with the knowledge of the facts and circumstances concerning the passage and potential impact’ ” of the statute. Id. (citation omitted).
In Rostker v. Goldberg, 453 U.S. 57, 101 S.Ct. 2646, 69 L.Ed.2d 478 (1981), similarly, the Court upheld Congress’s decision to authorize draft registration for men but not women because it found that the gender classification “was not the ‘ “accidental byproduct of a traditional way of thinking about females.” ’ ” Id. at 74, 101 S.Ct. at 2656. The justices sharply criticized the lower court for relying on the testimony of military experts to reject Congress’s conclusion that conscripting women would not free up men for battle. “[The] District Court was quite wrong in undertaking an independent evaluation of this evidence, rather than adopting an appropriately deferential examination of Congress’s evaluation of that evidence.” Id. at 82-83, 101 S.Ct. at 2661.
When the Court has struck down gender classifications, it has done so because they rested on impermissible stereotypes, whether or not they were supported by statistical evidence. My colleagues misread Califano v. Westcott, 443 U.S. 76, 99 S.Ct. 2655, 61 L.Ed.2d 382 (1979), which they say struck down an AFDC program “because” no empirical evidence supported Congress’s judgment. Ante at 396. But the Westcott Court said clearly that the gender classification was not substantially related to its objective because it was, “rather, part of the „‘baggage of sexual stereotypes’ that presumes that the father has the ‘primary responsibility to provide a home and its essentials,’ while the mother is the ‘ “center of home and family life.” ’ ” Id. at 89, 99 S.Ct. at 2663 (citations omitted). The absence of statistics was significant only because it showed that the challenged provision rested on stereotypes, not reason.
The Court has also struck down gender classifications that are supported by statistics, when it has concluded that the classifications also rested on stereotypes. In Weinberger v. Wiesenfeld, 420 U.S. 636, 95 S.Ct. 1225, 43 L.Ed.2d 514 (1975), for example, the Court found unconstitutional a provision of the Social Security Act authorizing benefits to widows but not widowers. Congress justified the statute on the ground that men are more likely than women to be primary supporters of their spouses and children, and the Court found the notion “not entirely without empirical support.” Id. at 645, 95 S.Ct. at 1232. It invalidated the provision nevertheless, holding that the statistics did not “justify” — in a normative sense — “the denigration of the efforts of women who do work.” Id. My colleagues cite Wiesenfeld to support their claim that Congress’s predictive judgments must be sustained by “meaningful evidence.” Ante at' 393. But the Wiesenfeld Court did not strike down the provision because the statistical evidence was not meaningful. It struck down the provision because it rested on an “ ‘archaic and over-broad generalization’ ” that men’s wages, but not women’s wages, are vital to family support. Id. at 643, 95 S.Ct. at 1231 (citation omitted).
My colleagues also read too much into Craig v. Boren, 429 U.S. 190, 97 S.Ct. 451, 50 L.Ed.2d 397 (1976). In striking down an Oklahoma law that established different drinking ages for men and women, the *408Craig majority found the evidence supporting the age differential “not trivial in a statistical sense,” id. at 201, 97 S.Ct. at 459. And in concurring opinions, Justices Powell and Stevens conceded that “the statistics ... do tend generally to support the view that young men drive more, possibly are inclined to drink more, and — for various reasons — are involved in more accidents than young women.” Id. at 211, 97 S.Ct. at 464. (Powell, J., concurring); see also id. at 213, 97 S.Ct. at 465 (Stevens, J. concurring). In the Court’s view, however, the statistics could “hardly ... form the basis for employment of a gender line as a classifying device.” Id. at 201, 97 S.Ct. at 459. Although the Court did review the statistical evidence, it struck down the law only after pointing out that the different drinking ages were based on “social stereotypes.” Id. at 202 n. 14, 97 S.Ct. at 459 n. 14. “The question,” as Justice Stevens explained, “is whether the traffic safety justification put forward by .the State is sufficient to make an otherwise offensive, classification acceptable.” Id. at 213, 97 S.Ct. at 465 (Stevens, J., concurring) (emphasis added).
Later cases made clear what was implicit in Craig: courts must find that gender classifications rest on offensive stereotypes, rather than reasoned analysis, before striking them down. Nothing in my colleagues’ long footnote calls this simple proposition into question. Ante at 393-395 n. 3. I agree, of course, that Congress’s gender classifications cannot be upheld if they rest on premises that are not true. But I cannot agree that a logical premise is presumptively false until Congress commissions statistics to support it. Statistics, obviously, are one way to support a premise; but the cases make clear that Congress may also rely on logic — on “reasoned analysis,” “permissible reasoning” or “legitimate inferences.” The debate, in this case, however, is essentially semantic, since when the dust has settled my colleagues do not point to any stereotypes on which Congress relied.
Rather than supporting my colleagues’ view that gender classifications must be supported by statistical evidence, Craig v. Boren warns of the dangers of their approach:
It is unrealistic to expect either members of the judiciary or state officials to be well versed in the rigors of experimental or statistical technique. But this merely illustrates that proving broad sociological propositions by statistics is a dubious business, and one that inevitably is in tension with the normative philosophy that underlies the Equal Protection Clause.
Id. at 204, 97 S.Ct. at 460.
Neither Metro, nor Westcott, nor Craig, nor Wiesenfeld, in short, say what my colleagues say they say: that courts can overturn congressionally mandated gender classifications whenever they are not convinced by the statistical evidence before Congress. On the contrary, Metro, following a long line of gender cases, says that the purpose of intermediate scrutiny is to ensure that Congress’s judgments are based on reasoned analysis rather than archaic stereo-' types.
C. The Government’s Objectives
Before deciding whether reasoned analysis supports Congress’s judgment that the gender preference is ■ substantially related to its objectives, we must be clear about which objectives, precisely, Congress endorsed. “Implicit in the government’s judgment,” my colleagues suggest, is the assumption that women are more likely than white men to broadcast “women’s programming” or other targeted programming. ■ Ante at 395. In fact, Congress did not emphasize the idea of “women’s” or “minority” or any other “targeted” programming when it endorsed the FCC’s gender preference. Congress’s repeated statements on the subject show that it thought increased female ownership would promote the broader (and less controversial) goal of increasing programming diversity in general. My colleagues acknowledge the broader goal sporadically in their opinion, but in applying the intermediate scrutiny test, they focus only on the narrower goal of increasing women’s or minority programming. And they strike down the pref*409erence based on a single study that never considered programming for general audiences.
As early as 1982, Congress recognized that women are “significantly underrepresented in the ownership of telecommunications facilities” and that “the American public will benefit by having access to a wider diversity of information sources” if the role of underrepresented groups in broadcasting, such as women, is increased. H.R. No. 97-765 (Conf.Rep.), 97th Cong., 2d Sess. 43, reprinted in 1982 U.S.Code Cong. & Adm.News 2288-89; cf. Random Selection/Lottery Systems—Third Notice of Proposed Rule Making, 95 F.C.C.2d 432, 438 (1983) (noting that women own a majority of only 2.8% of all televisions stations). The Metro Court itself cited the report accompanying Congress’s 1987 directive that the FCC preference policies stay in place. The report of the Appropriations Committee explained: “The Congress has expressed its support for such policies in the past and has found that promoting diversity of ownership of broadcast properties satisfies important public policy goals. Diversity of ownership results in diversity of programming and improved service to minority and women audiences.” S.Rep. No. 1982, 100th Cong., 1st Sess. 76 (1987) (quoted in Metro, 110 S.Ct. at 3016) (emphasis added). The Committee recognized the continuity of congressional action in approving the minority and the gender preferences, noting that “[i]n approving a lottery system for the selection of certain broadcast licensees, Congress explicitly approved the use of preferences to promote minority and women ownership.” Id. at 76-77 (also quoted in Metro, 110 S.Ct. at 3016) (emphasis added).
Two years later, Senator Hollings, chair of the authorizing committee and the appropriations subcommittee for the FCC, described the repeated inclusion of the appropriations rider in subsequent appropriations bills “as an indication of Congress’ continuing support for these policies.” Minority Ownership of Broadcast Stations: Hearing Before the Subcomm. on Communications of the Senate Comm. on Commerce, 101st Cong., 1st Sess. at 3 (1989). At the same time, the Senate heard testimony that “a specific definition for the term ‘female programming’ ” was not necessary because “[t]he purpose of the female preference is to increase female ownership in order to promote viewpoint diversity. ” Id. at 78 (emphasis added).
In the 1987 appropriations rider itself, Congress endorsed the rationale of the FCC’s gender preference. See Continuing Appropriations Act for Fiscal Year 1988, Pub.L. 100-202, 101 Stat. 1329, 1329-31 (1987) (specifically approving preference policies as developed by the Commission). And the FCC had “ ‘recognized female involvement as contributing to potential diversity of programming and awarded merit in a comparative proceeding on that basis.’ ” Horne Industries, Inc., 94 F.C.C.2d 815, 823 n. 32 (1983) (quoting Minority Ownership of Broadcast Facilities, 69 F.C.C.2d 1591, 1593 n. 9 (1978)) (emphasis added); see also Third Notice of Proposed Rulemaking, 95 F.C.C.2d at 435-36; Mid-Florida Television Corp., 69 F.C.C.2d at 652.
All this is to say that the “governmental objective” in this case, as in Metro Broadcasting, is not merely to increase programming specifically targeted at women or minorities, but to increase programming diversity in general. “Congress and the FCC have selected the minority ownership policies primarily to promote programming diversity,” the Court said in Metro, and “the interest in enhancing broadcast diversity is, at the very least, an important governmental objective.” 110 S.Ct. at 3010 (emphasis added). Although Congress did not provide a precise definition of “programming diversity,” the- Metro Court offered the following definition: the expectation that “varying perspectives will be more fairly represented on the airwaves.” 110 S.Ct. at 3018. Programming diversity includes, the Court made clear, the “selection of topics for news coverage and the presentation of editorial viewpoint,” as well as avoiding stereotypes on general programming. Id. at 3017; see also Steele v. FCC, 770 F.2d 1192, 1208-09 (D.C.Cir.1985) (Wald, J., dissenting).
*410D. The Nexus
1. Reasoned Analysis
In addition to obscuring the government’s objectives, my colleagues ignore the Metro Court’s central conclusion that the minority preferences were “substantially related” to the objectives because “the reasoning employed by the Commission and Congress is permissible,” 110 S.Ct. at 3018. The reasoning upheld in Metro is hard to distinguish from the reasoning underlying the gender preference. Noting that “the nexus between ownership and programming ‘has been repeatedly recognized by both the Commission and the courts,’ ” 110 S.Ct. at 3015 (quoting H.R.Conf.Rep. No. 97-765, p. 40 (1982)), the Metro Court went on to endorse Congress’s logic by comparing it to the logic of the jury selection cases:
We have recognized, for example, that the fair cross-section requirement of the Sixth Amendment forbids the exclusion of groups on the basis of such characteristics as race and gender from a jury venire because “[wjithout that requirement, the State could draw up jury lists in such manner as to produce a pool of prospective jurors disproportionately ill disposed towards one or all classes of defendants, and thus more likely to yield petit juries with similar disposition.”
110 S.Ct. at 3018 (quoting Holland v. Illinois, 493 U.S. 474, 110 S.Ct. 803, 807, 107 L.Ed.2d 905 (1990)). “It is a small step from this logic,” the Metro Court said, “to the conclusion that including minorities in the electromagnetic spectrum will be more likely to produce a ‘fair cross section’ of diverse content.” Id, It is an even smaller step from this logic to the conclusion that including women in the broadcast spectrum will be more likely to produce a “fair cross section” of diverse content. The first jury venire cases, after all, concerned women, not minorities. As early as 1946, the Supreme Court suggested that female participation in the jury system is necessary if that process is to reflect “a cross section of the community.” Ballard v. United States 329 U.S. 187, 191, 67 S.Ct. 261, 263, 91 L.Ed. 181 (1946). The Court stressed that while neither men nor women think or act as a class,
[t]he truth is that the two sexes are not fungible; a community made up exclusively of one is different from a community composed of both; the subtle interplay of influence one on the other is among the imponderables. To insulate the courtroom from either may not in a given case make an iota of difference. Yet a flavor, a distinct quality is lost if either sex is excluded.
Id. at 193-94, 67 S.Ct. at 264. My colleagues have no patience for “subtle interplay” or “distinct quality,” insisting on numbers instead.
The jury cases, combined with Metro, make it clear that Congress’s assumptions about ownership by women and diversity of programming are permissible rather than impermissible. It is clearly “a legitimate inference for Congress and the Commission to draw that as more [women] gain ownership and policymaking roles in the media, varying perspectives will be more fairly represented on the airwaves.” Metro, 110 S.Ct. at 3018. (citation omitted).
As a matter of logic, in fact, the mild gender preference that my colleagues say is unconstitutional seem much more directly related to the goal of programming diversity than the minority distress sale policy upheld in Metro. As Justice O’Connor pointed out in her Metro dissent, “the distress sale policy provides preferences to minority owners who neither intend nor desire to manage the station in any respect.” 110 S.Ct. at 3041-42. The FCC awards a gender credit, by contrast, only where the prospective female owners will devote “substantial amounts of time on a daily basis” to the management of the station. Policy Statement on Comparative Broadcast Hearings, 1 F.C.C.2d 393, 395 (1965). In fact, under FCC rules, credit is awarded only to the extent that female owners will hold a policy-making position and the credit is proportionate to the extent of her involvement. New Continental Broadcasting Co., 96 F.C.C.2d 544, 546-47 (Rev.Bd.1983). In practice, this means that well-qualified women have been rejected *411for licenses because the FCC thinks they will not be sufficiently involved in daily programming decisions. See, e.g., Alexander S. Klein, Jr., 69 F.C.C.2d 2134, 2146-47 (Rev.Bd.1978). Given the fact that female owners who benefit from, the gender preference are more likely than minority distress sale owners actually to influence programming, I do not understand how my colleagues can conclude that minority ownership is logically related to the goal of programming diversity, but female ownership is not.
The Court in Metro also emphasized the fact that the distress sale policy placed only a “slight” burden on non-minorities. 110 S.Ct. at 3026-27. “In practice,” the Court pointed out, “distress sales have represented a tiny fraction — less than four tenths of one percent — of all broadcast sales since 1979.” Id. at 3027. In the case of gender preference, the burden on men is equally slight: out of 8,720 stations responding to an FCC survey, only 81 stations said they had benefited from the minority or women’s preference program, less than 1%. Congressional Research Service, Minority Broadcast Station Ownership and Broadcast Programming: Is There a Nexus? 41 (June 29,1988) [hereinafter CRS Report ]. And since the gender preference does not exclude men while the distress sales the Supreme Court approved do exclude “non-minorities,” the logic of Metro suggests that the burden is even slighter.
2. Archaic Stereotypes
In light of Metro and the gender cases, unless statistics disproved the link, I think that my colleagues could strike down the preference policy only if they held that Congress’s assumptions about female ownership and programming diversity are based on archaic stereotypes. They never suggest this, unsurprisingly, because Congress has not relied on stereotypes of any kind. It has not assumed that women “share some cohesive, collective viewpoint,” Metro, 110 S.Ct. at 3018, or that female journalists will approach stories about the federal budget, school prayer, voting rights, or foreign relations any differently than male journalists would. Cf. Roberts v. United States Jaycees, 468 U.S. 609, 628, 104 S.Ct. 3244, 3255, 82 L.Ed.2d 462 (1984). Nor has Congress endorsed the similarly controversial proposition that men and women think differently about most questions. (If it had, the constitutional issue might be closer: the Fourteenth Amendment does not enact Ms. Betty Friedan’s The Second Stage). It has merely assumed that some female programmers will choose to emphasize different subjects — breast cancer, say, or glass ceilings in the workplace — than male programmers will. The assumption strikes me as innocuous to the point of being obvious.
The familiar response is that the editorial perspectives of the Washington Post and the New York Times are hard to distinguish even though the Post was, until recently, published by a woman, and the Times is published by a man. See, e.g., Steele v. FCC, 770 F.2d 1192, 1199 (D.C.Cir.1985). But if constitutional decisions must turn on anecdotes, there are plenty to the contrary. For example, when Jennifer Lawson was appointed to be the Public Broadcasting Service’s first executive vice president for national programming, she announced that her main goal was to “make PBS programming a better reflection of the cultural, political, and sexual diversity of the U.S.” David Kling-hoffer, All-Powerful Programming Czar Promises the Era of the Multicultural, Washington Times, March 6, 1991, at El. And the Supreme Court in Metro stressed that “Congressional policy does not assume that in every case minority-ownership and management will lead to more minority-oriented programming or to the expression of a discrete ‘minority viewpoint’ on the airwaves ... Rather both Congress and the FCC maintain simply that expanded minority ownership of broadcast outlets will, in the aggregate result in greater broadcast diversity.” 110 S.Ct. at 3016 (emphasis added).
It would be hard, in any case, for my colleagues to maintain that gender preference will promote archaic stereotypes, since the FCC and Congress have concluded that *412gender preference will reduce archaic stereotypes. In the late 1970s, the Civil Rights Commission published two reports concluding that the broadcast media often portrayed women and minorities in stereotypical ways, and that the problem could be addressed by increasing the number of women and minority decisionmakers in radio and television. See Window Dressing on the Set: Women and Minorities in Television, A Report of the United States Commission on Civil Rights (1977) (cited by the Commission in its Statement of Policy on Minority Ownership, 68 F.C.C.2d at 980 n. 9); see also Window Dressing on the Set: an Update, A Report of the United States Commission on Civil Rights (1979). A 1990 report finds that the stereotypes persist, and also concludes that female programmers would reduce them. National Commission on Working Women, What’s Wrong With This Picture? The Status of Women on Screen and Behind the Camera in Entertainment TV (1990). Just as the Metro Court concluded that minority owned stations tend to “avoid racial and ethnic stereotypes in portraying minorities,” 110 S.Ct. at 3017, it was both reasonable and permissible for Congress and the Commission to conclude that female owned stations would avoid gender stereotypes.
In their briefs and at oral argument, lawyers for the FCC tried to summarize the logic that supports Congress’s conclusions, but my colleagues have belittled their efforts at every turn. “The Commission’s brief cites nothing,” the majority asserts, “that might support its predictive judgment that women owners will broadcast women’s or minority programming at any different rate than will men.” Ante at 395. But the Commission’s brief cites precisely the kind of reasoned analysis that satisfied the Supreme Court in Metro. See Appellant’s Brief at 34-36 (citing, among other things, jury venire analogy and Window Dressing on the Set Report). My colleagues reproduce a long exchange at oral argument in which, they say, the Commission’s lawyer confirmed that the record in this case lacks evidence of a substantial relationship. But, as the transcript makes clear, the questioning judge repeatedly cut off the lawyer’s attempt to explain the government’s reasoning, demanding he produce “evidence” or a “basis in fact” for the conclusion. To my mind, the excerpt helps to show how starkly my colleagues have misunderstood the nature of the inquiry into substantial relationship, and in their hunger for statistical evidence, have overlooked reasoned analysis.
E. The Statistics
I am reluctant to accept my colleagues’ invitation to scrutinize the statistics strictly, since Metro tells us that Congress’s empirical conclusions deserve “great weight.” 110 S.Ct. at 3011. But my colleagues’ treatment of the data is so unconvincing that their conclusions cannot be defended, even on their own terms. And even though the Metro Court found empirical support for Congress’s conclusions outside of the administrative record, my colleagues make no attempt to consider the host of other studies that provide more of the statistics they demand.
1. The CRS Report
My colleagues have decided to reject Congress’s findings after parsing a single report prepared by the Congressional Research Service, which merited a single footnote in Metro, 110 S.Ct. at 3017 n. 31, and which received no more attention in the opinion than any other study. Their odd treatment of the report is obvious from the beginning of their analysis. The title of the report asks “Is There a Nexus?” and, according to my colleagues, the report, “answer[s] its own question, at least with respect to women. The answer it gives is 'no.’ ” Ante at 396. But the answer it gives is plainly “yes.” “Based on th[e report’s] findings,” the cover page of the CRS study concludes, “there is a strong indication that minority and women station ownership result in a greater degree of minority programming.” CRS Report at cover page (emphasis added); see also id. at 44 (“While stations with women owners lag slightly behind those with minority owners in programming for minorities gen*413erally, a substantial percentage [of women-owned stations] programs for blacks and Hispanic audiences.”) (emphasis added). The Metro Court took the study’s conclusions at face value. Yet my colleagues ignore the central conclusion printed on the face of the study.
I concede that if reliable data conclusively disproved Congress’s judgments, it might suggest that Congress’s analysis was not reasoned. But the CRS study not only fails to disprove Congress’s judgments, it clearly supports them. The report concludes, for example, that 35% of stations owned primarily by women are likely to broadcast “women’s programming,” while just 28% of stations owned by “non-women” are likely to do so. CRS Report at 18 (Fig. 9A). That means that the female owned stations are 20% more likely to broadcast women’s programming than stations owned by men.
The data also reveal a statistical correlation between female ownership and minority programming. Twenty-six percent of stations with female owners broadcast programming targeted to Blacks, while 20% of stations owned by “non-Blacks” do. CRS Report at 14 (Fig. 5A). In other words, stations owned in part by women are 30% more likely to broadcast Black programming than stations owned by people who are not Black. The statistical correlation between women owners and programming directed at other minority groups is about the same. See CRS Report at 15-17
To put the percentages in perspective: In the 1952 election, a landslide, Dwight Eisenhower won 55% of the popular vote, to Adlai Stevenson’s 44% (about 33.5 million votes to about 27 million). The difference between the votes received by the two candidates? Only 25%.
A correlation of that size does not satisfy my colleagues. But they do not tell us why that link does not satisfy the Constitution. The Fifth Amendment does not identify the mystical point at which an empirical correlation becomes — to use my colleague’s word — “meaningful;” and judges have no basis, except their own policy preferences, for telling Congress it could not conclude that a 20% or 30% increase in women’s or minority programming was enough to justify the gender preference.
Besides downplaying the significance of the statistical correlation in the CRS report, my colleagues make three other observations about the data. But the data do not prove what my colleagues say.
* My colleagues stress that stations in which women own less than 50% of equity are as likely to broadcast women’s programming as stations in which women own more than 50%. Ante at 397. But that is the wrong comparison. Because the preference policy goes only to women who will also manage (and who will therefore be in a position to influence programming), the more relevant comparison is between women owners who are managers and women owners who are not. And the report makes clear that there is no correlation between owning more than 50% equity and holding a management position. Of the stations reporting that at least one of their owners was also a manager, 55.8% percent said those owners held less than 50% interest in the station, while just 44.2% reported those owners held more than a 50% interest in the station. CRS Report at 39.
The right question is whether a correlation exists, on the whole, between female owner-managers and diverse programming. The report demonstrates that correlation and likely understates it, because, according to the statistics, only 18% of the female owners in the survey were also managers. CRS Report at 40. The report also understates the correlation because it asked only about programming specifically targeted at women, minorities, children and senior citizens, and not at general audiences. CRS Report at 54.
* My colleagues focus on data from five large cities — a subsection of the report— which they claim call into question the link between female owners and women’s programming. Ante at 397. But they fail to mention that data from the same cities suggests a significant link between female ownership and minority programming. About 30% of stations with female owners target Black audiences, compared to only *414about 23% of stations without minority owners — a difference of 30%. CRS Report at 25. About 27% of stations with female owners target Hispanic audiences, compared to only about 20% of stations without minority owners — a difference of 35%. Id. The differences are similar for programming directed at Asian, Pacific Islander, Native American and Alaskan audiences. Id.
* My colleagues rely on the report’s finding that radio stations with at least one women owner use broadcasting formats in nearly the same order as stations owned by “non-minorities” (which includes men and women). Ante at 397. That means that stations with female owners are about as likely to use an “All News” or “Talk” or “Golden Oldies” format as stations without any minority ownership. But my colleagues’ focus on formats strikes me as irrelevant at best, and unsettling at worst. They cannot possibly expect stations owned by women to program soft, “feminine,” music, or to replace a “Country Western” format with “Adult Contemporary.” The point is that more female owner-managers will likely enhance the diversity of programming within the existing formats, and the study certainly does not disprove it.
2. Other Studies
Finally, if my colleagues want more statistics, there are plenty to support Congress’s position. When the Supreme Court in Metro noted that Congress’s conclusion about the nexus between programming and diversity “is corroborated by a host of empirical evidence,” 110 S.Ct. at 3017, it did not limit itself to studies that Congress had consulted, or that appeared in the administrative record or FCC brief. Apart from the CRS report, in fact, every study the Supreme Court cited (including unpublished doctoral dissertations and presentations at obscure symposia) came directly from amicus briefs supplied by organizations like the NAACP and the Congressional Black Caucus. This Court, too, could have called for amicus briefs on the statistical question (or accepted the brief that was offered) and our failure to do so looks less than sporting, since our decision seems to turn on the absence of them.
Nevertheless, statistical support for the gender preference is not hard to find. Despite the fact that Metro, for example, concerned race rather than gender, the Court received one amicus brief showing a clear statistical correlation between women programmers and women’s programming. An organization called American Women in Radio and Television surveyed the radio programs that had received National Commendation Awards in 1986 for presenting issues of particular interest to women or for presenting women in a “positive and realistic light.” Women represented 58.5% of the producers, 84.2% of the writers, and 92.9% of the reporter/hosts. Brief Amicus Curiae of American Women in Radio and Television, Inc. in Support of Respondents at 20-21, Metro Broadcasting v. FCC, 497 U.S. 547, 110 S.Ct. 2997, 111 L.Ed.2d 445 (1991). Women, by contrast, accounted for only 29.8% of broadcast managers or professionals in general. See FCC, Cable and Broadcast Industry Equal Employment Opportunity Trend Report (1986).
The statistics for producers are relevant, rather than “irrelevant,” ante at 396 n. 7, because the gender preference goes only to owners/managers who are also involved in programming decisions. The majority persistently ignores this fact, except to suggest that female owner/managers can influence programming only by hiring female producers and writers. Id. In fact, however, FCC cases make clear that the female owner/managers who receive preferences are themselves involved in programming decisions. In Coastal Broadcasting Partners, 6 FCC Rcd. (No. 14) 4242 (1991), for example, the Commission gave a preference to a female owner/manager who proposed to work full-time as Director of Programming and Public Affairs, a role in which she would “focus on program development, program acquisitions, ascertainment of needs and public affairs programs and announcements.” Id. at 4244-45. She would also serve on a programming committee responsible for “set[ting] programming policy,” and would “see to it that *415those policies are followed and [would] carry them out.” Id. at 4245.
If this case is appealed to the Supreme Court, I have no doubt that a host of amici will submit “a host of empirical evidence” to “corroborate” Congress’s judgment. In the meantime, a quick skim of the bestseller lists, the computer banks, and the rejected amicus brief, suggests the range of the studies they will have at their disposal. Susan Faludi’s Backlash, for example, devotes an entire chapter to the proposition that female programmers in the 1980s who tried to portray single or working women as independent professionals rather than as miserable careerists or as sex symbols, were challenged by male programmers at every turn. She gives a host of examples, and cites several studies to prove her point. Susan Faludi, Backlash, 140-168 (1991) (iciting National Commission on Working Women, “Women Out of View: An Analysis of Female Characters on 1987-88 TV Programs,” (1987); Davis, Portrayals of Women in Prime-Time Network Television: Some Demographic Characteristics, 23 Sex Roles 325-30 (1990)). Other sources include: two symposia co-sponsored by the FCC and American Women in Radio and Television, Inc., “Women in the Telecommunications Marketplace” and “The Woman Entrepreneur”; National Commission on Working Women, What’s Wrong With This Picture? The Status of Women on Screen and Behind the Camera in Entertainment TV (1990); Herbert, Study Charges Sexism in Women’s Sports Coverage, L.A. Times, Aug 30, 1990, Part F, p. 2, col. 3 (describing a study prepared by professors at the University of Wisconsin and the University of Southern California); P. Koza, Kiddie TV Study: “One Smurfette Amid a Host of Smurfs ”, United Press International, July 14, 1982 (describing a study by Boston University professors finding that programs targeted at children presented outdated stereotypes of gender roles). Again, I hardly think these kinds of studies are necessary to decide the constitutional question, but since my colleagues do, there are plenty to go around.
I must also dissent, finally, from my colleagues’ paraphrased descriptions of my positions. I do not suggest, as they claim, “that the Supreme Court already has decided that the Commission’s sex-based policy passes constitutional muster,” ante at 386 n. 1. I do not suggest “that deference means not just limited factual review, but none at all,” ante at 392 n. 2. I do not suggest “that a sex based classification might survive intermediate scrutiny even if it rests upon unsupported generalizations about men or women as a group,” ante at 393 n. 3. And I do not suggest “that a relevant generalization is presumed true (even if unsupported) unless proved otherwise,” id. I welcome, of course, a vigorous debate. But I think we should debate the more modest propositions actually contained in this dissent.
“The blind use of statistics,” Chief Justice Rehnquist has noted, “cannot be permitted to undermine the policies of Congress or erode our decisions on substantive law.” Procter & Gamble Mfg. Co. v. Fisher, 449 U.S. 1115, 1118, 101 S.Ct. 929, 930, 66 L.Ed.2d 845 (1981) (Rehnquist, J., dissenting from denial of cert.). As a matter of policy, the decision to give a mild preference to women in the assignment of broadcast licenses is open to debate. But as a matter of law, the constitutionality of this affirmative action program is clear — at least until the Supreme Court overturns Metro and a long line of gender cases. One of the most unsettling trends in appellate jurisprudence is the tendency of judges who are devoted to the original intention of the framers of the Constitution to ignore the original intentions of elected representatives in Congress. Today my colleagues thwart not only the intentions of Congress and the Executive, but also the intentions of the Supreme Court. I think that the third branch has no business telling the first branch how to make national policy, or what policy to make, and I dissent.