Minority Television Project, Inc. v. Federal Communications Commission

OPINION

BEA, Circuit Judge:

A federal statute, 47 U.S.C. § 399b prohibits public broadcast radio and television stations1 from transmitting over the public airways: 1) advertisements for goods and services on behalf of for-profit entities, 2) advertisements regarding issues of public importance or interest (“public issues”), and 3) political advertisements. 47 U.S.C. § 399b(a). The statute is therefore a content-based ban on speech: public broadcasters may transmit many types of speech, but, unlike most other stations, they may not transmit those three classes of advertising messages. Plaintiff-Appellant Minority Television Project, a public broadcaster, contends that this ban violates the First Amendment. Applying intermediate scrutiny, we uphold the ban on the transmission of advertisements for goods and services by for-profit entities, but we strike down as unconstitutional the ban on public issue and political advertisements.2

I. Facts and Prior Proceedings

Appellant Minority Television Project is a nonprofit California corporation which operates the San Francisco television station KMTP-TV. KMTP-TV focuses on what it describes as “multicultural programming,” and it airs a wide variety of non-English language television shows. KMTP-TV is licensed as a public broadcast station by the FCC, which sets aside certain broadcast frequencies for public radio and television stations which transmit educational programming. Unlike commercial stations, public broadcast stations are expected not to rely on paid advertising, but on federal and state subsidies, individual donors, special events, foundation grants, and corporate contributions. See generally 47 C.F.R. § 73.621. KMTP-TV is one of the few public broadcast stations in the United States which does not receive funding from the Corporation for Public Broadcasting (a private, non-profit corporation created by Congress in 1967 to invest in educational programming on public broadcast stations). See 47 U.S.C. § 396(g)(2)(B).

However, because of its status as a public broadcast station, Minority is nonetheless subject to 47 U.S.C. § 399b, which prohibits public broadcast stations from transmitting any “advertisements.” Under § 399b, an “advertisement” is defined as: (1) a paid promotional message from a for-profit entity, (2) a paid message on any *873matter of public importance, or (8) a paid message in support of a political candidate. The relevant portion of the statute reads:

(a) “Advertisement” defined. For purposes of this section, the term “advertisement” means any message or other programming material which is broadcast or otherwise transmitted in exchange for any remuneration, and which is intended—
(1) to promote any service, facility, or product offered by any person who is engaged in such offering for profit;
(2) to express the views of any person with respect to any matter of public importance or interest; or
(8) to support or oppose any candidate for political office.

47 U.S.C. § 399b.

On August 9, 2002, pursuant to a complaint filed by another broadcaster, the FCC determined that Minority had violated § 399b approximately 1,900 times between the years 1999 and 2002. 7 FCC Red 15646 (2003). The FCC found that Minority had “willfully and repeatedly” violated § 399b when it broadcast paid promotional messages on KMTP-TV from for-profit corporations such as State Farm, Chevrolet, and U-Tron Computers.3 7 FCC Red 15646 (2003). Minority was fined $10,000 by the FCC. 7 FCC Red 15646 (2003). Minority paid its $10,000 fine, but also filed a complaint in federal district court for the Northern District of California in which it sought both reimbursement of the $10,000 and declaratory relief. In relevant part, Minority alleged that § 399b violates the First Amendment because its restriction on advertising was not narrowly tailored to the government’s interest in preserving the educational programs on public broadcast stations. Minority alleges that it has declined to broadcast public issue and political advertisements and would do so but for the fear of FCC fines and forfeitures similar to those previously imposed and paid. Minority contended § 399b is an unconstitutional content-based restriction on speech, because it bans all paid public issue and political speech while permitting paid promotional messages by non-profits.

In response, the government contended § 399b’s restrictions on advertising are necessary to preserve the educational nature of public broadcast programming. The government contended that because advertisers naturally wish to reach the largest possible audience, advertisers are more likely to buy commercials on television programs with high numbers of viewers. Thus, the government contended, advertiser-supported television and radio stations have an incentive to broadcast programs with mass-market appeal. According to the government, if public television and radio stations became financially dependant on advertising, such stations would replace their niche educational programs with more popular programs which have greater mass-market appeal, thus en*874dangering the broadcast of the educational programs for which public broadcast stations exist. See supra p. 872, n. 1.

After discovery, Minority and the FCC filed cross-motions for summary judgment on Minority’s facial challenges to § 399b. The district court applied intermediate scrutiny to § 399b, and determined the prohibitions on advertising were narrowly tailored to meet the substantial government interest in maintaining educational programming on public stations.

Minority timely appealed the district court’s grant of summary judgment. We have jurisdiction under 28 U.S.C. § 1291. We review de novo the district court’s grant of summary judgment. United States v. Alisal Water Corp., 431 F.3d 643, 651 (9th Cir.2005).

II. Determining What Level of Scrutiny Applies

As in all First Amendment cases, we must first determine the correct standard of scrutiny to apply to the challenged statute. Because First Amendment doctrine and the media landscape have changed substantially in recent years, this is no simple matter.

A. The Nature of the Restriction

At the threshold of the inquiry, we must determine whether this restriction is content based or content neutral. We have previously held that “whether a statute is content neutral or content based is something that can be determined on the face of it; if the statute describes speech by content, then it is content based.” G.K. Ltd. Travel v. City of Lake Oswego, 436 F.3d 1064, 1071 (9th Cir.2006). Here, § 399b imposes clear content-based restrictions on the station’s speech.

First, Minority may broadcast a wide variety of content for a wide variety of purposes, but the station may not air the three types of advertisements banned by § 399b. That is a content-based restriction, since it plainly restricts Minority’s speech based on the speech’s content.

Second, and equally important, § 399b discriminates within the class of speech it defines as “advertisements.” Public broadcast stations may not broadcast most types of advertising speech, but these stations may broadcast paid promotional messages for products and services of nonprofit corporations. See 47 U.S.C. § 399b(a)(l). For example: the record shows that the FCC allowed a public broadcast station in Indiana to broadcast a paid message which promoted Planned Parenthood’s “confidential, affordable reproductive health services” because Planned Parenthood “is a non-profit organization.” Nonetheless, a public broadcast station may not broadcast a paid message “to express the views of any person with respect to any matter of public importance” or “to support or oppose any candidate for political office” regardless whether the sponsoring entity is an individual, a nonprofit corporation, or a for-profit corporation. 47 U.S.C. § 399b(a)(2) and (a)(3). Thus, had Planned Parenthood sought to air a paid message in support of Presidential candidates who favored abortion rights, or sought to broadcast an “issue ad” on the importance of sex education in schools, a public broadcast station would have been prohibited from airing it under § 399b(a)(2) and (a)(3). Indeed, in its letter to Planned Parenthood, the FCC specifically stated that its proposed message did not violate § 399b because it did not support any candidate for political office, nor express any views with respect to a matter of public importance. But, as shown, Planned Parenthood could advertise to promote itself.

Thus, § 399b prohibits a public broadcast station from broadcasting any adver*875tisement which expresses views on a matter of public importance or on behalf of a political candidate regardless who sponsored the message—Planned Parenthood (a nonprofit), Apple, Inc. (a for-profit), or a committee to re-elect President Obama (a political group). But it allows a public broadcast station to transmit a paid promotional message from a nonprofit, so long as that message does not express views on public issues or political candidates. That is a further content-based restriction on speech, and this restriction in particular burdens speech on issues of public importance and political speech.

B. Intermediate Scrutiny Applies

Having identified § 399b’s speech restrictions as content-based on two levels, we must determine what level of scrutiny to apply in our analysis. Because government regulation of content is one of the primary evils contemplated by the First Amendment, content-based restrictions are strongly disfavored and are often subject to strict scrutiny. Indeed, in the typical case, “[c]ontent-based regulations are presumptively invalid.” R.A.V. v. City of St. Paul, 505 U.S. 377, 382, 112 S.Ct. 2538,120 L.Ed.2d 305 (1992).

Further, the bans on public issue and political advertisements appear at first glance to be especially strong candidates for strict judicial scrutiny because political speech is “entitled to the most exacting degree of First Amendment protection.” League of Women Voters, 468 U.S. 364 at 375, 104 S.Ct. 3106. Under strict scrutiny, the government would be required to “prove that the restriction furthers a compelling interest and is narrowly tailored to achieve that interest.” Citizens United v. FEC, — U.S.-, 130 S.Ct. 876, 898, 175 L.Ed.2d 753 (2010).

But this is not the typical case, because these particular content-based restrictions on speech apply to broadcasters. For decades now, the Supreme Court has held that content-based speech restrictions that apply to broadcasters are subject to a less demanding form of judicial scrutiny than similar restrictions that arise in other media contexts. See FCC v. Pacifica Found., 438 U.S. 726, 98 S.Ct. 3026, 57 L.Ed.2d 1073 (1978). Indeed, in FCC v. League of Women Voters, 468 U.S. 364, 104 S.Ct. 3106, 82 L.Ed.2d 278 (1984), the Court held that this intermediate level of scrutiny applies to regulations governing public broadcasters in particular. Id. at 376-77, 104 S.Ct. 3106.

Specifically, in League of Women Voters, the Court observed that “because broadcast regulation involves unique considerations, our cases have not followed precisely the same approach that we have applied to other media and have never gone so far as to demand that such regulations serve ‘compelling’ governmental interests.” League of Women Voters, 468 U.S. at 376, 104 S.Ct. 3106. Pursuant to the Commerce Clause, Congress regulates the broadcast spectrum—which is a “scarce and valuable national resource”— to ensure that stations which broadcast on those frequencies “satisfy the public interest, convenience, and necessity.”4 Id. *876Thus, when Congress acts pursuant to its regulation of the broadcast spectrum, it does not operate under the same First Amendment standards that apply to regulation of other forms of media. Instead, in light of the history behind Congressional regulation of the broadcast spectrum, the Supreme Court has held that laws enacted pursuant to Congressional broadcast regulation—even those which, as here, impose a content-based restriction on core political speech—are subject to intermediate First Amendment scrutiny. Under intermediate scrutiny, the government must prove a challenged statute is “narrowly tailored to further a substantial governmental interest.” Id. at 380,104 S.Ct. 3106.

Despite the Court’s pronouncement in League of Women Voters, which was a public broadcasting case, Minority urges us to apply strict scrutiny for two different reasons. First, citing a concurring opinion in FCC v. Fox Television Stations, Inc., 556 U.S. 502, 129 S.Ct. 1800, 173 L.Ed.2d 738 (2009), which questioned the continuing validity of the broadcast regulation precedents on which League of Women Voters relied, Minority contends that new technologies such as cable and the Internet have undermined the core “spectrum scarcity” rationale of broadcast regulation cases. Id. at 1821 (Thomas, J., concurring). Under this theory, because “traditional broadcast television and radio are no longer the ‘uniquely pervasive’ media forms they once were,” id., courts should no longer treat broadcast restrictions any differently from other restrictions on speech.

Minority is surely correct that much has changed in the media landscape since the Supreme Court, in the 1970s, first adopted a standard that treats broadcasters differently under the First Amendment. Indeed, it is possible that the Supreme Court itself may soon declare that the era of a special broadcast exemption from strict scrutiny is over. After briefing and argument in this case, the Supreme Court heard argument in a case in which a coalition of the nation’s major broadcasters have asked the Court to overrule Pacifica and its progeny and “announce firmly and finally that the time for treating broadcast speech differently than all other communications is over.” Br. of Respondents Fox Television Stations et al. in FCC v. Fox Television Stations, No. 10-1293, at 1.

But that case has not yet been decided. Thus, just as golfers must play the ball as it lies, so too we must apply the law of broadcast regulation as it stands today. A majority of the Supreme Court has not overruled Pacifica, League of Women Voters, and related cases. Intermediate broadcast scrutiny remains in vigor, and it governs this ease.

Second, pointing to the bans on public issue and political advertising in particular, Minority contends that § 399b should be subject to strict scrutiny in the wake of Citizens United v. FEC, — U.S.-, 130 S.Ct. 876, 886, 175 L.Ed.2d 753 (2010). Citizens United applied strict scrutiny to 2 U.S.C. § 441b, which prohibited corpora*877tions from engaging in “electioneering communications”5 within 30 days of a primary or 60 days of a general election, and held that the statute violated the First Amendment. Id. at 890. “The only reasonable conclusion that can be drawn from Citizens United,” contends Minority, “is that any restriction or prohibition of political speech on broadcast radio or television is subject to strict scrutiny.”

We disagree. Citizens United was not a broadcast regulation case, so the Court there had no reason to revisit League of Women Voters and related cases. Instead, the Court relied on its previous application of strict scrutiny in cases which challenged the constitutionality of restrictions on campaign expenditures, not broadcast spectrum regulation. See id. at 899 (citing FEC v. Wise. Right to Life, Inc., 551 U.S. 449, 464, 127 S.Ct. 2652, 168 L.Ed.2d 329 (2007), a previous case which analyzed § 441b, for the proposition that “laws that burden political speech are subject to strict scrutiny”). Thus, in Citizens United, the Court applied strict scrutiny to § 441b because that statute dealt with regulations on campaign expenditures generally. See, e.g., id. at 897 (listing, as acts that would be outlawed under § 441b, corporations running advertisements, publishing books, or creating websites). Citizens United in no way dealt with the “unique considerations” inherent in Congress’s regulation of the broadcast spectrum.

Moreover, Citizens United expressly overruled two of the Court’s prior decisions: Austin v. Michigan Chamber of Commerce, 494 U.S. 652, 110 S.Ct. 1391, 108 L.Ed.2d 652 (1990), which permitted a ban on speech based on corporate identity, and McConnell v. FEC, 540 U.S. 93, 124 S.Ct. 619, 157 L.Ed.2d 491 (2003), which relied on Austin to uphold a facial challenge to § 441b. Neither case involved regulation of public broadcasting. Thus, it is not surprising that the Court in Citizens United did not once mention League of Women Voters; it was neither overruled nor distinguished away. That is fatal to Minority’s contention, because in League of Women Voters, the Supreme Court specifically rejected the contention that content-based laws which burden political speech and are enacted pursuant to the broadcast spectrum require the application of strict judicial scrutiny. League of Women Voters, 468 U.S. at 376, 104 S.Ct. 3106.

We therefore apply intermediate scrutiny to the restrictions. As explained below, we keep in mind as we apply that standard that public issue and political speech in particular is at the very core of the First Amendment’s protection. We also must be mindful that the narrow tailoring prong of the intermediate scrutiny standard itself has undergone additional elaboration by the Supreme Court since League of Women Voters was decided in 1984. It is the details of that standard to which we now turn.

C. The Requirements of Intermediate Scrutiny

1. League of Women Voters

In determining what the application of intermediate scrutiny entails, League of Women Voters is our starting point. In that case, the Supreme Court considered a First Amendment challenge to a statute which forbade any public broadcasting station from transmitting editorials on “con*878troversial issues of public importance” if that station had received a grant from the Corporation for Public Broadcasting. The Court in League of Women Voters held that the ban on station editorials was “defined solely on the basis of the content of the suppressed speech.” Id. at 383, 104 S.Ct. 3106. “In order to determine whether a particular statement by station management constitutes an ‘editorial,’ ” the Court reasoned, “enforcement authorities must necessarily examine the content of the message that is conveyed to determine whether the views expressed concern ‘controversial issues of public importance.’ ” Id. Although the Court held that the statute at issue in League of Women Voters was viewpoint-neutral—i.e., it prohibited station editorials on all sides of an issue— the Court held the “First Amendment’s hostility to content-based regulation extends not only to restrictions on particular viewpoints, but also to prohibition of public discussion on an entire topic”; thus, the Court held the statute was a content-based restriction on speech. Id. at 384, 104 S.Ct. 3106.

In light of the First Amendment’s hostility towards content-based restrictions on speech touching on controversial issues of public importance on the one hand, and deference afforded to Congress’s regulation of the broadcast spectrum on the other, the Court in League of Women Voters held that a robust form of intermediate scrutiny applies to content-based restrictions on broadcast speech which burden political expression. Under the standard applied in League of Women Voters, a restriction on speech will be upheld only if the government proves “the restriction is narrowly tailored to further a substantial governmental interest.” Id. at 380, 104 S.Ct. 3106. The Court in League of Women Voters—while declining to require the government to prove a “compelling” interest under the more stringent strict scrutiny test—required judicial “wariness” within the standard it described. The Court did so because the statute at issue in that case restricted editorials, which are “precisely the form of speech which the Framers of the Bill of Rights were most anxious to protect—speech that is indispensable to the discovery and spread of political truth.” Id. at 383, 104 S.Ct. 3106. The Court said that it “must be particularly wary in assessing [the statute] to determine whether it reflects an impermissible attempt to allow the government to control ... the search for political truth.” Id. at 384,104 S.Ct. 3106 (emphasis added).

The Court held that the restriction there was not narrowly tailored. Id. at 395, 104 S.Ct. 3106. Rather, a “broad ban on all editorializing by every station that receives [Corporation for Public Broadcasting] funds far exceeds what is necessary to protect against the risk of governmental interference or to prevent the public from assuming that editorials by public broadcasting stations represent the official view of government.” Id. Although the Court recognized that “the Government certainly has a substantial interest in ensuring that the audiences of noncommercial stations will not be led to think that the broadcaster’s editorials reflect the official view of the Government,” the Court said that “this interest can be fully satisfied by less restrictive means that are readily available.” Id. For example, the Court stated that Congress could “simply require public broadcasting stations to broadcast a disclaimer every time they air editorials which would state that the editorial ... does not in any way represent the views of the Federal Government or any of the station’s other sources of funding.” Id. Thus, the Court held the ban on station editorials unconstitutional and affirmed the grant of summary judgment to the League *879of Women Voters. Id. at 402, 104 S.Ct. 3106.

For the purposes of application of the proper level of scrutiny, the statute at issue in this case is similar to the challenged statute in League of Women Voters,6 Section 399b makes content-based distinctions which, by their terms, burden speech in a similar manner to the provision at issue in League of Women Voters. Like the statute in League of Women Voters, § 399b was enacted pursuant to Congress’s regulation of public broadcast stations—stations which were explicitly set aside for educational programming. Moreover, subsections 399b(a)(2) and (a)(3) share the additional similarity with the provision at issue in League of Women Voters that the provisions burden public issue and core political speech.

2. Subsequent Elaboration: The Turner Cases and Discovery Network

We are conscious, of course, that First Amendment doctrine has not been stagnant in the nearly thirty years since League of Women Voters was decided. We must also consider further elaborations of the narrow tailoring inquiry under intermediate scrutiny. We thus take guidance in particular from two cases together known as “the Turner cases,” as well from select commercial speech cases that applied intermediate scrutiny, especially Cincinnati v. Discovery Network, 507 U.S. 410, 113 S.Ct. 1505, 123 L.Ed.2d 99 (1993).

In the mid-1990s, the Supreme Court had occasion comprehensively to describe intermediate broadcast scrutiny, albeit in a slightly different context from that here, in a pair of cases known as the Turner cases. In Turner Broadcasting System v. FCC, 512 U.S. 622, 114 S.Ct. 2445, 129 L.Ed.2d 497 (1994) (“Turner /”), the Supreme Court reversed, for further factfinding, the district court’s grant of summary judgment to the FCC on a First Amendment challenge to a statute which compelled cable companies to carry local broadcast stations. 512 U.S. at 667, 114 S.Ct. 2445. The Court held that there was not enough evidence in the record to determine whether local broadcast stations would go out of business if cable companies were not required by law to carry local broadcast stations. Id. at 668, 114 S.Ct. 2445. The Court revisited the dispute after additional discovery in district court in Turner Broadcasting System v. FCC, 520 U.S. 180, 117 S.Ct. 1174, 137 L.Ed.2d 369 (1997) (“Turner II”). In Turner II, the Court upheld the district court’s decision on remand in favor of the FCC because the additional record evidence supported Congress’s determinations. Id. at 224, 117 S.Ct. 1174.

As relevant here, the guiding principle of narrow tailoring under intermediate scrutiny is that the government must “demonstrate that the recited harms” to the substantial governmental interest “are real, not merely conjectural, and that the regulation will in fact alleviate those harms in a direct and material way.” Turner I, 512 U.S. at 664-65, 114 S.Ct. 2445. Furthermore, although a statute is “not invalid simply because there is some imaginable alternative that might be less burdensome on speech,” Turner II, 520 U.S. at 217, 117 S.Ct. 1174, the government must prove that the statute does not “burden substantially more speech than is necessary to further the government’s legitimate interests.” Turner I, 512 U.S. at 665, 114 S.Ct. 2445 (internal quotations *880omitted). Importantly, the government must prove both the reality of the recited harms and that the statute does not burden more speech than necessary “by substantial evidence.” Turner II, 520 U.S. at 211, 117 S.Ct. 1174. “Substantial evidence” must include “substantial evidence in the record before Congress” at the time of the statute’s enaction.7 Id.

Additional instruction on what narrow tailoring requires comes from Cincinnati v. Discovery Network, 507 U.S. 410, 113 S.Ct. 1505, 123 L.Ed.2d 99 (1993). In Discovery Network, the Court was faced with a content-based restriction on speech: a city ordinance banned sidewalk news-racks which distributed “commercial handbills,” but not newsracks which distributed “newspapers.” Id. at 429, 113 S.Ct. 1505. A group of publishers of commercial handbills challenged the statute as an impermissible content-based restriction on speech prohibited by the First Amendment. Id. at 412, 113 S.Ct. 1505. The city defended the ordinance by contending it furthered its “legitimate interest in ensuring safe streets and regulating visual blight.” Id. at 415, 113 S.Ct. 1505. Cincinnati contended newsracks in general undermined safety and esthetics in the public right of way; thus, the ban on news-racks which contained a certain type of content was justified because it necessarily reduced the total number of newsracks on sidewalks. Id. at 415,113 S.Ct. 1505.

The Supreme Court held the statute unconstitutional, because the “selective and categorical” content-based ban on news-racks containing handbills was not narrowly tailored to the city’s purported interest. Id. at 417, 113 S.Ct. 1505. Although the city’s “desire to limit the total number of newsracks is justified by its interests in safety and esthetics,” the statute was “unrelated to any distinction between ‘commercial handbills’ and ‘newspapers,’ ” and thus was not narrowly tailored. Id. at 429-30, 113 S.Ct. 1505 (emphasis added, some internal quotation marks omitted). The Court said:

The city has asserted an interest in esthetics, but respondent publishers’ news-racks are no greater an eyesore than the newsracks permitted to remain on Cincinnati’s sidewalks. Each newsrack, whether containing “newspapers” or “commercial handbills,” is equally unattractive .... [T]he city’s primary concern, as argued to us, is with the aggregate number of newsracks on the streets. On that score, however, all newsracks, regardless whether they contain commercial or noncommercial publications, are equally at fault.

Id. at 425-26,113 S.Ct. 1505.

Thus, the Court held the newsrack ordinance was not narrowly tailored, because there was no proof that newsracks containing handbills (banned) threatened the governmental interests in esthetics and safety to a greater degree than news-racks containing newspapers (permitted). Therefore, the Court held the costs and benefits of the statute had not been “carefully calculated” to meet the substantial governmental interest. See id. at 416 n. 12, 113 S.Ct. 1505. Notably, the ordinance did not regulate the number of newsracks *881permitted on the city’s sidewalks, regardless their content.8

D. Summary

Synthesizing three decades of First Amendment cases, then, we take heed of two key principles. First, for us to sustain any content-based restriction, the government must prove both the reality of the recited harms and that the statute does not burden more speech than necessary “by substantial evidence.” Turner II, 520 U.S. at 211, 117 S.Ct. 1174. “Substantial evidence” must include “substantial evidence in the record before Congress” at the time of the statute’s enaction. Id. Second, when Congress enacts a “selective and categorical” ban on speech, as here, the government must prove that the speech banned by a statute poses a greater threat to the government’s purported interest than the speech permitted by the statute. Discovery Network, 507 U.S. at 425,113 S.Ct. 1505.

III. Analysis of § 399b

Our final step is to determine whether the government has carried its burden to prove that § 399b passes intermediate First Amendment scrutiny. Intermediate scrutiny requires the government to prove a challenged statute is narrowly tailored to a substantial government interest. League of Women Voters, 468 U.S. at 380, 104 S.Ct. 3106. We hold that 399b(a)(l), which prohibits advertising by for-profit entities for their goods and services, meets this standard. However, 399b(a)(2) and (a)(3), which prohibit public issue and political advertising, do not.

A. The Subsections are Severable

We must decide at what level of generality to undertake the analysis. That is: do all provisions of § 399b stand or fall together, or should we analyze separately subsections 399b(a)(l), (a)(2), and (a)(3)? We conclude that it is appropriate to sever the provisions and analyze them separately.

Although neither § 399b nor the Public Broadcasting Amendments Act of 1981 contains an explicit severability clause, we are confident that we nevertheless may hold one section constitutional without so holding as to the others. Statutes are presumptively severable and “a court should refrain from invalidating more of the statute than is necessary.” Alaska Airlines, Inc. v. Brock, 480 U.S. *882678, 684, 107 S.Ct. 1476, 94 L.Ed.2d 661 (1987). We may sever the statute “[u]n-less it is evident that the Legislature would not have enacted those provisions which are within its power, independently of that which is not.” Id.

Here, we see no reason to think that Congress would not have enacted the ban on promotional advertising by for-profit entities if it could not also have enacted the bans on political or public issue advertising. Moreover, neither party contends the statute is not severable, and the district court analyzed the provisions separately. There is thus no warrant for departing from the general presumption of severability in this case.

Our conclusion is further supported by Congress’s decision to ban advertising selectively. Since Congress determined that these three classes of advertising should be banned, then Congress must “demonstrate that the recited harms” to the substantial governmental interest of each class of advertising “are real, not merely conjectural, and that the regulation will in fact alleviate those harms in a direct and material way.” Turner I, 512 U.S. at 664-65, 114 S.Ct. 2445. It is that inquiry to which we now turn.

B. The Government’s Overall Interest

The government asserts the same interest is furthered by all three of § 399b’s restrictions on advertising: that Congress may ban advertising on public television stations to ensure that high quality educational and noncommercial programming is broadcast on the public airwaves. The government contends that if public broadcast stations were permitted to transmit paid commercial, public issue, and political advertisements, public broadcast stations would attempt to attract advertising dollars by replacing niche educational programming with programming of greater mass-market appeal. In turn, the distinction between public broadcast and commercial stations would be blurred—and the breadth of quality educational and other noncommercial programming on public broadcast stations would be reduced.

As an initial matter, we hold the government has a substantial interest in ensuring high-quality educational programing on public broadcast stations—a conclusion Minority does not dispute. Even though cable, satellite, and the Internet have changed the nature of television and radio, the broadcast spectrum remains a finite national resource. Congress set aside broadcast frequencies for public stations to ensure Americans would have access to niche programming such as public affairs shows and educational programs for children. See 41 F.C.C. at 166, ¶57 (FCC reserves broadcast frequencies for public broadcast stations because they offer “programming of an entirely different character from that available on most commercial stations”).

Moreover, the government has submitted unrebutted evidence that public broadcast stations do broadcast substantially different types of programs than do commercial stations. For example, the Government Accountability Office has determined that 16 percent of all program hours broadcast by public television stations are devoted to educational children’s programming. By contrast, commercial broadcasters devote less than 2 percent of their program hours to educational or informational children’s programming. According to a Senate report submitted by the government, public television is “the primary source of educational children’s programming in the United States.” Children’s Television Act of 1990, S.Rep. No. 101-66 at 3, reprinted in 1990 U.S.C.C.A.N. 1628, 1633. Public broad*883cast stations regularly broadcast renowned children’s shows such as “Sesame Street,” “Mr. Roger’s Neighborhood,” and “Reading Rainbow,” which attempt to teach children to read and to do sums. Id. at 23-36, 1990 U.S.C.C.A.N. at 1631-33. Again, Minority does not dispute that the government’s interest in maintaining public broadcast stations’ niche programming is “substantial.” Instead, Minority contends that § 399b is not narrowly tailored to the asserted government interest.

C. Whether The Restrictions Are Narrowly Tailored

We thus turn to the “narrowly tailored” prong of the intermediate scrutiny test. Under the Turner cases, the government must “demonstrate that the recited harms” to the substantial governmental interest “are real, not merely conjectural, and that the regulation will in fact alleviate those harms in a direct and material way.” Turner I, 512 U.S. at 664-65, 114 S.Ct. 2445. The government must prove both the reality of the recited harms and that the statute does not burden more speech than necessary “by substantial evidence.” Turner II, 520 U.S. at 211, 117 S.Ct. 1174. “Substantial evidence” must include “substantial evidence in the record before Congress” at the time of the statute’s enaction. Id. Moreover, when Congress enacts a “selective and categorical” ban on speech, as here, the government must prove that the speech banned by a statute poses a greater threat to the government’s purported interest than the speech permitted by the statute. See Discovery Network, 507 U.S. at 425, 429, 113 S.Ct. 1505.

1. Subsection 399b(a)(l)

We first turn to 399b(a)(l), the restriction on paid advertisements for goods and services on behalf of for-profit corporations. As in the Turner cases, we hold that there was “substantial evidence”—including “substantial evidence in the record before Congress”—to support Congress’s conclusions that: a) the harm posed by advertising by for-profit entities on public broadcast stations was “real, not merely conjectural,” and b) banning advertising by for-profit entities does not “burden substantially more speech than necessary to further the government’s legitimate interest.” Turner I, 512 U.S. at 664-65, 114 S.Ct. 2445.

Prior to the enactment of § 399b in 1981, Congress considered eliminating all advertising restrictions on public broadcast stations. Congress’s decision to continue regulation of promotional advertising—at least by for-profit entities—on public broadcast stations was supported by substantial evidence presented to Congress that advertising would harm the educational mission of public broadcast stations. Congress heard testimony by a senior vice president from National Public Radio (“NPR”), who testified that an internal study conducted by NPR had shown that “on-air advertising” would hold “little or no promise for public radio,” because public radio stations broadcast precisely “what commercial stations choose not to broadcast because of insubstantial income potential.” (emphasis added). Congress also heard testimony from John C. DeWitt, from the American Foundation for the Blind, who testified that he was concerned a “commercialization of public broadcasting” would focus public broadcast programming towards the “lowest common denominator,” rather than “diverse audiences,” including minorities, women and the “print handicapped.” Finally, Congress had before it written testimony by the Association of Independent Video and Filmmakers, which stated unequivocally: “commercialization will make public television indistinguishable from the new com*884mercial or pay culture cable services. And public television will fail.”

In addition to the evidence which was before Congress in 1981, the government submitted a report from Roger 6. Noll, an emeritus professor at Stanford University, who had written several books on the economics of the television industry. Noll’s report concluded that because advertisers wish to air commercials on television programs with high numbers of viewers, “[a] competitive, advertiser-supported television system leads to an emphasis on mass entertainment programming.” Advertiser-supported television provides few “programs that serve a small audience, even if that audience has an intense desire to watch programs that differ from standard mass entertainment programs.” Commercial television stations, for example—which are primarily reliant on advertising—offer “children’s programming only if it can be used to market products to children,” despite parents’ desire to have their children watch educational television programming. The government buttressed Noll’s report with a declaration from Lance Ozier, the former president of a nonprofit organization which operates a number of public broadcast television stations in Massachusetts. Ozier stated that the educational programs on public broadcast stations “do not attract sufficient viewership to attract substantial advertising revenue.” Thus, “subjecting non-commercial stations to the same commercial pressures faced by commercial stations would make it economically impossible to provide such programming.”

Minority urges us to discount the evidence before Congress in 1981 as “the opinions, predictions and wishes of the witnesses] unencumbered by any evidence.” Minority is correct that Congress had no empirical data—statistics, academic studies, or otherwise—to support its 1981 conclusion that allowing commercial advertising on public broadcast stations would undermine niche programming on those stations. But we are unaware of any authority which requires a particular type of evidence in the record before Congress. Indeed, Turner I reminds us that “[s]ound policymaking often requires legislators to forecast future events and to anticipate the likely impact of these events,” Turner I, 512 U.S. at 665,114 S.Ct. 2445. We do not normally “substitute our judgment for the reasonable conclusion of a legislative body.” Turner II, 520 U.S. at 211, 117 S.Ct. 1174. Moreover, “Congress is not obligated, when enacting its statutes, to make a record of the type that an administrative agency or court does to accommodate judicial review.” Turner I, 512 U.S. at 666, 114 S.Ct. 2445.

We thus decline Minority’s invitation to second-guess Congress as to the quality of the evidence before it as to the probable effect commercial advertising by for-profit firms would have on program content. In light of the deference we afford to Congress’s legislative judgments, we conclude that Congress’s conclusion that paid promotional messages by for-profit entities pose a threat to extinguish public broadcast stations’ niche programming was supported by substantial evidence, including “substantial evidence in the record before Congress.” Turner II, 520 U.S. at 211, 117 S.Ct. 1174. Moreover, all of the evidence which was before Congress-—and which was submitted by the government to the district court—evinces a strong connection between the harm recited and the prevalence of commercial advertising. Thus, we cannot conclude that § 399b(a)(l) “burden[s] substantially more speech than is necessary to further the government’s legitimate interests.” Turner I, 512 U.S. at 665,114 S.Ct. 2445.

*885 2. Subsections 399b(a)(2) and (a)(3)

The outcome is different for subsections 399b(a)(2) and (a)(3), which proscribe public issue and political advertising. As we explain, there is simply no evidence in the record—much less “substantial evidence in the record before Congress” at the time of the statute’s enaction, Turner II, 520 U.S. at 211, 117 S.Ct. 1174—to connect the ban on this speech to the government’s interest in maintaining certain types of programming. Moreover, there is no evidence that public issue and political advertisements, which are banned, are more harmful than advertisements for goods and services by non-profits, which are allowed. Discovery Network, 507 U.S. at 425, 113 S.Ct. 1505.

As previously discussed, we accept Congress’s conclusion that commercial advertisers seek the largest audience possible, and that, were public broadcast stations permitted to transmit commercial advertisements without restriction, such stations would seek to make themselves more attractive to advertisers by broadcasting programs with mass-market appeal. But neither logic nor evidence supports the notion that public issue and political advertisers are likely to encourage public broadcast stations to dilute the kind of noncommercial programming whose maintenance is the substantial interest that would support the advertising bans.

To take two key examples: the government cites ample evidence that public television provides “more public affairs programming and children’s and family programming” than advertiser supported stations do. It is easy to see how the ban on commercial advertisements in subsection 399b(a)(l) is narrowly tailored to further the governmental interest in preserving such niche programming. But the connection between a ban on public issue and political advertisements and the interest of promoting niche programming is, to put it generously, tenuous—and a tenuous connection is not enough to survive intermediate scrutiny. The restriction must be “narrowly tailored” and there must be “substantial evidence” that supports the restrictions.

Consider first the effect of the ban on public issue and political advertisements on the nature and prevalence of children’s educational programming. There is virtually no way that these advertisements, if allowed, would negatively affect the nature of children’s programming on public television stations. After all, the large majority of viewers of these programs are legally prohibited from voting, so there is virtually no incentive for a station to alter its children’s programming to suit the preferences of a political candidate or issue group. At the outer reaches of one’s imagination, perhaps, lies a potential Saturday morning cartoon featuring an appearance by President Obama or Candidate Romney, Santorum, Paul, or Gingrich, wherein the political personality appears in the episode to fight crime alongside Superman or Batman. It is true that such cartoon would be more likely to exist on a station where the particular candidate is able to run a 30-second political advertisement before and after his world-saving derring-do than on a station where such advertisements are prohibited. But the possibility that such cartoons will replace “Sesame Street” anytime soon seems quite remote. At best, it is pure speculation, which was never mentioned before Congress. Upholding the ban on public issue and political advertising requires more than speculation.

The interest in maintaining public affairs programming of the sort currently seen on public television is a slightly closer case, but the government still fails to carry its *886burden. There are a few scattered remarks in the record that, with § 399b, Congress wished to “insulate public broadcasting from special interest influences— political, commercial, or any other kind.” 127 Cong. Ree. 13145 (June 22, 1981) (remarks of Rep. Gonzalez); see also H.R.Rep. No. 97-82, 97th Cong., 1st Sess. 16 (1981) (emphasizing the need for “insulation of program control and content from the influence of special interests-—-be they commercial, political, or religious”). The temptation of receiving advertising dollars from groups or individuals who wish to air public issue or political advertisements are indeed the types of special interest influences that could distort the nature of public affairs programming offered on public television. Especially in an election season, we see how a news broadcaster may be tempted to alter the content of its public affairs programming if it thinks it can garner additional advertising dollars from one or another campaign, SuperPAC, or advocacy group by doing so.

But speculation aside, there is no evidence in the record—much less evidence which was in the record before Congress—to support Congress’s specific determination that public issue and political advertisements impact the programming decisions of public broadcast stations to a degree that justifies the comprehensive advertising restriction at issue here. In Turner II, the Supreme Court stated that such evidence must at least include “substantial evidence in the record before Congress” at the time of the statute’s en-action, Turner II, 520 U.S. at 211, 117 S.Ct. 1174. Here, the government fails to point to evidence to support' the needed connection between the means of § 399b(a)(2) and (a)(3) (prohibition of public issue and political issue advertisements) and the ends to be achieved (survival of educational programming). For instance, no witness testified to Congress in 1981 as to the relative motivations of public issue and political advertisers when compared to other advertisers. Instead, the only evidence before Congress dealt with the motivations of commercial advertisers or advertisers generally. But the type of advertising proscribed by § 399b(a)(2) and (a)(3) is very different than that proscribed by (a)(1).

Further, the only evidence cited by the government in the district court and in its brief to support § 399b(a)(3)’s content-based restriction on political speech was a 2008 article in AdWeek Magazine which stated that $2.2 billion was spent on political campaign advertisements during 2008. The AdWeek article did not provide any figures for the amount of money spent on non-political advertising, so the significance of the $2.2 billion figure is unclear. Additionally, the fact that a large amount of money was spent on political television advertisements in 2008 is not “substantial evidence” that political advertisers seek a larger audience than do nonprofits advertisers, or that they distort programming decisions by buying advertising time. And a magazine article from 2008 certainly was not “substantial evidence [which was] in the record before Congress” 27 years earlier, in 1981, when § 399b was enacted. See Turner II, 520 U.S. at 211, 117 S.Ct. 1174. Moreover, there was no evidence presented by the government to justify § 399b(a)(2)’s restriction on public issue speech.

Indeed, of the Congressional testimony relied on by the district court in its determination that the statute withstood intermediate scrutiny, it is instructive that two of the three representatives referred explicitly to the threat to public broadcasters of “commercialization.” John C. DeWitt from the American Foundation for the Blind testified that he was concerned the “commercialization of public broadcasting” *887threatened to focus its programming towards the “lowest common denominator,” rather than “diverse audiences,” including minorities, women and the “print handicapped.” Hearings before the Subcomm. on Telecommns, Consumer Protection, and Finance of the H. Comm, on Energy and Commerce on H.R. 3238 and H.R. 2774, 97th Cong. 1st Sess. (1981) (“1981 House Hearings.”). This does nothing to support the bans in subsections (a)(2) and (a)(3), because public issue and political advertisements pose no threat of “commercialization.” By definition, such advertisements do not encourage viewers to buy commercial goods and services. A ban on such advertising therefore cannot be narrowly tailored to serve the interest of preventing the “commercialization” of broadcasting.

The district court also cited testimony of the Association of Independent Video and Filmmakers (AIVF). AIVF’s testimony stated unequivocally that “commercialization will make public television indistinguishable from the new commercial or pay culture cable services. And public television will fail.” But again, this concern simply does not implicate public issue and political advertisements.9

Ultimately, the most revealing statement in the government’s brief on this point is the following sentence, which contains no citations: “Political advertisers are no less capable of exerting influence on programmers than commercial advertisers, and, accordingly, political advertising has never been permitted in public broadcasting.” If that preliminary statement of fact about the ability of political advertisers to exert program influence were supported by some evidence—in particular, some evidence before Congress when it enacted the ban—the government could sustain its burden under intermediate scrutiny. But at such a critical point, the government makes only a bare assertion, unsupported by citation to any evidence. The government cannot simply assert its way out of the “substantial evidence” requirement of the First Amendment.10

*888The fact that Congress chose not to ban all advertisements, but left a gap for certain non-profit advertisements, is also fatal to its case under the analysis in Discovery Network, the commercial handbills case. Here, the banned speech (public issue and political advertisements) is analogous to Discovery Network’s handbills; the permitted speech (promotional advertisements by nonprofits) is analogous to the Discovery Network’s newspapers. And just as the city in Discovery Network was required to prove that handbill-dispensing news-racks posed a greater threat to public safety and esthetics than newspaper-dispensing newsracks, the government here must prove that public issue and political advertisements pose a greater threat to educational programming on public broadcast stations than promotional advertisements on behalf of non-profits. 507 U.S. at 424, 118 S.Ct. 1505. Indeed, the government’s burden is even higher here than in Discovery Network, because § 399b disadvantages political speech— bans on which we must be “particularly wary.” League of Women Voters, 468 U.S. at 384, 104 S.Ct. 3106.11 Recall that the newsracks banned in Discovery Network contained commercial handbills.

Applying the Discovery Network standard to § 399b’s “selective and categorical” content-based bans on speech, we find no basis for § 399b’s content-based discrimination against public issue and political advertisements. In fact, it stands to reason that both public issue and political advertisers, on the one hand, and nonprofits seeking to advertise for their goods and services, on the other, would generally seek the largest audience possible. A nonprofit university which seeks to attract students through television advertisements would want its advertisements seen by as many potential applicants as possible; so, too, would a Presidential candidate generally want his advertisement viewed by the largest possible audience of voters.

Of course, this is a generality: we do not doubt that many advertisers—political and nonpolitical—sometimes target niche markets. A nonprofit group seeking to raise funds for wildlife preservation may choose to spend its money to advertise during nature shows, the better to reach motivated donors. But so might an evangelical Presidential candidate choose to spend his money advertising on religious-themed shows in Iowa in advance of the Iowa caucuses, to increase voter turn-out to his advantage. The point is that in general, there is no reason to think that public issue and political advertisers have any greater propensity to seek large audiences than do non-profit advertisers. Yet Minority and other public television stations may broadcast one type of advertisement but not the other. That is the kind of picking-and-choosing among different types of speech that Congress may not do, absent evidence to show that Congress’s favoritism is necessary to serve its substantial interest.

*889Thus, because § 399b’s content-based ban on public issue and political advertisements bears “no relationship whatsoever to the particular interests that the [government] has asserted,” Discovery Network, 507 U.S. at 424, 113 S.Ct. 1505, the statute is not narrowly tailored. Here, as in Discovery Netivork, there is no basis for the content-based distinction drawn by § 399b. Just as newsracks containing “commercial handbills” and newsracks containing “newspapers” posed an identical threat to safety and esthetics, so too, for aught that appears—especially in the Congressional record—-do public issue and political advertisers, on the one hand, and non-political, non-profit advertisers on the other pose identical threats to displace niche programs on public broadcast television.

The dissent relies on the aforementioned Ozier declaration for the proposition that “the content and quantity of nonprofit advertising do not pose the same sort of threat to public broadcasting’s financial model as other sorts of advertising.” Dissent at 897. But even if the dissent is correct that non-profit advertising in general does not pose the sort of threat that for-profit advertising does, that fact would do nothing to justify 399b’s content-based speech restriction unthin the class of nonprofit advertising. The statute gives Minority the ability to broadcast an advertisement for Planned Parenthood’s goods and services but does not allow Minority to broadcast an advertisement stating Planned Parenthood’s view on a piece of proposed legislation or a political candidate. That is a crucial content-based distinction—and there is no evidence at all to support it.12

The government’s evidence in this case shows only the size and effect of one class of advertising: traditional commercial advertising. That is the content of speech proscribed in subsection § 399b(a)(l), which proscription we today hold passes “intermediate scrutiny” and which we uphold. But the government cannot point to evidence that its fear of harm to public television that would come from allowing stations to air public issue and political advertisements is “real, not merely conjectural,” much less that the portions of the statute which ban such political and public issue advertisements “alleviate those harms in a direct and material way.” Turner I, 512 U.S. at 664, 114 S.Ct. 2445. Thus, we strike down as unconstitutional subsections 399b(a)(2) and (a)(3).

Of course, following today’s decision, Congress is free to “try again.” If there truly is evidence that broadcast of public issue and political advertisements would cause substantial harm—that their broadcast would change program content as directly and substantially as would for-profits’ advertising—Congress could compile a record to show as much, and perhaps pass a law restricting such speech. That record would contain evidence, not mere conjecture and anecdote. It is evidence of harm *890to a substantial governmental interest— not mere conjecture—which the First Amendment requires. See Turner II, 520 U.S. at 211, 117 S.Ct. 1174.

The district court’s grant of summary judgment to the government is AFFIRMED in part, and REVERSED in part. We REMAND to the district court with instructions to enter an order granting summary judgment to Minority Television Project as to § 399b(a)(2) and § 399b(a)(3), and for further proceedings consistent with this opinion.13

. Public broadcast stations'—-alternatively called "noncommercial educational stations,” compare 47 U.S.C. § 399b with 47 C.F.R. § 73.621—are stations which are "used primarily to serve the educational needs of the community; for the advancement of educational programs; and to furnish a nonprofit and noncommercial television broadcast service.” 47 C.F.R. § 73.621.

. In an unpublished disposition filed concurrently with this opinion, we address Minority's contentions that § 399b is unconstitutionally vague, and its appeal of the district court's dismissal of its as-applied challenges to § 399b and its implementing FCC orders.

. Minority does not contest that the broadcasts for which it was fined were "advertisements” within the meaning of § 399b. The translated script for a representative advertisement, for Asiana Airlines, read as follows:

Female Character: "Did you get the surprising news Asiana Airlines sent to you? Now you can get American Airline [sic] free tickets using Asiana mileage.”
Male Character: "Asiana Air now combines mileage with American Airlines.”
Female Character: "Now you can travel free to America, Central or South America, and even Europe—to 270 cities around world [sic] earning mileage with Asiana Airlines. Although you travel with Asiana Airlines or with American Airlines.”
Male Character: "Now where do you want to go?”
Female Character: "Well.... (laughter).”
Male Character: “Mileage benefits with the best airline in the world. Asiana Airlines.”

. Since 1927, the federal government has required radio and television stations which wish to transmit over-the-air signals to obtain a license to broadcast on a particular frequency. See Radio Act of 1927, 44 Stat. 1162 (1927). Since 1939, the FCC has reserved certain frequencies for public television and radio broadcasting stations. 47 CFR §§ 4.131-4.133 (1939) (radio); 41 F.C.C. 148 (1952) (television). The FCC justified its reservation of frequencies for public broadcast television stations by noting that broadcast frequencies are a scarce national resource, and public broadcast stations would provide "programming of an entirely different charac*876ter from that available on most commercial stations.” Id. at 166, ¶ 57.

When the FCC set aside television frequencies for public broadcast stations, it noted that the "objective for which special educational reservations have been established— i.e., the establishment of a genuinely educational type of service—would not be furthered by permitting educational institutions te operate in substantially the same manner as commercial applicants.” Id. As a result, the FCC placed strict restrictions on advertising on public broadcast stations. Before the 1981 enaction of § 399b, public broadcast stations were barred from broadcasting any advertisement, and could identify program sponsors only by name. Id.

. "Electioneering communication” was defined as "any broadcast, cable, or satellite communication” that "refers to a clearly identified candidate for Federal office,” or any communication which is "publicly distributed,” regardless of the medium. Citizens United, 130 S.Ct. 876, 887.

. The statute in that case can be found at 47 U.S.C. § 399 (1982) (subsequently amended by Pub.L. 100-626 (1988)).

. While the Court in Turner I held that the regulation there was not content-based, see 512 U.S. at 652, 114 S.Ct. 2445, these cases nonetheless guide the analysis because they are the most comprehensive descriptions of intermediate scrutiny in the First Amendment context. The government’s brief draws heavily from these cases. Indeed, the D.C. Circuit recently drew heavily on the Turner cases when it applied Second Amendment intermediate scrutiny to a series of new gun regulations imposed by the District of Columbia. See Heller v. District of Columbia, 670 F.3d 1244, 1258-59 (D.C.Cir.2011).

. The Supreme Court's recent decision in Sorrell v. IMS Health Inc., - U.S. -, 131 S.Ct. 2653, 180 L.Ed.2d 544 (2011), reaffirms that § 399b would be subjected to rigorous analysis even if viewed through the lens of commercial speech. In that case, the Court struck down a Vermont law which restricted the sale, disclosure, and use of records created by pharmacies that reveal the prescribing habits of doctors. Such records are useful to drug companies who market their drugs to doctors. That case therefore involved paradigmatic commercial speech, but the Court nonetheless noted that "the First Amendment requires heightened scrutiny whenever the government creates 'a regulation of speech because of disagreement with the message it conveys.’ " Id. at 2664 (quoting Ward v. Rock Against Racism, 491 U.S. 781, 791, 109 S.Ct. 2746, 105 L.Ed.2d 661 (1989)). The Court continued by noting that ”[c]ommercial speech is no exception” to this principle in part because a "consumer’s concern for the free flow of commercial speech often may be far keener than his concern for urgent political dialogue.” Id. (quotation marks omitted).

However, because “the outcome is the same whether a special commercial speech inquiry or a stricter form of judicial scrutiny is applied,” the Court did not formally overrule any cases holding that commercial speech is subject to less protection than core public issue or political speech. Id. at 2667. Thus, after Sorrell, it is clear that commercial speech is subject to a demanding form of intermediate scrutiny analysis.

. A third piece of testimony before Congress in 1981 was by a representative from National Public Radio (NPR), and that testimony is on the whole unhelpful to the government. The NPR representative in fact suggested that the programs provided by public television stations could survive a statutory scheme which allowed paid advertising generally. The NPR representative testified that, per NPR’s internal study, "on-air advertising and pay cable fall into [a] category” which "may attract money for some public television stations, [but] hold little or no promise for public radio.” Thus, the NPR testimony cannot be "substantial evidence” that prohibitions on public issue or political advertisements were “necessary to preserve” public broadcasting stations' programming.

. The dissent contends that the Constitution imposes only two "procedural” requirements on a law’s passage, bicameralism and presentment, and therefore objects to the conclusion that "the constitutionality of a federal law might turn on the quantity and quality of evidence before Congress at the time of enactment.” Dissent at 897. But those are two different questions: one is what the Constitution requires for enactment of a law, and another is what the First Amendment to the Constitution requires for a law abridging speech to be valid. For provisions like subsections 399b(a)(2) and (a)(3), which undeniably restrict speech, the Supreme Court has stated that the law must pass the intermediate scrutiny test: the government must show that the statute "promotes a substantial governmental interest” and "does not burden substantially more speech than necessary to further that interest.” Turner II, 520 U.S. at 213, 117 S.Ct. 1174 (internal quotations omitted). In elaborating the intermediate scrutiny inquiry, the Court stated that courts must "assure that, in formulating its judgments, Congress has drawn reasonable inferences based on substantial evidence.” Id. at 195, 117 S.Ct. 1174. The Turner requirement is not a "procedural requirement” to enactment; rather, it is a substantive test for a *888provision’s constitutionality under the First Amendment.

. Accordingly, even if Judges Noonan and Paez are correct that Discovery Network is not directly applicable because it is a commercial speech case, see Concurrence at 890; Dissent at 895, that would do nothing to change the outcome because "commercial speech [handbills] can be subject to greater governmental regulation than noncommercial speech [political advertising].” Discovery Network, 507 U.S. at 426, 113 S.Ct. 1505. If commercial speech could not be banned in Discovery Network, perforce political speech cannot be banned here. Indeed, the fact that we find certain sections of § 399b to be invalid under a standard that would, if anything, be more favorable to the government reinforces today’s holding.

. The dissent claims that Congress in fact "intended to shield public programming” from undue "political” influence but not from "advertisements by nonprofit entities.” Dissent at 896. In support of this "Congressional intent,” the dissent cites to not a word of statutory text but only to a sentence fragment by a single House member contained in the congressional record. Id. (citing 127 Cong. Rec. 13145 (June 22, 1981) (remarks of Rep. Gonzalez)). But “[r]eliance on such isolated fragments of legislative history in divining the intent of Congress is an exercise fraught with hazards,” New England Power Co. v. New Hampshire, 455 U.S. 331, 342, 102 S.Ct. 1096, 71 L.Ed.2d 188 (1982), and Representative Gonzalez’s isolated statement cannot overcome the absence of evidence in the record justifying the content-based restriction.

. Because Minority was fined $10,000 for violations of § 399b(a)(l)—which survives our decision today—we affirm the district court’s denial of reimbursement. Moreover, we express no opinion as to whether subsection 399b(a)(l) would withstand strict scrutiny analysis.