Shays v. Federal Election Commission

Opinion for the court filed by Circuit Judge TATEL.

Congress passed the McCain-Feingold Act, formally known as the Bipartisan Campaign Reform Act of 2002 (BCRA), Pub.L. No. 107-155, 116 Stat. 81, in an effort to rid American politics of two perceived evils: the corrupting influence of large, unregulated donations called “soft money,” and the use of “issue ads” purportedly aimed at influencing people’s policy views but actually directed at swaying their views of candidates. The Federal Election Commission promulgated regula *917 tions implementing the Act, but in Shays v. FEC, 414 F.3d 76 (D.C.Cir.2005) (“Shays II”), we rejected several of them as either contrary to the Act or arbitrary and capricious, concluding that the Commission had largely disregarded the Act in an effort to preserve the pre-BCRA status quo. Now the FEC has revised the regulations we earlier rejected and issued several new ones, three of which are before us here: (1) a “coordinated communication” standard, the original version of which we rejected in Shays II; (2) definitions of “get-out-the-vote activity” and “voter registration activity”; and (3) a rule allowing federal candidates to solicit soft money at state party fundraisers. Although we uphold one part of the coordinated communication standard known as the “firewall safe harbor,” we reject the balance of the regulations as either contrary to the Act or arbitrary and capricious. We remand these regulations in the hope that, as the nation enters the thick of the fourth election cycle since BCRA’s passage, the Commission will issue regulations consistent with the Act’s text and purpose.

I.

Because both we and the Supreme Court have provided detailed histories of campaign finance regulation, see generally McConnell v. FEC, 540 U.S. 93, 115-32, 124 S.Ct. 619, 157 L.Ed.2d 491 (2003); Shays II, 414 F.3d at 79-82, here we provide only the background necessary to understand this case. Since long before BCRA, the Federal Election Campaign Act (FECA), 2 U.S.C. §§ 431-455, has regulated many aspects of campaign finance. Relevant here, FECA prohibits corporations and unions from making direct contributions or expenditures in connection with federal elections, id. § 441b, and it imposes dollar limits on individuals’ contributions to federal candidates, id. § 441a(a). FECA defines “contributions” as “any gift ... made ... for the purpose of influencing any election for Federal office,” id. § 431(8)(A)(i), and it defines “expenditures” as “any purchase, payment, distribution, ... or gift of money or anything of value, made by any person for the purpose of influencing any election for Federal office,” id. § 431(9)(A)(i). Over time, “contributions subject to [FECA’s] source, amount, and disclosure requirements” came to be known as “hard money,” Shays II, 414 F.3d at 80, while “[political donations made in such a way as to avoid federal regulations or limits” came to be known as “soft money,” The American Heritage Dictionary of the English Language 1652 (4th Ed.2006); see also Shays II, 414 F.3d at 80 (defining “soft money” as “[f]unds outside FECA’s sphere”).

In Buckley v. Valeo, 424 U.S. 1, 96 S.Ct. 612, 46 L.Ed.2d 659 (1976), the Supreme Court, invoking constitutional avoidance, construed FECA’s limitation on expenditures to apply only to funding of communications that “express[ly] ... advocate the election or defeat of a clearly identified candidate for federal office,” i.e., those that contain phrases such as “ ‘vote for,’ ‘elect,’ ‘support,’ ‘cast your ballot for,’ ‘Smith for Congress,’ ‘vote against,’ ‘defeat,’ [or] ‘reject.’ ” Id. at 43-44 & n. 52, 96 S.Ct. 612. Thus, by avoiding these “magic words,” organizations unable to make “expenditures” — such as corporations and unions — could fund so-called “issue ads” that were “functionally identical” to campaign ads and just as effective. McConnell, 540 U.S. at 126, 124 S.Ct. 619; see also FEC v. Mass. Citizens for Life, 479 U.S. 238, 249, 107 S.Ct. 616, 93 L.Ed.2d 539 (1986) (clarifying that the limited definition of “expenditures” applied to ads funded by corporations and unions). “Little difference existed, for example, between an ad that urged viewers to ‘vote against Jane Doe’ and one that condemned Jane Doe’s record on a particular issue before exhorting viewers to ‘call Jane Doe *918 and tell her what you think.’ ” McConnell, 540 U.S. at 126-27, 124 S.Ct. 619.

Following Buckley, the Commission repeatedly interpreted FECA to expand the permissible uses of soft money. In particular, because FECA only regulated contributions intended to influence elections “for Federal office,” 2 U.S.C. § 431(8)(A)(i) (emphasis added), “questions arose concerning the treatment of contributions intended to influence both federal and state elections.” McConnell, 540 U.S. at 123, 124 S.Ct. 619. As the Supreme Court explained:

Although a literal reading of FECA’s definition of “contribution” would have required such activities to be funded with hard money, the FEC ruled that political parties could fund mixed-purpose activities — including get-out-the-vote drives and generic party advertising — in part with soft money. In 1995 the FEC concluded that the parties could also use soft money to defray the costs of “legislative advocacy media advertisements,” even if the ads mentioned the name of a federal candidate, so long as they did not expressly advocate the candidate’s election or defeat.

Id. at 123-24, 124 S.Ct. 619 (footnote and citations omitted).

Because soft money could now be spent in so many ways that benefited federal candidates, and because it could be raised in massive amounts without any of FECA’s limitations or reporting requirements, federal candidates would often solicit such donations directed to their political party. The party would then spend the money on ads supporting the candidate-omitting the magic words — or on get-out-the-vote activity and voter registration activity aimed at helping the candidate. This “enabled parties and candidates to circumvent FECA’s limitations on the source and amount of contributions in connection with federal elections.” Id. at 126, 124 S.Ct. 619. “As the permissible uses of soft money expanded, the amount of soft money raised and spent by the national political parties increased exponentially,” from $22 million in 1984 to $498 million in 2000. Id. at 124, 124 S.Ct. 619. Thus, “the ‘soft money loophole’ had led to a ‘meltdown’ of the campaign finance system that had been intended ‘to keep corporate, union and large individual contributions from influencing the electoral process.’ ” Id. at 129, 124 S.Ct. 619 (quoting S.Rep. No. 105-167, vol. 4, at 4611 (1998); id. vol. 5, at 7515).

Recognizing these problems, Congress passed BCRA, the “central provisions” of which were “designed to address Congress’ concerns about the increasing use of soft money and issue advertising to influence federal elections.” Id. at 132, 124 S.Ct. 619. BCRA made a number of dramatic changes to campaign finance law to achieve these goals, including barring national political parties from soliciting soft money. 2 U.S.C. § 441i(a). Relevant here, the Act required the FEC to develop a new test for determining what advertisements count as “coordinated communications,” BCRA § 214(c), 116 Stat. at 95 (codified at 2 U.S.C. § 441a note); barred state parties from spending soft money on “federal election activity,” including “get-out-the-vote activity” and “voter registration activity,” 2 U.S.C. § 441i(b)(l); and prohibited federal candidates from soliciting soft money, id. § 441i(e).

The FEC first issued regulations implementing BCRA in 2003. Believing these regulations far too permissive, Representative Chris Shays, a prime BCRA sponsor, challenged nineteen of them in the United States District Court for the District of Columbia, arguing that they either violated the Act or were arbitrary and capricious. In Shays v. FEC, 337 F.Supp.2d 28 (D.D.C.2004) (“Shays /”), *919 the district court largely agreed with Shays, rejecting fifteen of the nineteen regulations. On appeal, the FEC challenged the district court’s decision as to five of the regulations, and in Shays II, 414 F.3d 76, we affirmed the district court.

In response, the Commission modified or more thoroughly justified its proposed regulations and reissued them, along with several new ones. Shays now challenges three of these regulations. The first is the Commission’s definition of “coordinated communications,” the original version of which we rejected in Shays II. This regulation includes three subparts: (1) a “content standard” providing that only ads containing certain content may be deemed coordinated, 11 C.F.R. § 109.21(c); (2) a “conduct standard” governing when campaign employees and vendors who go to work for outside organizations may share campaign information, id. § 109.21(d)(4)-(5); and (3) a “firewall safe harbor” provision that is also part of the conduct standard and protects groups hiring former campaign employees and vendors, id. § 109.21(h). The second challenged regulation defines “get-out-the-vote activity” and “voter registration activity,” id. § 100.24(a)(2)-(3), while the third allows federal candidates to solicit soft money at state party fundraisers, id. § 300.64.

In a thorough opinion, the district court rejected each of these rules except the last one, finding them either contrary to BCRA’s purpose or arbitrary and capricious. See Shays v. FEC, 508 F.Supp.2d 10 (D.D.C.2007) (“Shays III”). The FEC now appeals as to the rules the district court struck down, and Shays cross appeals as to the rule the district court upheld. Senator Russell Feingold, another prime BCRA sponsor, has filed an amicus brief supporting Shays.

We address each rule in turn, employing two familiar standards of review: Chevron and the Administrative Procedure Act. As we explained in Shays II:

[Bjecause the regulations at issue interpret statutes the FEC administers, we review them under the two-step analysis set forth in Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), asking first whether Congress has spoken directly to the precise question at issue, and second, if it has not, whether the agency’s interpretation is reasonable. At the same time, because the regulations reflect final agency action under the APA, we ask whether they are “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A).

414 F.3d at 96 (citation omitted). In applying Chevron’s second step and the APA, we “must reject administrative constructions of [a] statute ... that frustrate the policy that Congress sought to implement.” Cont’l Air Lines, Inc. v. Dep’t of Transp., 843 F.2d 1444, 1453 (D.C.Cir.1988) (quoting FEC v. Democratic Senatorial Campaign Comm., 454 U.S. 27, 32, 102 S.Ct. 38, 70 L.Ed.2d 23 (1981)). We review the district court’s Chevron and APA holdings de novo. See Am. Legion v. Derwinski, 54 F.3d 789, 795 (D.C.Cir.1995).

II.

The first and most important issue before us is the FEC’s revised “coordinated communication” standard. Federal election law “has long restricted coordination of election-related spending between official campaigns and outside groups. The reason ... is obvious. Without a coordination rule, politicians could evade contribution limits and other restrictions by having donors finance campaign activity directly,” e.g., by asking a donor to buy *920 air time for a campaign-produced advertisement. Shays II, 414 F.3d at 97.

To prevent such evasion, FECA defines “contributions” to include “expenditures made by any person in cooperation, consultation, or concert, with, or at the request or suggestion of, a candidate.” 2 U.S.C. § 441a(a)(7)(B)(i).

Under pre-BCRA regulations, the FEC determined whether public communications such as radio and television ads were “coordinated” based largely on whether the candidate had engaged in “substantial discussion or negotiation” with an outsider, resulting in “collaboration or agreement.” See Shays [I], 337 F.Supp.2d at 55-56 & n. 25 (quoting old regulation). Absent that degree of cooperation, the communication was considered uncoordinated and thus would not count as a FECA contribution. BCRA instructed the Commission to scrap this approach. “The regulations on coordinated communications ... are repealed,” Congress declared. “The Federal Election Commission shall promulgate new regulations on coordinated communications paid for by persons other than candidates, authorized committees of candidates, and party committees. The regulations shall not require agreement or formal collaboration to establish coordination.” BCRA § 214(c), 116 Stat. at 95. Apart from this negative command- — -“shall not require”— BCRA merely listed several topics the rules “shall address,” providing no guidance as to how the FEC should address them. See id.

Shays II, 414 F.3d at 97-98 (omission in original).

Responding to BCRA, the FEC issued new coordinated communication regulations. “Under its new test, communications count as ‘coordinated’ (and thus as contributions) if: (1) someone other than the candidate, party, or official campaign pays for them, (2) the communication itself meets specified ‘content standards,’ and (3) the payer’s interaction with the candidate/party satisfies specified ‘conduct standards.’ ” Id. at 98 (quoting 11 C.F.R. § 109.21). The conduct standard can be satisfied in several ways, e.g., if “[t]he communication is created, produced, or distributed at the request or suggestion of a candidate,” 11 C.F.R. § 109.21(d)(l)(i); if “[t]he communication is created, produced, or distributed after one or more substantial discussions about the communication between the person paying for the communication ... and the candidate who is clearly identified in the communication,” id. § 109.21(d)(3); or if the person paying for the communication hires a candidate’s vendor or former employee “to create, produce, or distribute” it and in doing so that vendor/employee uses “material” information about “campaign plans, projects, activities, or needs” or shares such information with the payer, id. § 109.21(d)(4)-(5).

Shays challenges the content standard and two features of the conduct standard. We address each challenge in turn.

The Content Standard

“Under the ‘content’ element” of the original rule, “communications made within 120 days of a general election or primary and ‘directed’ at the relevant electorate [could] qualify as ‘coordinated’ if they referred] to a political party or ‘clearly identified candidate for Federal office.’ ” Shays II, 414 F.3d at 98 (quoting 11 C.F.R. § 109.21(c)(4) (2003)). “Before the 120-day mark,” however, “the rule covered] only communications that either recycle[d] official campaign materials or ‘expressly advocate[d] the election or defeat of a clearly identified candidate for federal office.’ ” Id. (emphasis added) (quoting 11 C.F.R. § 109.21(c)(2)-(3) (2003)). Thus, more than 120 days before a federal elec *921 tion, the FEC’s original rule allowed candidates to coordinate with outside groups so long as the ads those groups funded did not include the magic words or recycle campaign materials.

Challenging the rule, Shays argued that limiting regulation outside the 120-day window only to advertisements containing certain types of content violated the Act’s plain language and purpose, and that the Commission had failed to provide any good reason for doing so. The district court rejected Shays’s Chevron step one argument but found that the regulation failed Chevron step two because “excluding] certain types of communications” based solely on their content “regardless of whether or not they are coordinated would create an immense loophole that would facilitate the circumvention of the Act’s contribution limits, thereby creating ‘the potential for gross abuse.’ ” Shays I, 337 F.Supp.2d at 65 (quoting Orloski v. FEC, 795 F.2d 156, 165 (D.C.Cir.1986)).

In Shays II, we agreed with the district court that the rule was invalid, but for slightly different reasons. Like the district court, we “reject[ed] Shays’s ... argument that FECA precludes content-based standards under Chevron step one,” Shays II, 414 F.3d at 99, but we “disagree^] with the district court’s suggestion that any standard looking beyond collaboration to content would necessarily ‘create an immense loophole,’ thus exceeding the range of permissible readings under Chevron step two,” id. at 99-100 (emphasis added) (quoting Shays I, 337 F.Supp.2d at 65). Rather, we saw no need to reach the Chevron step two question— whether this particular content standard violated BCRA — because “contrary to the APA, the Commission offered no persuasive justification for the provisions challenged ..., i.e., the 120-day time-frame and the weak restraints applying outside of it.” Id. at 100; see also id. at 97 (“[W]e need not decide whether [this rule] represents an] altogether impermissible interpretation[] of FECA and BCRA — the Chevron step two inquiry — because in any event the FEC has given no rational justification for [it], as required by the APA’s arbitrary and capricious standard.” (citation omitted)). Remanding the rule, we directed the FEC to provide “some cogent explanation” for it, “not least because” it effectively “allowed a coordinated communication free-for-all for much of each election cycle.” Id. at 100. As we explained:

Under the[se] ... rules, more than 120 days before an election or primary, a candidate may sit down with a well-heeled supporter and say, “Why don’t you run some ads about my record on tax cuts?” The two may even sign a formal written agreement providing for such ads. Yet so long as the supporter neither recycles campaign materials nor employs the “magic words” of express advocacy — “vote for,” “vote against,” “elect,” and so forth — the ads won’t qualify as contributions subject to FECA.

Id. at 98.

On remand, the Commission published a new notice of proposed rulemaking, took comments, held hearings, and analyzed extensive data on television advertising by candidates for federal office. It then issued a revised regulation identical to the original regulation except that it shortened the length of stricter regulation in congressional races to 90 days. The revised regulation prohibits coordinated advertisements “refer[ring] to a clearly identified House or Senate candidate ... in the clearly identified candidate’s jurisdiction 90 days or fewer before the clearly identified candidate’s general, special, or runoff election, or primary or preference election.” 11 C.F.R. § 109.21(c)(4)®. It prohibits coordinated advertisements “re- *922 ferfring] to a clearly identified Presidential or Vice Presidential candidate ... in a jurisdiction during the period of time beginning 120 days before the clearly identified candidate’s primary or preference election in that jurisdiction, or nominating convention or caucus in that jurisdiction, up to and including the day of the general election.” Id. § 109.21 (c)(4)(ii). Outside the 90/120-day windows, however, the regulation still prohibits only coordinated advertisements that “disseminate[ ], dis-tributee ], or republish[ ] ... campaign materials prepared by a candidate,” or “expressly advocate[] the election or defeat of a clearly identified candidate.” Id. § 109.21(c)(2)-(3).

Again challenging the rule, Shays argued that the 90/120-day windows were unsupported by the evidence, violating the APA, and that the lax standard applying outside the windows was both unexplained and contrary to BCRA’s purpose, violating the APA and failing Chevron step two review. The district court concluded that the FEC had adequately justified the 90/120-day windows because the record showed that the “vast majority of candidate advertising occurred within” those periods. Shays III, 508 F.Supp.2d at 42; see id. at 40-43. It also rejected Shays’s claim that the lax pre-window standard would undermine the Act’s purposes. The district court nonetheless struck down the revised regulation as arbitrary and capricious because the FEC “ma[de] no attempt whatsoever to justify the Commission’s continued reliance on the express advocacy standard” outside the windows, id. at 47, thus “failpng] to meet the APA’s standard of reasoned decisionmaking,” id. at 48-49.

The FEC appeals this finding, but before we can reach the merits, we face a jurisdictional question. In Shays II we held that Shays had standing to challenge the regulations at issue there because he satisfied standing’s three requirements, “demonstrat[ing] that he ha[d] suffered ‘injury in fact,’ that the injury [wa]s ‘fairly traceable’ to the actions of the defendant, and that the injury w[ould] likely be redressed by a favorable decision.” Bennett v. Spear, 520 U.S. 154, 162, 117 S.Ct. 1154, 137 L.Ed.2d 281 (1997) (quoting Lujan v. Defenders of Wildlife, 504 U.S. 555, 560-61, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992)). The “injury in fact” was the FEC’s “illegal structuring of [the] competitive environment” in which Shays ran for Congress, Shays II, 414 F.3d at 85, that injury was traceable to the Commission because it promulgated the challenged rules, id. at 92-95, and a favorable decision could redress the injury by striking down the rules, id. at 95.

The FEC suggests that this case is different, saying “[i]t is unclear whether the Court has jurisdiction to rule on Shays’[s] challenge to the portion of the regulation governing the presidential election because he has never been, or stated any intention to be, a candidate for president.” Appellant’s Reply Br. 25 n. 12. The Commission failed to mention this argument in its opening brief, first raising it in a footnote in its reply brief. Moreover, although the Commission assured us in its brief that it was “not challeng[ing] Shays’[s] standing,” but rather only highlighting this issue for the court because we have our “own obligation to determine that [we have] jurisdiction over each of [Shays’s] claims,” id., the Commission changed its tone at oral argument, asserting that Shays lacked standing to challenge the 120-day window applicable to presidential candidates, Oral Arg. at 47:16-:38. Normally we would not consider an argument first raised in a reply brief, Carter v. George Washington Univ., *923 387 F.3d 872, 883 (D.C.Cir.2004), much less one raised only in a footnote, Hutchins v. District of Columbia, 188 F.3d 531, 539-40 n. 3 (D.C.Cir.1999) (en banc). But because this argument goes to our jurisdiction, we must consider it, see United States v. Hylton, 294 F.3d 130, 136 (D.C.Cir.2002), though we are disappointed in the FEC for raising this issue so late that Shays had no adequate opportunity to respond.

That said, Shays plainly has standing under FEC v. Akins, 524 U.S. 11, 118 S.Ct. 1777, 141 L.Ed.2d 10 (1998). Indeed, after some prodding at oral argument, FEC counsel virtually conceded as much, Oral Arg. at 47:45-49:15. In Akins, the petitioners — a group of voters seeking information about the political activities of the American Israel Public Affairs Committee (AIPAC) — challenged the FEC’s determination that AIPAC did not qualify as a “political committee,” a decision that meant AIPAC had no obligation to report information about its “members, contributions, and expenditures.” Id. at 16, 118 S.Ct. 1777. The Court held that petitioners had suffered an injury in fact, namely “their inability to obtain information — lists of AIPAC donors ... and campaign-related contributions and expenditures — that, on [their] view of the law, the statute require[d] that AIPAC make public,” id. at 21, 118 S.Ct. 1777; see also id. (“[A] plaintiff suffers an ‘injury in fact’ when the plaintiff fails to obtain information which must be publicly disclosed pursuant to a statute.”) (citing Pub. Citizen v. DOJ, 491 U.S. 440, 449, 109 S.Ct. 2558, 105 L.Ed.2d 377 (1989)).

Here, as in Akins, Shays’s injury in fact is the denial of information he believes the law entitles him to. Specifically, under the FEC’s definition of coordinated communications, presidential candidates need not report as contributions many expenditures that Shays believes BCRA requires them to report. Thus, Shays claims that the regulation illegally denies him information about who is funding presidential candidates’ campaigns. We see no difference between this injury and the injury deemed sufficient to create standing in Akins. Here, as there, “the information would help [Shays] (and others to whom [he] would communicate it) to evaluate candidates for public office ..., and to evaluate the role that [outside groups’] financial assistance might play in a specific election.” Id. And here, as there, Shays’s “injury consequently seems concrete and particular.” Id. Finally, as in Akins, Shays’s injury is fairly traceable to the FEC because it is caused by the Commission’s rule, and the injury would be redressed were this court to invalidate the rule. Id. at 25, 118 S.Ct. 1777.

Assured of Shays’s standing to challenge this rule in its entirety, we turn to the merits. Shays claims that the rule suffers from two flaws. First, the FEC failed to justify the length of the 90/120-day windows, violating the APA. And second, the lax standard the Commission imposed outside those windows not only runs counter to BCRA’s purpose, but also was entirely unjustified, failing both Chevron step two and APA review. After describing the evidence before the Commission, we address each argument in turn.

On remand the Commission gathered extensive evidence about the timing of advertising in federal election campaigns. Reviewing data from the Campaign Media Analysis Group regarding television ads run by federal candidates in the 2004 election cycle, the Commission found that “Senate candidates aired only 0.87 percent and 0.39 percent of their advertisements more than 90 days before their primary and general elections, respectively,” while “House candidates aired only 8.56 percent and 0.28 percent of their advertisements more than 90 days before their primary and general elections, respectively.” *924 Coordinated Communications, 71 Fed.Reg. 33,190, 33,194 (2006). In the 2004 presidential campaign, 8.44 percent of all candidate TV ads in the primary ran outside the 120-day window, as did 16 percent of all candidate TV ads in Iowa before its crucial caucus. Shays III, 508 F.Supp.2d at 45. While these percentages are small, the total amount spent on pre-window ads was substantial, totaling into the millions of dollars. See id.

In addition to evidence about spending by candidates, the Commission had before it many examples of expenditures by outside groups before the 90/120-day windows. For example, in the 2004 Alaska Senate race, the U.S. Chamber of Commerce began running TV ads supporting Senator Lisa Murkowski nine months before the primary election. In the 2004 Florida Senate race, the illuminatingly-named “People for a Better Florida” began running ads attacking candidate Mel Martinez over five months before the primary. In the 2006 Pennsylvania Senate race, a group called “Americans for Job Security” spent $500,000 on TV ads supporting Senator Rick Santorum starting six months before the primary. In the 2004 South Dakota Senate race, the Club for Growth began running ads attacking Senator Tom Daschle fifteen months before the general election. The group ran similar ads against Rhode Island Senator Lincoln Chafee beginning nine months before his 2006 primary. Because none of these ads contained the “magic words” of express advocacy, all could have been coordinated with candidates under the Commission’s rule.

The record also reveals that the vast majority of campaign ads omit “express advocacy.” “In the 1998 election cycle, just 4% of candidate advertisements used magic words; in 2000, that number was a mere 5%.” McConnell, 540 U.S. at 127 n. 18, 124 S.Ct. 619. “Indeed, campaign professionals” told Congress while it was considering BCRA “that the most effective campaign ads ... avoid the use of the magic words.” Id. at 127, 124 S.Ct. 619. Because campaign advertisements rarely use magic words, the Supreme Court has declared the express advocacy test “functionally meaningless.” Id. at 193, 124 S.Ct. 619.

In sum, the record demonstrates several key points: (1) the vast majority of advertising by candidates occurs in the 90/120-day windows the FEC regulates more strictly; (2) candidates and outside groups nonetheless run a significant number of ads before the 90/120-day windows; and (3) very few ads contain magic words. These facts lead us to two inexorable conclusions: the FEC’s decision to regulate ads more strictly within the 90/120-day windows was perfectly reasonable, but its decision to apply a “functionally meaningless” standard outside those windows was not. Id. at 193, 124 S.Ct. 619.

Beginning with the windows, we made clear in Shays II that nothing in BCRA forbids the FEC from “dr[awing] distinctions based on content, time, and place”; its failure then was that it provided no evidence in support of the window it chose. 414 F.3d at 100. But given the record evidence showing that the vast majority of federal campaign advertisements run within the more strictly regulated windows, the FEC now “appears to have drawn the line in a reasonable place based on the data available to it.” Shays III, 508 F.Supp.2d at 43.

The next issue is whether the FEC’s decision to regulate only ads containing express advocacy outside the 90/120-day windows fails Chevron step two review or violates the APA. As our cases explain, these inquiries overlap, for “[wjhether a statute is unreasonably interpreted is close analytically to ... whether an agency’s *925 actions under a statute are unreasonable.” Gen. Instrument Corp. v. FCC, 213 F.3d 724, 732 (D.C.Cir.2000). At Chevron step two and under the APA, “[courts] must reject administrative constructions of [a] statute ... that frustrate the policy that Congress sought to implement.” Cont’l Air Lines, 843 F.2d at 1453 (quoting Democratic Senatorial Campaign Comm., 454 U.S. at 32, 102 S.Ct. 38). While that policy may sometimes be unclear, here it is not: “BCRA’s fundamental purpose [is] prohibiting soft money from being used in connection with federal elections.” McConnell, 540 U.S. at 177 n. 69, 124 S.Ct. 619; see also id. at 132, 124 S.Ct. 619 (“BCRA’s central provisions are designed to address Congress’ concerns about the increasing use of soft money and issue advertising to influence federal elections.”). Recall that “soft money” refers to political donations made in such a way as to avoid FECA’s restrictions. See Shays II, 414 F.3d at 80.

The question, then, is this: Does the challenged regulation frustrate Congress’s goal of “prohibiting soft money from being used in connection with federal elections”? McConnell, 540 U.S. at 177 n. 69, 124 S.Ct. 619. We think it does. Outside the 90/120-day windows, the regulation allows candidates to evade — almost completely'— BCRA’s restrictions on the use of soft money. As FEC counsel conceded at oral argument, Oral Arg. at 0:46-2:00, the regulation still permits exactly what we worried about in Shays II, i.e., more than 90/120 days before an election, candidates may ask wealthy supporters to fund ads on their behalf, so long as those ads contain no magic words. 414 F.3d at 98. Indeed, pressed at oral argument, counsel admitted that the FEC would do nothing about such coordination, even if a contract formalizing the coordination and specifying that it was “for the purpose of influencing a federal election” appeared on the front page of the New York Times. Oral Arg. at 7:34-8:03. Thus, the FEC’s rule not only makes it eminently possible for soft money to be “used in connection with federal elections,” McConnell, 540 U.S. at 177 n. 69, 124 S.Ct. 619, but also provides a clear roadmap for doing so, directly frustrating BCRA’s purpose. Moreover, by allowing soft money a continuing role in the form of coordinated expenditures, the FEC’s proposed rule would lead to the exact perception and possibility of corruption Congress sought to stamp out in BCRA, for “expenditures made after a ‘wink or nod’ often will be ‘as useful to the candidate as cash,’ ” id. at 221, 124 S.Ct. 619 (quoting FEC v. Colo. Republican Fed. Campaign Comm., 533 U.S. 431, 442, 446, 121 S.Ct. 2351, 150 L.Ed.2d 461 (2001)), and “[i]t is not only plausible, but likely, that candidates would feel grateful for such donations and that donors would seek to exploit that gratitude,” id. at 145, 124 S.Ct. 619.

The FEC offers four reasons why we should nonetheless uphold this lax standard. First, explaining that it chose the standard to protect the First Amendment rights of outside groups conducting independent expenditures, it argues that any standard more vague than “express advocacy” would unacceptably chill the speech of such groups. We applaud the Commission’s sensitivity to First Amendment values, but as we said in Shays II, “regulating nothing at all” would achieve the same purpose, “and that would hardly comport with the statute.” 414 F.3d at 101. Thus, “[n]otwithstanding its obligation to attempt to avoid unnecessarily infringing on First Amendment interests, the Commission must establish, consistent with APA standards, that its rule rationally separates election-related advocacy from other activity falling outside FECA’s expenditure definition,” id. at 101-02 (citation omitted), which, remember, defines “expenditure” as “any purchase, payment, ... or gift of money or anything of value, *926 made by any person for the purpose of influencing any election for Federal of fice.” 2 U.S.C. § 431(9)(A)(i) (emphasis added). Here the Commission failed to show that its rule rationally separates election-related advocacy from other speech, for many of the ads its rule leaves unregulated are plainly intended to “influencie] an[ ] election for Federal office.” Id. The FEC claims it has drawn a rational line because ads omitting magic words run by outside groups in coordination with candidates before the windows are generally not intended to influence federal elections. But this is absurd. Because the magic words test is “functionally meaningless,” McConnell, 540 U.S. at 193, 124 S.Ct. 619, and “expenditures made after a ‘wink or nod’ often will be ‘as useful to the candidate as cash,’ ” id. at 221, 124 S.Ct. 619 (quoting Colo. Republican Comm., 533 U.S. at 442, 446, 121 S.Ct. 2351), there is no question that coordinated ads omitting magic words are often intended to influence federal elections. This is true even outside the 90/120-day windows, for as the FEC itself found, “[ajny time a candidate uses campaign funds to pay for an advertisement, it can be presumed that this advertisement is aired for the purpose of influencing the candidate’s election.” 71 Fed.Reg. at 33,193 (emphasis added). We have no reason to think this is any less true of spending that candidates coordinate with outside groups. In sum, although the FEC, properly motivated by First Amendment concerns, may choose a content standard less restrictive than the most restrictive it could impose, it must demonstrate that the standard it selects “rationally separates election-related advocacy from other activity falling outside FECA’s expenditure definition.” Shays II, 414 F.3d at 102. Because the “express advocacy” standard fails that test, it runs counter to BCRA’s purpose as well as the APA.

Second, the FEC points to our decision in Orloski v. FEC, 795 F.2d 156 (D.C.Cir.1986), as support for the rule it chose. Orloski dealt with 2 U.S.C. § 441b(a), which prohibits corporations from making “contribution^] or expenditure^] in connection with any election to any political office.” The FEC interpreted this provision to allow corporations to fund events for federal officeholders so long as those events were “non-political,” i.e., “(1) there is an absence of any communication expressly advocating the nomination or election of the congressman appearing or the defeat of any other candidate, and (2) there is no solicitation, making, or acceptance of a campaign contribution for the congressman in connection with the event.” Orloski, 795 F.2d at 160. Upholding the regulation under Chevron, we explained that although it was “at the outer bounds of permissible choice,” it was “still a ‘reasonable choice within a gap left open by Congress.’ ” Id. at 167 (quoting Chevron, 467 U.S. at 866, 104 S.Ct. 2778).

The FEC urges us to reach the same conclusion here, but it ignores the crucial differences separating Orloski from this case. Most important, in Orloski we found that “the FEC’s interpretation does not create the potential for gross abuse” because “under the FEC’s interpretation, corporations can make little more than insignificant, indirect donations to a candidate’s political warchest, which are unlikely to give the corporations improper influence over candidates for federal office or to significantly increase the level of campaign spending.” Id. at 165-66. Here, by contrast, the coordinated expenditures the Commission’s rule allows “often will be ‘as useful to the candidate as cash,’ ” McConnell, 540 U.S. at 221, 124 S.Ct. 619 (quoting Colo. Republican Comm., 533 U.S. at 446, 121 S.Ct. 2351), and “[i]t is not only plausible, but likely, that candidates would feel grateful for such donations and that *927 donors would seek to exploit that gratitude,” id. at 145, 124 S.Ct. 619. This “create[s] the potential for gross abuse” that was absent in Orloski. Orloski, 795 F.2d at 165. Moreover, in Orloski we said “[i]f the FEC’s interpretation unduly compromises the Act’s purposes, it is not a ‘reasonable accommodation’ under the Act, and it would therefore not be entitled to deference.” Id. at 164 (quoting Chevron, 467 U.S. at 845, 104 S.Ct. 2778). Here, as we have explained, the rule “unduly compromises” the Act’s purpose of “prohibiting soft money from being used in connection with federal elections.” McConnell, 540 U.S. at 177 n. 69, 124 S.Ct. 619.

Third, the FEC disparages the many examples Shays provides of pre-window expenditures by candidates and outside groups, calling them mere “anecdotes” and saying Shays failed to offer any evidence of their relative significance. See Appellant’s Reply Br. 19-21. But the FEC’s own study showed that almost 10% of primary election advertisements by House candidates and presidential candidates in 2004 — plainly “aired for the purpose, of influencing the candidate’s election,” 71 Fed.Reg. at 33,193 — ran before the windows, and Shays provided numerous examples of pre-window ads funded by outside groups that were obviously intended to influence federal elections. Notably, many of Shays’s examples came from media markets excluded from the FEC’s study, and they suggest that the percentage of early advertising may be even greater than that captured by the FEC’s analysis. Shays’s evidence, combined with the FEC’s study, proves his point. Given the rule the FEC chose, which regulates virtually no coordinated pre-window ads, the Commission could demonstrate that it met its statutory obligation — “rationally separating] election-related advocacy from other activity falling outside FECA’s expenditure definition,” Shays II, 414 F.3d at 102 — only by showing that a truly insignificant number of ads intended to influence federal elections run before the windows. See Shays II, 414 F.3d at 99 (“[T]he FEC lacks discretion to exclude [communications intended to influence federal elections] from its coordinated communication rule.”). The evidence in the FEC’s own study, as well as the evidence Shays provided, refutes any such contention.

Finally, the FEC assures us that we have no reason to worry about lax regulation outside the 90/120-day windows because it has received very few complaints alleging that candidates are currently coordinating expenditures with outside groups before the windows, and there is no evidence that candidates will begin coordinating with outside groups if we uphold the regulation. This argument flies in the face of common sense. Of course the FEC hasn’t received many complaints: the challenged rule allows unlimited coordination so long as the resulting advertisements omit express advocacy. In other words, people have had no reason to report this type of coordination because it is perfectly legal under the FEC’s rule. Moreover, the Commission’s prediction about what will happen in the future disregards everything Congress, the Supreme Court, and this court have said about campaign finance regulation. In passing BCRA, Congress found that ads funded with soft money “were often actually coordinated with, and controlled by, the campaigns.” McConnell, 540 U.S. at 131, 124 S.Ct. 619 (citing S. Rep. No. 105-167, vol. 1, at 49 (1998); id. vol. 3, at 3997-4006). In McConnell, the Supreme Court said, “[m]oney, like water, will always find an outlet,” id. at 224, 124 S.Ct. 619, and BCRA reflects “the hard lesson of circumvention” Congress has learned from “the entire history of campaign finance regulation,” id. at 165, 124 S.Ct. 619. And in Shays II, we said, “if regulatory safe harbors permit what BCRA bans, we have no *928 doubt that savvy campaign operators will exploit them to the hilt, reopening the very soft money floodgates BCRA aimed to close.” 414 F.3d at 115. Common sense requires the same conclusion here. Under the present rules, any lawyer worth her salt, if asked by an organization how to influence a federal candidate’s election, would undoubtedly point to the possibility of coordinating pre-window expenditures. The FEC’s claim that no one will take advantage of the enormous loophole it has created ignores both history and human nature.

Conduct Standard: Campaign Vendors and Former Employees

BCRA directed the FEC, in issuing its revised coordinated communication rules, to address “payments for the use of a common vendor” and “payments for communications directed or made by persons who previously served as an employee of a candidate or a political party.” BCRA § 214(c), 116 Stat. at 95 (codified at 2 U.S.C. § 441a note). The FEC’s original post-BCRA regulations implemented these provisions by specifying that the “conduct prong” of the coordinated communication test would be satisfied if a candidate’s vendor or former employee “create[d], produce[d], or distribute[d]” a communication using “material” information about “campaign plans, projects, activities, or needs,” or shared such information with the person paying for the communication, throughout the “current election cycle.” 11 C.F.R. § 109.21(d)(4)-(5) (2003).

Shays chose not to challenge these original provisions, but the FEC nonetheless revisited them after we remanded other aspects of the coordinated communication rule in Shays II. Because campaign vendors and employees complained that the regulation was unnecessarily cumbersome — they claimed that the information they possess quickly loses value — the FEC decided to change the rule so that it only prohibits vendors and former employees from using “material information” about “campaign plans, projects, activities, or needs,” or sharing such information with the person funding the ad, for 120 days, rather than throughout the whole election cycle. 11 C.F.R. § 109.21(d)(4)-(5).

In the district court, Shays challenged the revised regulation, arguing that it ran counter to BCRA’s purpose and violated the APA. Although the district court rejected the Chevron step two argument, it found the revised regulation arbitrary and capricious because the FEC had failed to justify its policy change. Shays III, 508 F.Supp.2d at 49-52. We agree.

Explaining the new rule, the FEC reasoned that “[r]educing the temporal limit to 120 days will not undermine the effectiveness of the conduct standards and will not lead to circumvention of the Act” because “material information regarding candidate and political party committee campaigns, strategy, plans, needs, and activities ... does not remain ‘material’ for long periods of time during an election cycle.” 71 Fed.Reg. at 33,204. The Commission went on to say that “a limit of 120 days is more than sufficient to reduce the risk of circumvention of the Act.” Id. at 33,205. We see two flaws in this rationale.

First, as the district court pointed out, “the Commission’s generalization that material information does not remain material for long overlooks the possibility that some information — for instance, a detailed state-by-state master plan prepared by a chief strategist — may very well remain material for at least the duration of a campaign.” Shays III, 508 F.Supp.2d at 51. Indeed, the Commission’s own regulations recognize that some types of information retain value for longer than 120 days. For example, the Commission says that polling data — arguably the campaign information that most quickly becomes obsolete — re *929 tains some value for 180 days. See 11 C.F.R. § 106.4(g). Yet the Commission inexplicably asserts that other types of campaign information — including some far more durable information such as donor lists and lists of supportive voters — will have lost value within 120 days. As Shays points out, under the FEC’s regulation, a top presidential campaign staffer could leave a campaign after an early primary, wait 120 days, and then spend the entire general election working for an outside group on behalf of his former candidate, using that candidate’s donor lists, mailing lists, and long-term strategic plan. The Commission never explains why this type of coordination should go unregulated.

Second, the FEC has provided no explanation for why it believes 120 days is a sufficient time period to prevent circumvention of the Act. Though the Commission certainly has some discretion in choosing exactly where to draw a bright line such as this one, it must support its decision with reasoning and evidence, for “a bright line can be drawn in the wrong place.” Shays II, 414 F.3d at 101.

Conduct Standard: Firewall Safe Harbor

When it revised the conduct standard with regard to former employees and vendors following Shays II, the FEC created a new “firewall safe harbor” provision to protect vendors and organizations in which some employees are working on a candidate’s campaign and others — separated by a firewall — are working for outside groups making independent expenditures. Under the new regulation, “[t]he conduct standards ... are not met if the commercial vendor, former employee, or political committee has established and implemented a firewall that meets the requirements of paragraphs (h)(1) and (h)(2) of this section.” 11 C.F.R. § 109.21(h). Those requirements are: “(1) The firewall must be designed and implemented to prohibit the flow of information between employees or consultants providing services for the person paying for the communication and those employees or consultants currently or previously providing services to the candidate who is clearly identified in the communication ...; and (2) The firewall must be described in a written policy that is distributed to all relevant employees, consultants, and clients affected by the policy.” Id. According to the regulation, “[t]his safe harbor provision does not apply if specific information indicates that, despite the firewall, information about the candidate’s or political party committee’s campaign plans, projects, activities, or needs that is material to the creation, production, or distribution of the communication was used or conveyed to the person paying for the communication.” Id.

Shays challenged this regulation, arguing that it was so vague as to invite near-certain circumvention, undermining BCRA’s purpose, and that the Commission failed not only to justify it, but also to explain why it changed its mind after rejecting a similar provision in 2003, violating the APA. The district court agreed with both arguments. Shays III, 508 F.Supp.2d at 53-56.

Challenging the district court’s ruling and acknowledging that the regulation provides few details on what constitutes an acceptable firewall, the FEC argues that “a firewall is more effective if established and implemented by each organization in light of its specific organization, clients, and personnel.” 71 Fed.Reg. at 33,206. The Commission emphasizes that “[a]n organization cannot come within the firewall safe harbor simply by alleging that it has an internal firewall”; rather, “[a]n entity seeking to use the firewall safe harbor must be ‘prepared to provide reliable information ... about [its] firewall, and how and when the firewall policy was distributed and implemented.’ ” Appellant’s Open *930 ing Br. 33 (quoting 71 Fed.Reg. at 33,207). Moreover, the FEC insists, it provided a good reason for implementing the safe harbor: to make it easier for candidates and independent organizations to hire consultants, vendors, and former employees— thus facilitating protected speech — without fear of being falsely accused of improper coordination. See 71 Fed.Reg. at 33,206. And it claims it did explain why it has now adopted a firewall safe harbor despite rejecting a similar proposal in 2003, namely in the interim it approved a firewall created by EMILY’s List and found it sufficient to protect against coordination. See id.; Coordinated Communications: Proposed Rules, 70 Fed.Reg. 73,946, 73,955 (2005).

Though we think this a close question, we agree with the FEC. The district court and Shays are undeniably correct that the regulation is vague as to what constitutes an acceptable firewall, but “when Congress has not specified the level of specificity expected of the agency,” as here, “the agency is entitled to broad deference in picking the suitable level.” Cement Kiln Recycling Coal. v. EPA, 493 F.3d 207, 217 (D.C.Cir.2007) (citation omitted). Moreover, “[t]he APA does not require that all the specific applications of a rule evolve by further, more precise rules rather than by adjudication.” Shalala v. Guernsey Mem’l Hosp., 514 U.S. 87, 96, 115 S.Ct. 1232, 131 L.Ed.2d 106 (1995). Thus, there is no “basis for suggesting that the agency has a statutory duty to promulgate regulations that, either by default rule or by specification, address every conceivable question.” Id. Instead, the Commission has authority to flesh out its rules through adjudications and advisory opinions. In addition, the Commission’s sensible conclusion that firewalls will be “more effective if established and implemented by each organization in light of its specific organization, clients, and personnel,” 71 Fed.Reg. at 33,206, represents just the kind of agency expert judgment to which we owe deference. See, e.g., North Carolina v. FERC, 112 F.3d 1175, 1189 (D.C.Cir.1997) (“So long as the Commission has examined the relevant data and provided a reasoned explanation supported by a stated connection between the facts found and the choices made, we will defer to the agency’s expertise.” (citation omitted)). Shays doubts whether the Commission will enforce the safe harbor provision in a way that actually requires meaningful firewalls, but as a court reviewing this facial challenge we must presume that the Commission will enforce its rule in good faith. See Sullivan v. Everhart, 494 U.S. 83, 94, 110 S.Ct. 960, 108 L.Ed.2d 72 (1990) (holding that in facial challenges to regulations courts must presume agencies will implement them in good faith).

We also believe that the FEC adequately justified the rule and its departure from past practice. Hardly contrary to BCRA, the regulation makes it easier for candidates and organizations to engage in protected speech by helping them hire consultants and employees without fear of false accusations of coordination. Moreover, the Commission’s favorable experience with the EMILY’s List firewall represents a perfectly reasonable basis for its change of heart since the 2003 rulemaking.

III.

As part of its effort to reduce the influence of soft money, BCRA requires that all “federal election activity” be paid for with either hard money or “Levin funds” — limited contributions to state parties specifically earmarked for “federal election activity.” See 2 U.S.C. § 441i(b)(2); see also Shays II, 414 F.3d at 112-13. The statute defines federal election activity as including “get-out-the-vote activity” (GOTV activity) and “voter registration activity,” but it leaves these terms undefined. 2 U.S.C. § 431(20). In 2003 the Commission issued *931 regulations defining GOTV and voter registration activity:

(2) Voter registration activity means contacting individuals by telephone, in person, or by other individualized means to assist them in registering to vote. Voter registration activity includes, but is not limited to, printing and distributing registration and voting information, providing individuals with voter registration forms, and assisting individuals in the completion and filing of such forms.

(3) Get-out-the-vote activity means contacting registered voters by telephone, in person, or by other individualized means, to assist them in engaging in the act of voting. Get-out-the-vote activity includes, but is not limited to:

(i) Providing to individual voters information such as the date of the election, the times when polling places are open, and the location of particular polling places; and

(ii) Offering to transport or actually transporting voters to the polls.

11 C.F.R. § 100.24(a).

In Shays I, the district court invalidated these definitions on procedural grounds. See 337 F.Supp.2d at 101-07 (holding that the FEC violated the APA’s notice requirements in promulgating these definitions because interested parties could not reasonably have anticipated the final rule-makings from the notice of proposed rule-making). “On remand, the Commission re-promulgated its regulations defining voter registration activity and GOTV activity (with minimal alterations to the definition of GOTV activity), and issued an expanded [explanation and justification].” Shays III, 508 F.Supp.2d at 63.

Shays again challenged the regulation, and the district court found that the definitions survived Chevron step one but failed Chevron step two because both left unaddressed “vast gray areafs]” of possible GOTV and voter registration activity, making it possible for state parties to circumvent the statute and frustrate BCRA’s purpose. See id. at 63-70. According to the district court, the definitions also violated the APA because the Commission gave no good reason for leaving such large gray areas. See id. at 66, 69-70.

Challenging the district court’s ruling, the FEC again emphasizes the deference to which it is entitled “when Congress has not specified the level of specificity expected of the agency.” Cement Kiln, 493 F.3d at 217 (citation omitted). We agree with the FEC that its decision to promulgate a somewhat vague regulation, in and of itself, runs afoul of neither BCRA nor the APA, for there is no “basis for suggesting that the agency has a statutory duty to promulgate regulations that, either by default rule or by specification, address every conceivable question.” Guernsey Mem.'l, 514 U.S. at 96, 115 S.Ct. 1232. Thus, the Commission has discretion to leave a large gray area and fill it in later through adjudication and advisory opinions. That said, we reject the regulation for other reasons.

As Shays explains, the FEC’s definitions of GOTV activity and voter registration activity create “two distinct loopholes.” Appellee’s Opening Br. 41. First, both definitions require that the party contacting potential voters actually “assist” them in voting or registering to vote, 11 C.F.R. § 100.24(a)(2)-(3), thus excluding efforts that actively encourage people to vote or register to vote and dramatically narrowing which activities are covered. Second, both definitions require the contact to be “by telephone, in person, or by other individualized means,” thus entirely excluding mass communications targeted to many people. Id. (emphasis added). As Shays points out:

*932 under the Commission’s construction, a state party within days of a federal election can send out multiple direct mailings to every potential voter sympathetic to its cause urging them to vote, and can blanket the state with automated telephone calls by celebrities identifying the date of the election and exhorting recipients to get out to vote, without being deemed to be engaged in GOTV activity. Likewise, large-scale efforts encouraging potential supporters to register to vote and directing them how they may do so are not “voter registration activities” under the Commission’s definitions. Indeed, the more people that a communication is intended to reach, and the more money the party spends, the less likely it is that the communication will be an “individualized means” of “assistance” subject to BCRA’s restrictions on [federal election activity].

Appellee’s Opening Br. 43. These examples are not merely hypothetical. In a recent advisory opinion, the FEC decided that letters and pre-recorded telephone calls directed to registered Democrats in Long Beach, California, encouraging them to vote in an upcoming election, did not count as GOTV activity because they provided no individualized information to any particular recipient. See FEC Advisory Op. 2006-19 (June 5, 2006).

The FEC’s restrictive definitions of GOTV activity and voter registration activity run directly counter to BCRA’s purpose, and the Commission has provided no persuasive justification for them. Indeed, though Shays has not argued as much here, we question whether these definitions could even survive at Chevron step one, for we doubt whether the meaning of GOTV activity and voter registration activity can plausibly be limited to individualized assistance. In any event, the definitions fail at Chevron step two because they conflict with BCRA’s purpose of “prohibiting soft money from being used in connection with federal elections.” McConnell, 540 U.S. at 177 n. 69, 124 S.Ct. 619. The regulation will allow the use of soft money for many efforts that influence federal elections, for as the Supreme Court observed in McConnell, “[cjommon sense dictates” that “any efforts [by state or local parties] that increase the number of like-minded registered voters who actually go to the polls” will “directly assist [a] party’s candidates for federal office.” Id. at 167-68, 124 S.Ct. 619.

Moreover, the only rationales the Commission gave for adopting its limited constructions of GOTV activity and voter registration activity were: (1) to ensure that mere exhortations to get out and vote or register to vote made at the end of a political event or speech would not count as federal election activity; and (2) to give clear guidance to state and local party organizations so they know what activities they can engage in. Definition of Federal Election Activity, 71 Fed.Reg. 8,926, 8,928-29 (2006). The first rationale is unpersuasive. As Shays points out, “a definition could surely be crafted that would exempt such routine or spontaneous speech-ending exhortations without opening a gaping loophole permitting state parties to use soft money to saturate voters with unlimited direct mail and robocalls that unquestionably benefit federal candidates.” Appellee’s Opening Br. 45. And the second rationale doesn’t even amount to an argument for a limited definition of GOTV activity and voter registration activity; instead, it’s an argument for a clear and detailed definition. But because any clear definition would satisfy the FEC’s goal of providing precise guidance — one that forbade any activity designed to get people to register or vote would be just as easy to follow as one that allowed unlimited GOTV and voter registration efforts— the desire for a clear rule, in and of itself, *933 provides no justification for this limited definition.

IV.

The single regulation the district court upheld-as to which Shays cross appeals— deals with soft-money solicitations by federal candidates at state party fundraising events. BCRA prohibits those seeking or holding federal office from “solicit[ing] ... funds in connection with an election for Federal office, including funds for any Federal election activity, unless the funds are subject to the limitations, prohibitions, and reporting requirements of this Act,” i.e., the funds solicited must not be soft money. 2 U.S.C. § 441i(e)(l)(A). It also prohibits federal candidates and officeholders from “soliciting] ... funds in connection with any election other than an election for Federal office or disbursing] funds in connection with such an election unless the funds” are hard money or Levin funds. Id. § 441i(e)(l)(B). The statute specifies, however, that “[n]otwithstand-ing” these prohibitions, “a candidate or an individual holding Federal office may attend, speak, or be a featured guest at a fundraising event for a State, district, or local committee of a political party.” Id. § 441i(e)(3). Asserting that this latter provision made the statute ambiguous, the FEC issued a regulation allowing federal candidates and officeholders to solicit soft money at state and local party fundraisers. See 11 C.F.R. § 300.64 (“Candidates and individuals holding Federal office may speak at [state and local party] events without restriction or regulation.” (emphasis added)).

In Shays I the district court found that although the regulation survived Chevron review, the FEC had failed to provide an adequate justification for it, violating the APA. See Shays I, 337 F.Supp.2d at 92-93. Choosing not to appeal this aspect of Shays I, the FEC instead issued a new notice of proposed rulemaking, took additional comments, and issued the same regulation with an expanded explanation. This time the district court found the FEC’s explanation satisfactory. See Shays III, 508 F.Supp.2d at 60-61. Shays now appeals, arguing that the regulation violates BCRA and the APA.

In our view, the regulation fails because it allows what BCRA directly prohibits. As noted above, section 441i(e)(l)(A) expressly prohibits federal candidates and officeholders from soliciting soft money, yet the Commission’s rule allows federal candidates and officeholders to do just that at state and local party fundraisers. See Chevron, 467 U.S. at 842, 104 S.Ct. 2778 (“If the intent of Congress is clear, that is the end of the matter... .”).

Contrary to the Commission’s position, section 441i(e)(3) — “a candidate or an individual holding Federal office may attend, speak, or be a featured guest at a fundraising event for a State, district, or local committee of a political party” — does nothing to make the statute’s prohibition on soft-money solicitations ambiguous. Rather, section (e)(3) merely clarifies that despite the statute’s ban on soliciting soft money, federal candidates may still “attend, speak, or be a featured guest” at state party events where soft money is raised, which the statute might otherwise be read as forbidding. Indeed, several factors demonstrate that section (e)(3) cannot plausibly be read to allow federal candidates to solicit soft money at state party events. Most important, when Congress wanted to create an exception to the ban on federal candidates soliciting soft money, it did so explicitly. Section 441i(e) contains three express exceptions to section (e)(l)(A)’s general prohibition on raising soft money. See 2 U.S.C. § 441i(e)(2) (allowing candidates for federal office who are also candidates for local or state office *934 to solicit soft money authorized under state law for their state or local campaign); id. § 441(e)(4)(A) (authorizing federal candidates to solicit soft money for certain nonprofit groups); id. § 441i(e)(4)(B) (authorizing candidates to solicit up to $20,000 per individual to fund state party GOTV and voter registration activities). Given these express exceptions, we have no basis for reading section 441i(e)(3) as creating an implied fourth exception. “Where Congress explicitly enumerates certain exceptions to a general prohibition, additional exceptions are not to be implied, in the absence of evidence of a contrary legislative intent,” none of which is present here. TRW Inc. v. Andrews, 534 U.S. 19, 28, 122 S.Ct. 441, 151 L.Ed.2d 339 (2001) (citation omitted). Moreover, these exceptions expressly allow “solicitation” of soft money, yet section 441i(e)(3) says only that federal candidates may “attend, speak, or be a featured guest” at state party fundraisers. The difference in terminology matters, for “Congress’ choice of different verbs to characterize the two situations is a choice which we properly take as evidence of an intentional differentiation.” Nat’l Insulation Transp. Comm. v. ICC, 683 F.2d 533, 537 (D.C.Cir.1982) (citation omitted). This is especially true because Congress repeatedly used the term “solicit” and “solicitation” in section 441i — over a dozen times — yet chose not to do so in section 441i(e)(3). Reading section 441i(e)(3) as allowing solicitation in light of the clear differences between it and other sections of the statute that expressly allow solicitation “inverts the usual canon that when Congress uses different language in different sections of a statute, it does so intentionally.” Fla. Pub. Telecomms. Ass’n v. FCC, 54 F.3d 857, 860 (D.C.Cir.1995).

V.

For the foregoing reasons, we affirm the district court with respect to the content standard for coordinated expenditures, the rule for when former employees/vendors may share material information, and the definitions of GOTV activity and voter registration activity. With respect to the firewall safe harbor provision and the rule allowing soft-money solicitations at state party events, we reverse and remand for further proceedings consistent with this opinion.

So ordered