Section 11(c) of the Cable Television Consumer Protection and Competition Act of 1992, Pub. L. No. 102-385, 106 Stat. 1460 ("1992 Cable Act"), amends 47 U.S.C. § 533 to direct the Federal Communications Commission to set two types of limits on cable operators. The first type is horizontal, addressing operators' scale: "limits on the number of cable subscribers a person is authorized to reach through cable systems owned by such person, or in which such person has an attributable interest." 47 U.S.C. § 533(f)(a)(1)(A). The second type is vertical, addressing operators' integration with "programmers" (suppliers of programs to be carried over cable systems): "limits on the number of channels on a cable system that can be occupied by a video programmer in which a cable operator has an attributable interest." 47 U.S.C. § 533(f)(a)(1)(B). The FCC has duly promulgated regulations. See 47 C.F.R. § 76.503-04. Petitioners Time Warner and AT&T challenge the horizontal limit as in excess of statutory authority, as unconstitutional infringements of their freedom of speech, and as products of arbitrary and capricious decisionmaking which violate the Administrative Procedure Act. Time Warner similarly challenges the vertical limit. Together with AT&T, Time Warner also challenges as arbitrary and capricious the rules for determining what counts as an "attributable interest." Concluding that the FCC has not met its burden under the First Amendment and, in part, lacks statutory authority for its actions, we remand for further consideration of both limits. In addition we vacate specific portions of the attribution rules as lacking rational justification.

Consumers Union also files a petition for review, which need not detain us long. It objects to the Commission's action to the extent that it continued a stay on enforcement of the horizontal limit. See Implementation of Section 11(c) of the Cable Television Consumer Protection and Competition Act of 1992, 14 F.C.C.R. 19098, 19127-28 p p 71-73 (1999) ("Third Report"). The Commission issued the stay after a district court found the statute underlying that limit unconstitutional, see Daniels Cablevision, Inc. v. United States, 835 F. Supp. 1 (D.D.C. 1993), and provided that in the event of Daniels's reversal the stay would end. See Implementation of Sections 11 and 13 of the Cable Television Consumer Protection and Competition Act of 1992, 8 F.C.C.R. 8565, 8609 p 109 (1993) ("Second Report"). We did reverse Daniels in Time Warner Entertainment Co. v. United States, 211 F.3d 1313 (D.C. Cir. 2000) ("Time Warner I"), so the stay ended automatically.1 Thus the stay issue is moot unless the issue posed is capable of repetition yet evading review. Even if we assume that the issue evades review, its recurrence is not probable enough to qualify it as "capable of repetition." See Spencer v. Kemna, 523 U.S. 1, 17 (1998) (requiring "a reasonable expectation that the same complaining party [will] be subject to the same action again") (internal citations omitted). Although we find here that the regulations fail constitutional scrutiny, the specific condition that led to the stay--a pending challenge to the statute's constitutionality-is highly unlikely to recur. We therefore find Consumers Union's claim moot and dismiss the petition.

* * *The horizontal rule imposes a 30% limit on the number of subscribers that may be served by a multiple cable system operator ("MSO"). See 47 C.F.R. § 76.503; Third Report 14 F.C.C.R. at 19119 p 55. Both the numerator and denominator of this fraction include only current subscribers to multichannel video program distributor ("MVPD") services. See id. at 19107-10 p p 20-25. Subscribers include not only users of traditional cable services but also subscribers to non-cable MVPD services such as Direct Broadcast Satellite ("DBS"),2 a rapidly growing segment of the MVPD market. See id. at 19110-12 p p 26-35. The Commission pointed out that under this provision the nominal 30% limit would allow a cable operator to serve 36.7% of the nation's cable subscribers if it served none by DBS. See id. at 19113 p 37 & n.82.3 In an express effort to encourage competition through new provision of cable, the Commission excluded from any MSO's numerator all new subscribers signed up by virtue of "overbuilding," the industry's term for cable laid in competition with a pre-existing cable operator. See id. at 19112-13 p p 34, 37. Further, subscribers to a service franchised after the rule's adoption (October 20, 1999) do not go into an MSO's numerator, even if not the result of an overbuild. See id. at 19112 p 33. As a result, the rule's main bite is on firms obtaining subscribers through merger or acquisition.

The vertical limit is currently set at 40% of channel capacity, reserving 60% for programming by non-affiliated firms. See 47 C.F.R. § 76.504; Second Report, 8 F.C.C.R. at 859394 p 68; Implementation of Section 11(c) of the Cable Television Consumer Protection and Competition Act of 1992, 10 F.C.C.R. 7364, 7368 p 14 (1995) ("Reconsideration Order"). Channels assigned to broadcast stations, leased access, and for public, educational, or governmental uses are included in the calculation of channel capacity. See id. at 7371-73 p p 2027. Capacity over 75 channels is not subject to the limit, so a cable operator is never required to reserve more than 45 channels for others (.60 x 75 = 45). See id. at 7374-76 p p 31-35.

As cable operators, Time Warner and AT&T "exercise[ ] editorial discretion in selecting the programming [they] will make available to [their] subscribers," Time Warner I, 211 F.3d at 1316, and are "entitled to the protection of the speech and press provisions of the First Amendment," Turner Broadcasting System, Inc. v. Federal Communications Commission, 512 U.S. 622, 636 (1994) ("Turner I") (quoting Leathers v. Medlock, 499 U.S. 439, 444 (1991)). The horizontal limit interferes with petitioners' speech rights by restricting the number of viewers to whom they can speak. The vertical limit restricts their ability to exercise their editorial control over a portion of the content they transmit.

In Time Warner I we upheld the statutory provisions against a facial attack, after finding them subject to intermediate rather than, as the cable firms argued, strict scrutiny. Time Warner I, 211 F.3d at 1316-22. The regulations here present a related but independent set of questions. Constitutional authority to impose some limit is not authority to impose any limit imaginable.

In briefs written before the issuance of Time Warner I, petitioners argued here for strict scrutiny. At oral argument they withdrew from this position and said, euphemistically, that they were "happy to stand on intermediate scrutiny." Because of that concession and, in any event, not seeing any distinction between the statute and the regulations for levelof-scrutiny purposes, we apply intermediate scrutiny. Under the formula set forth in United States v. O'Brien, 391 U.S. 367, 377 (1968), and reaffirmed by Turner Broadcasting System, Inc. v. Federal Communications Commission, 520 U.S. 180, 189 (1997) ("Turner II"), a governmental regulation subject to intermediate scrutiny will be upheld if it "advances important governmental interests unrelated to the suppression of free speech and does not burden substantially more speech than necessary to further those interests." Id. (quoting O'Brien, 391 U.S. at 377).

The interests asserted in support of the horizontal and vertical limits are the same interrelated interests that we found sufficient to support the statutory scheme in Time Warner I: "the promotion of diversity in ideas and speech" and "the preservation of competition." Time Warner I, 211 F.2d at 1319; see also Turner I, 512 U.S. at 662-64 (concluding that both qualify as important governmental interests). After a review of the legislative history, we concluded that Congress had drawn "reasonable inferences, based upon substantial evidence, that increases in the concentration of cable operators threatened diversity and competition in the cable industry." Time Warner I, 211 F.3d at 1319-20. But the FCC must still justify the limits that it has chosen as not burdening substantially more speech than necessary. In addition, in "demonstrat[ing] that the recited harms are real, not merely conjectural," Turner I, 512 U.S. at 664, the FCC must show a record that validates the regulations, not just the abstract statutory authority.

* * *

The FCC asserts that a 30% horizontal limit satisfies its statutory obligation to ensure that no single "cable operator or group of cable operators can unfairly impede ... the flow of video programming from the video programmer to the consumer," 47 U.S.C. § 533(f)(2)(A), while adequately respecting the benefits of clustering4 and the economies of scale that are thought to come with larger size. See Third Report, 14 F.C.C.R. at 19123-24 p 61. It interpreted this statutory language as a directive to prohibit large MSOs--either by the action of a single MSO or the coincidental or collusive actions of several MSOs--from precluding the entry into the market of a new cable programmer. See id. at 19116 p 43. In setting the limit at 30%, it assumed there was a serious risk of collusion. See id., Part VI, at 19113-25 p p 36-65. But while collusion is a form of anti-competitive behavior that implicates an important government interest, the FCC has not presented the "substantial evidence" required by Turner I and Turner II that such collusion has in fact occurred or is likely to occur; so its assumptions are mere conjecture. See Turner II, 520 U.S. at 195 (citing Turner I, 512 U.S. at 666). The FCC alternatively relies on its supposed grant of authority to regulate the non-collusive actions of large MSOs. Congress may indeed, under certain readingsof Turner I and Turner II, have the power to regulate the coincidental but independent actions of cable operators solely in the interest of diversity, but "[w]here an administrative interpretation of a statute invokes the outer limits of Congress' power, we expect a clear indication that Congress intended that result." Solid Waste Agency v. United States Army Corps of Eng'rs, __ U.S. __, 121 S. Ct. 675, 683 (2001). The 1992 Cable Act, as we shall see, instead expresses the contrary intention.

Part VI of the Third Report lays out the calculations that lead the FCC to the 30% limit. See Third Report, Part VI, 14 F.C.C.R. at 19113-25 p p 36-65. First the FCC determines that the average cable network needs to reach 15 million subscribers to be economically viable. See id. at 19114-16 p p 40-42. This is 18.56% of the roughly 80 million MVPD subscribers, and the FCC rounds it up to 20% of such subscribers. The FCC then divines that the average cable programmer will succeed in reaching only about 50% of the subscribers linked to cable companies that agree to carry its programming, because of channel capacity, "programming tastes of particular cable operators," or other factors. Id. at 19117-18 p 49. The average programmer therefore requires an "open field" of 40% of the market to be viable (.20/.50 = .40). See id. at 19117-18 p p 46-50.

Finally, to support the 30% limit that it says is necessary to assure this minimum, the Commission reasons as follows: With a 30% limit, a programmer has an "open field" of 40% of the market even if the two largest cable companies deny carriage, acting "individually or collusively." Id. at 19119 p 53. A 50% rule is inadequate because, if a duopoly were to result, "[t]he probability of tacit collusion is higher with 2 competitors than 3 competitors." Id. at 19118-19 p 51. Even if collusion were not to occur, independent rejections by two MSOs could doom a new programmer, thwarting congressional intent as the Commission saw it. See id. A 40% limit is insufficient for the same reason: "two MSOs, ... representing a total of 80% of the market, might decline to carry the new network" and leave only 20% "open," which by hypothesis is not enough (because of the 50% success rate). Id. at 19119 p 52. Although the Commission doesn't spell out the intellectual process, it is necessarily defining the requisite "open field" as the residue of the market after a programmer is turned down either (1) by one cable company acting alone, or (2) by a set of companies acting either (a) collusively or (b) independently but nonetheless in some way that, because of the combined effect of their choices, threatens fulfillment of the statutory purposes. We address the FCC's authority to regulate each of these scenarios in turn.

The Commission is on solid ground in asserting authority to be sure that no single company could be in a position singlehandedly to deal a programmer a death blow. Statutory authority flows plainly from the instruction that the Commission's regulations "ensure that no cable operator or group of cable operators can unfairly impede, either because of the size of any individual operator or because of joint actions of operators of sufficient size, the flow of video programming from the video programmer to the consumer." 47 U.S.C. S 533(f)(2)(A) (emphasis added). Constitutional authority is equally plain. As the Supreme Court said in Turner II: "We have identified a corresponding 'governmental purpose of the highest order' in ensuring public access to a multiplicity of information sources.' " 520 U.S. at 190 (quoting Turner I, 512 U.S. at 663); see also Time Warner Entertainment Co. v. Federal Communications Commission, 93 F.3d 957, 969 (D.C. Cir. 1996). If this interest in diversity is to mean anything in this context, the government must be able to ensure that a programmer have at least two conduits through which it can reach the number of viewers needed for viabilityindependent of concerns over anticompetitive conduct.

Assuming the validity of the premises supporting the FCC's conclusion that a 40% "open field" is necessary (a question that we need not answer here), the statute's express concern for the act of "any individual operator" would justify a horizontal limit of 60%. To reach the 30% limit, the FCC's action necessarily involves one or the other of two additional propositions: Either there is a material risk of collusive denial of carriage by two or more companies, or the statute authorizes the Commission to protect programmers against the risk of completely independent rejections by two or more companies leaving less than 40% of the MVPD audience potentially accessible. Neither proposition is sound.

First, we consider whether there is record support for inferring a non-conjectural risk of collusive rejection. Either Congress or the Commission could supply that record, and we take them in that order. We give deference to the predictive judgments of Congress, see Turner II, 520 U.S. at 195-96 (citing Turner I, 512 U.S. at 665), but Congress appears to have made no judgment regarding collusion. The statute plainly alludes to the possibility of collusion when it authorizes regulations to protect against "joint actions by a group of operators of sufficient size." 47 U.S.C. S 533(f)(2)(A) (emphasis added). But this phrase, while granting the Commission authority to take action in the event that it finds collusion extant or likely, is not itself a congressional finding of actual or probable collusion. Such findings have not been made. No reference to collusion appears in the Act's findings or policy, see 1992 Cable Act S 2, 106 Stat. at 1460-63, nor in the legislative history discussing the horizontal or vertical limits. See H.R. Rep. No. 102-628, at 40-43 (1992) ("House Report"); S. Rep. No. 102-92, at 24-29, 32-34, reprinted in 1991 U.S.C.C.A.N. 1133, at 1156-62, 1165-67 ("Senate Report"). It was thus appropriate for the FCC to describe Congress's reference to "joint" action as merely a "legislative assumption." Third Report, 14 F.C.C.R. at 19116 p 43 (emphasis added).

The Commission's own findings amount to precious little. It says only:

The legislative assumption [about joint action] is not unreasonable given an environment in which all the larger operators in the industry are vertically integrated so that all are both buyers and sellers of programming and have mutual incentives to reach carriage decisions beneficial to each other. Operators have incentives to agree to buy their programming from one another. Moreover, they have incentives to encourage one another to carry the same non-vertically integrated programming in order to share the costs of such programming.

Id. None of these assertions is supported in the record. The Commission never explains why the vertical integration of MSOs gives them "mutual incentive to reach carriage decisions beneficial to each other," what may be the firms' "incentives to buy ... from one another," or what the probabilities are that firms would engage in reciprocal buying (presumably to reduce each other's average programming costs). After all, the economy is filled with firms that, like MSOs, display partial upstream vertical integration. If that phenomenon implies the sort of collusion the Commission infers, one would expect the Commission to be able to point to examples. Yet it names none. Further, even if one accepts the proposition that an MSO could benefit from sharing the services of specific programmers, programming is not more attractive for this purpose merely because it originates with another MSO's affiliate rather than with an independent.

The only justification that the FCC offers in support of its collusion hypothesis is the economic commonplace that, all other things being equal, collusion is less likely when there are more firms. See Third Report 14 F.C.C.R. at 19118-19 p 51. This observation will always be true, although marginally less so for each additional firm; but by itself it lends no insight into the question of what the appropriate horizontal limit is. Turner I demands that the FCC do more than "simply 'posit the existence of the disease sought to be cured.' " Turner I, 512 U.S. at 664 (quoting Quincy Cable TV, Inc. v. Federal Communications Commission, 768 F.2d 1434, 1455 (D.C. Cir. 1985). It requires that the FCC draw "reasonable inferences based on substantial evidence." Turner I, 512 U.S. at 666. Substantial evidence does not require a complete factual record--we must give appropriate deference to predictive judgments that necessarily involve the expertise and experience of the agency. See Turner II 520 U.S. at 196, citing Federal Communications Commission v. National Citizens Comm. For Broadcasting, 436 U.S. 775, 814 (1978). But the FCC has put forth no evidence at all that indicates the prospects for collusion.

That having been said, we do not foreclose the possibility that there are theories of anti-competitive behavior other than collusion that may be relevant to the horizontal limit and on which the FCC may be able to rely on remand. See 47 U.S.C. S 533(f)(1). Indeed, Congress considered, among other things, the ability of MSOs dominant in specific cable markets to extort equity from programmers or force exclusive contracts on them. See 1992 Cable Act S 2(a)(4)-(5), 106 Stat. at 1460-61; Senate Report at 3, 14, 23-29, 32-34, reprinted in 1991 U.S.C.C.A.N. at 1135, 1146-47, 1156-62, 1165-67; House Report at 40-43. A single MSO, acting alone rather than "jointly," might perhaps be able to do so while serving somewhat less than the 60% of the market (i.e., less than the fraction that would allow it unilaterally to lock out a new cable programmer) despite the existence of antitrust laws and specific behavioral prohibitions enacted as part of the 1992 Cable Act, see 47 U.S.C. S 536, and the risk might justify a prophylactic limit under the statute. See Time Warner I, 211 F.3d at 1322-23. So the absence of any showing of a serious risk of collusion does not necessarily preclude a finding of a sufficient governmental interest in preventing unfair competition. (We express no opinion on whether exploitation of a monopoly position in a specific cable market to extract rents that would otherwise flow to programmers alone gives rise to an "important governmental interest" justifying a burden on speech.) But the FCC made no attempt to justify its regulation on these grounds.

We pause here to address an aspect of petitioners' statutory challenge that is relevant to a showing of non-conjectural harm. Congress required that in setting the horizontal limit, the FCC "take particular account of the market structure ...

including the nature and market power of the local franchise." 47 U.S.C. S 533(f)(2)(C). Petitioners assert that the Commission's failure to take adequate account of the competitive pressures brought by the availability and increasing success of DBS make the horizontal limit arbitrary and capricious. Although DBS accounts for only 15.4% of current MVPD households, the annual increase in its total subscribership is almost three times that of cable (nearly three million additional subscribers over the period June 1999 to June 2000, as against one million for cable). See Seventh Annual Report p p 6-8. To the extent petitioners argue that the horizontal limit must fail because market share does not equal market power, they misconstrue the statutory command. The Commission is not required to design a limit that falls solely on firms possessing market power.5 The provision is directed to the Commission's intellectual process, and requires it, in evaluating the harms posedby concentration and in setting the subscriber limit, to assess the determinants of market power in the cable industry and to draw a connection between market power and the limit set.

It follows naturally from our earlier discussion that we do not believe the Commission has satisfied this obligation. Having failed to identify a non-conjectural harm, the Commission could not possibly have addressed the connection between the harm and market power. But the assessment of a real risk of anti-competitive behavior--collusive or not--is itself dependent on an understanding of market power, and the Commission's statements in the Third Report seem to ignore the true relevance of competition. In changing the calculation of the horizontal limit to reflect subscribers instead of homes at which a service is available, for instance, the Commission wrote:

[W]hether subscribership or homes passed data is used is largely a mechanical issue in terms of the market power issue.... As the market develops in terms of competition we believe ... that an operator's actual number of subscribers more uniformly and accurately reflects power in the programming marketplace.

Third Report, 14 F.C.C.R. at 19108 p 22.

But normally a company's ability to exercise market power depends not only on its share of the market, but also on the elasticities of supply and demand, which in turn are determined by the availability of competition. See AT&T Corp. v. Federal Communications Commission, 236 F.3d 729, 736 (D.C. Cir. 2001). If an MVPD refuses to offer new programming, customers with access to an alternative MVPD may switch. The FCC shows no reason why this logic does not apply to the cable industry. Indeed, its most recent competition report suggests that it does. According to the Commission, "several very small and rural cable systems have used a variety of schemes to add digital channels, expand their program offerings, and take preemptive action against aggressive DBS marketing." Seventh Annual Report p 67.

Given the substantial changes in the cable industry since publication of the Third Report in 1999 and our reversal on other grounds, there is little point in our reviewing the Commission's assessment of then-existing market power of cable MVPDs. But whatever conclusions are to be drawn from the new data, it seems clear that in revisiting the horizontal rules the Commission will have to take account of the impact of DBS on that market power. Already when the Third Report was written, DBS could be considered to "pass every home in the country." Third Report, 14 F.C.C.R. at 19107-08 p 20. The technological and regulatory changes since then appear only to strengthen petitioners' contention. See Seventh Annual Report p p 60-82, 140.

With the risk of collusion inadequately substantiated to support the 30% limit and no attempt to find other anticompetitive behavior, there remains the Commission's alternative ground--that programming choices made "unilaterally" by multiple cable companies, Third Report, 14 F.C.C.R. at 19118-19 p 51; see also id. at 19119 p 53 ("individually"), might reduce a programmer's "open field" below the 40% benchmark. The only support the Commission offered for regulation based on this possibility was the idea that every additional chance for a programmer to secure access would enhance diversity:

[T]he 30% limit serves the salutary purpose of ensuring that there will be at least 4 MSOs in the marketplace. The rule thus maximizes the potential number of MSOs that will purchase programming. With more MSOs making purchasing decisions, this increases the likelihood that the MSOs will make different programming choices and a greater variety of media voices will therefore be available to the public.

Id. p 54. Petitioners challenge the FCC's authority to regulate for this purpose on both constitutional and statutory grounds.

We have some concern how far such a theory may be pressed against First Amendment norms. Everything else being equal, each additional "voice" may be said to enhance diversity. And in this special context, every additional splintering of the cable industry increases the number of combinations of companies whose acceptance would in the aggregate lay the foundations for a programmer's viability. But at some point, surely, the marginal value of such an increment in "diversity" would not qualify as an "important" governmental interest. Is moving from 100 possible combinations to 101 "important"? It is not clear to us how a court could determine the point where gaining such an increment is no longer important. And it would be odd to discover that although a newspaper that is the only general daily in a metropolitan area cannot be subjected to a right of reply, see Miami Herald Publishing Co. v. Tornillo, 418 U.S. 241 (1974), it could in the name of diversity be forced to self-divide. Certainly the Supreme Court has not gone so far.

We need not face that issue, however, because we conclude that Congress has not given the Commission authority to impose, solely on the basis of the "diversity" precept, a limit that does more than guarantee a programmer two possible outlets (each of them a market adequate for viability). We analyze the agency action under the familiar framework of Chevron USA, Inc. v. National Resources Defense Council, Inc., 467 U.S. 837 (1984). If we find (using traditional tools of statutory interpretation) that Congress has resolved the question, that is the end of the matter. FDA v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 132 (2000); National Resources Defense Council, Inc. v. Browner, 57 F.3d 1122, 1125 (D.C. Cir. 1995). We must place the statutory language in context and "interpret the statute 'as a symmetrical and coherent regulatory scheme.' " Brown & Williamson, 529 U.S. at 133.

We begin with the statutory language. The relevant section requires the FCC to

ensure that no cable operator or group of cable operators can unfairly impede, either because of the size of any individual operator or because of joint actions by a group of operators of sufficient size, the flow of video programming from the video programmer to the consumer.

47 U.S.C. S 533(f)(2)(A).

The language addresses only "unfair[ ]" impediments to the flow of programming. The word "unfair" is of course extremely vague. Certainly, the action of several firms that is "joint," in the sense of collusive, may often entail unfairness of a conventional sort. The statute goes further, plainly treating exercise of editorial discretion by a single cable operator as "unfair" simply because that operator is the only game in town. (And Time Warner I authoritatively determines that the government is constitutionally entitled to impose limits solely on that ground.) But we cannot see how the word unfair could plausibly apply to the legitimate, independent editorial choices of multiple MSOs. A broad interpretation is plausible only for actions that impinge at least to some degree on the interest in competition that lay at the heart of Congress's concern.6 The Commission's reading of the clause effectively deletes the word "joint" and opens the door to illimitable restrictions in the name of diversity.

Looking at the statute as a whole does little to support the FCC's position. The "interrelated interests" of promoting diversity and fair competition run throughout the 1992 Cable Act's various provisions. Turner II, 520 U.S. at 189.7 But despite the duality of interests at work in this section, see Time Warner I, 211 F.3d at 1319, it is clear from the structure of the statute that Congress's primary concern in authorizing ownership limits is "fair" competition. The statute specifies, after all, that these regulations are to be promulgated "[i]n order to enhance effective competition." 47 U.S.C. S 533 (f)(1). In only two of the other sections of the 1992 Cable Act does Congress specify a dominant purpose.8 This statement of purpose supports a reading that sharply confines the authority to regulate solely in the interest of diversity.

The FCC points to the statutory findings that the "cable industry has become highly concentrated" and that "the potential effects of such concentration are barriers to entry for new programmers and a reduction in the number of media voices available to consumers." Third Report, 14 F.C.C.R. at 19118-19 p 51, 1992 Cable Act S 2(a)(4), 106 Stat. at 1460. But reference to a congressional finding cannot overcome the clear language and purpose of the actual provision. The quoted finding stands as little more than support for the proposition that Congress was concerned with the possibilities for market failure and the possible impact on new programmers. The legislative history also offers little. Again, the fact that Congress's interest in anti-competitive behavior may have been animated by an interest in preserving diversity doesn't give the FCC carte blanche to cobble cable operators in the name of the latter value alone. After all, Congress also sought to "ensure that cable operators continue to expand, where economically justified, their capacity," 1992 Cable Act S 2(b)(3), 106 Stat. at 1463, and it specifically directed the FCC, in setting the ownership limit, to take into account the "efficiencies and other benefits that might be gained through increased ownership or control." 47 U.S.C. S 533(f)(2)(D).

On the record before us, we conclude that the 30% horizontal limit is in excess of statutory authority. While a 60% limit might be appropriate as necessary to ensure that programmers had an adequate "open field" even in the face of rejection by the largest company, the present record supports no more. In addition, the statute allows the Commission to act prophylactically against the risk of "unfair" conduct by cable operators that might unduly impede the flow of programming, either by the "joint" actions of two or more companies or the independent action of a single company of sufficient size. But the Commission has pointed to nothing in the record supporting a non-conjectural risk of anticompetitive behavior, either by collusion or other means. Accordingly, we reverse and remand with respect to the 30% rule.

* * *The FCC presents its 40% vertical limit as advancing the same interests invoked in support of its statutory authority to adopt the rule: diversity in programming and fair competition. As with the horizontal rules the FCC must defend the rules themselves under intermediate scrutiny and justify its chosen limit as not burdening substantially more speech than necessary. Far from satisfying this test, the FCC seems to have plucked the 40% limit out of thin air.

The FCC relies almost exclusively on the congressional findings that vertical integration in the cable industry could "make it difficult for non-cable affiliated ... programmers to secure carriage on vertically integrated cables systems" and that "vertically integrated program suppliers have the incentive and the ability to favor their affiliated cable operators ... and program distributors." Second Report, 8 F.C.C.R. at 8583 p 41 (citing 1992 Cable Act S 2(a)(5), 106 Stat. at 1460). Regulatory limits in response to these consequences would "increase the diversity of voices available to the public." Second Report, 8 F.C.C.R. at 8583-84 p 42 (citing Senate Report at 80, reprinted in 1991 U.S.C.C.A.N. at 1213). In Time Warner I we thought these findings strong enough to overcome the First Amendment challenge to the relevant provision of the 1992 Cable Act. In doing so, we held that such a prophylactic rule was not "rendered unnecessary merely because preexisting statutes [such as the antitrust laws and the antidiscrimination provisions of the 1992 Cable Act] impose behavioral norms." Time Warner I, 211 F.3d at 1322-23. Beyond that we did not assess the appropriateness of the burden on speech. We upheld no specific vertical limit--none was before us.

We recognize that in drawing a numerical line an agency will ultimately indulge in some inescapable residue of arbitrariness; even if 40% is a highly justifiable pick, no one could expect the Commission to show why it was materially better than 39% or 41%. See Missouri Public Service Comm'n v. FERC, 215 F.3d 1, 5 (D.C. Cir. 2000). But to pass even the arbitrary and capricious standard, the agency must at least reveal " 'a rational connection between the facts found and the choice made.' " Dickson v. Secretary of Defense, 68 F.3d 1396, 1404-05 (D.C. Cir. 1995) (quoting Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983). Here the FCC must also meet First Amendment intermediate scrutiny. Yet it appears to provide nothing but the conclusion that "we believe that a 40% limit is appropriate to balance the goals." See Second Report, 8 F.C.C.R. at 8593-95 p 68. What are the conditions that make 50% too high and 30% too low? How great is the risk presented by current market conditions? These questions are left unanswered by the Commission's discussion.

The FCC argued before us that no MSO has yet complained that the 40% vertical limit has required it to alter programming. This is no answer at all, as it says nothing about plans that the rule may have scuttled. Petitioners responded that their subsidiaries frequently must juggle their channel lineups to stay within the cap. Furthermore, it appears uncontested that AT&T's merger with MediaOne brings the vertical limits into play. See In the Matter of Applications for Consent to the Transfer of Control of Licenses and Section 214 Authorizations from MediaOne Group, Inc. to AT&T Corporation, 15 F.C.C.R. 9816 (2000).

In fairness, the FCC does make an attempt to review some relevant conditions. See Second Report, 8 F.C.C.R. at 858385 p p 41-45. The FCC cites the House Report's conclusion that "some" vertically integrated MSOs favor their affiliates and "may" discriminate against others. Id. at 8583-84 p 42 (citing House Report at 43). But it also notes a report that none of the top five MSOs "showed a pattern" of favoring their affiliates. Id. at 8584 p 43. Indeed, the FCC concludes that "vertical relationships had increased both the quality and quantity of cable programming services." Id. p 44. But still it settled on a limit of 40%. There is no effort to link the numerical limits to the benefits and detriments depicted. Further, given the pursuit of diversity, one might expect some inquiry into whether innovative independent originators of programming find greater success selling to affiliated or to unaffiliated programming firms, but there is none.

Quite apart from the numerical limit vel non, petitioners attack the Commission's refusal to exclude from the vertical limit cable operators that are subject to effective competition. The FCC had proposed exempting cable operators who met the definition of effective competition provided by S 623 of the Communications Act of 1934. See Implementation of Sections 11 and 13 of the Cable Television Consumer Protection and Competition Act of 1992, 8 F.C.C.R. 6828, 6862 p 231 (1993) ("First Report"); see also 47 U.S.C. S 543(l )(1) (defining the categories of cable operators that are not subject to rate regulation under that section).9 Of course our decision in Time Warner I acknowledged the existence of incentives to use affiliated programming. 211 F.3d at 1322. For example, even where an unaffiliated supplier offered a better costquality trade-off, a company might be reluctant to ditch or curtail an inefficient in-house operation because of the impact on firm executives or other employees, or the resulting spotlight on management's earlier judgment. But petitioners argue, quite plausibly, that exposure to competition will have an impact on a cable company's ability to indulge in favoritism for in-house productions. After all, while reliance on in house suppliers offering an inferior price-quality trade-off will reduce a monopolist's profits, it may threaten a competitive firm's very survival. This analysis is not foreign to the Commission, which endorsed it when proposing the exemption:

We believe that this proposal is appropriate since effective competition will preclude cable operators from exercising the market power which originally justified channel occupancy limits. Where systems face effective competition, their incentive to favor an affiliated programmer will be replaced by the incentive to provide programming that is most valued by subscribers.

First Report, 8 F.C.C.R. at 6862 p 231.

The FCC makes two arguments to justify its refusal to exempt MVPDs that are subject to effective competition. First, it says that the definition of competition provided by 47 U.S.C. S 543 was "not adopted for this specific purpose" but rather for relief from rate regulation. See Reconsideration Order 10 F.C.C.R. at 7379 p 47. Indeed, we have recognized that one of the ways in which the statutory standard is met may be surprisingly defective as a mark of real competition. See Time Warner Entertainment Co., L.P. v. Federal Communications Commission, 56 F.3d 151, 166 (D.C. Cir. 1995) (MVPDs satisfying subsection (A) of 47 U.S.C. S 543(l )(1) (low penetration) may do so more as a result of geography than competition). But the Commission is free to carve out subsections that are truly pertinent to competition, as it had proposed. See First Report, 8 F.C.C.R. at 8662-63 p 232; Second Report, 8 F.C.C.R. at 8602 p 85.

Of course competition that is adequate to justify dispensing with rate regulation could still leave an undue likelihood of improper favoritism for affiliated programmers. But the possible failure of readily available criteria does not itself justify the use of so blunt a blade. Congress expressly directed the Commission to take "particular account of the market structure..., including the nature and market power of the local franchise." 47 U.S.C. S 533(f)(2)(C) (emphasis added). Because competition raises the stakes for a firm that sacrifices the optimal price-quality trade-off in its acquisition of programming, the issue seems to trigger the legislative directive. Yet the Commission seems to ignore its own conclusions about cable companies' incentives and constraints, and the dynamics of the programming industry. See First Report, 8 F.C.C.R. at 6862 p 231. If the criteria of S 543(l )(1) are unsuitable, the Commission can consider concepts of effective competition that it finds more apt for these purposes.

Second, the FCC comments that if a competing MVPD favored its own affiliated programmers, the presence of competition would have no tendency to create room for independent programmers. See Reconsideration Order 10 F.C.C.R at 7379 p 47. But this theory seems contradicted by the Commission's own observation, mentioned earlier, that no vertically integrated MPVD has complained of reaching the 40% limit. Vertically integrated MVPDs evidently use loads of independent programming. Further, although cable operators continue to expand their interests in programmers, "[t]he proportion of vertically integrated channels ... continue[d] to decline" for each of the last two years. Sixth Annual Report, 15 F.C.C.R. at 1058-59 p 181, Seventh Annual Report p 173 (emphasis added). Even if competing MSOs filled all of their channels with affiliates' products (as unlikely as that seems), the Commission nowhere explains why, in the pursuit of diversity, the independence of competing vertically integrated MVPDs is inferior to the independence of unaffiliated programmers. In any event, the Commission's point here does not respond to the intuition that competition spurs a firm's search for the best price-quality trade-off.

In its brief the Commission adds the argument that truly effective competition under S 543(l )(1) existed only for a tiny fraction of cable systems. Indeed, it said in its Sixth Annual Report that of the nation's 33,000 cable community units, only 157 satisfy the definition through being in a market offering more than one wireline MVPD. Sixth Annual Report, 15 F.C.C.R. at 1045-46 p 142. (In the Seventh Annual Report we learn that now 330, or 1% of the total, meet the competition standard through exposure to another MVPD; in this report the qualifier "wireline" is absent. See Seventh Annual Report p 138.) But in determining whether or not the regulations burden substantially more speech than necessary, it is a weak move to point to the paucity of MVPDs facing competition if, as seems the case, it is easy to exempt them from the limit.

We find that the FCC has failed to justify its vertical limit as not burdening substantially more speech than necessary. Accordingly, we reverse and remand to the FCC for further consideration.

* * *

We turn, finally, to several aspects of the rules for attributing ownership for purposes of the horizontal and vertical limits, recently revised by the FCC and challenged by petitioners. See Implementation of the Cable Television Consumer Protection and Competition Act of 1992, 14 F.C.C.R. 19014 (1999) ("Attribution Order"). Petitioners suggest that these rules affect their ability to "speak" to subscribers because of their connection to the horizontal and vertical limits. But petitioners' speech rights are implicated only where their interest allows them to exercise editorial control, in which case attribution would be proper and it is the horizontal or vertical limit that constrains speech. The only effect of the attribution rules where no control is exercised is to limit the extent of petitioners' investments in a particular class of companies. We therefore review the agency actions under the APA standards, to determine whether they are "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." See 5 U.S.C. S 706(2)(A).

The FCC adopted as its starting point the pre-existing rules for attributing ownership of broadcast television stations, finding that the purposes of the rules are the same. See Attribution Order, 14 F.C.C.R. at 19030 p 35; Second Report 8 F.C.C.R. at 8577-79, 8593-96 p p 30-35, 56-63. Under that standard, attribution is triggered by ownership of 5% of the voting shares of a company, with various exceptions. See Attribution of Ownership Interests, 97 FCC 2d 997 (1984). Because the decisions in the Attribution Order tracked, to a large degree, similar decisions related to the broadcast attribution rules, the FCC incorporated by reference much of the reasoning from the broadcast orders. See Attribution Order, 14 F.C.C.R. at 19015-16 p 1.

Petitioners challenge the sufficiency and relevance of the Commission's evidence in support of the 5% attribution rule and its failure to adopt an alternative proposed by cable industry interests. They begin by asserting that the FCC improperly relied on two studies that were mentioned neither in the FCC's notice nor in any party's submission. See Notice of Proposed Rulemaking, 13 F.C.C.R. 12990 (1998). Although it is true that an agency cannot rest a rule on data " 'that, [in] critical degree, is known only to the agency,' " Community Nutrition Institute v. Block, 749 F.2d 50, 57 (D.C. Cir. 1984) (quoting Portland Cement Ass'n v. Ruckelshaus, 486 F.2d 375, 393 (D.C. Cir. 1973); see also International Union, UAW v. OSHA, 938 F.2d 1310, 1324-35 (D.C. Cir. 1991) (approving reliance on documents not exposed to comment if not "vital" to agency's support for rule), obviously not every cited document is "critical."

Here, although petitioners assert that the studies were the sole evidence cited by the FCC, the Commission also relied on a survey, used to support the 1984 broadcast attribution rules, showing that in widely held corporations, an owner of 5% or more would ordinarily be one of the two or three largest shareholders. See Attribution Order, 14 F.C.C.R. at 19034 p 46; Block, 749 F.2d at 58 (1984) (new information "expanded on and confirmed information"). The earlier rulemaking had inferred that with such ownership a holder of 5% or more would be able "to potentially affect the outcome of elective or discretionary decisions and to command the attention of management." Attribution of Ownership Interests, 97 FCC 2d at 1005-06 p 14. This hardly seems implausible. Presumably an owner of 5% or more typically has enough of an interest to justify the burden of informing himself about the company's activities and trying to influence (or supplant) management, a fact that management would bear in mind in deciding to whose exhortations it should pay attention. Petitioners have not pointed to any evidence suggesting that the FCC's survey is no longer accurate, or that the conclusion they draw from it has been undermined.

Furthermore, in attacking the relevance of the new studies, the petitioners fail to acknowledge that the FCC sought a rule that would capture "influence or control," not just control. Attribution Order, 14 F.C.C.R. at 19015-16 p 1 (emphasis added). The Commission specifically noted that a "firm does not need actual operational control over ... a company in order to exert influence." Id. at 19030-31 p 36. This distinction also tends to rebut petitioners' critique of the Commission's reliance on the Securities and Exchange Commission's requirement that investors report to the SEC when their holdings exceed 5% of any class of a firm's shares. See 15 U.S.C. S 78m(d)(1). The FCC noted that the purpose of the SEC's requirement was to alert investors to potential changes in control, and reasoned that this was similar to its own purpose in the attribution rules, encompassing not merely control but influence. See Attribution Order, 14 F.C.C.R. at 19035 p 49 (citing Securities and Exchange Comm'n v. Savoy Indus., Inc., 587 F.2d 1149 (D.C. Cir. 1978)).

Finally, petitioners contend that it was arbitrary for the FCC to reject a "control certification" approach, such as it adopted for partnerships, under which a partner can avoid attribution if (but only if) it certifies to the absence of certain relationships that might betoken control. In this argument, petitioners make a classic apples-and-oranges mix, since the bases that they proposed for self-certification, see Attribution Order at 19024 p 22, are quite different from those adopted by the Commission for partnerships, see id. at 19038 p 57 n.163. Even if corporations and partnerships were virtually identical, the Commission would hardly be guilty of self contradiction if it rejected certification scheme A for corporations and accepted certification scheme B for partnerships. In any event, for corporations the Commission rejected a case-by-case approach on conventional grounds, observing that a bright-line rule was to be preferred because it "reduces regulatory costs, provides regulatory certainty, and permits planning of financial transactions." Id. at 19035 p 48; see also id. at 19031 p 38. Given an agency's very broad discretion whether to proceed by way of adjudication or rulemaking, see N.L.R.B. v. Bell Aerospace Co., 416 U.S. 267, 294 (1974), and the reasonableness of the 5% criterion, we doubt there was need to explain further. The Commission did, however, observe that the certification proposals offered did "not take into account the variety of ways that an investor may exert influence or control over a company." Id. at 19030-31 p 36. And it implicitly distinguished its treatment of partnerships when it said that a limited partner's influence may not be proportional to equity interest "because the extent of its power may be modified by contract." Id. at 19039 p 61. Indeed, the Commission's certification rules for partnerships require voting restrictions that would not normally, and perhaps could not, be paralleled in the corporate world (such as abnegation of any power to remove the general partner except under extremely limited circumstances, see id. at 19038 p 57 n.163). We find the Commission's discussion adequate.

We also uphold the FCC's adoption of an "equity-and-debt" rule to capture "nonattributable investments that could carry the potential for influence." Id. at 19047 p 83. The rule triggers attribution "to an investor that holds an interest that exceeds 33% of the total asset value (equity plus debt) of the applicable entity." Id. at 19046-47 p 82.10 Petitioners attack the sufficiency of evidence to support both the rule itself and the selection of 33% as limit. They observe in particular that the Commission's own claims seem to depend on combinations of debt and equity with contractual rights. See, e.g., id. at 19047 p 83. But the Commission explicitly relied on an earlier rulemaking, see, e.g., id. at 19047 p 83, citing Review of the Commission's Regulations Governing Attribution of Broadcast and Cable/MDS Interests, 14 F.C.C.R. 12559 (1999) ("Broadcast Attribution Order"), which in turn relied on academic literature, see id. at 12589 p 62 nn.132, 134. Petitioners offer no critique of that literature's relevance, and it is not our role to launch one on our own. So we must accept the Commission's basic finding.

Although petitioners independently attack the Commission's selection of 33% as the debt-and-equity limit, we are constrained in our review by the sketchy character of their attack on the basic theory. The Commission's choice of 33% certainly has modest support. It recited the numbers offered by various parties, which ranged from 10% to 50%, in some cases with variations dependent on the presence of special contract provisions. Attribution Order, 14 F.C.C.R. at 19048-49 p p 85-86. Obviously 33% is not far off the median, but, as the Commission says nothing to evaluate the numbers recited, that tells us little.

The Commission also cited its own past decisions, saying that it had used the same percentage for the parallel rule in its broadcast cross-interest policy, and that there it "does not appear to have had a disruptive effect," id. at 19048-49 p 86, though without indicating what (if any) assessment it had made. And it referred to two prior adjudications. Id. (citing Cleveland Television Corp., 91 FCC 2d 1129 (Rev. Bd. 1982), aff'd, Cleveland Television Corp. v. Federal Communications Commission, 732 F.2d 962 (D.C. Cir. 1984), and Roy M. Speer, 11 F.C.C.R. 18393 (1996)). In Cleveland Television it had simply held that a one-third preferred stock interest conferred " 'insufficient incidents of contingent control' " under various policies, Attribution Order, 14 F.C.C.R. at 1904849 p 86 (emphasis added). In Roy M. Spear, it relied on Cleveland Television to impose a 33% ownership on a creditor's purchase option, but deferred establishment of any general rule. See 11 F.C.C.R. 18393 p 126 n.26. These prior adjudications provide thin affirmative support for the choice of 33%, though they at least suggest that the Commission has not indulged in self-contradiction. But given the absence of a real probe of the Commission's underlying reasoning for having the restriction at all, the inevitable difficulty in picking such a number, and the deference due the Commission, we cannot find the choice of 33% arbitrary. See Cassell v. Federal Communications Commission, 154 F.3d 478, 485 (D.C. Cir. 1998).

Petitioners also challenge the Commission's elimination of an exemption that prevailed in the broadcast attribution rules at the time the cable attribution rules were promulgated. In the broadcast context, an otherwise covered minority shareholder in a company with a single majority shareholder was exempted, on the principle that in such a case the minority shareholder would ordinarily not be able to direct the activities of the company.11 See Attribution of Ownership Interests, 97 FCC 2d at 1008-09 p 21; Attribution Order, 14 F.C.C.R. at 19044-46 p p 74-81. There were contentions in the Broadcast Attribution Order proceeding that the majority shareholder exemption was being used evasively. See 14 F.C.C.R. at 12574-75 p 29. The Commission neither rejected nor accepted these claims, but retained the exemption. See id. at p 36. In dispatching the exemption here, the Commission cited only its concern that a minority shareholder might be able to exercise influence even in these circumstances, the "lack of a record ... that the exemption should be retained," and the fact that no one claimed to be using the exemption. Attribution Order, 14 F.C.C.R. at 19046 p 81.

The Commission argues here that petitioners lack standing because they have not shown that they are using the exemption. Again, the FCC disregards the impact the rule can have on investment plans. Petitioners say that they are continually reviewing investment opportunities and that they are constrained by the absence of the single majority exemption. See supra p. 1137. This is an actual "injury in fact" that is "fairly traceable" to the administrative action. See Lujan v. Defenders of Wildlife, 504 U.S. 555, 561 (1992); see also Committee for Effective Cellular Rules v. Federal Communications Commission, 53 F.3d 1309, 1315-16 (D.C. Cir. 1995). And of course the absence of current use is no reason to delete an exemption. Removal of the exemption is a tightening of the regulatory screws, if perhaps a minor one. It requires some affirmative justification, cf. State Farm, 463 U.S. at 41-42 (requiring justification for removal of a restriction), yet the Commission effectively offers none. Its "concern" about the possibility of influence would be a basis, if supported by some finding grounded in experience or reason, but the Commission made no finding at all. Accordingly, deletion of the exemption cannot stand.

Finally, petitioners object to one of the seven criteria that a cable operator must satisfy in order to be exempt from attribution of limited partnership. The general rule is that any partnership interest, no matter how small, leads to attribution, Attribution Order, 14 F.C.C.R. at 19039-40 p 61, but a limited partner can secure exemption if it certifies compliance with certain criteria intended to ensure that the partner "will not be materially involved in the media management and operations of the partnership." Id. The Commission interprets one of these criteria to bar exemption when a limited partner that is a vertically integrated MSO also sells programming to the partnership. See id. at 19055 p 106. This criterion applies even though the limited partner, to achieve exemption, must have certified that it does not "communicate with the licensee or general partners on matters pertaining to the day-to-day operations of its video programming business." Id. at 19040-41 p 64.

We agree with petitioners that the no-sale criterion bears no rational relation to the goal, as the Commission has drawn no connection between the sale of programming and the ability of a limited partner to control programming choices. Of course a programmer might secure contract terms giving it some control over a partnership's programming choices, but, given the independent criterion barring even communications on the video-programming business, see Attribution Order, 14 F.C.C.R. at 19040-41 p 64, exercise of that power would seem to be barred. Even if it weren't, the bargaining opportunity would depend on the desirability of the partner's programming, not on its status as a partner. The FCC does not even offer a hypothetical to the contrary.

* * *

To summarize, we reverse and remand the horizontal and vertical limits, including the refusal to exempt cable operators subject to effective competition from the vertical limits, for further proceedings. We also reverse and remand the elimination of the majority shareholder exception and the prohibition on sale of programming by an insulated limited partner. We uphold thebasic 5% attribution rule and the creation of a 33% equity-and-debt rule.

So ordered.