Realcomp II, Ltd. v. Federal Trade Commission

OPINION KAREN NELSON MOORE, Circuit Judge. Realcomp II, Ltd. (“Realcomp”) is an association of local real-estate boards and associations located in southeastern Michigan, with a membership composed of local real-estate agents and brokers. Real-comp’s primary service to its member brokers is its operation of the Realcomp Multiple Listing Service (“Realcomp MLS”), a database of property listings that can be viewed and searched by Realcomp members. Pursuant to its website policy, Realcomp prohibited information about exclusive agency and other nontraditional listings on Realcomp’s MLS from being distributed to public real-estate advertising websites through its MLS feeds. Reversing and vacating the Initial Decision of the Chief Administrative Law Judge (“ALJ”), the Federal Trade Commission (“Commission”) ruled that Real-comp violated Section 5 of the Federal Trade Commission Act (“FTC Act”), 15 U.S.C. § 45, by adopting anticompetitive policies — including the website policy— that limited the public distribution and display of limited-service property listings based on the nature of the listing contract. Realcomp petitions for review of the Commission’s opinion and order with respect to only the website policy. Under a full rule-of-reason analysis, we conclude that substantial evidence supports the Commission’s findings that: 1) Realcomp’s website policy gave rise to potential genuine adverse effects on competition due to Realcomp’s substantial market power and the website policy’s anticompetitive nature; 2) the website policy in fact caused actual anticompetitive effects; and 3) Realcomp’s proffered procompetitive justifications were insufficient to overcome a prima facie case of adverse impact. These findings establish that Realcomp’s website policy unreasonably restrained competition in the market for the provision of residential real-estate-brokerage services in southeastern Michigan and the Realcomp MLS area. Therefore, we DENY Realcomp’s petition for review. I. FACTS AND PROCEDURE A. Realcomp and the Real-Estate Market Realcomp is an association of local real-estate boards and associations located in southeastern Michigan, with a membership composed of real-estate agents and bro *820 kers. 1 Realcomp is affiliated with the National Association of Realtors (“NAR”), and its bylaws require it to abide by NAR’s rules. Realcomp’s approximately 14,000 member brokers compete with one another to provide residential real-estate-brokerage service to home buyers and sellers. Any licensed real-estate broker who is a member of a Realcomp shareholder board, including brokers who offer discount services, may become a Realcomp member. Every Realcomp member, including those who offer alternative business models, pays the same quarterly membership fees. Realcomp’s primary service to its member brokers is its operation of the Real-comp MLS, the largest MLS in Michigan. An MLS is “a database of information about properties for sale (exclusive of FSBO [For Sale By Owner] properties) that can be viewed and searched by all other local brokers who practice in the area and participate in the MLS.” Pet’r App. Vol. II at 68 (Initial Decision (“Dec.”) ¶ 14). By disseminating detailed listings, the Realcomp MLS facilitates the sharing of information among brokers representing buyers and brokers representing sellers. Real-estate listings on Realcomp’s MLS, which include property details and offers of compensation, can be viewed by Realcomp members through Realcomp’s online system, but not by the general public without the access of a member broker. The Realcomp MLS also disseminates listing information to selected public websites that can be searched by members of the public. Thus, in addition to access to and advertisement on the MLS database itself, the Realcomp MLS offers its members internet advertising on the approved websites to which Realcomp provides information. Approved websites include MoveInMichigan.com, Realcomp IDX participant websites, and Realtor.com. To disseminate listings, Realcomp provides an IDX (Internet Data Exchange) feed each day which can be loaded onto websites of member brokers and which assembles selected MLS listing data from all brokers who have requested that their listings be distributed. Through Realcomp’s IDX feed, brokers are able to display listing information from the Realcomp database on their individual websites so that consumers can search available properties on those websites. The Commission contends that technological developments like the MLS data feed are enabling consumers to self-supply certain services and are exerting competitive pressure on the traditional model for brokerage services. Under the traditional model, home sales involving the use of real-estate brokers incorporate both a listing broker, who works with home sellers, and a cooperating broker, who works with home buyers. Although representing one party in a particular transaction, brokers do not often specialize as either a cooperating or listing broker and may represent either buyers or sellers. The agreement between a listing broker and home seller, called a listing agreement, specifies the duration of the contract, the types of services to be provided by the listing broker, the compensation to be paid to the listing broker, and the offer of compensation to be paid to any cooperating broker who secures the home purchaser. A listing broker is compensated either by a flat fee paid up-front at the time of the listing agreement or by commission based on the selling price of the home, or by some combination of the two. The home seller *821 also compensates the cooperating broker, either directly or through payment to the listing broker. There are two common types of listing agreements governing the bundle of services provided by and compensation paid to residential real-estate brokers: Exclusive Right to Sell (“ERTS”) and Exclusive Agency (“EA”) agreements. Under an ERTS listing agreement, the listing broker is appointed as the seller’s exclusive agent for a specified period of time to sell the property on the owner’s stated terms, and is provided the same compensation when the property is sold even if the owner or another broker, and not the ERTS listing agent, secures the property’s sale. A cooperating broker, in contrast, typically is not paid directly by either the home seller or the home buyer, but instead is compensated indirectly by the home seller through the listing broker, who makes an offer of compensation carved from the listing broker’s own compensation to any cooperating broker who finds the buyer that ultimately purchases the home. A common ERTS compensation structure includes a 6% commission to the listing broker and an offer of compensation of 3% by the listing broker to the cooperating broker. There are also flat-fee ERTS listings that provide higher fees to the listing agent than do flat-fee EA agreements. Under an EA listing agreement, the listing broker acts as the exclusive agent of the home seller, but is paid less or no additional compensation if the property is sold without further assistance from the listing broker. Cooperating brokers are paid directly by the seller. EA contracts may offer a la carte, or unbundled, brokerage services, with a compensation structure characterized by an up-front fee to the listing broker rather than a commission, and a 3% offer of compensation from the seller directly to any cooperating broker. In a commission-based ERTS transaction, if the home is sold to an unrepresented buyer, the listing broker retains the compensation that otherwise would have been paid to the cooperating broker, and the cost to the home seller remains the same. Under an EA agreement, in contrast, if the home is sold to an unrepresented buyer, the compensation to the listing broker remains the same, and the compensation that would have been paid to the cooperating broker is retained by the home seller. ERTS agreements typically govern a traditional package of full brokerage services, while EA agreements and flat-fee ERTS agreements are conducive to providing discounted, limited brokerage services. The traditional set of services provided by a listing broker to the home seller include showing and marketing the property, presenting and evaluating offers to the seller, and negotiating counteroffers. Full-service listing brokers in Realcomp’s area typically charge commission rates around 6% and are compensated through commission-based ERTS contracts. The discount, limited-service brokerage model exemplified by EA listings offers a lower-cost alternative to the traditional full-service model. A listing broker in a limited-service listing may provide any, but not all, of the services provided under a traditional brokerage model, according to the preferences of the home seller as consumer. As described by the ALJ, unbundled brokerage services “meet a consumer demand for lower cost brokerage services where consumers are willing to carry out some of the home selling tasks themselves that otherwise would be performed by real estate professionals.” Pet’r App. Vol. II at 75 (Dec. ¶73) (internal quotation marks omitted). Home sellers may “purchase a *822 subset of the full range brokerage services (such as listing in an MLS), while self-supplying other services” such as “showing] the property, holding] open houses, negotiating] with buyers, or closing] the transaction ... without broker assistance.” Id. (Dec. ¶ 72). The expansion of the market share of limited-service brokers since 2003 has been attributed in part to the role of the internet in making it easier for brokers to market directly to home buyers and in enabling consumers to self-supply services. The development of the internet and MLS databases, the increase in the number of broker websites, and data feeds provided from the local MLS to public websites have enhanced the ability of brokers to share real-estate information and of members of the public to access it. As a result, the traditional brokerage model faces competitive pressure arising from the technological developments that enable consumers to self-supply certain services and from limited-service brokers who discount their fees in response to these developments. B. Complaint Against Realcomp On October 10, 2006, complaint counsel for the Commission issued an administrative complaint against Realcomp, alleging that Realcomp’s policies unreasonably restrained competition among brokers in the market for the provision of residential real-estate-brokerage services in southeastern Michigan or the Realcomp Service Area, and constituted unfair methods of competition in violation of Section 5 of the FTC Act. Specifically, the Commission alleged that Realcomp’s website policy and search-function policy injured consumers by explicitly limiting the publication and marketing of nontraditional listings, thereby eliminating certain forms of competition without cognizable and plausible efficiency justifications. The Commission argued that Realcomp had adopted restrictive policies in order to restrain the competition from limited-service brokers. According to complaint counsel, pursuant to Realcomp’s website policy, Real-comp prohibited information about EA listings and other nontraditional listings 2 on Realcomp’s MLS from being distributed to public real-estate advertising websites through the MLS feeds. Adopted in 2001, the website policy was first enforced in 2004 when Realcomp incorporated the requirement that members designate a listing type for all listings. The policy violated an NAR rule forbidding member MLSs from excluding EA listings from their IDX feeds. But the Realcomp board voted against adopting the NAR IDX policy and retained its data-feed exclusions. Pursuant to the search-function policy, adopted in 2003 and eliminated in 2007, EA and other nontraditional listings were excluded from the default search setting in the Realcomp MLS. As a result of the default settings on the MLS, a broker wanting to display EA listings in her search results had to select specifically to search all listings or the EA listings, or change permanently the search default by saving changes to her settings. In addition to requiring members to disclose each listing’s type, Realcomp implemented a minimum-service requirement which mandated that, in order for a listing to be labeled ERTS — and consequently included in data feeds to public websites and in the default search settings in the Real-comp MLS — brokers were required to provide full-service brokerage services in connection with the listing. The minimum- *823 service requirement was adopted in 2004 and eliminated in 2007. The Commission’s Chief ALJ held hearings over eight days in June 2007 evaluating these policies and summarized his findings in his Initial Decision, issued on December 10, 2007. The ALJ dismissed the Commission’s complaint upon finding that the Commission’s complaint counsel did not demonstrate that Realcomp’s policies unreasonably restrained or substantially reduced competition in violation of Section 5 of the FTC Act. Applying a traditional rule-of-reason analysis, the ALJ concluded that, although the website policy was likely anticompetitive in nature, and although Realcomp possessed substantial market power in the relevant markets, the Commission’s complaint counsel failed to show significant anticompetitive effects of the policies. In addition, the ALJ found that the website policy addressed a free-rider problem and a bidding-disadvantage problem whereby EA home sellers compete with Realcomp cooperating brokers for home buyers. The Commission unanimously reversed, concluding that the Realcomp policies unreasonably restrained competition, first on the grounds that the policies were inherently suspect and therefore presumptively unlawful. The Commission also concluded that the ALJ erred when it credited Real-comp’s proffered procompetitive justifications. Alternatively, under a full rule-of-reason analysis, the Commission further found that Realcomp’s substantial market power, combined with the likely anticompetitive tendencies of its policies, rendered the policies unreasonable due to their likely anticompetitive effects. The Commission also found that direct evidence established actual anticompetitive effects. The Commission then entered its cease-and-desist order, prohibiting Realcomp from “adopting or enforcing any policy, rule, practice or agreement ... that denies, restricts or interferes with the ability of Realcomp Members to enter into Exclusive Agency Listings or Other Lawful Listing agreements with the sellers of properties.” Pet’r App. Vol. I at 4 (Final Order at 4). Realcomp petitioned for review, claiming that a traditional rule-of-reason approach is required to evaluate Realcomp’s restrictions and that there was no substantial evidence that Realcomp’s website policy had anticompetitive effects. Realcomp repealed the search-function policy and the minimum-services requirement after the Commission’s complaint and does not challenge paragraph five of Part II of the order, which incorporated repeal of the search-function policy. II. ANALYSIS A. Standard of Review When we review a decision of the Federal Trade Commission, the legal issues are “for the courts to resolve, although even in considering such issues the courts are to give some deference to the Commission’s informed judgment that a particular commercial practice is to be condemned as ‘unfair.’ ” FTC v. Ind. Fed’n of Dentists, 476 U.S. 447, 454, 106 S.Ct. 2009, 90 L.Ed.2d 445 (1986). The Commission’s findings of fact are conclusive if supported by substantial evidence. Barnett Pontiac-Datsun, Inc. v. FTC (In re Detroit Auto Dealers Ass’n), 955 F.2d 457, 469 (6th Cir.1992); see 15 U.S.C. § 45(c). When we review the Commission’s findings, we may not “make [our] own appraisal of the testimony, picking and choosing for [ourselves] among uncertain and conflicting inferences.” Ind. Fed’n, 476 U.S. at 454, 106 S.Ct. 2009 (quoting FTC v. Algoma Lumber Co., 291 U.S. 67, 73, 54 S.Ct. 315, 78 L.Ed. 655 (1934)). Rather, under the substantial-evidence standard, we uphold the *824 Commission’s findings “if ... supported by ‘such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.’” Id. (quoting Universal Camera Corp. v. NLRB, 340 U.S. 474, 477, 71 S.Ct. 456, 95 L.Ed. 456 (1951)); see also In re Detroit Auto Dealers Ass’n, 955 F.2d at 469. That the Commission overruled the ALJ does not alter the standard of review. The statute governing our review, 15 U.S.C. § 45(c), provides that “[t]he findings of the Commission as to the facts, if supported by evidence, shall be conclusive.” 15 U.S.C. § 45(c) (emphasis added); see Arkansas v. Oklahoma, 503 U.S. 91, 113, 112 S.Ct. 1046, 117 L.Ed.2d 239 (1992) (“A court reviewing an agency’s adjudicative action should accept the agency’s factual findings if those findings are supported by substantial evidence on the record as a whole.”) (emphasis in original). Although “[t]he findings of the examiner” — in this case, the ALJ — “are to be considered” by a reviewing court, “[t]he ‘substantial evidence’ standard is not modified in any way when the [agency] and its examiner disagree.” Universal Camera Corp., 340 U.S. at 496, 71 S.Ct. 456. 3 Thus, because “we defer to the inferences that the [agency] derives from the evidence, not to those of the ALJ,” Varnadore v. Sec’y of Labor, 141 F.3d 625, 630 (6th Cir.1998) (internal quotation marks omitted), the relevant inquiry is whether substantial evidence supported the Commission’s conclusion that the website policy constitutes an unreasonable restraint of trade. B. Restraint of Trade Because “[t]he FTC Act’s prohibition of unfair competition and deceptive acts or practices ... overlaps the scope of § 1 of the Sherman Act ... aimed at prohibiting restraint of trade,” we rely upon Sherman Act jurisprudence in determining whether the challenged policies violated Section 5 of the FTC Act. Cal. Dental Ass’n v. FTC, 526 U.S. 756, 762 n. 3, 119 S.Ct. 1604, 143 L.Ed.2d 935 (1999); see Ind. Fed’n, 476 U.S. at 454-55, 106 S.Ct. 2009 (noting that the same analysis applies to both violations of Section 1 of the Sherman Act and Section 5 of the FTC Act). To determine whether Realcomp’s actions constitute a violation, we assess: (1) whether there was a contract, combination, or conspiracy — or, more simply, an agreement; and, if so, (2) whether the contract, combination, or conspiracy “unreasonably restrained trade in the relevant market.” See Worldwide Basketball & Sport Tours, Inc. v. NCAA, 388 F.3d 955, 959 (6th Cir.2004) (internal quotation marks omitted) (identifying elements of a violation of Section 1 of the Sherman Act); Bailey’s, Inc. v. Windsor America, Inc., 948 F.2d 1018, 1027 & n. 4 (6th Cir.1991) (same). With respect to the first element, Realcomp is a combination of its members with respect to the challenged policies: Realcomp is owned by seven associations of competing real-estate brokers, is governed by members of those associations, and claims a membership of brokers competing in the market for real-estate-brokerage services. The website policy constitutes an agree *825 ment governing the Realcomp MLS among the Realcomp members. Realcomp is, therefore, a contract, combination, or conspiracy. With respect to the second element, in evaluating whether Realcomp unreasonably restrained trade, the Supreme Court has explained that “a restraint may be adjudged unreasonable either because it fits within a class of restraints that has been held to be ‘per se ’ unreasonable, or because it violates what has come to be known as the ‘Rule of Reason.’ ” Ind. Fed’n, 476 U.S. at 457-58, 106 S.Ct. 2009. Under per se analysis, “certain agreements or practices are so ‘plainly anticompetitive,’ ... and so often ‘lack ... any redeeming virtue,’ ... that they are conclusively presumed illegal without further examination.” Broadcast Music, Inc. v. Columbia Broadcasting Sys., Inc., 441 U.S. 1, 8, 99 S.Ct. 1551, 60 L.Ed.2d 1 (1979) (internal citations omitted). “A court need not then inquire whether the restraint’s authors actually possess the power to inflict public injury ..., nor will the court accept argument that the restraint in the circumstances is justified by any procompetitive purpose or effect.” United States v. Realty Multi-List, Inc., 629 F.2d 1351, 1362 (5th Cir.1980) (internal citations omitted). When restraints are not per se unlawful, and their net impact on competition not obvious, the conventional rule-of-reason approach requires courts to engage in a thorough analysis of the relevant market and the effects of the restraint in that market. Ind. Fed’n, 476 U.S. at 461, 106 S.Ct. 2009. A full rule-of-reason inquiry “may extend to a ‘plenary market examination,’ ” Continental Airlines, Inc. v. United Airlines, Inc., 277 F.3d 499, 509 (4th Cir.2002) (quoting Cal. Dental Ass’n, 526 U.S. at 779, 119 S.Ct. 1604), which may include the analysis of “ ‘the facts peculiar to the business, the history of the restraint, and the reasons why it was imposed,’ ” id. (quoting Nat’l Soc’y of Prof'l Eng’rs v. United States, 435 U.S. 679, 692, 98 S.Ct. 1355, 55 L.Ed.2d 637 (1978)), “as well as the availability of reasonable, less restrictive alternatives,” id. If Realcomp’s challenged policies are shown to have an anticompetitive effect, or if Realcomp is shown to have market power and to have adopted policies likely to have an anticompetitive effect, then the burden shifts to Realcomp to provide procompetitive justifications for the policies. See infra Part U.C.; see also Worldwide Basketball, 388 F.3d at 959. An abbreviated or quick-look analysis, however, does not require “elaborate industry analysis,” and applies when “an observer with even a rudimentary understanding of economics could conclude that the arrangements in question would have an anticompetitive effect on customers and markets.” Cal. Dental Ass’n, 526 U.S. at 770, 119 S.Ct. 1604 (internal quotation marks omitted); see Gordon v. Lewistown Hosp., 423 F.3d 184, 209-10 (3d Cir.2005) (quick-look analysis applies when “no elaborate industry analysis is required to demonstrate the anticompetitive character” of an alleged restraint). Thus, when a restraint is not “conclusively presumed illegal,” Broadcast Music, Inc., 441 U.S. at 8, 99 S.Ct. 1551, but “the likelihood of anticompetitive effects is ... obvious,” the proponent of the restraint must provide “some competitive justification” for it, “even in the absence of a detailed market analysis” showing market power or market effects, Cal. Dental Ass’n, 526 U.S. at 769-71, 119 S.Ct. 1604 (internal quotation marks omitted). Under a quick-look analysis, once a restraint is deemed facially anticompetitive, the burden shifts to its proponent for justification on procompetitive grounds. Gordon, 423 F.3d at 210. *826 Despite these different methods, “no categorical line” separates those “restraints that give rise to an intuitively obvious inference of anticompetitive effect and those that call for more detailed treatment.” Cal. Dental Ass’n, 526 U.S. at 780-81, 119 S.Ct. 1604. Rather, the Supreme Court has emphasized that “whether the ultimate finding is the product of a presumption or actual market analysis, the essential inquiry remains the same' — • whether or not the challenged restraint enhances competition.” Id. at 779-80, 119 S.Ct. 1604 (quoting NCAA v. Bd. of Regents of Univ. of Okla., 468 U.S. 85, 104, 104 S.Ct. 2948, 82 L.Ed.2d 70 (1984)). Accordingly, the Court has moved “away from ... reliance upon fixed categories and toward a continuum,” Polygram Holding, Inc. v. FTC, 416 F.3d 29, 35 (D.C.Cir.2005), within which “the extent of the inquiry is tailored to the suspect conduct in each particular case,” id. at 34; see also 7 Phillip E. Areeda & Herbert Hovenkamp, Antitrust Law ¶ 1507 (3d ed. 2010) [hereinafter Areeda] (“[T]he quality of proof required should vary with the circumstances.”). Therefore, we must make “an enquiry meet for the case, looking to the circumstances, details, and logic of a restraint.” Cal. Dental Ass’n, 526 U.S. at 781, 119 S.Ct. 1604. In all cases, “the criterion to be used in judging the validity of a restraint on trade is its impact on competition.” NCAA, 468 U.S. at 104, 104 S.Ct. 2948. The FTC as a party and the Commission in its decision below both assert that the Realcomp website policy is “inherently suspect” and therefore subject to an abbreviated or quick-look review. 4 Because the Commission found that the website policy was inherently suspect, the Commission concluded that anticompetitive effects could be inferred without proof of Real-comp’s market power or a demonstration of actual adverse impact. But, notwithstanding its initial quick-look analysis, the Commission alternatively invalidated the challenged restraints under a more searching inquiry, which included an assessment of Realcomp’s market power and the actual, as well as likely, anticompetitive effects of its policies. Realcomp objects that the website policy should not be treated as inherently suspect, and that under any test, the website policy is not an unreasonable restraint of trade. We uphold the Commission on the basis of the more extended rule-of-reason analysis without reaching the question of whether to apply quick-look analysis. “The object” in determining whether to apply an abbreviated quick-look analysis “is to see whether the experience of the market has been so clear, or necessarily will be, that a confident conclusion about the principal tendency of a restriction will follow from a quick (or at least quicker) look, in place of a more sedulous one.” Cal. Dental Ass’n, 526 U.S. at 781, 119 S.Ct. 1604. 5 Courts have found potential *827 antitrust violations when MLS rules deny MLS membership to some brokers, see, e.g., Realty Multi-List, Inc., 629 F.2d at 1388-89, and when dealers have excluded discount brokers altogether from a venue, see, e.g., Denny’s Marina, Inc. v. Renfro Prods., Inc., 8 F.3d 1217, 1220-22 (7th Cir.1993). Here, in comparison, the challenged restraint is an internal rule within an MLS regarding its distribution of certain types of real-estate listings to the public. We need not and do not decide whether this policy is sufficiently analogous to practices already deemed by courts to be anticompetitive for it to qualify as a facially anticompetitive restraint. See Cal. Dental Ass’n, 526 U.S. at 775 n. 12, 119 S.Ct. 1604 (“[T]here must be some indication that the court making the decision has properly identified the theoretical basis for the anticompetitive effects and considered whether the effects actually are anticompetitive. Where ... the circumstances of the restriction are somewhat complex, assumption alone will not do.”). The Commission also based its holding on market analysis and evidence of anticompetitive effects, and we deny the petition for review on that basis. C. Rule of Reason Applying the rule of reason, we first look to see “whether [the] FTC has demonstrated ‘actual detrimental effects’ or ‘the potential for genuine adverse effects on competition.’ ” In re Detroit Auto Dealers Ass’n, 955 F.2d at 469 (quoting Ind. Fed’n, 476 U.S. at 460, 106 S.Ct. 2009). Market power and the anticompetitive nature of the restraint are sufficient to show the potential for anticompetitive effects under a rule-of-reason analysis, and once this showing has been made, Real-comp must offer procompetitive justifications. Realcomp contends that, to the contrary, “the requirement for proof of market power can be obviated by evidence of actual anticompetitive effects, not the other way around,” and so Realcomp urges that actual anticompetitive effects must be proven. Pet’r Br. at 25. The Supreme Court in Indiana Federation explained that “an inquiry into market power ... is ... a surrogate for detrimental effects.” Ind. Fed’n, 476 U.S. at 461, 106 S.Ct. 2009 (internal quotation marks omitted). If adverse effects are clear, inquiry into market power is unnecessary. Id. And the reverse is also true; joint ventures may be judged based on “an inquiry into market power and market structure designed to assess the combination’s actual effect.” Copperweld Corp. v. Independence Tube Corp., 467 U.S. 752, 768, 104 S.Ct. 2731, 81 L.Ed.2d 628 (1984). 6 *828 “[T]he purpose of the inquiries into market definition and market power is to determine whether an arrangement has the potential for genuine adverse effects on competition,” and this is so precisely because actual anticompetitive effects may be difficult to demonstrate. Ind. Fed’n, 476 U.S. at 460, 106 S.Ct. 2009 (emphasis added). Although the policy at issue in Indiana Federation was initially afforded abbreviated treatment, id. at 459-60, 106 S.Ct. 2009, the Supreme Court also analyzed the restriction as if it were not sufficiently “naked” automatically to necessitate “some countervailing procompetitive virtue,” id., and explained what must be shown to establish the illegality of less obviously anti-competitive restraints, id. at 460-64, 106 S.Ct. 2009. Notably, Indiana Federation did not require proof of an anticompetitive effect — specifically, in that case, that the policy at issue resulted in the provision of more costly services than consumers otherwise would have chosen — because: [a] concerted and effective effort to withhold (or make more costly) information desired by consumers for the purpose of determining whether a particular purchase is cost justified is likely enough to disrupt the proper functioning of the price-setting mechanism of the market that it may be condemned even absent proof that it resulted in ... the purchase of higher priced services]]] than would occur in its absence. Id. at 461-62, 106 S.Ct. 2009. Similarly, here, the website policy is alleged to be an “effort to withhold (or make more costly) information desired by consumers” that would result in “the purchase of higher priced [brokerage] services, than would occur in its absence.” Id. Thus, we next assess whether Realcomp’s website policy resulted in “ ‘actual detrimental effects’ or ‘the potential for genuine adverse effects on competition.’ ” In re Detroit Auto Dealers Ass’n, 955 F.2d at 469 (quoting Ind. Fed’n, 476 U.S. at 460, 106 S.Ct. 2009) (emphasis added). 7 Under either inquiry, substantial evidence supports the Commission’s findings. 1. Potential Adverse Effects a. Market Power The Commission adopted the ALJ’s findings that Realcomp possessed substantial market power in the relevant markets, and Realcomp does not dispute those findings. The ALJ defined the relevant product market as real-estate-brokerage services and found that, for most home sellers and buyers, no reasonable substitutes for such services exist because of the significant advantages of using a real-estate broker to sell a home. Because of the local nature of real-estate markets, the ALJ found that counties in southeastern Michigan define the geographic scope of competition for real-estate-brokerage services. Because of the lack of substitutes for brokerage services, the ALJ found that a broker monopolist could profitably increase commissions significantly above competitive levels. Defining the relevant input market as the supply of multiple listing services to real-estate brokers, the ALJ found that an MLS like Realcomp exhibits network effects, meaning that the value of the MLS *829 increases as the number of other users of the service increases. The value of an MLS to home sellers (or their representatives) increases with the number of home buyers (or their representatives) using the site, and, similarly, the value to home buyers increases as more home sellers list their properties on the MLS. “Brokers without full access to an MLS would ... be at a significant competitive disadvantage,” Pet’r App. Yol. II at 101 (Dec. ¶ 313), “listing services with fewer users are not economically viable substitutes,” id. (Dec. ¶ 310), and barriers to entry make it “improbable” for a rival MLS successfully to enter the market, id. at 104 (Dec. ¶ 333). Because the value of an MLS depends on the number of users, the ALJ observed that “market share is a good indicator of market power,” and found that Realcomp possessed a large market share in each relevant county. Id. at 103 (Dec. ¶ 329). In light of Realcomp MLS’s market share, network effects, and barriers to entry, the ALJ concluded that Realcomp possessed substantial market power in the relevant markets. Adopting these findings, the Commission agreed that “Realcomp possessed substantial market power in two relevant markets in Southeastern Michigan: the market for residential real estate brokerage services and the market for multiple listing services, which is a vital input into the brokerage services market.” Pet’r App. Vol. I at 42 (Comm’n Op. at 36). Given the extensive and undisputed market analysis undertaken by the ALJ and adopted by the Commission, substantial evidence supports the Commission’s findings that Realcomp possessed substantial market power. b. Anticompetitive Nature Because Realcomp possesses substantial market power, we next evaluate the anticompetitive tendencies of the Real-comp website policy. Realcomp does not regulate rates of commission, offers of compensation, or other price terms; thus, we examine the effect of Realcomp’s restrictions on consumer choice, specifically, the reduction in competitive brokerage options available to home sellers. The relevant output to be measured, therefore, is the share of non-ERTS listings in the Realcomp MLS, the exposure of these listings to consumers, and the relationship of these outcomes to the Realcomp website policy. In establishing that Realcomp’s policies “narrow consumer choice” and “hinder the competitive process,” the Commission made the following relevant findings: (1) because of its database of listings, the Realcomp MLS is the most effective tool for the sale of residential real estate in Southeastern Michigan; (2) brokers offering limited service and brokers offering traditional, full-service brokers’ services compete with one another for new listings; (3) limited service brokers’ services potentially cost less than the services of brokers offering only full-service listings (they not only unbundle the services offered but also unbundle the commission structure); (4) limited service brokers’ listings consequently exert ‘price pressure’ on full-service brokers’ listings; (5) Realcomp’s Website Policy, coupled with its Minimum Service Requirement, severely restricted consumers’ access to limited service listings because, as a result of those policies, the listings were not available on the most popular websites. Id. at 48 (Comm’n Op. at 42). These findings are supported by the evidence before the ALJ and the Commission that the website policy created barriers to the dissemination of discount listings to public websites — the entry point for many consumers in their online real-estate searches. *830 This evidence showed that EA listings can be posted on some public websites, despite Realcomp’s policies, but only by dual-listing with another MLS that does not impose similar restrictions. Moreover, such dual-listing raises costs for offering and advertising such discount services, and limited-service brokers testified that Real-comp’s policies placed them at a competitive disadvantage. In addition, two of the top four public websites used by consumers in the relevant market can be accessed only through the Realcomp MLS; therefore, due to the website policy, EA listings cannot be placed on these websites. Consumers using these websites are not informed that the websites display only ERTS listings. Realcomp’s website policy thus limited access to internet marketing and imposed additional costs on the marketing of discount listings. Furthermore, the ALJ and Commission both found that limited-service listings exert price pressure on the full-service brokerage model; and brokers testified that discounted online and limited-service models have led to customers asking agents to reduce their commissions. Evidence that the website policy limited exposure of discount listings thus reveals “a concerted refusal to deal with [EA listings] on substantially equal terms” and establishes that the website policy is likely to protect its full-service brokers from competitive pricing pressure. Northwest Wholesale Stationers, Inc. v. Pacific Stationery & Printing Co., 472 U.S. 284, 295 n. 6, 105 S.Ct. 2613, 86 L.Ed.2d 202 (1985). Combining these findings with Realcomp’s substantial market power, the Commission reasonably concluded that Realcomp’s website policy is likely to be anticompetitive. The ALJ observed that the Realcomp MLS alone reached about 80% of home buyers, and brokers could place EA listings on Realtor.com and reach approximately 90% of home buyers. The ALJ concluded that, consequently, the website policy prevented EA listings from reaching “only a relatively small additional percentage of home buyers” — the 10% who perused home listings on the inaccessible websites. Pet’r App. Yol. II at 162 (Dec. at 101). To the contrary, however, by reducing by 10% the number of home buyers that are exposed to discount listings, the website policy may very well constitute an unreasonable restraint. Restricting the online dissemination of home listings is especially pernicious because of the emerging competitive impact of the internet and of discounted brokerage services on the residential real-estate market. As the D.C. Circuit observed, “the exclusion of nascent threats is the type of conduct that is reasonably capable of contributing significantly to a defendant’s continued monopoly power.” United States v. Microsoft Corp., 253 F.3d 34, 79 (D.C.Cir.2001). The D.C. Circuit was analyzing a monopolization claim under Section 2 of the Sherman Act, rather than a horizontal restraint under Section 1. Id. at 80. Substantial evidence shows, though, that “the exclusion of nascent threats” such as discount brokerage services and consumer access to online listings “is reasonably capable of contributing significantly” to anticompetitive effects. As discussed in Part H.B., supra, courts have found potential antitrust violations when MLS rules deny MLS membership to some brokers, see, e.g., Realty Multi-List, Inc., 629 F.2d at 1388-89, and when dealers have excluded discount brokers altogether from a venue, see, e.g., Denny’s Marina, Inc., 8 F.3d at 1220-22. At issue here is a different sort of restriction than membership exclusion — internal rules within an MLS regarding its distribution of certain types of real-estate listings to the public. There is no rule of complete *831 exclusion from the MLS, and evidence shows that brokers are allowed to and do place limited-service listings on the Real-comp MLS. Nonetheless, similar to excluding discount brokers from the MLS altogether, the website policy limits exposure of discount listings. “[Listings [are] not ... distributed as widely as possible” due to the website policy, “resulting in inefficient sales prices,” which is the same kind of economic harm caused by MLS exclusions. Thompson v. Metropolitan Multi-List, Inc., 934 F.2d 1566, 1580 (11th Cir.1991); cf. Ind. Fed’n, 476 U.S. at 461-62, 106 S.Ct. 2009 (holding that proof of higher prices is not required in the context of “[a] concerted and effective effort to withhold (or make more costly) information desired by consumers for the purpose of determining whether a particular purchase is cost justified”); Polygram Holding, Inc., 416 F.3d at 37 (“[A]greements restraining autonomy in pricing and advertising impede the ‘ordinary give and take of the market place.’ ”). In particular, the website policy — like exclusions found to be anticompetitive — “reduces the competition among brokers and could result in less competition for brokerage fees.” Thompson, 934 F.2d at 1580; see Cantor v. Multiple Listing Service of Dutchess Cnty., Inc., 568 F.Supp. 424, 430 (S.D.N.Y.1983) (MLS rules “restrict[ed] [brokers’] ability to advertise their services” in order to prevent brokers from “obtaining] any competitive advantage”); see also Morales v. Trans World Airlines, Inc., 504 U.S. 374, 389, 112 S.Ct. 2031, 119 L.Ed.2d 157 (1992) (“[P]rice advertising surely relates to price.”) (internal quotation marks omitted). It is undisputed that the website policy restricted the public dissemination of Real-comp real-estate listings tending to offer consumers limited-brokerage services at reduced costs. The Commission concluded that “[the] finding of market power, coupled with [the] ... determination that the tendency of the challenged policies was to suppress competition, provide ‘indirect’ evidence that those policies have or likely will have anticompetitive effects.” Pet’r App. Vol. 1 at 41 (Comm’n Op. at 35). Evaluating the website policy, the ALJ also notably found that “the nature of the restraint is such that it is likely to be anticompetitive.” Pet’r App. Vol. II at 158 (Dec. at 97). 8 Given the significance of the Realcomp MLS to the advertising of real-estate listings, and given the role of the internet in providing consumers with the ability to self-provide certain real-estate services, substantial evidence supports the Commission’s conclusion that Realcomp’s website policy is likely to have an adverse impact on competition by restricting consumer access to discount listings. 2. Actual Adverse Effects The Commission also examined direct evidence of competitive effects and concluded that Realcomp’s policies adversely affected competition. We conclude that the Commission’s findings of actual adverse effects are supported by “relevant evidence that a reasonable mind might accept as adequate to support [its] conclusion.” In re Detroit Auto Dealers Ass’n, 955 F.2d at 469. Evidence described in Part II.C.l, supra, demonstrates that the website policy in fact limited access to internet marketing and imposed financial and administrative costs on brokers seeking to dual-list with other MLSs. Three quantitative analyses conducted by the Commission’s economic expert, Dr. Darrell *832 Williams, corroborated other record evidence showing substantial consumer harm — specifically, a reduction of the share of discount listings in the southeastern Michigan real-estate market. A time-series analysis conducted by Dr. Williams established that the share of EA listings declined by 50% after the introduction of the Realcomp restrictions. The analysis compared the share of EA listings in the Realcomp MLS before and after the challenged policies went into effect and showed that, from May 2004 to October 2006, the monthly average share of EA listings fell from about 1.5% of total listings to about 0.75%. As the Commission noted, although the study showed a reduction in only 0.75 absolute percentage points, the share of EA listings had dropped by half, revealing that “non-traditional arrangements [were] losing their toehold in the market.” Pet’r App. Vol. I at 51 (Comm’n Op. at 45). 9 While the time-series analysis documented the drop in EA listings, it could not rule out the influence of other economic factors that might have caused the decline. The conclusion that the drop was at least in part caused by Realcomp’s restrictive policies was supported by other evidence. A benchmark study conducted by Dr. Williams compared the share of EA listings in the local MLSs of Metropolitan Statistical Areas (MSAs) without restrictions (labeled Control MSAs) to that of MSAs with restrictions (labeled Restriction MSAs). 10 Control MSAs were selected based on a combination of economic and demographic factors that rendered the MSAs statistically similar to Detroit. The benchmark study revealed relatively low shares of EA listings in all the Restriction MSAs — averaging only 1.4% of MLS listings — despite differences among the MSAs with respect to other variables such as population size. Dr. Williams also found that the weighted average share of EA listings in Control MSAs was higher than in Restriction MSAs — 5.6% compared to 1.4% — and that Realcomp’s MLS had a significantly smaller share of EA listings than each of the MLSs without similar restrictions. 11 *833 For the period from 2002 to 2006, Dayton, the MSA most statistically similar to Detroit, was also the Control MSA with the lowest share of EA listings, at 1.24%. Over that same period, the share of EA listings on Realcomp’s MLS, by comparison, was 1.01%. However, even Dayton’s low share of EA listings is 22.7% greater than the share of EA listings on Realcomp. 12 These findings support Dr. Williams’s conclusion that restrictive policies like Real-comp’s Website Policy lead to a reduction in the share of EA listings. Attributing the decline in EA shares to a buyers’ market, the ALJ credited testimony that “in a declining or distressed market, where both the value of a home and the seller’s equity are declining, more home sellers would choose full service ERTS listings over EA listings because they want the professional marketing services of a full service broker.” Pet’r App. Vol. II at 164 (Dec. at 103). Indeed, the ALJ found that, between 2003 and 2005, EA listings grew from 2% to 15% of listings nationally, but between 2005 and 2006, fell from 15% to 8%. However, the ALJ also heard testimony that demand for the services of limited-service brokers increases in a softening housing market because the reduced cost appeals to home sellers without equity in their homes. Furthermore, as the Commission noted, prior to November 2006, NAR permitted members to adopt restrictive policies like Realcomp’s, suggesting that such policies could have interfered with the prevalence of EA listings in MLSs nationwide. This evidence is sufficient to provide substantial evidence for the Commission’s inferences. Dr. Williams also conducted ten statistical-regression analyses to evaluate the effects of different factors, including Realcomp’s policies, on the share of EA listings. Dr. Williams concluded that Realcomp’s policies are associated with a reduction in the share of EA listings of between 5.47 and 6.15 percentage points, leading him to predict that the percentage of EA listings in Realcomp would be higher, and the percentage of ERTS listings would be lower, in the absence of Realcomp’s policies. The ALJ found Dr. Williams’s analyses “instructive, though not conclusive,” and accepted the testimony of Realcomp’s expert, Dr. Eisenstadt, that Dr. Williams failed to include relevant variables in his regressions. Pet’r App. Vol. II at 171 (Dec. at 110). Dr. Eisenstadt testified that, when certain variables were measured at both the MSA and the local level, the effect of the Realcomp policies on the share of EA listings was not statistically different from zero. The Commission, however, concluded that MSA-level variables were properly excluded from Dr. Williams’s analysis because they were already captured by similar county and zip-code-level data. The Commission’s conclusion is strongly supported by the fact that, in his response to Dr. Eisenstadt’s report, Dr. Williams incorporated county and zip-code-level measures of the variables suggested by Dr. Eisenstadt — and still found that the restrictions were associated with a statistically significant decrease in non-ERTS contracts. Applying these variables to Dr. Eisenstadt’s sample, which excluded MLSs with restrictions other than Realcomp, Dr. Williams found that Realcomp’s restrictions are associated with a statistically significant 5.2% decrease in the share of non *834 ERTS contracts. The remaining difference was that Dr. Williams excluded MSA-level measures of Dr. Eisenstadt’s suggested variables when those variables were already measured at the county and zip-code levels. For example, Dr. Williams incorporated county-level median household income, but excluded MSA-level median household income. In other words, Dr. Williams excluded duplicative variables. In statistical terms, the MSA-level variables were excluded on the grounds that they were highly collinear with other explanatory variables and would render unreliable the results of the analysis. 13 Dr. Williams — and the Commission evaluating his work — thus explored the implications of including MSA-level data into the analysis, and concluded that the MSA-level data was already captured at the county level. 14 Dr. Williams’s time-series, benchmark, and statistical-regression analyses thus provide substantial evidence in support of the Commission’s findings of anticompetitive effects. And, even if the evidence of actual effects is inconclusive, the Commission also demonstrated the adverse potential of Realcomp’s website policy by establishing Realcomp’s market power and the anticompetitive tendencies of the website policy. 15 3. Procompetitive Justifications Realcomp might still prevail, despite evidence of actual or likely anticompetitive effects, by demonstrating “some countervailing procompetitive virtue — such as, for example, the creation of efficiencies in the operation of a market or the provision of goods and services.” Ind. Fed’n, 476 U.S. at 459, 106 S.Ct. 2009. Giving “some deference” to the Commission’s conclusion, In re Detroit Auto Dealers, 955 F.2d at 469, we conclude that the Commission properly rejected Realcomp’s proffered justifications as not “legitimate, plausible, substantial and reasonable.” Id. at 470. *835 The ALJ concluded that, without the website restrictions, home sellers with EA agreements “would free ride on the Real-comp members who invest and participate in the MLS through the payment of dues and who otherwise undertake to support the cooperative endeavor of the MLS.” Pet’r App. Vol. II at 182 (Dec. at 121). However, as the Commission found, the circumstances of this case do not establish free-riding. EA home sellers making use of the Realcomp MLS still must employ a listing broker who is a paying Realcomp member. Realcomp charges equal membership fees to all users. Therefore, Real-comp’s services to EA home sellers are compensated through payments of the EA seller to her listing broker, who in turn pays Realcomp for the benefit of participation in the MLS. Realcomp also erroneously argues that the EA home seller free-rides specifically on Realcomp cooperating agents because the EA home seller may act as her own cooperating broker. By not employing a cooperating broker, the argument goes, the EA home seller compensates Real-comp only through her payment to the listing broker, which does not cover all of Realcomp’s costs. These costs are shared by cooperating brokers who pay dues to Realcomp in order to benefit from the business generated by the MLS. Thus, Realcomp asserts, cooperating brokers “subsidize the cost that property owners would otherwise incur to procure buyers who do not use cooperating brokers.” Pet’r Br. at 52. Cooperating brokers are compensated, however, for whatever services they do provide to the EA home seller, and the EA home seller receives no free services. If an EA home seller does choose to transact with a cooperating buyer, which the EA home seller does in 80% of EA listing transactions, compensation is provided for the cooperating broker by the terms of the EA listing contract, of which the cooperating broker is aware before the sale. Moreover, the free-riding justification fails because a home seller may contract with an unrepresented buyer regardless of the type of listing contract, and under either EA or ERTS listing agreements, there is no compensation provided to cooperating brokers when the home seller chooses not to employ their services. ERTS listings include no requirement for the involvement of cooperating brokers, and, like EA listing agreements, flat-fee ERTS listings do not incorporate a cooperating broker’s commission into the listing broker’s compensation. Although there may be incentives for an EA home seller to transact with an unrepresented buyer in order to avoid additional payment to a cooperating broker, listing brokers in ERTS transactions also benefit from selecting an unrepresented buyer and retaining the commission that would otherwise be divided. Thus, Realcomp has not demonstrated a connection between the website policy and the prevention of free-riding. 16 The website policy also purportedly eliminates a “bidding disadvantage” faced by a buyer represented by a cooperating broker when bidding against an unrepresented buyer in an EA transaction. Pet’r Br. at 56. But rather than enhance competition, such a policy insulates cooperat *836 ing brokers’ commissions from competitive pricing pressure. As the Commission found, the bidding-disadvantage justification “reinforces the conclusion that [the policies] have an anti-competitive effect” by deliberately protecting established commissions and preventing the reduction in the cost of selling a home. Pet’r App. Vol. I at 39 (Comm’n Op. at 33). Even if there are financial incentives for a home seller to contract with an unrepresented buyer over a cooperating broker, Realcomp offers no meritorious procompetitive justification for protecting cooperating brokers from pressure to lower costs. 17 And, as with the free-riding justification, Realcomp fails to demonstrate how EA listings give rise to a greater “bidding disadvantage” than do ERTS listings given that a listing broker in an ERTS agreement presumably prefers to retain the cooperating broker’s commission for herself by transacting with an unrepresented buyer. III. CONCLUSION Under a full rule-of-reason analysis, we conclude that substantial evidence supports the Commission’s findings that: 1) Realcomp’s website policy gave rise to potential genuine adverse effects on competition due to Realcomp’s substantial market power and the website policy’s anticompetitive nature; 2) the website policy in fact caused actual anticompetitive effects; and 3) Realcomp’s proffered procompetitive justifications were insufficient to overcome a prima facie case of adverse impact. These findings establish that Realcomp’s website policy unreasonably restrained competition in the market for the provision of residential real-estate-brokerage services in southeastern Michigan and the Realcomp MLS area. Therefore, we DENY Realcomp’s petition for review. Footnotes [1] For simplicity, we refer to the agents and brokers who make up Realcomp’s membership as "brokers.” [2] The rules prohibit distribution to real-estate internet advertising sites of “Exclusive Agency, Limited Service and MLS Entry Only Listings.” Pet’r App. Vol. II at 107 (Dec. ¶ 358). [3] We have stated that "[w]hen the Commission overrules the ALJ and substitutes its own findings, we should carefully scrutinize the Commission's determinations of fact, and therefore its conclusions based upon those facts.” In re Detroit Auto Dealers Ass'n, 955 F.2d at 469. The Supreme Court has explained, however, that consideration of the examiner’s findings is only one of many "other factors which in sum determine whether evidence is 'substantial,' " Universal Camera, 340 U.S. at 497, 71 S.Ct. 456; thus, a court "need not limit its reexamination of the case to the effect” of those findings beyond "the relevance that they reasonably command,” id. [4] The Commission applied the "inherently suspect” or "quick-look” analytical framework upheld by the D.C. Circuit in Polygram Holding, Inc. v. FTC, 416 F.3d 29 (D.C.Cir.2005). In Polygram, the D.C. Circuit held that the challenged practices bore a "close family resemblance” to other practices "that already stand[ ] convicted in the court of consumer welfare.” Id. at 37. The court found unlawful “[a]n agreement between joint venturers to restrain price cutting and advertising with respect to products not part of the joint venture [which] looks suspiciously like a naked price fixing agreement between competitors.” Id. [5] In Detroit Auto Dealers, we expressed reservations about applying the "inherently suspect” framework to an agreement among car dealers to limit showroom hours because the impact on competition was not "immediately apparent,” and we examined actual and potential adverse effects on competition. In re Detroit Auto Dealers Ass’n, 955 F.2d at 469- *827 70. However, "in deference to the Commission, we [did] not ... hold the analogy [to inherently suspect practices] erroneous.” Id. at 471-72; see also Worldwide Basketball, 388 F.3d at 961 (declining to apply quick-look analysis "where the contours of the market ... are [not] sufficiently well-known or defined to permit the court to ascertain without the aid of extensive market analysis whether the challenged practice impairs competition”). We note that Detroit Auto Dealers preceded in time the Supreme Court’s decision in California Dental and the D.C. Circuit’s Polygram decision endorsing the quick-look analysis. [6] Other circuits have permitted an inference of adverse effects based on a showing of market power and anticompetitive tendencies. See, e.g., Craftsmen Limousine, Inc. v. Ford Motor Co., 491 F.3d 380, 388 (8th Cir.2007) (holding that plaintiff may prove "detrimental effects on competition by making an inquiry into market power and market structure designed to assess the [restrainfi’s actual effect.” (internal quotation marks omitted)); Tops Markets, Inc. v. Quality Markets, Inc., 142 F.3d 90, 97 (2d Cir.1998) ("A plaintiff seeking to use market power as a proxy for adverse effect must show market power, plus some other ground for believing that the challenged behavior could harm competition in the mar *828 ket, such as the inherent anticompetitive nature of the defendant's behavior or the structure of the interbrand market.”). [7] See 7 Areeda ¶ 1511 (Supp. 2010) (“The proof of actual market effects that the ALJ insisted on would certainly be appropriate in a private Sherman Act proceeding under which proof of private harm were required. ... But in the presence of market power and the absence of a convincing justification the FTC's treatment of highly suspicious restraints is warranted.”). [8] Despite this initial finding of anticompetitive tendencies, the AU ultimately credited Realcomp's procompetitive justifications and found insufficient the Commission’s direct econometric evidence to establish an unreasonable restraint of trade. [9] A decline in the share of EA listings would have real implications for consumers who would otherwise purchase those listings. Assuming an average home sale price of $200,000, and assuming that EA listings save home sellers half of the typical 6% commission, an EA listing would save an individual home seller, on average, $6,000. [10] The Control MSAs were Charlotte, North Carolina; Dayton, Ohio; Denver, Colorado; Memphis, Tennessee; Toledo, Ohio; and Wichita, Kansas. The Restriction MSAs were Green Bay/Appleton, Wisconsin; Boulder, Colorado; and Williamsburg, Virginia. [11] The ALJ faulted Dr. Williams for his selection of Control MSAs, noting that Midwestern Rust Belt cities omitted from the study were intuitively more similar to Detroit than the chosen MSAs, and therefore were instructive points of comparison. Increasing the sample size would have strengthened Dr. Williams’s findings. However, the ALJ pointed to no evidence that comparisons to Rust Belt cities would have yielded different results. Furthermore, the Control MSAs were chosen based on a ranking of variables plausibly relevant to choice of home listing contracts (namely, home price change over 1 and 5 years, median single-family home price, education levels as measured by percentage of high school and bachelor degrees, median income, population size, and population density). The ALJ also theorized that if Dr. Williams had correctly identified factors that determine the share of EA contracts and then had chosen MSA's similar to one another, the EA shares of the Control MSAs would be very similar, but instead varied between 1.24% and 14%. But, as the Commission noted, the values of the economic and demographic variables used as selection criteria could vary across MSAs in the control sample, even if related to the use of EA listings, and even if they rendered the selected MSAs more similar *833 to Detroit than other cities. That the shares of EA listings in the Control MSAs were different from one another does not mean that the MSAs were inappropriate for comparison. [12] Given that the difference in absolute terms between these shares is less than 1%, the ALJ found the difference insignificant. [13] In statistical terms, the MSA-level variables were omitted to avoid the problem of multicollinearity. Regression analysis measures the effects of different factors, called independent or explanatory variables, on a particular outcome, called the dependent variable. The more that independent variables are correlated with one another — such as when two measures capture the same information — the less unique information they will each contribute to the prediction of the dependent variable. When predictors become more correlated, the estimate of individual regression coefficients becomes more unreliable and, thus, the effects of the individual independent variables become difficult to measure accurately. See Jacob Cohen et al„ Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences 419-420 (3d ed. 2003). Dr. Eisenstadt himself admitted that, while not completely duplicative, the MSA-level and county-level data were highly correlated. [14] Realcomp may be correct that the measure of a variable like median household income is to some extent relevant at both the local and MSA-wide level, and that these measures convey different information relevant to decisions about how a home seller will approach the sale. If important independent variables are omitted from the analysis, their effects on the dependent variable may be attributed to independent variables that are included, and as a result, the causal relationship between included independent variables and the dependent variable may be biased. See id. at 143, 426. But, as discussed supra, including these variables introduced significant problems of multicollinearity, and substantial evidence supports the Commission’s interpretation of Dr. Williams’s statistical analysis. [15] In this appeal, Realcomp is specifically challenging the Commission's order on the basis of its conclusions with respect to the website policy. As the ALJ noted, Dr. Williams did not disentangle the effects of the search-function policy, website policy, and minimum-services requirement, all of which were at issue when complaint counsel filed the complaint. [16] The Commission's analysis is persuasive on this point, noting that EA listings disadvantage not the cooperating brokers, but rather "the listing broker who signs an EA contract for less compensation than an ERTS contract would have provided, and the listing broker who insists upon an ERTS contract and loses a listing as a result.... In other words, these two categories of listing brokers are not losing money through free-riding; they are losing money through competition.” Pet'r App. Vol. I at 37 (Comm’n Op. at 31). [17] The ALJ also noted that the website policy reflects the greater value of ERTS over EA contracts to the MLS because cooperating brokers often must deal directly with EA home sellers rather than listing brokers when engaging in EA contracts, and, as a result, may be required to provide transactional services that would otherwise be performed by full-service listing brokers. We agree with the Commission, however, that Realcomp has failed to carry its burden of demonstrating how meeting the preferences of cooperating brokers ultimately benefits consumers, why the price for EA listings does not incorporate these costs, or how this policy in particular offers efficiency benefits. It is also worth noting that, despite Realcomp's assertions of the website policy’s necessity, the NAR has ruled that MLSs may no longer exclude EA listings from their IDX feeds.