INTRODUCTION

The United States has asymmetric regulation of the provision of broadband Internet access service. A cable television system operator is not regulated in its sale of cable modem service. In contrast, an incumbent local exchange carrier (ILEC) that offers digital subscriber line (DSL) service faces price regulation as well as the obligation to offer competitors the use of its broadband network on a wholesale (or, unbundled) basis so that they may offer, in the retail market, DSL services that compete with the ILECs own retail offering to consumers. The social costs of asymmetric regulation are by now familiar. Such regulation leads not to deregulation, but to an enduring managed competition far more complex to administer than traditional regulation of a monopoly service provider ever was.1 The alternative to asymmetric regulation is either symmetric regulation or symmetric freedom from regulation.
1. See, e.g., J. GREGORY SIDAK & DANIEL F. SPULBER, DEREGULATORY TAKINGS AND THE REGULATORY CONTRACT: THE COMPETITIVE TRANSFORMATION OF NETWORK INDUSTRIES IN THE UNITED STATES 307-92, 403-26 (Cambridge Univ. Press 1997); DALE E. LEHMAN & DENNIS WEISMAN, THE TELECOMMUNICATIONS ACT OF 1996: THE COSTS OF MANAGED COMPETITION (Kluwer Academic Press 2000); ALFRED E. KAHN, WHOM THE GODS WOULD DESTROY, OR HOW NOT TO DEREGULATE (AEI-Brookings Joint Center on Regulatory Reform 2001).
(03,5,&$/ &$6( $*$,167 $6<00(75,& 5(*8/$7,21
Assuming that the latter alternative is preferred, what actual steps would be taken to abolish asymmetric regulation of ILEC provision of broadband Internet access? Congress could again amend the Communications Act of 1934.2 That step, however, would be a daunting one for all the reasons that it took many years for Congress to pass the Telecommunications Act of 1996.3 Alternatively, the Federal Communications Commission (FCC) could remove asymmetric regulation that the agency itself previously imposed. The FCC could declare that broadband Internet access service is not a telecommunications service, subject to numerous regulations applicable to ILECs, but rather an information service, which is free of such regulations. Amid considerable controversy, the FCC invited public comment in February 2002 on such a proposed reclassification.4 Or the FCC could use its power under section 10 of the Communications Act to forbear from regulating ILEC provision of broadband Internet access.5 A third, and more incremental, approach would be for the FCC to declare ILECs nondominant in the provision of advanced services, such as broadband Internet access. Nondominant global carriers are exempt from price-cap or rate-of-return regulation, as well as the obligation to file tariffs and to establish the reasonableness of those tariffs through the submission of cost data.6 Understandably, a dominant carrier will seek to be reclassified as nondominant as soon as market conditions will support such a conclusion by regulators. Much, if not all, of the economic analysis required to determine whether a carrier is nondominant also would be relevant to the FCCs decision whether to forbear from regulating a particular service or whether to reclassify the service in question as unregulated. Although the FCC did not receive its authority under section 10 to forbear from regulation until

own-price elasticity of demand for DSL service and the cross-price elasticity of demand for cable modem service with respect to DSL service. We find that the own-price elasticity of DSL service is high, which suggests that demand for DSL service is price-elastic. Moreover, the cross-price elasticity of demand for cable modem service with respect to a change in the price of DSL service is high as well, which is evidence that DSL and cable modems are in the same product market, and that DSL providers lack market power. Having concluded on the basis of both empirical and qualitative evidence that an ILEC does not now have market power in the provision of DSL service, we explain why an ILEC could not leverage its purported market power in the local exchange market into the massmarket broadband services market. We conclude that the FCC would advance the public interest by ruling that the ILECs are nondominant in the mass-market broadband services market. Doing so would be an important first step toward the elimination of asymmetric regulation of broadband Internet access service. II. THE FCCS FRAMEWORK FOR ANALYZING NONDOMINANCE
The FCC defines a dominant carrier as a carrier that possesses market power and a nondominant carrier as a carrier not found to be dominant (that is, one that does not possess market power).10 In the Fourth Report in the Competitive Carrier proceeding, the FCC defined market power as the ability to raise and maintain price above the competitive level without driving away so many customers as to make the increase unprofitable.11 The Commission affirmed this definition in the Bell Operating Company (BOC) Classification Order, concluding that the BOC interLATA affiliates should be classified as dominant carriers in the provision of inregion, interstate, domestic interLATA services only if the affiliates have the ability to raise prices of those services by restricting their own output.12 Notably, in affirming this definition, the FCC rejected arguments
10. 47 C.F.R. 61.3(o), 61.3(t). 11. Policy and Rules Concerning Rates for Competitive Common Carrier Services and Facilities Authorizations Therefore, Fourth Report and Order, 95 F.C.C.2d 554, (1983) (citing William M. Landes & Richard A. Posner, Market Power in Antitrust Cases, 94 HARV. L. REV. 937, 945-52 (1981); Valley Liquors, Inc. v. Renfield Importers, Ltd., 678 F.2d 742, 745 (7th Cir. 1982); H&B Equip. Co., Inc. v. Intl Harvester Co., 577 F.2d 239, 243 (5th Cir. 1978); ALFRED E. KAHN, THE ECONOMICS OF REGULATION: PRINCIPLES AND INSTITUTIONS 65-66 (MIT Press 1988)). 12. Regulatory Treatment of LEC Provision of Interexchange Services Originating in the LECs Local Exchange Area and Policy and Rules Concerning the Interstate, Inter-

same four-part nondominance methodology in both the AT&T International Nondominance Order19 and the COMSAT Nondominance Order.20 In the BOC Classification Order,21 the FCC revised its approach to defining product and geographic markets and amplified its nondominance analysis to account for firms that could be dominant in the provision of some, but not all, services. With respect to market definition, the Commission adopted the approach taken in the 1992 Merger Guidelines.22 Under that approach, a relevant market encompasses any service or group of services for which there are no close demand substitutes.23 The Commission noted that, under this approach, there could be multiple product markets for domestic, interstate, interexchange services, and that each route that allows a connection from one location to another would be a separate geographic market.24 Nevertheless, noting that it would be impracticable and inefficient to recognize and analyze every such market, the FCC concluded that it need only do so if there was credible evidence suggesting that there is or could be a lack of competitive performance with respect to a particular service or group of services or in a particular area.25 Using this approach, the FCC concluded that all domestic, interstate, interexchange services could continue to be treated as a single product market and that a
19. Motion of AT&T Corp. to be Declared Nondominant for International Service, 11 F.C.C. Rcd. 17,963, 17,(1996) [hereinafter AT&T International Nondominance Order]. 20. COMSAT Corp., Petition Pursuant to Section 10(c) of the Communications Act of 1934, as amended, for Forbearance from Dominant Carrier Regulation and for Reclassification as a Nondominant Carrier, 13 F.C.C. Rcd. 14,083, 14,118-(1998) [hereinafter COMSAT Nondominance Order]. 21. BOC Classification Order, supra note 12, at 18,042 28. 22. Although under the Merger Guidelines, markets are defined primarily based on demand elasticity, the F.C.C. will assume that two services are in the same product market if production substitution among these services is nearly universalthat is, if the suppliers of one service also supply the other service. See Application of WorldCom, Inc. and MCI Communications Corporation for Transfer of Control of MCI Communication Corporation to WorldCom, Inc., 13 F.C.C. Rcd. 18,025, 18,& n.66 (1998) (citing 1992 Horizontal Merger Guidelines, 57 Fed. Reg. at 41,557, 1.32 n.14) [hereinafter WorldCom/MCI Order]. 23. BOC Classification Order, supra note 12, at 18,042 29. 24. Id. at 18,042 30. 25. SBC/Ameritech Merger Order, supra note 7, at 14,(As we explained in the WorldCom/MCI Order, to define relevant product markets we can identify and aggregate consumers with similar demand patterns). See also id. at 14,747 n.147 (We can consider, as a whole, groups of point-to-point markets where customers face the same competitive conditions. We therefore treat as a geographic market an area in which all customers in that areas will like face the same competitive alternatives for a product).

BOCs in-region and out-of-region territory be treated as separate geographic markets. In 1998, the Commission amplified its nondominance analysis in the BOC Classification Order. The Commission began its analysis by applying the four-part test first articulated in the AT&T Streamlining Order and then applied in subsequent nondominance orders. Because the BOCs were new entrants in the long-distance market, the Commission readily concluded that the BOCs would not have market power in the provision of domestic in-region, interLATA services immediately upon entry into that market. Then, theorizing that this initial lack of market power could simply reflect the BOCs prior exclusion from the market, the Commission went on to address whether a BOC could leverage market power in local exchange and exchange access services to confer market power on its long-distance affiliate upon entry by the affiliate into the market, or shortly thereafter. As noted, though, the Commission specifically rejected claims that a BOC long-distance affiliate should be deemed dominant solely on the basis of a finding that it could derive advantages in the market through discrimination, cross-subsidization, or other anticompetitive actions by the ILEC.26 While acknowledging that such actions could distort the market and result in a misallocation of societal resources, the Commission found that dominant carrier regulation of the affiliate was a poor tool for dealing with those risks.27 Rather, the Commission found, dominant carrier status would only be appropriate if the advantages conferred by leveraging were so great that, upon entry or soon thereafter, the ILECs affiliate would be able to raise prices by restricting its own output.28 We now apply the FCCs four-part test to mass-market broadband services, which include residential customers and small businesses. III. NONDOMINANCE IN THE MASS-MARKET BROADBAND SERVICES MARKET
Mass-market broadband services are provided by cable providers, telephone companies, direct broadcast satellite (DBS), and other firms that provide mass-market consumers packet switched transport at speeds

Although some have argued that broadband services do not comprise a complete product market,52 their position has been that the market is broader in scope, not that individual broadband Internet services comprise discrete product markets. More specifically, the only debate of which we are aware regarding product market definitions for mass-market broadband services has centered around the question of whether such services are part of a larger market that also includes narrowband Internet access services. Whether one should include narrowband Internet services in the market is largely academic for present purposes. Even assuming, as some have claimed, that narrowband and broadband Internet access services are part of the same product market, consumers would necessarily have a regulated substitute service (narrowband service) available to them even after the detariffing of an ILECs broadband service. That regulated alternative would ensure that a substitute service is available at just and reasonable terms, and it would further constrain an ILECs ability profitably to raise DSL prices to supercompetitive levels. Any such increase would precipitate a migration of customers to alternative broadband platforms, and to narrowband access. 2. The Geographic Market Like long-distance voice traffic, a broadband connection to the Internet at its most fundamental level, involves a customer making a connection from one specific location to another specific location.53 As with a longdistance voice call, customers do not view broadband connections originating in different locations to be close substitutes for each other.54 In this
Gregory Sidak, A Consumer-Welfare Approach to the Mandatory Unbundling of Telecommunications Networks, 109 YALE L. J. 417, 475-77 (1999). For an elaborate discussion of Marshalls rules, see P.R.G. LAYARD & A.A. WALTERS, MICROECONOMIC THEORY 259 (McGraw-Hill 1978). 52. See, e.g., Applications for Consent to Transfer of Licenses and Section 214 Authorizations from MediaOne Group, Inc., Transferor, to AT&T Corp., Transferee, CC Docket No. 99-251, AT&T Reply Comments at 71-79 (arguing that high-speed and narrowband Internet access services constitute part of the same market); AOL/Time Warner Order, supra note 34, 68-74 (noting that MediaOne, AT&T, Time Warner, and AOL have argued, in various proceedings, that narrowband and broadband Internet access services constitute a single product market). Ironically, in the AT&T-TCI merger proceeding, AOL argued that broadband access services constituted a separate product market. See Applications for Consent to the Transfer of Control of Licenses and Section 214 Authorizations from Tele-Communications, Inc., Transferor, to AT&T Corp., Transferee, CS Dkt. No. 98-178, Comments of America Online at 16. 53. See BOC Classification Order, supra note 12, at 15,792-93 64. 54. Id.

Thus, it is likely that a very large percentage of households in an ILECs territory have access to cable modem service. That conclusion is supported by two recent analyst reports. One, issued by the Yankee Group, found that, as of year-end 2001, two thirds of all U.S. households will have access to cable modem service and that, by year-end 2002, 77 percent of U.S. households would have access to cable modem service.60 Another report, issued jointly by JP Morgan and McKinsey & Co., found even higher addressability: 74 percent of U.S. households at the end of 2000, and an estimated 82 percent at the end of 2001.61 Finally, the National Cable and Television Association reported in September 2001 that 83 percent of all U.S. households would be upgraded for cable modem service by the end of 2001.62 Nationally, fewer than half of all U.S. households have access to DSL service.63 For example, SBC is currently able to offer DSL service to only 50 percent of its customers.64 Given that cable penetration in an ILECs territory is likely to be at least as high as the national average, there are likely to be very few areas in which that ILEC offers DSL service but no cable provider offers cable modem service.65 Even if there are a few such nonoverlapping areas, there are alternative broadband platforms in those (and other) areas that compete with an ILECs DSL service. For example, despite the recent difficulties encountered by many CLECs, they collectively accounted for about 16 percent of the DSL market in December 2000.66 They can lease unbundled loops or the high-frequency portion of a loop anywhere in an ILECs territory to provide DSL service. Satellite and fixed wireless broadband services also provide consumers with a broadband alternative. Although wireless technologies still account for a relatively small share of the mass-market broadband services market, they are ubiquitously available and growing rapidly. For example, the Strategis Group predicts that the number of U.S.
60. YANKEE GROUP BROADBAND, supra note 40, at 4. 61. JP MORGAN BROADBAND, supra note 40, at 39, tbl.6. 62. Downloaded from National Cable Television Association web site on Sept. 25, 2001 at http://www.ncta.com/industry_overview/indStat.cfm?ind-OverviewID=2. 63. JP MORGAN BROADBAND, supra note 40, at 43 chart 25. 64. See SBC Communications, Inc., Strong Growth in Data, Wireless and Long Distance Highlights SBCs First-Quarter Results, Press Release, Apr. 23, 2001, available at http://www.sbc.com/Investor/Financial/Earning_Info/docs/-1Q_IB_FINAL.pdf. 65. JP MORGAN BROADBAND, supra note 40, at 43 chart 25 (as of first quarter of 2000, only 10 percent of residential households were addressable by DSL but not cable modem service). 66. ILECs account for about 84 percent of DSL nationwide. See COMM. DAILY, Aug. 14, 2001 at 6 (discussing Telechoice study).

satellite subscribers will grow to more than four million by 2005.67 It is reasonable to believe that satellite services could reach those (rare) areas that are served by DSL but not served by cable modems. To be sure, upload speeds for satellite broadband service are slow. But as Professors Janusz Ordover and Robert Willig have testified on behalf of AT&T, such concerns are irrelevant to the vast majority of users who, if they worry about speed at all, are primarily interested in fast download times and do not send significant amounts of information.68 Irrespective of whether satellite services experience the explosive growth that some predict, it is reasonable to conclude that providers of those services will fill a market niche, focusing their competitive efforts in any areas in which cable modem and/or DSL services are not available. Indeed, satellite broadband providers are likely to be most successful in areas in which cable service is not available, because it is in those areas that consumers are likely to use DBS service for video, and consumers who already use satellite service are likely to be the most receptive to satellite-based Internet access. For customers who cannot obtain cable modem access, fixed-wireless service is another option. Frost & Sullivan project that the number of fixed-wireless broadband subscribers will grow from 79,000 at the end of 2000 to over 400,000 at the end of 2001 and to almost one million at the end of 2002.69 Even if these predictions prove wrong, fixed-wireless services, at a minimum, can be expected to fill any niche in which competition between DSL and cable modem services is less vigorous. B. Application of the FCCs Framework to the Mass-Market Broadband Services Market
With respect to each of the Commissions criteria, we conclude that ILECs are nondominant in the mass-market broadband services. 1. Market Share When the FCC declared AT&T to be nondominant in interstate, domestic, interexchange telecommunications services in late 1995, AT&Ts market share was estimated to be 60 percent.70 Likewise, AT&Ts overall
67. Dan Miller, Who You Going to Call for Broadband?, INDUS. STANDARD, Apr. 30, 2001. 68. Decl. of Janusz A. Ordover and Robert Willig, Applications for Consent to the Transfer of Control of Licenses from MediaOne Group, to AT&T Corp., CC Dkt. No. 99251 (Sept. 17, 1999). 69. FROST & SULLIVAN, NORTH AMERICA BROADBAND WIRELESS ACCESS SERVICES MARKET 2001, at I-2. 70. AT&T Reclassification Order, supra note 17, at & n.173.

previous nondominance orders, the Commission relied solely on indirect evidence of demand elasticity. For example, in the AT&T Reclassification Order, the Commission relied heavily on high churn rates in concluding that long-distance customers had highly price-elastic demand. It also cited its finding in the AT&T Streamlining Order that business customers had highly price-elastic demand. This finding was based on evidence that business customers tend to be more sophisticated and knowledgeable purchasers of telecommunications services, are aware of the choices available to them, and have no strong bias towards AT&T versus other interexchange carriers. Here we present not only the type of indirect evidence of high demand elasticities upon which the Commission has relied in the past, but also direct quantitative evidence of the own-price elasticity of demand for massmarket broadband services. That quantitative evidence is derived from a study conducted in 2000 by Professors Paul Rappoport, Don Kridel, Lester Taylor, and Kevin Duffy-Demo. Using Marketing Science survey data for Internet use and TNS Telecoms survey data for broadband access availability and prices, Professors Rappoport, Kridel, Taylor, and Duffy-Demo calculated that the own-price elasticity of demand (the percentage change in demand for every one-percent increase in price) for DSL services is negative 1.462, which implies that for every one-percent increase in the price of DSL service, demand decreases by 1.462 percent.81 That result suggests that demand for DSL service is, by definition, price-elastic.82 Because the study by Professors Rappoport, et al., was based on data from the first quarter of 2000, we have updated it, using nearly the identical econometric model and data from the fourth quarter of 2000 and the first quarter of 2001. In its quarterly survey, TNS Telecoms obtains detailed Internet usage data from approximately 3,500 respondents. Each respondent is asked, among other things, (1) whether DSL and/or cable modem access is available in his or her neighborhood, and (2) whether he or she subscribes to dial-up access or broadband access, and, if so, at what price. Respondents also supply socio-economic information concerning
81. Paul Rappoport, et al., Residential Demand for Access to the Internet, University of Arizona Working Paper, Spring 2001, at Table 10; see also Paul Rappoport, et al., An Econometric Study of the Demand for Access to the Internet, in THE FUTURE OF THE TELECOMMUNICATIONS INDUSTRY: FORECASTING AND DEMAND ANALYSIS (D.G. Loomis & L. D. Taylor eds., Kluwer Academic Publishers 1999). 82. See, e.g., WILLIAM J. BAUMOL & ALAN S. BLINDER, MICROECONOMICS: PRINCIPLES AND POLICY 133 (Dryden Press 7th ed. 1994) (demand is elastic whenever a rise in the price will decrease total revenue).
their income, race, occupation, and other characteristics that might influence the decision to purchase Internet services. In the present application, standard regression analysis is not appropriate to estimate the consumer-choice model because the decision to purchase a broadband access technology is a binary (as opposed to continuous) variable. Hence, like Rappoport, et al., we estimated a nested logit model, which allows us to examine the discrete choice across all massmarket broadband alternatives, and the discrete choices within broadbandaccess alternatives. Details of the model are described in the Appendix. A brief description is provided here. The model involves two stages in the estimation procedure. In the first stage, we assume that the customer chooses between no Internet access, narrowband Internet access, and broadband Internet access based on the following variables: (1) the customers income, (2) the customers gender, (3) the customers age, and (4) the customers education. In the second stage, conditional on choosing broadband Internet access, we assume that the customer chooses between cable modem and DSL services based on the price of each service. Before estimating the model, we removed from the dataset any customer who did not have access to both cable modem and DSL services. The output from the regression model appears in the Appendix. Table 1 presents the updated elasticity estimates.

Table 1: Own-Price and Cross-Price Elasticities for DSL Service Service Price of DSL Price of Cable Modem Choice: DSL -1.184 0.415 Choice: Cable Modem 0.591 -1.220
As Table 1 shows, the own-price elasticity of DSL is still high (the comparable estimate from Rappoport et al. is -1.462), which suggests that demand for DSL service is price-elastic. Moreover, the cross-price elasticity of demand for cable modem with respect to a change in the price of DSL is high as well: for every 1 percent increase in the price of DSL service, the demand for cable modems rises by 0.591 percent. The comparable elasticity from Rappoport et al. was 0.766 percent. The high crossprice elasticity is further evidence that DSL and cable modems are in the same product market, and that DSL providers do not have market power. Other evidence underscores that DSL subscribers have highly priceelastic demand. The features that consumers most value from broadband services offerings are not unique to DSL service. As noted above, con-
sumer surveys show that consumers who choose a broadband service do so primarily (1) to increase the speed of ordinary web surfing, (2) to take advantage of applications that require greater bandwidth, (3) to obtain an always-on connection, and (4) to free up their telephone line when they are using the Internet. All of these benefits are available not only from DSL, but also from cable modem and other broadband options. Finally, surveys indicate that many consumers who are interested in broadband service are not predisposed towards cable or DSL service. They do not care about what platform they use to obtain broadband serviceonly that they obtain the features of a broadband connection.83 In short, econometric and qualitative evidence indicates that there is price-elasticity of demand for DSL service is high, supporting the conclusion that the ILECs are non-dominant in their provision of DSL services. 3. Supply Elasticities A third consideration in determining an ILECs nondominant status is supply elasticity. In evaluating supply elasticity in its AT&T Reclassification Order, the FCC focused on two factors: (1) the capacity of existing competitors to expand supply and (2) low entry barriers for new suppliers.84 It concluded that AT&Ts competitors can add significant numbers of new customers with their existing capacity and add incrementally to this capacity as new customers. are added to their networks.85 The same is clearly true of the ILECs competitors that provide massmarket broadband service. Those competitors could absorb immediately, and without additional investment, a significant number of an ILECs DSL subscribers and could eventually absorb an ILECs entire customer base with little or no additional investment. First, as discussed above, there are numerous providers of competitive mass-market broadband services throughout each ILECs region. For example, two-thirds or more of the households in the United States are served by cable facilities that are capable of providing mass-market broadband services. Yet the vast majority of this plant is not being used.

[I]mproper allocation of costs by a BOC is of concern because such action may allow a BOC to recover costs from subscribers to its regulated services that were incurred by its interLATA affiliate in providing competitive interLATA services. In addition to the direct harm to regulated ratepayers, this practice can distort price signals in those markets and may, under certain circumstances, give the affiliate an unfair advantage over its competitors. For purposes of determining whether the BOC interLATA affiliates should be classified as dominant, however, we must consider only whether the BOCs could improperly allocate costs to such an extent that it would give the BOC interLATA affiliates, upon entry or soon thereafter, the ability to raise prices by restricting their own output.105
The Commissions conclusion that dominant carrier regulation of a service is appropriate only if the BOC could quickly acquire market power in that service is sound. As the Commission noted, our dominant carrier regulations are generally designed to prevent a carrier from raising prices by restricting its output. We agree with the DOJ that applying dominant carrier regulation to an affiliate in a downstream market would be at best a clumsy tool for controlling vertical leveraging of market power by the
105. BOC Classification Order, supra note 12, at 15,815 103.
parent, if the parent can be directly regulated instead.106 Moreover, as the Commission noted, regulations associated with dominant carrier classification can. have undesirable effects on competition.107 Thus the Commission does not impose dominant carrier status on an entity or service unless the firm at issue can control price in the market by restricting its output of that service. The FCC does not impose dominant carrier status simply to ensure what some call a level playing field. It is inconceivable that any showing of leveraging could be made. As noted above, cable operators enjoy significant advantages in the massmarket broadband services market. To quickly acquire market power, an ILEC would not only have to overcome these advantages, but also would have to establish its own overwhelming advantages. Considering that the services in which an ILEC is ostensibly dominantlocal exchange and exchange access servicesare highly regulated, that outcome is most unlikely. Of course, the FCC need not speculate on this point. If, soon after entering the mass-market broadband services market, an ILEC could have acquired monopoly power in that market, it presumably would have done so. Yet the ILECs collective market share continues to be dwarfed by their cable competitors share. Those facts show that an ILEC cannot use its position in the local exchange market to obtain dominance in the massmarket broadband services market. In any event, an ILEC could not, even as a theoretical matter, quickly acquire market power in the mass-market broadband services market by leveraging any market power that it might retain in the local exchange market. In its past orders, the Commission has recognized three ways in which such leveraging could occur: cross-subsidization, discrimination, and the effectuation of a price squeeze. We address each below in the context of the relevant product market. In the BOC Classification Order, the Commission held that BOC longdistance affiliates could obtain the ability, through cross-subsidization, to raise prices by restricting their own output only if a BOCs improper allocation enabled a BOC interLATA affiliate to set retail interLATA prices at predatory levels (i.e., below the costs incurred to provide those services), drive out its interLATA competitors, and then raise and sustain retail interLATA prices significantly above competitive levels.108 Thus the issue here is whether, through cross-subsidization, an ILEC could set DSL

106. Id. at 15,804 85, 15,808 85, 91 (quoting DOJ Reply, Aug. 30, 1996, at 27). 107. Id. at 15,808 90. 108. Id. at 15,815 103.
prices at predatory levels, drive its broadband competitors out of the market, and then raise and sustain its prices significantly above competitive levels. Even in the unlikely event that an ILEC could drive a cable operator into bankruptcy, the bandwidth capacity of that carrier would remain intact, ready for another firm to use (after a liquidation sale) and immediately undercut an ILECs noncompetitive prices. If an ILEC were to attempt predatory pricing in the broadband market, it could not expect to recoup its investment in sales made below incremental cost. The FCC has expressly embraced this economic reasoning when it concluded that predation is implausible with respect to either long-distance fiber-optic networks109 or spectrum.110 The argument applies with equal force to the fixed broadband infrastructure of a cable operator. Clearly such cross-subsidization is not possible. The ILECs have, until quite recently, been treated as nondominant providers of DSL services, and their prices for DSL Internet access servicesfar from being predatoryare higher than prevailing prices for cable modem service, as are their costs.111 Moreover, an ILEC could not possibly finance a predatory pricing strategy through cross-subsidization. An ILECs basic local exchange rates are subject to rigorous price regulation, including price ceilings, in each of its states. Thus, an ILEC has no ability to raise basic local exchange prices to finance below-cost DSL prices. Similarly, an ILECs switched-access prices are capped, as a result of the CALLS proceeding, at 0.55 cents per minute and its special-access rates are constrained by price cap regulation in all areas that do not exhibit sufficient competition to
109. See Implementation of the Non-Accounting Safeguards of Sections 271 and 272 of the Communications Act of 1934, as Amended; and Regulatory Treatment of LEC Provision of Interexchange Services Originating in the LECs Local Exchange Area, Notice of Proposed Rulemaking, 11 F.C.C.R. 18,877, 18,(1996) (citing Daniel F. Spulber, Deregulating Telecommunications, 12 YALE J. ON REG. 25, 60 (1995)). See also SIDAK & SPULBER, supra note 1, at 93-94 (making same argument). 110. See Applications of Voicestream Wireless Corp., Powertel, Inc., Transferors, and Deutsche Telekom AG, Transferee, for Consent to Transfer Control of Licenses and Authorizations Pursuant to Sections 214 and 310(d) of the Communications Act and Petition for Declaratory Ruling Pursuant to Section 310 of the Communications Act and Powertel, Inc., Transferor, and Voicestream Wireless Corp., Transferee, for Consent to Transfer Control of Licenses and Authorizations Pursuant to Sections 214 and 310(d) of the Communications Act, etc., Memorandum Opinion and Order, 16 F.C.C.R. 9779, 90 (2001). 111. Those higher prices are a product both of the higher costs of DSL deployment and the Commissions asymmetric regulatory requirements, which further raise an ILECs cost of providing DSL service.

or cable modem service at the time of the survey. We also exclude from the sample observations with survey weights equal to zero. We calculate price information for Internet service using the billharvesting portion of the sample, a survey in which only a fraction of the sample participates. Even for consumers who are in the bill-harvesting sample, the price of Internet service is available only for the chosen alternative for each consumer. We impute missing data for dial-up prices using geographic matching within the sample. We impute missing data for DSL and cable modem service using the typical price charged by an ILEC (for DSL) and incumbent cable provider (for cable modem service) in the geographic area where the consumer is located. These prices were obtained from the companies web sites and news reports about price changes. The average price for dial-up was $19.25, for cable modem $41.80, and for DSL $43.08 in the sample for the fourth quarter of 2000 and first quarter of 2001. Table A1 presents the means of the independent variables used in the first stage of the nested logit model. In certain cases, the higher categories of the income and education variables were dropped, because they were not identified.
Table A1: Explanatory Variables in the Nested Logit Model Variable Mean Standard Error Income less than $15,000 0.0746 0.26 Income between $15,000 and $25,000 0.0946 0.29 Income between $25,000 and $35,000 0.1230 0.33 Education less than high school 0.0497 0.22 High school education 0.2404 0.43 Some college education 0.2626 0.44 Age 40.0000 11.58
We used the nested logit routine in the LIMDEP (Version 7) program to estimate the nested logit model and calculate the own-price and crossprice elasticities of demand for Internet access choices. LIMDEP requires the user to specify the tree structure for the model as well as the utility functions for each alternative at each stage. The nested logit routine can then formulate the likelihood function and estimate the nested logit model using maximum likelihood. Figure A1 shows the tree structure, and is followed by the utility functions that we specified. Table A2 presents the coefficient estimates.
Figure A1: Tree Structure for the Nested Logit Model

No Internet

Narrowband

Broadband

Cable Modem
The utility functions for the nested logit model were as follows: [1] [2] [3] [4] [5] [6] U (DSL) = "1 +$*price-DSL U (Cable) = "2 +$*price-Cable U (Dial-up) = "3 +$*price-Dial-up U (No Internet) = $*price-No Internet = 0 U (Broadband) = b1*income1 + b2 *income2 + b3*income3 + c1 *educ1+ c2 educ2+ c3 *educ3+ d *age U (Narrowband) = e1*income1 +e2*income2 + e3 *income3 + f1*educ1 + f2 *educ2 + f3 *educ3 + g *age

where U( ) are the utility functions for the relevant services; price reflects the price for each service; the three income variables are dummy variables reflecting different levels of household income; the three educ variables are dummy variables reflecting the level of education of the head of household; and age is the age of the head of household.
Table A2: Estimated Coefficients from the Nested Logit Model

P (Y = BROADBAND)

Variable Income less than $15,000 Income between $15,000 and $25,000 Income between $25,000 and $35,000 Education less than high school High school education Some college education Age Coefficient -1.4977 -1.1317 -0.9080 -1.3247 -1.0906 -0.3665 -0.0245 Standard Error 0.1680 0.1443 0.1265 0.2067 0.1074 0.0995 0.0036 T-Statistic -8.9170 -7.8440 -7.1790 -6.4100 -10.1500 -3.6830 -6.7910 P-Value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0002 0.0000

P (Y = NARROWBAND)

Variable Income less than $15,000 Income between $15,000 and $25,000 Income between $25,000 and $35,000 Education less than high school High school education Some college education Age Coefficient -1.3159 -0.7672 -0.6039 -0.7729 -0.5869 -0.1504 -0.0197 Standard Error 0.1021 0.0914 0.0859 0.1255 0.0751 0.0784 0.0025 T-Statistic -12.8870 -8.3930 -7.0260 -6.1600 -7.8170 -1.9170 -7.7950 P-Value 0.0000 0.0000 0.0000 0.0000 0.0000 0.0552 0.0000
P (TYPE OF INTERNET ACCESS)
Variable Price DSL Cable modem Dialup Coefficient -0.0284 0.0972 0.4374 1.7474 Standard Error 0.0069 0.2996 0.2899 0.1358 T-Statistic -4.1140 0.3240 1.5090 12.8700 P-Value 0.0000 0.7456 0.1314 0.0000
The estimates indicate that income below $35,000 and lack of a college degree significantly decreases a consumers propensity to choose a broadband access technology. Households that are headed by an older person are less likely to choose a broadband access technology. Finally, increases in the price of the Internet access technologyregardless of the typesignificantly decreases the consumers propensity to choose that access technology.