I fj)'-- Cable and DSL: Broadband Access for Residential Customers

By JERRY A. HAUSMAN, J. GREGORY SIDAK, AND HAL J. SINGER*

To date, most residential customers to the Currently, AT&T is the nation's largest cable Internet have used dial-up modems with a top multiple system operator (MSO). AT&T also speed of about 56.6 kbps [kilobits per second]. controls Excite@Home Corp., the largest pro­ In the past two years broadband access has vider of residential broadband service, with become available via cable modems offered by over 2.3 million subscribers in November 2000. the local unregulated cable provider and via Excite@Home has exclusive contract rights to digital subscriber lines (DSL) offered by the provide residential broadband service over the local regulated telephone company (the in­ cable facilities of its three principal equity cumbent local exchange carrier [ILEC]) and holders (AT&T, Corporation, and Cox competitors who resell DSL using the ILEC Communications, Inc.), which collectively ac­ facilities. Cable modems and DSL offer access count for over 35 percent of the nation's cable speeds about 10-30 times higher than dial-up subscribers. Similarly, Time-Warner is the access and are termed "broadband Internet ac­ second-largest cable provider and has an exclu­ cess." Although Federal Communication Com­ sive contract with Road Runner, the second­ mission (FCC) regulation requires ILEC' s to largest provider of broadband Internet service, sell the use of their facilities to competitors at with 1.1 million subscribers. The competitive below-cost prices, no regulation of cable com­ implication of the exclusive arrangements are panies has occurred. This outcome is curious straightforward: to access an alternative broad­ given that cable companies have a significantly band ISP instead of the ISP affiliated with the greater incentive to distort competition as a cable provider, a user of broadband cable access result of their unregulated monopoly profits has to "pay twice." from their cable operations. This asymmetric Alternative sources of delivery for pro­ regulation by the FCC has led to the "open­ gramming provide a competitive threat to the sig­ access" debate. The open-access debate in­ nificant market power of the cable industry. volves the question about whether the cable Previously, the cable industry has unsuccessfully providers should be required to provide ac­ attempted to control access through control of cess to competing broadband Internet service satellite delivery of video programming, the first providers (ISP's) or whether cable providers alternative medium for multichannel video pro­ can use exclusive contracts with their affili­ gramming. This attempted strategy was blocked ated ISP's. by the Department of Justice (DOJ). Control of Here, we consider the economic incentives broadband Internet delivery of video program­ and actions of the providers of broadband ac­ ming, the second alternative medium for mul­ cess with respect to limiting the usage of broad­ tichannel video programming, arises from band access, including the potential competitive cable-provider control of cable broadband ac­ effects for cable , a sector of the econ­ cess. Internet "video streaming" competes and omy where, to date, system operators have been will compete even more in the future with video able to exercise significant market power. 1 programming offered by cable systems, satellite companies, and television broadcasters.

* Department of Economics, Massachusetts Institute of I. Description of the Broadband Technology, 50 Memorial Drive, Cambridge, MA 02142, Internet Market American Enterprise Institute for Public Policy Research, and Criterion Economics, respectively. Please send corre­ spondence to J. Hausman (e-mail: [email protected]). A. Qualitative Description 1 By "market power" we use the antitrust definition of charging above the competitive price for a significant period Many of the services supported by broadband of time. connections are not available through narrow- 302 VOL. 91 NO. 2 INTERCONNECTION AND ACCESS IN TELECOM AND THE INTERNET 303 band connections. The demand for applications band customers. For Internet ac­ that can be supported only by high-bandwidth cess, we collected data from the ILEC's connections suggests that the product markets providing service in the areas served by the for narrowband and broadband access are dis­ local cable provider. tinct. Functionalities that are only supported by Prices for second telephone lines (used, for broadband connections include real-time video instance, by many AOL [America OnLine] cus­ programming, on-demand video, customized tomers) varied from $7.70 to $47.62 per month. music and video libraries, home networking, Installation costs for a second telephone line real-time programming, interactive multi­ varied from $16.90 to $55.30. Again we amor­ player gaming, high-speed telecommuting, and tized the installation cost for the second tele­ interactive advertising and e-commerce. phone line. Given that the "standard" price for the Excite@Home cable service is $40 per B. Quantitative Analysis month and the price for second lines for nar­ rowband access varies widely from $8 to $48 To answer the question of whether the price per month, plus the standard fee which is na­ of narrowband constrains the tionwide for narrowband ISP's (for example, price of broadband Internet access, it is useful to $21.95 per month for AOL), the data demon­ note that the data demonstrate wide variation in strate that the Merger Guideline test for market second-telephone-line prices across different definition places narrowband Internet access in regulatory jurisdictions. Because most narrow­ a separate market from broadband Internet ac­ band Internet users cannot tie up their first line, cess. The straightforward observation is that the price of a second line is a good estimate for narrowband access prices differ by a factor of the incremental price of narrowband access. It is over 300 percent, while broadband access prices also useful to note that the price of broadband do not vary in any way with these differences. Internet access, as measured by the price of Thus, variations in the price of narrowband ac­ cable service, remains relatively con­ cess cannot explain the variations in the price of stant across these different jurisdictions. Hence, broadband access. Otherwise, when the price of one can infer that narrowband Internet access is a second telephone line changes from $48 to $8 unlikely to constrain the price of broadband per month, we would expect to observe a de­ Internet access. crease in the price for the broadband access The question of market definition can be service. No significant decrease is found, which tested empirically. If it can be shown that nar­ demonstrates the existence of separate product rowband Internet access prices (including the markets for antitrust purposes.2 access charge plus the price of a second tele­ Table l shows the regression results, which phone line) do not constrain broadband Internet use the price of broadband access (either access prices, then a hypothetical monopoly Excite@Home or Road Runner) as the left-hand provider of broadband Internet access could side variable. The price variable is specified in more easily sustain a 5-percent price increase logarithms. The right-hand side variables are an above the competitive market price; hence, the intercept, an indicator variable for Road Runner, a existence of a separate broadband Internet ac­ variable for second-telephone-line prices from cess market is more plausible using the govem­ the ILEC, and variables for population charac­ ment' s Merger Guideline analysis. teristics and density, which could affect demand 3 To conduct an econometric analysis, we or cost characteristics of broadband cable. gathered price data in August 1999 from 41 states and 59 multiple system operators where Excite@Home and Road Runner were then cur­ 2 Some narrowband Internet customers do not use a rently being sold. For cable subscribers, the second telephone line. We have also analyzed the data using broadband access price varied from $34.95 per a weighted average of customers who use a first or month to $64.95 month. We also considered the second telephone line. The results do not differ signifi­ installation fee, which varied from $50 to $150. cantly. 3 The ILEC price of second-telephone-line service is We amortized this installation fee over different treated as predetermined in the regression specification be­ periods in various regression specifications, cause it is set by regulation, not by market forces. Also, a depending on the predicted chum rate for broad- Hausman (1978) specification test did not reject exogeneity. 304 0 (.,, AEA PAPERS AND PROCEEDINGS ·-< MAY 2001

TABLE ]-REGRESSION RESULTS (DEPENDENT VARI ABLE = II. Possible Future Developments Loo OF CABLE BROADBAND ACCESS PRICE PLUS AMORTIZED MONTHLY COST OF INSTALLATION) It is possible that at some point in the future Variable Coefficient SE new technologies will emerge, or existing tech­ nologies will be refined, in such a way that they Intercept 4.86 0.564 8.62 Log price of narrowband access -0.029 0.033 -0.877 will compete effectively with cable-based Inter­ Log population density 0.001 0.010 0.057 net services. However, we believe that, under Log median household income -0.D28 0.064 -0.433 Percentage of population age 65 the current regulatory framework, neither DSL and older -0.006 0.006 -1.16 nor satellite-based Internet service will be able Percentage of population age 35-54 -0.009 0.009 -0.979 to offer close substitutes for cable-based Inter­ Percentage of population under net service within the medium-range time hori­ age 5 -0.016 0.022 -0.757 Road Runner indicator -0.114 0.014 -8.07 zon. Hence, neither will be able to provide the price-disciplining constraint needed to protect Number of observations: 59 consumer welfare. SER: 0.002 R2: 0.600 The relatively slow deployment of DSL to date has limited its ability to discipline any price Notes: Broadband access price is the log of cable broadband access price plus amortized monthly cost of installation. Narrow­ increase by a cable-based provider of broadband band access price is the log of the price of a second telephone Internet access. DSL deployment is constrained line plus second-line fees plus amortization of the installation by technical impediments. DSL is sensitive to cost. the distance that transmissions must travel be­ tween the home and central office. DSL does not work (or work well) if the copper segment The estimated coefficient for the price of exceeds approximately 3-3.5 miles (4.84-5.65 narrowband access, as measured by the price km), which encompasses about 25-35 percent of a second line, is essentially zero, -0.029, of ILEC customers. Also, DSL cannot be pro­ which is extremely small and nowhere near vided where digital loop carrier (DLC) technol­ statistical significance. Thus, the hypothesis ogy has been employed, which includes a large that the price of narrowband access does not part of the southern United States. affect the price of broadband access (trans­ Even if DSL providers were to overcome port) and ISP service is not rejected. Our their technological limitations, significant regu­ finding is that lower narrowband access prices latory barriers prevent them from competing do not constrain the prices charged for broad­ effectively against the cable broadband provid­ band access. Because the price of AOL is not ers. The regional Bell operating companies included in any explanatory variable, its ef­ (RBOC's), which are the primary providers of fect is contained in the estimate of the inter­ DSL, operate within an entirely different regu­ cept coefficient. latory environment than their cable competitors. The findings are quite uniform across dif­ First, the RBOC's are excluded entirely from ferent specifications corresponding to differ­ the core backbone market. Also, the RBOC's ent definitions and amortization periods for face separate-subsidiary requirements that may installation costs. The estimated coefficient of make it more expensive to provide Internet the narrowband-access price variable is found search engines or content of any kind. Also, the to be very small and statistically insignificant Act requires RBOC's to across specifications. We find similar results unbundle their network services at rates below if we limit the sample to Excite@Home the costs of providing them. The FCC has in­ MSO's. We estimated an additional specifi­ dicated its policy of extending unbundling cation by including in the regression the me­ requirements to broadband Internet services, dian household income and the average which decreases the economic incentives to pro­ population density for the relevant markets. vides these services. 4 The asymmetric regula­ Thus, we conclude that the price of narrow­ tory treatment of the RBOC's with respect to band access does not constrain the price of cable providers prevents DSL from being an broadband access. Broadband Internet access is a separate relevant market for competitive analysis and for antitrust purposes. 4 See Hausman (I 997) for a further discussion. VOL 91 NO. 2 INTERCONNECT/ON AND ACCESS IN TELECOM AND THE INTERNET 305

TABLE 2-MARKET-SHARE ESTIMATES MADE BETWEEN percent in the third quarter of 1999 ( Cable 3 I MAY 2000 AND 31 DECEMBER 2000 Datacom News, August 1999, p. 2). It is impor­

Cable tant to note two items when considering broad­ Cable DSL Relevant share band market share. First, the relevant market for 0 Consultancy subscribers subscribers date (percent) the purpose of our discussion is the residential Yunkee Group/Kagan 3,500,000 900,000 31/12/2000 79.5 broadband access market. Because the above TelcChoice 3,000,000 1.400,000 29/8/2000 68.2 Forrester Research 2,227,000 869.000 3115/2000 71.9 numbers include both residential and business Avcmgc: 73.2 sectors, and because cable has little presence in the business sector, the market share for DSL 11 Oates arc reported as day/month/year. providers is overstated. According to an FCC study released in October 2000, the ratio of cable modems to DSL for "residential and small business high-speed lines" at the end of June effective competitor in the broadband Internet­ 2000 was 2.5 (FCC News, 31 October 2000 access market for residential customers. Al­ [table 3)). though we do not discuss potential competition Second, the threat of discrimination against from satellites here, we do not believe that they unaffiliated broadband conduits or broadband will provide significant constraining competi­ content providers remains with this level of tion to cable providers for broadband Internet market penetration by cable modems. By dis­ access in the next few years. criminating against (downstream) access rivals, Looking further in the future, one cannot the vertically integrated firm can distort compe­ ignore the potential impact of both current­ tition by forcing broadband customers who de­ and next-generation wireless Internet access. mand marquee content to choose cable over Third-generation (3G) wireless technology other forms of access. By discriminating against promises to deliver wireless Internet access (upstream) content rivals, the vertically inte­ speeds of up to 2 Mbps to indoor home users grated firm can distort competition by weaken­ and is expected to be implemented in the ing its access rivals' bargaining position with United States in a few years. However, these respect to nonaffiliated upstream suppliers. possible future technological developments do not have a significant current influence on B. Possible Anticompetitive Strategies broadband competition. by Cable Providers

III. Competitive Assessment Full-service broadband providers integrate four inputs of broadband service: (i) broadband A. Penetration Levels of Cable Modems content (e.g., streaming video and audio, mov­ and DSL ies, video conferencing, interactive games), (ii) the aggregation of broadband content and com­ As we described earlier, broadband Internet plementary services (e.g., chat rooms, instant services markets are local in nature. Measures messaging) by a broadband portal, (iii) connec­ of concentration at a local level are not readily tivity to the Internet supplied by a broadband available, however, because carriers only pro­ Internet service provider, and (iv) high-speed vide information on subscribers at the national transport from the home to the ISP supplied by level in their quarterly financial filings. It is only a cable provider, telephone company, or other possible to draw inferences about the average broadband conduit provider. local level of concentration based on a nation­ From these descriptions, there are two anti­ wide measure of concentration. Table 2 shows competitive strategies that a vertically inte­ several estimates of the market share for cable grated firm offering both broadband transport modems and DSL. • and portal services could profitably pursue. As Table 2 shows, cable's market share was, First, an integrated provider could engage in on average, estimated to be 73.2 percent as of conduit discrimination, insulating its own con­ the third quarter of 2000. DSL still lags far duit from competition by limiting its distribu­ behind cable modems and is not closing the gap tion of affiliated content and services over rival as quickly as expected: cable's share was 83.6 platforms. Conduit discrimination could involve 306 AEA PAPERS AND PROCEEDINGS MAY 2001 a range of anticompetitive strategies, from re­ on standards and network externalities provides fusing to distribute an affiliated portal over theoretical and empirical support for the conjec­ competing conduits to making popular content ture that AT&T could impose proprietary stan­ available only to customers using an affiliated dards that would raise the switching costs for its conduit. Second, an integrated provider could subscribers and stifle competition in vertically engage in content discrimination, insulating its related software markets. own affiliated content from competition by The traditional cable strategy of TCI (AT&T' s blocking or degrading the quality of outside predecessor) has been to use its market power in content. Content discrimination could involve a the delivery of programming to expand its con­ range of strategies, from blocking outside con­ trol over the programming itself. Time-Warner tent entirely to affording preferential caching has previously used a similar strategy to limit treatment to affiliated content. competition in programming. Both of these strategies are potentially costly, but the benefits could outweigh the IV. Regulatory Review of Open Access costs in certain situations. For example, a firm engaging in conduit discrimination will To date, the FCC has imposed no regulatory forgo revenues from content distribution over conditions on the provision of broadband access rival platforms. However, there are poten­ by cable providers. In considering the merger of tially countervailing benefits, because with AT&T and MediaOne, the FCC acknowledged conduit discrimination, customers will per­ that the merger could pose anticompetitive ceive the cable conduit as more valuable. threats to emerging markets for broadband This, in turn, will increase the demand for Internet services, but "those harms will be cable transport relative to other forms of avoided if: (a) consumers can choose among transport. Hence, a cable broadband provider various alternative broadband access providers, will engage in conduit (or content) discrimi­ such as DSL, wireless, and satellite; or (b) un­ nation if the gain from additional access rev­ affiliated ISP' s are permitted access to the enues from broadband users offsets the loss in merged firm's cable network." The FCC was content revenues from narrower distribution. satisfied that a competitive market for broad­ To the extent that cable transport providers band access already provided to the former, and compete against DSL and other broadband the agency was convinced that AT&T had com­ transport providers, the reduction in revenues mitted itself to providing the latter. With the from lost customers will be greater. exception of a small trial in Boulder, Colorado, There are several ways in which a vertically AT&T has not provided broadband access to integrated broadband provider can discriminate unaffiliated ISP's. against unaffiliated content providers. First, it The Federal Trade Commission (FTC) and can give preference to an affiliated content pro­ the FCC began their review of the proposed vider by caching its content locally. Such pref­ merger of AOL and Time-Warner in the sum­ erential treatment ensures that affiliated content mer of 2000. The FCC has asked AOL and can be delivered at faster speeds than unaffili­ Time-Warner for additional information on ated content. "open access" to platforms. In a Second, a vertically integrated broadband response to the FCC request, AOL and Time­ provider can limit the duration of streaming Warner promised that they would not discrimi­ of broadcast quality to such an extent nate against ISP's and would let multiple ISP's that they can never compete against cable pro­ use their cable lines. However, the FTC appears gramming. Stated more generally, a vertically reluctant to accept this guarantee. Instead, it integrated firm like AT&T can block any com­ is likely to require the merged AOL-Time­ peting content that it wants to. Currently, Warner to have at least one unaffiliated ISP AT&T and other cable providers limit video signed up before AOL is permitted• to deliver streaming to less than ten minutes. service using the Time-Warner cable network. Third, a vertically integrated firm such as The companies were in talks with the FTC at the AT&T or AOL-Time-Warner could impose time of this writing. proprietary standards that would render unaffil­ The approach of the FCC is especially curi­ iated content useless. The academic literature ous because it requires ILEC's to provide the VOL 91 NO. 2 INTERCONNECTION AND ACCESS JN TELECOM AND THE INTERNET 307 use of DSL facilities at below cost. 5 If the FCC To remedy the risks of conduit and content actually believes that the broadband access mar­ discrimination, regulators should subject any ket is competitive with cable modems, DSL, pending mergers to an open-access provision. and satellite delivery, then no reason exists for In particular, the regulatory agencies should the FCC decision in 1999 that required ILEC's require vertically integrated cable firms to to sell the use of their DSL facilities to compet­ afford unaffiliated ISP's equal and nondis­ itors at below-cost prices. Competition from criminatory access to the combined compa­ cable modems would restrain any anticompeti­ ny's cable modem platform. Doing so will tive actions of telephone companies in the pro­ ensure that the incumbent cable provider does vision of DSL. Because telephone companies not evade or retard the advent of open access, do not provide broadband content and do not will promote investment in the broadband have any market power in broadband (or narrow­ portal market by giving new entrants certain band) Internet portals, they have no economic access to the merged company's cable cus­ incentive to discriminate against downstream tomers, and will limit the cable firm's ability competition. If, on the other hand, competitive to engage in both conduit and content dis­ concerns exist, regulation of cable modems to crimination. require nondiscrimination seems appropriate, because cable companies have significantly REFERENCES greater economic incentive to discriminate against their rivals given their current market Cable Datacom News. Phoenix, AZ: Kinetic power in multichannel video programming. The Strategies, August 1999. current asymmetric regulatory treatment by the Crandall, Robert and Hausman,Jerry. "Compe­ FCC of cable modems and DSL seems espe­ tition in U.S. Telecommunications Services: cially curious. Moreover, the FCC's approach Effects of the 1996 Legislation," in S. Peltz­ does not follow from the goal of regulation, man and C. Winston, eds., Deregulation of which is to hinder the exercise of anticompeti­ network industries. Washington, DC: Brook­ tive market power. ings Institution Press, 2000, pp. 73-111. FCC News, "High Speed Services for Internet V. Conclusion Services-Subscribership as of June 30, 2000." Washington, DC: Federal Communi­ Cable firms are positioned to dominate the cations Commission, 31 October 2000. broadband industry as they have dominated Hausman,Jerry. "Specification Tests in Econo­ the delivery of multichannel video program­ metrics." Econometrica, November 1978, ming. With control of both the broadband 46(6), pp. 1251-71. content and the pipes, a large footprint en­ ___ . "Valuing the Effect of Regulation on courages cable firms to discriminate against New Services in Telecommunications." their unaffiliated content and conduit rivals. Brookings Papers on Economic Activity, Mi­ croeconomics 1997, pp. 1-38. Hausman,Jerry and Sidak, J. Gregory. "A Con­ sumer-Welfare Approach to the Mandatory 5 For a review of FCC policy see Hausman and Sidak ( 1999) and Robert Crandall and Hausman (2000). Below­ Unbundling of Telecommunications Net­ cost prices set by the FCC, due to failure to consider the works." Yale Law Journal, December 1999, effect of sunk costs, is discussed in Hausman (1997). 109(3), pp. 417-505.