N E U.S. ATURAL XPERIMENTS IN B R ROADBAND EGULATION Thomas W. Hazlett, Anil Caliskan George Mason University School of Law George Mason University Law and Economics Research Paper Series 08-04 This paper can be downloaded without charge from the Social Science Research Network at http://ssrn.com/abstract_id=1093393 Natural Experiments in U.S. Broadband Regulation Thomas W. Hazlett Professor of Law & Economics Director, Information Economy Project George Mason University [email protected] Anil Caliskan Research Economist Information Economy Project George Mason University [email protected] Nov. 29, 2007 The authors wish to thank Kathryn Zeiler, Howard Shelanski, and seminar participants at Georgetown Law Center for useful comments. They, of course, remain blameless for any remaining errors or omissions. Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation Abstract Network neutrality (NN) regulations governing how broadband Internet Service Providers (ISPs) package and price high-speed “last mile” access are being considered. Advocates say such rules are necessary to separate transport and application layers of the Internet, protecting “innovation at the edge.” Opponents argue that such rules are unnecessary and could block efficient forms of vertical integration. Either side points to future developments – NN advocates to anti-competitive foreclosure by ISPs, opponents to a deterrence of investment due to regulatory disincentives – as primary arguments. Such predictions are sharply contested. A natural experiment, however, may yield market evidence. U.S. residential broadband markets have been subject to “open access” rules, analogous to NN regulation, that have varied across time and technologies. These policy switches allow empirical measurement of consumer reactions. Using the simple metric of market penetration, which incorporates both demand and supply effects, the relative effectiveness of rival policy regimes can be appraised. This paper considers three distinct regimes governing the two leading technologies for residential broadband, cable modems (CM) and digital subscriber line (DSL) service. Prior to 1Q2003, CM service was unregulated (and has remained so), while DSL was subject to network unbundling mandates that included “line sharing” rules enabling rival Internet Service Providers to access last-mile loops at incremental cost. CM service enjoyed nearly a two-to-one market share advantage during this period. Following 1Q2003, when line-sharing rules were eliminated by the Federal Communications Commission (FCC), effectively raising wholesale access prices, DSL subscribership sharply increased relative to trend and to DSL trend and to contemporaneous CM subscriber growth. This pattern continued unabated following further deregulation in 3Q2005, when the FCC classified DSL as an “information service”. By year-end 2006, DSL subscribership was about 65% above the linear growth trend established in the regulated pre-1Q2003 era, some eight to ten million more households than predicted. This evidence, in sum, suggests “open access” broadband regulation deters subscriber growth, potentially important empirical input into the ongoing regulatory debate. 2 Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation I. INTRODUCTION A debate is taking place over the optimal economic structure of computer networks. Citing the dominance of two rival broadband networks in residential markets, net neutrality (NN) advocates argue that Internet Service Providers (ISPs) will exploit “gatekeeper” positions to decrease competition in complementary markets, punishing consumers. These broadband networks are seen to be in a position to extract rents from customers or content suppliers by blocking (and unblocking) the flow of e-commerce, a variant of the vertical foreclosure argument (Farrell & Weiser 2003; Economides 2007).1 Government regulation is advanced as a remedy. To anti-competitive behavior, rules are proposed2 to prevent broadband ISPs from preferentially supplying improved access to particular applications (particularly those owned by, or paying, the ISP). Whatever the merits of these policies, such limits carry costs. The reduced flexibility afforded consumers, ISPs, caching services, and content providers in contracting for services constrains the range of business models over which markets may optimize. The costs of regulation are substantial if the benefits of vertical integration are large. Such integration is commonly observed in the development of communications networks (Owen 2007; Hahn & Litan 2007). Setting aside potential efficiency gains sacrificed by vertical limits, NN rules designed to constrain ISPs prohibit what is presumably (by revealed preference) profit-maximizing behavior. This directly reduces incentives for investment in broadband networks, and forms the crux of the argument against mandated NN (Hahn & Wallsten 2006). These direct effects may be offset in part or in whole by indirect effects that encourage innovation in applications and thereby increase demand for Internet access. Where the increase is sufficient in scale to dominate the costs of regulation, ISPs will face stronger incentives to invest in network capacity, despite being constrained from achieving a local profit maximum. This rationale is provided in the “end to end” argument supplied by Lessig & Lemley (2000), Wu (2003) and others. Their view posits that transactional and organizational efficiencies result from separating ownership of basic transport functions (via physical networks connecting backbones and ISPs) from content. Specifically, removing ISP control over applications facilitates competitive entry into creation of the latter. By removing the prospect of vertical integration (by ownership or contract), traffic is said to flow freely “end-to-end,” stimulating innovation “at the edge.” This view is disputed in its portrayal of Internet architecture (Yoo 2004). Whatever differences exist in this lively discussion, there is a fundamental consensus that the gulf between policy positions needs to be bridged with empirical evidence. The ostensible purpose of NN rules is to increase efficiency and promote conditions that improve the products consumers use. Such rules entail trade-offs, 1 For an elaboration on the argument for NN regulation, see Wu (2003) 2 It is not clear precisely what NN rules would be imposed (Peha 2007). 3 Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation however, that may entirely offset intended effects, reducing consumer welfare. Simply stated, incentives for network investment decline when network owners lose a set of valuable property rights, as in the case where regulation is imposed to block certain pricing or bundling models. Whether a corresponding intensification of demand via edge product stimulation occurs remains a fact question. This point has been made by Lawrence Lessig, who has articulated the case for net neutrality regulation. In a 2006 forum, he characterized the argument that such rules “would burden investment” a “fallacy,” and wedged this verdict on observed marketplace outcomes. “Of course, this is an empirical claim.”3 The source for Lessig’s conclusion was attributed to Gerald Faulhaber (2002), which evaluated the impact of “open access” regulation for deployment of broadband services. Such rules, similar to NN policies, mandate that a broadband network permit rivals to lease its facilities in order to supply competitive retail service. Open access can be applied both to cable modem (CM) access and to digital subscriber line (DSL) service provided over telephone networks. In the latter instance, where it has been implemented in the U.S., it can also involve “line sharing,” where entrants are permitted to send high-speed data over local loops simultaneously used to make voice phone calls over standard (circuit switched) technology. Like NN rules, “open access” has been offered as a policy protecting “end to end” (Lemley & Lessig, 2000), eliminating gatekeepers by yielding subscribers and application vendors a choice of alternative ISPs. Hence, in arguing for NN, Lessig offered Faulhaber’s conclusion as empirical justification: [I]t is difficult to sustain the argument that regulatory policy regarding open access for cable or line sharing for digital [subscriber line] service has in any way been an impediment to broadband deployment. If there have been impediments, to deployment, they have been overwhelmingly on the supply side (p. 241). The more telling component of this analysis, however, specified the conclusion and contained a crucial caveat. Given that cable TV operators were then, as now, unregulated (i.e., not subject to open access mandates), Faulhaber focused on the effect of open access on DSL carriers, primarily incumbent local exchange carriers (ILECs), of which the largest are the Bell companies. He wrote: [T]he regional Bells are deploying broadband as fast as they can as a competitive necessity, and they have been willing to suffer substantial internal inefficiencies to do so. It is likely that the cost increase due to the line-sharing mandate is small compared to these other costs and will have no effect on the deployment speed of digital subscriber lines. Ultimately, this is an empirical issue, and the hard evidence of what these costs are 3 Key Issues in Telecommunications Policy, AEI-Brookings Joint Center for Regulatory Studies (May 10, 2006), http://www.aei-brookings.org/events/page.php?id=155. 4 Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation and how they compare to the relevant market incentives is not yet available (Ibid.). This interpretation motivates further empirical analysis. Since Faulhaber’s conclusion was offered, more systematic deployment data have become available.4 Of particular importance is that two fundamental policy regime switches occurred in 2003 and 2005, both with key implications for residential broadband access suppliers. While NN has not specifically been imposed, open access regulations have, carrying similar implications for consumers and producers. NN and open access are close policy substitutes, and the market’s reaction to open access serves as a proxy for the likely effects of NN. This logic flows directly from Lawrence Lessig’s use of Faulhaber’s conclusion, gleaned from implementation of open access obligations on telephone carriers offering DSL, as factual support for the position that NN will not deter broadband deployment. Hence, open access regulation in U.S. broadband markets offer a potentially useful natural experiment. Testing the deployment effects of regulatory changes already occurring in broadband markets yields evidence on the likely effects of NN mandates. Given that the policy debate hinges on empirical verification, these data should prove crucial to the policy debate ongoing. II. BROADBAND REGULATION 1. General Approach Rules were initially imposed on telephone company-delivered digital subscriber line (DSL) services, and then largely removed by regulators. Residential broadband markets then offer a test bed for the effectiveness of open access policies, one that incorporates the potential market power of broadband internet service providers (ISPs). To the degree that existing providers inefficiently restrict consumers’ choices, regulatory constraints more easily improve outcomes. This inquiry focuses on the broadband deployment as proxied by CM and DSL residential subscribership. This is the general approach suggested by Faulhaber (2002) and elsewhere. Residential subscribership is a simple, highly-aggregated metric, signaling an outcome of key interest to policy makers. While examining the effects of regulation on DSL growth, general broadband trends in the U.S. and technological developments that potentially affect DSL subscribership are proxied by U.S. CM and Canadian DSL and CM subscriberships. A number of alternative specifications are used 4 Interestingly, Faulhaber has also rendered an empirical assessment that a related regulatory policy for voice (narrowband) services, unbundling local loops, was ineffective in advancing local telephone competition (Faulhaber 2003). 5 Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation to test the general hypothesis that broadband regulation (as in open access rules) increase efficiency. 2. Three Regimes for Regulating DSL Three identifiable regulatory regimes have governed DSL. As CM service has been consistently unregulated (no open access mandates) from inception in the mid- 1990s, the episodic variation of DSL regulation enables empirical analysis comparing DSL v. CM outcomes. Regime I: Regulated DSL with Line Sharing Obligations (pre-2003) Cable TV operators began offering CM services in 1995 (Rosston 2006, p. 6) without any obligation to share network infrastructure with rival service providers. A cable system operator such as TCI or AT&T (which purchased TCI in 1997) could offer broadband Internet access to retail customers via vertical integration, or via an exclusive contract with an ISP such as @Home. Alternatively, an independent ISP such as Earthlink could negotiate an agreement to sell Internet access to retail customers using the cable operator’s network, but such deals were unregulated, A campaign requesting mandatory wholesale access, at reasonable terms and conditions, quickly materialized, led by America Online (AOL) and GTE (a large local telephone carrier later acquired by Verizon) (Esbin 1998, 2000). These efforts were unsuccessful.5 A 1999 FCC report concluded that access regulation would risk deterring investment in the rapidly evolving market (FCC 1999). The FCC later classified CM access as an interstate information service, categorically exempting it from common- carrier or open access obligations at the federal, state, or local level. The U.S. Supreme Court upheld the FCC’s determination in June 2005 (Brand X, 2005). In contrast to unregulated CM services, DSL services were regulated from their inception in the mid-1990s. In particular, incumbent local exchange carriers (ILECs) supplying DSL faced three major obligations. First, under the Computer III regime, telephone companies were required to provide the broadband transmission component of DSL services on a common-carrier basis (FCC 2005, pp. 19-20). Second, under the Commission’s unbundling rules, telephone companies were mandated to provide the copper loops used to provide DSL service on an unbundled basis.6 This enabled competitive local exchange carriers (CLECs, or dCLECs for those entrants specializing in data services) to place their switches in a telephone company’s central office and to connect that equipment to an unbundled loop supplying DSL services to end users. Third, the FCC’s “line sharing” rules required telephone companies to lease just the high- 5 The AOL/Time Warner merger, consummated in early 2000, imposed unique third party access obligations. The provisions required the merged firm to offer AOL Broadband only after permitting two independent ISPs to utilize Time Warner Cable infrastructure. The rules did not regulate wholesale prices, nor did they regulate Time Warner’s Road Runner broadband ISP. 6 47 C.F.R. § 51.319(a)(1). 6 Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation frequency portion of the loop (“HFPL”) used to provide DSL services (FCC 1999). Federal and state regulations then set the price for the HFPL far below the price for an unbundled loop as a whole, substantially reducing dCLEC costs.7 Regime II: DSL Partially Deregulated (ex-Line Sharing) In February 2003, the FCC eliminated DSL line sharing rules.8 This meant that, in order to supply DSL service to customers over an ILEC’s lines, dCLECs would have to pay for the entire local loop or strike a commercial agreement with the carrier to share a loop. The rationale for the reform was that, with lessened network sharing obligations, telephone carriers would invest more heavily in bringing broadband services to residential customers (Crandall 2005, p. 127). FCC member (later Chairman) Kevin Martin voted for the measure, noting “that competition--not regulation--is the best method of delivering the benefits of choice, innovation and affordability to consumers." 9 The policy switch was criticized by others, however, as a measure that would unleash market power, slowing broadband deployment. “For consumers, it's a bleak day, according to the Consumer Federation of America. There are about 40 million DSL subscribers now in the United States, and today's decision will likely choke off any future growth, one consumer group said.”10 Hence, a natural experiment was created. As developments in the residential broadband markets have played out, subscriber growth trends across the regime switches can now be observed, shedding light the potential effects of regulation. Regime III: Unregulated Cable/Unregulated DSL In August 2005, the FCC eliminated the Computer III rules with respect to DSL services. This removed the remaining open access regulations when Internet connections are bundled with transport. With the Commission determining that DSL fell under Title I of the Communications Act, broadband Internet access became treated as an “information service” exempt from common-carrier regulation.11 This essentially put DSL services on regulatory parity with CM service. The rationale, as with the line-sharing deregulation, was that reducing network sharing mandates would improve operator incentives to deploy, upgrade, and market broadband access. Once again, proponents of regulation argued that the FCC’s deregulatory steps would have negative consequences for broadband customers and the Internet generally.12 7 47 C.F.R. § 51.319(a)(1)(i). 8 FCC, FCC Adopts News Rules for Network Unbundling Obligations Of Incumbent Local Phone Carriers (Feb. 20, 2003); http://hraunfoss.fcc.gov/edocs_public/attachmatch/DOC-231344A1.pdf. 9 Ben Charny, FCC Loosens Broadband Rules, CNET NEWS.COM (Feb. 20, 2003), http://news.com.com/FCC+loosens+broadband+rules/2100-1033_3-985313.html. 10 Ibid. The CFA attributed the slowdown in growth to “higher prices and fewer choices for consumers in high-speed services.” 11 Federal Communications Commission (Sep. 23, 2005), op cit. 12 Andrew Jay Schwartzman of the Media Access Project was quoted as saying: “This is a bad day for the Internet. I think it means higher prices and less competition and threatens the growth of the Internet.” 7 Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation III. EMPIRICAL OUTCOMES The premise for open access rules is that consumer welfare will increase when cable modem (CM) and digital subscriber line (DSL) network owners are legally constrained to permit rival ISPs to serve retail broadband subscribers via their facilities. Via such independent, non-integrated ISPs, end users may obtain unfettered access to Internet content and applications, as ISPs unaffiliated with the underlying broadband network have no business interest in inefficient blocking. This not only parallels the theoretical argument for Net Neutrality (NN), it operates as a similar policy intervention, setting regulatory terms for broadband service providers by mandating user access to vertical services. In either case, common carrier regulation supersedes market forces (Owen 2007). Indeed, in the policy debate open access and NN are portrayed as substitutes. Advocates for NN frequently base their argument for regulation on the elimination of DSL’s open access regime. For instance, THE NEW REPUBLIC editorialized for NN legislation in June 2006 on the grounds that, during the previous year, the FCC “exempted telecoms that provide Internet connections from [open-access] restrictions, dealing a blow to both entrepreneurship and political discourse."13 Similarly, Vinton Cerf, co-developer of the IP/TCP protocol and “Chief Technology Evangelist” for Google, recommends implementation of NN due to elimination of open access. “Cerf said that he thought things were better before 2005 when broadband providers were controlled by common carriage rules that prevented providers from discriminating in terms of what traffic was carried. ‘It protected the Internet,’ he said.”14 The parallel nature of these policy approaches allows empirical evaluation of NN prior to its adoption. Herein, we examine how broadband subscribership responds to changes in broadband regulations, testing the implications of open access rules on the evidence yielded by subscriber choices. We, in particular, test whether the U.S. DSL subscriber growth decreases with deregulation after controlling for the concurrent U.S. CM and Canadian DSL and CM penetrations. The main focus is on the effects of the “line sharing” deregulation in 1Q2003, but we also provide an analysis of the 3Q2005 deregulation even though our dataset contains few observations from this second deregulatory period. The pre-deregulation trends are examined as well. FCC Reclassifies DSL as Data Service - Week of 9/20/05, Northern Light; http://www.centerformarketintelligence.com/analystviews/20050920-WeeklyReport.htm (visited April 5, 2007). 13 The Editors, Open Net, THE NEW REPUBLIC (June 19, 2006); http://www.tnr.com/doc. mhtml?i=20060626&s=editorial062606. 14 Wayne Rash, Net Neutrality Advocates Face Off, eWeek (July 17, 2006); http://www.eweek.com/article2/0,1895,1990357,00.asp. 8 Thomas W. Hazlett & Anil Caliskan, Natural Experiments in U.S. Broadband Regulation If open access promotes efficiency, the following hypotheses cannot be rejected: Period/Regime/Null Prediction (1) pre-1Q2003: CM unregulated, DSL regulated with line sharing Prediction: DSL subscriber growth will exceed CM subscriber growth. (2) 1Q2003-4Q2006: DSL line sharing eliminated 1Q2003 Prediction: DSL subscriber growth will decline from trend . (3) 3Q2005-4Q2006: DSL classified “information service” 3Q2005 Prediction: DSL subscriber growth will further decline from trend. 1. Growth of DSL vs. Cable Modem Prior to 1Q 200315 While DSL and CM technologies were developed at roughly the same time, unregulated cable companies expanded the availability and penetration of their services more aggressively than regulated telephone companies. By year-end 1999 CM dominated the emerging residential broadband market: residential and small business DSL lines totaled just 0.29 million, while CM subscribers numbered 1.40 million. CM continued its dominance through year-end 2002, when it served 11.34 million, double the number of DSL lines (5.53 million) according to the FCC. See Figure 1. [Insert Figure 1 here] While factors unrelated to regulation may help explain CM’s dominance, these data -- absent further evidence – are inconsistent with the hypothesis that open access promotes broadband deployment. Moreover, some Wall Street analysts concluded that regulatory factors did play an important causative role in the relatively quick deployment of cable modem services. According to Blake Bath of Lehman Brothers: The reason that the cable companies really stepped up their investment in 1997 and beyond was they were not regulated, they weren’t forced to open up their networks. There were multiple revenue streams that they could address. They could price the services however they wanted.16 Investor behavior provides further evidence of the regulatory linkage. In a study of 29 events which, between Jan. 1998 and Oct. 2000, significantly impacted the 15 Research on the “broadband race” under this regime has been evaluated in Bittlingmayer & Hazlett (2002), which is summarized here. 16 Quoted in Adam Thierer, Broadband and the Markets: Perspectives from the Investment Community, Cato Institute Policy Forum (July 24, 2001), p. 15; http://www.cato.org/events/transcripts/010724et.pdf. Industry analysts routinely cited a more favorable regulatory climate for cable operators vs. telephone carriers when assessing broadband markets. See, e.g., Jupiter Research, US Broadband Household Projections: Tier Services and Pricing to Drive Demand (Feb. 3, 2003). 9