A Future Friendly Digital Economy Strategy: Submission to the Government of Canada's Digital Economy Consultation — Appendix 2
IV. The Berkman Study's Conclusions are not Supported by Qualitative Analysis
64. To advance its case for an expanded network unbundling regime in the United States, the Berkman Study attempts to buttress its quantitative analysis with extensive discussions of unbundling policies in various OECD countries. While it does not suggest that every country that has pursued unbundling has achieved success through such a policy, its general conclusion is that unbundling has generally had favorable effects. As we explain below, like the quantitative evidence, the qualitative evidence does not support the Berkman Study's conclusions.
A. The Alleged Successes
65. The Berkman Study concludes that network unbundling has generated competition, lower prices, and greater broadband penetration in several Nordic countries, the Netherlands, France, the United Kingdom, and Japan. A careful examination of available data for each of these countries suggests that the Berkman Study's conclusions are either incorrect or, at best, vastly overstated.
1. The Nordic Countries and Netherlands
66. A cursory look at the OECD broadband penetration rankings reveals that a large share of the best performers are cold, northern countries – Denmark, Norway, Sweden, Finland, Netherlands, and Iceland occupy the top six positions. Countries with more moderate climates, such as Italy, Spain, New Zealand, and Australia are far down the OECD list. Equally important, despite substantial differences in policy across these Nordic countries, all are converging to a very similar level of broadband penetration as Figure 1 shows.Footnote 70 This might suggest that cultural, weather, or other demand factors explain a large share of the variance in penetration, but the Berkman Study chooses to focus on network access policies. Its discussion of even the "best" performers is not persuasive.
Source: OECD [Description of figure 1]
67. In discussing the Nordic countries, the Berkman Study is somewhat equivocal about the effects of network unbundling. With respect to Denmark, for example, it simply concludes that
Competition developed both within each technological platform, and across platforms; to some extent benefiting from unbundling when it was available, and to some extent benefiting from relatively low levels of investment necessary to upgrade an existing infrastructure like cable….Footnote 71
The report does not point out to the reader that cable modems exceed the number of unbundled and shared lines in Denmark,Footnote 72 or that cable modem competition in Denmark was highly developed before the recent growth of unbundled lines to deliver DSL.Footnote 73 Thus, intra-platform competition appears to have followed inter-platform competition in Denmark, and the available empirical evidence suggests that such inter-platform competition is far more important than intra-platform competition.
68. In Sweden, platform competition has also been responsible for more broadband subscriptions than has network unbundling. As of the first quarter of 2009, cable connections and non-incumbent telco fiber to the home, provided largely over government fiber networks, accounted for 1.2 million connections while DSL over unbundled and shared loops for only 600,000 connections, and the number of lines delivered over unbundled/shared lines is actually declining.Footnote 74 Much of Sweden's "success' derives from its government-provided fiber, not from network unbundling. Nevertheless, the Berkman Study relies on the views of the Swedish regulator:
Convinced by the perceived success of unbundling in fostering competition, investment, and innovation in its broadband markets, concerned about managing the transition to next generation networks, and possibly smarting from the long fought battle over bitstream access, the Swedish regulator PTS concluded that it would best manage the transition to next generation connectivity by imposing functional separation on its incumbent.Footnote 75
69. This is a curious conclusion, given that the incumbent has not been among the leaders in the EU in network investment and only has 12 percent of the country's fiber connections. The transition to next generation networks in Sweden has been over networks built largely by municipal governments, not by the incumbent operator.
70. The Berkman Study's dismissal of the effect of unbundling in Finland is similarly curious. It asserts that "Finland was the first Nordic country to introduce unbundling, in 1996. Unbundling seems to have had little or no effect in the Finnish market, however."Footnote 76 But according to latest ECTA data, broadband delivered over unbundled and shared loops accounted for 25 percent of all broadband lines while they accounted for only 18 percent in Sweden, a country that the Berkman Study argues is a "perceived success of unbundling." Notably, Finland – an early adopter of network unbundling – has not deployed any fiber to the home according to the ECTA data.
71. Finally, in the case of Netherlands, the Berkman Study suggests that network unbundling plays a critical role in allowing facilities-based entrants, the cable companies, to address customers in areas where they lack facilities. But unbundled and shared lines accounted for only about 680,000 lines in March 2009 while broadband over cable accounted for 2.2 million lines according to the latest ECTA data. And despite the Berkman Study's glowing description of the incumbent's (KPN's) decision to offer unbundled fiber to its competitors, the latest ECTA data show no such unbundling. Perhaps more important, the ECTA data show that entrants have 140,000 subscribers over their own fiber, but KPN has none, and the latest data from IDATE show that KPN has yet to even deploy FTTH.Footnote 77
72. The Berkman Study makes much of differences between France's and Germany's unbundling policies. It paints the French policy since 2002 in very rosy terms, pointing out that an EC infringement action forced the French regulator to mandate lower rates for unbundled and shared loops. Subsequently, in 2003 new entrants expanded rapidly. By contrast, the German regulator is described as unwilling to enforce an unbundling regime that works.Footnote 78 These differences in policy lead the Berkman Study to opine that: "Despite having roughly similar GDP per capita (Germany slightly higher) and population concentration (France slightly higher), the two countries present quite different trajectories."Footnote 79 What trajectories is it referencing? A quick look at Figure 2 shows that the trajectory of broadband penetration since 2003 is very similar in the two countries. Why worry about differences in unbundling policy, given this obvious fact? Figure 2 shows that Germany actually experienced slightly greater broadband growth after 2003, a result that is largely due to the greater growth of inter-platform competition from cable television. Germany now has about 2 million cable broadband subscribers – more than twice the number in France.Footnote 80
73. Moreover, both France and Germany remain far below the Nordic countries and Netherlands in terms of broadband penetration. Neither has yet to deploy much fiber to the subscriber. France had only 40,000 fiber loops in the first quarter of this year, out of a total of more than 18 million broadband connections; Germany had none. By contrast, Verizon alone had 3.3 million FTTH broadband connections at the end of the third quarter of 2009. Thus, while unbundling has created competition among suppliers of DSL in France, it has not led entrants or incumbents to invest much in fiber to the subscriber. Despite the Berkman Study's assertion that France Telecom has responded with "higher investment," its capital spending as a share of revenues remains mired in the very lowest echelons of capital spending among EU-15 incumbents.Footnote 81
3. The United Kingdom
74. Two of the authors of this declaration were co-authors of an earlier study of network separation policies that is being published this year in the Federal Communications Law Journal. In that article, we reviewed in great depth the results of Ofcom's new (2005) unbundling and functional separation policy. We concluded that under this policy the rate of growth of the UK's broadband subscribers had declined relative to the growth in the overall EU-15.
75. Specifically, we showed that:
A comparison of UK broadband growth with growth in the EU-15 yields a similarly bleak conclusion about the effects of functional separation. According to ECTA data, between September 2002 and September 2005, when the new Ofcom policy went into effect, UK broadband lines increased at an annual rate of 76 percent while EU-15 broadband lines rose at a rate of 54 percent. Thus, prior to the change in policy, the rate of increase in UK broadband lines was 41 percent greater than the rate of increase in the EU-15. In the three years following the implementation of the new Ofcom policy, UK broadband line growth fell to 21 percent, and EU-15 broadband line growth fell to 23 percent.Footnote 82 …Thus, the new policy has been associated with a severe decline in UK growth relative to the growth in the EU-15. Indeed, the UK broadband growth rate is now less than the average rate for the entire EU-15, and broadband penetration in the UK has fallen relative to EU-15 penetration in the three years that the policy has been in place.Footnote 83
76. The recent ECTA data do not alter this conclusion; UK broadband subscriptions have grown more slowly than EU-15 subscriptions over the 3.25 years since the new unbundling regime went into effect in the third quarter of 2005. Nor do the ECTA data show any fiber deployment in the United Kingdom through the first quarter of this year. The policy has simply not worked to accelerate the growth of broadband or encourage deployment of a Next Generation Network.
77. The authors of the Berkman Study would have us believe that Japan's policy of network unbundling has been an unqualified success, although it admits that facilities-based competition is driving fiber deployment. It claims that the Japanese regulator's decision to require fiber unbundling has not deterred fiber investment: "…the overall level of investment in the fiber market questions the argument that open access deters investment."Footnote 84 However, we can find no evidence that NTT has actually delivered a significant number – or, for that matter, any number – of unbundled fiber connections. Fiber "unbundling" has not deterred NTT's investment in FTTH for a simple reason: fiber unbundling has not been implemented in a meaningful manner.
78. Contrary to the impression created by the Berkman Study, Softbank (Yahoo!BB) states in its most recent annual report that it is losing broadband DSL customers steadily, and it reports no customers over NTT's fiber plant. Indeed, the Softbank 2009 annual report downplays any discussion of the decline in its wireline broadband business, simply stating that it is reducing capital expenditures in its fixed-wire operations: "…the segment's profitability and cash flow are showing marked increases. This reflects a lower depreciation from the end of major capital expenditures, combined with a structural transformation toward businesses."Footnote 85 It is not investing in fiber to the home or any other form of residential connections, but rather shifting its attention to the business market.
79. In short, the unbundling era in Japan appears to be fading rapidly. DSL subscriptions have been declining for three years. Fiber connections, supplied by NTT and a host of other platform-based competitors now substantially exceed DSL connections. The principal competitors of NTT in delivering broadband over fiber optics are the country's power companies, not the companies who have used NTT's copper loops to deliver DSL service.Footnote 86 At this point, these power companies are deploying fiber without any apparent concern that they will have to unbundle their fiber networks. Platform competition, not network unbundling, is driving fiber deployment in Japan.
B. The Doubtful Cases or Outright Regulatory "Failures"
80. The Berkman Study reviews the regulatory developments in several OECD countries, citing their failure to embrace network unbundling, or the failure to implement it assertively, as a regulatory mistake. These countries include South Korea, Germany (already discussed), Italy, New Zealand, and Canada.
1. South Korea
81. No matter how the Berkman Study tries to characterize it, the South Korean broadband policy is a triumph for platform-based competition. The Study argues that:
The South Korean experience is more ambiguous on access, pointing more toward heavy government investment. Both (Japan and South Korea) suggest that a strong, professional regulator, exercising effective power over incumbent providers, can foster significant market development and competition.Footnote 87
82. But this "strong" regulation essentially amounted to subsidizing capital investments by the cable and electric utility companies, but discouraging these companies from offering retail broadband Internet services over their own platforms. As a result, new entrants sprang up offering broadband services over their own platforms and over the cable companies' and electric companies' networks.
83. As the Berkman Study correctly observes, this platform competition led to rapid growth in Korean broadband before any unbundling requirements were imposed on the incumbent telco, KT, in 2002. By that time, Korea had by far the highest broadband penetration in the OECD –21.8 percent versus an average of 4.8 percent in the OECD. Since 2002, Korea's penetration has grown very slowly, expanding by less than 50 percent to 32 percent in December 2008 while the average OECD country has nearly quintupled its penetration to 22.4 percent. Four Nordic countries and the Netherlands have surpassed Korea since its new unbundling requirement took effect in 2002.Footnote 88 Thus, unbundling had nothing to do with South Korea's spectacular early broadband growth.
84. The Italian regulator required network unbundling in 2001 and then required functional separation of Telecom Italia in 2006-08. As of March 2009, entrants using unbundled or shared lines accounted for 27 percent of all lines. The principal problem in Italy had been the absence of platform competition since it had virtually no cable television. Platform competition did emerge in the form of Fastweb'sFootnote 89 fiber deployment in large cities, but this deployment has now been halted. Fastweb has chosen to grow only through leasing Telecom Italia's loops now that functional separation of TI assures them of rapid access to TI's loops as TI's own retail division.Footnote 90 Thus, contrary to the Berkman Study's assertion that Fastweb will continue to deploy fiber and share it with TI, Fastweb has chosen to abandon further fiber investment because of recently improved access to TI's unbundled loops. Fastweb's recent decision is evidence of the adverse effect of mandated network sharing on investment in new technologies. But Italy's relatively low broadband penetration probably has little to do with its regulatory policy. Rather, it is likely a reflection of low personal computer penetration in the country.Footnote 91
3. New Zealand
85. New Zealand had decided not to mandate network unbundling in 2003, but the Communications Ministry changed course in 2006. The government imposed network unbundling and structural separation of Telecom New Zealand in December 2006, undoubtedly because broadband penetration was relatively low, and there was very little platform competition. Since that time, New Zealand's broadband penetration has grown more rapidly than the OECD average, but New Zealand's penetration is still below the OECD average. Moreover, there is no sign of fiber to the premises deployment by Telecom New Zealand or its competitors.
86. Canada's broadband policy has been similar to that in the United States. Canada has very strong platform competition from cable, a fact that propelled it to second place in OECD broadband penetration in 2001. As in the United States, cable broadband connections have exceeded DSL connections from the very beginning. Thus, inter-platform competition was extremely strong in Canada from the outset. Canada has had network unbundling regulations in place since 1997, but few entrants have succeeded in using unbundled loops for delivering broadband or other telecommunications services to residential customers perhaps because of the strength of inter-platform competition in most regional Canadian markets. Most of the local entrants using unbundled loops were unable to compete in this environment and either failed or were acquired. The Berkman Study is incorrect, however, when it asserts that "[t]here are no smaller entrants of note."Footnote 92 MTS-Allstream has been using Bell Canada's and Telus's local loops for years, but with little apparent effect on broadband penetration because its principal focus has been the business market, not the residential market. Thus, while unbundled loops have been available in Canada for more than a decade, they have not been used successfully by entrants to compete in the mass market for broadband services.
87. The Berkman Study has failed to unearth any systematic anecdotal information that mandated network unbundling has increased broadband penetration in OECD countries. This is hardly surprising given the preponderance of empirical studies which conclude that unbundling has limited or no effect. Its narrative does often suggest, however, that platform competition is important in driving competition in countries such as South Korea, the Nordic countries, Canada, and the Netherlands. This too is consistent with the published empirical evidence.
88. The Berkman Study's focus on unbundling of copper loops to promote intraplatform competition is misplaced. The issue for U.S. regulators today is not how to increase competition in DSL services, but rather how to encourage the development of new, faster networks. Even if unbundling copper loops had been a mild success in promoting DSL, this result would not suggest that a similar policy would be successful in promoting the development of new fiber-based networks. Indeed, we have shown that the empirical literature on network investment and access regulation shows that mandated access regulation reduces the incentive to invest in new network technologies.
89. The Berkman Study is critical of the United States and Canadian policies for not being sufficiently aggressive in promoting network sharing and is correspondingly approving of a general European regulatory bias towards encouraging such sharing. Unfortunately for the Europeans, their more aggressive regulatory policy has resulted in far less deployment of fiber in the European Union than in North America as demonstrated by Figures 3 and 4.
Sources: Canadian Households from www40.statcan.ca/l01/cst01/famil55a-eng.htm; U.S. Households, from http://quickfacts.census.gov/qfd/states/00000.html; European Households, from http://w3.unece.org/pxweb/Dialog/varval.asp?ma=08_GEFHPrivHouse_r&ti=Private+households+by+Household+Type%2C+Measurement%2C+Country+and+Year&path=../DATABASE/Stat/30-GE/02-Families_households/&lang=1; FTTH subscribers Europe, from ECTA; FTTH subscribers North America, from "North American FTTH/FTTP Status," Fiber-to-the-Home Council: North America (2009) at 4. [Description of figure 3]
Sources: Household data, same as Figure 3; FTTH Homes Passed North America, from "North American FTTH/FTTP Status," Fiber-to-the-Home Council: North America (2009) at 2; FTTH Homes Passed Europe 2006 from "FTTH Situation in Europe," Idate Consulting & Research (February 2007 at 20; FTTH Homes Passed Europe 2008 from "FTTH European Panorama," Idate Consulting & Research (December 2008) at 10. [Description of figure 4]
- 70 We show the growth in broadband penetration on a logarithmic scale in this and the next figure to illustrate the rate of growth of broadband over time in these two countries. (back to footnote reference 70)
- 71 Berkman Study at 91. (back to footnote reference 71)
- 72 ECTA Broadband Scorecard, 2009, 1st Quarter. (back to footnote reference 72)
- 73 ECTA Broadband Scorecards, various editions. (back to footnote reference 73)
- 74 ECTA Broadband Scorecard, 2009 1st Quarter, and earlier editions. (back to footnote reference 74)
- 75 Berkman Study at 92. (back to footnote reference 75)
- 76 Berkman Study at 93. (back to footnote reference 76)
- 77 ECTA, Broadband Scorecard, 1st quarter 2009; IDATE, FTTH European Panorama (February 11, 2009) at 10. (back to footnote reference 77)
- 78 It is also important to recognize that neither country has had aggressive competition from cable systems until recently. Cable's share of broadband is about 5 percent in France and 9 percent in Germany according to the most recent ECTA data. (back to footnote reference 78)
- 79 Berkman Study at 95. (back to footnote reference 79)
- 80 ECTA, Broadband Scorecard, 1st quarter 2009. (back to footnote reference 80)
- 81 Company annual reports for 2008. Among the 12 incumbents for which we have data for 2008, only Belgacom has lower capital spending as a share of revenue (13.1% versus 13.2% for FT). By contrast, DT's 2008 CapEx was 14.1% of revenues. (back to footnote reference 81)
- 82 The most recent ECTA data are for the first quarter of 2009. (back to footnote reference 82)
- 83 Crandall, Eisenach and Litan at 21-22. (back to footnote reference 83)
- 84 Berkman Study at 85. (back to footnote reference 84)
- 85 Softbank, Annual Report 2009, p. 41. (back to footnote reference 85)
- 86 The Ministry of Internal Affairs and Communications (MIC) reported in March 2009 that the power companies, including the company bought by KDDI, had nearly 16 percent of fiber connections, NTT had approximately 74 percent, and USEN, the largest cable company, had 4 percent. " See MIC, Disclosure of Quarterly Data Concerning Competition Review in the Telecommunications Business Field," March 25, 2009. (back to footnote reference 86)
- 87 Berkman Study at 83 (emphasis added). (back to footnote reference 87)
- 88 OECD Broadband Statistics, various years. (back to footnote reference 88)
- 89 Fastweb is an Italian company now owned by Swisscom. (back to footnote reference 89)
- 90 Meeting of Robert Crandall with Fastweb, Milan, Italy, September 23, 2009. (back to footnote reference 90)
- 91 Italy ranked 13th among EU-15 countries in household personal computer penetration in 2005. Fewer than 50 percent of Italian households had a personal computer. By contrast, nearly 80 percent of households in Nordic countries have a PC. See OECD, 2007 Communications Outlook at 134. (back to footnote reference 91)
- 92 Berkman Study at 110. (back to footnote reference 92)
- Date modified: