A Future Friendly Digital Economy Strategy: Submission to the Government of Canada's Digital Economy Consultation — Appendix 2

Submission (continued)

V. The Berkman Study's Analysis is Flawed in Other Significant Respects

90. In addition to the errors noted above, the Berkman Study relies on erroneous analyses or makes misleading assertions on a number of fronts. In the sections below, we focus on four of these: (A) its over-reliance on bivariate analysis and "scatter plots" to support its conclusions; (B) its mischaracterization of the evolution of U.S. broadband policy; (C) its mischaracterization of the policy consensus abroad, especially in the European Union.

A. The Berkman Study Relies Almost Exclusively on Bivariate Analysis

91. Bivariate analysis – the comparison of two sets of data to determine the extent to which one variable is correlated with another – can be a useful means of describing the relationships between variables, and as a starting point for more rigorous multivariate (i.e., regression) analysis. It cannot, however, take the place of multivariate analysis, and can almost never support valid findings of causality. Yet, the Berkman Study relies heavily on such analysis to support its conclusions, presenting scatter plot after scatter plot, depicting everything from the relationship between broadband penetration and entrepreneurship (Figure 2.3) to the relationship between average download speeds and median download speeds (Figure 3.191-i). Identifying and critiquing each and every instance in which the Berkman Study relies inappropriately (explicitly or implicitly) on such comparisons to support its conclusions would require far more space than we can justify here.

92. We do, however, point out one particularly egregious (and significant) example of how bivariate analysis can be used to present misleading information: Figure 4.2 of the Berkman Study (reproduced below) presents a scatter plot which shows the results of one of the two quantitative analyses contained in the study, a survey of broadband price and speed offerings by 59 companies in OECD countries. The Berkman Study relies heavily on this analysis for its conclusions, stating that the figure

shows that prices and speeds at the highest tiers of service follow a clear pattern. The highest prices for the lowest speeds are overwhelmingly offered by firms in the United States and Canada, all of which inhabit markets structured around "inter-modal" competition – that is, competition between one incumbent owning a telephone system, and one incumbent owning a cable system.Footnote 93

Moreover, subsequent to the study's release, author Yochai Benkler (in the course of responding to various criticisms of the study) posted a blog entry in which he stated that Figure 4.2 shows results that "are consistent with the open access hypothesis, and inconsistent with the inter-modal competition hypothesis," and that it supports "the basic storyline," that "Where a regulator rolled up its sleeves and really implemented open access, new and innovative entrants used the opportunity to invest in new service models and new electronics, introduced bundled voice over IP, or IPTV, or nomadic access."Footnote 94

Figure 5 : Berkman Study Scatter Plot of Broadband Prices and Speeds by Different Providers
Berkman Study Scatter Plot of Broadband Prices and Speeds by Different Providers

Source: Berkman Center analysis of OECD Broadband Statistics and Global Comms 3.0
Note: Includes highest speed offerings from US players with minimum 2 million subscribers [Description of figure 5]

93. While the Berkman Study's Figure 4.2 may appear visually to support the Berkman Study's "basic story line," the figure hardly constitutes "analysis" and provides virtually no basis for any conclusions about the relationship between unbundling and investment. For example:

  1. Most obviously, the figure fails to account for other variables that likely affect the cost (and hence price) of broadband services, such as loop lengths, urbanicity, dwelling type (multi-dwelling units vs. single-family homes), and topology;
  2. By the same token, the figure fails to account for policy variables other than unbundling (e.g., government subsidies) which likely affect both costs and prices;
  3. The data presented shows only each carrier's "highest speed offering," which may or may not be utilized by a significant number of consumers;Footnote 95
  4. The figure compares data points for firms (such as Verizon and Qwest) that serve large, low-density areas with firms (such as Fastweb) that serve only dense urban areas.

94. Thus, even leaving aside issues of measurement (e.g, whether advertised speeds faithfully reflect delivered speeds), the "analysis" in the Berkman Study's Figure 4.2 is in fact nothing more than a description of the relationship between two sets of variables, from which no conclusions about causality (let alone policy) can appropriately be drawn.

B. The Berkman Study Mischaracterizes the Evolution of U.S. Broadband Policy

95. The Berkman Study is being offered at a time when the Commission and the National Telecommunications and Information Administration (NTIA) are developing a broadband plan. At the same time, the FCC is considering new rules governing "network neutrality" on broadband networks. We are not aware, however, that the FCC is planning to revisit the rules that it has established to implement the network access requirements in the 1996 Telecommunications Act. Yet, the Berkman Study is openly critical of the evolution of these rules, seeing them as a reflection of incumbent "resistance" and "skeptical" courts.Footnote 96 In the course of its criticism, the Berkman Study mischaracterizes the evolution of U.S. policy towards unbundling.

96. At one point, the United States had the most ambitious and aggressive unbundling policy among the OECD countries. In implementing the 1996 Act, the FCC decided to require incumbents to unbundle every network element, not just the "local loop" that connects the subscriber to the carrier's wire center. Indeed, the incumbents were required to offer all of these elements as an "unbundled network element platform" or UNE-P at regulated rates based on forward-looking costs. No other country required such a wholesale offering. In addition, the FCC required the incumbents to share their subscriber loops with entrants seeking to offer only broadband services through the higher frequencies on the loop. The incumbents would be left with the lower frequencies to offer traditional telephone service.

97. This aggressive unbundling policy was contested in the federal courts as inconsistent with the requirements of the 1996 Act. The DC Circuit Court of Appeals and the Supreme Court forced the FCC to revise its unbundling rules in several ways. First, the Supreme Court overturned the FCC's approach to unbundling as overly broad because the Commission had failed to consider whether an entrant could self-provision the various unbundled elements or acquire them from a third party and because the Commission had considered any increase in costs a source of "impairment" that, in turn, required access to unbundled elements at regulated prices.Footnote 97

98. Second, the DC Circuit reversed the FCC's rules on line sharing in United States Telecommunications Association v. FCC, 290 F.3d 415 (2002) because the Commission had failed to evaluate the effects of inter-platform competition. The FCC complied in 2003 by dropping the line-sharing requirement. However, entrants would still be able to lease the entire local loop –a requirement that remains in place today.

99. Third, the DC Circuit was also skeptical of the FCC's ruling that virtually every network element, including mass-market switching, had to be offered by incumbents to competitors at regulated rates. The Court instructed the Commission to examine whether the unbundling of switching was required in every geographic market given the investment in switching by many new competitors in many urban markets. When the FCC failed to drop the requirement for unbundling switching services in its revised rules in 2003, the DC Circuit once again instructed the FCC to reconsider in a strongly-worded opinion. (United States Telecommunications Association v. FCC, 259 F3d. 572 (2003)) As a result, in 2004 the FCC was forced to eliminate unbundling of mass-market switching and, therefore, the UNE-P. However, competitors could still lease, among other network elements, the copper loop at regulated rates if they wished to offer broadband and any other telecommunications services.

100. Thus, the Berkman Study is incorrect when it asserts that:

By the fall of 2001 a new FCC had changed course. Between that fall and the spring of 2002, the FCC passed a series of decisions that abandoned the effort to implement open access, and shifted the focus of American policy from the idea of regulated competition within each wire – competition over the copper plant of the telephone company and over the coaxial cable of the cable company – to competition between the owners of the two wires.Footnote 98

To the contrary, neither the FCC nor state regulators in the U.S. "abandoned" local loop unbundling in 2002 or at any time since then. Competitors may still lease the local loop at regulated rates. However, they do not have the right to share the loop at even lower regulated rates.

101. It is true, however, that as a result of these changes in U.S. policy, competitors' use of incumbents' unbundled network elements began to decline. A large share of this decline was due to the ending of the UNE-P caused by the FCC's decision to drop network switching from the list of required mandatory unbundled elements. These lines were not being used to offer broadband services, but rather to compete in the provision of traditional telephone services. Many entrants – including Covad, NorthPoint, Rhythms, and Net 2000 – had tried to offer DSL services over unbundled incumbent loops and even shared incumbent loops and had failed. Indeed, given the aggressive competition between cable companies and incumbent telephone companies in offering broadband, and despite the best efforts of state and federal regulators to create a favorable regulatory environment and low rates for unbundled elements, no U.S. broadband entrant that has relied on unbundled loops has been able to avoid bankruptcy.Footnote 99

102. The Berkman Study also presents an incomplete (and, as a result, misleading) discussion of the U.S. decision not to impose open access policies on cable broadband providers, implying that the FCC failed to actively consider imposing such rules on cable operators. As the Berkman Study tells it,

…around 1999-2000, as AT&T purchased major cable systems, a new question emerged – whether cable should be subject to the same kind of open access regulation. In several instances cable franchising authorities tried to do this; but the power to impose open access on cable operators was seen as residing in the FCC, not local authorities.Footnote 100

103. In fact, the FCC actively considered imposing open access as a merger condition at the time of the AT&T-TCI merger. After considering the matter, its Chairman at the time, William Kennard, explained the grounds for his decision to forego such a vast expansion of open access regulation onto the cable pipe. As he said in a September 1999 speech,

It is easy to say that government should write a regulation, to say that as a broad statement of principle that a cable operator shall not discriminate against unaffiliated Internet service providers on the cable platform. It is quite another thing to write that rule, to make it real and then to enforce it…. So, if we have the hope of facilitating a market-based solution here, we should do it, because the alternative is to go to the telephone world, a world that we are trying to deregulate and just pick up this whole morass of regulation and dump it wholesale on the cable pipe. That is not good for America.Footnote 101

C. The Berkman Study Mischaracterizes the Policy Consensus Abroad

104. The Berkman Study also goes out of its way to characterize foreign views of unbundling, seeking to present a picture of virtual unanimity among foreign regulators about unbundling's virtues, with the U.S. (along with a few others) as "recalcitrant" outliers.

105. While it is undeniably true that unbundling has played a larger role in some other countries than in the U.S., the reason for this result is not – as the Berkman Study suggests – because their regulators agree that intramodal competition (supported by unbundling) is generally superior to intermodal competition. To the contrary, there is a wide and deep consensus that conditions among and within countries vary widely, and that intermodal competition is superior wherever it is possible.

106. Consider, for example, the June 2009 report of the European Regulators Group (ERG),Footnote 102 which is cited numerous times by the Berkman Study for the proposition that foreign regulators support the unbundling of broadband networks. In fact, the ERG Report emphasizes that economic conditions vary from country to country, that the "economic viability of roll-out strategies is largely influenced by specific local characteristics,"Footnote 103 and that "[o]verall, the country cases support the result … that there are significant differences between and within countries, which reflect differences in the economics of NGA networks, resulting from e.g. different densities or loop length but also from the relevance of competitive pressure from cable networks."Footnote 104

107. For example, the ERG Report finds with respect to Korea and Japan that "Due to the high population density and greater proportion of aerial cables… it is relatively cheap to deploy fibre in these countries."Footnote 105 The ERG also notes approvingly that "Outside of Europe, AT&T is rolling out a FttCab network and Verizon offers FttH in certain US cities. This is in part because of the poor quality of existing DSL services (due to long copper lines) and because of strong competition from cable operators."Footnote 106

108. While the ERG Report acknowledges that various factors affect broadband penetration and investment, and that these factors vary across countries, it is unambiguous in its support for infrastructure competition, finding that "Competition, in particular infrastructure competition, promotes investment as can be seen in countries with high cable penetration. Regulation should set the framework in such a way that it promotes competition."Footnote 107 Thus, "Where it is practically and economically feasible to promote infrastructure based competition, this should be the aim of national regulatory authorities (NRAs)."Footnote 108

109. The Berkman Study does not stop, however, at mischaracterizing foreign regulators' opinions: It also mischaracterizes their actions. For example, at page 76, the Berkman Study lists as a core finding that "Access rules are now being applied to the next generation transition, particularly to fiber," and cites the ERG Report in support of this conclusion. In fact, however, the ERG Report states plainly that among European countries, "Unbundled fibre access is only available on regulated terms in the Netherlands."Footnote 109

Moreover, as noted above, the most recent ECTA and IDATE data show that the incumbent in the Netherlands (KPN) currently has no residential fiber loops to unbundle!


  1. 93 Berkman Study at 12. (back to footnote reference 93)
  2. 94 See http://cyber.law.harvard.edu/node/5751. (back to footnote reference 94)
  3. 95 The Berkman Study acknowledges that the U.S. has among the lowest "entry level" prices, and that this fact should lead to higher levels of penetration. It never reconciles this fact with its contention that penetration in the U.S. is low as a result of the absence of unbundling, which could only increase penetration to the extent it reduces prices. (back to footnote reference 95)
  4. 96 Berkman Study at 78. The Berkman Study's highly opinionated characterization of U.S. policy during the first decade of the 21st century is consistent with the previously expressed views of its principal investigator, Dr. Yochai Benkler, who (for example) has previously characterized the FCC's decision to classify cable modem service as an "information service" (a decision upheld by the Supreme Court in the Brand X case) as a "legally admissible policy error." See, Yochai Benkler, The Wealth of Networks (2006) at 400 (available at www.benkler.org/Benkler_Wealth_Of_Networks.pdf). (back to footnote reference 96)
  5. 97 AT&T Corp. v. Iowa Utilities Board, 525 U.S. 366 (1999). ("The FCC cannot, consistent with the statute, blind itself to the availability of elements outside the incumbent's network. In addition, the FCC's assumption that any increase in cost (or decrease in quality) imposed by denial of a network element renders access to that element 'necessary,' and causes the failure to provide that element to 'impair' the entrant's ability to furnish its desired services, is simply not in accord with the ordinary and fair meaning of those terms.") (back to footnote reference 97)
  6. 98 Berkman Study at 82. (See also at 77: "Some form of open access regulation has at this point been adopted by every country in the OECD except the United States, Mexico, and the Slovak Republic (which has been in the process of passing unbundling requirements for over two years, but has not yet done so)." The Berkman Study then goes on to note that broadband penetration in both Mexico and the Slovak Republic is relatively low, apparently implying that there is a connection between low broadband penetration and the absence of unbundling in these countries. Yet, the Berkman Study provides no evidence whatsoever for this conclusion; to the contrary, its regression analysis of the determinants of broadband penetration suggests that penetration in these countries is explained by their relatively low incomes and population densities.) (back to footnote reference 98)
  7. 99 For a discussion of the causes of the CLECs' demise, see Larry F. Darby, Jeffrey A. Eisenach and Joseph S. Kraemer, "The CLEC Experiment: Anatomy of a Meltdown," Progress on Point 9.23, The Progress & Freedom Foundation (September 2002); see also Jeffrey A. Eisenach and Janusz R. Mrozek, "Do UNE Rates Reflect Underlying Costs?" CapAnalysis LLC (December 2003). (back to footnote reference 99)
  8. 100 Berkman Study at 82. (back to footnote reference 100)
  9. 101 William E. Kennard, "Consumer Choice Through Competition: Remarks Before the National Association of Telecommunications Officers and Advisors, 19th Annual Conference," (September 17, 1999) (available at http://www.fcc.gov/Speeches/Kennard/spwek931.html). (back to footnote reference 101)
  10. 102 European Regulators Group, Report on Next Generation Access - Economic Analysis and Regulatory Principles (June 2009) (hereafter ERG Report). (back to footnote reference 102)
  11. 103 ERG Report at 4. (back to footnote reference 103)
  12. 104 ERG Report at 5. (back to footnote reference 104)
  13. 105 ERG Report at 7. (back to footnote reference 105)
  14. 106 ERG Report at 6. (back to footnote reference 106)
  15. 107 ERG Report at 16. (back to footnote reference 107)
  16. 108 ERG Report at 1. (back to footnote reference 108)
  17. 109 ERG Report at 13. (back to footnote reference 109)