Free Notice (Other) - District Court of Colorado - Colorado

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 1 of 26

THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF COLORADO Civil Action No. 02-cv-01977-RPM (Consolidated with Civil Action No. 02-cv-01978-RPM for pretrial purposes) SPA UNIVERSAIRE and VACATION TAN & TRAVEL, Plaintiffs, v. QWEST COMMUNICATIONS INTERNATIONAL, INC. and QWEST CORPORATION, Defendants. REBUTTAL REPORT OF JOHN. B. HAYES

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 2 of 26

REBUTTAL REPORT OF DR. JOHN B. HAYES

1. Affidavit of John B. Hayes 2. -2-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 3 of 26

I. II.
A.

Scope and Organization of Rebuttal Report .............................................. 1 Points of Agreement...................................................................................... 1
Reduced Interconnection Costs Tend to Increase Competition and Reduce Prices Paid for Local Exchange Service...................................................................................1 B. CLECs Would "Opt In" to Lower Interconnection Prices.............................................2

III.
A. B. C. D.

Misunderstandings, Mischaracterizations, and Errors ............................ 4
Economic Model of Competitive Effects ......................................................................4 Econometric Measurement of Damages ........................................................................7 Availability of Data for Econometric Measurement of Damages ...............................12 Benchmark Measurement of Damages ........................................................................14

IV.
A. B.

Disputes Regarding Proper Economic Analysis ...................................... 15
"Too Much" Variation .................................................................................................15 The But-For World and Measurement of Damages.....................................................21

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 4 of 26

I.

Scope and Organization of Rebuttal Report

1. My name is John B. Hayes. I am the author of Expert Report of John B. Hayes filed March 9, 2006 in this matter on behalf of counsel for the plaintiffs. In this supplemental report I respond to the views expressed in the Expert Report of Dr. Gary J. Dorman filed on April 10, 2006 on behalf of defendant Qwest and to the opinions expressed by Dr. Dorman in his deposition of April 25, 2006. 2. My rebuttal report is organized as follows. First, I identify important areas of agreement between myself and Qwest's economic expert, Dr. Dorman. Second, I explain how a number of the remaining differences between Dr. Dorman and myself flow from errors in his testimony and misunderstandings and mischaracterizations of my testimony. I then focus on what appear to be genuine differences of opinion regarding the economic analysis of class-wide impact and damages.

II.

Points of Agreement A. Reduced Interconnection Costs Tend to Increase Competition and Reduce Prices Paid for Local Exchange Service

3. Dr. Dorman agrees with my conclusion that reducing the fee(s) charged by Qwest for interconnection services would cause lower prices to be paid for local exchange service by some members of the proposed class.1 Specifically, Dr. Dorman has not disputed any of my conclusions regarding the effects of reduced interconnection fees on (1) entry by CLECs, (2) the variety of services offered to class members by local exchange carriers, or (3) prices of local exchange services sold by Qwest and CLECs. Rather, Dr. Dorman's concerns relate to (1) whether the disfavored CLECs would have "opted in" to the secret interconnection agreements if those agreements had been available, and (2) my proposed methodologies for measuring the effects of reduced interconnection fees on local telephone competition. I address the first issue in the next section of this report, and the remainder of this report addresses the second issue. 4. The fact that reducing costs will cause firms to enter markets (or expand output) and reduce prices is a staple of economic analysis.2 An assessment of CLEC business strategies summed the issue up nicely, stating: "[t]he CLEC must know that its entry into the industry will place downward pressure on retail prices because entry increases

1

As in my prior report (¶ 19), I refer to all of the various services and equipment referenced in interconnection agreements that are subject to the filing requirements of the 1996 Act as interconnection services and to the cost of these interconnection services and equipment as interconnection costs. See, for example, Robert S. Pindyck and Daniel L. Rubinfeld, Microeconomics, Sixth Edition (Pearson Prentice Hall, 2005), pp. 274 and 283-284.

2

Rebuttal Report of Dr. John B. Hayes -1-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 5 of 26

competition."3 (Emphasis added.) State public utilities commissions ("PUCs") investigating Qwest's unfiled agreements reached similar conclusions in assessing the impact of higher interconnection costs to disfavored CLECs. For example, the Minnesota PUC concluded that Qwest's conduct "generally harms competition and the growth of CLECs in Minnesota."4 The staff of the Colorado PUC concluded that "[i]n order to introduce more firms into the market, it is necessary to take steps to ease entry, including reducing entry barriers and entry costs. ... With fewer companies in the marketplace, consumer choice is artificially limited, and price and other pressures on both ILECs and new entrants are reduced."5 5. Numerous empirical studies have also found evidence that reduced interconnection fees promotes entry into the provision of local exchange service and that additional entry lowers prices for that service.6

B. CLECs Would "Opt In" to Lower Interconnection Prices
6. Dr. Dorman agrees that CLECs, given the opportunity to purchase interconnection on more attractive terms, will accept the offer, all else equal.7 The reason is straightforward: accepting an offer of reduced interconnection fees unambiguously increases a CLEC's profits.

3

Crandall, Robert W., "An Assessment of the Competitive Local Exchange Carriers Five Years After the Passage of the Telecommunications Act," January 2002, p. 23. Minnesota Public Utilities Commission, Findings of Fact, Conclusions and Memorandum, 6-2500-14782-2, September 20, 2002, ¶ 375. Initial Comments of Staff of the Colorado Public Utilities Commission, Docket No. 02I-572T, February 27, 2004, pp. 49-50. Braunstein, Yale M., "The Role of UNE-P in Vertically Integrated Telephone Networks: Ensuring Healthy and Competitive Local, Long-Distance and DSL Markets," May 2003, available at http://www.sims.berkeley.edu/~bigyale/UNE/UCB_Study_UNE_May_2003.pdf; Braunstein, Yale M., "UNE-P Benefits in Verizon's New Jersey Territory," March 2004, available at http://www.sims.berkeley.edu/~bigyale/UNE/UCB_NJ_UNE_study_Mar_2004.pdf; Greenstein, S., and Mazzeo, M., "Differentiation Strategy and Market Deregulation: Local Telecommunication Entry in the Late 1990s," NBER Working Paper 9761, June 2003, available at http://www.nber.org/papers/W9761; Roycroft, Trevor R., "Empirical Analysis of Entry in the Local Exchange Market: The Case of Pacific Bell," Contemporary Economic Policy, Vol. 23, No. 1, January 2005, pp. 107-115; Economides, Nicholas, Seim, Katja, and Viard, V. Brian, Quantifying the Benefits of Entry into Local Phone Service, NYU Center for Law and Economics, Working Paper No. 05-23, October 2005, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=825984#PaperDownload; TRAC, "Projected Residential Savings in California's Telephone Market," November 2001; Lehman, Dale D., "(How) do Regulated Prices Affect Competitive Entry?," Info: the Journal of Policy, Regulation and Strategy for Telecommunications, Information and Media, Vol. 5, No. 4, 2003, pp. 20-27; Michigan Alliance for Competitive Telecommunication, "Consumer Savings from Local Phone Competition in Michigan," May 2003, available at http://www.miact.org/pdf/Mich_ConsumerSavings.pdf. Deposition of Gary Dorman, April 25, 2006 (hereafter "Dorman Deposition"), pp. 144-145.

4

5

6

7

Rebuttal Report of Dr. John B. Hayes -2-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 6 of 26

7. In making this point, I have not ignored the concern that some disfavored CLECs may not have found the specific interconnection terms offered to McLeod or Eschelon more attractive than the terms they were already receiving.8 Thus I agree that one economic issue in this litigation is the extent to which the disfavored CLECs would have found the terms available in the secret agreements more attractive than those in their existing agreements.9 I noted this issue in my prior report (¶¶ 16, 41). It is also important to recognize, however, that the terms of the McLeod and Eschelon agreements very likely were better than those available to other CLECs. 8. My initial report provided the Minnesota Public Utilities Commission's conclusion on this point, which bears repeating:
"In each of the twelve interconnection agreements cited by the Department, Qwest provided terms, conditions, or rates to certain CLECs that were better than the terms, rates and conditions that it made available to the other CLECs and, in fact, it kept those better terms, conditions, and rates a secret from the other CLECs. In so doing, Qwest unquestionably treated those select CLECs better than the other CLECs."10

Thus at least one authority has already addressed this issue and found that the terms in the secret agreements were, in fact, more attractive than those available to the disfavored CLECs. 9. Dr. Dorman specifically questions whether some disfavored CLECs would have found the quantities purchased under the McLeod and Eschelon agreements prohibitively large and, therefore, those CLECs would not have "opted in" to the more attractive pricing in those agreements. Implicit in Dr. Dorman's question is a particular interpretation of the "pickand-choose" rules. Specifically, he appears to understand those rules to require that CLECs wanting to opt in to the pricing provisions of the secret interconnection agreements also accept the volume terms.11 Dr. Dorman (¶ 18) further suggests that CLEC decisions to "opt in" to the secret interconnection agreements will vary for each geographic location within which each CLEC operates. 10. While I do not know whether Dr. Dorman's interpretation of the "pick-and-choose" rules is correct, I note that the volume commitments in the McLeod and Eschelon agreements applied to all purchases by those CLECs. Thus any CLEC that "opted in" to either of these agreements would receive the discounts region-wide. Therefore, there would not be

8

Expert Report of Dr. Gary J. Dorman, April 7, 2006 (hereafter "Dorman Report"), ¶¶ 6-7. Dorman Deposition, pp. 144 and 171. Dorman Deposition, pp. 170-171. Before the Minnesota Public Utilities Commission, In the Matter of the Complaint of the Minnesota Department of Commerce Against Qwest Corporation Regarding Unfiled Agreements, "Order Assessing Penalties," Docket No. P-421/C-02-197, February 28, 2003, pp. 3-4. Dorman Deposition, pp. 99 and 171.

9 10

11

Rebuttal Report of Dr. John B. Hayes -3-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 7 of 26

any geographic variation in the discount received by any CLEC that "opted in" to one of these agreements.

III.

Misunderstandings, Mischaracterizations, and Errors A. Economic Model of Competitive Effects

11. Dr. Dorman (¶ 15) criticizes the economic model of competition in a local telecommunications market that I presented in Appendix D to my initial report, saying it utilizes a homogenous product while the actual product--local telephone service--is differentiated.12 The purpose of an economic model is to capture the essential factors that influence the problem at hand--here, whether extending a discount to disfavored competitors would affect market prices and quantities--and to focus attention on how changes in those key factors affect outcomes. As virtually all products are differentiated to some degree, economists necessarily make informed judgments about whether the differentiation matters to the question at hand. In this case, the use of a model with homogenous products is inconsequential to an analysis of whether extending the terms available in the secret agreements to the disfavored CLECs would reduce the prices paid for local telephone service. I note that Dr. Dorman does not claim that any of the key results I derived in Appendix D would change if I had used a model with differentiated products instead of one with homogenous products, and I know of no reason to believe that they would.13 12. Moreover, if subsequent analysis of the economic issues in this litigation indicates that product differentiation is an important factor in explaining how extending a discount to disfavored competitors would affect market prices and quantities, the dominant firmcompetitive fringe ("DF/CF") model that I used in Appendix D can be extended to accommodate differentiated products.14 Indeed, there are several published papers that utilize a differentiated products version of the DF/CF to study competition in

12

Dr. Dorman agrees that the econometric method I proposed for measuring class-wide damages can account for differentiated products. (Dorman Deposition, pp. 154-155.) Thus his concern regarding this issue is apparently limited to the economic model I presented in Appendix D. The "homogenous product" model I used in Appendix D has been used to study entry into local telephone markets. See, for example, Abel, J., "Entry into Regulated Monopoly Markets: The Development of a Competitive Fringe in the Local Telephone Industry," Journal of Law and Economics, Vol. XLV (October 2002). "[T]he pure DF/CF model that assumes perfect product homogeneity may be modified to account for any tangible sources of product differentiation that may be part of the industry in question." (Kahai, Simran, Kaserman, David L., and Mayo, John W., "Is the `Dominant Firm' Dominant? An Empirical Analysis of AT&T's Market Power," Journal of Law and Economics, Vol. XXXIX (October 1996), pp. 499-517 at 504.)

13

14

Rebuttal Report of Dr. John B. Hayes -4-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 8 of 26

telecommunications and other markets.15 Dr. Dorman appears to not recognize these facts, as he expresses doubt that the DF/CF model could be extended to account for differentiated products, saying it would not make sense based on "pure economics."16 He also fails to acknowledge the use of the differentiated DF/CF model to study competition in telecommunications markets.17 13. Dr. Dorman (¶ 16) further criticizes my use of a DF/CF model because he apparently believes Qwest is not "dominant" in (at least some of) the local telephone markets included in the proposed class. Among the key assumptions of the DF/CF model are that (1) there is one firm that holds a relatively large share of the market (the dominant firm), and (2) there are a number of smaller competitors that take the dominant firm's price as given when making their supply decisions (the competitive fringe).18 The size of the "dominant" firm relative to its smaller, "fringe" competitors is not crucial: The key point is that the fringe competitors are price takers and the dominant firm is not. Kahai, et al., for example, used the DF/CF model to study competition in long-distance telecommunications markets where the share of the dominant firm (AT&T) ranged from 84% to 59%, with an average share of 72%.19 In addition, a widely used textbook suggests that a 40% market share held by the dominant firm is sufficient for use of the DF/CF model.20 Qwest's market share easily exceeds this threshold levels in the vast majority, if not all, of the local telephone service markets at issue in this litigation. 14. Exhibit 1 to my initial report showed that Qwest provided local exchange service to the vast majority of the wireline customers in its service areas during the class period. Qwest's share exceeds 70% in each of the 8 states included in the table. These data report Qwest's share of all local telephone lines. Qwest's share of lines to customers served by a twisted pair line is likely even higher than reported in the table because CLEC lines are

15

Blank, Larry, Kaserman, David L., and Mayo, John W., "Dominant Firm Pricing with Competitive Entry and Regulation: The Case of IntraLATA Toll," Journal of Regulatory Economics, Vol. 14 (1998), pp. 35-53; Kahai, Simran, Kaserman, David L., and Mayo, John W., "Is the `Dominant Firm' Dominant? An Empirical Analysis of AT&T's Market Power," Journal of Law and Economics, Vol. XXXIX (October 1996), pp. 499517. Another example of how the basic DF/CF model has been used to account for differentiated products is Suslow, Valerie Y., "Estimating Monopoly Behavior with Competitive Recycling: An Application to Alcoa," Rand Journal of Economics, Vol. 17, No. 3 (Autumn 1986), pp. 389-403. Dorman Deposition, pp. 155-158. Dorman Deposition, p. 162. See, for example, Kahai, Simran, Kaserman, David L., and Mayo, John W., "Is the `Dominant Firm' Dominant? An Empirical Analysis of AT&T's Market Power," Journal of Law and Economics, Vol. XXXIX (October 1996), pp. 499-517. Kahai, Simran, Kaserman, David L., and Mayo, John W., "Is the `Dominant Firm' Dominant? An Empirical Analysis of AT&T's Market Power," Journal of Law and Economics, Vol. XXXIX (October 1996), pp. 499517 at 502. Scherer, F. M. and Ross, D., Industrial Market Structure and Economic Performance, Third Edition (Houghton Mifflin, 1990), p. 221.

16 17 18

19

20

Rebuttal Report of Dr. John B. Hayes -5-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 9 of 26

disproportionately concentrated among business customers.21 Larger business customers, which are not within the proposed class, are often served by DS1 or larger facilities.22 15. Dr. Dorman (¶ 17) states that there are cities in the proposed class where Qwest is not dominant. He specifically identifies Albuquerque, New Mexico and Council Bluffs, Iowa as examples of cities where Qwest is not dominant. The document that Dr. Dorman cites to support his conclusion regarding Albuquerque does not report Qwest's share in that city, and it does not indicate that Qwest was not dominant. Instead, that document states only that at least one competitor has more than a de minimus share of local exchange service in Albuquerque.23 16. Exhibit R1 to this report contains data on Qwest's share of local telephone lines for 8 cities where Qwest is the incumbent local exchange carrier for the period 4Q2002 ­ 4Q2004. These data show that Qwest's share exceeds 70% in 7 of the 8 cities throughout the reference period. Council Bluffs stands out as an exception: Qwest's share in Council Bluffs is well below its share in the other cities.24 17. On balance, the available evidence shows that Qwest had a very large share of local service lines in most locations within the Qwest service territory during the class period. It follows that the DF/CF model is an appropriate tool for analysis of the competitive effects of extending the secret discounts to the disfavored competitors. 18. In addition, Dr. Dorman (¶ 13) claims my DF/CF model does not take into account CLEC costs other than interconnection costs, which means, according to Dr. Dorman, that the model cannot be used to estimate the magnitude of the competitive effects of extending the secret discounts to the disfavored competitors. Dr. Dorman admitted in deposition that he had not studied the DF/CF model in Appendix D.25 He also has not shown that my analytic results depend in any way on this simplification. As I stated in my initial report (¶ 25), "the analytic results do not depend on this assumption. In particular, my conclusions

21

For example, in its "Local Competition Report" as of December 2002, the FCC reported that 58% of switched access lines in service to CLEC customers served residential and small business customers. The comparable figure for ILECs was 78%. (Federal Communications Commission, "Local Telephone Competition: Status as of December 31, 2002," June 2003, p. 1, Table 2, and Table 11. See also, Crandall, Robert W., "An Assessment of the Competitive Local Exchange Carriers Five Years After the Passage of the Telecommunications Act," January 2002, p. 19.) A DS1 line contains 24 separate channels, each of which is the size of a standard twisted pair line. Dr. Dorman cites an FCC document that states: "[We] find that Cricket Communications, a PCS provider, serves more than a de minimis number of residential users over its own facilities and, for purposes of section 271 compliance, represents an actual commercial alternative to Qwest for residential telephone exchange services." (Federal Communications Commission, Memorandum Opinion and Order, FCC 03-81/WC Docket No. 03-11, April 15, 2003, p. 9.) The data for Exhibit R1 were produced by Qwest during discovery. Albuquerque was not included in these data. Dorman Deposition, pp. 127-131.

22 23

24

25

Rebuttal Report of Dr. John B. Hayes -6-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 10 of 26

regarding the directional effects on competition arising from the secret agreements would not be altered if CLECs had additional costs." In fact, the DF/CF model in Appendix D is easily altered to account for costs other than interconnection costs, and it can be used to estimate the magnitude of the competitive effects at issue in this litigation. 19. Moreover, the purpose of my economic model is not to measure the magnitude of classwide damages. The econometric and benchmark methods discussed in my initial report serve that purpose. The economic model establishes the competitive impact on the market arising from Qwest's alleged conduct. Specifically, the model shows that local telephone service prices would have been lower if additional CLECs had received the favorable interconnection terms at issue in this case. 20. Dr. Dorman (¶ 19) claims I assume that CLEC savings would be passed through to class members. Passthrough of the reduction in interconnection fees to class members is not an assumption: I explicitly derived this result through the calculations shown in Appendix D. Passthrough of reduced interconnection fees to final customers was also shown empirically in Braunstein's analyses of the data for California and New Jersey.26 21. Dr. Dorman (¶ 19) further asserts incorrectly that I provide no model or method to measure passthrough.27 The econometric method employed by Economides, et al. provides one framework to measure passthrough.28 Passthrough can also be measured using the method employed by Professor Braunstein. Appendix D shows how one can estimate passthrough of a reduction in interconnection fees to a group of disfavored CLECs given an estimate of passthrough by all CLECs.

B. Econometric Measurement of Damages
22. The essence of Dr. Dorman's critique of my reliance on the Economides paper is that the price, quantity, and variety effects reported in that study are measured too imprecisely to

26

I discussed both of these studies in my initial report. Braunstein, Yale M., "The Role of UNE-P in Vertically Integrated Telephone Networks: Ensuring Healthy and Competitive Local, Long-Distance and DSL Markets," May 2003, available at http://www.sims.berkeley.edu/~bigyale/UNE/UCB_Study_UNE_May_2003.pdf; Braunstein, Yale M.," UNE-P Benefits Update: SBC's California Territory, 2004," May 2004, available at http://www.sims.berkeley.edu/~bigyale/UNE/CA_UNE_competition_update_2004.pdf; Braunstein, Yale M., "UNE-P Benefits in Verizon's New Jersey Territory," March 2004, available at http://www.sims.berkeley.edu/~bigyale/UNE/UCB_NJ_UNE_study_Mar_2004.pdf. I understand that the passthrough analysis is only relevant to the claims of CLEC customers under Communications Act § 202, since I understand that the antitrust claims in this litigation are brought only by Qwest customers. See Economides, Nicholas, Seim, Katja, and Viard, V. Brian, Quantifying the Benefits of Entry into Local Phone Service, NYU Center for Law and Economics, Working Paper No. 05-23, October 2005, available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=825984#PaperDownload.

27

28

Rebuttal Report of Dr. John B. Hayes -7-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 11 of 26

rely upon for the purpose of calculating damages.29 Dr. Dorman then says that if classwide damages in this action were measured using the method described in Economides, they also would be unreliable.30 In this section I discuss at length how Dr. Dorman's conclusion that the Economides study inaccurately measured welfare gains is unfounded and rests on a serious error. That error also undermines his assertion that class-wide damages cannot reliably be measured using the method described by Economides. A correct analysis of the Economides results shows that welfare gains were measured with considerable precision. It is reasonable to anticipate that class-wide damages calculated using this method could also be reliably estimated. 23. Dr. Dorman calculates what he calls a "statistical confidence interval" around the average welfare gain reported by Economides in Table 8:
"[The] estimated welfare gain per customer per month was $3.86 with a confidence interval of plus or minus $10, meaning that the true value of the estimated welfare gain for the average customer ranged from $13.86 to a loss of roughly $6 per month. [...] The second paper [...] came up with an estimated welfare gain of 88 cents per customer per month with a plus or minus $7.92."31

Dr. Dorman (¶ 24) then says these "extremely wide confidence intervals raise serious doubts about whether the Economides methodology can be used to develop a reliable measure of alleged classwide damages." 24. To construct his "confidence interval," Dr. Dorman erroneously used the standard deviation of the sample welfare gain.32 To correctly construct a confidence interval around the average welfare gain, Dr. Dorman should have used the standard error of the average welfare gain. The "confidence interval" that Dr. Dorman constructs does not provide any useful information about the precision with which the welfare gains reported in the Economides paper were measured by the authors' econometric model. The amount of variation (plus or minus $7.92) simply indicates that different customers in the sample enjoyed a different level of benefits. In fact, the "confidence interval" that Dr. Dorman considered is unrelated to the precision with which the welfare gains were measured, and Dr. Dorman's core criticism of the Economides method is unfounded. 25. Economides estimates individual dollar welfare gains from increased competition for local telephone service in New York State. The authors' sample contains more than 5000 residential telephone customers, and they estimate the welfare gain for each customer in

29 30 31 32

Dorman Report, ¶ 24. Dorman Deposition, pp 64-65, 113-118 and 150-152. Dorman Report, ¶ 24; Dorman Deposition, pp. 113-115. Dorman Deposition, p. 118. See also, Dorman Report, ¶ 24. Dr. Dorman constructed the upper bound for his supposed "statistical confidence interval" by adding two times the standard deviation of the sample welfare gains to the average welfare gain. The lower bound was constructed by subtracting two times the standard deviation of the sample welfare gains from the average welfare gain.

Rebuttal Report of Dr. John B. Hayes -8-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 12 of 26

the sample. In Table 8, they report the simple arithmetic average of the welfare gains in the sample. This average welfare gain is not an estimate of the welfare gain for each customer in the sample. Indeed, it may not equal the estimated welfare gain for any individual customer in the sample. Consider a hypothetical example with a random sample of 5,000 customers, with 2,500 of each of two types of customers. Suppose the estimated welfare gains for those two types of customers were $5 per month and $10 per month respectively. The average welfare gain in this hypothetical sample would then be $7.50, but no customer has an estimated welfare gain equal to the average welfare gain. Further, the variation in my sample around this average welfare gain has nothing to do with the explanatory power of the underlying econometric model that was used to estimate the individual welfare gains of $5 and $10, as I now discuss. 26. Table 8 in Economides reports the standard deviation of the welfare gains in the sample. This standard deviation is simply a measure of the heterogeneity in the population of customers which the authors have successfully modeled. Returning to my simple example, the standard deviation in my sample is $2.50. I could use my sample standard deviation to calculate a "confidence interval" around my sample mean similar to that calculated by Dr. Dorman. In this hypothetical example, the "confidence interval" so constructed would be from $2.50 to $12.50. This "confidence interval" tells me that the probability that the welfare gain associated with any specific observation in the sample falls within $7.50 ± $5.00 is 95%.33 27. I could also compute a confidence interval for the average welfare gain in the population from which my sample was drawn. An approximate confidence interval for the average welfare gain in the population would depend on the standard error of the sample mean, which is equal to the sample standard deviation divided by the square root of the sample size. In my hypothetical sample, the sample standard deviation is $2.50, and the size of the sample is 5,000. The standard error of the sample mean is therefore $0.035. Thus, an approximate 95% confidence interval for the average welfare gain in the population would be from $7.43 to $7.57. This confidence interval tells me that, for the population that my sample was drawn from, there is approximately a 95% probability that the true value of the average welfare gain falls within a range of $7.50 ± $0.07 . 28. This simple example shows that even though there is considerable heterogeneity in the welfare gains, one can estimate the average welfare gain in the population within an error of approximately 1%.34 The actual sample size used by Economides was more than 5,000,

33 34

In fact, 100% of the observations in my hypothetical sample fall within this range. It follows that we can accurately estimate the total welfare gain, since the total welfare gain is simply the average welfare gain multiplied by the total size of the population.

Rebuttal Report of Dr. John B. Hayes -9-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 13 of 26

so the actual confidence interval for their average welfare gain in the population would be less than one-seventieth of the size suggested by Dr. Dorman.35 29. I noted above that the confidence interval around the average welfare gain in the population that I calculated was only approximate. This is so because that calculation did not account for estimation error in the underlying econometric model that was used to estimate the welfare gains. The standard errors of the parameter estimates that were used to calculate the welfare gains provide information about how well the econometric model fits the sample data. For example, if the individual welfare gains of $5 and $10 in my example were themselves measured with standard errors of $0.10, then the confidence intervals for the two types of customers would be from $4.80 to $5.20 and $9.80 to $10.20 respectively. These standard errors reflect how well the econometric model that was used to estimate the welfare gains fits the data. This additional source of uncertainty also increases the confidence interval for the average welfare gain in the population. In this example, the precise confidence interval for the average welfare gain in the population would be $7.50 ± $0.21 (i.e., $7.29 to $7.71).36 30. Economides reports the parameter estimates for his econometric model along with the associated standard errors of these estimates in Table 7. Of the 38 parameter estimates reported, 31 are statistically different from zero at the 1% level of significance. This indicates that the parameters are measured with a high degree of statistical reliability. Furthermore, the demand slope coefficient estimates for local toll usage and regional toll usage each have t-statistics that are greater than 80. These large t-statistics show that two of the key parameters used to compute demand elasticities are estimated with a very high degree of statistical precision. Another way to express this is to note that these parameters are estimated with standard errors that are less than 1.25% of the magnitudes of the parameters. This means that the effects of these parameters, such as demand responses to price changes, are measured with an error (standard deviation) of less than 1.25%. Dr. Dorman (¶¶ 34-36) claims there is no reason to think that reliable estimates of demand elasticities can be achieved. These estimates reported in Economides show otherwise.37 Additionally, the coefficients that vary by customer characteristics are generally measured

35

The square root of 5,000 is 71. As the sample increases, the number by which we would divide the sample standard deviation to obtain the standard error of the sample mean grows larger. For the purposes of this calculation, I have assumed for illustrative purposes that the estimation error is independent of the sampling error and that the estimation error is perfectly correlated across customers. In general, the exact size of the modified confidence interval for the average welfare gain in the population would depend on some additional statistics that could be easily computed in practice. Demand elasticities for local telephone service have been estimated successfully many times. See, for example, Train, Kenneth E., McFadden, Daniel L., and Ben-Akiva, Moshe, "The Demand for Local Telephone Service: A Fully Discrete Model of Residential Calling Patterns and Service Choices," Rand Journal of Economics, Vol. 18, No. 1, Spring 1987; Cain, Paul and MacDonald, James M., "Telephone Pricing Structures: The Effects of Universal Service," Journal of Regulatory Economics, Vol. 3, No. 4 (December 1991), pp. 293-308.

36

37

Rebuttal Report of Dr. John B. Hayes -10-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 14 of 26

with a high degree of statistical significance. This means the model is very successful at estimating how demand varies across customers and measuring how welfare gains from increased local telephone competition vary across customers. 31. The foregoing shows unequivocally that Dr. Dorman is wrong to claim that the Economides study does not yield reliable measures of the welfare gains from increased local competition. In fact, the method described in Economides does provide a reliable basis for assessing the damages suffered by individual class members. 32. Dr. Dorman (¶ 24) further faults the Economides paper as a guide for analysis of classwide damages in this matter because it has not yet been published in a peer-reviewed journal. He makes the same criticism (¶ 28) of the Braunstein studies. However, the method that Economides uses, and that I propose to use, has been published in peerreviewed journals, including journals that Dr. Dorman agrees are "authoritative."38 More fundamentally, it is common and accepted practice in the economics profession to review and reference unpublished working papers. For example, the first three articles published in the March 2006 issue of The American Economic Review, the leading publication of the American Economic Association, referenced more than 20 papers that were unpublished manuscripts or working papers.39 33. Dr, Dorman (¶ 20) states that I have not provided an econometric model that could be used to estimate damages to class members. However, I have discussed at length an existing study--the Economides paper--that sets out in great detail an econometric methodology that could be used to estimate damages in this litigation. The general features of the demand model that I propose to use in this litigation are quite similar to those described in Economides. In particular, I propose to estimate a model where customers (indexed by i) choose from among a set of service plans (indexed by j) offered by local service providers in distinct local telephone markets (indexed by m). Customers choose from among the set of plans offered by service providers in their local market. Service plans consist of a fixed fee ( Pjm ) and per-call pricing for regional toll calling ( pT ) and for local calling ( p L ).40 jm jm
T Given these prices, customers select a plan and quantities of regional toll calling ( qim ) and L local calling ( qim ). All of these items can be observed in the available data, such as the data from TNS Telecoms or from Qwest's business records. Customer preferences determine which plan each customer selects and how many calls it makes. Variation in customer preferences can be measured by observable characteristics of each customer. For residential customers, those characteristics could be such factors as the monthly income of the household and the number of persons living in the household. For business customers,

38

See, for example, Haneman, Michael W., "Discrete/Continuous Models of Consumer Demand," Econometrica, Vol. 52, No. 3 (May 1984), pp. 541-561. Dr. Dorman identified Econometrica as an "authoritative" economic journal. (Dorman Deposition, p. 123.) The American Economic Review, Vol. 96, No. 1 (March 2006), pp. 23-29, 49-50, and 80-81. The per-call prices for local and toll calling my be equal to zero for flat-rate plans.

39 40

Rebuttal Report of Dr. John B. Hayes -11-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 15 of 26

those characteristics could be factors such as the number of employees and the type of business (e.g., insurance, banking, manufacturing). These customer characteristics are also observed in the available data. The dependent variable in the demand model is the service plan selected by the customer along with the quantities of local and toll calls made. The explanatory variables are the prices for local telephone service and household and business demographic characteristics. 34. I also expect to model supply relationships in the econometric model. Local service providers make supply decisions based on local service costs, which consist of both fixed ( Cm ) and variable ( cm ) costs, the intensity of competition in the market ( N T , N L ), jm jm regulations ( RT , R L ), and the strength of demand in each market ( Z m ). The strength of jm jm demand in each market indicates how attractive each local telephone market is to local service providers. Previous studies have accounted for the strength of demand for local telephone service with data on the types of businesses located in the region, household incomes, and other similar information to describe the local telephone market. The dependent variables in the supply relations are the various service prices and intensity of competition in each market. The intensity of competition is itself a function of interconnection fees and downstream service prices. The independent variables are local service costs, regulation, and the strength of demand in each market. 35. The appropriate geographic scope for the local telephone markets depends, in part, on the locations of service providers and the plans offered by those providers. Previous studies have used zip codes or cities to define local markets. The data that are ultimately made available to estimate the model on locations of service providers and plans offered by those providers will inform my analysis going forward, and I will make determinations about the appropriate size of geographic markets as more data and other information become available. One guiding principle in this effort is that within relevant markets all customers face essentially the same set of plans and prices.

C. Availability of Data for Econometric Measurement of Damages
36. Dr. Dorman agrees that the variables identified for econometric analysis in my initial report (¶ 48) are "the kinds of independent variables that would go into the model" of the economic effects of extending the secret discounts to the disfavored CLECs.41 His concern is that some of these variables may be difficult to measure. 37. First, Dr. Dorman (¶ 9) says that I do not have data on customer preferences.42 If he means that one needs literally to have the data in hand to reach conclusions regarding the ability of common methods to measure damages to class members, that position is

41 42

Dorman Deposition, p. 76. See also, pp. 146-148. See also, Dorman Report at ¶ 22 and Dorman Deposition, pp. 84, 86.

Rebuttal Report of Dr. John B. Hayes -12-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 16 of 26

unwarranted. It is perfectly reasonable, common and accepted academic practice to review the economic literature for examples of methods and data used to analyze closely related problems. I mentioned two sources for data on preferences in my initial report and deposition.43 Data from TNS Telecoms, a third party that collects data on local telecommunications services, is widely used to study local telephone competition.44 Economides used data from TNS to estimate his model. The defendant's business records are another source of data reflecting customer preferences. 38. Second, Dr. Dorman (¶¶ 21, 23) claims I have not identified data on the cost to provide local telephone service. There are several sources for these data. Data on interconnection fees are publicly available from state PUCs and other sources, and these data have been used to measure local service costs in several studies of local telephone competition.45 The FCC's Hybrid Cost Proxy Model ("HCPM") is another potential source of cost data that has been used to measure local service costs in several studies of local telephone competition.46 The HCPM was developed as a potential means of establishing Universal Service Fund support levels, cost-based access charges, and prices for interconnection and unbundled network elements.47 HCPM data are available for each wire center in the

43

Deposition of John B. Hayes, March 23, 2006, pp. 36, 57, 92, 95, and 98; Expert Report of John B. Hayes, March 9, 2006, ¶¶ 54-55. Expert Report of John B. Hayes, March 9, 2006, ¶¶ 54-55; Bill Harvesting Data, available from TNS Telecoms, is collected from a quarterly survey of consumer telephone bills. TNS collects surveys from approximately 30,000 households each quarter, of which approximately 9,000 households provide copies of their most recent telephone bills, including local, long-distance, wireless, and Internet access services. Each quarter, TNS draws a balanced random sample of households across U.S. states from a panel of approximately 60,000 households that have previously indicated their willingness to be surveyed. In the fourth quarter of 2002, TNS collected bills from 851 households in Qwest's 14-state service area. In the fourth quarter of 2004, TNS collected bills from 1,042 households in Qwest's 14-state service area. (Qwest, Consumer Market Share Quarterly Summary Report 3Q05, December 2005, Q08334-08434 at Q08338.) I referenced some of these studies in my initial report. See, for example, Roycroft, Trevor R., "Empirical Analysis of Entry in the Local Exchange Market: The Case of Pacific Bell," Contemporary Economic Policy, Vol. 23, No. 1 (January 2005), pp. 107-115; Greenstein, Shane, and Mazzeo, Michael, "Differentiation Strategy and Market Deregulation: Local Telecommunication Entry in the Late 1990s," National Bureau of Economic Research Working Paper 9761, June 2003, available at http://www.nber.org/papers/W10482. An additional example is Declaration of Dale E. Lehman in FCC 03-173, filed on behalf of Qwest Communications International Inc., "Investment and the Level of UNE-P Rates: A Critique of the Willig Study," January 30, 2004. I referenced some of these studies in my initial report. See, for example, Clarke, Richard N., Hassett, Kevin A., Ivanova, Zoya, and Kotlikoff, Laurence, "Assessing the Economic Gains from Telecom Competition," National Bureau of Economic Research Working Paper 10482, May 2004, available at http://www.nber.org/papers/W10482. Some additional examples are Rosston, Gregory L., and Wimmer, Bradley S., "Local Telephone Rate Structures: Before and After the Act," Information Economics and Policy," Vol. 17, No. 1 (January 2005), pp. 13-34; Eisner, James, and Lehman, Dale E., "Regulatory Behavior and Competitive Entry," June 2001, available at http://www.aestudies.com/library/elpaper.pdf. Bush, C.A., Kennet, D. M., Prisbrey, J., and Sharkey, W. W. "Computer Modeling of the Local Telephone Network," October 1999, p. 2, available at www.fcc.gov.

44

45

46

47

Rebuttal Report of Dr. John B. Hayes -13-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 17 of 26

ILEC's local telephone network, and they have been used to control for variations in the cost to provide local service across locations. Dr. Dorman testified that he is unfamiliar with the numerous studies that have used the HCPM to study local telephone competition.48 39. Third, Dr. Dorman claims I need data reflecting each individual CLEC's costs to measure damages reliably.49 This again is incorrect. One does not need to specify each CLEC's individual supply curve to produce reliable estimates of the effects of reduced interconnection costs on market prices, quantities, and variety of choices. As a practical matter, it is sufficient to model the supply curve for the entire competitive fringe within each local telephone market.50 Therefore, there is no need for detailed data on each individual CLEC's costs. The model can be estimated using data on the costs to provide local telephone service in each market. Data on the interconnection and other costs to provide local telephone service in each market are available from state PUCs (for interconnection fees) and the HCPM (for other costs), as noted above. 40. Fourth, Dr. Dorman claims I have not identified data to measure market shares of CLECs and Qwest.51 There are several potential sources of data for this purpose. The FCC collects data on CLEC and ILEC market shares. I reported some of that data in Exhibit 1 to my initial report. TNS Telecoms is a third party source for data on local telephone service market shares. Exhibit R1 reports shares in 8 cities located in the Qwest service area that are based on TNS data.

D. Benchmark Measurement of Damages
41. Dr. Dorman (¶¶ 11, 26) first claims I have not identified a benchmark for measurement of class wide damages. For purposes of this litigation, useful benchmarks are locations that experienced a reduction in interconnection fees and are otherwise similar to the Qwest service areas included in the proposed class. I have examined data on interconnection fees in the 8 states included in the proposed class and have determined that UNE-P rates were reduced substantially in 7 of these 8 states. Oregon, where UNE-P rates were the lowest of any of the 8 states at the outset, is the only state in the proposed class where UNE-P rates were not reduced. Exhibit R2 contains UNE-P rates for the 8 states in the proposed class

48 49 50

Dorman Deposition, pp. 139-141. Dorman Deposition, pp. 134-135. See, for example, Kahai, Simran, Kaserman, David L., and Mayo, John W., "Is the `Dominant Firm' Dominant? An Empirical Analysis of AT&T's Market Power," Journal of Law and Economics, Vol. XXXIX (October 1996), pp. 499-517; Blank, Larry, Kaserman, David L., and Mayo, John W., "Dominant Firm Pricing with Competitive Entry and Regulation: The Case of IntraLATA Toll," Journal of Regulatory Economics, Vol. 14 (1998), pp. 35-53; Suslow, Valerie Y., "Estimating Monopoly Behavior with Competitive Recycling: An Application to Alcoa," Rand Journal of Economics, Vol. 17, No. 3 (Autumn 1986), pp. 389-403. Dorman Deposition, pp. 86, 210-211, 216-217.

51

Rebuttal Report of Dr. John B. Hayes -14-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 18 of 26

during the period April 2001 through August 2004. These data indicate that the actual states that are included in the proposed class could serve as useful "before and after" benchmarks for measurement of damages. 42. Second, Dr. Dorman (¶ 28) points out that Braunstein assumed a distribution of customer types to measure customer benefits from reduced interconnection fees instead of using actual customer data. Actual customer data identifying customer types for purposes of measuring damages for the proposed class are available from the defendant's business records and from third party sources such as TNS Telecoms.52 There would be no need to rely on an assumed distribution of customer types to measure damages in this litigation, because actual data are readily available. 43. Third, Dr. Dorman (¶ 28) points out that Professor Braunstein found some variation in the benefits from reduced interconnection fees. The fact that there was variation in the benefits is not, by itself, a problem for the measurement of class wide damages. The key point is that the sources of variation can be observed and accounted for in the analysis. Professor Braunstein found that the benefits varied depending on how intensively customers used their local telephone service. In general, customers that purchased larger quantities of local telephone service benefited more than those that purchased less. In this instance, we would be able to observe how much local service each class member purchased, and we would be able to take those differences in usage into account when measuring individual damages. 44. Fourth, Dr. Dorman (¶ 28) points out that Professor Braunstein did not include business customers in his analysis. There is no reason why the method for assessing the dollar benefits to local service customers from reduced interconnection fees that was discussed in the Braunstein paper could not be used to measure damages to business customers, and Dr. Dorman cites none.

IV.

Disputes Regarding Proper Economic Analysis A. "Too Much" Variation

45. Dr. Dorman (¶ 30) claims there is too much variation among class members to measure damages to individual class members reliably.53 In this section I first discuss each of the specific sources of variation that Dr. Dorman identifies. I conclude that each of these sources of variation has been widely studied and successfully addressed in the existing economic literature. Thus the sources of variation that Dr. Dorman identifies do not present novel or uniquely complex issues. For this and other reasons, I anticipate that these sources of variation can be considered, so that reasonable estimates of damages may

52 53

See n. 44 for a summary of customer data available from TNS Telecoms. See also, Dorman Deposition, p. 113.

Rebuttal Report of Dr. John B. Hayes -15-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 19 of 26

be formulated for individual class members. Following my analysis of these sources of variation, I discuss more generally Dr. Dorman's conclusion that there is "too much" variation present in the proposed class to calculate reasonable estimates of damages. I conclude that Dr. Dorman fails to distinguish two distinct types of variation that have different implications for the measurement of class-wide damages. His failure to distinguish these sources of variation causes him to misinterpret the econometric results reported by Professor Economides. See my discussion of this error in section III.B. After reviewing these deficiencies in Dr. Dorman's analysis, I conclude that class wide impact and damages can be measured reliably using the common methods and evidence that I discussed in my initial report. 1. Sources of Variation in the Proposed Class 46. Dr. Dorman (¶¶ 14, 31-39) specifically references four sources of variation that he believes are present in the class. First, Dr. Dorman (¶¶ 31-33) claims there is too much variation in the regulations applicable to local telephone service to measure damages reliably. I addressed this issue in my initial report (¶¶ 35-38, 63-64) where I noted that (1) most of the regulations that apply to local telephone service would not affect calculation of damages to class members, and (2) for those regulations that might affect damages, there is limited variation. More fundamentally, variation in regulations that can be observed and measured will not adversely impact damages calculations. The regulations that could affect damages in this litigation can be observed and their effects measured. Several economic analyses of competition in telecommunications markets have observed, measured, and successfully accounted for the effects of variation in regulations on outcomes. I referenced several of these studies in my initial report,54 and I provide additional references here.55 For example, Blank, et al. estimate an econometric model of IntraLATA toll pricing.56 The model accounts for the impact of variation across states in

54

Blank, et al (1998) control for the effects of regulation in their study of competition in intraLATA toll markets. (Blank, Larry, Kaserman, David L., and Mayo, John W., "Dominant Firm Pricing with Competitive Entry and Regulation: The Case of IntraLATA Toll," Journal of Regulatory Economics, Vol. 14 (1998), pp. 35-53.) Greenstein and Mazzeo (2003) control for the effects of regulation in their study of CLEC differentiation. (Greenstein, S., and Mazzeo, M., "Differentiation Strategy and Market Deregulation: Local Telecommunication Entry in the Late 1990s," NBER Working Paper 9761, June 2003, available at http://www.nber.org/papers/W9761.) Abel, J., "Entry into Regulated Monopoly Markets: The Development of a Competitive Fringe in the Local Telephone Industry," Journal of Law and Economics, Vol. XLV (October 2002), pp. 297; Declaration of Dale E. Lehman in FCC 03-173, filed on behalf of Qwest Communications International Inc., "Investment and the Level of UNE-P Rates: A Critique of the Willig Study," January 30, 2004; Eisner, James and Lehman, Dale E., "Regulatory Behavior and Competitive Entry," June 2001, available at http://www.aestudies.com/library/elpaper.pdf; Zolnierek, J., Eisner, J., and Burton, E., "An Empirical Examination of Entry Patterns in Local Telephone Markets," Journal of Regulatory Economics, Vol 19, No. 2 (March 2001), pp. 143-159. Blank, Larry, Kaserman, David L., and Mayo, John W., "Dominant Firm Pricing with Competitive Entry and Regulation: The Case of IntraLATA Toll," Journal of Regulatory Economics, Vol. 14 (1998), pp. 35-53.

55

56

Rebuttal Report of Dr. John B. Hayes -16-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 20 of 26

the type of regulation. In particular, the model accounts for the different effects of incentive regulation (e.g. price cap regulation) and rate-of-return regulation. Exhibit 2 in my initial report showed that these are the relevant forms of regulation present in the states included in the class. These existing economic studies demonstrate that the effects of variation in telecommunications regulations routinely and successfully are addressed in the economic literature. 47. Second, Dr. Dorman (¶¶ 34-36) claims the differences among customer purchases are too great to measure damages reliably. The essence of his argument is that local telephone service is a mix of "plain old telephone service" (i.e., the ability to place and receive local telephone calls) and "other" local telephone services such as vertical features (e.g., call waiting and call forwarding), voice mail, and local toll calling, and that variations in customer purchases among this mix are too great to measure accurately. This issue has successfully been addressed in the economic literature for many years. There have been several studies of the demand for local telephone service that take into account the mix of local telephone services, and there are well-developed and widely-accepted econometric methods of analysis that are applicable to this issue. For example, Train, et al. (1987) estimate demand for local telephone service by modeling choice of a calling plan and calling pattern (number and distance of calls).57 The Economides study is a more recent example of the literature on estimating the demand for local telephone service. Economides extends the analysis in Train by modeling service plans as a combination of a discrete choice (which service plan the customer selects) and a continuous choice (how many calls the customer makes). These existing economic studies demonstrate that the effects of variation in customer purchases routinely and successfully are addressed in the economic literature. 48. Third, Dr. Dorman (¶¶ 37-39) claims that differences in the intensity of competition across markets for local telephone services are too great to measure damages. In fact, the effects of variation in the intensity of competition on market prices have been a central focus of economic research for many years, and economists have developed a large number of methodological tools to account for this type of variation. Bresnahan is one widely cited survey of the older literature on this topic.58 The economic literature addressing the effects of variation in the intensity of competition across local telephone markets is more recent, since there was virtually no competition in these markets prior to 1996. I referenced several studies that have explored this issue in my initial report,59 and I provide an

57

See Train, Kenneth E., McFadden Daniel L., and Ben-Akiva, Moshe, "The Demand for Local Telephone Service: A Fully Discrete Model of Residential Calling Patterns and Service Choices," Rand Journal of Economics, Vo. 18, No. 1 (Spring 1987), pp. 109-123. Bresnahan, Timothy F., "Empirical Studies of Industries with Market Power," in Schmalensee, Richard, and Willig, Robert, Handbook of Industrial Organization (North Holland, 2000), pp. 1011-1057. Blank, et al (1998) control for variations in the intensity of competition in their study of competition in intraLATA toll markets. (Blank, Larry, Kaserman, David L., and Mayo, John W., "Dominant Firm Pricing with Competitive Entry and Regulation: The Case of IntraLATA Toll," Journal of Regulatory Economics,

58

59

Rebuttal Report of Dr. John B. Hayes -17-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 21 of 26

additional reference here.60 For example, Abel ( 2002) utilizes an econometric model to explain the factors contributing to entry into the local telephone markets in LATAs spanning 25 states.61 He adjusts for differences in the profitability of each LATA using FCC data to calculate profit per line.62 These existing economic studies demonstrate that the effects of variation in the intensity of competition routinely and successfully are addressed in the economic literature. 49. Fourth, Dr. Dorman (¶¶ 14, 21) claims that the structure and amount of CLEC supply costs are likely to vary widely geographically and across CLECs. I agree with Dr. Dorman that there are potentially important sources of geographic variation in CLEC costs that the analysis should accommodate. However, the general issue of controlling for variation in supply costs, has been widely studied and successfully addressed in prior research. Bresnahan is one widely cited survey of the older literature on this topic.63 I also cited several studies that successfully address this issue in the context of local telephone service in my prior report, and I provide additional references here.64 For example, Greenstein and Mazzeo (2002) model CLEC entry into local telephone markets and adjust for differences in interconnection costs using data from Gregg's (2002) survey of UNE prices.65

Vol. 14 (1998), pp. 35-53.) Greenstein and Mazzeo (2003) control for variations in the intensity of competition in their study of CLEC differentiation. (Greenstein, S., and Mazzeo, M., "Differentiation Strategy and Market Deregulation: Local Telecommunication Entry in the Late 1990s," NBER Working Paper 9761, June 2003, available at http://www.nber.org/papers/W9761.)
60

Abel, J., "Entry into Regulated Monopoly Markets: The Development of a Competitive Fringe in the Local Telephone Industry," Journal of Law and Economics, Vol. XLV (October 2002). Abel, J., "Entry into Regulated Monopoly Markets: The Development of a Competitive Fringe in the Local Telephone Industry," Journal of Law and Economics, Vol. XLV (October 2002). Profit is closely related to the "intensity of competition" in that more competitive markets tend to exhibit lower profit margins. See, for example, Robert S. Pindyck and Daniel L. Rubinfeld, Microeconomics, Sixth Edition (Pearson Prentice Hall, 2005), pp. 283-284. Bresnahan, Timothy F., "Empirical Studies of Industries with Market Power," in Schmalensee, Richard, and Willig, Robert, Handbook of Industrial Organization (North Holland, 2000), pp. 1011-1057. Roycroft, Trevor R., "Empirical Analysis of Entry in the Local Exchange Market: The Case of Pacific Bell," Contemporary Economic Policy, Vol. 23, No. 1 (January 2005), pp. 107-115; Abel, J., "Entry into Regulated Monopoly Markets: The Development of a Competitive Fringe in the Local Telephone Industry," Journal of Law and Economics, Vol. XLV (October 2002); Declaration of Dale E. Lehman in FCC 03-173, filed on behalf of Qwest Communications International Inc., "Investment and the Level of UNE-P Rates: A Critique of the Willig Study," January 30, 2004; Rosston, Gregory L., and Wimmer, Bradley S., "Local Telephone Rate Structures: Before and After the Act," Information Economics and Policy," Vol. 17, No. 1 (January 2005), pp. 13-34; Greenstein, S., and Mazzeo, M., "Differentiation Strategy and Market Deregulation: Local Telecommunication Entry in the Late 1990s," NBER Working Paper 9761, June 2003, available at http://www.nber.org/papers/W9761. Greenstein, S., and Mazzeo, M., "Differentiation Strategy and Market Deregulation: Local Telecommunication Entry in the Late 1990s," NBER Working Paper 9761, June 2003, available at

61

62

63

64

65

Rebuttal Report of Dr. John B. Hayes -18-

Case 1:02-cv-01977-RPM

Document 95

Filed 05/11/2006

Page 22 of 26

2. How Much Variation is "Too Much"? 50. Dr. Dorman (¶ 30) emphasizes that there is "too much" variation in the proposed class in this litigation to measure damages reliably on a class-wide basis.66 But virtually all class action litigation involves and accommodates variation among the class members. A central purpose of an econometric analysis is to identify those independent variables that can explain the variation in the dependent variable. 51. At deposition, Dr. Dorman explained that economists routinely apply formal standards to evaluate whether an analysis adequately models the sources of variation in the dependent variable after the analysis has been completed.67 Of course, those standards generally cannot be used to determine whether there is "too much" variation for purposes of class certification, because detailed analyses of class-wide impact and damages generally have not been conducted at this stage in the litigation. That is certainly the situation in this particular litigation. When asked how he reached the conclusion that my methodology would fail due to "too much variation," Dr. Dorman simply adverted to his (mistaken) interpretation of wide confidence intervals in the Economides study.68 In fact, concluding that there is "too much" variation in input data based on standards applicable to final analyses would put the cart before the horse. 52. For purposes of class certification, economists rely on prior studies of related issues, together with their training and experience, to determine whether class-wide impact and damages can be reasonably estimated. In this instance we have a large number of prior studies addressing the effects of increased competition in local telephone markets, and raising issues that are closely related to the issues raised in this litigation, to inform our opinions. I have reviewed these studies and provided numerous references to support my conclusion that the sources of variation at issue in this litigation are amenable to common analysis of class-wide impact and damages. In contrast, Dr. Dorman has not identified any economic literature that casts doubt upon that conclusion. 53. Instead, Dr. Dorman largely bases his claim that there is "too much" variation on his faulty analysis of the Economides study.69 Specifically, Dr. Dorman claims that Economides was unable to measure the welfare gains caused by increased competition for local telephone service with a reasonable degree of precision.70 He then reasons from this (incorrect) result that the econometric methods used in the Economides study could not

http://www.nber.org/papers/W9761; Gregg, Billy Jack,