1 2 3 4 5 6 7 Robert A. Mittelstaedt #60359 ramittelstaedt@jonesday.com Craig E. Stewart #129530 cestewart@jonesday.com David C. Kiernan #215335 dkiernan@jonesday.com JONES DAY 555 California Street, 26th Floor San Francisco, CA 94104 Telephone: (415) 626-3939 Facsimile: (415) 875-5700 Attorneys for Defendant APPLE INC. 8 UNITED STATES DISTRICT COURT 9 NORTHERN DISTRICT OF CALIFORNIA 10 11 12 THE APPLE iPOD iTUNES ANTI-TRUST LITIGATION 13 Lead Case No. C 05-00037 JW APPLE�S RESPONSE TO PROFESSOR NOLL�S JULY 18 DECLARATION 14 15 16 17 18 Apple Inc. submits this response to Professor Noll�s July 18 declaration. This response is 19 supported by the accompanying Supplemental Report of Dr. Michelle Burtis dated July 22, 2011. 20 Professor Noll continues to rely on the preliminary regression model that he previously 21 submitted to try to address the impact of iTunes 4.7. As demonstrated at the June 27 hearing, 22 Noll testified that his preliminary regression was unreliable for numerous reasons. Although he 23 has now added a �dummy� variable for iTunes 7.0, he has not purported to fix the fundamental 24 defects in his model. Thus, for the same reasons that forced him to admit that he was not relying 25 on his model to prove anything with respect to iTunes 4.7, he cannot rely on his model to prove 26 anything with respect to iTunes 7.0. 27 Even putting aside the fundamental reasons why his model continues to be unworkable, it 28 does not show that iTunes 7.0 had any classwide impact. If it could be used to show anything, it -1SFI-703833v3 Case No. C 05-00037 JW APPLE�S RESP TO NOLL�S JUL 18 DECL 1 would show that iTunes 7.0 lowered iPod prices�the opposite of Plaintiffs� unsupportable 2 impact and damages theory. 3 All of this underscores the importance of obtaining Noll�s supplemental expert report on 4 class certification�which is the next step ordered by the Court. Then Noll should be deposed 5 and an evidentiary hearing should be held. 6 I. 7 THE COURT�S TWO-STEP PROCESS SHOULD BE ADHERED TO. This Court�s June 27 Order set up a two-step process: � 8 9 a declaration by Professor Noll��stating that he can show antitrust impact and a methodology for measuring damages� for iTunes 7.0, with an optional response by 10 Apple, to be followed by � 11 12 a Supplemental Expert Report by Professor Noll and a Supplemental Rebuttal Expert Report by Dr. Burtis, on a schedule to be agreed on by the parties. 13 Plaintiffs refuse to produce any materials relating to the regression analysis until Professor 14 Noll submits his Supplemental Expert Report. (Declaration of David C. Kiernan In Support of 15 Apple�s Response to Professor Noll�s July 18 Declaration, Ex. 1.) So it is impossible to know 16 precisely how the regression was performed. For the reasons shown below, it is essential that 17 Professor Noll produce his working materials and be subjected to cross-examination before the 18 Court takes any action based on his declaration. 19 II. PROFESSOR NOLL�S JULY 18 DECLARATION FAILS TO SHOW ANTITRUST 20 IMPACT OR DAMAGES. 21 A. 22 23 Professor Noll has not remedied the defects that he admitted made the model unreliable as applied to iTunes 4.7. In his report submitted last March (Dkt. 551), Noll stated that he had run a preliminary 24 regression that produced �highly significant� results showing a price effect of $6 from Harmony 25 being disabled by iTunes 4.7 in October 2004. He claimed that the adjusted R2 for the equations 26 was 0.98, which he said meant that his equations explain �virtually all of the variations in prices 27 across models of iPods.� In his report, the only qualification he placed on the reliability of the 28 results was that there are some �remaining data problems.� SFI-703833v3 -2- Case No. C 05-00037 JW APPLE�S RESP TO NOLL�S JUL 18 DECL 1 At his deposition, however, he disavowed any reliance on the regression to determine 2 whether iTunes 4.7 had any impact on iPod prices. He admitted that �I drew no causal inferences 3 from that regression,� that he couldn�t say �whether the launch of Harmony or the disabling of 4 Harmony had any effect on iPod prices,� and that he had not even attempted to determine �when, 5 if ever, the disabling of Harmony had an impact on iPod prices.� 1 The regression model�s 6 problems, acknowledged by Noll, were not limited to incomplete or inaccurate data. Rather, he 7 admitted that, even apart from data issues, he hadn�t done the required work to determine whether 8 the model itself was valid. 2 9 With his latest report, Noll engages in the same exercise. Relying on the same admittedly 10 unreliable preliminary model, he asserts that, by adding a variable for iTunes 7.0, the model 11 indicates that iTunes 7.0 caused iPod prices to be elevated by $4.85. As before, he describes 12 these findings as �highly significant� and points to an adjusted R2 of 0.98 as showing that the �fit 13 of the equation is very high.� 3 As explained below, however, his regression shows no such 14 thing�as Noll will have no choice but to admit when his deposition is taken again or an 15 evidentiary hearing is held, just as he admitted last time. In fact, if there were any validity to his 16 regression model, which he has already admitted there is not, it would indicate that iTunes 7.0 17 caused iPods to be priced lower than they otherwise would have been, by $1.69. See 18 Supplemental Report of Michelle Burtis, filed herewith, ¶ 3-5. 19 20 Although he is using essentially the same model as before, Noll does not attempt to address the most fundamental problems that made it unreliable, either the problems 21 22 23 24 25 26 27 28 1 Kiernan Decl., Ex. 2, pp. 92, 113, 202. Among the numerous reasons why the 4.7 regression model was not reliable, Noll admitted that he may not have included all the factors that influenced iPod prices (pp. 92-93, 123); he could have been measuring a price change that occurred before and unrelated to iTunes 4.7 (pp. 94-98); he �ran out of time� to determine whether his model was invalid due to lack of sufficient independent price observations (pp. 112-13); he was looking at average iPod prices, not prices for individual models (pp. 128-29, 146); and he failed to include a variable for the Harmony relaunch in April 2005 (pp. 146-48). That�s why he was not �relying on it for anything� (p. 134). The foregoing pages are included in Exhibit 2 to the Kiernan Declaration. 3 As demonstrated by the Burtis Supplemental Report (¶¶ 6-10), Noll�s statements are misleading, which is why he distanced himself from them at deposition. 2 SFI-703833v3 -3- Case No. C 05-00037 JW APPLE�S RESP TO NOLL�S JUL 18 DECL 1 acknowledged by him at his last deposition or those identified by Dr. Burtis. Burtis Supp. Report, 2 ¶¶ 6-15. Instead, pushing the bounds of advocacy, he asserts (p. 2) that �my prior analysis found 3 that the 4.7 update elevated iPod prices.� There is no way to square that assertion with his 4 specific deposition admissions that his model was unreliable and did not show �whether the 5 disabling of Harmony had any effect on iPod prices.� Kiernan Decl., Ex. 2, p. 113. This is not a 6 �battle between experts� as much as the difference between Noll�s written work and his sworn 7 testimony on cross-examination. 8 In short, Noll�s July 18 declaration fails to provide any basis for showing antitrust impact, 9 as required by this Court�s June 27 Order. As he admitted with respect to the iTunes 4.7 version, 10 11 he cannot rely on it for anything. B. If Noll�s regression model shows anything (and it doesn�t), it would show that 12 iTunes 7.0 did not cause any classwide antitrust impact or damages. Rather, 13 it �caused� lower iPod prices. 14 For reasons previously discussed (Doc. 633, pp. 3-5; Doc. 646, pp. 1-2), it remains 15 implausible that iTunes 7.0 had any conceivable effect on iPod prices. Under Plaintiffs� theory, 16 that update simply required the additional step of �burning/ripping� to play music from one of 17 multiple sources on some new iPod models. Consumers could still play their CD collections 18 directly on iPods and could still directly play all of their music (including RealNetworks� music) 19 on existing iPods and on new iPod shuffles. They could also play RealNetworks� music on other 20 new iPod models (e.g., iPod nano second generation) simply by copying (�burning�) the music to 21 a CD and loading (�ripping�) it back to a computer for loading to an iPod. As Professor Noll has 22 admitted, there is no theoretical basis for assuming that preventing RealNetworks� music from 23 playing directly on certain iPods had any effect on iPod prices. Kiernan Decl., Ex. 2, pp. 151-53. 24 Even now, Noll repeats (p. 2) that whether Apple set iPod prices above levels that would have 25 otherwise existed �is an empirical matter.� 4 26 27 28 4 Nor does Noll�s purported conclusion that, without iTunes 7.0, iPods would have been priced $4.85 lower pass a basic reality test. That would mean that a 2GB iPod nano second generation shuffle priced at $149 supposedly would have been priced at $144.15. But that never would have happened given Apple�s �aesthetic� pricing, e.g., at $149 or $199 rather than at SFI-703833v3 -4- Case No. C 05-00037 JW APPLE�S RESP TO NOLL�S JUL 18 DECL 1 Noll�s analysis fails to show empirically that Apple elevated iPod prices due to iTunes 7.0. 2 As submitted last March, his regression model used various �dummy� or indicator variables (e.g., 3 the iTunes store, Harmony, 4.7, iTS competitors becoming DRM-free, etc.) and attempted to 4 measure the impact, if any, of those factors on iPod prices. For his latest declaration, Noll adds a 5 new dummy variable for iTunes 7.0 to the same model. If he had treated all the variables the 6 same, the result for the iTunes 7.0 variable (again putting aside all the other defects that render 7 the model invalid) would be negative $1.69, indicating that 7.0 caused iPod prices to be lower 8 than they otherwise would have. Burtis Supp. Report, ¶¶ 3-5. 9 To mask that result, Noll resorts to an inexcusable sleight-of-hand. He turns off (p. 2) the 10 variable for 4.7 at the same time he turns on the variable for 7.0. This is contrary to the way that 11 Noll treats all the other variables and is contrary to the way he treated the iTunes 4.7 variable in 12 his previous regression. Burtis Supp. Report, ¶¶ 3-5. 5 By turning off the variable despite his 13 theory that it would continue to cause �lock in,� Noll creates the false impression that iTunes 7.0, 14 rather than iTunes 4.7, is causing the positive price effect. 15 Exhibit 1 to Noll�s declaration shows (assuming arguendo that the regression model is 16 valid) the price effect of iTunes 4.7 from October 2004 to September 2006 is $6.54 and that the 17 price effect of iTunes 7.0 from September 2006 to March 2009 is $4.85. If the model were able 18 to show anything, the difference between those two numbers would indicate that iTunes 7.0 19 caused average iPod prices to drop by $1.69. Burtis Supp. Report, ¶¶ 3-5. So the accurate way to 20 21 22 23 24 25 26 27 28 (continued�) interim levels. Doc. 512, p. 12. Noll acknowledged as much at his deposition. Kiernan Decl, Ex. 2, pp. 208-13. 5 In his previous regression, all variables including 4.7 are �left on� for the entire period. That is because they reflect the ongoing impact of those events. The variable for iTMS, for example, would (at least, in theory) measure the total impact of iTMS on iPod price from the April 2003 launch to the end of the alleged class period, March 2009. If that variable were turned off at, say, the date that the iPod shuffle was launched, the variable for the iPod shuffle would pick up the continuing impact of iTS as well (assuming the regression model were otherwise appropriate). Noll does not explain why he didn�t turn off the 4.7 variable in his previous regression. Nor does he explain why he doesn�t turn off that variable in his new regression at the time earlier iTunes updates like 4.8 , 4.9, 5.0 or 6.0 were released. SFI-703833v3 -5- Case No. C 05-00037 JW APPLE�S RESP TO NOLL�S JUL 18 DECL 1 depict the �findings� of his analysis would be to say that iTunes 7.0 caused average iPod prices to 2 drop by $1.69 compared to what they would have been without the update. Id. 3 At page 2 of his July 18 declaration, Noll tries to justify his trick by saying that turning off 4 the 4.7 variable permits a test of whether iTunes 7.0 �perpetuated elevated prices� for iPods. But 5 when he says �perpetuate,� he does not mean that the model measures whether iTunes 7.0 caused 6 the allegedly elevated but lawful prices attributed to iTunes 4.7 to remain elevated when they 7 otherwise would have been declining. Instead, he is saying that the allegedly elevated prices 8 persisted even after iTunes 7.0 was released. That is because the continued existence of a price 9 effect, rather than any causal effect from 7.0, is all that his model could possibly have been 10 measuring. But that is not antitrust impact because iTunes 4.7 was lawful and any price effect 11 that may have persisted was likewise lawful. 12 Finally, at page 4, Noll states that the coefficient for the 7.0 variable �indicates that it 13 caused the wholesale price of iPods to be elevated by $4.85 compared to a price elevation of 14 $6.54 due to update 4.7.� When Noll is cross-examined at deposition and/or an evidentiary 15 hearing, he must retract that statement. It is an inaccurate report of his results. What he is really 16 saying is that iTunes 7.0�s negative effect on prices caused the price effect of iTunes 4.7 to drop 17 from $6.54 to $4.85. See Burtis Supp. Report, ¶¶ 3-5. Again, that is not antitrust injury or 18 damages. 19 20 Dated: July 22, 2011 21 Respectfully submitted, Jones Day 22 23 By: /s/ Robert A. Mittelstaedt Robert A. Mittelstaedt 24 Counsel for Defendant 25 26 27 28 SFI-703833v3 -6- Case No. C 05-00037 JW APPLE�S RESP TO NOLL�S JUL 18 DECL