For Public Inspection
FOR PUBLIC INSPECTION
channels, local, state, and regional news and public affairs channels, Spanish-language news channels,
an international news channel - CNN International - and a single on-demand news channel. These
were determined to be news, business news, and/or public affairs channels, based on individual
examination of channel names and information on the types of programming provided on specific
channels, by both me and my support team as well as counsel for Bloomberg. I include a list of all
channels designated as possible news, business news, and/or public affairs channels as Appendix B
at the end of this declaration.
15. I have examined the websites and other information regarding the local news
networks listed in Appendix B. \'Vith one exception, this material confirms that these networks
concentrate on local news programming and in particular focus on local news reporting and analysis
between 6:00 a.m. and 4:00 p.m. in the U.S. Eastern Time Zone.! I have also examined the websites and other information regarding the local, state, and regional public affairs networks listed in
Appendix B. This material confirms that these networks concentrate on public affairs programming
and in particular focus on public affairs reporting and analysis between 6:00 a.m. and 4:00 p.m. in
the U.S. Eastern Time Zone.
16. For convenience, I will refer to the list of news, business news, and public affairs
channels in Appendix B as "possible news channels" in the balance of this declaration.
17. An initial examination of the channels provided by each headend on each channel position illustrated an important problem with the raw TMS data: there were many instances of multiple channels being offered on a single channel position. This was an important problem, as
I The exception is the Virginia News Channel. Counsel for Bloomberg has informed me that, at one time, the Virginia News Channel appeared to rebroadcast local news programming from the Roanoke CBS affiliate, Channel 7, on a continuous loop. I have not been able to determine if it continues to do so. While I have included it in Appendix B because it is still included in certain channel lineups in the TMS data, its presence or absence does not affect any of the calculations regarding news neighborhoods in the 35 most populated DMAs set forth below. FOR PUBLIC INSPECTION
defining a channel neighborhood necessarily requires accurate information on which channel(s) are being provided on which channel positions.
18. Further examination of the data indicated that most of these instances were due to headends providing multiple channel lineups according to the device households were using to receive the programming. Channels being delivered to different devices naturally shared channel positions as, for example, channel 5 for a headend's Analog device need not have the same channel as channel 5 for that headend's Digital device.
19. To address this problem, I first defined a channel lineup as the set of channel positions provided by each device on a particular headend. For example, there were two channel lineups on the headend: one delivered to Analog devices (with 73 channel positions) and one delivered to Digital (non-rebuild) devices (with 536 channel positions).
20. Overall, there were 1,909 lineups across the 1,014 Comcast headends. The vast majority of these were Analog and Digital (non-rebuild) lineups: all 1,014 headends offered an
Analog lineup and 849 offered a Digital (non-rebuild) lineup. The remaining Comcast lineups were
''Analog (rebuild)" (20 lineups), "Cable-ready TV" (14 lineups), "Digital (rebuild)" (10 lineups), and
"Cable.N' and "Cable B" (1 lineup each). 164 Comcast headends offered only one lineup (all of which were Analog), 813 offered two (all but one of which were Analog and Digital (non-rebuild)), and the remaining 37 offered either three or four. For simplicity, I decided to keep one lineup per
Comcast headend. For each headend that provided more than one lineup that included a Digital
(non-rebuild) lineup, I kept that lineup. This decision was made because Comcast is migrating quickly to all-digital systems, and digital lineups are therefore more relevant for the future than are analog lineups. For the one headend that had an Analog and Cable-ready TV lineup, I kept the FOR PUBLIC INSPECTION
Analog lineup. The result was 1,014 channel lineups for the 1,014 Comcast headends, 849 of which were Digital (non-rebuild) lineups and 165 of which were Analog lineups.
21. Including a single lineup per Comcast headend resolved many, but not all, of the instances of multiple channels per channel position. After limiting the sample to one lineup per headend, there remained 2,471 (of 346,740, or 0.71%) channel positions in which there was more than one listed channel. Given the importance of accurately identifying which channel was in a channel position for the purposes of defining channel neighborhoods, I determined to resolve all of these instances. I did so in three steps.
22. First, I determined if the duplicated channels on a given channel position were effectively the same channel. I determined this was so if they shared the same channel name and then dropped all but one of the repeated channels for each channel position. This resolved 422 of the 2,471 duplicate channel positions, leaving 2,049 (or 0.59% of all headend-channel positions) remalnlng.
23. Second, I determined if at least one of the channels on the duplicated channel position was a possible news channel. If none were a possible news channel, then it was not going to influence how I later calculated neighborhoods of news channels and it didn't matter if it was a duplicated channel position. This resolved 1,957 of the remaining 2,049 remaining duplicate positions, leaving 92 (or 0.03%) duplicated positions. These 92 duplicated positions were on 69
Comcast headends.
24. Third, I resolved these last remaining duplicates. For 12 headends, the problems arose due to different feeds of CNN en Espanol (Mexico and US vs. United States) being provided in the same channel position. I decided to treat these as the same channel by dropping one of them and renaming the other, "CNN en Espanol - All Feeds". For the remaining 57 headends, the FOR PUBLIC INSPECTION
appropriate channel lineup was found by comparing the headend id and community name from the
TMS data to channel lineups for that community on the Comcast website. The Comcast website was reviewed, and it was determined which of the channels listed for the duplicated channel positions appeared to be listed in that position according to the lineup for that community. The decisions made for these duplicated channel positions are listed as Appendix C to this declaration.
25. All of these steps resulted in the final dataset on which I performed my analysis of
Comcast's neighborhoods of news and public affairs channels. It contained 346,740 channel positions on 1,014 Comcast headends.
26. I next analyzed whether neighborhoods of news, business news, and public affairs channels were evident on the cleaned Comcast lineups in the TMS data. For convenience, I will refer to neighborhoods of news, business news, and public affairs channels as "news neighborhoods" in the balance of this declaration.
27. I defined news neighborhoods in three steps. First, I defined the set of news channels which would be counted in the definition of news neighborhoods. Second, I determined how calculations of news neighborhoods should handle blank channel positions. Third, I defined neighborhoods based on alternative definitions of adjacent relative channel positions. I describe each of these steps in what follows.
28. I first defined the set of news channels which would be counted in the definition of news neighborhoods. This consisted of all national news, public affairs, and business news channels as well as all local, state, and regional news and public affairs channels. I will refer to this set of channels as "news channels" for the balance of this declaration. Another way of describing the choice is that it consisted of all possible news channels defined previously less High Definition,
Spanish-language and other foreign-language news channels, and the on-demand news channel. HD FOR PUBLIC INSPECTION
news feeds were omitted from the definition because HD channels are a higher-quality product that
largely replicate the content of standard definition feeds. Spanish-language and other foreign
language news channels were omitted from the definition because they are typically located in
Spanish-language and/or foreign-language neighborhoods rather than with English-language
channels (including English-language news channels) and are identified by Comcast as
"Multicultural" programming in their channel lineups. The on-demand news channel was omitted
because it does not provide standard, linear television programming. The set of channels which were included in my calculations of neighborhoods is given by the starred (*) channel groups in
Appendix B.
29. Second, I determined how calculations of neighborhoods should handle blank
channels between channel positions reported in the TMS data. I considered two possibilities. First,
I considered the actual channel position as reported by TMS in the definition of news
neighborhoods. This treated a blank channel as a gap between adjacent channels. For example,
news channels in positions 48, SO, 51, and 52, with channel 49 blank (as in
_ would not have qualified as a neighborhood under a neighborhood definition that required four news channels to be contiguous. Alternatively, I considered relative channel positions in the definition of a news neighborhood. In this case, I simply listed all the channels on a headend in order of their reported channel position and assigned to them a new index, which I called their relative channel position, when calculating a neighborhood. For example, the same four channels in
_ occupied relative channel positions 46, 47, 48, and 49. Under this alternative definition of channel position, they did qualify as a neighborhood under a neighborhood definition that required four news channels to be contiguous. FOR PUBLIC INSPECTION
30. I use the definition of a neighborhood based on the second method, relative channel position, in the balance of this declaration. I did so to reflect what I felt households are likely to believe to be a neighborhood based on their television usage. Many households likely decide what to watch by flipping through television channels. If so, then what matters to their perception of a neighborhood are the relative positions of channels, not their absolute channel numbers.
31. Having defined the news channels which would count in the calculation of news neighborhoods and the metric - relative channel position - that would define adjacency of channels in a channel lineup, I next defined news neighborhoods.
32. In the FCC's Order approving the Comcast-NBC Universal Merger, the Commission defined a neighborhood as a "significant number or percentage of news and/or business news channels [that are] substantially adjacent to one another in a system's channel lineup." Because the relevant definition uses the term "substantially adjacent" rather than "adjacent," I did not employ a definition of neighborhood that required absolute contiguity (e.g., four channels in a row). Rather, I defined both four contiguous news channels and four news channels within five relative channel positions to be a news neighborhood, as well as any longer similar pattern. For example, five or six contiguous news channels and/or five news channels within six relative channel positions or six news channels within seven relative channel positions are also defined to be news neighborhoods.
In essence, I defined a news neighborhood to be all groups of channels that, based on their relative channel position in a Comcast headend's channel lineup, included (1) at least four contiguous news channels or (2) at least four news channels in a group of five channel positions.
33. In what follows, I briefly describe how I implemented these definitions. Locating news neighborhoods was easiest for groups of news channels that were contiguous in Comcast's channel lineups. To do so, I defined a "pod" of channels as a collection of like channels (news or FOR PUBLIC INSPECTION
non-news) that were adjacent in a Comcast channel lineup. The type of the first pod in any
channel lineup depends on whether the first channel in that lineup is a news channel or a non-news
channel. Most lineups begin with a non-news channel and so the first pod was typically a non-news
pod. For expositional convenience, suppose that the first pod is indeed a non-news pod. I then
examined the next channel in the channel lineup. If it, too, was a non-news channel, then it was
added to the first pod, making a pod of two non-news channels. I then examined the next channel
in the channel lineup.
34. This process continued until I came across a channel of a type different than that of
the current pod. Suppose for convenience this happened in the eighth channel position. Since the
first pod was a non-news pod, the eighth channel must have been a news channel. This defines the
end of the first, non-news, pod (which was a pod of seven channels) and the beginning of the first news pod. As always, I continued by examining the next channel in the channel lineup. If it was a news channel, it was added to the second pod, making it a news pod of two channels. If it was a
non-news channel, the second pod was a news pod with a only one channel, and the non-news channel marked the beginning of the third pod (or the second non-news pod). This process continued, adding channels to the previous pod if the channel was the same type as the previous channel in the channel lineup or defining a new pod if the channel was a different type as the previous channel in the channel lineup, until the end of the channel lineup.
35. By construction, pods must alternate between news pods and non-news pods. I defmed a news neighborhood (based on contiguous channel groups) to be a news pod of at least
four channels.
36. I next extended the definition of a news neighborhood to allow for a single non- news channel within a group of at least four news channels. Based on the definition of FOR PUBLIC INSPECTION
neighborhoods using pods as defined above, a group of news channels on positions 32, 33, 35, and
36 with a non-news channel on position 34 would not qualify as a news neighborhood. The
previous definition of pods would define that group of channels as a news pod of two channels
(channels 32 and 33) followed by a non-news pod of one channel (channel 34) followed by a news
pod of two channels (channels 35 and 36).
37. To allow for news neighborhoods with a single non-news channel among news
channels, I created an additional (broader) definition of news pods and re-calculated neighborhoods.
I began by identifying all non-news pods of one channel that were not at the beginning or end of a
channel lineup; by construction, each of these one-channel non-news pods was located between two
news pods. I then examined whether the sum of news channels in the two pods on either side of
the (singleton) non-news pod contained at least 4 news channels. If so, I called the combination of
the two news pods surrounding the singleton non-news pod a "news pod allowing one non-news
channel" and also defined this as a news neighborhood.
38. In the example above, this meant that channels 32-36 would now qualify as a news
neighborhood even though one of those channels (channel 34) was a non-news channel.
39. Taking these two conditions together yielded the complete definition of a news
neighborhood I used in my analysis. A news neighborhood is defined to be all groups of channels
that, based on their relative channel position in a Comcast headend's channel lineup, were part of (1) a contiguous news pod of at least four news channels or (2) a news pod allowing one non-news channel but having at least four news channels.
40. Using this definition of news neighborhood, I analyzed how many Comcast
headends contain news neighborhoods. I found that out of Comcast's 1,014 headends, 676 (or
66.7%) have news neighborhoods. FOR PUBLIC INSPECTION
41. I also analyzed how many Comcast headends that carry B'N have news neighborhoods. Of Comcast's 1,014 headends, 759 (or 74.9%) headends carry BTv. And of these
759 headends, I found that 599 (or 78.9%) have news neighborhoods.
42. I also focused on those Comcast headends that both are located in the 35 most- populous DMAs in the country and carry BTv. 604 Comcast headends are located in the 35 most populous DMAs, and 485 of these headends carry BTV. And of Comcast's 485 headends that are located in the 35 most populous DMAs and carry BTv, I found that 418 headends (or 86.2%) have news neighborhoods.
43. I next examined whether BTV was being carried in those news neighborhoods that exist on Comcast headends. I found that of the 418 headends that are located in the 35 most populous DMAs, carry BTV, and have a news neighborhood, 368 headends have a news neighborhood that does not include BTv. These 368 headends are located in 21 DMAs. A list of these headends may be found in Exhibit G that is attached to Bloomberg's Complaint. A list of the news channels on the channel lineups for these headends may be found in Exhibit H that is attached to Bloomberg's Complaint. A list of the complete channel lineups for these headends was too large to include as a printed attachment and may instead be found as Exhibit L (an electronic attachment).
44. I also found that 365 of these 418 headends have news neighborhoods that include
CNBC but do not include BTV whereas only 17 of these 418 headends have news neighborhoods that include both CNBC and BTV.
45. I also found that a small minority of Comcast headends have two news neighborhoods. For example, in 48 of the 368 headends that are located in the 35 most populous
DMA and carry BTV but have a news neighborhood that does not include BTv, there is a news FOR PUBLIC INSPECTION
neighborhood located between channel 1-100 that includes CNBC, and a separate neighborhood
located above channel 100 that includes BTV
46. I then looked at what news channels were included in news neighborhoods that
excluded BTV. I found that of the 368 headends in the 35 most-populous DMAs that carry BTV
and have a news neighborhood that does not include BTV, 99.7% (or 367) include HLN in a news
neighborhood that excludes BTV, 99.2% include CNBC (or 365) ,97.3% (or 358) include CNN,
93.5% (or 344) include Fox News, and 61.7% (or 227) include MSNBC.
47. I next analyzed the size of news neighborhoods present on Comcast's headends. In
the 368 Comcast headends that carry BTV in the most populous 35 DMAs and have news
neighborhoods that do not include BTV, I found that the average news neighborhood that does not
include BTV contains 5.05 news channels. The majority of these headends, 268, have
neighborhoods with five news channels whereas 46 of these headends have news neighborhoods with four news channels, 46 of these headends have news neighborhoods with six news channels,
and 8 of these headends have news neighborhoods with seven or more news channels.
48. I next calculated the total number of channels located on Comcast headends that were defined as news channels as set forth above. I found that in the 368 Comcast headends that
carry BTV in the 35 most-populous DMAs and have news neighborhoods that do not include BTV,
the average headend carries 11.03 news channels. A total of 151 headends carry exactly 11 news
channels. A total of 78 headends carry 12 news channels. A total of 93 headends carry 10 news
channels. A total of 19 headends carry 9 or fewer news channels, and 27 carry 13 or more news
channels.
49. I next calculated the percentage of channels defined as news channels that were carried in news neighborhoods that did not include BTV I found that in the 368 Comcast headends FOR PUBLIC INSPECTION
that carry BTV in the 3S most populous DMAs and have news neighborhoods that do not include
BTV; the average headend carries 46.2% of news channels in a news neighborhood that does not include BTV; and that 362 of these 368 headends carry 33% or more of news channels in a neighborhood that does not include BTV.
50. I then analyzed the probability that a news neighborhood would be located on a
Comcast headend if channels lineups were determined in a random manner. As above, I focused on
the 485 Comcast headends that both are located in the 35 most populous DMAs and carry BTV.
51. I calculated the probability that a news neighborhood would be located on a
Comcast headend if channel lineups were determined randomly in two steps. First, I calculated, for each of these 485 headends, the number of news channels carried on that headend, which I will
denote M h, and the number of total channels on that headend, which I will denote N h• I counted as
news channels, M h, the same channels I counted in my calculations of news neighborhoods above.
These are given by the starred channel groups in Appendix B. I counted as total channels, N h, all channels offered on a Basic or Expanded Basic tier of service that were not one of (1) broadcast television channels (e.g. ABC, CBS, Fox, etc.), (2) HD channels, (3) On Demand channels, or (4)
Public, Educational, or Government (PEG) channels. I excluded broadcast television channels and
PEG channels because FCC regulations constrain cable operators' flexibility in selecting their channel positions. I excluded HD channels and On Demand channels because they are not linear, standard definition channels and cable operators typically place them in their own channel groupings for non-genre-related reasons. On average across the 485 headends, there were 11.15 news channels and 157.42 total channels per headend.
52. I next estimated the probability, for each headend, that there would be a news
neighborhood of at least 4 news channels in 5 channel positions on that headend if the M h news FOR PUBLIC INSPECTION
channels offered there were randomly distributed among the N h (non broadcast/HD/OnDemand/PEG) total channels offered on that headend. I did so by constructing
100,000 random channel lineups for each headend, calculating the number of these lineups that had a news neighborhood of at least 4 news channels in 5 channel positions, and dividing that number by 100,000. Using numerical simulation to estimate probabilities is a widespread practice in econometrics. As long as there are a large number of simulation draws (I used 100,000 per headend), such methods can yield very precise estimates of the true underlying probability of an event.
53. I found that the average probability that a news neighborhood of at least 4 news channels in 5 channel positions would occur by chance for those Comcast headends that are in the
35 most populous DMAs and carry BTV is 1.20%. The average probability excluding four headends that have particularly low numbers of total channels relative to news channels is 0.93%. Moreover, the probability that news neighborhoods would occur by chance on at least 418 of the 485 headends that carry BTV in the 35 most-populous DMAs is so infinitesimal that it cannot be calculated with precision by a computer. I estimate that chance to be on the order of 10 to the negative seven hundredth power (i.e., a decimal point followed by 699 zeros and then a "1"). Dated: June 13,2011
Gregory S. Crawford
5174638 FOR PUBLIC INSPECTION
APPENDIX A FOR PUBLIC INSPECTION
Gregory S. Crawford
Business Address Home Address Department of Economics Brookfield House University of Warwick 2 Beauchamp Hill Coventry CV4 7AL, UK Leamington Spa Email: [email protected] CV32 5NP, UK Phone: +44 (0)2476 523470 UK Mobile: +44 7549 948788 US Mobile: (520) 977-4589
Education
Ph.D. in Economics, Stanford University, Stanford, CA, 1998 B.A., Economics (with Honors), University of Pennsylvania, Philadelphia, PA, 1991 Professional Experience
University of Warwick, Department of Economics
Professor of Economics, September 2008-present
Director of Research, September 2009-present Courses taught: Graduate: Empirical Industrial Organization (MScjPhD), Empirical Methods. Undergraduate: Introductory Econometrics (time series, limited dependent variables, panel data), Undergraduate Business Strategy.
Centre for Economic Policy Research (CEPR)
Research Fellow, Industrial Organization Programme, February 2011-present
Federal Communications Commission (FCC)
Chief Economist, September 2007 - August 2008
Reported to the then-FCC Chairman, Kevin Martin. Primary responsibilities were to advise the Chairman and his staff regarding the economic issues facing the Commission, to formulate and implement desired policies, to communicate and discuss these policies with senior Commission staff, and to assist as needed the 40+ staff economists. Main workstreams focused on the cable and satellite industries, including bundling and tying in wholesale and retail cable and satellite television markets and the economic analysis of XMjSirius satellite radio merger. Also consulted on spectrum auction design, net neutrality, access pricing, ownership rules, and various international policy issues. Previous to joining the Commission, wrote a sponsored study analyzing media ownership and its impact in television markets.
University of Arizona, Department of Economics
Associate Professor of Economics, September 2008-August 2009 (on leave)
Gregory Crawford cv, June 2011 1 FOR PUBLIC INSPECTION
Assistant Professor of Economics, September 2002-August 2008 (on leave, 2007-08)
Courses taught: Graduate: Empirical Industrial Organization (2nd-year PhD), Business Strategy (MBA) Undergraduate: Introductory Econometrics (cross-section).
Duke University, Department of Economics
Assistant Professor of Economics, September 1997-August 2002
Courses taught: Graduate: Empirical Industrial Organization (2nd-year PhD), Graduate Econometrics (1st-year PhD), Undergraduate: Introductory Econometrics (cross-section), Introductory Microeconomics, The Economics and Statistics of Sports.
Other Academic Appointments
Visiting Professor, European School of Management and Technology, Berlin, Summer 2007.
Visiting Professor, Fuqua School of Business, Duke University, 2000-2001
Consulting Experience (Country)
Evaluating switching costs in fixed voice telephony markets (UK), 2010-11 - Designed and executed reports for Office of Communication (Ofcom) evaluating the impact of automatically renewable contracts (ARCs) introduced by British Telecommunications (BT) in the UK fixed voice telephony market. Ofcom subsequently challenged the practice (March, 2011).
Evaluating competitive harms, Comcast-NBCU (US), 2010, consulting expert- Worked closely with lead expert to design and execute economic and econometric analyses in support of Bloomberg (Television) L.P.s opposition to Comcast-NBCU merger. Analysis included business news market definition and quantifying the potential harms of the merger, including those related to "neighborhooding" of television channels and refusal to carry (foreclosure). Report submitted to media regulator (FCC). FCC conditions required merged firm not to favor their content in general, with specific provisions for the neighborhooding of news (including business news) channels.
Analysis of advertising market regulations (UK), 2009-10, consulting expert - Advised project team on analysis of demand for advertising for the purpose of evaluating changes in regulation of advertising minutes on public-service broadcasters in the United Kingdom. Designed econometric model and supervised implementation and description of results. Report submitted to media regulator (Ofcom).
Distribution of cable copyright royalties (US), 2009-10, testifying expert - Submitted rebuttal testimony to copyright royalty judges regarding relative market value of programming provided on the distant broadcast signals carried by U.S. cable systems. Testified before judge panel.
Blockbuster/Hollywood Video (US), 2005, consulting expert - Supported BatesWhite team to establish liability for FTC challenge of proposed merger.
Echostar/DirecTV (US), 2002-03, consulting expert - Supported analysis by AES Consulting (now Compass) of liability for proposed merger. Helped design econometric model of pay-television
Gregory Crawford cv, June 2011 2 FOR PUBLIC INSPECTION
demand and participated in conference calls with opposing lawyers and experts.
Advisory roles (US): AMD/Intel, 2009; DRAM Litigation, 2009; German media market, 2007; AT&T/BellSouth, 2006; Auto-finance merger, 2005; Death-care industry merger, 2005; Vitamins price-fixing litigation, 1999-2001
Bates White LLC, Academic Affiliate, 2005-present
Publications
"The Welfare Effects of Bundling in Multi-channel Television Markets," (with Ali Yurukoglu), University of Warwick, April 2011, forthcoming, American Economic Review.
"Cable Regulation in the Satellite Era," Chapter 5 in Rose, N., ed, "Economic Regulation and Its Reform: What Have We Learned?", forthcoming, University of Chicago Press.
"Economics at the FCC: 2007-2008," (with Evan Kwerel and Jonathan Levy), Review of Industrial Organization, v33n3 (November 2008), 187-210.
"The Discriminatory Incentives to Bundle: The Case of Cable Television," Quantitative Marketing and Economics, v6n1 (March 2008), 41-78. - Winner, 2009 Dick Wittink Prize for the best paper published in the QME
"Bidding Asymmetries in Multi-Unit Auctions: Implications of Bid Function Equilibria in the British Spot Market for Electricity, (with Joseph Crespo and Helen Tauchen), International Journal of Industrial Organization, v25n6 (December 2007), 1233-1268.
"Bundling, Product Choice, and Efficiency: Should Cable Television Networks Be Offered A La Carte?," (with Joseph Cullen), Information Economics and Policy, v19n3-4 (October 2007), 379-404.
"Monopoly Quality Degradation and Regulation in Cable Television," (with Matthew Shum), Journal of Law and Economics, v50n1 (February 2007), 181-209.
"Uncertainty and Learning in Pharmaceutical Demand," (with Matthew Shum), Econometrica, v73n4 (July 2005), 1137-1174.
"Recent Advances in Structural Econometric Modeling: Dynamics, Product Positioning, and Entry," (with J.-P. Dube, K. Sudhir, A. Ching, M. Draganska, J. Fox, W. Hartmann, G. Hitsch, B. Viard, M. Villas-Boas, and N. Vilcassim), Marketing Letters, v16n2 (July 2005).
"The Impact of the 1992 Cable Act on Household Demand and Welfare," RAND Journal of Economics, v31n3 (Autumn 2000), 422-449.
Reports
Gregory Crawford cv, June 2011 3 FOR PUBLIC INSPECTION
"Empirical analysis of BT's automatically renewable contracts," (with ESMT Competition Analysis, Commissioned Research Study for the Office of Communications), August 2010. Also Supplementary Report, February 2011.
"Television Station Ownership Structure and the Quantity and Quality of TV Programming," (Commissioned Research Study for the Federal Communications Commission), July 2007.
Work in Progress
Working Papers
"The Empirical Consequences of Advertising Content in the Hungarian Mobile Phone Market," (with Jozsef Molnar), University of Arizona, March, 2008.
"Estimating Price Elasticities in Differentiated Product Demand Models with Endogenous Characteristics," (with Dan Ackerberg), mimeo, University of Arizona, March 2007.
"The Welfare Effects of Endogenous Quality Choice: The Case of Cable Television," (with Matthew Shum), mimeo, University of Arizona, March, 2006
"A Virtual Stakes Approach to Measuring Competition in Product Markets," (with R. Michael Black, Shihua Lu, and Hal White), mimeo, University of Arizona, May 2004.
Work In Progress
"Robust Instrumental Variables," (with Dan Ackerberg), mimeo, UCLA, March 2007.
"An Empirical Analysis of Manufacturer-Retailer Interaction: What Determines Wholesale Prices?" (with Zsolt Macskasi), May 2006.
"Storability, Competition, and Sales: Do Firms Cut Prices to Steal Demand from Rivals or Themselves?," (with James J. Anton), April 2005.
"A Dynamic Model of Quality Competition in Subscription Television Markets," (with Alex Shcherbakov), March 2007.
"The Impact of Ratings and Word-of-Mouth on DVD Rentals: An Analysis of the Netflix Data," (with Ivan Maryanchyk), February 2007.
Grants
"Endogenous Product Characteristics in Empirical Industrial Organization," Economic and Social Research Council, £140,000 C$220,000), 2010-2012.
"The Empirical Consequences of Advertising Content" (with Jozsef Molnar), Hungarian
Gregory Crawford cv, June 2011 4 FOR PUBLIC INSPECTION
Competition Commission, 10,000,000 Hungarian Forint C$50,000), 2007-2008
Other Professional Activities
Associate Editor, International Journal of Industrial Organization, October 2005 - present.
Editorial Board, Information Economics and Policy, December 2007 - present.
Referee for Econometrica, American Economic Review, Review of Economics Studies, RAND Journal of Economics, Review of Economics and Statistics, Quantitative Marketing and Economics, National Science Foundation, International Journal of Industrial Organization, Journal of Industrial Economics, Journal of Applied Econometrics, Information Economics and Policy, Management Science, Southern Economic Journal
2010 Presentations: LBS (1/10), UCL (4/10), Oxford (5/10), Invitational Choice Conference (5/10), Manchester University (9/10), EIEF (Rome, 10/10), University of Venice (10/10), University College Dublin (11/10). 2009 Presentations: ESMT, Berlin (5/09), CEPR 10, Mannheim (5/09), University of Leuven (9/09), University of Toulouse (Econometrics Workshop and Competition Policy Workshop), (11/09) 2008 Presentations: UK Competition Commission (1/08), Oxford University (1/08), University of Warwick (1/08), University of Virginia (3/08), Industrial Organization Society (5/08), NBER Summer Institute, 10 Group (6/08), 6th Workshop in Media Economics, Zurich (10/08), Network of Industrial Economics, London (12/08) 2007 Presentations: University of Pennsylvania (Wharton, 3/07), ESMT (Berlin, 4/07), Northwestern University (5/07), Bates White Antitrust/Merger Conference (6/07), University of Wisconsin, Madison (10/07), Duke University (Fuqua, 11/07) 2006 Presentations: AEA Meetings, Boston (1/06), Columbia (3/06), University of Chicago Marketing (3/06), Bates White Antitrust/Merger Conference (6/06), EARlE Amsterdam (8/06) 2005 Presentations: NBER Conferences on Regulation (2/05, 6/05), Econometric Society World Congress, London (8/05) 2004 Presentations: Stanford University (3/04), CEPR "The Role of Competition in the New Economy", Greece (5/04), Invitational Choice Conference (6/04), FCC Symposium on 'A La Carte" MVPD Pricing (7/04)
Conference Organization: Triangle Applied Micro Conference, April 2000, Triangle Applied Micro Conference, May 1999 (co-organizer)
Gregory Crawford cv, June 2011 5 FOR PUBLIC INSPECTION
APPENDIXB FOR PUBLIC INSPECTION
Appendix B List of (46) Television Networks Included as Possible News and Public Affairs Networks (TMS Channel Names)
1. National News Networks (4)*2 Cable News Network Fox News Channel HLN (Formerly Headline News) MSNBC
2. Business News Networks (4)* CNBC Bloomberg Business Television CNBC World Fox Business
3. National Public Affairs Networks (3)* CSPAN CSPAN2 CSPAN3
4. Local News Networks (7)* 10 News 2 Chicagoland Television News Local News (9 News Colorado) New England Cable News News 12 New Jersey News Channel 3 Anytime News Channel 5 Newschannel 207 (KMGH News) Northwest Cable News Pittsburgh Cable News Channel Virginia News Channel
2 All groups are included in possible news channels totals for a given headend. Starred (*) network groups (1)-(5) + (8) are included in the definition of news neighborhoods. FOR PUBLIC INSPECTION
News Channel 8 Ohio News Network News 12 Westchester (WEST12) SNNNews 6 Texas Cable News News First
5. Local, State, or Regional Public Affairs Networks (5)* CTN Connecticut Public Affairs California Channel Pennsylvania Cable Network New York State Legislative Channel TV WASHINGTON
6. HD News Networks (6) CNBCHD CNNHD Fox Business HD Fox News Channel HD HLNHD MSNBCHD
7. Spanish-Language News Networks (5) CABLENOTICIAS CNN En Espanol Latin America CNN En Espanol- Mexico and US CNN en Espanola United States CNNEn Espanol - All Feeds
8. Other News Networks Included in the Definition of News Neighborhoods (1)* CNN International
9. Other News Networks Excluded from the Definition ofNews Neighborhoods (1) CNN On Demand FOR PUBLIC INSPECTION
APPENDIXC FOR PUBLIC INSPECTION
REDACTED PURSUANT TO REQUEST FOR CONFIDENTIAL TREATMENT. FOR PUBLIC INSPECTION
EXHIBIT G