Competing with Free: The Impact of Movie Broadcasts on DVD Sales and Internet Piracy

Author(s): Michael D. Smith and Rahul Telang

Source: MIS Quarterly , Jun., 2009, Vol. 33, No. 2 (Jun., 2009), pp. 321-338
Published by: Management Information Systems Research Center, University of
Minnesota

Stable URL: https://www.jstor.org/stable/20650294

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Management Information Systems Research Center, University of Minnesota is collaborating

with JSTOR to digitize, preserve and extend access to MIS Quarterly

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Research Article

C^afterly
Competing with Free: The Impact
of Movie Broadcasts on DVD
Sales and Internet Piracy1

By: Michael D. Smith distribution in one channel on sales through paid channels,
H. John Heinz III School of Public Policy and and the impact of illegitimate free distribution in pirated
Management channels on sales through paid channels. We do this by
Carnegie Mellon University studying the impact ofmovie broadcasts on D VD demand and
Pittsburgh, PA 15213 the impact of piracy availability at the time of broadcast on
U.S.A. DVD demand. Our data include all movies shown on over
mds@andrew.cmu.edu the-air and cable television during an eight-month period in
2005-2006.
Rahul Telang
H. John Heinz III School of Public Policy and With respect to the impact of movie broadcasts on piracy and
Management sales, we find that movie broadcasts on over-the-air networks
Carnegie Mellon University result in a significant increase in both DVD sales at Amazon,
Pittsburgh, PA 15213 com and illegal downloads for those movies that are available
U.S.A. on BitTorrent at the time of broadcast. With respect to the
rtelang@andrew.cmu.edu impact of piracy on sales, we use the television broadcast as
an exogenous demand shock andfind that the availability of
pirated content at the time of broadcast has no effect on post
broadcast DVD sales gains.
Abstract
Together our results suggest that creative artists can use
The creative industries have frequently expressed concern product differentiation and market segmentation strategies to
that they can't compete with freely available copies of their compete with freely available copies of their content. Speci
content. Competing with free is particularly concerning for fically, the post-broadcast increase in DVD sales suggests
movie studios, whose content may be more prone to single that giving away content in one channel can stimulate sales
use consumption than other industries such as music. This in a paid channel if the free content is sufficiently differen
issue has gained renewed importance recently with the advent tiated from its paid counterpart. Likewise, our finding that
of new digital video recording and distribution technologies, the presence ofpirated content does not cannibalize sales for
and the widespread availability of Internet piracy. the movies in our sample suggests that if free and paid pro
ducts appeal to separate customer segments, the presence of
We examine competition between 'free " and paid video con free products need not harm paid sales.
tent in two important contexts: the impact of legitimate free
Keywords: Information goods, movie broadcasts, movie pro
motion, DVD sales, movie piracy, broadcast flag, consumer
1 Chris Kemerer was the accepting senior editor for this paper. surplus

MIS Quarterly Vol. 33 No. 2, pp. 321-338/June 2009 321

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Teiang/Competing with Free

Introduction free, unencrypted high definition television broadcasts will

harm the marketability of the studios' content. For example,
"We can't compete with free. That's an economic in testimony before the Federal Communication Commission
paradigm that doesn't work." regarding the need for federally mandated broadcast flag
James Gianopulos, Co-chairman, Twentieth content protection in high definition broadcasts, Viacom made
Century Fox Filmed Entertainment (quoted in the following statement:
Thompson 2003)
Viacom believes that [digital television] sales and
As noted in the above quote, members of the creative indus broadband subscriptions have reached the "tipping
tries have long expressed the belief that they are unable to point" at which it can no longer afford to expose its
compete with "free" copies of their content made available content to piracy. A broadcast flag regime is needed
through new information technologies. Their argument is now to protect the value of our important assets or
intuitive: Once a consumer is able to consume and potentially we must withhold our quality HD digital content
retain a copy of free content, why would they consider pur [from over-the-air broadcasts].
chasing that content? Viacom comments before the Federal Commu
nications Commission in the matter of Digital
Sales cannibalization from free distribution may be parti Broadcast Copy Protection, December 6, 2002
cularly salient in the movie industry for two reasons. First, (in Lucey 2002, p. 8)
movie content may be more prone to single-use consumption
than other intellectual property categories such as music or These concerns are driven by two main factors. First, that the
software. Second, movie studios are particularly reliant on ability of consumers to easily record, edit, and retain digital
revenue from media sales: Media sales (primarily DVD sales) television broadcasts will reduce demand for paid content.
made up 46 percent ($14.9 billion) of total movie revenue in And second, that the ability of (disreputable) consumers to
2002 (Epstein 2005, p. 20; PBS 2005), a little over twice that post high quality copies of movies shown on television will
of theater revenue, and margins on these media sales are increase the supply of pirated content and reduce demand for
higher than margins in many of the studios' other lines of legitimate media sales.
business.2
At the same time, in the face of these concerns it is possible
With these issues in mind, the goal of this research is to to see "competing with free" as a special case of price compe
analyze the impact of free distribution of movies on paid con tition. In this context, the academic literature has shown that,
sumption in two important contexts. First, the impact of free
in spite of initial concerns of fierce price competition in
television broadcasts of movies on consumer demand for
Internet markets, some Internet retailers are able to maintain
DVDs. Second, the impact of piracy availability at the time both high market share and high margins through product and
of broadcast on post-broadcast DVD demand. These two
service differentiation and customer segmentation (e.g.,
empirical questions highlight two important areas of competi
Brynjolfsson and Smith 2000; Smith and Brynjolfsson 2001).
tion between free and paid content: the impact of legitimate
free distribution in one channel on demand in a paid channel,
Thus, as is outlined in more detail below, it is unclear from a
and the impact of illegitimate "free" pirated distribution on
theoretical perspective what impact these two types of free
demand in a paid channel.
goods might have on subsequent demand through legitimate
channels. Because of this, we address these questions empi
These questions have also become salient from a business and
rically by gathering a new data set including all movies shown
public policy perspective with the development of new tech
on over-the-air television networks and the four most popular
nologies such as digital video recorders, high definition
advertising supported cable networks (hereafter ad-cable)
digital television (HDTV), high bandwidth Internet access,
from July 12,2005, to March 3,2006. For each movie in our
and a proliferation of tools facilitating Internet piracy.
sample, we collect data on its sales level at Amazon.com and
Specifically, with the development of new HDTV standards
piracy levels at two prominent BitTorrent tracker sites.
and the prevalence of piracy on the Internet, studios have
expressed concern that consumers' ability to make copies of
Our results show that, contrary to fears about competing with
free content, neither type of free content analyzed in this
study seems to reduce demand for paid content. In the case
2For example, according to a studio executive we spoke to, studios currently of free movie broadcasts on television, we find that the broad
pay only 20% of DVD revenues to the various artist and production unions,
keeping the remaining 80%. cast acts as a strong, short-term stimulus to demand for

322 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

DVDs. In our sample, over-the-air movie broadcasts result in net effect of television broadcasts is to increase media sales,
an increase in DVD sales at Amazon.com by an average of and that the presence of pirated content does not reduce post
118 percent during the first week after broadcast. broadcast sales of movies shown on television.

With respect to the impact of free pirated content, studies of For academics, our research presents a new empirical strategy
the impact of piracy face the endogeneity concern that for tracking piracy levels on the BitTorrent network and a
(unobserved) popularity influences both sales (left-hand-side new strategy for analyzing the impact of piracy on media
variable) and piracy levels (right-hand-side variable). In this sales. Specifically, in settings where the decision to promote
study, we attempt to address this endogeneity concern by or distribute a product (through broadcast in our case) is
using the promotional stimulus from the movie broadcast as uncorrelated with the availability of the product on pirate
an exogenous shock, and comparing the post-broadcast pro networks, the promotional stimulus can be used as a natural
motional gain for movies that have pirated versions readily experiment to compare the response of products with and
available on BitTorrent networks at the time of broadcast and without pirated copies available.
those that do not. If piracy is harming sales for these movies,
movies that have pirated copies readily available on The remainder of this paper proceeds as follows. In the next
BitTorrent will exhibit a smaller post-broadcast promotional section, we review the relevant literature pertaining to the
stimulus than those that do not because some consumers who impact of broadcasts and piracy on product sales, and on the
would otherwise have purchased a DVD will (illegally) effectiveness of product differentiation and market segmen
download the free BitTorrent version instead. However, we tation strategies in Internet markets. In the third section, we
observe in our data that movies that have pirated copies present our main empirical tests and briefly discuss the
readily available on BitTorrent networks at the time of broad theoretical basis for each test. We then discuss our data and
cast have statistically the same increase in DVD sales as those present our empirical models and results. Finally, we discuss
that do not. the implication of our findings, limitation of our analysis, and
areas for future research.
For movie studios, our results suggest that competing with
free is possible through product differentiation and customer
segmentation. With regard to differentiation, our results
suggest that the television broadcast of a movie is sufficiently Literature HHHH
differentiated from the DVD version (in terms of conveni
ence, usability, and content) that, not only does it not appear Our work most closely pertains to the lit
to cannibalize sales, it has a net promotional effect on sales? of piracy in markets for information go
even though nearly the entire copy of the movie is shown on in this area has focused on software
television and even though movies are thought to be single particularly on peer-to-peer file sharin
use consumption products. With respect to segmentation, our impact on firm profitability. A prominen
results suggest that, at the time of broadcast, pirates and literature has been to show that piracy
purchasers represent two different market segments. The firms. Prasad and Mahajan (2003) argue
movie broadcast stimulates demand for DVDs and demand for good for a new product if the firm needs
piracy. However, the presence of pirated content does not user base to speed up diffusion. Gu
cannibalize DVD sales at the point of time a movie is shown show that because piracy removes the m
on television. This is conceptually similar to well-known buyers from the market it can reduce p
examples of price discrimination where a lower priced benefiting sellers. Finally, Peitz and Wa
product (in this case a free pirated product) need not canni that piracy can act as a free "sample,
balize sales from higher priced products if the two products awareness.
appeal to different customer segments. The difference, in this
case, is that rights holders have only limited control over the The empirical work on piracy has focused on estimating the
availability and "price" of pirated content as compared to impact of piracy on demand for legitimate content. The
price setting and product differentiation strategies available to majority of this literature has focused on the music industry,
firms in more traditional settings. addressing three related sets of empirical questions. The first
question is the degree to which the emergence of peer-to-peer
For policy-makers, we find no evidence to indicate an imme file sharing in 1999 can explain the steady decline in record
diate need for "broadcast flag" style copy protection of movie sales from 1999 to 2003. In addressing this question, Liebo
broadcasts. In contrast, our results suggest that at present the witz (2008) finds that increased Internet penetration can

MIS Quarterly Vol. 33 No. 2/'June 2009 323

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Telang/Competing with Free

explain the significant reduction in album sales from 1999 to supply of piracy, Byers et al. (2003) show that the majority of
2003, while Hong (2004) and Peitz and Waelbroeck (2004) movies available on file sharing networks originate from
find that approximately 20 percent of the decline in record studio leaks, as opposed to copies from DVDs or other post
sales can be explained by piracy using data from 2000 and market sources.
1998 to 2002 respectively.
From the perspective of empirical methods, our analysis
The second, and related, major question addressed in the relates to the growing empirical literature using Amazon's
literature is the degree to which the consumption of pirated sales rank data to estimate the company's product-level sales.
content displaces sales of legitimate content. Here, estimates While Amazon.com does not provide product-level sales
range from 42 percent displacement in an international sample information for its products, the company does provide
from 1994 to 1998 (Hui and Png 2003), to 33 percent dis information about the sales ranking of products within a
placement among U.S. sales in 2003 (Blackburn 2007), to 30 particular product category. Researchers have used this sales
percent displacement among 15,000 European consumers in rank data to estimate Amazon's sales through direct empirical
2001 (Zentner 2006), to 20 percent displacement among a estimation (Brynjolfsson et al. 2003) and experimental cali
sample of University of Pennsylvania students (Rob and bration (Chevalier and Goolsbee 2003). Subsequent papers
Waldfogel 2006), to finally no displacement among U.S. in the literature have used Chevalier and Goolsbee's experi
downloaders in late 2002 (Oberholzer and Strumpf 2007). mental calibration technique in a variety of contexts (e.g.,
Chevalier and Mayzlin 2004; Ghose et al. 2006; Ghose and
The third major question addressed in the literature is the Sundararajan 2005; Smith and Telang 2004).
degree to which harm from piracy affects popular and less
popular artists. Here Blackburn (2007) finds that piracy has Finally, we note that the impact of piracy on product markets
a stronger impact on popular artists while Bhattacharjee et al. is conceptually similar to the impact of used goods markets on
(2007) and Rob and Waldfogel (2006) seem to find the new product sales (Ghose et al. 2006), the impact of increased
opposite effect: that less popular CDs face higher piracy TV and radio penetration on the movie and music industries
risks. Thus, each of the papers in the literature?with one (Liebowitz 2004), competition between traditional print
notable exception?has found some level of harm from music copies of books and PDF copies of books (Kannan and Jain
piracy in the late 1990s and early 2000s, but there is a fair 2002), consumers' decisions to rent or purchase movies
amount of disagreement as to the degree of harm from piracy. (Knox and Eliashberg 2005), and international movie release
windows (Elberse and Eliashberg 2003).
However, while there is now a great deal of literature in the
context of music piracy, we are aware of only two papers to
address movie piracy. First, Rob and Waldfogel (2007) use
survey data from 500 University of Pennsylvania under Theoretical Framework
graduates and find that piracy displaces paid consumption by
nearly 100 percent on the first viewing and 20 percent on the In this section, we outline the main empirical questions ad
second viewing. Second, Smith and Telang (2007) find that dressed in this paper and discuss the theoretical rationale
increases in broadband Internet penetration from 2000 to 2003 underlying each question.
led to a $ 1.3 billion increase in DVD sales. Moreover, it may
be particularly important to analyze the impact of video piracy
separately from music piracy because of differences in size, The Impact of Movie Broadcasts
download speed, digital rights protection, and consumption on DVD Sales
patterns between the two types of content.
On one hand, it is possible that the dominant impact of
Another stream of the literature analyzes piracy from a policy "giving away" a movie through an unencrypted, freely avail
perspective. In this context, Gopal and Sanders (1998) show able medium such as broadcast television would suppress
that government enforcement of intellectual property rights DVD sales. In this view, consumers who would have other
depends on the robustness of the domestic software industry. wise purchased the movie on DVD would be less inclined to
With respect to copyright policy, Png and Wang (2006) show do so if they could instead watch and retain copies of movies
that copyright extensions enacted by OECD countries from shown on free television. The movie studios first raised this
1991 to 2002 were associated with an increase in movie argument in 1982 as part of the development of the first
production?and that this increase was stronger in countries analog videocassette recorders. At that time, the movie
where piracy was lower. Finally, from the perspective of the industry argued before the United States Congress and

324 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Supreme Court that home recording of television programs The Impact of Pirated Content
infringed the studios' copyright and that manufacturers of on After-Broadcast DVD Sales
home video equipment should be held liable of all resulting
instances of copyright infringement. This argument was most If movie broadcasts serve to stimulate DVD sales, will this
famously advanced by Jack Valenti's statement before sales stimulus be lower for movies that have pirated content
Congress that "the VCR is to the American film producer and available at the time of broadcast? This is a critical empirical
the American public as the Boston strangler is to the woman question for movie studios looking to protect their valuable
home alone."3 As noted above, studios have again raised this content.
concern in the context of HDTV broadcasts and digital video
recorders, noting that it is easier for consumers to retain, edit, On one hand, the majority of the empirical literature has
and share digital broadcasts than analog broadcasts and that, shown that, at least in the context of music and software, the
unlike analog broadcasts, digital storage, editing, and sharing availability of pirated content reduces, at least somewhat,
can occur without loss of signal quality. legitimate demand. It would be natural to expect that the
same rationale would carryover to movies as well.
On the other hand, it is possible that television broadcasts of
movies could have no effect on DVD sales, or even stimulate However, it is also possible that the availability of pirated
sales. The "no effect" view is consistent with Liebowitz content for movies has a negligible effect on legitimate
(1985), who concluded that there was no detrimental impact consumption. This view is consistent with the notion that
of the VCR on TV content providers. In the "stimulate sales" pirated content for music and software is a much stronger
view, the television broadcast would serve as advertising for substitute for paid content than pirated copies of movies.
the movie, allowing consumers who otherwise would not have Pirated music and software have nearly the same quality and
purchased the DVD to become aware of (or reacquainted usability as the legitimate content. In the case of music,
with) its content. These consumers might decide to purchase pirated content may have even higher usability as, unlike
the DVD even after seeing the movie on television because many legitimate digital downloads, pirated music does not
DVDs offer more information (e.g., deleted scenes, director's contain restrictions associated with Digital Rights Manage
commentary), higher convenience (e.g., no commercials, easy ment, and unlike CDs, pirated music does not require a
portability), and higher video quality than television broad separate step to be played on portable digital music players.
casts do. A similar idea was espoused, though not empirically
tested, by Liebowitz (1985). In contrast, pirated movies frequently have significantly lower
quality than legitimate media due to the compression neces
We also note that these effects might exist side-by-side, with
sary to facilitate easy Internet downloads. Likewise, from a

some consumers deciding not to purchase DVDs because they usability standpoint, it is harder to play pirated movies on
can view and retain the television broadcasts, and some most home theater systems than it would be to play a
consumers deciding to purchase the DVD on the basis of legitimate DVD.
seeing the broadcast. While our data do not allow us to
separately identify these two effects, we are able to identify
The impact of piracy on legitimate demand also critically
the net effect of the television broadcast on DVD sales. depends on how loosely (or tightly) coupled the user seg
ments in these markets are. On one hand, it is possible that
these segments are tightly coupled and that a significant
number of users would forgo the purchase of a DVD if pirated
content were available. On the other hand, these segments
might be loosely coupled such that potential DVD buyers
would not consider the availability of pirated content in their
purchase decision, and potential pirates would not consider
purchasing the DVD if pirated content were not available.

Thus while the economic theory underlying our empirical

analysis is well established, the actual effects critically
depend on the market structure and user choices associated
3"Hearings Before the Subcommittee on Courts, Civil Liberties, and the with the provision of free media products. Therefore, we
Administration of Justice of the Committee on the Judiciary House of
believe that these issues are inherently empirical and in the
Representatives, Ninety-Seventh Congress, Second Session, on Home
Recording of Copyrighted Works," 1982, Serial No. 97, Part I, U.S.
next two sections we outline the data gathered to address
Government Printing Office, 15-1680, Washington, D.C. these questions and our empirical results based on this data.

MIS Quarterly Vol. 33 No. 2/June 2009 325

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Telang/Competing with Free

Data additional control for "popularity" of the movie, we collected

data from Nielsen media research on television viewership for
We address these empirical questions using data collected each movie at the time of broadcast.
from July 12, 2005, through November 23, 2005,4 and from
January 1,2006, through March 3,2006. We have eliminatedWe also collected information about DVD characteristics and
all observations during the Christmas 2005 holiday season to sales rank for each version of the movie available at
avoid any potential counter-explanations that might occur Amazon.com. Many movies have separate wide screen and
during this time period (e.g., increased sales of DVD,full screen editions, and in some cases separate special or
reduction in piracy, systematic changes in preferences for unrated editions.5 For each of the DVD versions, we col
piracy versus legitimate purchases owing to holiday giftlected product characteristics including list price, release date,
purchases). MPAA rating, aspect ratio, number of discs, and sound
quality (e.g., Stereo, Dolby Surround, Dolby THX). We also
Our data consist of information on all movies shown on over collected Amazon marketplace information including the
the-air broadcast channels and major advertising supported Amazon price, the Amazon users' star rating of the movie,
cable channels. With respect to over-the-air channels, weand the movie's sales rank. We collected this information
gathered data on all movies shown in national broadcasts on hourly for two weeks before and after the movie was broad
the major broadcast networks during our sample: ABC, CBS,cast, and daily thereafter. We do not include observations
NBC, FOX, UPN, and WB. We used only national broad that occur after the second showing of a movie in our data or
casts as a partial control for audience size as local affiliates for movies that had shown on television during the six-month
have the option of slotting movies that will only be shown inperiod prior to July 2005. This allows us to focus our
a local region. We determined that a broadcast was national attention on the sales gain from the first showing.
if it was shown in both the New York City and Los Angeles
affiliates during the same time slot. Finally, we eliminated any movies in our sample that had
sequels that appeared in movie theaters or were released on
We also collected data from the four most popular DVD during the period of March 2005 through September
advertising-supported cable networks (hereafter ad-cable): 2006 (i.e., six months before and after our data collection
TBS, TNT, USA, and Lifetime. We selected these four netperiod). This is done to control for endogenous promotional
works based on Nielsen Media Research viewership estimates effects associated with theatrical and DVD release dates. Our
(as reported by TelevisionWeek magazine) for the six-monthfinal sample contains 522 broadcast movies and 759 DVD
period from March to August 2005. The four most popular titles. The summary statistics for our sales data are shown in
Table
channels were the same whether we considered total daily l.6
viewers or prime time viewers.
We use Amazon's DVD sales rank as a proxy for the number
We collected data for each of these movie broadcasts from
of products sold at Amazon. Amazon.com lists the rank of
three primary sources. We collected broadcast information?products sold in each product category, with 1 corresponding
broadcast date and time, broadcast duration, movie name andto the highest selling product, 2 to the second highest selling
description, and whether the movie was shown in high definiproduct, and so on. Following Brynjolfsson et al. (2003) and
tion format?from TitanTV.com. We used TitanTV becauseChevalier and Goolsbee (2003), we assume that the rela
it is easily searchable and provided a 14-day advance notice tionship between sales and sales rank follows a Pareto
before a movie's broadcast date. This advance notice allowed distribution:7
us to obtain a baseline level of sales and piracy before
broadcast. We used the Internet Movie Database (imdb.com) Quantity = a Rantf (i)
to obtain information on the theatrical release date, rental
revenue, gross revenue, gross budget, and IMDB user star
rating for each of the movies in our study. Finally, as an
5We did not collect data on box sets that contain multiple different movies,
even if one of the movies in the box set was present in our sample.

November 23,2005, was the day before Thanksgiving, the traditional start 6This table includes all data from 14 days before broadcast date through 28
of the Christmas shopping season. Our results are not sensitive to this choice days after the broadcast date consistent with our regressions below.
of dates. For example, a more conservative approach of eliminating all
observations from November 1, 2005, to January 1, 2006, would result in a 7This technique has also been applied in a variety of other studies, including
slight (and statistically insignificant) increase in the post-broadcast sales Chevalier and Mayzlin (2004), Ghose et al. (2006), and Ghose and
gains reported here. Sundararajan (2005).

326 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Table 1. Sales Data Summary Statistics

Variable Obs. Mean St. Dev. Min
Amazon Rank 39,013 12,853 13,100
Amazon Price 39,013 12.95 4.06 4.98
IMDB Star Rating 39,013 5.95 1.24 1.80
Broadcast Duration (Hours) 39,013 2.21 0.41 1.35
Broadcast Network 39,013 0.18 0.38
Number of Discs in DVD 39,013 1.10 0.36
Ln(Gross Revenue ($ Million)) 31,974 17.36 1.32 10.53
IMDB User Votes 39,013 12,103 19,987
No of Viewers (in millions) 522 2.04 1.97 0.28
Minutes Edited from Broadcast 39,013 9.90 10.50

This relationship can be parameterized using either direct (i.e., leechers), and the number of cumulative downloads.
observation of sales and ranks for select titles, typically ob This characteristic makes BitTorrent tracker sites particularly
tained from product suppliers (see Brynjolfsson et al. 2003) useful for empirical analysis of piracy levels, and we believe
or by means of an experiment (see Chevalier and Goolsbee that the use of BitTorrent tracker sites in this way represents
2003). an additional contribution of this paper to the literature.

Lacking supplier data, we apply Chevalier and Goolsbee's Among BitTorrent trackers, we selected Piratebay and Mini
experimental technique by finding two DVDs with high rank nova as data sources because they were among the most
(low sales) and observing their rank over the course of several popular BitTorrent tracker sites during our study period,9 and
weeks to estimate the number of daily sales. We then ordered these sites also listed the current number of seeds, leechers,
seven copies of the DVDs in an hour, observing the initial and and downloads for each of their trackers at the time of our
final rank. This allows us to obtain two points on the curve, study (Figure 1 shows a sample screen from mininova.org).
which we can then use to determine the slope (?) of equation
(1) in log-log space. We performed this experiment on July For each of the movies in our data set, we use an automated
1 and July 8, 2004, for two separate DVDs and found the ? script to search for movie torrents matching the movie title.
parameter for equation (1) was equal to -1.61 in both cases. We collect this data daily starting before the movie's broad
We performed this experiment again on February 8,2006, for cast date and continuing after the broadcast date. This allows
two additional DVDs, and found ? parameters of-1.76 and us to track both (1) any activity on torrents that existed before
-1.81 respectively. We use the average of the four ? esti the broadcast and (2) any new torrents that might be added
mates (-1.70) in our subsequent calculations.8 after the broadcast date. For all trackers that match the movie
title and general description, we collect the date the tracker
Our piracy data come from piratebay.org and mininova.org, was added to the respective tracker site, the file size, and daily
two public tracker sites for the BitTorrent protocol. We observations of the number of seeds, leechers, and cumulative
selected BitTorrent as a proxy for piracy levels for two downloads.
reasons. First, BitTorrent is currently the most popular
protocol for sharing large files, such as movie files (which Our final data set covers the period of October 28, 2005,
typically range from one to six gigabytes in size for content through March 3,2006. As above, we exclude the Christmas
sourced from DVDs). Second, the design of the BitTorrent holiday period (November 23,2005, through January 1,2006)
protocol is such that all nodes participating in a file download to avoid the possibility that piracy levels are systematically
report their status to the tracker every 20 seconds. Thus, different during this time period. The summary statistics for
tracker sites such as Piratebay and Mininova can report in our piracy data are shown in Table 2.10
near real-time the number of users providing the entire file
(i.e., seeds), the number of users actively downloading the file 9
For example, Gil (2006) lists both Piratebay and Mininova among the five
most popular BitTorrent tracker sites.

Our results would be qualitatively the same if we used either the July 2004 10Our summary statistics only include observations from 14 days before
or February 2006 coefficients. broadcast to 7 days after broadcast, consistent with our regressions below.

MIS Quarterly Vol. 33 No. 2/June 2009 327

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Telang/Competing with Free

^t* H h?P /***** m.mnova org/to<7??l$9* _ * * ?

OmlOKQIMJWl) Stiren Staren ?stnci wt? ita/th

Harn? Brawn* cat?gorte$ ? Upload a tornwu Atfvawiart starch COfmmyjifcy * Searrf? doud

Location; Mininova > >Cataaartaft

AOlfiO > Th> ft?iif?? IdtrWy 2002 DVDft*D AC3 XviD

[ General information ||.j>it*^

The Bourne Identity 2002 DVDRip AC3 XvtO

iNT-PFa[www.torTentsforall.net]
Download thi? torrent I or us* i

To ?art tru? ?2P downJodd, yoy have to t^stati a 8<? Torrent ctoern Kk* ?jTorr?nt or W

Category: Stet** > Mi? Shar? ratio:

Total siz? 2 38 exabyte 2 liours ago
4 days a<go Downloads: 391
Subtitle*:

Do**?

Figure 1. Mininova.org Screen

Table 2. Piracy Data Summary Statistics

Variable Obs. Mean St. Dev. Min
Broadcast Network 22,798 0.44 0.49
Leechers 22,798 11.61 29.50
Seeds 22,798 4.33 13.98
Daily Downloads 21,826 4.27 13.81

Results time the broadcast started on the East Coast of the United
States. Likewise, D(l) equals 1 for the first week after the
start of the broadcast in Eastern time zone.
The Impact of Movie Broadcasts
on DVD Sales
We then estimate a model with DVD-level fixed effects to
examine how sales change after a movie is broadcast on over
To estimate the effect of movie broadcasts on DVD sales, we
the-air or cable television. A fixed effect model ensures that
create a set of weekly time dummy variables that control for
the sales levels before and after the broadcast. For notational changes in sales are captured within DVDs. The fixed effect
model we estimate is
simplicity, the dummy variable D(x) will be equal to one for
x weeks before or after the broadcast. Thus, D(-l) equals 1
for the time period from one week before broadcast until the Ln(Rankit) = ? Priceit + ?Dt + et (2)

328 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Table 3. Impact of Movie Broadcasts on Sales Rank (Fixed Effects)

Independent Variables Broadcast Ad-Cable
D(-1) 0.000 (0.3) 0.000 (0.3)
D(1) -0.008** (14.1) -0.012** (20.2)
D(2) -0.005** (7.8) -0.002** (2.0)
D(3) -0.002** (3.4) 0.000 (0.3)
D(4) -0.000 (0.8) 0.003** (3.5)
Amazon Price 0.129** (11.3) 0.090** (20.4)
Constant 8.649** (0.282) 7.825** (38.5)
Number of Observations 3,063 14,551
Number of of Groups 93 678
The dependent variable is ln(sales rank). T-statistics are listed in parenthesis; ** and * denote significance at 0.01 and 0.05, respectively. All
models use DVD-level fixed effects and AR(1) serial correlation correction. Results are normalized per 100,000 viewers.

where / indexes a movie and / indexes time. ? and ? are the vision, the DVD sales rank decreases (DVD sales increase)
vectors of coefficients to be estimated, where S captures the significantly for both movies shown on broadcast networks
effect of the movie broadcast on DVD sales, our variable of and movies shown on cable networks. Similarly, D(2) and
interest. Due to time series effects, we control for AR(1) D(3) are also negative and significant (except for D(3) in the
disturbances in the error term. case of ad-cable, which is insignificant), although the magni
tudes are decreasing with respect to D(l). Thus the sales in
As noted above, we start observing the sales rank for a DVD weeks two and three are also higher than pre-broadcast levels,
two weeks before its television broadcast. Using this data, the although they are not as high as in week one. Finally, the
left out variable in this regression is the average sales level estimate on D(4) is small, positive, and insignificant (except
two weeks before broadcast. Since the number of viewers in the case of ad-cable, where it is positive and significant).
differs significantly across movies and across channels, we Thus by week four, DVD sales reach approximately the same
interact our time dummy variable Dt with the number of level as they were two weeks prior to the broadcast. To focus
viewers reported by Nielsen media research (in units of on the event of interest, we do not include dummies beyond
100,000 viewers). This allows us to control for differences in week four, although the estimates on D(5) are economically
viewership and for movie popularity effects. Thus, the impact and statistically insignificant. Also note that, over time, DVD
of Dt should be interpreted as the change in rank in week / per sales show a declining trend. If we were to control for it (by
100,000 viewers. Our results for this regression are shown in including age of the DVD or by including a control group of
Table 3. movies that were not broadcast) our estimates on the weekly
dummies would be even stronger. In summary, broadcasting
movies on television?essentially giving away the content for
The crucial variables in our model are the dummy variables
free?provides a strong short-term stimulus to DVD sales.
on time. As noted above, the left out category is the time
period two weeks prior to broadcast. First note that D(-l) is
Based on these estimates, we can quantify the percentage
insignificant. Thus, in the week prior to the movie broadcast
increase in sales due to a movie broadcast. To do this, we
there is little change in the rank (sales) of DVDs relative to
interpret the values of the dummy variables in terms of overall
their sales two weeks before broadcast. This suggests that our
changes in sales. Recall that Ln(Sales) =? x Ln(Rank) where
results are not driven by consumers delaying their purchases
? = -1.70. From this, it is straightforward to show that the
until after the movie is broadcast or by responses to pre
percentage increase in sales resulting from a coefficient 8t is
broadcast promotion of the movie.11

Asales = ^3-1 (3)

Next note that after broadcast D(l) is negative and highly
significant. In the week after the movie is shown on tele
Recall that our estimates in Table 3 are normalized to be per
100,000 viewers. Since we know the viewership numbers for
1 ^ote that advertising for movie broadcasts typically occurs in the week each movie, we can calculate the percentage increase in
prior to broadcast. weekly sales due to movie broadcasts (Table 4).

MIS Quarterly Vol. 33 No. 2/June 2009 329

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Table 4. Percentage Increase in Weekly Sales Due to Movie Broadcast

Weeks Since Broadcast Broadcast Ad-Cable
DM) -1.4% 0.03%
D(1) 118.9%* 27.2%*
D(2) 55.8%* 3.2%*
D(3) 21.5%* 0.5%
D(4) 4.9% -5.7%*
Average Number of Viewers 5.6 million 1.2 million

denotes statistical significance at 0.01

From Table 4, we can see that movies shown on over-the-air increase in sales typically persists for three to four weeks
broadcast networks experience a 119 percent increase in DVD before returning to its baseline level. Thus, our findings show
sales in week one, a 56 percent increase in week two and a 22 that the sales promotion benefits of digital television
percent increase in week three. As noted above, by week broadcasts far outweigh any short-term cannibalization effect.
four, DVD sales return to levels that are statistically the same We next turn our attention to measuring the impact of
as the levels before the movie was broadcast. broadcasts on the supply of and demand for pirated content.

We also note that the weekly percentage sales increase for

movies shown on ad-cable networks is significantly lower The Impact of Pirated Content
than the percentage increase for broadcast networks. This is on After-Broadcast DVD Sales
because fewer viewers watch movies on ad-cable (viewership
on ad-cable is approximately one-fifth of over-the-air viewer In the second part of our analysis, we examine how the
ship for movies in our sample). Thus, while Table 3 shows presence of free pirated content at the time of broadcast
that the per viewer increase in sales is comparable for ad impacts DVD sales. To do this, we first analyze how tele
cable and broadcast networks, Table 4 shows that the aggre vision broadcasts impact the demand for pirated content on
gate increase in DVD sales is far higher for broadcast two prominent BitTorrent file-sharing networks at the time of
networks. broadcast. The models we estimate are of the form

To test whether the sales gains differed systematically across

{downloads ip leechersit, seeds J =
different movies' characteristics, we interacted the weekly
dummies with movie characteristics in a random effects
XDt+ ? tracker age + eit (4)
specification. We did not find strong evidence of an inter
action effect. However, interaction with box office revenues where our dependent variables include, separately, the daily
download rate, the number of leechers, and the number of
and the "star rating" given to movies by IMDB voters were
seeders for each movie tracker i on day t. Our independent
statistically (but not economically) significant and in the
variables include weekly time dummy variables for weeks
expected direction.12 We also tested whether sales changes
after broadcast as in the sales models, and the age of the
are different between high definition and standard broadcasts,
and did not find any difference between the two.
tracker measured in days since it was first posted on the
BitTorrent network. This controls for changes in the popu
In summary, our results show that both ad-cable and over-the
larity of individual tracker files over time. We include
weekly dummy variables through week five as the movie
air movies experience a large, statistically significant increase
downloads show a statistically significant increase through
in sales immediately following their broadcast, and that this
the fifth week after broadcast. In many cases, due to data
collection limitations, the tracker data was not available for a
12 full two weeks prior to the broadcast. Therefore, the omitted
To keep the paper within page limits, we do not show these results. In variable in this regression is the number of seeds, leechers,
Table Al of the Appendix, we show how change in sales differs across and downloads before broadcast. We also did not have
movies with different initial ranks, finding that percentage sales gains after
broadcast are statistically the same across high and low ranked DVDs. viewership data for all trackers and to avoid dropping some

330 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Table 5. Impact of Movie Broadcasts on Piracy (Fixed Effects) for Broadcast Channels

Independent Variables Daily Downloads Leechers Seeds

D(1) 5.32** (1.3) 4.03* (2.174) 2.38** (1.01)
D(2) 6.24** (1.45) 5.79** (2.45) 2.95** (1.15)
D(3) 8.72** (1.57) 2.86 (2.67) 1.68(1.26)
D(4) 7.35** (1.67) 3.31 (2.8) 217(1.33)
D(5) 7.01** (1.8) 2.30 (3.1) 1.81 (1.44)
Tracker Age -5.13* (2.72) -2.29 (5.3) -1.08 (2.56)
Constant 25.51** (4.53) 22.78** (7.4) 2.69** (1.25)
Number of Observations 3654 3866 3866
Number of of Groups 161 165 165
The dependent variable is daily downloads (column 1), number of leechers (column 2), and number of seeds (column 3). Standard errors are listed
in parenthesis; ** and * denote significance at 0.01 and 0.05, respectively. All models use tracker-level fixed effects.

data we do not interact viewership numbers with time It is important to note that these increases are driven by
dummies as done in the previous section. We first estimate demand-side effects as opposed to supply-side effects. To test
the impact of over-the-air broadcasts on piracy (Table 5). supply-side effects, we used our BitTorrent tracker data to
analyze the names and sizes of all trackers added in the month
Our results show a significant increase in piracy immediately after the movie's broadcast date and found no evidence that
after movies are broadcast on over-the-air channels. From our television broadcasts (whether digital or analog) serve as the
estimates, we quantify the magnitude of these changes in source material for pirated content in our sample. That is, the
Table 6, where the baseline levels for daily downloads, increase in downloads, seeds, and leechers is driven by
leechers, and seeders were 8.8, 15.7, and 5.4 per week per increased interest in the existing trackers for these movies
tracker respectively. (based on similar affects to those driving increased DVD
sales), as opposed to an increased supply of copies of the
The magnitude of these changes is nontrivial. For example, movies taken from the (unencrypted) over-the-air or cable
our results suggest that daily downloads increase by 60 to 100 broadcasts.
percent in the four weeks after broadcast. Similarly, seeds
and leechers increase by between 25 and 55 percent in the Given these empirical findings, we are now able to analyze
first two weeks after broadcast, with smaller (and statistically whether the availability of pirated content on prominent
insignificant) increases in weeks three and four. BitTorrent networks at the time of broadcast is associated
with smaller increases in DVD sales after broadcast than for
We ran the same piracy regressions as above on the movies movies where no BitTorrent tracker is available at the time of
shown on cable channels in our sample. Our results are broadcast. One might wonder if the availability of pirated
shown in Table 7. The cable results reveal a slight increase content at the time a movie is broadcast on television reduces
in downloads, which is statistically significant only in week the number of DVD purchases that otherwise would have
three. The regressions show no statistical change in the levels occurred. In short, does movie piracy adversely impact DVD
of leechers or seeders after broadcast. As in the previous sales for movies at the point in time where they are shown on
section, a significant reason for the low estimates on piracy television? To analyze this question, we use the television
levels in these regressions is that viewership levels for ad broadcast of movies as an exogenous demand shock and
cable movies are significantly smaller than those for over-the compare the DVD sales gain for movies that have BitTorrent
air broadcasts. trackers at the time of broadcast to the DVD sales gain among
movies that do not have BitTorrent trackers at the time of
In summary, we find that over-the-air movie broadcasts tend broadcast. If the presence of pirated content harms sales, we
to stimulate both DVD sales and piracy, and these increases should see a smaller increase in post-broadcast sales for
are substantially higher for over-the-air broadcasts than they movies with pirated copies available than for those with no
are for cable broadcasts. pirated copies available.

MIS Quarterly Vol. 33 No. 2/June 2009 331

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Telang/Competing with Free

Table 6. Percentage Gain in Piracy

Weeks Since % Increase in Daily % Increase in % Increase in
Broadcast Downloads Leechers Seeds
D(1) 60%* 25% 45%*
D(2) 71%* 55%*
D(3) 99%* 17% 31%
D(4) 83%* 21% 40%
D(5) 80%* 15% 33%
! and* denote statistical significance at 0.01 and 0.05 respectively.

Table 7. Impact of Movie Broadcasts on Piracy (Fixed Effects) for Cable Channels
Independent Variables Daily Downloads Leechers Seeds
D(1) 0.31 (0.42) 0.34 (0.33) 0.25* (0.13)
D(2) 1.15** (0.57) -0.45 (0.45) 0.02 (0.17)
D(3) 2.1** (0.63) -0.55 (0.51) -0.04 (0.19)
D(4) 0.69 (0.68) -0.07 (0.54) -0.07 (0.20)
D(5) 0.11 (0.73) 0.17(0.58) 0.006 (0.21)
Tracker Age 0.22 (1.23) -1.24** (0.55) -0.91** (0.23)
Constant 3.91 (3.29) 16.71** (1.26) 7.82** (0.61)
Number of Observations 5628 6070 6070
Number of Groups 388 390 390

The dependent variable is daily downloads (column 1), number of leechers (column 2), and number of seeds (column 3). Standard errors are listed
in parenthesis; ** and * denote significance at 0.01 and 0.05, respectively. All models use tracker-level fixed effects.

We have data for 160 movies that were available on in terms of the change in rank after broadcast (see Table Al
BitTorrent at the time of broadcast and 107 movies that in the Appendix for these estimates). Moreover, in our
were not available on BitTorrent at the time of broadcast. regressions we control for movie popularity by including
One potential concern with the data is that more popular the number of viewers as a control variable. The fact that
movies might be more likely to be available on BitTorrent we see no differences between popular and less popular
than less popular movies.13 However, note that we are movies in terms of percentage change in rank after broad
interested in changes in sales rather than the absolute sales cast, combined with our use of movie-level fixed effects and
level (the fixed effect model measures changes in rank controls for the number of viewers (popularity), should
within movies). Thus, the actual starting rank is less of a control for selection effects when analyzing the change in
concern. Rather, if popular movies show a larger increase sales for movies available on BitTorrent versus the change
in sales after broadcast than do less popular movies, we in sales for movies that are not available on BitTorrent.
would have cause for concern due to a selection problem. However, below we also use a propensity score method
(Rosenbaum and Dubin 1983) as an additional check on the
However, we see no evidence in the data that there is a
possibility of selection bias, again finding no evidence of
difference between more popular and less popular movies selection bias in our results.

To avoid the additional notation of four weekly dummies

13This is supported by the data: The average Amazon sales rank of movies
available on BitTorrent is about 10,000, while the average rank of movies and four additional interaction terms, we simply use an after
that are not available on BitTorrent is about 16,000. broadcast dummy variable instead of four weekly dummies.

332 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Table 8. The Impact of BitTorrent Availability on After Broadcast DVD Sales

Log (Rank) Estimate
Price 0.073** (7.5)
After Broadcast -0.005** (3.9)
After Broadcast x BT 0.0005 (0.4)
Constant 8.14(150.0)
Number of Observations 5247
Number of Groups 266
The dependent variable is Log(Rank). T-statistics are listed in parentheses; ** and * denote significance at
0.01 and 0.05, respectively. All models use tracker-level fixed effects. Results are normalized per 100,000
viewers.

Thus, the dummy variable estimates the average change in our results. To address this possibility, we reestimate our
DVD sales per 100,000 viewers over the four-week period piracy regressions using a propensity score matching method.
after the movie broadcast. To capture the effect of BitTorrent, Propensity score matching has been used extensively in
we also interact the BitTorrent dummy variable with the "after economics and statistics to overcome the problem of selection
broadcast" dummy variable. Note that we cannot include a bias (Dehejia and Waba 2002). The basic principle of the
separate BitTorrent dummy variable in our estimation propensity score is to use some observable variables (e.g., box
because, in the fixed effect estimation, this dummy variable office revenues, imdb.com user ratings) to predict the
cannot be identified. We show the results for both over-the probability of a movie being on the BitTorrent network. This
air and ad-cable movies in Table 8. allows the direct comparison of movies that have similar
characteristics (propensity scores), where one movie is
We first note that the estimate on the after broadcast dummy available on BitTorrent while the other is not. Matching
is negative and significant, which is consistent with our movies in this way should substantially reduce any remaining
finding above that sales rank decreases (sales increase) in the selection bias issues.
month after a movie is broadcast on television. However, we
also note that the estimate on the interaction dummy variable Propensity scores are calculated using the standard Probit
is positive but statistically and economically insignificant. function with observed explanatory variables (see Table A2
This suggests that the increase in sales after broadcast is in the Appendix for the Probit results). We plot the pro
statistically the same for movies that are available on pensity scores for movies on BitTorrent (BT =1) and not on
BitTorrent at the time of broadcast (BT =1) and those that are BitTorrent (BT = 0) in Figure 2.
not (BT = 0).
Propensity score analysis techniques rely on being able to find
One potential concern about this result is that over-the-air movies with similar propensity scores in both groups (BT =
movies seem to show a much stronger increase in piracy than 0 and BT = 1). Based on this, it is important to note that the
ad-cable movies do, and thus over-the-air movies may plots in Figure 2 have a similar shape and most importantly
experience more harm from piracy. To address this issue, in that for any given propensity score it is possible to find
Table 9 we run this regression again, but this time with only movies with similar propensity scores in both the BT = 1 and
movies that were shown on over-the-air broadcast networks. BT = 0 groups.

The results of this regression are similar to those in Table 8, Once the propensity score is calculated, the analysis reduces
with slightly lower significance on the after broadcast dummy to comparing the sales changes of movies in the treatment
variable. Likewise, the interaction term is still statistically (BT =1) and control groups (BT = 0) with appropriately
and economically insignificant, although the sign is now matched propensity scores. For this test, the estimate on the
negative. difference in sales changes for movies on BitTorrent (as
compared to movies not on BitTorrent) is -0.038 with a
Finally, despite the controls for movie popularity outlined standard error of 0.061, making this coefficient statistically
above, it is still possible that a selection problem is driving insignificant.

MIS Quarterly Vol. 33 No. 2/June 2009 333

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Telang/Competing with Free

Table 9. The Impact of BitTorrent Availability on After Broadcast DVD Sales

(Over-the-Air Only)
Log (Rank) Estimate
Price 0.075** (3.6)
After Broadcast -0.003** (1.7)
After Broadcast x BT -0.0002 (0.0)
Constant 7.39 (71.9)
Number of Observations 823
Number of Groups 28

The dependent variable is Log(Rank). T-statistics are listed in parentheses; ** and * denote significance at
0.01 and 0.05, respectively. All models use tracker-level fixed effects. Results are normalized per 100,000
viewers.

.4 .6
Propensity Score (BT=1) Propensity Score (BT=0)

Figure 2. Propensity Scores for Movie

Thus, using both the regression analysis and Discussion

matching methods, we find no evidence tha
ability on BitTorrent Inat
this study,
the we analyze
time the ability of
of moviebroadcast
studios to
broadcast increase in DVD
compete with free copies sales. Put
of their content made available anot
television broadcasts through
ofboth movies increase
television broadcasts and pirate networks. The both
movie piracy, it seems creative industries
that have long argued that they
these two can't compete
user s
with free, and these
mate buyers and pirates) are concernsseparate.
may be particularly salient forThe te
movie that
cast acts as a stimulus studios, whose content
affects may be more pronebothto single segm
buyers order more DVDsuse consumption from
than other industries such as music.
Amazon afte
pirates download more copies of the movies f
networks as well. But there
We address this isdata(statistical
question by collecting from all movies
between shown on over-the-air
the two groups in terms and advertising of
supported cable
pirates p
that are unavailable onbroadcastsBitTorrent
from October 28, 2005, through March
or 3, 2006.potent
Our data include DVD sales
choosing instead to consume ainformation
pirated from Amazon.com
copy o
available on BitTorrent.and data tracking
We the supply
discussof and demand for pirated
these fin
detail below. content through two prominent BitTorrent networks.

334 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

We find that after a movie is shown on broadcast television, Second, our finding that the availability of pirated content
there is a strong and immediate increase in sales of the does not seem to impact the demand for legitimate content
corresponding DVD through Amazon.com. Similarly, we suggests that, at least at the point in time where a movie is
find that after broadcast there is a strong increase in the shown on television, demand from legitimate consumers and
demand for pirated content of these movies through two pirates is relatively segmented. That is, we do not see evi
prominent BitTorrent tracker sites. However, there is no dence that the availability of pirated content causes consumers

corresponding after-broadcast increase in the supply of who would have otherwise purchased a DVD after broadcast
pirated content. That is, movie broadcasts in our sample to consume pirated content instead. This result suggests that
studios may wish to focus their scarce antipiracy resources on
promote the consumption of pirated material but do not serve
recent theatrical and DVD releases where the availability of
as the source material for pirated content.
pirated material may have a stronger negative impact on sales.

We then use these empirical observations to analyze the

Finally, our findings may inform the recent debate on digital
impact of piracy on DVD demand by using the broadcast as
television content protection, such as the proposed broadcast
an exogenous shock to movie sales. In this analysis, we find flag legislation. Specifically, we find no empirical evidence
that movies that have pirated content available on BitTorrent to support the need for broadcast flag protection in digital
at the time of broadcast have statistically the same post television broadcasts, at least for movie content.14 In our data,
broadcast increase in sales as those that do not have pirated the dominant impact of unprotected over-the-air movie broad
content available at the time of broadcast.
casts is to increase DVD sales, the presence of pirated content
at the time of broadcasts does not impact DVD sales, and
Our results have several managerial and policy implications. digital television broadcasts do not serve as the source
First, for movie studios and broadcasters, our finding that material for pirated content.
movie broadcasts act as a strong complement to downstream
content sales should be encouraging for broadcasters who However, we also note that there are several important data
have long feared that the dominant impact of consumer analog and econometric limitations associated with this study. First,
and digital recording devices would be reduced demand for and most importantly, while our piracy regressions attempt to
control for differences between movies that are and are not
subsequent media purchases. From the introduction of the
available on BitTorrent networks at the time of broadcast
VCR to the more recent introduction of digital broadcast
(e.g., viewership, movie-level fixed effects, propensity score
television, movie studios have expressed concern that if a
consumer can record and retain a copy of a movie, TV analysis, and the use of proportional as opposed to absolute
sales changes), like any observational study, we cannot
broadcasts of movies will serve as a substitute for subsequent
completely rule out the possibility of selection bias.
purchases of the movie content. However, in a digital world,
this argument may ignore the increased opportunities for
Second, our sales results are based entirely on sales at
studios to differentiate their digital media products from
Amazon.com. While Amazon.com has an estimated 90 per
content shown over TV. For example, the increased capacity
cent share of the online DVD market (Netherby 2005), DVD
and random-access capabilities of the DVD format (and News (2006) estimates that, overall, Amazon.com is the
nascent Blu-ray format) allow studios to include extra content
fourth largest seller of DVDs in the United States behind
such as commentary tracks, deleted scenes, "behind-the brick-and-mortar giants WalMart, Target, and Best Buy.
scenes" documentaries, and music videos. It is also possible Nonetheless, we believe that Amazon is an appropriate sales
that the inconvenience consumers face in copying and storing reference point in our context for two reasons. First,
the broadcast content is sufficiently large to make the com WalMart, Target, and Best Buy (and most other brick-and
mercial purchase of media an attractive option. mortar retailers) typically carry a very limited selection
focused on recently released movies (see Brynjolfsson et al.
Similarly, the finding that TV broadcasts primarily serve as 2003). Since movies are typically shown on broadcast
complements to subsequent media purchases should also be television 12 to 18 months after their DVD release date, it
encouraging for studios increasingly looking to monetize their seems likely that at the time a movie is broadcast on tele
content through digital download services such as the iTunes vision, consumer demand will be focused on Internet retailers
video store, Amazon Unbox, and other similar services. such as Amazon as opposed to brick-and-mortar retailers.
Indeed the immediate spike in media purchases after a movie
is shown on television suggests there might be an opportunity
14Episodic or sports programming may have different behaviors and would
for in-program promotion of broadcast content. be a fruitful area for future research.

MIS Quarterly Vol. 33 No. 2/June 2009 335

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Telang/Competing with Free

Second, we believe that, at present, online retailers such as comments from conference and workshop participants at the 2007
Amazon.com are the most appropriate reference point for International Conference on Information Systems (ICIS), 2007
measuring the trade-offs consumers make between satisfying Workshop on the Economics of ICT, 2007 International Industrial
Organization Conference (HOC), 2006 Workshop on Information
the demand for movies through legitimate outlets and online
Systems and Economics (WISE), the 2006 INFORMS Marketing
pirate networks.
Science Conference, the 2006 Statistical Challenges in Electronic
Commerce Conference, the 2006 Telecommunications Policy
Third, our piracy results come from two public BitTorrent Research Conference (TPRC); and from seminar participants at
trackers and are not a comprehensive measure of the avail MIT, the Ohio State University, Purdue University, Temple Univer
ability of pirated material. Rather, we are using these data as sity, the University of California at Irvine, the University of Florida,
proxies for overall content availability and piracy levels. the University of Illinois at Urbana-Champaign, the University of
However, we believe these measures serve as valid proxies Minnesota, the University of Texas at Austin, the University of
based on the popularity of the BitTorrent protocol for movie Texas at Dallas, the University of Washington, and Carnegie Mellon
piracy and given the relative popularity of these two sites for University. We thank Pisit Chartbanchachai, Samita Dhanasobhon,
posting trackers related to movie piracy. Lawrence Gioia, Reshma Kane, Sang Dong Lee, Thomas Oliver,
Amita Pimple, Jocelyn Sabruno, and Kelly West for excellent
research assistance. Both authors acknowledge the National Science
Finally, it is possible that the post-broadcast sales increase
Foundation for generous financial support provided through
observed in our data is driven by promotion unrelated to the
CAREER award IIS-0118767 (Smith) and CAREER award
television broadcast. However, we also note that we believe
CNS-0546009 (Telang). Both authors also acknowledge generous
this is unlikely given that our results show a strong increase financial support through the Center for the Analysis of Property
in movie sales the week after broadcast and no statistical
Rights and Innovation (CAPRI) at the University of Texas at Dallas.
change in the week before broadcast. Author names are in alphabetical order.

In addition to these limitations, we also note that our results

should be viewed in their proper context. First, our results do References
not speak to the impact of piracy in the earlier part of a
movie's lifecycle, where the availability of pirated content Bhattacharjee, S., Gopal, R., Marsden, J., and Telang, R. 2007.
may have a negative impact on sales (see Rob and Waldfogel "The Effect of Digital Sharing Technologies on Music Markets:
2006 for example). Second, our findings may change in the A Survival Analysis of Albums on Ranking Charts," Manage
future if the environment surrounding piracy changes. It is ment Science (53:10), pp. 1359-1374.
possible that the increasing penetration of digital video Blackburn, D. 2007. "The Heterogeneous Effects of Copying: The
recorders, computer-based digital television recording and Case of Recorded Music," Working Paper, Harvard University,
Cambridge, MA.
editing products, and an increasing integration between com
Brynjolfsson, E., and Smith, M. 2000. "Frictionless Commerce?
puting equipment and television viewing devices will change
A Comparison of Internet and Conventional Retailers," Manage
consumers' preferences for recorded television broadcasts ment Science (46:4), pp. 563-585.
relative to purchased content. Similarly, it is possible that Brynjolfsson, E., Hu, Y., and Smith, M. 2003. "Consumer Surplus
increases in broadband Internet speeds and penetration will in the Digital Economy: Estimating the Value of Increased Pro
change consumers' preferences for purchased content relative duct Variety," Management Science (49:11), pp. 1580-1596.
to pirated content (see Smith and Telang 2007). Third, our Byers, S., Cranor, L., Korman, D., McDaniel, P., and Cronin, E.
results should not be viewed as a policy impact study as we 2003. "Analysis of Security Vulnerabilities in the Movie Produc
do not observe what would happen to DVD sales in the tion and Distribution Process," in Proceedings of the 3rd ACM
presence of content protection on digital television broadcasts Workshop on Digital Rights Management, Washington DC, pp.
1-12.
such as the proposed broadcast flag regulations. Indeed, each
Chevalier, J., and Goolsbee, A. 2003. "Measuring Prices and Price
of these topics would represent a useful area for future
research. Competition Online: Amazon and Barnes and Noble," Quantita
tive Marketing and Economics (1:2), pp. 203-222.
Chevalier, J., and Mayzlin, D. 2004. "The Effect of Word of
Mouth on Sales: Online Book Reviews," Working Paper, Yale
Acknowledgments School of Management, New Haven, CT.
Dehejia, R., and Wahba, S. 1999. "Propensity Score Matching
The authors thank Erik Brynjolfsson, Brett Danaher, Rajiv Dewan, Methods for Nonexperimental Causal Studies," The Review of
Stan Liebowitz, Hank Lucas, Marvin Sirbu, Koleman Strumpf, Economics and Statistics (84:1), pp. 151-161.
Lowell Taylor, Patrick Wagstrom, and the editors of this journal for DVD News. 2006. "Top DVD Retailers: Market Shares," DVD
valuable comments on this research. We also received many helpful News Magazine (10:41), November 30.

336 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
 Smith & Telang/Competing with Free

Elberse, A., and Eliashberg, J. 2003. "Demand and Supply shows/hollywood/business/windows.html; accessed January 5,
Dynamics for Sequentially Released Products in International 2006).
Markets: The Case of Motion Pictures," Marketing Science Peitz, M., and Waelbroeck, P. 2003. "Piracy of Digital Products:
(22:3), pp. 329-354. A Critical Review of the Economics Literature," CESifo
Epstein, E. J. 2005. The Big Picture: The New Logic of Money and Working Paper Series No. 1071.
Power in Hollywood, New York: Random House. Peitz, M, and Waelbroeck, P. 2004. "The Effect of Internet Piracy
Ghose, A., Smith, M. D., and Telang, R. 2006. "Internet Exchanges on Music Sales: Cross-Section Evidence," Review of Economic
for Used Books: An Empirical Analysis of Product Cannibali Research on Copyright Issues (1:2), pp. 71-79.
zation and Welfare Impact," Information Systems Research Png, I., and Wang, Q. 2006. "Copyright Duration and the Supply
(17:1), pp. 3-19. of Creative Work: Evidence from the Movies," Working Paper,
Ghose, A., and Sundararajan, A. 2005. "Software Versioning and National University of Singapore.
Quality Degradation? An Exploratory Study of the Evidence," Prasad, A., and Mahajan, V. 2003. "How Many Pirates Should a
Working Paper, New York University. Software Firm Tolerate? An Analysis of Piracy Protection on the
Gil, P. 2006. "What Are the Good BitTorrent Search Engines," Diffusion of Software," International Journal of Research in
About.com: Internet for Beginners, February 12 (available from Marketing (20:4), pp. 337-353.
http://netforbeginners.about.eom/od/peersharing/f/ Rob, R., and Waldfogel, J. 2006. "Piracy on the High C's: Music
torrentsearch.htm; accessed February 19, 2006). Downloading, Sales Displacement, and Social Welfare in a
Gopal, R., and Sanders, L., 1998. "International Software Piracy: Sample of College Students," Journal of Law and Economics
Analysis of Key Issues and Impacts," Information Systems (49:1), pp. 29-62.
Research (9:4), pp. 380-397. Rob, R., and Waldfogel, J. 2007. "Piracy on the Silver Screen,"
Gu, B., and Mahajan, V. 2005. "How Much Anti-Piracy Effort Is Journal of Industrial Economics (55:3), pp. 379-393.
Too Much? A Study of the Global Software Industry," Working Rosenbaum, P., and Rubin, D. 1983. "The Central Role of the
Paper, McCombs School of Business, University of Texas, Propensity Score in Observational Studies for Causal Effects,"
Austin, TX (available at http://ssrn.com/abstract=825165). Biometrika (70:1), pp. 41-55.
Hui, K., and Png, I. 2003. "Piracy and the Legitimate Demand for Smith, M. D., and Brynjolfsson, E. 2001. "Customer Decision
Recorded Music," Contributions to Economic Analysis and Making at an Internet Shopbot: Brand Still Matters," The
Policy (2:1), pp. 1-24. Journal of Industrial Economics (49:4), pp. 541-558.
Hong, S. 2004. "The Effect of Napster on Recorded Music Sales: Smith, M. D., and Telang, R. 2004. "Internet Exchanges for Used
Evidence from the Consumer Expenditure Survey," Stanford Goods: A Comparison of Digital and Analog Information
Institute for Economic Policy Research Working Paper 03-18, Goods," paper presented at the Workshop on Information
Stanford University, Stanford, CA. Systems and Economics (WISE), Washington DC, December
Kannan, P. K., and Jain, S. 2002. "Pricing Product Lines of Digital 11-12.
Content: A Model and Application to Online Content Provision," Smith, M. D., and Telang, R. 2007. "Piracy or Promotion? The
Working Paper, University of Maryland, College Park, MD. Impact of the Broadband Internet Penetration on DVD Sales,"
Knox, G., and Eliashberg, J. 2005. "The Consumer's Rent Versus Working Paper, Carnegie Mellon University, Pittsburgh, PA.
Buy Decision: The Case of Home Video," Working Paper, The Thompson, A. 2003. "Tinseltown Follies," The New York
Wharton School, The University of Pennsylvania, Philadelphia, Magazine, April 28 (available at http://nymag.com/nymetro/
PA. movies/columns/holly wood/n_8667/).
Liebowitz, S. L. 1985. "The Economics of Betamax: Unauthorized Zentner, A. 2006. "Measuring the Effect of File Sharing on Music
Copying of Advertising Based Television Broadcasts," Working Purchases," Journal of Law and Economics (49:1), pp. 63-90.
Paper, University of Chicago.
Liebowitz, S. L. 2004. "The Elusive Symbiosis: The Impact of
Radio on the Record Industry," Review of Economic Research About the Authors
on Copyright Issues (1:1), pp. 93-118.
Liebowitz, S. L. 2008. "Testing File-Sharing's Impact by Exam Michael D. Smith is an associate professor of Information Systems
ining Record Sales in Cities," Management Science (54:4), pp. and Marketing at Carnegie Mellon University, with appointments
852-859. at the Heinz College's School of Information Systems and Manage
Lucey, A. 2002. "Comments of Viacom Before the Federal Com ment and the Tepper School of Business. He received his Ph.D. in
munications Commission, in the Matter of Digital Broadcast Management Science and Information Technology from the Sloan
Copy Protection," MB Docket No. 02-230, December 6. School of Management at MIT. Michael's research relates to
Netherby, J. 2005. "Amazon.com Dominates in Online DVD analyzing and designing efficient information exchanges. His
Sales," Reed Business Information Inc. Gale Group, Farmington research in this area has been published in leading Management
Hills, MI, December 19. Science, Economics, and Marketing journals and covered by popular
Oberholzer, F., and Strumpf, K. 2007. "The Effect of File Sharing outlets including The Economist, The Wall Street Journal, Sloan
on Record Sales: An Empirical Analysis," Journal of Political Management Review, The New York Times, Time Magazine, and
Economy (115:1), pp. 1-42. Business Week. He also jointly conducted some of the first
PBS. 2005. "NowPlaying...andPlaying...andPlaying,"PBSFront academic research into the social welfare impact of increased
line (available from http://www.pbs.org/wgbh/pages/frontline/ product variety in Internet markets. This work was cited in Chris

MIS Quarterly Vol. 33 No. 2/June 2009 337

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms
Smith & Telang/Competing with Free

Anderson's best-selling, and artfully titled, book The Long Tail. various security/privacy laws, compliance, and standards on con
Michael has received several awards for his teaching and research sumer welfare. He received the NSF Career Award for his work in
including the National Science Foundation's prestigious CAREER the domain. His second stream of research is focused on interaction
Award, the best published paper award runner-up for Information of various digital media distribution platforms with particular focus
Systems Research in 2006, and the Best Teacher Award in the on online piracy. He has empirically investigated the impact on
Master's of Information Systems Management program. used book markets, P2P file sharing for music, movies and television
shows. He is the recipient of the Alfred P. Sloan foundation
Rahul Telang is an associate professor of Information Systems and industry studies fellow for his work in this domain. Rahul has
Management at the Heinz College, Carnegie Mellon University. He published many articles in various journals including Management
earned his doctorate from the Tepper School of Business, Carnegie Science, Information Systems Research, Journal of Marketing
Mellon University in 2002. Rahul's main research interests are in Research, and IEEE Transactions on Software Engineering. He
economics of information security. He has examined software serves on the editorial boards of Management Science
vulnerability disclosure, patch release policies, and impact of vari and Information Systems Research.

Appendix
Table A1. Estimates with Different Starting Ranks
Independent Variables Starting Rank 1,000-8,000 Starting Rank 10,000-20,000
DM) -0.000 (-.013) 0.002(1.6)
D(1) -0.010** (-21.7) -0.008** (-7.7)
D(2) -0.004** (-7.4) -0.005** (3.8)
D(3) -0.001** (-2.6) -0.001** (-0.7)
D(4) -0.001 (-1.0) -0.003 (2.2)
Amazon Price 0.115** (18.2) 0.083** (9.3)
Constant 6.872 ** (87.1) 8.401** (71.6)
Number of Observations 7,191 4,020
Number of Groups 332 170
Dependent variable is ln(sales rank). T-statistics are listed in parenthesis;1 ' and * denote significance at 0.01 and
0.05, respectively. All models use DVD-level fixed effects.

Table A2. Propensity Score Estimation (Probit)

Independent Variables Estimate (Standard Error)
Average Price Before Broadcast 0.076** (-0.03)
Gross Revenues(log) 0.240** (0.09)
IMDB ratings 0.284** (0.1)
Movie Duration -0.002 (0.01)
Number of discs -0.543** (0.23)
Minutes edited -0.007 (0.008)
DVD age 0.006 (0.03)
Constant -5.43** (1.4)
Number of Observations (N) 231
The dependent variable is availability of BitTorrent at the time of broadcast (0/1). Notice
that all the estimates are sensible: DVDs with higher price, higher box-office revenues,
higher IMDB ratings, and fewer discs are more likely to be available on BitTorrent.

338 MIS Quarterly Vol. 33 No. 2/June 2009

This content downloaded from

104.28.244.85 on Thu, 30 Mar 2023 19:46:05 UTC
All use subject to https://about.jstor.org/terms