Rent or Buy: The Effectiveness of Netflix and Redbox Windows

Document technical information

Format pdf
Size 1.6 MB
First found Jun 9, 2017

Document content analysis

Language
English
Type
not defined
Concepts
no text concepts found

Persons

Hal Varian
Hal Varian

wikipedia, lookup

Bob Iger
Bob Iger

wikipedia, lookup

Organizations

Places

Transcript

Rent or Buy: The Effectiveness of Netflix and Redbox Windows
Mickey Ferri∗
(Job Market Paper)
Economics Department
University of Chicago
First Draft: September 15, 2011
This Draft: January 24, 2013
Abstract
The movie industry has been in turmoil for several years, as distributors try to profit
from new legal exhibition methods and dissuade consumers from illegal consumption.
Threatened by low-priced rentals and increasing digital options, DVD and Blu-ray sales
have been falling every year since 2007. I make use of exogenous variation in Netflix
and Redbox availability to show that when movies are not available on Netflix and
Redbox during the first four weeks of release, DVD sales, Blu-ray sales, and Video
on Demand rentals are unaffected. Instead of substituting to higher-priced options,
consumers time-shift their consumption and wait until Week Five or later for a Netflix
or Redbox rental. This implies that consumer preferences for method of consumption
outweigh the decay in utility from postponing, and that a windowing strategy yields
lower short-run profits. Despite lower short-run profits, the windowing strategy may
benefit studios in the long run by slowing the growth of Netflix and Redbox and
stimulating growth of the downstream exhibition channels that studios own.
∗
I am deeply grateful to Ali Hortacsu, Chad Syverson, Steve Levitt, and Pradeep Chintagunta for invaluable feedback and guidance. I thank Amanda Agan, Dennis Carlton, Yiwei Chen, Tal Cohen Chaiken, Tony
Cookson, Sebastien Gay, Pranav Jindal, Anup Malani, Kevin Murphy, Emily Oster, Marina Niessner, Laura
Pilossoph, Julian Reif, Allen Sanderson, Jesse Shapiro, and University of Chicago Applied Microeconomics
Lunch participants for useful discussions and feedback.
1
1
Introduction
The movie industry has been in turmoil for several years, as distributors try to profit from
new legal exhibition methods and dissuade consumers from illegal consumption. Threatened
by low-priced rentals and increasing digital options, DVD and Blu-ray sales have been falling
every year since 2007. Industry experts have argued over how much of the fall in DVD and
Blu-ray sales should be attributed to Netflix and Redbox cannibalization and how much
should be attributed to other factors like illegal downloading and increasing legal digital
options. To the best of my knowledge, this paper is the first academic attempt to answer
that question.
I build a general model of consumer preferences for renting and buying DVDs, Blu-rays,
and digital movies. To estimate my model, I use several sources to construct a novel dataset
that includes national domestic weekly DVD sales, Blu-ray sales, physical rentals, and Video
on Demand (VOD) rentals. The most detailed specifications of my model explain 80-90% of
the variation in rentals and sales.
In early 2010, three of the six major movie distributors, Fox, Universal, and Warner
Brothers, took severe actions in an effort to boost disc sales and high-priced rentals. Rather
than giving low-priced rental companies Netflix and Redbox immediate access to DVDs, they
withheld movies from these two companies for a 28 day window after the initial DVD release
date.1 This studio specific variation in Netflix and Redbox availability is exogenous to movie
characteristics, providing a natural experiment to evaluate the effectiveness of the windowing
policy and analyze consumer substitution patterns. Contrary to the hopes of Fox, Universal,
and Warner Brothers, I find that when movies are not available on Netflix and Redbox for
a 28 day window, DVD sales, Blu-ray sales, and VOD rentals are unaffected. Instead of
substituting to higher-priced options, consumers either 1. time-shift their consumption and
wait until Week Five or later for a Netflix or Redbox rental or 2. drop out of the market for
1
see section 3.2 and table 2 for a detailed discussion on the histories of Netflix and Redbox.
2
that movie entirely. I estimate that the average major DVD release2 would earn $5 million
more if it were available at Netflix and Redbox on the DVD release date rather than after a
28 day window, and that the windowing strategy costs studios an average of $3.33 million
in profits per movie.
Technology, movie exhibition, and the overall economy have evolved rapidly over the past
five years, and one of the major challenges of this paper is understanding and accounting
for all of the market forces at work. I identify six factors that have led to major shifts in
movie exhibition demand and supply since 2007: 1. the innovative use of DVD-by-mail and
kiosk technology by Netflix and Redbox to take over the physical rental market,3 2. the
introduction of Blu-ray,4 3. better online legal sources for buying and renting movies (e.g.
Netflix and iTunes), 4. the economic recession - DVD and Blu-ray sales are generally normal
goods, while rentals are inferior goods, 5. increased ease of internet movie piracy,5 and 6.
the emergence of alternative forms of entertainment (youtube, facebook, etc.).
The six major studios, Disney, Fox, Paramount, Sony, Universal, and Warner Brothers,
control the distribution of about 80% of the industry. They are all parts of large media
and entertainment conglomerates that have varying levels of interest in different parts of
the industry.6 Fox, Universal, and Warner Brothers have significant ownership and long
term contracts with downstream exhibition channels like HBO, Comcast, and Hulu, so these
three studios see Netflix and Redbox as huge threats to existing business models. The other
three major studios, Disney, Paramount, and Sony, have less ownership in movie exhibition
channels, so they see Netflix and Redbox as new market opportunities rather than threats
to their existing business model.
2
From June 1, 2011 through June 30, 2012
At their respective peaks, Blockbuster employed about 60,000 people and Netflix employed about 4,000
people to rent about the same number of total units. Average rental prices at Netflix and Redbox are close
to $2, while their competitors average $5.
4
There was a short-lived Blu-ray HD-DVD format war, which likely had a short run negative impact on both
demand and supply. The long run effect of Blu-ray would be into increase disc sales.
5
See Ferri (2012) for a detailed description of the illegal movie downloading market. According to a 2006
study by LEK Consulting, U.S. studios lost $6.1 billion in global wholesale revenue in 2005 due to movie
piracy (Mcbride and Fowler, 2006).
6
See section 3, as well as tables 4 and 5 for a detailed description of the movie industry.
3
3
The main finding of this paper is that when movies are delayed on Netflix and Redbox
for a 28 day window, DVD sales, Blu-ray sales, and VOD rentals are unaffected. Instead of
substituting to higher-priced options, consumers either 1. time-shift their consumption and
wait until Week Five or later for a Netflix or Redbox rental or 2. drop out of the market
for that movie entirely. This substitution pattern implies that consumers exhibit strong
loyalty to method of consumption and that there is a positive but relatively small disutility
of waiting an additional 28 days. The windowing strategy yields lower short run profits, but
at the same time it greatly reduces the quality of titles on Netflix and Redbox. The logical
conclusion is that Fox, Universal and Warner Brothers forego short term profits to protect
their downstream exhibition channels, such as Comcast, HBO, and Hulu, which is especially
important in light of the strong loyalty to method of consumption. In the future, studios are
sure to continue altering their release and pricing strategies. One top Paramount strategist
foresees a scenario in which “a movie opens on 25,000 screens around the world in a single
weekend, and within a week it’s available for downloads, Netflix, video stores, and cable
television” (Epstein, 2010). In the latest studio attempt to control the entire vertical supply
chain, a consortium of firms including Walmart and the five major studios aside from Disney
is promoting a digital rights licensing system called Ultraviolet, which officially launched in
October, 2011.
The rest of this paper is structured as follows. In section 2, I discuss related work. In
section 3, I describe all of the relevant markets within the movie industry. In section 4,
I present a model of consumer tastes for renting and buying movies. Section 5 describes
the various data sources I use for the analysis, and section 6 contains results, and section
7 concludes. Appendix A contains tables and figures, appendix B includes miscellaneous
information, examples, and derivations, and appendix C is a data appendix in a separate file
available on my website.
4
2
Related Work
This is the first academic paper as far as I know to analyze the impact of low cost
physical rentals on the rest of the movie market. There are several existing studies that
have used proprietary data, many of which were conducted by the movie studios themselves,
which I discuss in section 2.1. I make a contribution to this literature by constructing a
comprehensive dataset of all major studio releases from 2006 to present and using the dataset
to answer some questions that are very important to the success of the movie industry. There
are two strands of literature in economics and marketing to which I hope to contribute in
this paper. The sequential release of products literature (section 2.2) includes (Lehmann and
Weinberg, 2000), (Luan and Sudhir, 2006), and (Chiou, 2008), and the consumer decision to
rent or buy literature (section 2.3) includes Varian (2000), Mortimer (2007), Rao (2011).
2.1
Home Media Market
Compared to the box office, there is scant academic research on the home viewing market.
The main reason for this is a lack of good data. Box office data is widely available and well
organized, while sale and rental data is much more difficult to find. Some data is publicly
available from early 2006, it is not well organized, incomplete, and in many cases needs to
be manually transcribed from the original source. Much of the current industry discussion
is based on proprietary consultancy reports and consumer surveys for which the underlying
data, assumptions, and analyses are not open for verification.
For example, Nielsen EDI conducted a study on the relationship between box office
and DVD sales in 2005 (Snyder, 2005). More recently, Lang et al. (2011) use a subset of
the data used in my paper to look at the factors determining sales of individual DVDs,
with an emphasis on the MPAA rating. In 2012, BTIG Research surveyed respondents
on the question, “If new movies released in the theater were offered simultaneously in the
home via cable/satellite/VOD or electronic sellthrough for $20-$25 per movie, would it
5
increase/decrease/have no impact on your households movie expenditures and movie industry
piracy?” (Gruenwedel, 2012).
All of the research I have seen on the impact of low cost rentals comes from movie
studios, consultancies, and rental companies themselves. Mitch Lowe, CEO of Redbox, has
consistently dismissed worries about the cannibalization of sales. During August, 2009, he
cited internal research indicating that 20 percent of Redbox’s volume is additive - people
who did not previously buy or rent DVDs - and that partners like Wal-Mart have had only
a 1 percent decline in sales after Redbox machines have been installed at their entrances.7
Also citing internal research, Lowe commented that “About a third of those customers end
up buying a movie after seeing how great it is.”8 Speaking September 23, 2010, Time
Warner CFO John Martin said that Warner Brothers’ 28 day delay of new release movies
to Netflix and Redbox had helped increase packaged-media sales 15% and improved cable
video-on-demand by 20% to 30%.9
Viacom CEO Philipe Dauman noted that “Unlike some of our competitors who were
in litigation with Redbox, we had an agreement that gave us actual data over a 10 month
period...The data showed us that there was extremely little degradation in DVD sales, and
the financial terms offered by Redbox for an earlier window far outweighed the degradation.
Our competitors who have 28 day windows are getting much lower pricing in the Redbox
deals.”10
In June, 2010, Disney CEO Bob Iger told an investor group that “we have not seen
any significant cannibalization from the $1 Redbox rental window...And with that in mind,
selling them units at 50% reduction in cost, even for the 28-day window, was not an equation
that made sense to us.” (Gruenwedel, 2010) In early 2012, Disney joined Fox, Universal, and
7
Article available at http://www.nytimes.com/2009/09/07/business/media/07redbox.html? r=1&ref=technology
(Last accessed April 18, 2012)
8
Article available at http://www.homemediamagazine.com/redbox/mitch-lowe-redbox-took-fox-court%E2%80%98reluctantly%E2%80%99-16718 (Last accessed May 12, 2012)
9
Source: Warner Execs Say Windows Up Revenue, Erik Gruenwedel, Home Media Magazine, September
27-October 3, 2010 issue
10
Viacom’s Dauman: Video Rev Not Impaired By Kiosks, By Chris Tribbey, Home Media Magazine, August
4, 2010
6
Warner Brothers in imposing a window on Netflix and Redbox. When asked why Disney
had waited nearly three years to impose the delay, Iger said the studio had worried there
was nothing to be gained by the delays and would actually hurt sellthrough.11
Redbox received a lot of attention after a worse than expected fourth quarter of 2010,
during which analyst Michael Pachter said that revenue “missed” by $30 million, reflecting
about 14 million fewer rental transactions than the company had expected. Pachter “believe[s] that approximately 25% of these “lost” transactions reflected the purchase of DVDs;
... around 25% were market share losses to VOD on cable or the Internet; ... around 25%
reflected a shift to Blockbuster (which aggressively promoted its 28-day advantage during
the quarter); and ... the remaining 25% were “walk away” customers who didnt find the
breadth of product they were seeking when visiting a Redbox kiosk.” Pachter estimated the
net gain of DVD sellthrough at around 3 million discs, or around $50 million, out of approximately $5 billion in DVD sales for the quarter. He estimated the net gain to Blockbuster
to be around 4 million transactions, or around 3% of its quarterly sales, and the net gain to
VOD to be around 4 million transactions, or around 5% of quarterly sales.12
2.2
Sequentially Released Products
There are large economic and marketing literatures, both theoretical and empirical, on
sequentially released products. We see sequential releases in many industries other than the
movie industry: for example books and electronic games. In these industries, some of the
most important decisions producers have to make are which distribution channels to use and
when to release their product in each distribution channel. In the movie industry, the two
most obvious sequential distribution channels historically are and the box office and DVD
release. Lehmann and Weinberg (2000) and Nelson et al. (2007) analyze the optimal timing
of a DVD release, given box office performance, although both papers suffer from a lack of
11
Article available at http://www.homemediamagazine.com/disney/iger-disney-implementing-28-daydelays-taking-wait-and-see-approach-toward-ultraviolet-26354 (Last accessed June 8, 2012)
12
Article available at http://www.homemediamagazine.com/redbox/analysts-kiosk-business-sound-despiteredbox-missteps-21698 (Last accessed May 9, 2012)
7
DVD and VHS data. Luan and Sudhir (2006) develops a structural model to analyze the
optimal DVD release timing problem, then estimates the model with proprietary DVD data
from 1999-2003. Chiou (2008) also analyzes the optimal timing of DVD releases, using data
from 1999 to 2003. Hennig-Thurau et al. (2007) uses survey data to study the timing and
order of movie distribution channels, coming to the conclusion that the simultaneous release
of movies in theaters and on rental home video generates maximum revenues for studios, but
has devastating effects on other players, such as theater chains.
Since the publication of these papers, the internet has radically altered the movie industry,
both lowering the costs of delivery and increasing the consumption options. There is an
entirely new decision to be considered, i.e. whether to release a movie for sale and rental at
the same time or stagger these dates. In this paper, I focus on that decision rather than the
optimality of the gap between box office and DVD release dates.
2.3
Consumer Decision to Rent or Buy
This paper also contributes to the literature on the consumer decision to rent or buy. In
his seminal paper, Varian (2000) considers several different models for consumer preferences
and sharing costs, and derives the optimal sale and rental prices in each situation. Mortimer
(2007) analyzes the relationship between upstream and downstream sellers in the DVD and
VHS rental markets using data from January, 2000 through June, 2002. In her job market
paper, Rao (2011) considers the optimal prices for the online purchase and rental prices for
a movie, and she designs an experiment to recover consumer preferences.
3
Market Description
In section 3.1, I give a broad overview of the movie industry. I describe the physical
rental market in section 3.2 and the video on demand, digital sale and digital rental markets
in section 3.3. I leave discussions on premium cable (section B.1) and premium video on
8
demand (section B.3) in Appendix B.
3.1
Movie Industry Overview
The supply side of the movie industry consists of three vertical components: 1. production, 2. distribution, and 3. downstream exhibition, retail, and rental. Most production
and distribution is vertically integrated, and integration with the downstream companies is
increasing.13 The industry is fairly to highly concentrated at all levels, and the six major
movie studios14 are all parts of large media and entertainment conglomerates. Table 4 is
summary of the major movie studios and their affiliations, and Table 5 is a summary of the
market leaders in each downstream segment. Table 1 is meant to give an illustration of all
possible methods of movie consumption, in roughly decreasing order of average consumer
willingness to pay.
Major box office releases generally open on Fridays or Wednesdays on the same date all
over the world. The average movie makes over half its revenue during the first two weeks
of wide release15 and remains in wide release for five or six weeks. Toward the end of the
box office run (generally 8-12 weeks in), the distributor announces a DVD release date, often
referred to as “street date.” On average, there is an 18 week gap between the box office wide
release date and street date, and that gap has been shortening during recent years.
The fact that distributors cannot own theaters, together with the large market power of
the major theater chains, has kept the long box office to DVD gap in place. Evidence and
common sense suggest that the long gap harms overall industry profitability and consumer
13
In the early 20th century, the major Hollywood studios had integrated all three components. However,
in United States v. Paramount Pictures, Inc., (1948), vertical integration between distributors and movie
theaters was ruled as anticompetitive behavior. Recent examples of increasing integration include Warner
Brothers-Time Warner-HBO, Hulu (owned by Disney, Fox, and Universal), ABC-Disney, Ultraviolet (See
section 3.3), and the NBC-Universal-Comcast merger.
14
The six major studios distribute their own content, so I will often use the words “studio” and “distributor”
interchangeably.
15
Movies are released in theaters under wide, limited, or regional releases. A wide release is defined to be
a movie showing in 600 or more theaters in the U.S. Limited or regional releases will generally start in
10-200 theaters, and then if they are successful, the distributor will expand to a wide release. This can act
as both advertising and market research for the film, and I consider the “release date” to be the date at
which the movie expanded to wide release.
9
surplus, while benefitting movie theater owners.16
The DVD and Blu-ray17 sale market is relatively straightforward. DVDs and Blu-rays
generally go on sale on Tuesdays, and there is a sharp decay in unit sales each week, as
shown in Table 9a. Walmart, Target, Amazon, and Best Buy are the four major retailers of
new release DVDs and Blu-rays and together make up almost the entire market. Walmart
and Target use DVDs and Blu-rays as loss leaders to drive traffic into stores.
Concurrent with the DVD and Blu-ray release, there is a physical rental market, digital
sale and rental markets, and video on demand. Again, most digital sales and rentals are
within the first few weeks of release. Well after these markets are saturated, premium cable
channels and basic cable channels acquire the rights to show the movie. I include a discussion
of premium and basic cable in appendix B, Section B.1.
3.2
Physical Rental Market
The U.S. movie rental industry was very stable from 2003 through 2007. Annual rental
revenues were constant at $8 billion, rental prices for new releases were $5, and traditional
brick and mortar stores dominated the market. Blockbuster and Movie Gallery together
comprised about half of the U.S. rental market. Netflix, after going public in 2002, had been
growing very slowly, and Redbox, founded in 2002, remained a trivial part of the market.
Starting in 2008, Netflix and Redbox experienced massive growth at the expense of brick
and mortar rental stores, as shown in Figures 1 and 2. Movie Gallery filed for chapter
11 bankruptcy in October, 2007 and closed the last of its 4,700 stores in August, 2010.
Blockbuster filed for bankruptcy in September, 2010 and was bought by Dish Network in
April, 2011. There were approximately 900 U.S. Blockbuster locations as of June, 2012,
16
In 2005, J.P. Morgan issued a study with the result that if studios released films simultaneously in theaters
and on DVD, box office revenue would fall by 49% and DVD revenues would increase by 76%, leading to
a 36% increase in overall studio revenues from $14.9 billion to $20.2 billion per year (Snyder, 2005).
17
Blu-ray was introduced in 2006 as a higher quality format than DVD. A single layer Blu-ray disk can hold
25 GB of information, compared to 4.7 GB for a DVD. The extra space allows for better sound and picture
quality, as well as more special features. There was a short lived format war between Blu-ray and HD
DVD, but HD DVD production ended in 2008.
10
down from a peak of over 5,800 in 2004.18
Before 2011, most of Netflix revenues came from its DVD-by-mail product, which allows
consumers to have a certain amount of movies out at a time for a monthly fee. The most
popular Netflix DVD-by-mail package is priced at about $8 per month, and the average
Netflix by mail subscriber rents four movies a month, yielding an average rental price of
approximately $2. Redbox places DVD kiosks at locations such as supermarkets, pharmacies,
and convenience stores, and charges $1 for each night that the consumer holds the DVD.19
The average consumer holds a Redbox DVD for approximately 2 nights, yielding an average
price of approximately $2.20
Due to their innovative business models21 and low prices, Netflix and Redbox thrived
in the recession economy. As they took over the physical rental market, average prices
plummeted. Physical rental volume remained stable, so overall physical revenues dropped,
as shown in Figure 1. With distributor profits from the physical rental market dropping,22
some of the distributors began to take severe actions. Fox, Universal, and Warner Brothers
refused to sell DVDs to Redbox, and Redbox countered with lawsuits. A timeline of Redbox
history is available in Table 2b. Throughout the legal battles, Redbox employees bought
copies of the DVDs from Walmart and other retailers to stock their machines as soon as
possible after street date.23 While fighting legal battles with Fox, Universal, and Warner
Brothers, Redbox signed same day distribution deals for movies from Disney, Paramount,
Sony, Lionsgate, and Summit. Under these distribution agreements, Redbox typically buys
18
Movie Gallery, Blockbuster, and Dish Network 10-K reports
Redbox increased their prices for DVD rentals in late 2011. See Table 2b for details.
20
The average Redbox transaction during the first quarter of 2009 was $2.01, and it was $2.16 during the
first quarter of 2010. (Article available at http://www.homemediamagazine.com/redbox/coinstar-posts64-million-profit-19249 (Last accessed April 18, 2012)
21
With no physical stores and minimal labor costs, both companies had much lower costs than their competitors. Redbox allows consumers to check online for availability, has significantly more locations than their
competitors, and allows consumers to return to any location. Netflix allows consumers to browse and order
from its website and delivers directly to people’s homes through the U.S. Postal service. As of March 28,
2011, Netflix had 58 shipping locations throughout the U.S. (Source: http://en.wikipedia.org/wiki/Netflix)
22
Netflix and Redbox used fewer copies of the DVDs than other companies to generate the same number of
rentals
23
See Section B.2 for details.
19
11
DVDs directly from the distributors and is required to return or destroy them rather than
reselling after their rental lives are up. At the time, Sony DVDs were expected to account
for 20% of Redbox purchases, and Lionsgate were expected to account for 7.4%. Analysts
speculated that both Sony and Paramount signed agreements with Redbox in part because
their home entertainment units were under pressure to meet financial targets that had been
set before the DVD decline.24 In early 2010, the legal battles ended as Redbox agreed to
28 day windows for Fox, Universal, and Warner titles in exchange for lower DVD prices.
The studios aligned similarly when it came to Netflix; Disney and Paramount signed sameday distribution agreements, Fox, Universal, and Warner Brothers signed 28 day window
agreements, and Sony worked with Netflix on a title-by-title basis.
Despite only having half of all titles on street date, Netflix and Redbox continued to grow
during 2010 and 2011. One of the keys to Netflix’s success was its low cost of Starz streaming
content.25 With the Starz contract set to expire in October, 2011, Netflix anticipated a
dramatic increase in its costs and needed to raise prices to maintain profitability. Effective
September 1, 2011, Netflix stopped offering its most popular product, which was a bundle of
one DVD out at a time and unlimited streaming for $9.99 a month. Each service was priced
separately at $7.99, so this was effectively a 60% price increase of its most popular product.
At the time of the announcement, the vast majority of Netflix’s 24 million consumers were
using the bundled plan. Less than a year later at the end of June, 2012, the number of DVD
subscribers had fallen below 10 million, as shown in Figure 2.
Although Redbox failed to meet revenue expectations during the fourth quarter of 2010,
they continued to grow through 2011 despite 28 day delays on half their titles. Since Netflix’s
price hike in late 2011 and Redbox’s early 2012 buyout of their biggest kiosk rival, NCR’s
Blockbuster Express, Redbox is by far the dominant firm in the physical rental market.
24
Article available at http://www.nytimes.com/2009/09/07/business/media/07redbox.html? r=1&ref=technology
(Last accessed April 18, 2012)
25
See section 3.3 for a discussion.
12
3.3
Video On Demand, Digital Sale and Rental
In this section, I describe the digital market for movie sale and rental. The best way to
understand the digital movie market is to break it up into five submarkets, which I will call:
1. video on demand rental (VOD), 2. digital sale, 3. digital rental, 4. subscription video
on demand, and 5. premium video on demand rental. I discuss the first four submarkets
in detail in this section, and leave the discussion of premium VOD in Appendix B. The
leading suppliers and their market shares can be seen in Table 5. VOD and premium VOD
are available through multi service operators like Comcast and DirecTV, and consumers pay
on a per movie basis. Digital sale and rental are available through other companies like
Apple, Google, and Amazon, and consumers pay on a per movie basis. Subscription VOD
providers typically charge users for monthly subscriptions. Major blockbuster movies are
generally available for VOD, digital sale, and digital rental on street date, premium VOD
before street date, and subscription VOD after street date. Street date rentals are typically
priced at $4.99, ($5.99 for high definition), and sales are priced at $15-$20 ($20-$25 for HD).
Premium VOD rentals are priced at $30, and subscriptions are generally $8 per month.
Digital delivery has many advantages over physical delivery. It is much less costly to
manufacture, transport, stock, and deliver to consumers. The consumer does not have to
leave his or her home and can start watching a movie within seconds of ordering. VOD profit
margins for the studio can reach 70%, compared to 30% for DVD rental, which translates to
$2 more per rental.26
Video On Demand
Despite the higher margins, studios were reluctant to implement street date VOD for
years. In 2006, the window between street date and VOD was typically 45 days (Gruenwedel,
2010). A likely explanation is that they used the threat of street date VOD to negotiate
higher prices to brick and mortar rental companies. When the brick and mortar rental
26
Article available at http://staging.homemediamagazine.com/vod/economy-drives-%E2%80%98risky
%E2%80%99-studio-embrace-electronic-rental-16015 (Last accessed June 6, 2012)
13
market started to collapse, the studios turned to street date VOD. Warner Brothers was the
first to offer a street date VOD title through Comcast in 2007, and by 2009 the majority
of Warner titles were available through Comcast on street date. Fox, Universal, Summit,
and MGM had also begun to make titles available on street date, and by 2009, over half of
the new release movies Comcast VOD offered were available on street date.27 During 2010,
the movie studios and MSOs launched a $30 million marketing campaign, “The Video Store
Just Moved In,” intended to underscore the ease of renting VOD movies through the TV
remote control.28 In May, 2010, DirecTV began to offer movies on street date through VOD,
initially from Universal, Fox, and Warner Brothers. By the middle of 2010, the average
window between street date and digital availability date was five days (Gruenwedel, 2010).
At a price of $5 per rental, VOD has taken the place of brick and mortar rental as the high
priced streetdate rental, and it is a significantly more efficient method of delivery than its
predecessor.
Digital Sale and Rental
With total revenues of about $500 million in 2011,29 the digital sale and rental market
is small and growth prospects are uncertain because there are so many close substitutes.
Apple has always been the market leader, but now is facing increasing competition, including
Google, Amazon, Best Buy (Cinemanow), and Ultraviolet. Apple started to offer rentals from
all major studios in January, 2008 and sales from all major studios in May, 2008.30 Google
began offering online movie rentals through Youtube in January, 2010 with five independent
films from the Sundance film festival. They expanded to include Lionsgate on April 22,
27
Article available at http://blog.comcast.com/2009/06/day-and-date-vod-movie-release-windows-and-48hour-rentals.html (Last accessed October 20, 2012)
28
Article available at http://eon.businesswire.com/news/eon/20100317005555/en (Last accessed October 20,
2012)
29
Article available at http://appleinsider.com/articles/12/06/01/netflix online movie streaming revenue
explodes slices apple share in half (Last accessed October 20, 2012)
30
Article available at http://www.apple.com/pr/library/2008/01/15Apple-Premieres-iTunes-Movie-RentalsWith-All-Major-Film-Studios.html (Last accessed August 27, 2012)
14
2010,31 started to offer movies from Universal, Sony, and Fox on May 9, 2011,32 Disney on
November 23, 2011,33 and Paramount on April 4, 2012, bringing their total library to nearly
9,000 titles.34
A big reason for the slow growth of this market is the lack of a common format and
transferability of files. A consortium of firms led by Walmart and the five major distributors
apart from Disney35 has been trying to solve this problem by developing Ultraviolet, a
digital rights authentication and cloud-based licensing system that allows users to stream
and download purchased content to multiple platforms and devices.36 Accounts are free of
charge, and users are able to buy or rent movies, which they are then be able to play from
multiple devices, similar to the iTunes model.
Ultraviolet launched in the October, 2011. As a way of advertising, studios started to
include ultraviolet rights with the purchase of some DVDs and Blu-Rays in January, 2012.
By June, 2012, Ultraviolet had over 3 million registered accounts, and over 6,000 titles
enabled with the buy once, play anywhere cloud-based content ecosystem. It grew by about
1 million accounts from April to May, after Walmart launched its disc-to-digital initiative,
which allows consumers to bring all old DVDs in to be converted to an ultraviolet copy for
$2 each.37
Subscription Video on Demand
The subscription VOD market is dominated by Netflix, and Hulu is far behind in second. The product that these two companies offer is somewhere in between digital rental and
premium cable. These services are imperfect substitutes to other methods of movie consump31
Article available at http://www.homemediamagazine.com/vod/youtube-quietly-expands-movie-rentalservice-19183 (Last accessed April 18, 2012)
32
Article available at http://bits.blogs.nytimes.com/2011/05/09/youtube-expands-video-rental-service/
(Last accessed April 18, 2012)
33
Article available at http://www.deadline.com/2011/11/youtube-pacts-to-rent-100s-of-disney-filmsincluding-pixar-dreamworks/ (Last accessed April 18, 2012)
34
Article
available
at
http://www.reuters.com/article/2012/04/04/net-us-youtube-idUSBRE8330H
120120404 (Last accessed October 20, 2012)
35
Disney is developing its own competing product, which they call Keychest.
36
http://en.wikipedia.org/wiki/UltraViolet (system) (Last accessed June 6, 2012)
37
Article available at http://www.homemediamagazine.com/digital-evolution/ultraviolet-hits-3-millionaccounts-6000-titles-27590 (Last accessed October 20, 2012)
15
tion, and they have seen the fastest growth in the industry in 2011 and 2012. The content is
comparable to premium cable channels like HBO and Starz, the method of watching online
is comparable to Youtube and Apple, and the method of payment is unique: monthly subscription. One of the keys to Netflix’s success from 2008 to 2011 was that it had the rights to
Starz streaming content, including Disney and Sony movies, for approximately $30 million
a year. This translates to $0.23 per subscriber per month, compared to the estimated $4
per subscriber per month Starz gets for its premium movie service from MSOs. By the end
of the Starz contract in October, 2011, Netflix’s subscriber base had increased to about 24
million, so it was only paying $0.10 per subscriber per month for access to the same Starz
content that MSOs got: about 2,500 pieces, split evenly between TV and movies.38
During 2010 and 2011, it was becoming clear that Netflix would not be able to compete
with Redbox in the physical rental business due to the high cost of postage,39 so they started
to emphasize the SVOD segment of their company, most famously by splitting the company
into two separate operating segments and announcing and withdrawing plans to rename the
DVD-by-mail segment Qwikster.40
In late 2010, Netflix streaming had become so big that there was a debate about how to
deal with the bandwidth. Level 3 Communications, the largest competitive local exchange
carrier (CLEC) in North America, signed a deal with Netflix in November, 2010, to host and
deliver Netflix’s content to internet providers like Comcast. After signing the deal, Level 3
asked Comcast if they could send twice the amount of traffic as they had done before without
paying any extra. After a debate in the press41 Level 3 agreed to Comcast’s terms.
In March 2012, an analyst estimated that 70-80% of the time consumers spent on Netflix
38
The differences are that the MSO could offer content in HD, and the average MSO consumer spends
significantly more time watching Starz content than the average Netflix subscriber (Bond, 2011)
39
An analyst estimated that Netflix would spend $700 million in postage and $110 million in DVD acquisition
expenses for 2010 (Tribbey, 2010).
40
Article available at http://online.wsj.com/article/SB10001424052970203499704576622674082410578.html
(Last accessed October 20, 2012)
41
Article available at http://money.cnn.com/2010/11/30/technology/netflix level3 comcast traffic/index.htm
(Last accessed May 10, 2012)
16
streaming was watching TV shows, rather than movies.42 In early 2011, an analyst estimated
the value of the Netflix and Disney contract at $183 million per year, including $45 million
for all six seasons of Lost, $26 million for Scrubs $18 million for Hannah Montana, and $12
million each for Desperate Housewives and Wizards of Waverly Place. All six seasons of
Nip/Tuck (Time Warner) went to Netflix for $20 million a year (Bond, 2011).
In September, 2011, Netflix and DreamWorks announced an agreement that presumably
will replace any potential TV rights for Dreamworks movies, beginning in 2013. Analysts
estimate that the deal is worth $30 million per picture for an unspecified period of years. It
is the first time a major Hollywood supplier has chosen Web streaming over pay television.
In the past, HBO has paid steep licensing fees of about $20 million per picture for exclusive
rights a few months after films arrive on DVD.43 A television station paid $2 million per
episode for syndication rights to The Big Bang Theory, and Netflix wanted it for about
$100,000 per episode.44 For more comparisons, see section B.1.
Hulu launched in March, 2008 and rose to prominence as a free website, generating its
revenue through advertisements. Beginning in late 2010, they started to offer a subscription
based service, Hulu Plus, which is comparable to Netflix’s streaming service. Hulu has always
emphasized television content rather than movie content, but it is important to consider as a
force in the movie industry, as it is owned by NBCUniversal (32%), Fox (31%), ABC-Disney
(27%), and Providence Equity Partners (10%).
4
Model
In this section, I describe the model of consumer behavior. I begin with a simple model
in section 4.1, where movies are parameterized by two values: V1 , the utility the average
42
Article available at http://www.homemediamagazine.com/market-share/dvd-and-blu-ray-rentals-topped56-billion-2011-26778 (Last accessed October 20, 2012)
43
Article available at http://www.nytimes.com/2011/09/26/business/media/netflix-secures-streaming-dealwith-dreamworks.html?pagewanted=all (Last accessed April 12, 2012)
44
Source: “Studio Execs Differ on Netflix, Premium VOD,” Erik Gruenwedel and Chris Tribbey. Home
Media Magazine, December 13-19, 2010 issue
17
consumer derives from the first watch of a movie, and V0 , the additional utility the average
consumer gets from owning the movie. To illustrate the model, I first allow consumers to buy
or rent the movie in a one period world. Then I introduce decay parameters and extend the
model to multiple periods. I introduce the model in the specific context of the movie market,
but the same framework can be used to describe consumer behavior in any marketing setting
involving buying and renting decisions across multiple periods.
In section 4.2, I build a nested logit model, which will be the workhorse model for
the paper. I discuss each parameter of the model in detail in section 4.3 and run several
simulations to illustrate the model. In section 4.4, I introduce competition and seasonality
into the model, and I also move to a version of the model where V1 and V0 are functions of
movie characteristics. I discuss decay rates in section 4.5.
The model is inspired by Varian (2000), and has some features in common with Luan and
Sudhir (2006), Hennig-Thurau et al. (2007), Chiou (2008), and Rao (2011). It is important
to note the distinction between these models and those that are inspired by (Einav, 2007),
in which consumers are deciding between multiple movies during a given week.
4.1
Simple Model
Consider a consumer trying to decide whether to buy or rent a movie in a one period
model. Buying is superior to renting for three basic reasons:
1. If he rents the movie, he will only have access to it for a limited time (generally 24 or
48 hours), whereas if he buys it, he will own it forever.
2. Some consumers get utility from owning a library of DVDs and putting it on display
in their homes.
3. Often, there are special features that that are only available on for-sale copies of the
movies, such as interviews with cast members and deleted scenes.
18
Let UiB be consumer i’s value of buying the movie, UiR be his value of renting it, and Ui0 be
his value of not consuming the movie. Decompose these numbers as follows:
UiB = V0 + V1 + B
i
UiR = V1 + R
i
Ui0 = 0i
Written this way, V1 is the average consumer’s utility from seeing the movie at home once,45
and V0 is the additional utility from owning the movie. In section 4.4, I write V0 and V1
as functions of a vector of observable movie characteristics, X. For now, a movie is only
R
0
represented by its values of V0 and V1 . B
i , i and i are idiosyncratic shocks, which I discuss
in detail in sections 4.2 and 4.3. This is a simple and convenient framework for the consumer
decision problem, and everything in the model will stem from here.46 Let P B be the price
of buying and P R be the price of renting. It is straightforward to show that the consumer’s
solution is as follows:
R
0
B
Buy if B
> P B − P R − V0 and B
i − i
i − i > P − V0 − V1
R
0
R
Rent if B
< P B − P R − V0 and R
i − i
i − i > P − V1
and otherwise do not consume the movie.
Inter temporal substitution is a major concern in this paper and the movie industry as a
whole, so I now extend the model to multiple periods. I do this by assuming two decay
rates, λB and λR , which represent the average decay in utility of buying and renting from
45
This assumes that all consumers only watch rented movies once. More precisely, it is the utility he derives
from renting the movie
46
More complicated ways of decomposing the buy or rent decision are beyond the scope of this paper and
would require a more detailed dataset than I have available. One very interesting question, for example,
is how much having seen the movie in theaters affects a consumer’s UiB and UiR . With a panel dataset,
I could make these both functions of having seen the movie, as Rao (2011) does with her survey data.
However, as I only have access to aggregate data, I leave V0 and V1 as functions of movie characteristics
and allow all the consumer heterogeneity to present itself in the error.
19
one period to the next over all movies and weeks, and introducing error terms ξtB and ξtR ,
which should be interpreted as movie-week deviations from the common decay patterns.47
Assume there are a finite number of periods during which the movie is available for sale and
rental: the movie is available for sale from t = T1B , ..., TNB , and the movie is available for
rental from t = T1R , ..., TNR . The consumer is only allowed to consume the movie zero or one
times. For example, he cannot rent the movie in period 1, then buy or rerent it in period
2.48
The consumer can buy or rent the movie in whichever period he pleases and get the
associated utility:
B
B
UitB = V0 + V1 − λB t + ξtB + B
it , t ∈ {T1 , ..., TN }
(4.1)
R
R
UitR = V1 − λR t + ξtR + R
it , t ∈ {T1 , ..., TN }
(4.2)
Ui0 = 0i
The consumer’s problem can be written as:
max{
max
B}
t∈{T1B ,...,TN
max
B
V0 + V1 − λB t + ξtB + B
it − Pt ,
R}
t∈{T1R ,...,TN
R 0
V1 − λR t + ξtR + R
it − Pt , i }
Note that the movie-week deviations, idiosyncratic time preference shocks, and prices are
all indexed by t, where t is equal to the number of periods that have passed since the DVD
release date. In sections 4.2 and 4.4, I develop the model by adding the following features:
1. putting a nested logit structure on the error term (section 4.2)
2. writing utility as a function of the log of price, so that the coefficient on log(price) can
47
There are several alternate ways to incorporate decay into the model. See section 4.5 for a detailed
discussion on decay rates.
48
This is a common approach in the discrete choice literature, as it greatly simplifies the math and estimation.
This model is appropriate for the purposes of this paper, as evidence suggests that the only common type of
repeat movie consumption is box office then second stage consumption; repeat consumption in the second
stage is rare. See Rob and Waldfogel (2007) for a discussion.
20
be interpreted as a price elasticity (section 4.2)
3. including observed demand shocks for each period to account for seasonality and competition (section 4.4)
4. expressing V0 and V1 as functions of a vector of movie characteristics, X (section 4.4)
Before moving to the full model, it is useful to consider the example in Appendix B, section
B.4 to understand some of the tradeoffs consumers and distributors will face.
4.2
Nested Logit Model
In this section, I build upon the model introduced in section 4.1 and the example from
Appendix B, section B.4 to create a nested logit model for buying and renting movies. For
this section, utility from buying, renting, and nonconsumption will be expressed as:
V0 + V1 − λB t − αp ln pB
+ ξtB
t
B
B
+ B
=
it , t ∈ {T1 , ..., TN }
µ
V1 − λR t − αp ln pR
+ ξtR
t
R
R
R R
+ R
Uit (pt ) =
it , t ∈ {T1 , ..., TN }
µ
UitB (pB
t )
(4.3)
(4.4)
Ui0 = 0i
These are the same as expressions (4.1) and (4.2) in section 4.1, with two exceptions. First,
the log of price enters into utility, which is a constant price elasticity assumption where the
elasticity is αp . Second, I have divided everything but the idiosyncratic error term by a
scale parameter, µ. Having µ as a free parameter allows flexibility in the variance of the
idiosyncratic error term.49
The idiosyncratic error structure is a very important feature of the model, so it is important that the reader fully understands it before moving on. To make the model realistic, I
49
See Train (2009), chapter 3.2 and Appendix B, section B.5
21
use the following nested logit structure:
B
B
B
B
B
it = ζi + (1 − σ) νit , t ∈ {T1 , ..., TN }
R
R
R
R
R
it = ζi + (1 − σ) νit , t ∈ {T1 , ..., TN }
0i = ζi0 + (1 − σ) νi0
where νitg , g ∈ {B, R} and νi0 are distributed type 1 extreme value, and ζig , g ∈ {B, R, 0}
have distributions C(1 − σ) such that git , g ∈ {B, R} and 0i are also distributed type 1
extreme value. See Cardell (1997) and Berry (1994). The unobservable ζiB , ζiR , and ζi0
represent people’s idiosyncratic preferences for buying, renting, and nonconsumption for the
movie. There are two fundamental segments of consumer attributes these terms are meant
to capture. The first segment is the consumer’s relationship to the movie, which includes
whether he saw it in theaters, how much he enjoyed seeing it in theaters, whether he saw
a trailer, and anything his friends and family told him about the movie before the DVD
release date. The second is the consumer’s general movie consumption preferences. In the
words of Thomas K. Arnold, publisher of Home Media Magazine, there are three classes of
consumers of home media: 1. collectors, 2. minimalists, and 3. “quick and easy.” Collectors
enjoy owning movies on disc, organizing them on shelves, and showing them off with pride
to guests. Minimalists abhor clutter, and they were the first to embrace video rental, on
demand TV, and Netflix streaming. The third group prefers the path of least resistance. If
they happen to be in Walmart, they’ll pick up a cheap movie or two in the $5 dump bin;
if, after about 20 seconds of browsing, they see nothing they like, they’ll check out the new
releases. If there is nothing there, they’ll stop by the Redbox on the way out.50
The unobservable νitB , νitR , and νi0 represent idiosyncratic time shocks. These capture
things such as a consumer’s busy schedule, movie date nights, or a movie loving friend’s
birthday. σ, restricted such that σ ∈ (0, 1), is known as the correlation coefficient. As σ
50
Article available at http://www.homemediamagazine.com/thomas-k-arnold/three-faces-consumerism
(Last accessed September 21, 2012)
22
approaches 1, within group correlation of errors becomes stronger. Higher sigma values mean
that people’s preferences for buying vs. renting are strong relative to their preference shocks
each period, which gives rise to more inter temporal substitution rather than substitution between buying and renting. As sigma approaches zero, the model approaches the simple logit
model, which suffers from the IIA (Independence of irrelevant alternatives) assumption.51
For readers unfamiliar with the IIA assumption, It is useful to visualize the nested logit
model in a tree structure, depicted in Figure 4. Suppose given a set of rental and sale
prices in each available period, there is some consumer whose utility maximizing option is
to purchase the movie in period k. Now suppose that the distributor increases PkB enough
that this is no longer the consumer’s best option. If σ is high, the consumer is more likely
to substitute to buying the movie in another period, rather than substitute to renting. If σ
is 0, the consumer is equally likely to substitute to any other option.
The above expressions for utility and the nested logit error structure imply the following
equations for market shares:
ξB
V0 + V1 − λB t − αp ln pB
t
+ σ ln st|B + t
ln (sBt ) − ln (s0 ) =
µ
µ
R
R
R
ξ
V1 − λ t − αp ln pt
ln (sRt ) − ln (s0 ) =
+ σ ln st|R + t
µ
µ
(4.5)
(4.6)
See Appendix B, section B.5 for a full derivation of the market shares. In the N period
model, these equations and the adding up constraint yield a system of 2N+1 equations and
51
Readers unfamiliar with logit models will want to consult Train (2009), especially chapters 3 and 4.
23
2N+1 unknown market shares. In the two period model, the five equations are:
ln (sB1 ) − ln (s0 ) =
ln (sR1 ) − ln (s0 ) =
ln (sB2 ) − ln (s0 ) =
ln (sR2 ) − ln (s0 ) =
B
V0 + V1 − αp ln pB
sB1
ξ
1
+ σ ln
+ 1
(4.7)
µ
sB1 + sB2
µ
R
V1 − αp ln pR
sR1
ξ
1
+ σ ln
+ 1
(4.8)
µ
sR1 + sR2
µ
B
V0 + V1 − λB − αp ln pB
sB2
ξ
2
+ σ ln
+ 2
(4.9)
µ
sB1 + sB2
µ
R
V1 − λR − αp ln pR
sR2
ξ
2
+ σ ln
+ 2
(4.10)
µ
sR1 + sR2
µ
sB1 + sB2 + sR1 + sR2 + s0 = 1
(4.11)
This two period model is a useful analytical tool for two reasons. First, it is easier to
conceptualize tradeoffs between two periods rather than N periods. Second, the two model
is a good approximation of distributor windowing strategies, which is the current state of
the movie industry. The convenience of the nested logit model is that it allows me to write
the market shares as functions of independent variables and one endogenous variable, the
within group share. Given a suitable instrument, I can estimate the model using IV, which
I do in section ??.
4.3
Simulations
To understand some of the nuances of this model, I will now run through some simulations. There are seven parameters of the model, and four prices that the producers are able
to choose. Before looking deeply into the data, I will calibrate the model using observed
prices and sample means. The observed prices, as described above, are P1B = P2B = 17,
P1R = 2 or 5, and P2R = 2. First, consider the two parameters that describe a movie: V0 ,
V1 , and the two decay rates: λB and λR . A reasonable starting point might be to think of
the consumer at the center of the distribution as being indifferent between all consumption
and nonconsumption options, similar to where we started in section 4.2. For this to be the
case, we would set V1 = log(5), V0 = log(17) − log(5), λR = log(5) − log(2), and λB = 0.
24
With these parameters, approximately 20% of the population would choose each of the five
options. However, this would not be an accurate depiction of reality at all. The average
movie in my sample is rented approximately four million times in the first four weeks and
two million times in weeks five and later. In addition, approximately two million consumers
buy the movie in the first four weeks, and another 500,000 buy the movie in weeks five and
later. The tricky unknown question, which turns out to be a very important question, is
what the overall market size should be. There are approximately 115 million households in
the United States, and approximately 88% have a DVD player.52 Therefore, the maximum
market size might be 100 million households. I will run simulations with a market size of
25, 50, and 100 million households.
I will begin with a market size assumption of 50 million households. This means that
the market shares for the average movie are sB1 = 4%, sB2 = 1%, sR1 = 8%, sR2 = 4%, and
s0 = 83%. I will start with σ = .5, µ = 1, and αP = 1. The remainder of this section is a
description of what happens as each parameter changes. Accompanying figures are included
in the Appendix A.
µ ≈ variance of the error terms
As µ increases, the individuals’ error terms become more meaningful, and the four products look more similar to each other. As µ → ∞, all four the market shares converge to the
same value. Because the market shares are small, the consumers purchasing and renting are
to the far right of the distribution. Thus, the market shares are mostly increasing in µ, the
outside share is certainly decreasing in µ, so that is why profits are increasing in µ.
As µ gets larger, the windowing strategy becomes more attractive. For small values of
µ, there are so few people consuming anything, so the early release strategy induces people
to consume the cheapest version of the product and it dominates. For higher values of µ,
the early release strategy has a strong market stealing effect, and less of a market expand52
Article available at http://blog.nielsen.com/nielsenwire/media entertainment/more-than-half-the-homesin-us-have-three-or-more-tvs/ (Last accessed June 27, 2012)
25
ing effect, so windowing dominates. µ captures the following: “The more heterogeneous
consumers are, the more attractive windowing is.”
σ ≈ correlation of within group errors
As σ increases, the correlation of the within group error terms increases. In the two period
B
R R
model, that means that corr(B
i1 , i2 ) and corr(i1 , i2 ) are increasing. Low σ means that we
are closer to a multinomial logit model, higher σ means that we are closer to the world
where buyers and renters are “two separate groups.” As σ increases, profits are decreasing,
because the share of the outside good is increasing. As σ increases, the early release strategy
becomes more attractive for a time, because the buyers will still buy, so the market expanding
is outweighing the fact that rentals are less expensive. When σ is very high, however, the
windowing strategy dominates again, because for high σ, market expanding is small. A lot
of this depends on the price elasticity also.
αp ≈ price elasticity
The higher αp is, the more sensitive to prices consumers are. Profits are decreasing in
αp , as are all market shares aside from that of the least expensive option. When αp is low,
windowing is more attractive, as more consumers are willing to pay for the higher priced first
period rental and purchase. For higher values of αp , a large share of consumers are already
waiting for the low priced rental in the second period, so the early release strategy does not
have a major market stealing effect, and market expanding outweighs market stealing.
λR ≈ disutility of waiting until second period to rent
For very low values of λR , a lot of consumers are willing to wait until the second period
to rent, so the windowing strategy does not work all that well. As λR increases, windowing
becomes more attractive. After a certain point, windowing profits are increasing in λR , while
early release profits are decreasing in λR . It is interesting that for very low values of λR , the
26
windowing strategy is working so poorly that profits are decreasing in λR . Note the graphs
as λR varies for different values of σ. Focus only on the solid lines, which represent market
shares with a windowing strategy. When σ=.75 and λR increases, most of the substitution
is from renting in period 2 to renting in period 1. When σ=.5, there is a lot of substitution
from period 2 renting to period 1 renting, but there is some substitution to buying. When
σ=.25, there is a good amount of substitution to all three alternatives. Also note that s0
increases at a faster rate when σ is lower
λB ≈ disutility of waiting until second period to buy
Profits are decreasing in λB , because as λB increases, people are less willing to wait to buy
the movie, so the period 2 buyers substitute to the other four options. Some buy in period
1, but others rent or drop out of the market, which mark decreases in profits. Windowing
is more attractive for low values of λB , because there are more period 2 buyers who will
convert to period 1 renters. For high values of λB , there are few buyers in the market, so
market expanding outweighs market stealing.
4.4
Movie Characteristics, Seasonality, and Competition
Movie Characteristics
Until now, I have left V1 and V0 , the average first watch utility and additional utility
from owning the movie respectively, as unexplained and seemingly arbitrary values. In fact,
they are highly dependent on movie characteristics and vary a lot from movie to movie. For
example, G and PG movies earn on average 30.20% of revenues from DVD and Blu-ray sales
and 16.69% of revenues from rentals, while R rated movies earn only 23.22% from sales and
38.41% from rentals. The likely explanation is that the consumers of G and PG movies are
parents with children who want to see the movies again and again, while the consumers of R
rated movies generally only want to see movies once. Table 7 shows the revenue breakdown
by movie characteristics for two different cuts of the overall sample.
27
Let X be a vector of observable movie characteristics that are known before the DVD
release date and remain unchanged throughout the DVD sale and rental period. X includes
log(US box office revenue), log(first week box office revenue)53 , MPAA rating (PG, PG-13,
or R)54 , genre, log(production budget), log(the maximum number of theaters in which the
movie played during its theatrical run), user reviews, gap (between box office wide release
date and DVD release date), and gap squared.
I obtained user reviews from rottentomatoes, IMDB, metacritic, yahoo, and fandango. I
use the data from rottentomatoes in my estimation because it covers the most movies in the
sample and movies are rated on several different metrics. There are three types of people
who give ratings: critics, top critics, and fans. The six metrics that I use from their website
are 1. all critics log(number of reviews), 2. all critics average rating on a scale from 1 to
10, 3. all critics percent liked (percent that gave the movie “yes” on a yes or no scale), 4.
fan log(number of reviews), 5. fan average rating on a scale from 1 to 5, and 6. fan percent
liked.
There is a decent industry and academic literature on the gap between box office and
DVD release dates, including Luan and Sudhir (2006). The gap is endogenously determined
by the distributor, which means there is a potential problem in estimation if the gap is
correlated with the error term. In reality, the gap is mostly determined by seasonality
and coordination with other DVD releases, and it seems to be uncorrelated with other
observables. Furthermore, estimates imply that the gap has little sale and rental revenues,
conditional on other observables. Thus, I find it unnecessary to instrument for gap.
One of the assumptions of my model is that DVD demand can be influenced by the
gap, but the gap has no effect on box office demand. This might seem problematic, as we
often hear of consumers “waiting for a movie to come out on DVD.” There is an ongoing
debate between distributors and theater owners about this issue, and I argue in this paper
53
To enable inclusion of both Friday and Wednesday releases, I use first week revenue rather than first
weekend revenue
54
G and Open movies are included in PG, NC-17 movies are included in R
28
that within the 16-20 week period that most movies fall in, variation in gap has little to no
impact on the theatrical demand. For the sake of argument, consider a model where the
gap affects theatrical demand. A consumer is deciding whether to see a movie during its
first week in theatrical release, its second week in theatrical release, or not see it in theaters
at all. The consumer must first forecast the DVD release date, which on average is 17-18
weeks after the box office release and usually announced towards the end of the movie’s
theatrical run. In an extreme case, her forecast of the gap might be 16 weeks rather than 20,
and this would marginally reduce her demand for seeing the movie in theaters but probably
not alter it significantly. If in reality gaps were short enough or variation was large enough
to influence theatrical demand, this would be a feature worth considering in the model.
Furthermore, even if the gap had an effect on box office revenue, it would be extremely
difficult to estimate. What matters is not the actual gap, which is unknown ex-ante, but the
consumer expectations of gap, which are unobservable to the econometrician.
When movie characteristics are incorporated into the model, V1 and V0 should be thought
of as
V0 (X) = ν00 (log(US B.O. rev)) + ν01 (log(first week B.O. rev)) + ν02 (MPAA dummy)
+ ν03 (genre dummy) + ν04 (log(prod. budget)) + ν05 (log(max theaters))
(4.12)
+ ν06 (user reviews variables) + ν07 (gap) + ν08 (gap squared)
V1 (X) = ν10 (log(US B.O. rev)) + ν11 (log(first week B.O. rev)) + ν12 (MPAA dummy)
+ ν13 (genre dummy) + ν14 (log(prod. budget)) + ν15 (log(max theaters))
(4.13)
+ ν16 (user reviews variables) + ν17 (gap) + ν18 (gap squared)
Seasonality
There is tremendous seasonality in the movie industry, especially in the box office and
DVD and Blu-ray sales. I do not find any noticeable pattern in seasonality for rentals. My
29
panel is not long enough to estimate seasonality using weekly dummies, as in (Einav, 2007),
so I construct SEASON variables to measure the impact of the holiday season on sales and
rentals. All of the observable seasonality in DVD and Blu-ray sales occurs during weeks 51
through 55 of the 56 week calendar year,55 which are the weeks between Thanksgiving and
Christmas. I assume that there is an identical demand shock to the market in weeks 51
through 55, so when estimating the model using a period length of one week, incorporating
seasonality is as simple as inserting a dummy equal to one for weeks 51 through 55. To
estimate the two period model, I assume that a movie makes roughly
revenue during week 1,
1
4
during week 2, and
assumption that a movie makes roughly
during week 6, and
1
8
1
2
1
8
1
2
of its first period
during weeks 3 and 4. I make the similar
of its second period revenue during week 5,
1
4
during weeks 7 and 8. These assumptions yield the simple and very
nice pattern of SEASON variables shown in Table C5. SEASON1 is meant to capture the
full effect that a holiday release can have on sales and rentals during period 1, and SEASON2
captures the full effect on period 2. For example, a movie that is released during week 51
or 52 gets a SEASON1 value of 1 because all of its first four weeks fall during the holiday
season, while a movie released during week 55, which usually ends right around Christmas
day, gets a SEASON1 value of
1
2
because it will only experience its first week during the
holiday rush.
Competition
To account for competition from other movies, I construct two sets of COMP variables:
COMP BO to account for box office competition and COMP DVD to account for competition from other DVDs.56 The COMP BO variable allows the demand for a DVD to be
influenced by the set of movies that are playing at the box office while the DVD is accumulating revenue. To properly account for all of the box office competition in week t, I first
P
construct a set of variables, WEEK COMP BOt = RBO =t BO REVj , (where RjBO is the
j
55
See Einav (2007), which introduced a 56 week calendar year to incorporate holidays. In the 56 week year,
Thanksgiving is defined to be the Thursday of week 51.
56
Movies playing at the box office today will not appear on DVD for another four months, so there is no
possibility for a contemporaneous effect of a movie’s own box office revenue on its rental and sale revenue.
30
box office release week of movie j), i.e. the sum of total box office revenue of all movies
released during week t. Then, I use these variables to construct another set of variables,
ALL COMP BOt =
1
1
(WEEK COMP BOt−2 ) + (WEEK COMP BOt−1 )
4
2
+ WEEK COMP BOt
+
(4.14)
1
1
(WEEK COMP BOt+1 ) + (WEEK COMP BOt+2 )
2
4
In this way, I account for strong box office releases that happened within two weeks of the
DVD release. The logic behind this method is that a movie released at the box office during
the same week a DVD is released will have the largest effect, but box office releases close to
the DVD release will also have an effect.
Similarly, I construct WEEK COMP DVDt =
P
RjDV D =t
BO REVj to be the sum of total
box office revenue of all movies with DVD release dates during week t. I use the box office
revenue rather than the DVD revenue to construct this variable because DVD revenue is
endogenous to the model. Then I construct
ALL COMP DVDtj =
1
1
(WEEK COMP DVDt−2 ) + (WEEK COMP DVDt−1 )
4
2
+ WEEK COMP DVDt − BO REVj
+
(4.15)
1
1
(WEEK COMP DVDt+1 ) + (WEEK COMP DVDt+2 )
2
4
When the period length is one week, ALL COMP BOt and ALL COMP DVDtj are included
as demand shifting COMP variables for week t and movie j. When estimating the two period
model, I want to account separately for competition during periods 1 and 2. I construct the
31
following variables:
1
ALL COMP BO 1t = ALL COMP BO(t) + ALL COMP BO(t+1)
(4.16)
2
1
1
ALL COMP BO(t+2) + ALL COMP BO(t+3)
+
4
8
1
ALL COMP BO 2t = ALL COMP BO(t+4) + ALL COMP BO(t+5)
(4.17)
2
1
1
+
ALL COMP BO(t+6) + ALL COMP BO(t+7)
4
8
1
(4.18)
ALL COMP DVD 1tj = ALL COMP DVD(t)j + ALL COMP DVD(t+1)j
2
1
1
+
ALL COMP DVD(t+2)j + ALL COMP DVD(t+3)j
4
8
1
ALL COMP DVD 2tj = ALL COMP DVD(t+4)j + ALL COMP DVD(t+5)j (4.19)
2
1
1
ALL COMP DVD(t+6)j + ALL COMP DVD(t+7)j
+
4
8
Note that a movie’s own box office revenue gets subtracted 1 time in ALL COMP DVD(t)j ,
1
1
times in ALL COMP DVD(t+1)j , and 14 14 times in ALL COMP DVD(t+2)j , for a
2
2
5
total of 1 16
times.
Here is a good point to emphasize the difference between my model and models that stem
from Einav (2007). In my model, for each movie that comes out, each consumer chooses a
method of consumption and a period during which to consume. A consumer’s choices are
independent of each other in a sense,57 and a consumer with very high shocks in a given
period could theoretically buy one copy of every DVD in release that period. In Einav
(2007), consumers choose to see zero or one movies each period, but could theoretically see
the same movie in theaters during each week the movie is playing. The real world is a mix
of both, so each model has its limitations. People sometimes see more than one movie in
theaters during a week, and people sometimes rent the same movie twice or rent and buy
the same movie. That being said, Einav’s model seems to be a very good proxy for the
57
Consumers are not identified in any way except for their shocks. In a world where consumers are budget
or time constrained, it is reasonable to think of a consumer who is high in the distribution for one movie
or one period as being low in the distribution for a different movie that period or for the next period.
32
consumer decision to see movies in theaters, and I believe my model is a very good proxy
for the consumer decision to rent and buy movies on DVD and Blu-ray.
When movie characteristics, seasonality, and competition are accounted for, the utility
that consumer i derives from buying or renting movie j in period t go from expressions (4.3)
and (4.4) to:
B
Uijt
(Xj , ΩB
tj )
B
V0 (Xj ) + V1 (Xj ) − λB t − αp ln pB
+ ΩB
t
tj + ξt
B
B
+ B
=
ijt , t ∈ {T1 , ..., TN }
µ
(4.20)
R
R
Uijt
(Xj , ΩR
tj ) =
V1 (Xj ) − λR t − αp ln pt
µ
R
+ ΩR
tj + ξt
R
R
+ R
ijt , t ∈ {T1 , ..., TN }
(4.21)
and the market shares in the N period model become:
B
B
ξB
V
(X
)
+
V
(X
)
−
λ
t
−
α
ln
p
+ ΩB
0
j
1
j
p
t
tj
ln (s0Bt ) − ln (s00 ) =
+ σ ln s0t|B + t (4.22)
µ
µ
R
R
R
ξtR
V1 (Xj ) − λ t − αp ln pt Ωtj
0
0
0
ln (sRt ) − ln (s0 ) =
+ σ ln st|R +
(4.23)
µ
µ
where
B
B
B
ΩB
tj = ω1 ALL COMP BOt + ω2 ALL COMP DVDtj + ω3 SEASONt
R
R
R
ΩR
tj = ω1 ALL COMP BOt + ω2 ALL COMP DVDtj + ω3 SEASONt
4.5
Decay Rates
One of the key identifying assumptions of the model is that there is a homogeneous decay
in utility of renting and buying from period to period, which is captured by the parameters
λB and λR . First, I discuss the reasons for decay in average utility from buying and renting,
then I defend the assumption of a homogeneous decay rate. There are two distinct reasons
for a decay in buying and renting utility. The first is that people generally do not buy or
33
rent the same movie twice, so everyone who has seen the movie prior to period t (the highest
value consumers) is not in the distribution for week t, lowering the average utility of the
remaining consumers. Second, there is an advertising component: consumers like being the
first to buy or rent movies, when there is a lot of buzz about the DVD release, commercials,
talk show appearances, billboards, etc. Distributor and word-of-mouth advertising are both
very powerful in the movie industry, and both are much more prevalent early in a movie’s
DVD run.58 Without a more detailed micro dataset, I cannot separate these two effects.
It is possible that there are heterogeneous decay rates, so I ran a robustness check allowing
decay rate to vary by MPAA rating and genre, neither of which seemed to affect decay rate.
5
Data
The data collection process for this paper was very time consuming and tedious, and the
end result is a comprehensive dataset on weekly box office, DVD sale, Blu-ray sale, physical
rental, and Video on Demand units and revenues from early 2006 to present day. The
dataset includes numerous descriptive characteristics described in section 5.1, movie reviews
from various websites, and information on release dates for rental and retail companies. I
was able to collect most of the data by webscraping, but much of the data needed to be
manually transcribed, and merging datasets was often tedious. In section 5.1, I describe
the data from The Numbers, section 5.2 gives an account of the DVD and Blu-ray sales
data from Home Media Magazine (HMM), section 5.3 gives an account of the rental data
from HMM, and section 5.4 gives a description of the Video on Demand data. A large
portion of the Blu-ray dataset does not include units or prices, so in section C.1, I describe
how I transformed the original Blu-ray dataset into the dataset I used in the estimation.
Similarly, much of the original rental and VOD datasets do not include units or prices,
58
When modeling decay in box office utility, there is a third reason which should be considered, which is a
network effect. The experience of viewing a movie in a crowded theater during opening weekend is very
different from seeing it in an empty one on Tuesday night during week 4. The week in release could be a
good proxy for the number of other consumers in the theater, which could affect the average consumer’s
utility.
34
so in sections C.2 and C.3, I describe how I transformed the original datasets into the
datasets I use in the estimation. Appendix C is available in a separate file on my website at
http://home.uchicago.edu/ mferri/mf research.html.
5.1
The Numbers
The Numbers (www.the-numbers.com) is one of the leading websites that publishes movie
industry data, and it is maintained by Nash Information Services. I obtained several datasets
from The Numbers, including movie descriptive characteristics, box office data, DVD revenue and units data, and Blu Ray revenue and units data. The descriptive characteristics
include but are not limited to release date, production budget, MPAA rating, running time,
keywords, distributor, source, genre, country of production, production method, creative
type, and a list of castmembers. U.S. domestic box office data has been public for years,
and it is available at the weekly and daily levels. The dataset also includes international
weekend box office data. For each period, the box office dataset includes revenue and number
of theaters in which the movie was playing.
The Numbers has a separate dataset of DVD sales data. Starting February 12, 2006,
The Numbers lists the top 30 selling DVDs on a weekly basis (Monday through Sunday).
This dataset includes estimated weekly units, weekly revenue, total units, total revenue,
and the number of weeks in release for the top 30 titles. According to Nash Information
Services, “Precise information on DVD sales is not generally available. Our DVD sales
figures are estimates based on studio figures, publicly available data, and private research on
retail sales carried out by Nash Information Services. The figures include estimated sales at
Wal-Mart and other retailers that do not publicly release sales information.”
The third dataset from The Numbers is the top 10 selling Blu-ray discs each week. The
variables in the dataset are the same as those in the DVD dataset. The Numbers does
not have a public archive of the Blu-ray dataset, so I only have the dataset each week
starting January 8, 2012. I was able to obtain five random weeks from May 1, 2011 through
35
September 18, 2011 through an internet archive. See Table C1 for a complete description of
the availability of DVD and Blu-ray data.
5.2
Home Media Magazine DVD and Blu-ray Sales Data
Home Media Magazine (HMM) is one of the leading sources for data and analysis of the
home media market. I obtain several datasets from HMM, details of which are shown in
Table C1. HMM uses figures from Nielsen VideoScan and conducts some of its own internal
research to compute its estimates. The magazine publishes weekly charts of combined sales
(DVD+Blu-ray) and Blu-ray sales, examples of which are in Appendix C as Figures C2 and
C3. The lists generally include the top 20 in each category, sometimes include up to 50,
and sometimes (very few times) only include the top ten. See Table C1 for details on which
dates are included in the dataset and how many titles are included each week. In addition
to publishing weekly charts, HMM publishes a weekly report of the entire market, often
displayed as one or more infographics. An example infographic is available as Figure C1.
This report includes total units and consumer spending on DVDs and Blu-rays during the
week. The data is often expressed as this year’s number and percent changes from that week
the previous year. Finally, HMM often publishes quarterly charts on the weekly Blu-ray
units sold each week for the quarter.
The weekly charts are available in html format starting May 15, 2011 for the combined
chart and February 20, 2011 for the Blu-ray chart. For weeks before these dates, a research
assistant and I transcribed each chart from the online magazine. Html versions of the weekly
report are available starting December 10, 2011, and we manually transcribed this data for
weeks prior to that date. HMM sometimes releases corrections of past data, and I thank
Kosty59 for compiling these updated numbers in his forum.
Unlike The Numbers, HMM does not publish revenue or units for each title-week. Instead,
the magazine publishes the “index” of each title, which is the units sold as a percentage of
59
Data available at http://forums.highdefdigest.com/high-definition-smackdown/112110-blu-ray-salescharts-graphs-reference-temporary-process-updates-thread.html (Last accessed September 21, 2012)
36
the top selling title. There are three important pieces of information that I pulled from the
HMM charts for each title-week: the combined index (DVD+Blu-ray units as a percentage
of the top selling title that week), the Blu-ray share (Blu-ray units divided by combined
units for each title-week), and the Blu-ray index (Blu-ray units as a percentage of the top
selling title that week). Where I do not have data on Blu-ray units and revenues, I impute it
using the data I have available. I present details of the imputation process in section C.1. As
shown in the last row of Table C2, I am able to extend my sample of Blu-ray units estimates
from 340 to 10,040 title-week observations, and DVD units estimates from 11,103 to 29,162
title-week observations.
5.3
Home Media Magazine Physical Rental Data
The rental data collection process is similar to the DVD and Blu-ray sale data collection
process, and I impute part of the dataset using a different imputation method than I use for
the Blu-ray data. Details of the imputation are in Appendix C, section C.2. Home Media
Magazine publishes weekly charts of the most rented titles, an example of which is shown in
Figure C4. Details on the raw data are available in Table C1. Depending on the time period,
HMM publishes different parts of the dataset, and it is sometimes published in places other
than Home Media Magazine.
The first part of the dataset, which I obtain from IMDB60 and Boxofficemojo,61 includes
the rental indeces of the top 50 titles each week from September 5, 1999 through May 9,
2010. For over two years (April 9, 2006 through June 1, 2008), the revenues of the top 50
titles each week are also available. From May 16, 2010 through October 3, 2010, HMM
published the indeces of the top 20 titles, and I obtain this data from IMDB. Starting
October 10, 2010, HMM stopped publishing the rental indeces, but continued to publish the
top 20 titles in order from 1 through 20. This data is available on IMDB and also on HMM’s
website. In addition, HMM publishes the total weekly rental revenue for the entire market
60
61
http://www.imdb.com/boxoffice/rentals
http://boxofficemojo.com/dvdrentals/ (This data is no longer available on boxofficemojo)
37
in an infographic from February 5, 2006 through October 3, 2010, also shown in Figure
C4. Where the data is not available in html format, a research assistant and I manually
transcribe the data.
The original dataset includes the top 20 or 50 rented titles from all weeks from September
5, 1999 through July 22, 2012, with the exceptions of December 24, 2000, December 22, 2002,
December 28, 2003, April 3, 2005, April 10, 2005, December 19, 2010, and January 23, 2011.
I use the weekly total revenue and title-week observations of index and rank to impute
revenue and units for 88,446 title-week observations, as shown in Table C3. Details of the
imputation are available in section C.2.
5.4
Home Media Magazine Video On Demand Data
The raw Video on Demand data comes from Rentrak, via Home Media Magazine. In
some issues, the magazine publishes a list of the top 10 or top 5 VOD titles by units rented.
I use the same basic imputation method that I used for the physical rental data to impute
units and revenue for 9,252 title-week observations. Details of the imputation are available
in section C.3.
6
Results
The general question that motivated this paper is: “When a distributor releases a movie
for sale and rental across several different platforms, what is the optimal release and pricing
strategy?” The distributors only control a small share of downstream exhibition, so they do
not have direct control over most pricing decisions. Furthermore, there is extremely little
price variation in the data, and the price variation that does exist is highly endogenous
and limited to DVD and Blu-ray sales. For these reasons, I shift the focus of the empirical
analysis to optimal release strategy.
There is exogenous variation in Netflix and Redbox availability starting in early 2010
38
due to external forces such as contracts with HBO and Starz. This exogenous variation
provides an excellent environment for analyzing optimal release strategy under certain constraints. Furthermore, by studying substitution patterns in the data, I can provide insight
into consumer preferences over renting and buying movies.
The dependent variables I have available are weekly units for the four major downstream
revenue sources during the DVD release period: physical rentals, DVD sales, Blu-ray sales,
and VOD rentals. The most obvious question to ask first is: “When a movie is delayed at
Netflix and Redbox for four weeks, what happens to revenues from each of these sources?”
I estimate the simple model:
log (unitsgτ ) = 1{Window}βWindow + (6.1)
for g ∈ {Physical Rent, DVD, Blu-ray VOD}, τ ∈ {1(weeks 1-4), 2(weeks 5-8)},62 and I present
results in column 1 of Tables 10a through 10d. I limit the sample to titles that had street
date releases on both Netflix and Redbox or four week delays on both Netflix and Redbox
from January 1, 2010 through July 22, 2012, which is almost all movies that were released
during this time period.63
A window has a huge negative impact on physical units rented in period 1 and a huge
positive impact on physical units rented during period 2, which is exactly what is expected.
The interesting result is that the window does not affect DVD sales, Blu-ray sales, or VOD
rentals during periods 1 or 2 at all.
I refine the analysis in column 2 by adding a time trend and allowing the effect of the
window to vary with time:
log (unitsgτ ) = 1{Window}βWindow + tβt + (1{Window} x t) βW x t + 62
(6.2)
I ran alternate specifications where period 2 was weeks 5 through 10 and 5 through ∞. The coefficients
are unaffected, but the standard errors on some variables increase.
63
For titles that appeared in some charts but never appeared in others, I assign to them 90% of the minimum
observed value in the dataset. This is better than excluding them from the analysis, which would bias the
results.
39
The time trends have the expected signs in all regressions, and the Window x t trend had
an effect on physical rentals, but small and insignificant effects on the other three revenue
sources. The magnitude of the Window x t trend for physical rentals is huge, which is
expected.64 Figure 7 shows the expected physical units rented in periods 1 and 2, as well as
expected revenue65 of the average movie in the sample if it were released under each regime on
a given date. Through most of 2010, a windowing strategy yielded higher estimated revenue,
but starting right around January 1, 2011, the early release strategy began to dominate.
I include log(DVD price) in column (3), which by itself has a positive and significant
effect on all dependent variables. This is the classic case of price endogeneity - better movies
have higher prices. As soon as I control for movie characteristics in columns (4) and (5),
the coefficients on log(DVD price) look much more reasonable, and the signs flip where they
should. Finally, in specifications (6) and (7), I control for seasonality and competition.
The two main findings of this paper are that the street date strategy yields higher revenue
from physical rentals, and that the window does not affect any revenue sources other than
physical rentals. The most detailed specification of the model explains 87% of the variation
in period 1 physical rentals, 86% of the variation in DVD sales, 79% of the variation in Bluray sales, and 58% of the variation in VOD rentals, and the results are robust to alternative
specifications.
As shown in Figure 7, the difference between physical rental revenue for the average
movie is about $5 million for movies released in the third quarter of 2011 or later. An even
split of revenues between rental companies and studios implies that windowing decreases
studio profitability by $2.5 million per movie. Upstream prices that rental companies pay to
studios are not generally disclosed, but numerous industry reports state that studios charge
significantly higher prices to Netflix and Redbox for early releases. This implies that the
loss per movie is more than $2.5 million. Using industry averages, I perform a back of the
64
Netflix and Redbox has increasing market shares over the sample period, so movies released later should
have a more pronounced Window effect.
65
I estimate revenue by assuming a price of $5 in period 1 for windowed rentals and a price of $2 for all
other rentals.
40
envelope calculation to estimate that the average movie released on June 30, 2011 or later
would do $3.33 million better under an early release regime than a windowing regime.66 For
the six major studios that each release approximately 30 movies a year, this translates into
a lost annual opportunity of $100 million from using a windowing strategy.
7
Conclusions
The main finding of this paper is that when movies are delayed on Netflix and Redbox
for a 28 day window, DVD sales, Blu-ray sales, and VOD rentals are unaffected. Instead of
substituting to higher-priced options, consumers either 1. time-shift their consumption and
wait until Week Five or later for a Netflix or Redbox rental or 2. drop out of the market
for that movie entirely. This substitution pattern implies that consumers exhibit strong
loyalty to method of consumption and that there is a positive but relatively small disutility
of waiting an additional 28 days.
The windowing strategy yields lower short run profits by $3.33 million per movie, but
at the same time it greatly reduces the quality of titles on Netflix and Redbox. The logical
conclusion is that Time Warner (2011 rev: $29 billion), Fox-News Corp. (2011 rev: $33 billion), NBC-Universal-Comcast (2011 rev: $56 billion) are willing to forego short term profits
of about $100 million to protect their downstream exhibition channels, such as Comcast,
HBO, and Hulu.67 This strategy may be especially important in light of the strong loyalty
to method of consumption. Disney (2011 rev: $41 billion), Sony (2011 rev: $87 billion),
Paramount-Viacom (2011 rev: $15 billion), and the minor distributors have less ownership
of downstream exhibition, so they use street date releases to make higher short run profits.
The exogenous variation of Netflix and Redbox availability provided a natural experiment
66
The average movie would rent 5 million units and make $10 million in revenue in an early release regime,
2.5 million units and $5 million in revenue in a windowing regime. Assuming the average movie rents
15 times (Source: http://en.wikipedia.org/wiki/Redbox), rental companies would purchase 333,333 copies
from the distributors for early releases and 166,667 copies for windowed releases. Assuming an early release
price of $15 and a windowed price of $10, studio profits are $5 million from an early release strategy and
$1.67 million from a window strategy for the average movie.
67
See Figure 3
41
framework for this study, so the reduced form model provided very meaningful results. The
structural model could prove useful in a welfare calculation or merger analysis in a future
work, but it does not add much to the analysis in this paper. As shown in Table 11, estimates
of σ are close to 1, but they are not precise. This is likely due to the very small amount of
substitution across methods of consumption and the high variation in unobserved consumer
characteristics. The structural model does seem to capture other variables well, such as λB ,
λR , and how movie characteristics factor into utility.
To fully understand substitution patterns between renting and buying in the movie industry, one would need a panel dataset on household movie consumption. A comprehensive
panel would include household level Netflix subscriptions, cable and satellite subscriptions,
premium cable subscriptions, premium cable use, VOD rentals, DVD and Blu-ray purchases,
Redbox, DVD-by-mail, and brick and mortar rentals. Future work with that dataset would
include modeling the household choice to subscribe to Comcast, Netflix, or HBO, where the
model treats these services as different bundles of content.
42
References
Steven T. Berry. Estimating discrete choice models of product differentiation. The RAND
Journal of Economics, 25(2):242–262, 1994.
Paul Bond. What hollywood execs privately say about netflix. Hollywood Reporter, 2011.
N. Scott Cardell. Variance components structures for the extreme value and logistic distributions with application to models of heterogeneity. Econometric Theory, 13(2):185–213,
1997.
Lesley Chiou. The timing of movie releases: Evidence from the home video industry. International Journal of Industrial Organization, 26, 2008.
Liran Einav. Seasonality in the u.s. motion picture industry. RAND Journal of Economics,
38:127–145, 2007.
Jay Edward Epstein. The Hollywood Economist. E.J.E. Publications, Ltd. Inc, 2010.
Mickey Ferri. A detailed look inside the illegal movie market. Working Paper, August 2012.
Daniel Frankel. Hbo holding up the movie-download future. The Wrap, 2010.
Erik Gruenwedel. Dvd/vod window nearly shut. Home Media Magazine, 2010.
Erik Gruenwedel. Survey: No risk to collapsed theatrical-to-home entertainment window.
Home Media Magazine, 2012.
Thorsten Hennig-Thurau, Victor Henning, Henrik Sattler, Felix Eggers, and Mark B. Houston. The last picture show? timing and order of movie distribution channels. Journal of
Marketing, 71:63–83, 2007.
David M. Lang, David M. Switzer, and Brandon J. Swartz. Dvd sales and the r-rating
puzzle. Journal of Cultural Economics, 35:267–286, 2011.
Donald R. Lehmann and Charles B. Weinberg. Sales through sequential distribution channels: An application to movies and videos. Journal of Marketing, 64:18–33, 2000.
Jackie Y. Luan and K. Sudhir. Optimal inter-release time between sequentially released
products. Working Paper, September 2006.
Sarah Mcbride and Geoffrey Fowler. Studios see big rise in estimates of losses to movie
piracy. The Wall Street Journal, 2006.
Julie Holland Mortimer. Price discrimination, copyright law, and technological innovation:
Evidence from the introduction of dvds. Quarterly Journal of Economics, 122(3):1307–
1350, 2007.
Randy A. Nelson, Clifford E. Reid, and Owen Gilmore. An analysis of the out-of-market
gap for dvds in the u.s. Review of Industrial Organization, 30:303–323, 2007.
43
Anita Rao. Online content pricing: Purchase and rental markets. Job Market Paper, 2011.
Rafael Rob and Joel Waldfogel. Piracy on the silver screen. The Journal of Industrial
Economics, LV(3):379–395, 2007.
Gabriel Snyder. Who’s on top? Variety Magazine, 2005.
Kenneth E. Train. Discrete Choice Methods With Simulation. Cambridge University Press,
second edition edition, 2009.
Chris Tribbey. Redbox still stocks uni discs. Home Media Magazine, 2008.
Chris Tribbey. Analyst: Netflix cant keep profits up in digital world. Home Media Magazine,
2010.
Hal R. Varian. Buying, sharing, and renting information goods. Journal of Industrial Economics, XLVIII:473–488, 2000.
44
A
Appendix A: Tables and Figues
Notes: Overview of all possible ways to consume a movie in approximately decreasing order of average consumer willingness to
pay. Prices are approximations for new releases. Rows in yellow are the options available at the DVD release date, which are
the focus of this paper. Rows in green are digital options. MSO = multi service operator, also referred to as cable and satellite
provider.
45
46
Notes: All box office releases for approximately four years (DVD release dates from February 7, 2006 to January 1, 2010). This
is the cut of the sample when all studios employed an early release strategy. Revenue = US Box Office + DVD Sales + Blu-ray
Sales + Physical Rental for the entire period. Gap = mean(DVD release date - box office release date) in weeks. The bottom
9 rows are the percentage of titles in each MPAA and genre category.
Sources: The Numbers and Home Media Magazine
47
There is a wide range of vertical integration among movie and television suppliers. The three level vertical chain has production
at the top level, followed by distribution, then downstream exhibition. The six major movie studios distribute about 85% of
the revenue in the industry. Studios are arranged roughly in increasing order of downstream movie exhibition ownership.
1 Release date on Redbox and Netflix relative to DVD release date for most of 2010 and 2011. See Tables 2a and 2b for details
2 Sony delayed some movies at Netflix in late 2010
3 Disney switched to a 28 day window starting May 22, 2012
4 Warner Brothers tried to impose a 56 day window effective February 1, 2012. Netflix accepted, but Redbox rejected. Redbox
now uses workaround methods (See section B.2)
5 Before 2005, Viacom and CBS were one company. They are currently separate companies, but Sumner Redstone is the
Executive Chair of both Viacom and CBS Corporation
6 NBCUniversal was formed in 2004 with the merger of GE’s NBC and Vivendi’s Universal. Comcast and GE announced a
buyout agreement for Vivendi in 2009
7 Epix is a joint venture between Viacom, MGM, and Lionsgate
Notes: Overview of all legal submarkets for movie exhibition and the market leaders in each submarket. All submarkets are
moderately to highly concentrated(CR4 range from 46% to 100%.) Sources: See Table Sources at the end of Tables section
48
Notes: Overview of the example in Appendix B, section B.4. I construct four consumers with preferences close to the indifference
points of a general distribution, each of whom deviates on a different margin. I compute their optimal actions in a windowing
regime and an early low-priced regime. If the distribution heavily favors type 1 buyers and renters, a windowing strategy yields
higher profits. If the distribution heavily favors type 2 buyers and renters, the early low-priced strategy yields higher profits.
Notes: This table includes all box office releases with DVD release dates between February 7, 2006 and January 1, 2010. I stop
at this date to show the cut of the sample when all studios employed an early release strategy. Revenue ($Millions) = US Box
Office + DVD Sales + Blu-ray Sales + Physical Rental
Sources: The Numbers and Home Media Magazine
49
Notes: Left side includes raw data from DVD weekly top 30 charts from February 12, 2006 through July 15, 2012. Right side
includes Blu-ray units estimates from my imputation process in Appendix C, section C.1. Raw Blu-ray data comes from top 20
and top 10 charts from January 18, 2009 through July 15, 2012. Units and revenue are the average for all titles that appeared
in the chart during that week in release. Decay is the units sold that week divided by the units sold in the previous week when
both observations are available. SE Decay is the standard error of all decay rates for that week in release.
Sources: Home Media Magazine and The Numbers
Notes: Raw data from Physical Rental weekly top 20 charts from January 1, 2010 through July 22, 2012. Windowed releases
have either 25, 27, or 28 day windows to both Netflix and Redbox. Observations include rank and previous rank for all titles
that appear in top 20. Index for weeks after October 10, 2010 is my estimate, computed in step 6 of section C.2. Decay =
(previous week index / this week index). Entries for rank, index, and decay are means.
Source: Home Media Magazine
50
Notes: Raw data from Video on Demand weekly top 10 and top 5 charts from January 1, 2010 through May 13, 2012. Windowed
releases have either 25, 27, or 28 day windows to both Netflix and Redbox. Sample only includes titles that were released on
VOD the same week they were released on DVD. Observations include rank for all titles that appear in top 10 or top 5 weekly
VOD charts. Index is my estimate, computed in step 2 of appendix C, section C.3. Decay = (previous week index / this week
index). Entries for rank, index, and decay are means.
Source: Home Media Magazine
51
52
53
54
55
Notes: The table includes regression results from the two period nested logit model presented in section 4. γ B and γ R are
coefficients on time trends, which capture the increasing market share of Netflix and Redbox over the sample period. Period
1 is weeks 1 through 4, and period 2 is weeks 5 through 8. *** Significant at the 1% level. ** Significant at the 5% level. *
Significant at the 10% level.
56
Table sources
1. http://en.wikipedia.org/wiki/Netflix (Last accessed September 21, 2012)
2. http://www.engadget.com/2008/10/01/starz-play-content-added-to-netflixs-streaming-library/ (Last accessed September 21,
2012)
3. http://www.homemediamagazine.com/netflix/%E2%80%98lying%E2%80%99-and-%E2%80%98whiteout%E2%80%99-firstdelayed-warner-titles-netflix-18131 (Last accessed November 1, 2012)
4. http://articles.latimes.com/2010/apr/10/business/la-fi-ct-netflix10-2010apr10 (Last accessed November 1, 2012)
5. http://www.homemediamagazine.com/netflix/analyst-netflix-cant-keep-profits-digital-world-20352 (Last accessed April 25,
2012)
6. http://allthingsd.com/20101006/netflix-adds-another-studio-sony-agrees-to-window/ (Last accessed November 1, 2012)
7. http://gigaom.com/video/netflix-splits-dvd-streaming-plans/ (Last accessed May 10, 2012)
8. http://latimesblogs.latimes.com/entertainmentnewsbuzz/2012/01/warner-bros-netflix-deal-includes-delay-in-queues.html (Last
accessed May 18, 2012)
9. http://www.homemediamagazine.com/disney/disney-implements-28-day-embargo-all-physical-rentals-27470 (Last Accessed
September 21, 2012)
10. http://en.wikipedia.org/wiki/Redbox (Last accessed November 1, 2012)
11. http://www.homemediamagazine.com/redbox/redbox-inks-deals-with-summit-ncircle-17461 (Last accessed November 1,
2012)
12. http://www.californiabusinesslitigation.com/2008/10/david v goliath redbox sues un 1.html (Last accessed November 1,
2012)
13. http://www.homemediamagazine.com/news/redbox-still-stocking-universal-dvds-14029 (Last accessed November 1, 2012)
14. http://www.homemediamagazine.com/news/sony-deal-with-redbox-16441 (Last accessed June 6, 2012)
15. http://www.homemediamagazine.com/kiosk/lionsgate-inks-redbox-deal-16688 (Last accessed July 8, 2012)
16. http://www.homemediamagazine.com/redbox/redbox-files-second-suit-against-a-studio-16705 (Last accessed May 12, 2012)
17. http://www.homemediamagazine.com/warner/update-warner-takes-aim-redbox-netflix-16749 (Last accessed May 12, 2012)
18. http://www.homemediamagazine.com/redbox/paramount-comes-terms-with-redbox-16835 (Last accessed April 18, 2012)
19. http://www.homemediamagazine.com/redbox/paramount-redbox-extend-relationship-17849 (Last accessed November 1,
2012)
20. http://www.homemediamagazine.com/retailers/group-says-wal-mart-target-mandate-new-release-dvd-purchase-limits-18226
(Last accessed November 1, 2012)
21. http://gigaom.com/video/redbox-makes-nice-with-warner-bros-agrees-to-28-day-window/ (Last accessed April 18, 2012)
22. http://gigaom.com/video/redbox-agrees-to-28-day-windows-for-fox-universal-titles/ (Last accessed April 18, 2012)
23. http://www.homemediamagazine.com/kiosk/redbox-paramount-extend-distribution-deal-19712 (Last accessed November
1, 2012)
24. http://news.cnet.com/8301-13506 3-20012252-17.html (Last accessed November 1, 2012)
25. http://www.homemediamagazine.com/redbox/redbox-raises-dvd-rates-120-25491 (Last accessed November 1, 2012)
26.
http://www.homemediamagazine.com/warner/warner-extends-disc-embargo-56-days-26084 (Last accessed November 1,
2012)
27. http://www.homemediamagazine.com/redbox/redbox-acquires-ncr-dvd-kiosk-business-26340 (Last accessed May 9, 2012)
28. http://latimesblogs.latimes.com/entertainmentnewsbuzz/2012/03/universal-keeps-peace-with-redbox-unlike-warner-bros.html
(Last accessed November 1, 2012)
29. http://www.webpronews.com/redbox-extends-sony-agreement-to-keep-same-day-releases-2012-08 (Last accessed November
1, 2012)
30 Share of Households. National Cable & Telecommunications Association
31 Market Share by Revenue. Data taken from company 10-K filings and Adams Media Research
57
32 Market Share by Revenue. Available at https://www.npd.com/wps/portal/npd/us/news/press-releases/pr 120119/ (Last
accessed on October 1, 2012)
33 Available at http://www.homemediamagazine.com/tks-take/walmart-ultraviolet-deal-bold-innovative-way-launch-entertainmentproduct (Last accessed on October 1, 2012)
34 Wikipedia pages for each premium cable channel
35 Available at http://articles.latimes.com/2010/aug/11/business/la-fi-ct-netflixfolo-20100811 (Last accessed October 1, 2012)
36 IHS Screen Digest - Available at http://appleinsider.com/articles/12/06/01/netflix online movie streaming revenue explodes slices apple share i
(Last accessed October 1, 2012)
37 National Association of Theater Owners (NATO) - Available at http://www.natoonline.org/ statisticscircuits.htm (Last
accessed on October 1, 2012)
38 AMC Networks 2012Q3 10-Q report. Dish Network stopped carrying AMC in its basic package on July 1, 2012. AMC went
from 96.9 million subscribers on June 30, 2012 to 84.3 million subscribers on September 30, 2012
39 Available at http://www.latimes.com/entertainment/envelope/cotown/la-et-ct-netflix-disney-ted-sarandos-20121205,0,5013585.story
(Lastaccessed December 11, 2012)
58
Figure 1: Sources of Revenue in the Movie Industry
Notes: Revenue sources not pictured: international box office, cable, premium cable, digital sale, and digital rental. 2012
entries are annual estimates after third quarter. Physical Rentals = Brick & Mortar+Physical Subscription+Kiosk. Blockbuster
is the market leader for Brick & Mortar, Netflix is the market leader for Physical Subscription and Subscription VOD. Redbox
is the market leader for Kiosk. VOD is mostly through cable and satellite MSOs. 2009 entries in bottom right figure are annual
revenue smoothed over all quarters
Sources: The Numbers, Home Media Magazine, Adams Media Reserach, Digital Entertainment Group
59
Figure 2: Netflix and Redbox Growth
Notes: On July 12, 2011, Netflix announced that effective September 1, 2011, they would no longer offer the bundled package
for $10. Each service is individually priced at $8. This was effectively a 60% price increase on their most popular product.
Sources: Netflix and Coinstar 10-Q and 10-K reports
Figure 3: Major Cable and Satellite Providers in the U.S.
Notes: This figure shows the four largest cable and satellite providers in the U.S. Due to economies of scale, this industry is
consolidating. At the same time, many consumers are “chord cutting” - dropping video in favor of internet. Comcast and Time
Warner Cable operate primarily in urban areas, offering three separate services. DirecTV and Dish are available in both urban
and rural areas, and they offer video services through satellite.
Sources: Comcast, Time Warner Cable, DirecTV, and Dish Network 10-Q and 10-K reports
60
Figure 4: N Period Nested Logit Model
Figure 4a: Rent or Buy
Figure 4b: Four Different
Methods of Consumption
Notes: This figure shows two of the model specifications I consider. The difference between the two specifications is the structure
of the unobserved idiosyncratic shocks. In Figure 4a, the consumer receives one rental shock each period, which will factor into
his utility for both a VOD and physical rental. In Figure 4b, the consumer receives a VOD shock, a physical rental shock, and
additional shocks each period. See section 4.2 for details.
61
Figure 5: Mean Units Rented and Sold by Week Since DVD Release
Physical Rentals
DVD Sales
VOD Rentals
Blu-ray Sales
Notes: Observations include all titles from March 28, 2010 through June 30, 2012 that had either a 25, 27, or 28 day windowed
release on both Netflix and Redbox or a street date release at both. Details of imputation are available in Appendix C.
62
Figure 6: Mean Units Rented by Week Since DVD Release
Start Date
End Date
Windowed Releases
Early Releases
Redbox Market Share
Netflix Market Share
Blockbuster Market Share
Period 1
Period 1
Mar 23, 2010
Dec 14, 2010
47
69
19.5%
23%
24%
Period 2
Period 2
Dec 17, 2010
Aug 30, 2011
48
72
23%
37%
17%
Period 3
Sept 6, 2011
Jun 19, 2012
49
90
41%
27%
11%
Period 3
Notes: Mean physical units rented by week since DVD release. Observations include all titles from March 28, 2010 through
June 30, 2012 that had either a 25, 27, or 28 day windowed release on both Netflix and Redbox or a street date release at both
outlets. Details of imputation are available in Appendix C.
Figure 7: Estimated Revenue and Units For Different Release Strategies
Notes: Estimated revenue and units for the average movie in the sample if it were released on a given release date under
a windowing regime and and early release regime. Estimates are based on the constant from regression 1 in table 10a and
coefficients on time, window, and window x time from regression 7. This figure only includes physical rentals. Period 1 is weeks
1 through 4, and period 2 is weeks 5 through 8.
63
Figure 8: Results From Simulations (Figure Notes Forthcoming)
64
65
B
B.1
Appendix B
Premium Cable Networks: HBO, Showtime, Starz, and Epix
There are essentially four premium cable networks that show movies, HBO, Showtime,
Starz, and Epix.68 These channels generally are not included in standard cable and satellite
packages, so consumers have to pay extra monthly fees to access them. Because they are
premium channels, they need to offer premium content. For years, this has meant procuring
the exclusive rights to show theatrical releases during a special premium cable window, which
generally begins 10-12 months after the theatrical release and lasts for about 12 months.69
Each network allies with different movie studios to procure content. HBO has the exclusive
right to films from Warner Brothers, Fox, and Universal during the premium window, Starz
has the exclusive rights to movies from Disney and Sony, and Epix gets the Paramount
movies (See table 4).
The premium cable networks pay millions of dollars for the exclusive right to show movies
during their window, and it is largely this exclusivity that attracts subscribers. Fox, for
example, agreed to an output deal with HBO in 2007 for what was reported to be a sum of
$1 billion covering 10 years (Frankel, 2010). HBOs output agreements with Warner Brothers,
Fox and Universal prohibit download sales while top new releases play on the premium cable
channel - effectively blocking efforts to grow the digital movie marketplace (Frankel, 2010).
Even with a year long exclusive window, more availability to consumers before and after
the window devalue the content during the window. Not only that, alternatives with different
libraries of movies during the window also devalue the content being offered by premium
cable networks. Therefore, Netflix, Redbox, iTunes, and all of the other competitors have
presented a serious threat to the premium cable networks. In 2008, Starz sued Disney,
claiming the studios distribution of the Pirates of the Caribbean and other films on iTunes
violated its exclusive output deal. The resulting out-of-court settlement enabled Disney,
which is working to establish its own digital-locker system known as KeyChest, to keep
distributing its movies on iTunes. Sony, which also has an output deal with Starz, can do
the same. (Frankel, 2010)
According to an HBO executive in January, 2011, “HBO believes in content exclusivity,
especially for high-value content. That’s our rationale for not selling streaming rights to a
competing subscription service.” While HBO licenses shows to such pay-as-you-go streaming services as iTunes and Amazon, it has “no intention of making its content available for
streaming on Netflix.” A Time Warner executive said that if Netflix expects to get a meaningful amount of HBO content, it would have to raise the price of its streaming-only service
from $7.99 a month to $20 before the economics made sense (Bond, 2011).
There are two major differences between HBO and the rest of these networks. One is
that HBO has much more content, and thus much more market power, than any of the other
ones. The other is that HBO is owned by Time Warner, which also owns Warner Brothers.
68
Encore and Starz are under the same ownership, The Movie Channel and Showtime are under the same
ownership
69
In recent years, premium cable networks have started to generate revenue by producing their own content.
For example, HBO has now produced many successful series, including The Sopranos and Sex and the
City, while Showtime is producing successful series like Dexter.
66
When a consumer watches a movie on HBO, all of that revenue is going to Time Warner,
whereas profits for Disney and Sony don’t depend on whether consumers watch their movies
on Netflix or Starz.
Starz currently has the rights to Sony movies through 2016, and is paying (i) a total of
$190 million in four equal annual installments beginning in 2011 for a contract extension
through 2013, and (ii) a total of $120 million in three equal annual installments beginning
in 2015 for the rights through 2016. Starz has the rights to Disney movies through 2012.
B.2
The First Sale Doctrine and Redbox’s Workaround Policy
Under U.S. copyright law, the “First Sale Doctrine” grants copyright holders the exclusive
right to reproduce and sell their copyrighted goods. However, once a copyrighted good has
been sold to a buyer, the copyright holder no longer has jurisdiction over subsequent use of
that product.70 When Fox, Universal, and Warner Brothers refused to sell DVDs to Redbox
in 2008 and 2009, Redbox continued to stock DVDs using “new distribution arrangements,”
which consisted of sending Redbox employees to buy copies at Walmart and other DVD
retailers. For example, Redbox stocked more than 100,000 copies of Universal’s Wanted
within three days of its December 2, 2008 street date and rented the movie more than
230,000 times during the first week (Tribbey, 2008). The first Fox title Redbox stocked
themselves was Ice Age: Dawn of the Dinosaurs, released October 27, 2009. On October
31, the DVD was available at over two thirds of Redbox machines, and on November 14th
it was in 94% of machines. A similar study looked at the Redbox availability of Warner’s
My Sister’s Keeper and Universal’s Bruno, both released November 17, 2009. None of the
Redbox machines had these two titles on street date, but 11.1% had Bruno the day after
street date and 8.1% had My Sister’s Keeper.71
In one instance a Redbox employee was escorted out of a store for trying to buy multiple
copies of a Universal DVD.72 The economics of Walmart’s stance toward Redbox are very
interesting. The driving factors are that many Redbox locations are at the front of Walmart
stores, and that Walmart uses DVD sales as a loss leader, i.e. they price DVDs at or below
cost to attract consumers into the store. Walmart also uses Redbox machines to attract
consumers to the store, so there is a very interesting tradeoff when Redbox buys DVDs from
the back of the store to stock the Redbox machine at the front of the store. If the DVD is
in the Redbox machine, it will be rented 10-15 times by 10-15 different consumers, each of
whom has to come to the store once to rent it and once to return it73 The DVD in the store
can only be sold once to one consumer. That being said, consumers who buy DVDs actually
have to go into the store, and probably spend significantly more on other products during a
Walmart trip.
70
See U.S. Copyright Act of 1976 and the 1984 Supreme Court case, Universal v. Sony
Article available at http://www.homemediamagazine.com/redbox/analyst-redbox-improves-workaroundfor-fox-warner-titles-17601 (Last accessed June 8, 2012)
72
Article available at http://www.homemediamagazine.com/universal/universal-says-it-didn%E2%80%99tdirect-retailers-to-keep-dvds-from-redbox-16895 (Last accessed August 27, 2012)
73
assuming that each kiosk gets approximately the same DVDs in as DVDs out.
71
67
B.3
Premium Video on Demand
For years, movie studios have been trying to implement a premium VOD window, which
would feature very high priced rentals ($30-$60 for a 48 hour rental) between the box office
release date and DVD street date. Thus far, premium VOD has not seen much success, due
in part to a backlash from movie theaters. The target market is housebound dual-income
couples with young children, for whom a trip to the movie theater costs $60-$70 including
the babysitter.
On April 21, 2011, DirecTV became the first and thus far only MSO to offer premium
VOD for a major studio release. The first movie to appear was Sony’s Just Go With It. It
was available for $29.99 for a 48 hour rental 69 days after the theatrical opening and 47 days
before DVD street date. The other movies that DirecTV offered on premium VOD were
Sony’s Battle: Los Angeles and Soul Surfer, Fox’s Cedar Rapids, Diary of a Wimpy Kid:
Rodrick Rules, and Water for Elephants, Warner’s Hall Pass, Red Riding Hood, and Sucker
Punch, and Universal’s The Adjustment Bureau, Paul, and Your Highness.74 The studios
received from $21 to $24 of the $30 rental price.75 However, DirecTV did not effectively
promote premium VOD, viewership was extremely limited, and they have ceased to offer
the service. At the time it was launched, only about 6 million of DirecTV’s 19.2 million
subscribers had the necessary setup to participate.76
Other MSOs have announced plans for premium VOD for major studio releases, but they
have yet to come to fruition. The attempted release of Universal’s Tower Heist sparked a
notable controversy. With the theatrical release date set for November 4th, Universal Studios
(whose majority owner is Comcast) announced in early October that they would release the
movie to Comcast VOD customers in select markets a mere three weeks after the theatrical
release. The movie would be available to approximately 500,000 Comcast customers in
Portland, Oregon and Atlanta, Georgia for a 48 hour rental for a price of $59.99.77 In
response, several movie theater chains announced that they would boycott the movie. Rafe
Cohen, president of Galaxy Theatres, commented: “We just feel it’s a time to draw a line
in the sand. This is virtually a simultaneous release that we don’t think will be helpful
to anyone. We’re standing on principle that it’s best to preserve the theatrical window.”
Universal withdrew the plan, and the movie was not released in a premium window.
While major studios have not been able to effectively use premium VOD, independent
distributors will often release low-budget films to VOD for $8-$10 on or before the theatrical
release date. For example, Amazon started to offer Magnolia’s The Hunter for $9.99 for a
48 hour rental two weeks prior to its April 6, 2012 theatrical release date.78 The price is
significantly lower than what the major studios have in mind, but for movies with low box
office expectations and limited production and advertising budgets, this has proved successful
74
Home Media Magazine, July 2-8, 2012, page 20
Article available at http://latimesblogs.latimes.com/entertainmentnewsbuzz/2011/04/directv-to-launchpremium-video-on-demand-thursday-with-just-go-with-it.html (Last accessed May 9, 2012)
76
Article available at http://latimesblogs.latimes.com/entertainmentnewsbuzz/2011/07/not-much-demandyet-for-premium-video-on-demand.html (Last accessed June 6, 2012)
77
Article available at http://www.movieweb.com/news/tower-heist-vod-plan-sparks-theater-boycott (Last
accessed May 9, 2012
78
Article available at http://www.homemediamagazine.com/vod/news-analysis-premium-vod-doa-or-beingprimed-sequel-27692 (Last accessed September 21, 2012)
75
68
as both a low cost form of delivery and advertising for the box office run.
B.4
Two Period Example
This section is meant to provide a simple example to illustrate some of the major tradeoffs
with which this paper is concerned. See table 8 for a tabular representation of the example.
Consider a movie that is available for sale in two periods for $17, i.e. P1B = P2B = $17.
It is available for rental in period 1 for $5 and period 2 for $2. This is meant to illustrate
the windowing strategy used by Warner Brothers, Fox, and Universal, where the movie is
available at physical rental stores and on VOD for 4 weeks before it is available at Netflix
and Redbox. Assume that the movie has a V0 of 12 and V1 of 5, and also assume that
λB = λR = 3. These numbers are chosen so that the average consumer will be indifferent
between all choices (buy in period 1, rent in period 1, rent in period 2, or not consume.)79
Consider four consumers: buyer 1, who values the first watch more, buyer 2, who values
owning more, renter 1, a high value renter who has a slight preference for seeing the movie
early, and renter 2, a low value renter. They are differentiated by their error terms, as shown
in the table below. 0i =0 for all consumers.
Cares about
B
i1
B
i2
R
i1
R
i2
Windowing
Decision
P1R = 5,
Revenue
R
P2 = 2
C.S.
Street date
Decision
P1R = 2,
Revenue
R
P2 = 2
C.S.
Buyer 1
1st Watch
3
3
1
1
Buy
17
3
Rent at 1
2
4
Buyer 2
Own
4
4
0
0
Buy
17
4
Buy
17
4
Renter 1
Renter 2
High value Low value
0
-2
0
-2
1
-1
0
-1
Rent at 1
Nothing
5
0
1
0
Rent at 1 Rent at 1
2
2
4
2
Note that when the distributor uses the windowing strategy described above, P1R =
$5, P2R = $2, both buyers buy the movie, the high value renter rents in period 1 for $5, and
the low value renter does not rent. An extended version of this table with more details is
available in Appendix A as table 8.
Now consider an alternate pricing strategy, where the movie is available for sale in both
periods for P1B = P2B = $17, and available for rental in both periods for P1R = P2R = $2.
This is meant to illustrate the strategy implemented by Disney, Sony, Paramount, and most
minor distributors, who allowed their movies to be available at Netflix and Redbox on street
date. The decisions of each consumer when this pricing strategy is used are depicted in the
last row of the above table. Note the following:
1. With the lower period one rental price, buyer 1 will substitute from buying to renting,
which is a loss of $15 in revenue. This is one type of cannibalization that distributors
are worried about.
79
Buying in period 2 would only happen with a very high B
2
69
2. Buyer 2’s behavior is unchanged. Note the large the difference between buying and
renting utility for this buyer. We can think of him as a DVD collector.
3. Renter 1’s behavior is unchanged, but the lower rental price in period 1 represents a
loss of $3 in revenue. This is the other type of cannibalization.
4. Renter 2 is now consuming, whereas he was not with the windowing strategy, so this
represents a revenue gain of $2. This is the market expanding effect that makes Redbox
appealing to some distributors.
The main takeaway from this example is that if most people are type 1, the windowing
strategy dominates, but if most people are type 2, the street date strategy dominates. To
complete this example, I will do a short back of the envelope calculation. Assume that
the entire market is comprised of these four types of consumers. Also assume that the
distributor is revenue maximizing.80 Let b1 , b2 , r1 , and r2 denote the respective populations
of each group. The profits from a windowing strategy are π(W ) = 17b1 + 17b2 + 5r1 + 0r2 ,
and the profits from a street date strategy are π(SD) = 2b1 + 17b2 + 2r1 + 2r2 . Thus, a street
date strategy dominates if r2 > 7.5b1 + 1.5r1 .
B.5
Nested Logit Market Derivation
From section 4.2, utility from buying, renting, and nonconsumption are:
UitB =
UitR =
!
B
V0 + V1 − λB t − αp ln pB
ξ
t
B
B
+ ζi + (1 − σ) νit + t , t ∈ {T1B , ..., TNB(B.1)
}
µ
µ
!
R
V1 − λR t − αp ln pR
ξ
t
R
R
+ ζi + (1 − σ) νit + t , t ∈ {T1R , ..., TNR }
(B.2)
µ
µ
Ui0 = ζi0 + (1 − σ) νi0
(B.3)
Without loss of generality, write the probability of consumer i choosing group g, g ∈ {B, R} in
period t as the product of his probability of choosing group g and his conditional probability
of choosing period t, given that he chooses group g:
Probi (gt) = Probi (g) ∗ Probi (t|g)
80
(B.4)
The distributors share revenues with rental and retail companies under various arrangements, the details of
which are mostly unknown to the public. See Appendix C for a discussion. When considering distributor
profits, it is important to remember that movie rentals and purchases can be complementary to other
goods, for example Apple and Walmart’s use of DVD sales as loss leaders and the sale of movie related
merchandise like clothes and toys.
70
Since all consumers come from the same distribution, it is equivalent to write the above
expression in terms of market shares
sgt = sg ∗ st|g
Let δtB =
B
V0 +V1 −λB t−αp ln(pB
t )+ξt
µ
and δtR =
(B.5)
R
V1 −λR t−αp ln(pR
t )+ξt
µ
. For g ∈ {B, R}, t ∈
{T1g , ..., TNg },
st|g = Prob (t|g) =
=
=
st|g = Prob (t|g) =
where Dg =
g
g
g
g
g
g
Prob δt + ζi + (1 − σ) νit =
max g δs + ζi + (1 − σ) νis
s∈{T1g ,...,TN
}
g
g
g
g
Prob δt + (1 − σ) νit =
max g δs + (1 − σ) νis
s∈{T1g ,...,TN
}
g
δt
δsg
g
g
Prob
+ νit =
max g
+ νis
(1 − σ)
s∈{T1g ,...,TN
} (1 − σ)
g δt
exp 1−σ
(B.6)
Dg
g X
δs
exp
,
(B.7)
1−σ
g
g
s∈{T1 ,...,TN }
B
iid type 1 extreme value for all s ∈ {T1g , ..., TNg }. The probability of choosing
because νis
group g ∈ {B, R, 0} is
Dg1−σ
sg = Prob (g) = P
h∈{B,R,0}
(B.8)
Dh1−σ
Plugging in (B.6) and (B.8) to (B.5), we can now write
sgt = P
exp
Dg1−σ
h∈{B,R,0}
Dh1−σ
∗
δtg
1−σ
Dg
=
exp
P
σ
Dg
δtg
1−σ
h∈{B,R,0}
Dh1−σ
(B.9)
There is only one member of the outside group, 0, with δ 0 = 0, so D0 = 1, and
s0 = P
D01−σ
g∈{B,R,0}
Dg1−σ
=P
71
1
g∈{B,R,0}
Dg1−σ
(B.10)
Taking logs and subtracting,
ln (sgt ) − ln (s0 ) =
ln (sgt ) − ln (s0 ) =

δtg
1−σ
δtg
1−σ
− σ ln (Dg ) − ln 

X

Dg1−σ  − 0 + ln 
g∈{B,R,0}

X
Dg1−σ 
g∈{B,R,0}
− σ ln (Dg )
(B.11)
In order to get an expression for ln(Dg ), we can take logs of equation (B.8) and rearrange,
yielding:

ln (sg ) = (1 − σ) ln (Dg ) − ln 

X
Dg1−σ 
(B.12)
g∈{B,R,0}
P
Taking logs of (B.10), we can see that − ln( g∈{B,R,0} Dg1−σ ) is just ln(s0 ), so this means
that:
ln (sg ) = (1 − σ) ln (Dg ) + ln (s0 )
ln (Dg ) =
ln (sg ) − ln (s0 )
1−σ
(B.13)
Plugging back in to (B.11),
ln (sg ) − ln (s0 )
δtg
ln (sgt ) − ln (s0 ) =
−σ
1−σ
1−σ
(1 − σ) (ln (sgt ) − ln (s0 )) = δtg − σ ln (sg ) + σ ln (s0 )
ln (sgt ) − ln (s0 ) = δtg − σ ln (sg ) + σ ln (sgt )
ln (sgt ) − ln (s0 ) = δtg + σ ln st|g
where st|g =
sgt
sg
is the within group share, defined above.
72
(B.14)

Similar documents

×

Report this document