HEDONIC CONSUMER DECISION MAKING AND IMPLICATIONS FOR THE MARKETING
Test (2010)
Available from
Victor Henning's profile on Mendeley.
or
Abstract
This cumulative dissertation investigates aspects of consumer decision making in hedonic contexts and its implications for the marketing of media goods through a series of three empirical studies. All three studies take place within a common theoretical framework of decision making models (shown in Figure 1), applying parts of the framework in novel ways to solve real-world marketing research problems (study 1 and 2), and examining theoretical relationships between variables within of the framework (study 3). One notable way in which the studies differ is their theoretical treatment of the hedonic component of decision making, i.e. the role and conceptualization of emotions.
Available from
Victor Henning's profile on Mendeley.
Page 1
HEDONIC CONSUMER DECISION MAKING AND IMPLICATIONS FOR THE MARKETING
HEDONIC CONSUMER DECISION MAKING AND
IMPLICATIONS FOR THE MARKETING
OF MEDIA GOODS
Dissertation zur Erlangung des Grades
Doctor rerum politicarum (Dr. rer. pol.)
an der Bauhaus-Universität Weimar
Verfasser:
Dipl.-Kfm. Victor Henning
115 Sutherland Avenue, Flat 18
London W9 2QJ
United Kingdom
victor.henning@gmail.com
Betreuer:
Prof. Dr. Thorsten Hennig-Thurau
Lehrstuhl für Marketing und Medien
Westfälische Wilhelms-Universität Münster
IMPLICATIONS FOR THE MARKETING
OF MEDIA GOODS
Dissertation zur Erlangung des Grades
Doctor rerum politicarum (Dr. rer. pol.)
an der Bauhaus-Universität Weimar
Verfasser:
Dipl.-Kfm. Victor Henning
115 Sutherland Avenue, Flat 18
London W9 2QJ
United Kingdom
victor.henning@gmail.com
Betreuer:
Prof. Dr. Thorsten Hennig-Thurau
Lehrstuhl für Marketing und Medien
Westfälische Wilhelms-Universität Münster
Page 2
IContents
I. Preface ....................................................................................................................... 1
I.1 Integrative Framework and Summary of Results ...................................................... 1
I.2 Publication of Studies and Contribution of Co-Authors ............................................ 9
1 The Last Picture Show? Timing and Order of Movie Distribution Channels ................ 11
1.1 Introduction ......................................................................................................... 11
1.2 Sequential Distribution of Motion Pictures: Literature and Conceptual Framework . 14
1.2.1 Overview of Channel Timing and Order Research ......................................... 14
1.2.2 Conceptual Framework for Studio-Revenue Optimization ............................. 15
1.3 A Net Present Value Model of Movie Studio’s Sequential Distribution Revenues ... 19
1.3.1 General Considerations ................................................................................ 19
1.3.2 Formal Model Description ............................................................................ 22
1.3.3 Model Assumptions ..................................................................................... 25
1.4 Research Design ................................................................................................... 26
1.5 Results ................................................................................................................. 32
1.5.1 Estimation and Validation of Conjoint Data .................................................. 32
1.5.2 Sequential Distribution Chain Optimization: A Stepwise Approach ............... 34
1.5.3 Sensitivity Analysis ..................................................................................... 40
1.5.4 Accounting for Heterogeneity: The Impact of Movie Genres ......................... 42
1.6 Discussion and Implications ................................................................................. 44
1.6.1 Implications for Research and the Motion Picture Industry ............................ 44
1.6.2 Limitations, Future Research Opportunities, and Conclusion ......................... 48
2 Consumer File Sharing of Motion Pictures ................................................................ 52
2.1 Introduction ......................................................................................................... 52
2.2 Motion Picture File Sharing Literature .................................................................. 53
I. Preface ....................................................................................................................... 1
I.1 Integrative Framework and Summary of Results ...................................................... 1
I.2 Publication of Studies and Contribution of Co-Authors ............................................ 9
1 The Last Picture Show? Timing and Order of Movie Distribution Channels ................ 11
1.1 Introduction ......................................................................................................... 11
1.2 Sequential Distribution of Motion Pictures: Literature and Conceptual Framework . 14
1.2.1 Overview of Channel Timing and Order Research ......................................... 14
1.2.2 Conceptual Framework for Studio-Revenue Optimization ............................. 15
1.3 A Net Present Value Model of Movie Studio’s Sequential Distribution Revenues ... 19
1.3.1 General Considerations ................................................................................ 19
1.3.2 Formal Model Description ............................................................................ 22
1.3.3 Model Assumptions ..................................................................................... 25
1.4 Research Design ................................................................................................... 26
1.5 Results ................................................................................................................. 32
1.5.1 Estimation and Validation of Conjoint Data .................................................. 32
1.5.2 Sequential Distribution Chain Optimization: A Stepwise Approach ............... 34
1.5.3 Sensitivity Analysis ..................................................................................... 40
1.5.4 Accounting for Heterogeneity: The Impact of Movie Genres ......................... 42
1.6 Discussion and Implications ................................................................................. 44
1.6.1 Implications for Research and the Motion Picture Industry ............................ 44
1.6.2 Limitations, Future Research Opportunities, and Conclusion ......................... 48
2 Consumer File Sharing of Motion Pictures ................................................................ 52
2.1 Introduction ......................................................................................................... 52
2.2 Motion Picture File Sharing Literature .................................................................. 53
Page 3
II
2.2.1 File Sharing Consequences ........................................................................... 53
2.2.2 File Sharing Determinants: The Rochelandet–Le Guel Model ........................ 56
2.3 Consequences and Determinants of Motion Picture File Sharing ............................ 57
2.3.1 Motion Picture File Sharing as the Focal Construct ....................................... 57
2.3.2 The Effects of Motion Picture File Sharing on Commercial Channels ............ 57
2.3.3 Determinants of Motion Picture File Sharing ................................................ 59
2.4 Testing File Sharing Consequences ....................................................................... 65
2.4.1 Data Collection and Sample ......................................................................... 65
2.4.2 Measures of File Sharing and Commercial Consumption ............................... 68
2.4.3 Descriptive File Sharing Statistics ................................................................ 69
2.4.4 Method ........................................................................................................ 70
2.4.5 Theater-Related Results ............................................................................... 71
2.4.6 DVD-Related Results ................................................................................... 76
2.5 Testing File Sharing Determinants ........................................................................ 79
2.5.1 Data, Method, and Measures ........................................................................ 79
2.5.2 Results ........................................................................................................ 83
2.6 Discussion, Implications, and Limitations ............................................................. 86
3 Augmenting the Expectancy-Value Model with a Dimensional Model of Emotion: Does
It Matter if the Product Is Hedonic or Utilitarian? ...................................................... 92
3.1 Introduction ......................................................................................................... 92
3.2 The Link between the Expectancy-Value Model and Emotion in Extant Research ... 94
3.2.1 The Influence of the Expectancy-Value Model .............................................. 94
3.2.2 EVM and Measures of Emotion .................................................................... 95
3.2.3 The Role of Emotions for Attitude and Behavior ........................................... 96
3.3 Augmenting the Expectancy-Value Model: Hypothesis Development ..................... 99
3.4 Empirical Test of the Augmented EVM Model .................................................... 102
2.2.1 File Sharing Consequences ........................................................................... 53
2.2.2 File Sharing Determinants: The Rochelandet–Le Guel Model ........................ 56
2.3 Consequences and Determinants of Motion Picture File Sharing ............................ 57
2.3.1 Motion Picture File Sharing as the Focal Construct ....................................... 57
2.3.2 The Effects of Motion Picture File Sharing on Commercial Channels ............ 57
2.3.3 Determinants of Motion Picture File Sharing ................................................ 59
2.4 Testing File Sharing Consequences ....................................................................... 65
2.4.1 Data Collection and Sample ......................................................................... 65
2.4.2 Measures of File Sharing and Commercial Consumption ............................... 68
2.4.3 Descriptive File Sharing Statistics ................................................................ 69
2.4.4 Method ........................................................................................................ 70
2.4.5 Theater-Related Results ............................................................................... 71
2.4.6 DVD-Related Results ................................................................................... 76
2.5 Testing File Sharing Determinants ........................................................................ 79
2.5.1 Data, Method, and Measures ........................................................................ 79
2.5.2 Results ........................................................................................................ 83
2.6 Discussion, Implications, and Limitations ............................................................. 86
3 Augmenting the Expectancy-Value Model with a Dimensional Model of Emotion: Does
It Matter if the Product Is Hedonic or Utilitarian? ...................................................... 92
3.1 Introduction ......................................................................................................... 92
3.2 The Link between the Expectancy-Value Model and Emotion in Extant Research ... 94
3.2.1 The Influence of the Expectancy-Value Model .............................................. 94
3.2.2 EVM and Measures of Emotion .................................................................... 95
3.2.3 The Role of Emotions for Attitude and Behavior ........................................... 96
3.3 Augmenting the Expectancy-Value Model: Hypothesis Development ..................... 99
3.4 Empirical Test of the Augmented EVM Model .................................................... 102
Page 4
III
3.4.1 Pretest ....................................................................................................... 103
3.4.2 Experimental Procedure ............................................................................. 105
3.4.3 Manipulation Checks and Scale Validation ................................................. 106
3.4.4 Results for Hypothesis 1............................................................................. 109
3.4.5 Results for Hypothesis 2............................................................................. 114
3.5 Discussion.......................................................................................................... 117
3.6 Limitations and Future Research ......................................................................... 120
4 Summary and Implications ...................................................................................... 124
5 References .............................................................................................................. 129
6 Appendix................................................................................................................ 149
3.4.1 Pretest ....................................................................................................... 103
3.4.2 Experimental Procedure ............................................................................. 105
3.4.3 Manipulation Checks and Scale Validation ................................................. 106
3.4.4 Results for Hypothesis 1............................................................................. 109
3.4.5 Results for Hypothesis 2............................................................................. 114
3.5 Discussion.......................................................................................................... 117
3.6 Limitations and Future Research ......................................................................... 120
4 Summary and Implications ...................................................................................... 124
5 References .............................................................................................................. 129
6 Appendix................................................................................................................ 149
Page 5
1I. Preface
I.1 Integrative Framework and Summary of Results
This cumulative dissertation investigates aspects of consumer decision making in
hedonic contexts and its implications for the marketing of media goods through a series of
three empirical studies. All three studies take place within a common theoretical framework
of decision making models (shown in Figure 1), applying parts of the framework in novel
ways to solve real-world marketing research problems (study 1 and 2), and examining
theoretical relationships between variables within of the framework (study 3). One notable
way in which the studies differ is their theoretical treatment of the hedonic component of
decision making, i.e. the role and conceptualization of emotions.
The role of emotions excepted, the framework in Figure 1 largely corresponds to the
information processing view of behavior (Edwards 1954; Newell, Shaw, and Simon 1958;
Howard and Sheth 1969; Bettman 1970), which describes humans as boundedly rational
decision makers who try to maximize their expected utilities which each decision. Perhaps
the most prominent representation of this view is the logical flow model developed by
Howard and Sheth (1969), which in its original form is not a testable mathematical
formulation, but rather an encompassing, global paradigm of buyer behavior that specifies
the relationship between input variables, intervening response variables, and output
variables (Hunt and Pappas 1970). According to this model, consumers recognize their
needs and wants, search for information externally and internally, process the information
under given constraints, and then choose the option which will deliver the highest expected
utility. In the framework in Figure 1, this chain of evaluation, attitude formation, intention,
and behavior relates to the information processing paradigm’s output variables, while
I.1 Integrative Framework and Summary of Results
This cumulative dissertation investigates aspects of consumer decision making in
hedonic contexts and its implications for the marketing of media goods through a series of
three empirical studies. All three studies take place within a common theoretical framework
of decision making models (shown in Figure 1), applying parts of the framework in novel
ways to solve real-world marketing research problems (study 1 and 2), and examining
theoretical relationships between variables within of the framework (study 3). One notable
way in which the studies differ is their theoretical treatment of the hedonic component of
decision making, i.e. the role and conceptualization of emotions.
The role of emotions excepted, the framework in Figure 1 largely corresponds to the
information processing view of behavior (Edwards 1954; Newell, Shaw, and Simon 1958;
Howard and Sheth 1969; Bettman 1970), which describes humans as boundedly rational
decision makers who try to maximize their expected utilities which each decision. Perhaps
the most prominent representation of this view is the logical flow model developed by
Howard and Sheth (1969), which in its original form is not a testable mathematical
formulation, but rather an encompassing, global paradigm of buyer behavior that specifies
the relationship between input variables, intervening response variables, and output
variables (Hunt and Pappas 1970). According to this model, consumers recognize their
needs and wants, search for information externally and internally, process the information
under given constraints, and then choose the option which will deliver the highest expected
utility. In the framework in Figure 1, this chain of evaluation, attitude formation, intention,
and behavior relates to the information processing paradigm’s output variables, while
Page 6
2subjective/social norms and resource constraints are among the paradigm’s (more
peripherally treated) “exogenous variables”.
As the label “information processing view” suggests, the consumer’s processing of
available information in order to make a rational choice lies at the heart of this paradigm.
The foundation for this was laid by Edwards (1954) in his influential article The Theory of
Decision Making. Based on economic theories of rationality and utility, he introduced the
so-called “expectancy-value models” to the psychological literature. In his Subjective
Expected Utility model, the likelihood of an event’s occurrence when an action is taken is
the subjective probability SP of an outcome, and the desirability of this outcome is its
subjective utility U. The product of subjective probability and desirability equals the
subjective expected utility SEU from the action. The SEU of different alternative behaviors
are compared, and the alternative with the highest SEU is chosen:
(1)
n
i i
i 1
SEU SP U
=
= ·å
In the realm of social psychology, Fishbein (1967) adapted this expectancy-value
model to form the backbone of his theory of reasoned action. In Fishbein’s variant - today
considered “the most widely applied representation of attitude across many disciplines”
(Bagozzi , Gürhan-Canli, and Priester 2002: 7) - beliefs bi about the probability of the
presence of attributes in an object are multiplied with evaluations ei of these attributes. The
product of belief bi and evaluation ei then can be summed over n attributes to determine
global attitude toward the object Aobj. In turn, Aobj determines the intention to act, which
should trigger the corresponding behavior:
(2)
n
Obj i i
i 1
A b e
=
·=å
peripherally treated) “exogenous variables”.
As the label “information processing view” suggests, the consumer’s processing of
available information in order to make a rational choice lies at the heart of this paradigm.
The foundation for this was laid by Edwards (1954) in his influential article The Theory of
Decision Making. Based on economic theories of rationality and utility, he introduced the
so-called “expectancy-value models” to the psychological literature. In his Subjective
Expected Utility model, the likelihood of an event’s occurrence when an action is taken is
the subjective probability SP of an outcome, and the desirability of this outcome is its
subjective utility U. The product of subjective probability and desirability equals the
subjective expected utility SEU from the action. The SEU of different alternative behaviors
are compared, and the alternative with the highest SEU is chosen:
(1)
n
i i
i 1
SEU SP U
=
= ·å
In the realm of social psychology, Fishbein (1967) adapted this expectancy-value
model to form the backbone of his theory of reasoned action. In Fishbein’s variant - today
considered “the most widely applied representation of attitude across many disciplines”
(Bagozzi , Gürhan-Canli, and Priester 2002: 7) - beliefs bi about the probability of the
presence of attributes in an object are multiplied with evaluations ei of these attributes. The
product of belief bi and evaluation ei then can be summed over n attributes to determine
global attitude toward the object Aobj. In turn, Aobj determines the intention to act, which
should trigger the corresponding behavior:
(2)
n
Obj i i
i 1
A b e
=
·=å
Page 7
3Figure 1: Integrative Framework of Decision Making Models
Attribute
Evaluation a, b, c, f /
Symbolic
Evaluation e
Subjective
Expected
Utility a, d /
Attitudes b, c, f
Behavioral
Intention a, b, c, d, e, f
Behavior a, b, c, d, e, f
Emotions e
Resource
Constraintsc, e /
Perceived
Behavioral
Control f
Subjective / Social
Norms c, d, e, f
a) Subjective Expected Utility Model (Edwards 1954)
b) Expectancy-Value Model, Theory of Reasoned Action
(Fishbein 1967; Fishbein and Ajzen 1975)
c) Part of Information Processing View of Consumer Behavior
(Howard and Sheth 1969)
d) Theory of Social Interactions (Becker 1974)
e) Part of Experiential View of Consumer Behavior
(Holbrook and Hirschman 1982)
f) Theory of Planned Behavior (Ajzen 1991)
Context of Study 3
Context of Study 1
Attributes evaluated:
Movie distribution
channel, release timing,
price, bonus material,
language options
Context of Study 2
Costs and utilities measured:
Movie original: Gross utility, price, transaction costs;
Pirated copy: Moral costs, legal costs, technical costs,
transaction utility, mobility utility, storage utility, anti-
industry utility, social utility, collection utility
Attributes evaluated:
DVD: Story, actors, price, genre, cover design, bonus material, director, title; Calculator:
Functions, price, design, brand, quality of the display, ease of use, energy source, overall size
Emotions measured:
Positive: Relaxation, contentedness, calmness, enthusiasm, elation,
excitement; Negative: Boredom, dullness, sluggishness, sadness,
depression, nervousness, anxiety, annoyance, anger
Perceived behavioral control measure:
Technical file sharing knowledge
Attribute
Evaluation a, b, c, f /
Symbolic
Evaluation e
Subjective
Expected
Utility a, d /
Attitudes b, c, f
Behavioral
Intention a, b, c, d, e, f
Behavior a, b, c, d, e, f
Emotions e
Resource
Constraintsc, e /
Perceived
Behavioral
Control f
Subjective / Social
Norms c, d, e, f
a) Subjective Expected Utility Model (Edwards 1954)
b) Expectancy-Value Model, Theory of Reasoned Action
(Fishbein 1967; Fishbein and Ajzen 1975)
c) Part of Information Processing View of Consumer Behavior
(Howard and Sheth 1969)
d) Theory of Social Interactions (Becker 1974)
e) Part of Experiential View of Consumer Behavior
(Holbrook and Hirschman 1982)
f) Theory of Planned Behavior (Ajzen 1991)
Context of Study 3
Context of Study 1
Attributes evaluated:
Movie distribution
channel, release timing,
price, bonus material,
language options
Context of Study 2
Costs and utilities measured:
Movie original: Gross utility, price, transaction costs;
Pirated copy: Moral costs, legal costs, technical costs,
transaction utility, mobility utility, storage utility, anti-
industry utility, social utility, collection utility
Attributes evaluated:
DVD: Story, actors, price, genre, cover design, bonus material, director, title; Calculator:
Functions, price, design, brand, quality of the display, ease of use, energy source, overall size
Emotions measured:
Positive: Relaxation, contentedness, calmness, enthusiasm, elation,
excitement; Negative: Boredom, dullness, sluggishness, sadness,
depression, nervousness, anxiety, annoyance, anger
Perceived behavioral control measure:
Technical file sharing knowledge
Page 8
4The methodological background of study 1 in this dissertation, titled The Last
Picture Show? Timing and Order of Movie Distribution Channels, can be traced to these
attribute- and utility-based views of decision making. The study’s research question centers
on determining the optimal timing and order of motion picture releases across sequential
distribution channels in terms of either producer- or industry-revenue maximization. Movies
as well as other media goods are traditionally distributed across distinct sequential channels
(e.g., theaters, home video, video-on-demand). This so-called “release windowing” has
become one of the most contentious issues of debate within the film industry, with
stakeholders fearing cannibalization of their respective distribution channel revenues and
escalating the conflict to open threats and strategic boycotts. The reported study is the first
to simulate the effects of timing, order, and pricing variations on consumer choices - and
hence revenues - across four sequential distribution channels.
To achieve this, we draw on a particular consumer choice modeling technique,
conjoint analysis. Based on prior research in mathematical psychology, conjoint analysis in
marketing was developed in the late 1960s with the idea of estimating utility functions and
component (i.e. attribute) utilities of objects, given a set of rank-order choices (Green and
Rao 1971). In line with this idea, study 1 estimates consumer utility functions and attribute
utilities of movie distribution channels by presenting participants with choices between
consuming movies in different settings, systematically varying the underlying channel
attributes, and capturing their first choice of distribution channel. Based on the estimated
utility functions, we then systematically simulate consumer choices for novel distribution
scenarios and integrate these choices with a behavioral model that takes into account
success-breeds-success effects, repeat purchases, and hedonic saturation. Variables such as
emotions, social norms, or resource constraints are not explicitly considered in this model,
though they may implicitly influence the choice process. As such, out of the three studies
presented in this dissertation, the choice-based conjoint approach in study 1 is
Picture Show? Timing and Order of Movie Distribution Channels, can be traced to these
attribute- and utility-based views of decision making. The study’s research question centers
on determining the optimal timing and order of motion picture releases across sequential
distribution channels in terms of either producer- or industry-revenue maximization. Movies
as well as other media goods are traditionally distributed across distinct sequential channels
(e.g., theaters, home video, video-on-demand). This so-called “release windowing” has
become one of the most contentious issues of debate within the film industry, with
stakeholders fearing cannibalization of their respective distribution channel revenues and
escalating the conflict to open threats and strategic boycotts. The reported study is the first
to simulate the effects of timing, order, and pricing variations on consumer choices - and
hence revenues - across four sequential distribution channels.
To achieve this, we draw on a particular consumer choice modeling technique,
conjoint analysis. Based on prior research in mathematical psychology, conjoint analysis in
marketing was developed in the late 1960s with the idea of estimating utility functions and
component (i.e. attribute) utilities of objects, given a set of rank-order choices (Green and
Rao 1971). In line with this idea, study 1 estimates consumer utility functions and attribute
utilities of movie distribution channels by presenting participants with choices between
consuming movies in different settings, systematically varying the underlying channel
attributes, and capturing their first choice of distribution channel. Based on the estimated
utility functions, we then systematically simulate consumer choices for novel distribution
scenarios and integrate these choices with a behavioral model that takes into account
success-breeds-success effects, repeat purchases, and hedonic saturation. Variables such as
emotions, social norms, or resource constraints are not explicitly considered in this model,
though they may implicitly influence the choice process. As such, out of the three studies
presented in this dissertation, the choice-based conjoint approach in study 1 is
Page 9
5philosophically closest to a “pure” information processing view of the rational decision-
maker exclusively focused on attributes and utility maximization.
In terms of the real-world implications of study 1, the empirical results suggest that
the studios that produce motion pictures can increase their revenues by more than 16 percent
through sequential distribution chain timing and order changes when applying a common
distribution model for all movies in a country, and that revenue-maximizing structures differ
strongly between countries. Under the conditions of the study, we find that a simultaneous
release of movies in theaters and on rental home video generates maximum revenues for
movie studios in the U.S., while having devastating effects on other players such as theater
chains. We discuss different scenarios and their implications for movie studios and other
industry players, and critically reflect on barriers for an implementation of the revenue-
maximizing distribution models.
Study 2 of this dissertation, titled Consumer File Sharing of Motion Pictures,
represents a follow-up to a research question left open by study 1. It examines the economic
impact of illegal peer-to-peer file sharing on movie consumption choices, and thus
distribution channel revenues. Similarly, the theoretical model we choose to study this
phenomenon represents a follow-up development in economic and psychological decision
making research. By 1975, Fishbein and Ajzen had extended Fishbein’s (1967) earlier
expectancy-value model into the Theory of Reasoned Action, which now accounted for
subjective norms, i.e. a person’s perception of how others want him/her to behave, and
his/her motivation to comply. In the realm of economics, Becker (1974; 1992) had turned
his attention to research areas that had traditionally been the domain of sociology, studying
crime, drug addiction, discrimination, or family relationships from an individual-utility
maximization perspective. In his framework, social feelings of guilt, obligation, duty,
altruism, or love have positive or negative utility, and can therefore be subjected to an
economic cost-benefit analysis of “social income”. To assess the behavioral drivers of
maker exclusively focused on attributes and utility maximization.
In terms of the real-world implications of study 1, the empirical results suggest that
the studios that produce motion pictures can increase their revenues by more than 16 percent
through sequential distribution chain timing and order changes when applying a common
distribution model for all movies in a country, and that revenue-maximizing structures differ
strongly between countries. Under the conditions of the study, we find that a simultaneous
release of movies in theaters and on rental home video generates maximum revenues for
movie studios in the U.S., while having devastating effects on other players such as theater
chains. We discuss different scenarios and their implications for movie studios and other
industry players, and critically reflect on barriers for an implementation of the revenue-
maximizing distribution models.
Study 2 of this dissertation, titled Consumer File Sharing of Motion Pictures,
represents a follow-up to a research question left open by study 1. It examines the economic
impact of illegal peer-to-peer file sharing on movie consumption choices, and thus
distribution channel revenues. Similarly, the theoretical model we choose to study this
phenomenon represents a follow-up development in economic and psychological decision
making research. By 1975, Fishbein and Ajzen had extended Fishbein’s (1967) earlier
expectancy-value model into the Theory of Reasoned Action, which now accounted for
subjective norms, i.e. a person’s perception of how others want him/her to behave, and
his/her motivation to comply. In the realm of economics, Becker (1974; 1992) had turned
his attention to research areas that had traditionally been the domain of sociology, studying
crime, drug addiction, discrimination, or family relationships from an individual-utility
maximization perspective. In his framework, social feelings of guilt, obligation, duty,
altruism, or love have positive or negative utility, and can therefore be subjected to an
economic cost-benefit analysis of “social income”. To assess the behavioral drivers of
Page 10
6movie file sharing, we turn to a Beckerian utility analysis that, among other factors,
includes the moral costs of unethical behavior, the emotional cost of fearing legal action, the
“Schadenfreude” utility of harming movie studios perceived as greedy, and the social utility
of impressing one’s friends with freely acquired movies. The perceived behavioral control
variable in Figure 1 is captured through measuring the consumer’s technical knowledge,
which is a prerequisite for engaging in peer-to-peer filesharing. Our results suggest that
“Schadenfreude” utility and technical knowledge (which results in lowered search costs) are
indeed among the significant drivers of file sharing behavior, while moral costs are among
the significant deterrents.
In relation to the framework in Figure 1 it should be noted that while our analysis
takes a selection of emotions into account, it follows the traditional Beckerian approach,
subsuming feelings of fear, guilt, gloating and pride under the same utility-maximizing cost-
benefit analysis that also includes economic transaction, search, substitution, and purchasing
cost. As such, unlike study 3, it is not grounded systematically in either appraisal theories or
dimensional theories of emotion.
As for the economic effects of file sharing behavior, these had been hotly contested
prior to our study. Whereas industry advocates and some scholars postulated a cannibalistic
effect on commercial forms of movie consumption, other researchers denied this effect,
though evidence was lacking on both sides. Our study estimates the economic effects based
on data from a controlled longitudinal panel study of 1,075 German consumers. The data
contains information on the consumers’ behavioral intentions and actual behaviors toward
consuming 25 new motion pictures, allowing us to study more than 10,000 individual file
sharing opportunities. Using a series of ReLogit regression analyses and applying partial
least squares structural equation modeling, we find evidence of substantial cannibalization
of theater visits, DVD rentals, and DVD purchases, responsible for annual revenue losses of
$300 million in Germany.
includes the moral costs of unethical behavior, the emotional cost of fearing legal action, the
“Schadenfreude” utility of harming movie studios perceived as greedy, and the social utility
of impressing one’s friends with freely acquired movies. The perceived behavioral control
variable in Figure 1 is captured through measuring the consumer’s technical knowledge,
which is a prerequisite for engaging in peer-to-peer filesharing. Our results suggest that
“Schadenfreude” utility and technical knowledge (which results in lowered search costs) are
indeed among the significant drivers of file sharing behavior, while moral costs are among
the significant deterrents.
In relation to the framework in Figure 1 it should be noted that while our analysis
takes a selection of emotions into account, it follows the traditional Beckerian approach,
subsuming feelings of fear, guilt, gloating and pride under the same utility-maximizing cost-
benefit analysis that also includes economic transaction, search, substitution, and purchasing
cost. As such, unlike study 3, it is not grounded systematically in either appraisal theories or
dimensional theories of emotion.
As for the economic effects of file sharing behavior, these had been hotly contested
prior to our study. Whereas industry advocates and some scholars postulated a cannibalistic
effect on commercial forms of movie consumption, other researchers denied this effect,
though evidence was lacking on both sides. Our study estimates the economic effects based
on data from a controlled longitudinal panel study of 1,075 German consumers. The data
contains information on the consumers’ behavioral intentions and actual behaviors toward
consuming 25 new motion pictures, allowing us to study more than 10,000 individual file
sharing opportunities. Using a series of ReLogit regression analyses and applying partial
least squares structural equation modeling, we find evidence of substantial cannibalization
of theater visits, DVD rentals, and DVD purchases, responsible for annual revenue losses of
$300 million in Germany.
Page 16
12
channels. For example, Warner Bros. Entertainment chairman Barry Meyer publicly
envisions major movies debuting “on DVD simultaneously with their theatrical release,”
proposing that future premieres “will be in Wal-Mart” (Bond 2005) and that theater
revenues will be mere “added value.” As a result, the window between the theatrical and
home video release of a motion picture is shrinking (Saccone 2005), with consumers
being able to (pre)order the DVD of a movie even before it has opened theatrically in
some major export markets. Such fundamental shifts in sequencing strategies would
almost certainly affect players such as theater owners (Eliashberg, Elberse, and Leenders
2006; Vogel 2004). John Fithian, president of the National Association of Theater
Owners, considers timing and order changes as “the biggest threat to the viability of the
cinema industry today” (in CBC 2006, p. 1). So, with the growth of alternative ways to
watch films, will movie theaters soon see their “last picture show?”
The impact that timing and order changes would have on movie studio revenues
and profits is unclear. The current industry discussion is clearly dominated by
speculation based on proprietary consultancy reports for which the underlying data,
assumptions, and analyses are not open for verification. For example, a J.P. Morgan
report suggests that a simultaneous release of a film in theaters and on DVD would lead
to an overall 36% increase in studio revenues (Snyder 2005a). In terms of scholarly
research, a limited number of researchers have studied the effect that changes in
sequential distribution timing could have for studios (e.g., Lehmann and Weinberg 2000),
but extant studies present either theoretical models of specific aspects of the sequential
distribution process (Prasad, Bronnenberg, and Mahajan 2004) or empirical models that
are based on aggregated past market data (Frank 1994; Lehmann and Weinberg 2000).
No research has yet modeled the multi-stage sequential chains that reflect normal
marketplace conditions, i.e., involving three or more channels and two or more release
windows that have to be optimized simultaneously, and none has modeled the effects that
channels. For example, Warner Bros. Entertainment chairman Barry Meyer publicly
envisions major movies debuting “on DVD simultaneously with their theatrical release,”
proposing that future premieres “will be in Wal-Mart” (Bond 2005) and that theater
revenues will be mere “added value.” As a result, the window between the theatrical and
home video release of a motion picture is shrinking (Saccone 2005), with consumers
being able to (pre)order the DVD of a movie even before it has opened theatrically in
some major export markets. Such fundamental shifts in sequencing strategies would
almost certainly affect players such as theater owners (Eliashberg, Elberse, and Leenders
2006; Vogel 2004). John Fithian, president of the National Association of Theater
Owners, considers timing and order changes as “the biggest threat to the viability of the
cinema industry today” (in CBC 2006, p. 1). So, with the growth of alternative ways to
watch films, will movie theaters soon see their “last picture show?”
The impact that timing and order changes would have on movie studio revenues
and profits is unclear. The current industry discussion is clearly dominated by
speculation based on proprietary consultancy reports for which the underlying data,
assumptions, and analyses are not open for verification. For example, a J.P. Morgan
report suggests that a simultaneous release of a film in theaters and on DVD would lead
to an overall 36% increase in studio revenues (Snyder 2005a). In terms of scholarly
research, a limited number of researchers have studied the effect that changes in
sequential distribution timing could have for studios (e.g., Lehmann and Weinberg 2000),
but extant studies present either theoretical models of specific aspects of the sequential
distribution process (Prasad, Bronnenberg, and Mahajan 2004) or empirical models that
are based on aggregated past market data (Frank 1994; Lehmann and Weinberg 2000).
No research has yet modeled the multi-stage sequential chains that reflect normal
marketplace conditions, i.e., involving three or more channels and two or more release
windows that have to be optimized simultaneously, and none has modeled the effects that
Page 17
13
order changes would have on studio revenues. Also, previous research has not looked at
regional differences, despite the influence that cultural variables can have on the
consumption of entertainment products (Hennig-Thurau, Walsh, and Bode 2004) and the
importance of export markets for U.S. entertainment industries (about half of motion
picture revenues come from non-U.S. markets; OMSYC 2002).
The goal of this paper is to identify sequential distribution configurations that
maximize movie studio revenues. The approach employed here extends the existing
literature in three ways. We (1) consider multiple channels that consumers face in reality,
(2) use individual-level discrete choice consumer data which enables us to model
potential market configurations such as simultaneous releases in theaters and other
channels (e.g., home video) whose economic appeal cannot be assessed by past market
data, and (3) account for country differences. Drawing from the existent literature on
sequential distribution, we develop an integrative framework of sequential distribution’s
impact on studio revenues and use this framework to present a sequential distribution net
present value model. Combining a discrete-choice conjoint design with self-reported
customer data, we apply our model to three leading motion-picture markets (the U.S.,
Japan, and Germany) by drawing on random samples for each of these markets and a
total of 1,770 consumers to allow for market-specific effects. We use the model to
systematically test the effects that changes in the timing and order of the windows of the
sequential distribution chain would have on consumer choices and, subsequently, movie
studio revenues in the different countries. We isolate configurations of the sequential
distribution chain that, under the given assumptions, provide optimal payoffs to the
movie studio and differentiate our findings for different movie genres. We discuss these
results and highlight potential obstacles that studios might face when changing the
existing distribution structure.
order changes would have on studio revenues. Also, previous research has not looked at
regional differences, despite the influence that cultural variables can have on the
consumption of entertainment products (Hennig-Thurau, Walsh, and Bode 2004) and the
importance of export markets for U.S. entertainment industries (about half of motion
picture revenues come from non-U.S. markets; OMSYC 2002).
The goal of this paper is to identify sequential distribution configurations that
maximize movie studio revenues. The approach employed here extends the existing
literature in three ways. We (1) consider multiple channels that consumers face in reality,
(2) use individual-level discrete choice consumer data which enables us to model
potential market configurations such as simultaneous releases in theaters and other
channels (e.g., home video) whose economic appeal cannot be assessed by past market
data, and (3) account for country differences. Drawing from the existent literature on
sequential distribution, we develop an integrative framework of sequential distribution’s
impact on studio revenues and use this framework to present a sequential distribution net
present value model. Combining a discrete-choice conjoint design with self-reported
customer data, we apply our model to three leading motion-picture markets (the U.S.,
Japan, and Germany) by drawing on random samples for each of these markets and a
total of 1,770 consumers to allow for market-specific effects. We use the model to
systematically test the effects that changes in the timing and order of the windows of the
sequential distribution chain would have on consumer choices and, subsequently, movie
studio revenues in the different countries. We isolate configurations of the sequential
distribution chain that, under the given assumptions, provide optimal payoffs to the
movie studio and differentiate our findings for different movie genres. We discuss these
results and highlight potential obstacles that studios might face when changing the
existing distribution structure.
Page 19
15
media “buzz” for openings in retail stores, “[i]t isn’t that radical a proposition that
movies could follow that same path” (Gentile 2005). Consistent with these arguments,
Eliashberg, Elberse, and Leenders (2006, p. 27) conjecture “that new movies on PPV
[pay-per-view] or VOD prior to the theatrical release could be sold to millions of viewers
[..].” Overall, these contrasting views suggest that an empirical examination of sequential
channel order changes is merited.
1.2.2 Conceptual Framework for Studio-Revenue Optimization
Drawing on extant research on sequential distribution, we now present a
conceptual framework for sequential distribution optimization. As illustrated in Figure 3,
the framework postulates that maximum studio revenues depend upon three optimization
variables: the timing of distribution channels, the order in which these channels open,
and the price for which the product is made available in each channel. Further, it
proposes that these optimization variables are influenced by a number of micro-level and
macro-level factors.
Micro-level factors. We argue that the revenue-maximizing channel configuration
essentially depends upon six micro-level characteristics of sequential distribution chains.
These factors include four that are suggested by the extant literature: inter-channel
cannibalization, perishability, customer expectations, and success-breeds-success effects,
as well as two specific financial factors: the industry-specific discount rate and the
channel-specific revenue allocation.
media “buzz” for openings in retail stores, “[i]t isn’t that radical a proposition that
movies could follow that same path” (Gentile 2005). Consistent with these arguments,
Eliashberg, Elberse, and Leenders (2006, p. 27) conjecture “that new movies on PPV
[pay-per-view] or VOD prior to the theatrical release could be sold to millions of viewers
[..].” Overall, these contrasting views suggest that an empirical examination of sequential
channel order changes is merited.
1.2.2 Conceptual Framework for Studio-Revenue Optimization
Drawing on extant research on sequential distribution, we now present a
conceptual framework for sequential distribution optimization. As illustrated in Figure 3,
the framework postulates that maximum studio revenues depend upon three optimization
variables: the timing of distribution channels, the order in which these channels open,
and the price for which the product is made available in each channel. Further, it
proposes that these optimization variables are influenced by a number of micro-level and
macro-level factors.
Micro-level factors. We argue that the revenue-maximizing channel configuration
essentially depends upon six micro-level characteristics of sequential distribution chains.
These factors include four that are suggested by the extant literature: inter-channel
cannibalization, perishability, customer expectations, and success-breeds-success effects,
as well as two specific financial factors: the industry-specific discount rate and the
channel-specific revenue allocation.
Page 20
16
Figure 3: A Conceptual Framework of Sequential Distribution Revenue
Maximization
With regard to inter-channel cannibalization, we assume that the release of a
movie in a second channel has the potential to cannibalize revenues from an existing
channel due to consumers’ willingness to switch between channels. Inter-channel
cannibalization has been first discussed by Frank (1994) who, modeling the interrelations
between theater visits and home video rental revenues, found that cannibalization takes
place if a film is released on video “too early” (Frank 1994). Lehmann and Weinberg
(2000) also considered channel cannibalization between theater and video releases,
suggesting that the size of each market should determine the delay period. In addition,
cannibalization is reflected by industry thinking that “[a] good movie is a good movie,
regardless of where it's shown” (Bregman, in Arnold 2005). As argued by Prasad,
Channel order
Inter-channel
cannibalization
Channel-specific
revenue allocation
Consumer
expectations
Success breeds
success
Channel timingChannel pricing
Consumer preferences
for distribution channels
Micro-level characteristics
Country
characteristics
Macro-level characteristics
Industry-specific
discount rate
Perishability
Optimization variables
Maximum
producer revenues
Figure 3: A Conceptual Framework of Sequential Distribution Revenue
Maximization
With regard to inter-channel cannibalization, we assume that the release of a
movie in a second channel has the potential to cannibalize revenues from an existing
channel due to consumers’ willingness to switch between channels. Inter-channel
cannibalization has been first discussed by Frank (1994) who, modeling the interrelations
between theater visits and home video rental revenues, found that cannibalization takes
place if a film is released on video “too early” (Frank 1994). Lehmann and Weinberg
(2000) also considered channel cannibalization between theater and video releases,
suggesting that the size of each market should determine the delay period. In addition,
cannibalization is reflected by industry thinking that “[a] good movie is a good movie,
regardless of where it's shown” (Bregman, in Arnold 2005). As argued by Prasad,
Channel order
Inter-channel
cannibalization
Channel-specific
revenue allocation
Consumer
expectations
Success breeds
success
Channel timingChannel pricing
Consumer preferences
for distribution channels
Micro-level characteristics
Country
characteristics
Macro-level characteristics
Industry-specific
discount rate
Perishability
Optimization variables
Maximum
producer revenues
Page 21
17
Bronnenberg, and Mahajan (2004), cannibalization effects can be either complete or
partial, depending on consumers’ perceptions of substitutability between movie channels.
Concerning perishability, we draw on Frank (1994), Lehmann and Weinberg
(2000), and Prasad, Bronnenberg, and Mahajan (2004) who propose a ‘wear out’ effect,
which exists if a film is “too old” when released in secondary channels. Adapting their
argument, we assume that the revenues generated by movies in subsequent channels
should be affected by the time elapsed since the movie was first available, with demand
declining over time. This assumption is shared by industry executives, such as Bob
Chapek, the president of Buena Vista Home Entertainment, who compared a movie “to a
melting ice cube. The longer it sits, the smaller it becomes” (Dutka 2005).
Regarding customer expectations, Prasad, Bronnenberg, and Mahajan (2004) have
argued that, as studios shorten the time between a film’s theatrical run and its rental
availability, consumers will strategically defer their consumption of the movie in the first
channel because they expect the movie to be available soon in another channel that they
prefer for certain reasons (e.g., lower price or multiple viewings). Building on this, we
assume that consumers have expectations regarding the release of a motion picture in
subsequent channels, and that these expectations will influence channel choice, such as
passing up a theater visit in lieu of a later rental or purchase (Prasad, Bronnenberg, and
Mahajan 2004). These expectations can be based on experience, but also on information
from retailers and media (e.g. movie-related websites). For example, STAR WARS:
EPISODE III was the bestselling DVD on Amazon.com in Germany the week before the
movie was released to German theaters, with customers receiving emails from the online
retailer inviting them to preorder the DVD for the new movie.
With regard to success-breeds-success (SBS) effects, Prasad, Bronnenberg, and
Mahajan (2004) demonstrate the existence of complementary effects between channels
by linking the success of the movie in theaters to video revenues. We distinguish between
Bronnenberg, and Mahajan (2004), cannibalization effects can be either complete or
partial, depending on consumers’ perceptions of substitutability between movie channels.
Concerning perishability, we draw on Frank (1994), Lehmann and Weinberg
(2000), and Prasad, Bronnenberg, and Mahajan (2004) who propose a ‘wear out’ effect,
which exists if a film is “too old” when released in secondary channels. Adapting their
argument, we assume that the revenues generated by movies in subsequent channels
should be affected by the time elapsed since the movie was first available, with demand
declining over time. This assumption is shared by industry executives, such as Bob
Chapek, the president of Buena Vista Home Entertainment, who compared a movie “to a
melting ice cube. The longer it sits, the smaller it becomes” (Dutka 2005).
Regarding customer expectations, Prasad, Bronnenberg, and Mahajan (2004) have
argued that, as studios shorten the time between a film’s theatrical run and its rental
availability, consumers will strategically defer their consumption of the movie in the first
channel because they expect the movie to be available soon in another channel that they
prefer for certain reasons (e.g., lower price or multiple viewings). Building on this, we
assume that consumers have expectations regarding the release of a motion picture in
subsequent channels, and that these expectations will influence channel choice, such as
passing up a theater visit in lieu of a later rental or purchase (Prasad, Bronnenberg, and
Mahajan 2004). These expectations can be based on experience, but also on information
from retailers and media (e.g. movie-related websites). For example, STAR WARS:
EPISODE III was the bestselling DVD on Amazon.com in Germany the week before the
movie was released to German theaters, with customers receiving emails from the online
retailer inviting them to preorder the DVD for the new movie.
With regard to success-breeds-success (SBS) effects, Prasad, Bronnenberg, and
Mahajan (2004) demonstrate the existence of complementary effects between channels
by linking the success of the movie in theaters to video revenues. We distinguish between
Page 23
19
Macro-level factors. The micro-level characteristics and the revenue-maximizing
channel structure that can be derived from them are influenced by the macro-level factors
of channel preference and country. Specifically, micro-level characteristics are
influenced by the consumers’ preferences towards distribution channels such as movie
theaters, DVD purchases, DVD rentals, and online downloading (Vogel 2004), all of
which must be considered simultaneously. Channel preferences clearly differ; while some
consumers prefer going to the movies (“I love the mythos of the darkened theater”,
customer statement in Puig 2005), others argue that “There’s no place like home” (Clark
2005). This channel preference determines, among others, the extent of inter-channel
cannibalization and perishability, because strong preferences for a certain channel limit
the degree of cannibalization between channels and reduce the impact that perishability
might have on channel revenues. The second macro-level factor that we consider is
country characteristics. A wealth of research suggests that consumers across countries
differ in their decision-making processes. In a film context, cultural factors (e.g., Hennig-
Thurau, Walsh, and Bode 2004) and informational factors (e.g., Elberse and Eliashberg
2003) might explain these differences. Such country characteristics affect the
expectations consumers will have towards the opening of secondary channels, as well as
the extent of multiple purchasing and the role of word-of-mouth and charts for movie
consumption. They might also affect the financial parameters of our framework.
1.3 A Net Present Value Model of Movie Studio’s Sequential Distribution
Revenues
1.3.1 General Considerations
Using the sequential distribution framework described above, we now develop a
net present value model of movie studio revenues. In contrast to studies that focus on
overall industry revenues and other shared outcomes (Frank 1994; Luan 2005), we
Macro-level factors. The micro-level characteristics and the revenue-maximizing
channel structure that can be derived from them are influenced by the macro-level factors
of channel preference and country. Specifically, micro-level characteristics are
influenced by the consumers’ preferences towards distribution channels such as movie
theaters, DVD purchases, DVD rentals, and online downloading (Vogel 2004), all of
which must be considered simultaneously. Channel preferences clearly differ; while some
consumers prefer going to the movies (“I love the mythos of the darkened theater”,
customer statement in Puig 2005), others argue that “There’s no place like home” (Clark
2005). This channel preference determines, among others, the extent of inter-channel
cannibalization and perishability, because strong preferences for a certain channel limit
the degree of cannibalization between channels and reduce the impact that perishability
might have on channel revenues. The second macro-level factor that we consider is
country characteristics. A wealth of research suggests that consumers across countries
differ in their decision-making processes. In a film context, cultural factors (e.g., Hennig-
Thurau, Walsh, and Bode 2004) and informational factors (e.g., Elberse and Eliashberg
2003) might explain these differences. Such country characteristics affect the
expectations consumers will have towards the opening of secondary channels, as well as
the extent of multiple purchasing and the role of word-of-mouth and charts for movie
consumption. They might also affect the financial parameters of our framework.
1.3 A Net Present Value Model of Movie Studio’s Sequential Distribution
Revenues
1.3.1 General Considerations
Using the sequential distribution framework described above, we now develop a
net present value model of movie studio revenues. In contrast to studies that focus on
overall industry revenues and other shared outcomes (Frank 1994; Luan 2005), we
Page 27
23
which equals an annual industry-specific discount rate of r = 10% .3 The monthly
equivalent to r was represented by rM which is used to discount channel revenues to the
opening of the first channel, t is the time difference in months between the opening of the
first channel and the opening of the channel under consideration (i.e., window length),
and b is the percentage of revenues allocated to the studio for each channel.
Revenues are generated through consumers’ choices between different channels,
with choices x being a function of the channel attributes x = f(p,t,m,p), where p is the
price consumers have to pay to see a movie, m is the medium (or channel), and p is a
vector that reflects other factors such as the language in which the movie is shown and
the presence of bonus material. We model a consumer’s individual choice given a set of
channel alternatives via the multinomial logit model:
(2)
å
=
+++
+++=
J
j
jjmjtjp
iimitip
mtp
mtp
Jix
1
)exp(
)exp(
)|(
pqqqq
pqqqq
p
p
with x(i | J) being a consumer’s choice share for channel i in a specific scenario
with J movie consumption alternatives (including an option not to see the movie in one
of the given channel alternatives, i.e., to wait for the movie to be made available on
television for free) and q being a parameter vector that reflects the consumer’s preference
structure for the channel attributes.
Individual-level choice shares are complemented with individual-level SBS
information. It is important to model multiple-purchase SBS and information-cascading
3 While information on suitable discount rates for the valuation of movie studios is scarce,
available sources cite annual discount rates of 9.0% for Sony Pictures (Sony 1997), 9.1% for Disney,
11.0% for MGM, and 11.8% for Pixar (Chalmers 2002). Thus, a discount rate of 10% seemed reasonable.
which equals an annual industry-specific discount rate of r = 10% .3 The monthly
equivalent to r was represented by rM which is used to discount channel revenues to the
opening of the first channel, t is the time difference in months between the opening of the
first channel and the opening of the channel under consideration (i.e., window length),
and b is the percentage of revenues allocated to the studio for each channel.
Revenues are generated through consumers’ choices between different channels,
with choices x being a function of the channel attributes x = f(p,t,m,p), where p is the
price consumers have to pay to see a movie, m is the medium (or channel), and p is a
vector that reflects other factors such as the language in which the movie is shown and
the presence of bonus material. We model a consumer’s individual choice given a set of
channel alternatives via the multinomial logit model:
(2)
å
=
+++
+++=
J
j
jjmjtjp
iimitip
mtp
mtp
Jix
1
)exp(
)exp(
)|(
pqqqq
pqqqq
p
p
with x(i | J) being a consumer’s choice share for channel i in a specific scenario
with J movie consumption alternatives (including an option not to see the movie in one
of the given channel alternatives, i.e., to wait for the movie to be made available on
television for free) and q being a parameter vector that reflects the consumer’s preference
structure for the channel attributes.
Individual-level choice shares are complemented with individual-level SBS
information. It is important to model multiple-purchase SBS and information-cascading
3 While information on suitable discount rates for the valuation of movie studios is scarce,
available sources cite annual discount rates of 9.0% for Sony Pictures (Sony 1997), 9.1% for Disney,
11.0% for MGM, and 11.8% for Pixar (Chalmers 2002). Thus, a discount rate of 10% seemed reasonable.
Page 29
25
movies watched in the channel indicated by the subscript of d that were previously seen
in other channels (here: xFC). dDVD-S is also zero if the movie is first made available
through this specific channel (i.e., DVD sales). However, as the consumer’s desire to own
a movie is formed immediately after viewing it in a different channel and remains
constant thereafter in our model until fulfilled, the multiple-purchase parameter for the
DVD sales channel is time-invariant at dDVD-S = aDVD-S.
The overall channel-specific revenues are calculated by taking the arithmetic
mean of each channel’s complemented choice quantity across all consumers (X') and
multiplying it with the respective channel price. For example, theater revenues can be
calculated by THTHTH XpR ¢×= , with this information enabling us to calculate the weekly
return and the NPV of studio revenues. Appendix A contains an illustrative application of
the model.
1.3.3 Model Assumptions
It is important to note that the model described above is based on a number of
assumptions. In line with our studio perspective, we focus on studio-produced motion
pictures and the conditions under which such movies are distributed. Specifically, we
assume motion pictures to be released widely in theaters (the dominant distribution
model) and do not distinguish between producers and distributors of motion pictures with
regard to revenue maximization, as most movies produced by a major Hollywood studio
are distributed by sister companies over which the studio has complete control (e.g.,
Warner Bros. Pictures, Warner Bros. Pictures Domestic Distribution, and Warner Home
Video are all subsidiaries of Time Warner Inc.). Related, we assume that consumers who
want to see a movie in a channel that is already open are able to do so -- there are no
shortages of screens at the theater or of DVD copies in rental stores and at retailers to
limit consumption, with all movies being available through any channel. This is in line
movies watched in the channel indicated by the subscript of d that were previously seen
in other channels (here: xFC). dDVD-S is also zero if the movie is first made available
through this specific channel (i.e., DVD sales). However, as the consumer’s desire to own
a movie is formed immediately after viewing it in a different channel and remains
constant thereafter in our model until fulfilled, the multiple-purchase parameter for the
DVD sales channel is time-invariant at dDVD-S = aDVD-S.
The overall channel-specific revenues are calculated by taking the arithmetic
mean of each channel’s complemented choice quantity across all consumers (X') and
multiplying it with the respective channel price. For example, theater revenues can be
calculated by THTHTH XpR ¢×= , with this information enabling us to calculate the weekly
return and the NPV of studio revenues. Appendix A contains an illustrative application of
the model.
1.3.3 Model Assumptions
It is important to note that the model described above is based on a number of
assumptions. In line with our studio perspective, we focus on studio-produced motion
pictures and the conditions under which such movies are distributed. Specifically, we
assume motion pictures to be released widely in theaters (the dominant distribution
model) and do not distinguish between producers and distributors of motion pictures with
regard to revenue maximization, as most movies produced by a major Hollywood studio
are distributed by sister companies over which the studio has complete control (e.g.,
Warner Bros. Pictures, Warner Bros. Pictures Domestic Distribution, and Warner Home
Video are all subsidiaries of Time Warner Inc.). Related, we assume that consumers who
want to see a movie in a channel that is already open are able to do so -- there are no
shortages of screens at the theater or of DVD copies in rental stores and at retailers to
limit consumption, with all movies being available through any channel. This is in line
Page 30
26
with the market efficiency hypothesis which matches the reality of movie distribution
quite well for wide studio releases (Hennig-Thurau, Houston, and Walsh 2006).
Moreover, we assume that studio advertising is effective, with consumers being aware of
new studio releases and making their channel choices deliberately, and that its
effectiveness is the same for all channels. Consistent with the early announcement policy
of new movies by studios and retailers, consumers are assumed to have homogeneous
expectations (i.e., knowledge) about the timing of new studio movies’ releases in
different channels, with these expectations matching the actual release dates.
Furthermore, we assume that customers watch a movie in theaters only once, which
correspondents with norms reported in industry information (Hindes 1998). We also
assume that success-breeds-success is not exclusive to theatrical releases, but exists for
any channel in which a new movie is made available for the first time, and assume the
allocation of revenues between studios and other players to be constant over the course of
a movie release (i.e., the studio’s share is identical in week 1 and the weeks that follow).
Moreover, with our focus being on customer preferences, we do not consider potential
market barriers caused by other players such as movie theaters that might hinder studios
to implement certain distribution models (but discuss their impact later in the paper).
Finally, we exclude piracy from our model, as the effect of such illegal consumption
options on traditional distribution channels of motion pictures remains an unanswered
question.
1.4 Research Design
To account for the existence of country factors and because of the enormous
relevance of export markets for U.S. motion pictures (in 2005, cumulative foreign box
office exceeded domestic theatrical revenues by 60%; MPAA 2006b), we applied our
model not only to the U.S. market, but also to Japan and Germany, two film markets that
with the market efficiency hypothesis which matches the reality of movie distribution
quite well for wide studio releases (Hennig-Thurau, Houston, and Walsh 2006).
Moreover, we assume that studio advertising is effective, with consumers being aware of
new studio releases and making their channel choices deliberately, and that its
effectiveness is the same for all channels. Consistent with the early announcement policy
of new movies by studios and retailers, consumers are assumed to have homogeneous
expectations (i.e., knowledge) about the timing of new studio movies’ releases in
different channels, with these expectations matching the actual release dates.
Furthermore, we assume that customers watch a movie in theaters only once, which
correspondents with norms reported in industry information (Hindes 1998). We also
assume that success-breeds-success is not exclusive to theatrical releases, but exists for
any channel in which a new movie is made available for the first time, and assume the
allocation of revenues between studios and other players to be constant over the course of
a movie release (i.e., the studio’s share is identical in week 1 and the weeks that follow).
Moreover, with our focus being on customer preferences, we do not consider potential
market barriers caused by other players such as movie theaters that might hinder studios
to implement certain distribution models (but discuss their impact later in the paper).
Finally, we exclude piracy from our model, as the effect of such illegal consumption
options on traditional distribution channels of motion pictures remains an unanswered
question.
1.4 Research Design
To account for the existence of country factors and because of the enormous
relevance of export markets for U.S. motion pictures (in 2005, cumulative foreign box
office exceeded domestic theatrical revenues by 60%; MPAA 2006b), we applied our
model not only to the U.S. market, but also to Japan and Germany, two film markets that
Page 31
27
are important and culturally diverse. These three countries comprise 56.4 % of the
worldwide theatrical market (MPAA 2003), and Japan and Germany are the world’s
third- and fourth-largest theatrical export markets, respectively. Further, Japan is the
second largest home video market with annual revenues of $5.5 billion; Germany is fifth
with $1.7 billion (IVF 2004).
Stratified random samples of the U.S., Japanese and German population were
drawn in cooperation with a global marketing research company. With age and gender as
interlocked strata, 5,094 consumers (U.S. = 1,701; Japan = 1,802; Germany = 1,591)
were randomly selected from the research company’s database which mirrors each
country’s overall population, and were invited by email to fill out an Internet
questionnaire, and offered $1 for participation. A total of 1,859 consumers responded.
For quality reasons we eliminated respondents who completed the questionnaire in less
than five minutes, leaving a sample of 1,770 (n = 588 in the U.S., a response rate of 34.6
%; n = 593 in Japan, 32.9 %; n = 589 in Germany, 37.0 %). Demographic characteristics
of the subsamples are available upon request.
The questionnaire required respondents to participate in a number of discrete-
choice tasks and to answer rating-scaled questions. To increase the realism of the choice
tasks, respondents were first presented nine upcoming motion pictures and asked to
choose the movie they were most interested to see.4 Short descriptions of the nine
movies’ plots, directors, and stars were provided, as were posters and trailers. An
additional option for respondents was to wait until all nine movies are shown on
4 The nine studio-produced movies, which cover a wide range of genres, were: Harry Potter and
the Goblet of Fire, Jarhead, King Kong, Perfume - The Story of a Murderer, Pink Panther, The Chronicles
of Narnia, The DaVinci Code, Wallace & Gromit: The Curse of the Were-Rabbit, and X-Men 3. None had
been released at the time of the data collection.
are important and culturally diverse. These three countries comprise 56.4 % of the
worldwide theatrical market (MPAA 2003), and Japan and Germany are the world’s
third- and fourth-largest theatrical export markets, respectively. Further, Japan is the
second largest home video market with annual revenues of $5.5 billion; Germany is fifth
with $1.7 billion (IVF 2004).
Stratified random samples of the U.S., Japanese and German population were
drawn in cooperation with a global marketing research company. With age and gender as
interlocked strata, 5,094 consumers (U.S. = 1,701; Japan = 1,802; Germany = 1,591)
were randomly selected from the research company’s database which mirrors each
country’s overall population, and were invited by email to fill out an Internet
questionnaire, and offered $1 for participation. A total of 1,859 consumers responded.
For quality reasons we eliminated respondents who completed the questionnaire in less
than five minutes, leaving a sample of 1,770 (n = 588 in the U.S., a response rate of 34.6
%; n = 593 in Japan, 32.9 %; n = 589 in Germany, 37.0 %). Demographic characteristics
of the subsamples are available upon request.
The questionnaire required respondents to participate in a number of discrete-
choice tasks and to answer rating-scaled questions. To increase the realism of the choice
tasks, respondents were first presented nine upcoming motion pictures and asked to
choose the movie they were most interested to see.4 Short descriptions of the nine
movies’ plots, directors, and stars were provided, as were posters and trailers. An
additional option for respondents was to wait until all nine movies are shown on
4 The nine studio-produced movies, which cover a wide range of genres, were: Harry Potter and
the Goblet of Fire, Jarhead, King Kong, Perfume - The Story of a Murderer, Pink Panther, The Chronicles
of Narnia, The DaVinci Code, Wallace & Gromit: The Curse of the Were-Rabbit, and X-Men 3. None had
been released at the time of the data collection.
Page 32
28
television and can be watched free of charge; consumers who voted for this option were
excluded from the remainder of the questionnaire (Gilbride and Allenby 2004).
For the movie selected, seven choice sets were presented to the respondents
embedded in a choice-based conjoint design (Louviere and Woodworth 1983; for
conjoint work in channels contexts see for example Wuyts et al. 2004). Each choice set
contained four hypothetical channel options for watching the movie (i.e., conjoint
stimuli), as well as a “no-consumption” option (Figure 5). Regarding conjoint attributes,
each conjoint stimuli was described by four (U.S.) or five (Japan and Germany)
attributes, with attribute levels varied systematically (Table 1). Specifically, the attributes
used to generate conjoint stimuli in the U.S. questionnaire were (1) the channel through
which the movie was consumed, (2) the timing of availability, (3) the price a consumer
has to pay to watch the movie, and (4) any additional content (e.g., deleted scenes,
commentaries, etc.) made accessible to the consumer. As a result of pretesting and depth
interviews with industry experts, the latter attribute was included to increase realism. In
Japan and Germany, identical attributes and levels were used (with price levels
transformed into Yen and Euro, respectively). As motion pictures are often presented in
“dubbed” versions in theaters in these countries (i.e., movies are translated into
Japanese/German), language was included in both cases as an additional attribute.
Attribute level combinations which might have resulted in improbable alternatives and
respondent confusion were modeled as prohibited pairs. Stimuli and conjoint choice sets
were created according to a computer-generated randomized design that accounted for
the design principles minimal overlap, level balance and orthogonality (Huber and
Zwerina 1996).
television and can be watched free of charge; consumers who voted for this option were
excluded from the remainder of the questionnaire (Gilbride and Allenby 2004).
For the movie selected, seven choice sets were presented to the respondents
embedded in a choice-based conjoint design (Louviere and Woodworth 1983; for
conjoint work in channels contexts see for example Wuyts et al. 2004). Each choice set
contained four hypothetical channel options for watching the movie (i.e., conjoint
stimuli), as well as a “no-consumption” option (Figure 5). Regarding conjoint attributes,
each conjoint stimuli was described by four (U.S.) or five (Japan and Germany)
attributes, with attribute levels varied systematically (Table 1). Specifically, the attributes
used to generate conjoint stimuli in the U.S. questionnaire were (1) the channel through
which the movie was consumed, (2) the timing of availability, (3) the price a consumer
has to pay to watch the movie, and (4) any additional content (e.g., deleted scenes,
commentaries, etc.) made accessible to the consumer. As a result of pretesting and depth
interviews with industry experts, the latter attribute was included to increase realism. In
Japan and Germany, identical attributes and levels were used (with price levels
transformed into Yen and Euro, respectively). As motion pictures are often presented in
“dubbed” versions in theaters in these countries (i.e., movies are translated into
Japanese/German), language was included in both cases as an additional attribute.
Attribute level combinations which might have resulted in improbable alternatives and
respondent confusion were modeled as prohibited pairs. Stimuli and conjoint choice sets
were created according to a computer-generated randomized design that accounted for
the design principles minimal overlap, level balance and orthogonality (Huber and
Zwerina 1996).
Page 33
29
Figure 5: Example of Conjoint Task
Figure 5: Example of Conjoint Task
Page 34
30
Table 1: Attributes and Levels Included in Conjoint Study
Attribute Description Levels US Levels Japan Levels Germany
Channel The channel (or
medium) through
which the movie
is consumed
Movie theater, DVD
purchase, DVD
rental, legal Internet
download
as in US design as in US design
Timing Time that has
passed since the
movie was first
available for
consumers
through a legal
channel
0 months, 3 months,
6 months, 12 months
as in US design as in US design
Fee The price a
consumer has to
pay to get access
to the movie of
his or her choice
$3, $7.75, $12.50,
$17.25, $22
400 Yen, 1,175 Yen,
1,950 Yen, 2,725
Yen, 3,500 Yen
3 Euro, 7.75 Euro,
12.50 Euro, 17.25
Euro, 22 Euro
Bonus
material
The existence (or
absence) of
background
information
about a motion
picture
Movie only, movie
with a limited
amount of bonus
material (i.e.
making-of
featurette), movie
with extensive bonus
material (i.e., several
making-of
featurettes, deleted
scenes, multiple
audio-commentaries)
as in US design as in US design
Language
options
The language
options the
consumer can
choose between
not included Choice between
Japanese and English
audio track, Japanese
audio track only
Choice between
German and English
audio track, German
audio track only
Finally, respondents were asked to provide movie consumption-related responses
that were used as proxies for the SBS parameters. To calculate the multiple consumption
parameters d , respondents were asked what percentage of movies they had seen in
theaters they had later bought or rented on DVD/home video or downloaded from the
Table 1: Attributes and Levels Included in Conjoint Study
Attribute Description Levels US Levels Japan Levels Germany
Channel The channel (or
medium) through
which the movie
is consumed
Movie theater, DVD
purchase, DVD
rental, legal Internet
download
as in US design as in US design
Timing Time that has
passed since the
movie was first
available for
consumers
through a legal
channel
0 months, 3 months,
6 months, 12 months
as in US design as in US design
Fee The price a
consumer has to
pay to get access
to the movie of
his or her choice
$3, $7.75, $12.50,
$17.25, $22
400 Yen, 1,175 Yen,
1,950 Yen, 2,725
Yen, 3,500 Yen
3 Euro, 7.75 Euro,
12.50 Euro, 17.25
Euro, 22 Euro
Bonus
material
The existence (or
absence) of
background
information
about a motion
picture
Movie only, movie
with a limited
amount of bonus
material (i.e.
making-of
featurette), movie
with extensive bonus
material (i.e., several
making-of
featurettes, deleted
scenes, multiple
audio-commentaries)
as in US design as in US design
Language
options
The language
options the
consumer can
choose between
not included Choice between
Japanese and English
audio track, Japanese
audio track only
Choice between
German and English
audio track, German
audio track only
Finally, respondents were asked to provide movie consumption-related responses
that were used as proxies for the SBS parameters. To calculate the multiple consumption
parameters d , respondents were asked what percentage of movies they had seen in
theaters they had later bought or rented on DVD/home video or downloaded from the
Page 37
33
predicted choice in 66.0% of the U.S. sample cases and in 73.0% and 64.4% of the
Japanese and German cases, respectively.
Table 2: CBC Prediction Accuracy for the Three Samples
Predicted shares
Actual shares
(holdout)
Chance Estimated
shares (Logit)
Movie theater 32.48
32.04
51.61
20.00
20.00
20.00
25.61
29.96
50.75
DVD purchase 16.84
5.73
9.85
20.00
20.00
20.00
17.13
5.91
8.49
DVD
rental
36.05
47.72
22.75
20.00
20.00
20.00
40.70
51.24
25.09
Legal online 4.59
4.89
7.81
20.00
20.00
20.00
3.37
1.58
4.90
None 10.03
9.61
7.98
20.00
20.00
20.00
13.20
11.31
10.76
Chance model Logit model
Average
Attribute
Importance
MAE 11.4140
15.9060
13.7440
3.2399
2.1574
2.0475
Channel 36.96
25.98
30.70
RMSE 2.7572
3.8122
3.7455
0.8972
0.5528
0.4922
Timing 12.96
13.90
16.62
Chi-Square 38.0104
72.6629
70.1427
3.5839
7.5524
2.8917
Fee 42.44
44.36
41.96
Bonus
material
7.65
7.50
8.14
Language
options
n.a.
8.26
2.58
Values in the top row belong to the U.S. sample, in the middle row to the Japanese sample and in the
bottom row to the German sample. MAE = mean absolute error; RMSE = root mean squared error.
Comparing our results with real-world market data enables us to examine the
external validity of our model. We applied our model and U.S. data to a situation that
reflects actual market conditions observed at the time our analysis was conducted (U.S.
predicted choice in 66.0% of the U.S. sample cases and in 73.0% and 64.4% of the
Japanese and German cases, respectively.
Table 2: CBC Prediction Accuracy for the Three Samples
Predicted shares
Actual shares
(holdout)
Chance Estimated
shares (Logit)
Movie theater 32.48
32.04
51.61
20.00
20.00
20.00
25.61
29.96
50.75
DVD purchase 16.84
5.73
9.85
20.00
20.00
20.00
17.13
5.91
8.49
DVD
rental
36.05
47.72
22.75
20.00
20.00
20.00
40.70
51.24
25.09
Legal online 4.59
4.89
7.81
20.00
20.00
20.00
3.37
1.58
4.90
None 10.03
9.61
7.98
20.00
20.00
20.00
13.20
11.31
10.76
Chance model Logit model
Average
Attribute
Importance
MAE 11.4140
15.9060
13.7440
3.2399
2.1574
2.0475
Channel 36.96
25.98
30.70
RMSE 2.7572
3.8122
3.7455
0.8972
0.5528
0.4922
Timing 12.96
13.90
16.62
Chi-Square 38.0104
72.6629
70.1427
3.5839
7.5524
2.8917
Fee 42.44
44.36
41.96
Bonus
material
7.65
7.50
8.14
Language
options
n.a.
8.26
2.58
Values in the top row belong to the U.S. sample, in the middle row to the Japanese sample and in the
bottom row to the German sample. MAE = mean absolute error; RMSE = root mean squared error.
Comparing our results with real-world market data enables us to examine the
external validity of our model. We applied our model and U.S. data to a situation that
reflects actual market conditions observed at the time our analysis was conducted (U.S.
Page 38
34
benchmark model: tTH = 0; tDVD-R = 6; tDVD-S = 6; tVOD = 12; pDVD-S = $17.25; Epstein
2005). We found that the studio revenues in this benchmark model match actual studio
revenues per channel closely. Specifically, 23.7% of studio revenues are generated by
theaters in our simulated benchmark model, while the studio shares of the actual
theatrical revenues accounted for 25.3 % (or $4.5 billion) of the studios’ revenues in the
U.S. in 2005. 19.2 % of our benchmark model studio revenues stem from DVD rentals
mirrored in real-world DVD rental studio revenues of 19.2 % ($3.4 billion), and 57.1% of
the benchmark model studio revenues are generated by DVD sales, while actual DVD
sales revenues constitute 55.5 % ($9.8 billion) of the major studios’ combined theatrical
and home viewing revenues (MPAA 2006b; EMA 2006). This ability to reproduce
current revenue patterns suggests reasonable external validity of the model and the
applied conjoint procedure.
1.5.2 Sequential Distribution Chain Optimization: A Stepwise Approach
This research is the first to consider the timing of sequential distribution systems
as a multiple window problem that requires simultaneous optimization. As several
channel participants are involved, each of whom impose restrictions on the
implementation of distribution chain changes, we decided to use a stepwise approach
when applying our model to the data. Specifically, we test three different groups of
scenarios which differ in terms of restrictedness.
Scenario group I retains the traditional order of movie distribution (i.e., tTH < tDVD-
R and tDVD-S and tVOD,; tDVD-R £ tDVD-S ; tVOD > tDVD-R and tDVD-S), paralleling previous work
on sequential distribution in the film industry (e.g., Lehmann and Weinberg 2000). This
benchmark model: tTH = 0; tDVD-R = 6; tDVD-S = 6; tVOD = 12; pDVD-S = $17.25; Epstein
2005). We found that the studio revenues in this benchmark model match actual studio
revenues per channel closely. Specifically, 23.7% of studio revenues are generated by
theaters in our simulated benchmark model, while the studio shares of the actual
theatrical revenues accounted for 25.3 % (or $4.5 billion) of the studios’ revenues in the
U.S. in 2005. 19.2 % of our benchmark model studio revenues stem from DVD rentals
mirrored in real-world DVD rental studio revenues of 19.2 % ($3.4 billion), and 57.1% of
the benchmark model studio revenues are generated by DVD sales, while actual DVD
sales revenues constitute 55.5 % ($9.8 billion) of the major studios’ combined theatrical
and home viewing revenues (MPAA 2006b; EMA 2006). This ability to reproduce
current revenue patterns suggests reasonable external validity of the model and the
applied conjoint procedure.
1.5.2 Sequential Distribution Chain Optimization: A Stepwise Approach
This research is the first to consider the timing of sequential distribution systems
as a multiple window problem that requires simultaneous optimization. As several
channel participants are involved, each of whom impose restrictions on the
implementation of distribution chain changes, we decided to use a stepwise approach
when applying our model to the data. Specifically, we test three different groups of
scenarios which differ in terms of restrictedness.
Scenario group I retains the traditional order of movie distribution (i.e., tTH < tDVD-
R and tDVD-S and tVOD,; tDVD-R £ tDVD-S ; tVOD > tDVD-R and tDVD-S), paralleling previous work
on sequential distribution in the film industry (e.g., Lehmann and Weinberg 2000). This
Page 42
38
increase by 2.1% compared to the benchmark configuration. Because consumer
expectations now incorporate the higher DVD retail price, choices shares shift slightly
away from retail DVDs towards theaters, rental DVDs, and VOD.
Scenario group II results (U.S.). Removing all order constraints for home
entertainment channels, except for not opening earlier than theaters, we observe major
changes in terms of the channel structure that maximizes studio revenue. Under these
conditions, studio revenues are maximized when movies are released simultaneously in
movie theaters, on rental DVD, and in VOD, with DVDs being released for sale after a
three-month window for a price of $22. In this scenario, studio revenues increase by
16.2% compared to the benchmark constellation. However, these studio revenue gains
impose a heavy cost on movie theaters which lose 40.1% of their revenues due to
cannibalization. Besides movie studios, the beneficiaries of this scenario are DVD
retailers whose revenues increase by 49.6%.
Examining the next-best scenarios under this constraint set, common patterns
exist. The four revenue-maximizing configurations for studios all involve a simultaneous
release in theaters and on rental DVD, with a DVD sales channel window of three
months. Finally, the retail DVD price of $22 is common to the nine best scenarios,
suggesting that DVDs are currently priced too low to maximize studio revenue. This
result is consistent with the notion that “Wal-Mart, Best Buy and other mass marketers
are happily using DVDs and CDs as loss leaders and slashing prices to a level where
even [rental chain] Blockbuster acknowledges it can’t compete” (Amdur 2004).
Scenario group III results (U.S.). Allowing theatrical releases to occur after other
channels have been opened, we find that the most economically attractive scenarios
remain unchanged from scenario group II. Consequently, the results suggest that a
delayed theater release is not optimal for studios, since the loss of shared revenues due to
severe losses by movie theaters are not offset by increases in shared revenues from gains
increase by 2.1% compared to the benchmark configuration. Because consumer
expectations now incorporate the higher DVD retail price, choices shares shift slightly
away from retail DVDs towards theaters, rental DVDs, and VOD.
Scenario group II results (U.S.). Removing all order constraints for home
entertainment channels, except for not opening earlier than theaters, we observe major
changes in terms of the channel structure that maximizes studio revenue. Under these
conditions, studio revenues are maximized when movies are released simultaneously in
movie theaters, on rental DVD, and in VOD, with DVDs being released for sale after a
three-month window for a price of $22. In this scenario, studio revenues increase by
16.2% compared to the benchmark constellation. However, these studio revenue gains
impose a heavy cost on movie theaters which lose 40.1% of their revenues due to
cannibalization. Besides movie studios, the beneficiaries of this scenario are DVD
retailers whose revenues increase by 49.6%.
Examining the next-best scenarios under this constraint set, common patterns
exist. The four revenue-maximizing configurations for studios all involve a simultaneous
release in theaters and on rental DVD, with a DVD sales channel window of three
months. Finally, the retail DVD price of $22 is common to the nine best scenarios,
suggesting that DVDs are currently priced too low to maximize studio revenue. This
result is consistent with the notion that “Wal-Mart, Best Buy and other mass marketers
are happily using DVDs and CDs as loss leaders and slashing prices to a level where
even [rental chain] Blockbuster acknowledges it can’t compete” (Amdur 2004).
Scenario group III results (U.S.). Allowing theatrical releases to occur after other
channels have been opened, we find that the most economically attractive scenarios
remain unchanged from scenario group II. Consequently, the results suggest that a
delayed theater release is not optimal for studios, since the loss of shared revenues due to
severe losses by movie theaters are not offset by increases in shared revenues from gains
Page 44
40
retailer revenues and theater revenues jump up by 28.3% and 14.6%, respectively, while
the rental chains see their earnings plummet by 30.9%. Interestingly, the timing of the
VOD release varies across the different revenue maximizing scenarios, ranging from an
immediate opening to a 12 month delay. Although the VOD channel performs better with
a shorter release window, it does not exert much influence on the studios’ revenues due
to limited cannibalization. As with the U.S. market, lifting the final constraint in scenario
group III does not change the results in Japan and Germany. The best scenarios remain
those found in scenario group II, with the one exception that the new second-best
scenario in Japan suggests an exclusive VOD premiere, followed by a three-month
window for theaters and DVD retail and a 12-month window for DVD rentals.
1.5.3 Sensitivity Analysis
Because some of the information used for model estimation was self-reported by
respondents regarding their behavior under the current channel structure, we conducted a
set of sensitivity analyses to see how robust our results are with regard to these measures.
Specifically, we systematically varied the individual responses for all self-reported
behaviors (multiple consumption SBS, word-of-mouth-based SBS, and charts-based SBS)
for each channel by +/-20%. Table 4 provides the results of these analyses, showing how
variations in the measures affect the respective group-best scenario’s NPV change in
relation to the benchmark scenario. For example, under scenario group II conditions, a
20% increase of the multiple consumption parameter for DVD purchases in Germany
would result in a studio NPV increase of 15.2% compared to the benchmark model
(instead of 14.2% when the multiple consumption parameter for DVD purchases is not
manipulated), while a reduction of the same parameter by 20% would result in an
increase of 13% in studio NPV.
retailer revenues and theater revenues jump up by 28.3% and 14.6%, respectively, while
the rental chains see their earnings plummet by 30.9%. Interestingly, the timing of the
VOD release varies across the different revenue maximizing scenarios, ranging from an
immediate opening to a 12 month delay. Although the VOD channel performs better with
a shorter release window, it does not exert much influence on the studios’ revenues due
to limited cannibalization. As with the U.S. market, lifting the final constraint in scenario
group III does not change the results in Japan and Germany. The best scenarios remain
those found in scenario group II, with the one exception that the new second-best
scenario in Japan suggests an exclusive VOD premiere, followed by a three-month
window for theaters and DVD retail and a 12-month window for DVD rentals.
1.5.3 Sensitivity Analysis
Because some of the information used for model estimation was self-reported by
respondents regarding their behavior under the current channel structure, we conducted a
set of sensitivity analyses to see how robust our results are with regard to these measures.
Specifically, we systematically varied the individual responses for all self-reported
behaviors (multiple consumption SBS, word-of-mouth-based SBS, and charts-based SBS)
for each channel by +/-20%. Table 4 provides the results of these analyses, showing how
variations in the measures affect the respective group-best scenario’s NPV change in
relation to the benchmark scenario. For example, under scenario group II conditions, a
20% increase of the multiple consumption parameter for DVD purchases in Germany
would result in a studio NPV increase of 15.2% compared to the benchmark model
(instead of 14.2% when the multiple consumption parameter for DVD purchases is not
manipulated), while a reduction of the same parameter by 20% would result in an
increase of 13% in studio NPV.
Page 47
43
drawing on genre classifications by IMDBpro. This resulted in five genres (action,
comedy, drama, fantasy, thriller) with two movies in each genre (one movie was assigned
to two genres). Second, we repeated the optimization process used to identify general
revenue-maximizing distribution models for each of the five genres, considering only the
respective subsample (e.g., only respondents who selected fantasy movies).
There appear to be differences in consumer preferences. In the U.S., preferences
toward rental channels are somewhat higher for comedies, while preferences toward
theaters and DVD purchases are higher for action and fantasy movies, which implies
moving forward rental channels for comedies and moving back the DVD rental channel
behind the DVD purchase channel for action and fantasy movies. However, as a whole,
genre effects on NPV outcomes are quite moderate, surpassing the general distribution
model revenues by only .8% (U.S.), 1.6% (Japan), and 2.1% (Germany). Out of (3
countries*4 scenario groups*5 genres =) 60 constellations, we found only one in which a
genre-specific model outperforms the general model by more than five percent (scenario
group 3 in Japan for action movies, outperformance by 5.5%).
Given these relatively small revenues gains and considering that the
implementation of genre-specific distribution models would likely cause consumer
confusion (e.g., when new movies combine elements of two or more genres which have
different distribution patterns -- EVAN ALMIGHTY, the $250 million sequel to BRUCE
ALMIGHTY, is described by its studio as “a spectacle fantasy and also a comedy”, Muñoz
2006), we will focus on the general distribution approach when discussing potential
implications for the movie industry.
drawing on genre classifications by IMDBpro. This resulted in five genres (action,
comedy, drama, fantasy, thriller) with two movies in each genre (one movie was assigned
to two genres). Second, we repeated the optimization process used to identify general
revenue-maximizing distribution models for each of the five genres, considering only the
respective subsample (e.g., only respondents who selected fantasy movies).
There appear to be differences in consumer preferences. In the U.S., preferences
toward rental channels are somewhat higher for comedies, while preferences toward
theaters and DVD purchases are higher for action and fantasy movies, which implies
moving forward rental channels for comedies and moving back the DVD rental channel
behind the DVD purchase channel for action and fantasy movies. However, as a whole,
genre effects on NPV outcomes are quite moderate, surpassing the general distribution
model revenues by only .8% (U.S.), 1.6% (Japan), and 2.1% (Germany). Out of (3
countries*4 scenario groups*5 genres =) 60 constellations, we found only one in which a
genre-specific model outperforms the general model by more than five percent (scenario
group 3 in Japan for action movies, outperformance by 5.5%).
Given these relatively small revenues gains and considering that the
implementation of genre-specific distribution models would likely cause consumer
confusion (e.g., when new movies combine elements of two or more genres which have
different distribution patterns -- EVAN ALMIGHTY, the $250 million sequel to BRUCE
ALMIGHTY, is described by its studio as “a spectacle fantasy and also a comedy”, Muñoz
2006), we will focus on the general distribution approach when discussing potential
implications for the movie industry.
Page 48
44
1.6 Discussion and Implications
This study uses a multi-indicator approach that features Hierarchical Bayes
choice-based conjoint information for the intertemporal prediction of market shares. We
apply an NPV model of movie studio revenues across complex and multi-window
sequential distribution chains and find that by adjusting the configuration of distribution
channels and the price of DVDs, motion picture studios could, all other things being
equal, boost their revenues by 16.2% (or $3.5 billion) in the United States alone.
Moreover, we demonstrate that consumers’ channel preferences and movie-consumption
decisions clearly differ among three major markets (the U.S., Japan, and Germany), thus
offering insights on how studios might fine-tune distribution strategy by country.
1.6.1 Implications for Research and the Motion Picture Industry
Our results suggest that the movie industry’s current distribution model is not
optimal in terms of revenue generation. Our key implication is that studio revenues can
be increased by changing both the timing and order of distribution windows. The channel
configuration that performs best in the U.S. includes making a film simultaneously
available in theaters, DVD rental, and through VOD, followed three months later in the
DVD sale channel at a price of $22. If this configuration was to be used to distribute
motion pictures in the United States, studios would receive only 12.2% of their total
revenues from theaters (versus 25.3% in 2005) and only 14.1% from DVD rentals (versus
19.2% in 2005), while contributions from DVD sales would soar to 73.6% (from 55.5%
in 2004), according to our findings.
Our results suggest that recent industry speculation about simultaneous channel
releases, called a “death threat” by theater owners (Stanley 2005), would indeed be
devastating for movie theaters. However, such a change might be financially attractive to
1.6 Discussion and Implications
This study uses a multi-indicator approach that features Hierarchical Bayes
choice-based conjoint information for the intertemporal prediction of market shares. We
apply an NPV model of movie studio revenues across complex and multi-window
sequential distribution chains and find that by adjusting the configuration of distribution
channels and the price of DVDs, motion picture studios could, all other things being
equal, boost their revenues by 16.2% (or $3.5 billion) in the United States alone.
Moreover, we demonstrate that consumers’ channel preferences and movie-consumption
decisions clearly differ among three major markets (the U.S., Japan, and Germany), thus
offering insights on how studios might fine-tune distribution strategy by country.
1.6.1 Implications for Research and the Motion Picture Industry
Our results suggest that the movie industry’s current distribution model is not
optimal in terms of revenue generation. Our key implication is that studio revenues can
be increased by changing both the timing and order of distribution windows. The channel
configuration that performs best in the U.S. includes making a film simultaneously
available in theaters, DVD rental, and through VOD, followed three months later in the
DVD sale channel at a price of $22. If this configuration was to be used to distribute
motion pictures in the United States, studios would receive only 12.2% of their total
revenues from theaters (versus 25.3% in 2005) and only 14.1% from DVD rentals (versus
19.2% in 2005), while contributions from DVD sales would soar to 73.6% (from 55.5%
in 2004), according to our findings.
Our results suggest that recent industry speculation about simultaneous channel
releases, called a “death threat” by theater owners (Stanley 2005), would indeed be
devastating for movie theaters. However, such a change might be financially attractive to
Page 52
48
months. The outcome here would be a 7.6% revenue increase for studios, revenue growth
of 19.1% for DVD retailers, marginal benefits for rental chains, and no changes for
theaters and VOD. Even though these scenarios promise no negative effects for all parties
involved, implementation would likely be met with resistance because it requires
breaking with the industry tradition of opening the rental channel before (or simultaneous
to) the retail channel. Rental chains would likely resist a change that promises no gains
for them, but moves them further down the distribution chain. However, with DVD
retailers being the co-beneficiaries in every studio revenue-maximizing configuration
identified in our analyses, the studios should have powerful allies in retailing giants such
as Wal-Mart (U.S.) and the Metro Group (Europe).
Altogether, this study integrates the sparse research on inter-channel effects
relevant to the optimization of sequential distribution chains into a coherent model. Our
model builds on characteristics of sequential distribution systems that have been
identified by prior research. Any industry that relies on distribution windowing could
tailor our framework and empirical approach to their context. For example, major record
label SonyBMG recently has started to introduce sequential distribution to the music
industry, a strategy recommended by Booz-Allen & Hamilton consultants (Bhatia, Gay,
and Honey 2001). Other entertainment good producers that already employ windowing,
such as book publishers and computer game developers, may benefit financially from
examining the general characteristics derived in our study to gain insights into how to
refine their distribution models and to increase revenues.
1.6.2 Limitations, Future Research Opportunities, and Conclusion
In addition to our modeling assumptions, this study has some limitations. The
impact of distribution chain changes on piracy is not considered. Next to sequential
months. The outcome here would be a 7.6% revenue increase for studios, revenue growth
of 19.1% for DVD retailers, marginal benefits for rental chains, and no changes for
theaters and VOD. Even though these scenarios promise no negative effects for all parties
involved, implementation would likely be met with resistance because it requires
breaking with the industry tradition of opening the rental channel before (or simultaneous
to) the retail channel. Rental chains would likely resist a change that promises no gains
for them, but moves them further down the distribution chain. However, with DVD
retailers being the co-beneficiaries in every studio revenue-maximizing configuration
identified in our analyses, the studios should have powerful allies in retailing giants such
as Wal-Mart (U.S.) and the Metro Group (Europe).
Altogether, this study integrates the sparse research on inter-channel effects
relevant to the optimization of sequential distribution chains into a coherent model. Our
model builds on characteristics of sequential distribution systems that have been
identified by prior research. Any industry that relies on distribution windowing could
tailor our framework and empirical approach to their context. For example, major record
label SonyBMG recently has started to introduce sequential distribution to the music
industry, a strategy recommended by Booz-Allen & Hamilton consultants (Bhatia, Gay,
and Honey 2001). Other entertainment good producers that already employ windowing,
such as book publishers and computer game developers, may benefit financially from
examining the general characteristics derived in our study to gain insights into how to
refine their distribution models and to increase revenues.
1.6.2 Limitations, Future Research Opportunities, and Conclusion
In addition to our modeling assumptions, this study has some limitations. The
impact of distribution chain changes on piracy is not considered. Next to sequential
Page 57
53
10,000 movie file sharing opportunities. We use this data to investigate whether illegal
movie file sharing influences revenues generated through theatrical visits, DVD rentals,
and DVD purchases and, if so, how strong the effects are. In addition, we present, for the
first time, a comprehensive, theory-based model of the factors that drive consumers’
movie file sharing activity. This model offers the movie industry a more thorough
understanding of why consumers engage in file sharing, suggesting more effective
antipiracy strategies.
The paper is structured as follows. After reviewing the relevant literature, we
derive a set of hypotheses regarding the consequences and determinants of movie file
sharing from extant research and utility theory. We then report our data set and use
ReLogit regression analysis and partial least squares structural equation modeling to test
the hypotheses. We conclude by discussing the results and implications.
2.2 Motion Picture File Sharing Literature
2.2.1 File Sharing Consequences
Industry representatives unanimously argue that illegal motion picture file sharing
has a negative impact on other kinds of movie consumption, and industry-commissioned
studies, such as FFA (2006a) and MPAA (2004c), support their claims. For example, in a
study of movie piracy by the German Federal Film Board (FFA), respondents indicated
how movie downloading or copying movies with a CD/DVD burner had influenced their
consumption of motion pictures through other channels; 42% of the respondents reduced
their number of movie theater visits (though 8% stated they went to the movies more
often), 45% said they rented fewer DVDs, and 44% replied that they bought DVDs less
often (FFA 2006a). Similarly, the findings of an eight-country study commissioned by
the MPAA (2004c) indicate that “about one in four internet users (24%) have
10,000 movie file sharing opportunities. We use this data to investigate whether illegal
movie file sharing influences revenues generated through theatrical visits, DVD rentals,
and DVD purchases and, if so, how strong the effects are. In addition, we present, for the
first time, a comprehensive, theory-based model of the factors that drive consumers’
movie file sharing activity. This model offers the movie industry a more thorough
understanding of why consumers engage in file sharing, suggesting more effective
antipiracy strategies.
The paper is structured as follows. After reviewing the relevant literature, we
derive a set of hypotheses regarding the consequences and determinants of movie file
sharing from extant research and utility theory. We then report our data set and use
ReLogit regression analysis and partial least squares structural equation modeling to test
the hypotheses. We conclude by discussing the results and implications.
2.2 Motion Picture File Sharing Literature
2.2.1 File Sharing Consequences
Industry representatives unanimously argue that illegal motion picture file sharing
has a negative impact on other kinds of movie consumption, and industry-commissioned
studies, such as FFA (2006a) and MPAA (2004c), support their claims. For example, in a
study of movie piracy by the German Federal Film Board (FFA), respondents indicated
how movie downloading or copying movies with a CD/DVD burner had influenced their
consumption of motion pictures through other channels; 42% of the respondents reduced
their number of movie theater visits (though 8% stated they went to the movies more
often), 45% said they rented fewer DVDs, and 44% replied that they bought DVDs less
often (FFA 2006a). Similarly, the findings of an eight-country study commissioned by
the MPAA (2004c) indicate that “about one in four internet users (24%) have
Page 58
54
downloaded a movie” (MPAA 2004c, p. 1) and that, on a global level, 26% of
downloaders purchase movies “much less” or “a little less” often than in the past
(excluding the outlier Korea lowers the unweighted mean from 26% to 14%). The
insights generated by these and other industry studies are limited by their methodological
approaches and lack of transparency. In all cases, the results rely on an ex-post “what-if”
approach that asks consumers who have already seen movies as illegal copies (and
therefore know the cinematic quality) to speculate if they would have paid for the movies
if they had not been available as illegal copies.
To the best of our knowledge, no scholarly research addresses the effects of
sharing illegal movie copies on commercial distribution channels. In the related context
of music file sharing studies, researchers are split into two opposing groups. The first
group reports a negative impact of music file sharing on industry sales (Liebowitz 2006;
Michel 2006; Montero-Pons and Cuadrado-García 2006; Peitz and Waelbroeck 2004;
Zentner 2006), but these studies all rely on aggregate household Internet penetration in a
given city as a proxy for file sharing and do not monitor file sharing on an individual
basis. Obviously, this approach raises serious questions regarding spurious correlations
and paves the way for alternative explanations.
The second group of researchers question these findings and argue that file
sharing has either no or a positive impact on industry revenues. Specifically, Gopal,
Bhattacharjee, and Sanders (2005) propose a model of online music sharing economics
and derive implications for consumer surplus and producer profits. Following the train of
thought that consumer file sharing represents a form of “sampling” for experience goods,
they conclude that file sharing networks lower the total costs of evaluating and acquiring
experience goods, which increases purchases and industry profits. In other words, file
downloaded a movie” (MPAA 2004c, p. 1) and that, on a global level, 26% of
downloaders purchase movies “much less” or “a little less” often than in the past
(excluding the outlier Korea lowers the unweighted mean from 26% to 14%). The
insights generated by these and other industry studies are limited by their methodological
approaches and lack of transparency. In all cases, the results rely on an ex-post “what-if”
approach that asks consumers who have already seen movies as illegal copies (and
therefore know the cinematic quality) to speculate if they would have paid for the movies
if they had not been available as illegal copies.
To the best of our knowledge, no scholarly research addresses the effects of
sharing illegal movie copies on commercial distribution channels. In the related context
of music file sharing studies, researchers are split into two opposing groups. The first
group reports a negative impact of music file sharing on industry sales (Liebowitz 2006;
Michel 2006; Montero-Pons and Cuadrado-García 2006; Peitz and Waelbroeck 2004;
Zentner 2006), but these studies all rely on aggregate household Internet penetration in a
given city as a proxy for file sharing and do not monitor file sharing on an individual
basis. Obviously, this approach raises serious questions regarding spurious correlations
and paves the way for alternative explanations.
The second group of researchers question these findings and argue that file
sharing has either no or a positive impact on industry revenues. Specifically, Gopal,
Bhattacharjee, and Sanders (2005) propose a model of online music sharing economics
and derive implications for consumer surplus and producer profits. Following the train of
thought that consumer file sharing represents a form of “sampling” for experience goods,
they conclude that file sharing networks lower the total costs of evaluating and acquiring
experience goods, which increases purchases and industry profits. In other words, file
Page 59
55
sharing reduces consumers’ risk in evaluating new music (an argument that easily
extends to movies), a major obstacle in consumer decision making.
Using a different argument, Boldrin and Levine (2002) and Grgeta (2004) model
competition with sunk costs and argue that, with certain assumptions, the decreasing
costs of reproduction that result from file sharing make it easier, not harder, for the
producer to recoup his or her investment and that as the rate of reproduction increases,
competitive rents increase. Their conclusion is based on the concept of indirect
appropriability, which assumes an original product attains greater consumer utility when
it can be copied and that this utility increase can be captured by the producer through a
price increase. However, like Gopal, Bhattacharjee, and Sanders (2005), they do not
provide empirical findings to substantiate their conclusions.
Oberholzer-Gee and Strumpf (2005) present empirical results that show no
negative impact of file sharing on traditional music distribution channels. Over the
course of four months, they monitor 1.75 million file downloads on file sharing networks
and then match the downloads to U.S. album sales data. Their empirical analysis shows
that music file sharing has no significant impact on album sales. Again, however, the
generalizability of their findings is somewhat limited as the authors use the “number of
German school kids on vacation” as an instrumental variable for file sharing activity to
bypass endogeneity problems caused by the simultaneity of downloading and purchasing
activity in their aggregate level data.
In summary, movie industry representatives argue that file sharing serves as a
substitute for commercial movie consumption, while no peer-reviewed research has
studied this relationship for movies, and the results from music file sharing research are
inconclusive and limited by methodological constraints. Moreover, no existing study
sharing reduces consumers’ risk in evaluating new music (an argument that easily
extends to movies), a major obstacle in consumer decision making.
Using a different argument, Boldrin and Levine (2002) and Grgeta (2004) model
competition with sunk costs and argue that, with certain assumptions, the decreasing
costs of reproduction that result from file sharing make it easier, not harder, for the
producer to recoup his or her investment and that as the rate of reproduction increases,
competitive rents increase. Their conclusion is based on the concept of indirect
appropriability, which assumes an original product attains greater consumer utility when
it can be copied and that this utility increase can be captured by the producer through a
price increase. However, like Gopal, Bhattacharjee, and Sanders (2005), they do not
provide empirical findings to substantiate their conclusions.
Oberholzer-Gee and Strumpf (2005) present empirical results that show no
negative impact of file sharing on traditional music distribution channels. Over the
course of four months, they monitor 1.75 million file downloads on file sharing networks
and then match the downloads to U.S. album sales data. Their empirical analysis shows
that music file sharing has no significant impact on album sales. Again, however, the
generalizability of their findings is somewhat limited as the authors use the “number of
German school kids on vacation” as an instrumental variable for file sharing activity to
bypass endogeneity problems caused by the simultaneity of downloading and purchasing
activity in their aggregate level data.
In summary, movie industry representatives argue that file sharing serves as a
substitute for commercial movie consumption, while no peer-reviewed research has
studied this relationship for movies, and the results from music file sharing research are
inconclusive and limited by methodological constraints. Moreover, no existing study
Page 60
56
surveys actual consumer decision making on an individual level, and no study uses
longitudinal data.
2.2.2 File Sharing Determinants: The Rochelandet–Le Guel Model
Related to the consequences of movie file sharing for commercial channels are the
factors that drive consumer file sharing. Research into these factors is also rare; we are
not aware of a single academic study that directly addresses this question. Again, some
scholars have researched file sharing determinants in the related context of music. Most
authors focus on the role of individual constructs for file sharing (ethical predispositions,
Gopal et al. 2004; consumer expertise, social networking, and moral judgments, Huang
2005), while Rochelandet and Le Guel (2005) attempt to integrate different drivers of
sharing illegal music copies in a comprehensive model.
Building on the Beckerian consumer utility framework, Rochelandet and Le Guel
(2005) propose that consumers prefer illegal copies of music over the original product
(i.e., a CD) when consuming the illegal copy offers greater utility. More specifically,
they argue that three groups of factors influence consumers’ utility perceptions of the
original and the illegal copy: (1) the utility derived from buying an original (including
both gross utility and costs), (2) the costs of the illegal copy (mainly transaction costs),
and (3) the degree of substitution between an original and its illegal copies. Rochelandet
and Le Guel (2005) find partial support for their model from a convenience sample of
2,500 French consumers. With an ordered logit approach, the factors in their model
explain 10% of the music file sharing intensity.
surveys actual consumer decision making on an individual level, and no study uses
longitudinal data.
2.2.2 File Sharing Determinants: The Rochelandet–Le Guel Model
Related to the consequences of movie file sharing for commercial channels are the
factors that drive consumer file sharing. Research into these factors is also rare; we are
not aware of a single academic study that directly addresses this question. Again, some
scholars have researched file sharing determinants in the related context of music. Most
authors focus on the role of individual constructs for file sharing (ethical predispositions,
Gopal et al. 2004; consumer expertise, social networking, and moral judgments, Huang
2005), while Rochelandet and Le Guel (2005) attempt to integrate different drivers of
sharing illegal music copies in a comprehensive model.
Building on the Beckerian consumer utility framework, Rochelandet and Le Guel
(2005) propose that consumers prefer illegal copies of music over the original product
(i.e., a CD) when consuming the illegal copy offers greater utility. More specifically,
they argue that three groups of factors influence consumers’ utility perceptions of the
original and the illegal copy: (1) the utility derived from buying an original (including
both gross utility and costs), (2) the costs of the illegal copy (mainly transaction costs),
and (3) the degree of substitution between an original and its illegal copies. Rochelandet
and Le Guel (2005) find partial support for their model from a convenience sample of
2,500 French consumers. With an ordered logit approach, the factors in their model
explain 10% of the music file sharing intensity.
Page 64
60
Figure 7: Structural Model of File Sharing Determinants
CO = illegal movie copy; OR = original movie.
Degree of substitution. A direct implication of the utility-theoretic approach is that
the degree to which a consumer perceives illegal movie copies to provide the same utility
as watching the original movie in a theater or on DVD determines the intensity of
consumer file sharing. This perceived degree of substitution influences the utility of the
illegal copy (Rochelandet and Le Guel 2005) and therefore should have a positive effect
on the intensity with which consumers obtain and watch illegal movie copies.
H4: The degree to which a consumer judges illegal movie copies as substitutes for
movies in commercial channels correlates positively with the number of illegal movie
copies a consumer obtains and the number of illegal copies she or he watches.
,4D3;?>7?D
@8 ;==793=
>@F;7 5@A;7C
23D5:;?9
@8 ;==793=
>@F;7 5@A;7C
/73B5: 5@CDC
!’,"
+@B3= 5@CDC
!’,"
*793= 5@CDC
!’,"
075:?;53= 5@CDC
!’,"
/E4CD;DED7
0B3?C35D;@? ED;=;DH
!’,"
+@4;=;DH ED;=;DH
!’,"
&?D;$;?6ECDBH ED;=;DH
!’,"
/@5;3= ED;=;DH
!’,
’@==75D;@? ED;=;DH
!’,"
’@?CE>7B
8;=7 C:3B;?9
<?@G=7697
’
@C
DC
@8
D
:7
;==
79
3=
5
@A
H
(
79
B7
7
@8
CE
4C
D;
DE
D;
@?
/A
75;8;5
ED;=;DH
@8 D:7
;==793= 5@AH
’
@?CE>
7B
<?@G
=7697
/D@B397 ED;=;DH
!’,"
)B@CC ED;=;DH
!,."
-B;57
!,."
0B3?C35D;@? 5@CDC
!,."
1
D;
=;D
H
@8
D
:7
@B
;9
;?
3=
Figure 7: Structural Model of File Sharing Determinants
CO = illegal movie copy; OR = original movie.
Degree of substitution. A direct implication of the utility-theoretic approach is that
the degree to which a consumer perceives illegal movie copies to provide the same utility
as watching the original movie in a theater or on DVD determines the intensity of
consumer file sharing. This perceived degree of substitution influences the utility of the
illegal copy (Rochelandet and Le Guel 2005) and therefore should have a positive effect
on the intensity with which consumers obtain and watch illegal movie copies.
H4: The degree to which a consumer judges illegal movie copies as substitutes for
movies in commercial channels correlates positively with the number of illegal movie
copies a consumer obtains and the number of illegal copies she or he watches.
,4D3;?>7?D
@8 ;==793=
>@F;7 5@A;7C
23D5:;?9
@8 ;==793=
>@F;7 5@A;7C
/73B5: 5@CDC
!’,"
+@B3= 5@CDC
!’,"
*793= 5@CDC
!’,"
075:?;53= 5@CDC
!’,"
/E4CD;DED7
0B3?C35D;@? ED;=;DH
!’,"
+@4;=;DH ED;=;DH
!’,"
&?D;$;?6ECDBH ED;=;DH
!’,"
/@5;3= ED;=;DH
!’,
’@==75D;@? ED;=;DH
!’,"
’@?CE>7B
8;=7 C:3B;?9
<?@G=7697
’
@C
DC
@8
D
:7
;==
79
3=
5
@A
H
(
79
B7
7
@8
CE
4C
D;
DE
D;
@?
/A
75;8;5
ED;=;DH
@8 D:7
;==793= 5@AH
’
@?CE>
7B
<?@G
=7697
/D@B397 ED;=;DH
!’,"
)B@CC ED;=;DH
!,."
-B;57
!,."
0B3?C35D;@? 5@CDC
!,."
1
D;
=;D
H
@8
D
:7
@B
;9
;?
3=
Page 67
63
Mobility utility. Illegal copies enhance consumers’ mobility, because they can be
stored on mobile devices (e.g., laptop computers, video iPods, PDAs), which enables
consumers to carry extensive movie libraries in minimal space when traveling. Because
this mobility is not possible with regular DVDs, it represents a specific utility of the copy
to consumers.
Storage utility. Related, due to their non-physical character, illegal copies require
less physical storage space in the consumer’s domicile than purchased DVDs, which can
represent a benefit for consumers.
Anti-industry utility. The movie industry is a frequent target of consumer criticism
for its treatment of movies as mere commercial products rather than art, as well as for the
prices it charges for movies in legal channels (e.g., Graham 2004)—an attitude which is
shared by certain industry insiders (e.g., director M. Night Shyamalan calls studios
“greedy, heartless, soulless, and disrespectful”; Guardian 2005). Consumers might
consider “stealing” a movie by watching an illegal copy a legitimate kind of revenge on
the industry and derive a benefit from this.
Social utility. Accumulating illegal movie copies enables consumers to establish
social links with relevant others. Consumers can interact with their peers about illegal
movie copies and related technology and thereby become part of a “social copying
network.” This allows the consumers to demonstrate their expertise and receive social
rewards for that expertise from others. Huang (2005) provides initial empirical support
for such social utility.
Collection utility. The availability of illegal movie copies enables consumers to
collect large numbers of movies, regardless of their financial resources. Consumer
behavior literature reports that consumers derive a utility from such collecting behavior
Mobility utility. Illegal copies enhance consumers’ mobility, because they can be
stored on mobile devices (e.g., laptop computers, video iPods, PDAs), which enables
consumers to carry extensive movie libraries in minimal space when traveling. Because
this mobility is not possible with regular DVDs, it represents a specific utility of the copy
to consumers.
Storage utility. Related, due to their non-physical character, illegal copies require
less physical storage space in the consumer’s domicile than purchased DVDs, which can
represent a benefit for consumers.
Anti-industry utility. The movie industry is a frequent target of consumer criticism
for its treatment of movies as mere commercial products rather than art, as well as for the
prices it charges for movies in legal channels (e.g., Graham 2004)—an attitude which is
shared by certain industry insiders (e.g., director M. Night Shyamalan calls studios
“greedy, heartless, soulless, and disrespectful”; Guardian 2005). Consumers might
consider “stealing” a movie by watching an illegal copy a legitimate kind of revenge on
the industry and derive a benefit from this.
Social utility. Accumulating illegal movie copies enables consumers to establish
social links with relevant others. Consumers can interact with their peers about illegal
movie copies and related technology and thereby become part of a “social copying
network.” This allows the consumers to demonstrate their expertise and receive social
rewards for that expertise from others. Huang (2005) provides initial empirical support
for such social utility.
Collection utility. The availability of illegal movie copies enables consumers to
collect large numbers of movies, regardless of their financial resources. Consumer
behavior literature reports that consumers derive a utility from such collecting behavior
Page 69
65
2.4 Testing File Sharing Consequences
In this section, we test the hypotheses that address the consequences of movie file
sharing (i.e., H1–H3) using data from a controlled longitudinal sample and ReLogit
logistic regression.
2.4.1 Data Collection and Sample
Understanding the effect that movie file sharing has on commercial channel usage
requires a controlled longitudinal study design, which avoids biases from a priori
differences in movie consumption intentions between file sharers and non–file sharers as
well as speculative ex-post “what-if” questions. We collected information from a quota
sample of 1,075 German consumers, using gender, age, and occupation as quota criteria.
The sample mirrors the German movie-going population in terms of key demographic
variables and movie consumption (see Table 5). Respondents filled out three different
Internet questionnaires over the course of eight months, for which they used personalized
identification numbers, so that we could connect the information provided by a
respondent at different points in time and avoid multiple responses on the same
questionnaire from the same respondent. Respondents received a €10 present for
completing all three questionnaires, and we also raffled off additional prices to
participants.
2.4 Testing File Sharing Consequences
In this section, we test the hypotheses that address the consequences of movie file
sharing (i.e., H1–H3) using data from a controlled longitudinal sample and ReLogit
logistic regression.
2.4.1 Data Collection and Sample
Understanding the effect that movie file sharing has on commercial channel usage
requires a controlled longitudinal study design, which avoids biases from a priori
differences in movie consumption intentions between file sharers and non–file sharers as
well as speculative ex-post “what-if” questions. We collected information from a quota
sample of 1,075 German consumers, using gender, age, and occupation as quota criteria.
The sample mirrors the German movie-going population in terms of key demographic
variables and movie consumption (see Table 5). Respondents filled out three different
Internet questionnaires over the course of eight months, for which they used personalized
identification numbers, so that we could connect the information provided by a
respondent at different points in time and avoid multiple responses on the same
questionnaire from the same respondent. Respondents received a €10 present for
completing all three questionnaires, and we also raffled off additional prices to
participants.
Page 70
66
Table 5: Sample Characteristics
Criterion Sample German Movie Consumer Population b
Gender (%)
Female 52.7 51
Male 47.1 49
Age Groups (%)
≤ 29 57.1 52
30-39 20.9 20
40-49 8.9 14
³ 50 13.1 14
Occupation (%)
Student/in education 40.8 35
Worker .7 7
Employee 40.1 36
Civil servant 5.5 6
Self-employed 6.1 3
Homemaker 2.7 3
Pensioner 1.4 10
Other 2.6 -
Movie consumption (per year)
Theater visits 8.2 (6)a 5.2
DVD purchases 4.1 (1)a 5.2
DVD rentals 10.3 (5)a 11.1
a Number in parentheses is the median.
b Percentages are for 2003 (FFA 2004); more recent data are not available for individual categories.
We contacted participants first in February 2006 and asked about their intentions
to watch between 10 and 15 new motion pictures in a movie theater or as an illegal copy.
The movies were a subset of a total of 25 movie titles covering all major studio releases
in Germany in the following months, with none of the movies having been available in
theaters or on DVD at that point. Five of the movies were action films, five comedies,
five dramas, five children’s movies, and five thrillers (the individual titles appear in
Table 5: Sample Characteristics
Criterion Sample German Movie Consumer Population b
Gender (%)
Female 52.7 51
Male 47.1 49
Age Groups (%)
≤ 29 57.1 52
30-39 20.9 20
40-49 8.9 14
³ 50 13.1 14
Occupation (%)
Student/in education 40.8 35
Worker .7 7
Employee 40.1 36
Civil servant 5.5 6
Self-employed 6.1 3
Homemaker 2.7 3
Pensioner 1.4 10
Other 2.6 -
Movie consumption (per year)
Theater visits 8.2 (6)a 5.2
DVD purchases 4.1 (1)a 5.2
DVD rentals 10.3 (5)a 11.1
a Number in parentheses is the median.
b Percentages are for 2003 (FFA 2004); more recent data are not available for individual categories.
We contacted participants first in February 2006 and asked about their intentions
to watch between 10 and 15 new motion pictures in a movie theater or as an illegal copy.
The movies were a subset of a total of 25 movie titles covering all major studio releases
in Germany in the following months, with none of the movies having been available in
theaters or on DVD at that point. Five of the movies were action films, five comedies,
five dramas, five children’s movies, and five thrillers (the individual titles appear in
Page 71
67
Appendix C). Each respondent began by indicating his or her preferred genres and
then answered questions with regard to the movies assigned to those genres. The
maximum of 15 movies (i.e., three genres) per respondent prevents cognitive overload;
we also set a minimum condition of 10 movies (i.e., two genres). Participants viewed a
poster of each movie, information about the director and cast, and a short synopsis of the
movie’s content.
We then contacted the respondents for the second time in May 2006, after each
surveyed movie had been theatrically released but before they were available on DVD for
either purchase or rental. In the second questionnaire, we collected information about
whether respondents had seen the surveyed movies in theaters and whether they had
obtained and/or watched illegal copies of the movies. Respondents also indicated whether
they intended to rent and/or buy certain movie titles on DVD after these DVDs would
have become available, and whether they intended to watch illegal copies of the movies.
For the second questionnaire, 813 panel members responded, a satisfactory retention rate
of 76%.
Finally, we provided the third questionnaire in October 2006, when 18 of the 25
surveyed movies had been available on DVD for at least four weeks, which reflects the
period when studios collect approximately two-thirds of a movie’s eventual total DVD
rental and sales revenues.9 This questionnaire mainly consisted of questions about
respondents’ rentals and purchases of the surveyed movies on DVD, and respondents
9 This estimate is based on proprietary information on the weekly revenue distribution of studio
movies, which we collected from Video Business Magazine (weekly DVD rental revenues) and Nielsen
VideoScan (DVD purchase revenues). Also see Figure 4.
Appendix C). Each respondent began by indicating his or her preferred genres and
then answered questions with regard to the movies assigned to those genres. The
maximum of 15 movies (i.e., three genres) per respondent prevents cognitive overload;
we also set a minimum condition of 10 movies (i.e., two genres). Participants viewed a
poster of each movie, information about the director and cast, and a short synopsis of the
movie’s content.
We then contacted the respondents for the second time in May 2006, after each
surveyed movie had been theatrically released but before they were available on DVD for
either purchase or rental. In the second questionnaire, we collected information about
whether respondents had seen the surveyed movies in theaters and whether they had
obtained and/or watched illegal copies of the movies. Respondents also indicated whether
they intended to rent and/or buy certain movie titles on DVD after these DVDs would
have become available, and whether they intended to watch illegal copies of the movies.
For the second questionnaire, 813 panel members responded, a satisfactory retention rate
of 76%.
Finally, we provided the third questionnaire in October 2006, when 18 of the 25
surveyed movies had been available on DVD for at least four weeks, which reflects the
period when studios collect approximately two-thirds of a movie’s eventual total DVD
rental and sales revenues.9 This questionnaire mainly consisted of questions about
respondents’ rentals and purchases of the surveyed movies on DVD, and respondents
9 This estimate is based on proprietary information on the weekly revenue distribution of studio
movies, which we collected from Video Business Magazine (weekly DVD rental revenues) and Nielsen
VideoScan (DVD purchase revenues). Also see Figure 4.
Page 75
71
releases, so that p > t. We calculate the t parameters on the basis of publicly available
information, with tTheater = .0126, tRental = .0103, and tPurchase = .0040.11
2.4.5 Theater-Related Results
To account for potential differences between consumers’ obtaining and watching
illegal copies, we run three ReLogit models to test the impact of illegal file sharing on
movie theater visits. In each model, we include the respondents’ intentions to watch an
illegal copy of a movie (measured in the first questionnaire) and their actual file sharing
behavior (dichotomous factor, measured in the second questionnaire) as regressors and
actual theater-going behavior as the binary dependent variable. To rule out potential
endogenous effects which have troubled previous research on file sharing, we exclude
those cases in which theatrical consumption precedes file sharing (n = 10), taking
advantage of our individual-level longitudinal empirical design (in contrast to the
aggregate level, cross-sectional design of previous studies). As a result, the independent
variables in our ReLogit analyses are unaffected by the dependent variable (i.e., the
consumer’s theater visit).12 In the first model (the “overall model”), we set file sharing
11 We calculate tTheater by dividing the number of theater visits in Germany in 2005 (127.3
million) by the product of the number of movies released in Germany (372) and the number of German
movie consumers (27.2 million). This calculation provides the percentage of all movie-going decisions
that lead to a theater visit. Analogously, we calculate tRental based on 102.9 million rentals of current
feature film DVDs and tPurchase based on 39.8 million new feature film DVDs sold, with 369 new feature
film DVD releases in 2005. We obtain all data used to calculate the t parameters from SPIO (2006) and
BAM (2006). pTheater is .083, pRental.063, and pPurchase.013. In addition, we apply the Zelig version of
ReLogit, which offers minor advantages over other versions.
12 To provide empirical evidence for the absence of endogenous effects, we conduct the Durbin-
Wu-Hausman augmented regression test for endogeneity (Davidson and MacKinnon 1993). Consistent
with our theoretical argument, we find the error term of the file sharing regression to be clearly non-
significant in the theater visits regression equations, which means that file sharing is indeed an exogenous
variable as specified and that the results are unbiased by endogeneity. We conduct the same test for the
DVD rental and DVD purchase equation and again find file sharing to be exogenous.
releases, so that p > t. We calculate the t parameters on the basis of publicly available
information, with tTheater = .0126, tRental = .0103, and tPurchase = .0040.11
2.4.5 Theater-Related Results
To account for potential differences between consumers’ obtaining and watching
illegal copies, we run three ReLogit models to test the impact of illegal file sharing on
movie theater visits. In each model, we include the respondents’ intentions to watch an
illegal copy of a movie (measured in the first questionnaire) and their actual file sharing
behavior (dichotomous factor, measured in the second questionnaire) as regressors and
actual theater-going behavior as the binary dependent variable. To rule out potential
endogenous effects which have troubled previous research on file sharing, we exclude
those cases in which theatrical consumption precedes file sharing (n = 10), taking
advantage of our individual-level longitudinal empirical design (in contrast to the
aggregate level, cross-sectional design of previous studies). As a result, the independent
variables in our ReLogit analyses are unaffected by the dependent variable (i.e., the
consumer’s theater visit).12 In the first model (the “overall model”), we set file sharing
11 We calculate tTheater by dividing the number of theater visits in Germany in 2005 (127.3
million) by the product of the number of movies released in Germany (372) and the number of German
movie consumers (27.2 million). This calculation provides the percentage of all movie-going decisions
that lead to a theater visit. Analogously, we calculate tRental based on 102.9 million rentals of current
feature film DVDs and tPurchase based on 39.8 million new feature film DVDs sold, with 369 new feature
film DVD releases in 2005. We obtain all data used to calculate the t parameters from SPIO (2006) and
BAM (2006). pTheater is .083, pRental.063, and pPurchase.013. In addition, we apply the Zelig version of
ReLogit, which offers minor advantages over other versions.
12 To provide empirical evidence for the absence of endogenous effects, we conduct the Durbin-
Wu-Hausman augmented regression test for endogeneity (Davidson and MacKinnon 1993). Consistent
with our theoretical argument, we find the error term of the file sharing regression to be clearly non-
significant in the theater visits regression equations, which means that file sharing is indeed an exogenous
variable as specified and that the results are unbiased by endogeneity. We conduct the same test for the
DVD rental and DVD purchase equation and again find file sharing to be exogenous.
Page 82
78
(with a control for file sharing intentions) but do not watch it, their probability of
purchasing the DVD is higher than it is for consumers who have not obtained an illegal
copy. In such cases, the copy does not serve as a substitute for the DVD but rather
stimulates consumers’ desire to see the movie in a legal channel.
In summary, we find support for H1b and H1c (which state file sharing intentions
to diminish DVD rentals and purchases), but neither for H2b and H2c,(which posit a
negative effect of obtaining illegal copies on the two DVD channels) nor for H3b, or H3c
(which argue that the watching of copies cannibalizes DVD rentals and purchases). In the
case of H2c, we even find a significant positive effect instead of the proposed negative
effect. As an aside, the three movie characteristics play lesser roles for DVD
consumption than in the theater channel. Although in the DVD rental context, the user
rating positively influences decisions to rent a specific movie on DVD, screens and
theater attendance are not significant; for DVD purchase decisions, none of the movie
characteristics is significant. A likely explanation is that once movies have appeared in
theaters, extensive quality-related information becomes available, which is then
incorporated into the consumers’ intention to rent or purchase the movie on DVD.
As in the case of theater visits, we use the ReLogit estimations to speculate about
the strength of the industry-wide effect of movie file sharing on DVD rentals and
purchases.16 In a fictitious constellation without any illegal movie copies (but file sharing
intentions remaining unchanged), DVD rentals would increase by only 0.1% (from 103.4
million to 103.5 million transactions), producing approximately $0.5 million of
16 When calculating the industry-wide effect of file sharing on DVD rentals and purchases, we use
the same approach as in the case of theater visits (see footnotes 13 and 14).
(with a control for file sharing intentions) but do not watch it, their probability of
purchasing the DVD is higher than it is for consumers who have not obtained an illegal
copy. In such cases, the copy does not serve as a substitute for the DVD but rather
stimulates consumers’ desire to see the movie in a legal channel.
In summary, we find support for H1b and H1c (which state file sharing intentions
to diminish DVD rentals and purchases), but neither for H2b and H2c,(which posit a
negative effect of obtaining illegal copies on the two DVD channels) nor for H3b, or H3c
(which argue that the watching of copies cannibalizes DVD rentals and purchases). In the
case of H2c, we even find a significant positive effect instead of the proposed negative
effect. As an aside, the three movie characteristics play lesser roles for DVD
consumption than in the theater channel. Although in the DVD rental context, the user
rating positively influences decisions to rent a specific movie on DVD, screens and
theater attendance are not significant; for DVD purchase decisions, none of the movie
characteristics is significant. A likely explanation is that once movies have appeared in
theaters, extensive quality-related information becomes available, which is then
incorporated into the consumers’ intention to rent or purchase the movie on DVD.
As in the case of theater visits, we use the ReLogit estimations to speculate about
the strength of the industry-wide effect of movie file sharing on DVD rentals and
purchases.16 In a fictitious constellation without any illegal movie copies (but file sharing
intentions remaining unchanged), DVD rentals would increase by only 0.1% (from 103.4
million to 103.5 million transactions), producing approximately $0.5 million of
16 When calculating the industry-wide effect of file sharing on DVD rentals and purchases, we use
the same approach as in the case of theater visits (see footnotes 13 and 14).
Page 84
80
three latent variables for the different facets of the original movie’s utility (gross utility,
price, and transaction costs), four latent variables to address the different kinds of
transaction costs associated with the copy (search, moral, legal, and technical costs), and
one latent variable for each of the six specific utilities of the copy (transaction, mobility,
storage, collecting, anti-industry, and social utility). The model also contains the degree
of substitution and the consumer’s file sharing knowledge as determinants of watching
and obtaining illegal copies, and links from obtaining to watching illegal copies and from
file sharing knowledge to search costs.
We measure both obtaining and watching illegal copies with reflective, three-item
scales that combine respondents’ actual file sharing behavior with regard to the movies in
our study with two more global measures. Specifically, we measured the obtainment of
illegal movie copies as the number of copies of surveyed movies the consumer had
actually obtained, the total number of illegal copies obtained within the year preceding
the first questionnaire, and the answer to the same question from the second
questionnaire. For watching illegal movie copies, we used the number of surveyed
movies a respondent watched as illegal copies and the total number of watched illegal
copies within the 12 months preceding the first and the second questionnaire,
respectively. To measure file sharing determinants, we use existing scales when available
and develop new scales for the rest, most of which take a formative nature. Except for the
six specific utility variables, which we measure with one item each due to space
restrictions, we use multiple items for all constructs (see Appendix D).
The reliability of the reflective scales is generally satisfactory. Obtainment and
watching of illegal copies achieve alpha scores of .72 and .67, respectively, acceptable
for a combination of surveyed and general past behavior, as well as the lack of
established scales in the researched domain (Peter 1979). On the other reflective scales,
three latent variables for the different facets of the original movie’s utility (gross utility,
price, and transaction costs), four latent variables to address the different kinds of
transaction costs associated with the copy (search, moral, legal, and technical costs), and
one latent variable for each of the six specific utilities of the copy (transaction, mobility,
storage, collecting, anti-industry, and social utility). The model also contains the degree
of substitution and the consumer’s file sharing knowledge as determinants of watching
and obtaining illegal copies, and links from obtaining to watching illegal copies and from
file sharing knowledge to search costs.
We measure both obtaining and watching illegal copies with reflective, three-item
scales that combine respondents’ actual file sharing behavior with regard to the movies in
our study with two more global measures. Specifically, we measured the obtainment of
illegal movie copies as the number of copies of surveyed movies the consumer had
actually obtained, the total number of illegal copies obtained within the year preceding
the first questionnaire, and the answer to the same question from the second
questionnaire. For watching illegal movie copies, we used the number of surveyed
movies a respondent watched as illegal copies and the total number of watched illegal
copies within the 12 months preceding the first and the second questionnaire,
respectively. To measure file sharing determinants, we use existing scales when available
and develop new scales for the rest, most of which take a formative nature. Except for the
six specific utility variables, which we measure with one item each due to space
restrictions, we use multiple items for all constructs (see Appendix D).
The reliability of the reflective scales is generally satisfactory. Obtainment and
watching of illegal copies achieve alpha scores of .72 and .67, respectively, acceptable
for a combination of surveyed and general past behavior, as well as the lack of
established scales in the researched domain (Peter 1979). On the other reflective scales,
Page 86
82
Table 7: Descriptive Statistics and Correlations
Ma SDa (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)
1 Obtainment of illegal movie copies 14.04 29.76 .72
2 Substitute 11.40 3.85 .16 n.a.
3 Gross utility (OR) 25.35 7.75 -.14 -.07 n.a.
4 File sharing knowledge 16.14 10.68 .31 -.00 -.15 n.a.
5 Watching illegal movie copies 11.82 21.43 .89 .19 -.11 .29 .67
6 Price (OR) 24.79 9.26 .09 -.10 .13 .09 .09 n.a.
7 Transaction costs (OR) 18.78 11.49 .12 .07 -.01 -.04 .11 .08 n.a.
8 Moral costs (CO) 11.18 3.76 -.12 -.05 .07 -.06 -.12 .05 -.01 .84
9 Legal costs (CO) 7.56 2.49 .03 .06 -.02 -.01 .01 .06 -.00 .28 .71
10 Technical costs (CO) 6.34 2.93 -.11 -.02 .11 -.25 -.14 -.03 -.02 .32 .34 .86
11 Search costs (CO) 8.22 6.57 -.16 -.00 .16 -.24 -.15 -.04 .05 .22 .12 .33 n.a.
12 Transaction utility (CO) 3.15 1.77 .17 .29 -.03 .13 .19 .05 .09 -.07 .02 -.01 -.05 n.a.
13 Collection utility (CO) 1.53 1.00 .32 .25 -.06 .26 .29 .02 .05 -.05 .01 -.08 -.12 .22 n.a.
14 Anti-industry utility (CO) 1.53 1.07 .22 .02 -.12 .30 .21 .06 .06 -.04 .02 .05 -.06 .21 .18 n.a.
15 Storage utility (CO) 2.09 1.51 .24 .13 -.08 .33 .25 .04 .03 -.02 .10 -.06 -.09 .30 .34 .28 n.a.
16 Social utility (CO) 1.41 .87 .24 .10 -.01 .25 .21 .01 .06 .04 .10 .10 .01 .29 .28 .42 .39 n.a.
17 Mobility utility (CO) 2.65 1.81 .21 .14 -.07 .31 .23 .04 .01 -.01 .06 -.05 -.17 .29 .33 .26 .50 .30 n.a.
Numbers on the diagonal are Cronbach’s alpha scores. n.a. = no alpha score calculated because the construct is measured by a formative scale or single item. CO =
illegal movie copy; OR = original movie.
a Means and standard deviations are calculated for the sum of construct items.
Table 7: Descriptive Statistics and Correlations
Ma SDa (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)
1 Obtainment of illegal movie copies 14.04 29.76 .72
2 Substitute 11.40 3.85 .16 n.a.
3 Gross utility (OR) 25.35 7.75 -.14 -.07 n.a.
4 File sharing knowledge 16.14 10.68 .31 -.00 -.15 n.a.
5 Watching illegal movie copies 11.82 21.43 .89 .19 -.11 .29 .67
6 Price (OR) 24.79 9.26 .09 -.10 .13 .09 .09 n.a.
7 Transaction costs (OR) 18.78 11.49 .12 .07 -.01 -.04 .11 .08 n.a.
8 Moral costs (CO) 11.18 3.76 -.12 -.05 .07 -.06 -.12 .05 -.01 .84
9 Legal costs (CO) 7.56 2.49 .03 .06 -.02 -.01 .01 .06 -.00 .28 .71
10 Technical costs (CO) 6.34 2.93 -.11 -.02 .11 -.25 -.14 -.03 -.02 .32 .34 .86
11 Search costs (CO) 8.22 6.57 -.16 -.00 .16 -.24 -.15 -.04 .05 .22 .12 .33 n.a.
12 Transaction utility (CO) 3.15 1.77 .17 .29 -.03 .13 .19 .05 .09 -.07 .02 -.01 -.05 n.a.
13 Collection utility (CO) 1.53 1.00 .32 .25 -.06 .26 .29 .02 .05 -.05 .01 -.08 -.12 .22 n.a.
14 Anti-industry utility (CO) 1.53 1.07 .22 .02 -.12 .30 .21 .06 .06 -.04 .02 .05 -.06 .21 .18 n.a.
15 Storage utility (CO) 2.09 1.51 .24 .13 -.08 .33 .25 .04 .03 -.02 .10 -.06 -.09 .30 .34 .28 n.a.
16 Social utility (CO) 1.41 .87 .24 .10 -.01 .25 .21 .01 .06 .04 .10 .10 .01 .29 .28 .42 .39 n.a.
17 Mobility utility (CO) 2.65 1.81 .21 .14 -.07 .31 .23 .04 .01 -.01 .06 -.05 -.17 .29 .33 .26 .50 .30 n.a.
Numbers on the diagonal are Cronbach’s alpha scores. n.a. = no alpha score calculated because the construct is measured by a formative scale or single item. CO =
illegal movie copy; OR = original movie.
a Means and standard deviations are calculated for the sum of construct items.
Page 88
84
Table 8: Impact of Determinants of File Sharing Behavior
Effect of On Path Coefficient (t-
Value)
Total Effect
(t-Value)
Utility of the original
Gross utility (OR) Obtainment of illegal movie copies -.071 (1.84)* -.071 (1.84)*
Price (OR) Obtainment of illegal movie copies .075 (1.06) .075 (1.06)
Transaction costs (OR) Obtainment of illegal movie copies .100 (2.03)** .100 (2.03)**
Gross utility (OR) Watching illegal movie copies .012 (.82) -.0503 (1.29)
Price (OR) Watching illegal movie copies .009 (.03) .075 (1.02)
Transaction costs (OR) Watching illegal movie copies -.009 (.41) .079 (1.68)*
Costs of the illegal copy
Search costs (CO) Obtainment of illegal movie copies -.063 (1.79)* -.063 (1.79)*
Moral costs (CO) Obtainment of illegal movie copies -.087 (2.62)** -.087 (2.62)**
Legal costs (CO) Obtainment of illegal movie copies .044 (1.00) .044 (1.00)
Technical costs (CO) Obtainment of illegal movie copies -.019 (.57) -.019 (.57)
Search costs (CO) Watching illegal movie copies .015 (.82) -.040 (1.15)
Moral costs (CO) Watching illegal movie copies .007 (.41) -.069 (2.14)**
Legal costs (CO) Watching illegal movie copies -.016 (.80) .022 (.50)
Technical costs (CO) Watching illegal movie copies -.040 (2.00)** -.056 (1.78)*
Degree of substitution
Substitute Obtainment of illegal movie copies .089 (2.66)** .089 (2.66)**
Substitute Watching illegal movie copies .040 (2.16)** .120 (4.01)**
Specific utility of the illegal copy
Transaction utility (CO) Obtainment of illegal movie copies .012 (.34) .012 (.34)
Collection utility (CO) Obtainment of illegal movie copies .178 (2.90)** .178 (2.90)**
Mobility utility (CO) Obtainment of illegal movie copies -.002 (.04) -.002 (.04)
Storage utility (CO) Obtainment of illegal movie copies .035 (.78) .035 (.78)
Anti-industry utility (CO) Obtainment of illegal movie copies .064 (1.70)* .064 (1.70)*
Social utility (CO) Obtainment of illegal movie copies .085 (1.46) .085 (1.46)
Transaction utility (CO) Watching illegal movie copies .019 (.93) .030 (.80)
Collection utility (CO) Watching illegal movie copies -.018 (.69) .138 (2.07)**
Mobility utility (CO) Watching illegal movie copies .036 (1.56) .035 (.92)
Storage utility (CO) Watching illegal movie copies .029 (1.39) .060 (1.32)
Anti-industry utility (CO) Watching illegal movie copies .010 (.52) .066 (1.73)*
Social utility (CO) Watching illegal movie copies -.022 (.90) .053 (.97)
File sharing knowledge
File sharing knowledge Obtainment of illegal movie copies .172 (4.83)** .187 (5.28)**
File sharing knowledge Watching illegal movie copies -.003 (.17) .156 (4.60)**
File sharing knowledge Search costs (CO) -.236 (2.12)** -.236 (2.12)**
Additional path
Obtainment of illegal
movie copies
Watching illegal movie copies .875 (26.27)** .875 (26.27)**
Notes: OR = original commercial movie consumption, CO = illegal movie copy. T-values are calculated
through a bootstrapping routine with 813 cases and 500 samples.
* p < .05 (one-sided).
** p < .01 (one-sided).
Table 8: Impact of Determinants of File Sharing Behavior
Effect of On Path Coefficient (t-
Value)
Total Effect
(t-Value)
Utility of the original
Gross utility (OR) Obtainment of illegal movie copies -.071 (1.84)* -.071 (1.84)*
Price (OR) Obtainment of illegal movie copies .075 (1.06) .075 (1.06)
Transaction costs (OR) Obtainment of illegal movie copies .100 (2.03)** .100 (2.03)**
Gross utility (OR) Watching illegal movie copies .012 (.82) -.0503 (1.29)
Price (OR) Watching illegal movie copies .009 (.03) .075 (1.02)
Transaction costs (OR) Watching illegal movie copies -.009 (.41) .079 (1.68)*
Costs of the illegal copy
Search costs (CO) Obtainment of illegal movie copies -.063 (1.79)* -.063 (1.79)*
Moral costs (CO) Obtainment of illegal movie copies -.087 (2.62)** -.087 (2.62)**
Legal costs (CO) Obtainment of illegal movie copies .044 (1.00) .044 (1.00)
Technical costs (CO) Obtainment of illegal movie copies -.019 (.57) -.019 (.57)
Search costs (CO) Watching illegal movie copies .015 (.82) -.040 (1.15)
Moral costs (CO) Watching illegal movie copies .007 (.41) -.069 (2.14)**
Legal costs (CO) Watching illegal movie copies -.016 (.80) .022 (.50)
Technical costs (CO) Watching illegal movie copies -.040 (2.00)** -.056 (1.78)*
Degree of substitution
Substitute Obtainment of illegal movie copies .089 (2.66)** .089 (2.66)**
Substitute Watching illegal movie copies .040 (2.16)** .120 (4.01)**
Specific utility of the illegal copy
Transaction utility (CO) Obtainment of illegal movie copies .012 (.34) .012 (.34)
Collection utility (CO) Obtainment of illegal movie copies .178 (2.90)** .178 (2.90)**
Mobility utility (CO) Obtainment of illegal movie copies -.002 (.04) -.002 (.04)
Storage utility (CO) Obtainment of illegal movie copies .035 (.78) .035 (.78)
Anti-industry utility (CO) Obtainment of illegal movie copies .064 (1.70)* .064 (1.70)*
Social utility (CO) Obtainment of illegal movie copies .085 (1.46) .085 (1.46)
Transaction utility (CO) Watching illegal movie copies .019 (.93) .030 (.80)
Collection utility (CO) Watching illegal movie copies -.018 (.69) .138 (2.07)**
Mobility utility (CO) Watching illegal movie copies .036 (1.56) .035 (.92)
Storage utility (CO) Watching illegal movie copies .029 (1.39) .060 (1.32)
Anti-industry utility (CO) Watching illegal movie copies .010 (.52) .066 (1.73)*
Social utility (CO) Watching illegal movie copies -.022 (.90) .053 (.97)
File sharing knowledge
File sharing knowledge Obtainment of illegal movie copies .172 (4.83)** .187 (5.28)**
File sharing knowledge Watching illegal movie copies -.003 (.17) .156 (4.60)**
File sharing knowledge Search costs (CO) -.236 (2.12)** -.236 (2.12)**
Additional path
Obtainment of illegal
movie copies
Watching illegal movie copies .875 (26.27)** .875 (26.27)**
Notes: OR = original commercial movie consumption, CO = illegal movie copy. T-values are calculated
through a bootstrapping routine with 813 cases and 500 samples.
* p < .05 (one-sided).
** p < .01 (one-sided).
Page 89
85
8 of the 15 determinant constructs in the model have significant impacts.
Specifically, as we propose in H4, the degree of substitution between illegal copies and
movie originals increases both obtainment and watching of illegal copies. Regarding the
utility of the original, we find that the original’s transaction costs raise the extent of
obtainment, as proposed in H5b, in addition to the negative effect of the original’s gross
utility mentioned above. The latter effect might result from the lower reference point for the
utility of the original for consumers who possess more file sharing knowledge. In other
words, file sharing skills might reduce the utility consumers derive from seeing a movie in a
commercial channel, because they know how to get the same movie free of charge. In
support of this argument, when we add a path from file sharing knowledge to gross utility,
the path from gross utility to file sharing becomes insignificant.
With regard to the transaction costs of the copy, three individual drivers are
significantly correlated with file sharing, in support of H6. Whereas both search and moral
costs provide hurdles to the consumer obtaining illegal copies, technical costs directly
reduce the probability that a customer watches such copies. Two specific utilities of the
copy enhance obtainment: perceptions of illegal movie copies as collectibles (the strongest
direct effect of all determinants) and the consumer’s anti-industry attitude, which makes file
sharing a kind of revenge action. These findings support H7. The consumers’ file sharing
knowledge facilitates obtainment of illegal copies directly, as well as by lowering search
costs, as we hypothesize in H8a and H8b.
As we expected, watching illegal movie copies correlates strongly with the extent of
obtainment. Except for technical costs and degree of substitution, which also exhibit
significant direct paths to watching illegal movie copies, all determinant constructs in the
model influence illegal watching not directly but only through obtainment, which serves as
a full mediator.
8 of the 15 determinant constructs in the model have significant impacts.
Specifically, as we propose in H4, the degree of substitution between illegal copies and
movie originals increases both obtainment and watching of illegal copies. Regarding the
utility of the original, we find that the original’s transaction costs raise the extent of
obtainment, as proposed in H5b, in addition to the negative effect of the original’s gross
utility mentioned above. The latter effect might result from the lower reference point for the
utility of the original for consumers who possess more file sharing knowledge. In other
words, file sharing skills might reduce the utility consumers derive from seeing a movie in a
commercial channel, because they know how to get the same movie free of charge. In
support of this argument, when we add a path from file sharing knowledge to gross utility,
the path from gross utility to file sharing becomes insignificant.
With regard to the transaction costs of the copy, three individual drivers are
significantly correlated with file sharing, in support of H6. Whereas both search and moral
costs provide hurdles to the consumer obtaining illegal copies, technical costs directly
reduce the probability that a customer watches such copies. Two specific utilities of the
copy enhance obtainment: perceptions of illegal movie copies as collectibles (the strongest
direct effect of all determinants) and the consumer’s anti-industry attitude, which makes file
sharing a kind of revenge action. These findings support H7. The consumers’ file sharing
knowledge facilitates obtainment of illegal copies directly, as well as by lowering search
costs, as we hypothesize in H8a and H8b.
As we expected, watching illegal movie copies correlates strongly with the extent of
obtainment. Except for technical costs and degree of substitution, which also exhibit
significant direct paths to watching illegal movie copies, all determinant constructs in the
model influence illegal watching not directly but only through obtainment, which serves as
a full mediator.
Page 91
87
(controlling for file sharing intentions) significantly impacts legal consumption only when
the consumer has actually watched the copy. In addition, consumers’ intentions to watch a
movie copy significantly reduce the number of DVD rentals and purchases. Obtainment of
illegal copies does not affect rental transactions and exerts a positive impact on DVD
purchases when the consumer has not watched the copy. The latter effect means that an
illegal movie copy can function as a cue for purchasing the DVD of a movie. In cases where
the copy obtained by the consumer is broken or of a low quality, it can be argued that the
consumer’s positive anticipation of watching the movie is re-routed into a purchasing act. If
the copy is working, the mere presence and resulting salience of the copy seems to heighten
the consumer’s emotional and intellectual involvement with the movie title, which
subsequently stimulates the consumer to purchase the DVD of the movie (i.e., to “go for the
original”). However, the positive impact of obtainment on DVD purchases is clearly less
strong than the negative impact of file sharing intentions. We calculate an overall annual
industry loss of $300 million in Germany, which represents approximately 9.4% of the total
industry revenues in 2005. Even when taking into account the assumptions of our method
and sample, we consider these numbers substantial.
Three major implications arise from these results. First, the movie industry is right to
proclaim that consumer file sharing destroys a significant amount of its revenues. Second,
consumers’ intentions to engage in file sharing cause them to forgo theater visits, legal
DVD rentals, or legal DVD purchases. Therefore, decreasing consumers’ intention to watch
illegal movie copies may be the most powerful way to fight movie piracy. A reduction in the
number of illegal copies would have a much lesser (or even no) impact on piracy, as long as
intentions remain unaffected. Third, though our nationwide estimates represent bold
numbers, they also demonstrate that recent industry claims exaggerate the true impact of file
sharing. Some industry representatives argue that each illegal copy represents a lost theater
(controlling for file sharing intentions) significantly impacts legal consumption only when
the consumer has actually watched the copy. In addition, consumers’ intentions to watch a
movie copy significantly reduce the number of DVD rentals and purchases. Obtainment of
illegal copies does not affect rental transactions and exerts a positive impact on DVD
purchases when the consumer has not watched the copy. The latter effect means that an
illegal movie copy can function as a cue for purchasing the DVD of a movie. In cases where
the copy obtained by the consumer is broken or of a low quality, it can be argued that the
consumer’s positive anticipation of watching the movie is re-routed into a purchasing act. If
the copy is working, the mere presence and resulting salience of the copy seems to heighten
the consumer’s emotional and intellectual involvement with the movie title, which
subsequently stimulates the consumer to purchase the DVD of the movie (i.e., to “go for the
original”). However, the positive impact of obtainment on DVD purchases is clearly less
strong than the negative impact of file sharing intentions. We calculate an overall annual
industry loss of $300 million in Germany, which represents approximately 9.4% of the total
industry revenues in 2005. Even when taking into account the assumptions of our method
and sample, we consider these numbers substantial.
Three major implications arise from these results. First, the movie industry is right to
proclaim that consumer file sharing destroys a significant amount of its revenues. Second,
consumers’ intentions to engage in file sharing cause them to forgo theater visits, legal
DVD rentals, or legal DVD purchases. Therefore, decreasing consumers’ intention to watch
illegal movie copies may be the most powerful way to fight movie piracy. A reduction in the
number of illegal copies would have a much lesser (or even no) impact on piracy, as long as
intentions remain unaffected. Third, though our nationwide estimates represent bold
numbers, they also demonstrate that recent industry claims exaggerate the true impact of file
sharing. Some industry representatives argue that each illegal copy represents a lost theater
Page 93
89
illegal copies than that achieved by previous studies (Rochelandet and Le Guel 2005). The
PLS results highlight that each driver category contributes to consumer file sharing, though
to differing extents. The three drivers that exert the strongest direct impact are the collecting
utility of the copy, consumers’ file sharing knowledge, and transaction costs of the original;
we present the first two drivers for the first time here.
Our analysis also shows that file sharing occurs because of various factors, several of
which offer antipiracy organizations very specific starting points for countermeasures.
Specifically, stressing the unethical element of appropriating copyrighted content without
compensating the copyright owner in marketing campaigns could increase the moral costs of
illegal file sharing and lower file sharing activities. Similarly, because the transaction costs
of commercial channels motivate consumer file sharing, movie producers should think about
ways to reduce them. When watching a movie in theaters during its opening weekend is the
only way to access a new movie legally, customers must pay the accompanying transaction
costs that go far beyond the ticket price (e.g., babysitters and concession prices can make a
single movie easily cost $50; Puig 2005) and therefore feel pushed toward illegal channels
such as file sharing. Making movies available through new channels, such as video-on-
demand, that involve lower transaction costs for the consumers and shortening the time gap
between the theater and home entertainment channels might be an appropriate way to win
back transaction cost–sensitive consumers. However, this strategy could cause other
problems, such as increased interchannel cannibalization (Lehmann and Weinberg 2000).
Yet another starting point for reducing file sharing considers the degree of substitution
perceived by the customer. Although substitutability lies in the eye of the beholder, studios
may want to stress the uniqueness of legal movie consumption or add features and elements
to legal movie consumption that can hardly be included in illegal copies. Such elements
might include events in the theater that stress the social element of movie-going or attractive
illegal copies than that achieved by previous studies (Rochelandet and Le Guel 2005). The
PLS results highlight that each driver category contributes to consumer file sharing, though
to differing extents. The three drivers that exert the strongest direct impact are the collecting
utility of the copy, consumers’ file sharing knowledge, and transaction costs of the original;
we present the first two drivers for the first time here.
Our analysis also shows that file sharing occurs because of various factors, several of
which offer antipiracy organizations very specific starting points for countermeasures.
Specifically, stressing the unethical element of appropriating copyrighted content without
compensating the copyright owner in marketing campaigns could increase the moral costs of
illegal file sharing and lower file sharing activities. Similarly, because the transaction costs
of commercial channels motivate consumer file sharing, movie producers should think about
ways to reduce them. When watching a movie in theaters during its opening weekend is the
only way to access a new movie legally, customers must pay the accompanying transaction
costs that go far beyond the ticket price (e.g., babysitters and concession prices can make a
single movie easily cost $50; Puig 2005) and therefore feel pushed toward illegal channels
such as file sharing. Making movies available through new channels, such as video-on-
demand, that involve lower transaction costs for the consumers and shortening the time gap
between the theater and home entertainment channels might be an appropriate way to win
back transaction cost–sensitive consumers. However, this strategy could cause other
problems, such as increased interchannel cannibalization (Lehmann and Weinberg 2000).
Yet another starting point for reducing file sharing considers the degree of substitution
perceived by the customer. Although substitutability lies in the eye of the beholder, studios
may want to stress the uniqueness of legal movie consumption or add features and elements
to legal movie consumption that can hardly be included in illegal copies. Such elements
might include events in the theater that stress the social element of movie-going or attractive
Page 95
91
mirrors the German movie consumer population in demographics, we concede it is not a true
random sample. However, post hoc comparisons show that other criteria, such as movie
consumption patterns, are similar between the sample and the relevant population. Fourth,
our measurement approach enables us to separate the effects of consumer file sharing
intentions and behaviors on movie consumption on the basis of a controlled longitudinal
study, but the survey method means we must rely on consumer self-reported data instead of
on “objective” data. We believe this limitation does not strongly affect the results though,
because we use actual movie titles, measure specific behavioral variables, and avoid any
kind of moral bias in the questionnaires. Fifth, we acknowledge that the consumer x movie
observations in our data are not completely independent, which, however, reflects reality as
some consumers will watch several movies in a given period while other consumers will
watch only one. Sixth and final, we had to develop several scales ourselves because of the
limited extant research on movie file sharing. Although these scales indicate solid reliability
and validity, further research into their quality would certainly be helpful. This
recommendation is particularly applicable to those determinant variables that we measure
using single items.
mirrors the German movie consumer population in demographics, we concede it is not a true
random sample. However, post hoc comparisons show that other criteria, such as movie
consumption patterns, are similar between the sample and the relevant population. Fourth,
our measurement approach enables us to separate the effects of consumer file sharing
intentions and behaviors on movie consumption on the basis of a controlled longitudinal
study, but the survey method means we must rely on consumer self-reported data instead of
on “objective” data. We believe this limitation does not strongly affect the results though,
because we use actual movie titles, measure specific behavioral variables, and avoid any
kind of moral bias in the questionnaires. Fifth, we acknowledge that the consumer x movie
observations in our data are not completely independent, which, however, reflects reality as
some consumers will watch several movies in a given period while other consumers will
watch only one. Sixth and final, we had to develop several scales ourselves because of the
limited extant research on movie file sharing. Although these scales indicate solid reliability
and validity, further research into their quality would certainly be helpful. This
recommendation is particularly applicable to those determinant variables that we measure
using single items.
Page 97
93
from rational expectancy-value decision making, such as preference reversals when
focusing on anticipated emotions instead of focusing on product attributes (Caruso & Shafir,
2006; Shiv & Huber, 2000), the reliance on affect versus cognition under processing
constraints (Shiv & Fedorikhin, 1999), the impact of body feedback, meta-cognitive feelings
and moods on choice (Förster, 2004; Lee, 2004; Pham, 2004; Schwarz, 2004), or the
importance of contextual factors in determining affect and preferences (Bateman, Dent,
Peters, Slovic, & Starmer, 2007; Simonsohn, 2007).
As happens in most new research streams though, there have been setbacks too. The
proliferation of research on seemingly contextual affective influences on behavior and the
limited integration of new findings into established information processing frameworks
have led to growing concerns among decision-making researchers. Such concerns have
prompted questions such as the one cited by Schwarz (2006, p. 20): “Whatever happened to
Fishbein and Ajzen’s theory of rational behavior and other such models? All we hear about
from psychologists these days is how funny little things make people feel one way or
another, influencing what they like and do.”
This research attempts to address such concern by assessing the compatibility of the
flourishing emotion research stream with cognitively dominated attitude-theory decision
making models. The manuscript begins with a theoretical discussion of whether Fishbein
and Ajzen’s (1975) expectancy-value model (EVM) of attitude, “the most widely applied
representation of attitude across many disciplines” (Bagozzi et al., 2002, p. 7), is sufficient
to capture the influence of emotion on decision making. Then, the EVM is augmented with
anticipatory and anticipated emotion constructs (Bagozzi, Baumgartner, Pieters, &
Zeelenberg, 2000), drawing on Larsen and Diener’s (1992) circumplex model of emotion.
With a controlled experiment involving 308 college students faced with actual purchase
decisions, the authors test whether the augmented EVM performs better than the traditional
from rational expectancy-value decision making, such as preference reversals when
focusing on anticipated emotions instead of focusing on product attributes (Caruso & Shafir,
2006; Shiv & Huber, 2000), the reliance on affect versus cognition under processing
constraints (Shiv & Fedorikhin, 1999), the impact of body feedback, meta-cognitive feelings
and moods on choice (Förster, 2004; Lee, 2004; Pham, 2004; Schwarz, 2004), or the
importance of contextual factors in determining affect and preferences (Bateman, Dent,
Peters, Slovic, & Starmer, 2007; Simonsohn, 2007).
As happens in most new research streams though, there have been setbacks too. The
proliferation of research on seemingly contextual affective influences on behavior and the
limited integration of new findings into established information processing frameworks
have led to growing concerns among decision-making researchers. Such concerns have
prompted questions such as the one cited by Schwarz (2006, p. 20): “Whatever happened to
Fishbein and Ajzen’s theory of rational behavior and other such models? All we hear about
from psychologists these days is how funny little things make people feel one way or
another, influencing what they like and do.”
This research attempts to address such concern by assessing the compatibility of the
flourishing emotion research stream with cognitively dominated attitude-theory decision
making models. The manuscript begins with a theoretical discussion of whether Fishbein
and Ajzen’s (1975) expectancy-value model (EVM) of attitude, “the most widely applied
representation of attitude across many disciplines” (Bagozzi et al., 2002, p. 7), is sufficient
to capture the influence of emotion on decision making. Then, the EVM is augmented with
anticipatory and anticipated emotion constructs (Bagozzi, Baumgartner, Pieters, &
Zeelenberg, 2000), drawing on Larsen and Diener’s (1992) circumplex model of emotion.
With a controlled experiment involving 308 college students faced with actual purchase
decisions, the authors test whether the augmented EVM performs better than the traditional
Page 100
96
· Conceptual breadth. Affect encompasses the entire spectrum of human moods and
emotions, whereas evaluative liking or disliking is widely considered just a tiny subset
of this broad spectrum (Allen, Machleit, & Kleine, 1992).
· Possibility versus probability. Whereas affect is sensitive to mere possibility and can
influence intentions, even when the probability of an outcome is nearly zero, attitudes
usually are conceptualized as a direct function of probability and thus are very weak
when the probability is close to zero (Loewenstein, Weber, Hsee, & Welch, 2001;
MacInnis & de Mello, 2005).
· Dynamic appraisals versus static predispositions. Attitudinal evaluations are defined as
a consumer’s learned, static predispositions that are activated when the consumer is
confronted with the stimulus object. Emotional reactions depend instead on context-
sensitive dynamic appraisals (Bagozzi et al., 2003).
· Temporal focus. Whereas attribute evaluations are traditionally measured as
preconsumption judgments, affective reactions include the consumer’s actual and
expected emotions before, during, and after consumption (Bagozzi, Dholakia, &
Basuroy, 2000; Richard et al., 1996).
3.2.3 The Role of Emotions for Attitude and Behavior
While emotions and evaluation can be theoretically (and empirically) distinguished,
as shown above, there is considerable debate about how emotions affect consumers’
decision making—by functioning as an antecedent of attitude, by influencing behavior in
addition to attitudes, or by both.
Regarding emotions as attitude antecedents, Cohen and colleagues (2008, p. 309)
perceive an emerging consensus that emotions are “one of several potential antecedents or
determinants of overall evaluation or attitude.” Early evidence for this position was
provided by Breckler and Wiggins (1989), who showed in the context of blood donations
· Conceptual breadth. Affect encompasses the entire spectrum of human moods and
emotions, whereas evaluative liking or disliking is widely considered just a tiny subset
of this broad spectrum (Allen, Machleit, & Kleine, 1992).
· Possibility versus probability. Whereas affect is sensitive to mere possibility and can
influence intentions, even when the probability of an outcome is nearly zero, attitudes
usually are conceptualized as a direct function of probability and thus are very weak
when the probability is close to zero (Loewenstein, Weber, Hsee, & Welch, 2001;
MacInnis & de Mello, 2005).
· Dynamic appraisals versus static predispositions. Attitudinal evaluations are defined as
a consumer’s learned, static predispositions that are activated when the consumer is
confronted with the stimulus object. Emotional reactions depend instead on context-
sensitive dynamic appraisals (Bagozzi et al., 2003).
· Temporal focus. Whereas attribute evaluations are traditionally measured as
preconsumption judgments, affective reactions include the consumer’s actual and
expected emotions before, during, and after consumption (Bagozzi, Dholakia, &
Basuroy, 2000; Richard et al., 1996).
3.2.3 The Role of Emotions for Attitude and Behavior
While emotions and evaluation can be theoretically (and empirically) distinguished,
as shown above, there is considerable debate about how emotions affect consumers’
decision making—by functioning as an antecedent of attitude, by influencing behavior in
addition to attitudes, or by both.
Regarding emotions as attitude antecedents, Cohen and colleagues (2008, p. 309)
perceive an emerging consensus that emotions are “one of several potential antecedents or
determinants of overall evaluation or attitude.” Early evidence for this position was
provided by Breckler and Wiggins (1989), who showed in the context of blood donations
Sign up today - FREE
Mendeley saves you time finding and organizing research. Learn more
- All your research in one place
- Add and import papers easily
- Access it anywhere, anytime
Start using Mendeley in seconds!
Readership Statistics
3 Readers on Mendeley
by Discipline
33% Psychology
33% Social Sciences
by Academic Status
33% Student (Bachelor)
33% Post Doc
33% Ph.D. Student
by Country
33% United Kingdom
33% Brazil
33% United States