Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

http://rme.sagepub.

com/
(English Edition)
Recherche et Applications en Marketing
http://rme.sagepub.com/content/22/3/43
The online version of this article can be found at:

DOI: 10.1177/205157070702200304
2007 22: 43 Recherche et Applications en Marketing (English Edition)
Fabrice Larceneux
Buzz and Recommendations on the Internet. What Impacts on Box-Office Success?

Published by:
http://www.sagepublications.com
On behalf of:

Association Franaise du Marketing


can be found at: Recherche et Applications en Marketing (English Edition) Additional services and information for

http://rme.sagepub.com/cgi/alerts Email Alerts:

http://rme.sagepub.com/subscriptions Subscriptions:
http://www.sagepub.com/journalsReprints.nav Reprints:

http://www.sagepub.com/journalsPermissions.nav Permissions:

http://rme.sagepub.com/content/22/3/43.refs.html Citations:

What is This?

- Sep 1, 2007 Version of Record >>


by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from by slaheddine dardouri on October 19, 2013 rme.sagepub.com Downloaded from
INTRODUCTION
Producing new films is an extremely risky ven-
ture. Out of ten major productions, six are considered
unprofitable (Vogel, 2001). This state of affairs can
be partially explained by the increasingly competi-
tive nature of film distribution. In France, for
example, the number of screens available per movie
fell steadily from 12.2 in 1990 to 9.7 in 2004. Thus,
the big issue in the movie industry is not only a suc-
cessful distribution during a new releases first week
(De Vany and Walls, 1999) a films ability to stay on
the circuit is also a crucial financial factor. On ave-
rage, 60% of a films takings are made during the
first week (cf. cbo-office.com). From an artistic pers-
pective, this concentration can create problems for
quality films, since they do not have long to make a
Recherche et Applications en Marketing, vol. 22, n 3/2007
Buzz and recommendations on the Internet.
What impacts on box-office success?
Fabrice Larceneux
Research Fellow
CNRS
Greghec HEC School of Management, Paris (UMR 2959)
The author would like to thank the anonymous referees, as well as Florence Jusot, for their highly pertinent and constructive comments,
which have helped to improve this paper significantly.
The author can be contacted at the following e-mail address: larceneux@hec.fr
ABSTRACT
Potential moviegoers may find information on the Internet on movie critics recommendations and web users comments on
movies they have seen. The more a recommendation is shared by a large number of individuals, the greater is the buzz. This
research explores to what extent this online buzz can be a predictor and/or modifier of box-office results, particularly the
week following the launch.
A field experiment on 534 movies shows that the buzz, and more specifically critics recommendations appearing on the
allocine.com website, is significantly correlated with box-office results after the first week. Various developments show that these
online recommendations may not only predict but also influence box-office results.
Key words: Buzz, Internet, recommendations, web users, critics, box office, word of mouth.
03Larceneux 7/02/08 14:06 Page 43
difference. Lady Chatterleys Lover, for example,
was taken out of circulation after just one week, des-
pite the fact that many cinema operators wanted to
keep it running because of its favorable reviews
(Frodon, 2006). Squeezed by financial imperatives
and unsure of how large an audience they could get,
word of mouth was not given enough time to deve-
lop, condemning this film to wither away within two
weeks. Conversely, movies such as Basic Instinct II
or Cabaret Paradis, despite enjoying a relatively
widespread distribution week in week out, actually
brought in very few people. In summary, one of the
main challenges facing movie professionals is how to
identify new audience predictors.
Recent technological developments have led to
the creation of online communities that go well
beyond participants personal relationships. People
use such groups to share the experiences they have
had with a product or service (Dellarocas, 2003).
Such electronic word of mouth spreads via sites like
epinions.com and directly impacts the movie indus-
try. In France, for example, there is a benchmark
website called allocine.com that posts magazine critic
and general viewer assessments of every movie sho-
wing at a particular time. These ratings, which range
from 0 to 4 stars, are tantamount to recommendations
that, when shared by a large enough number of
people, can generate a buzz that will influence whe-
ther or not a film will succeed.
The managerial aim of the present study is to esti-
mate the extent to which the two kinds of buzz
derived from allocine.com the buzz created by
movie critics and the buzz created by a films initial
audiences can help improve box-office forecasts,
particularly after a movies first week.
More specifically, theoretical studies on the
effects of electronic word of mouth have tended to
distinguish between the level of recommendation and
the number of people originally involved (Liu, 2006)
without delving more deeply into the need for a
variable that is capable of aggregating both
dimensions. Yet, the greater the satisfaction someone
gets from a product, the likelier he/she is
to talk about it (Anderson, 1998). The main theoretical
aim of the present paper, therefore, consists of
exploring the pertinence of the buzz concept,
defined here as the combination of a recommenda-
tions more or less favorable overtones with its
reach.
Note that online reviews are sources of information
that can be accessed at all times, without any financial
or cognitive costs. People find this particularly useful
(Wang, 2005). With more than four million hits a
month in France, allocine.com is the leader, all website
categories combined, in the target 20-25 age group,
which is the bracket whose members go most often
to the movies (cf. Mdiamtrie net rating, June
2006). Potentially, its recommendations can
influence the film choices that site visitors make.
Thus, our second theoretical aim will be to study,
from a causal perspective, the exact nature of the
influence of the buzz being generated by both critics
and web users, to ascertain the impact on a films
success. To satisfy these objectives, a field experi-
ment was carried out on 534 films, using information
found on allocine.com.
LITERATURE REVIEW
Word of mouth, buzz and third parties
A word-of-mouth effect occurs in interpersonal
communications between a sender and a receiver
whenever the latter is the recipient of non-commer-
cial information about a product, brand or service
(Bristor, 1990). Word of mouth among peers is often
considered a very efficient mechanism for exerting
influence advertising pressure only has an effect
because of the recommendation that a product be
assessed and adopted (Martilla, 1971; Murray, 1991;
Anderson and Salisbury, 2003). Thus, the more a
product receives favorable recommendations, the
likelier consumers are to buy it (Ardnt, 1967; Price
and Feick, 1984) and given that films constitute
experiential entertainment products, people talk
about them a great deal (Chaffee, 1982; Harrison-
Walker, 2001). Word of mouth reaches members of
the public with information about a film, influencing
the choices they make (Faber and OGuinn, 1984;
Neelamegham and Chintagunta, 1999). Nowadays,
potential viewers can obtain information about a film
online. They can also take part in forums to discuss
with their peers matters of mutual interest, thereby
Fabrice Larceneux 44
03Larceneux 7/02/08 14:06 Page 44
building a sort of electronic proximity (Mann and
Stewart, 2000). Analysts often refer to the success of
The Blair Witch Projet, which generated more than
75 million hits on film-related sites, as the example
of an electronic word-of-mouth effect (Benghozi and
Paris, 2001).
This new kind of word-of-mouse can be appre-
hended via a construct called the buzz (Stambouli
and Briones, 2002). Usually defined in relatively
general terms, the buzz refers to the self-generated
explosive demand (Dye, 2000) that occurs when initial
marketing efforts are amplified by third parties active
or passive influence (Thomas, 2004). Two kinds of
buzz are identifiable: a relatively commercial variant,
grounded in marketing strategies that seek to initiate,
amplify and maintain information exchanges by get-
ting opinion leaders to exert an active influence (Dye,
2000; Vernette and Flores, 2004); and a seemingly
less commercial kind of buzz, grounded in more pas-
sive influences and driven by the desire of some
people usually independently-minded individuals
who wield influence more than they lead opinion
(Vernette, 2007) to communicate to potential vie-
wers visiting a site their own opinions of a product,
service or film they have experienced.
Within the framework of this second perspective,
we can find online two independent types of buzz,
generated by two different kinds of third parties
(Rosen and Olshavsky, 1987; Senecal and Nantel,
2004). The first category comprises relatively tradi-
tional professional experts, i.e., film industry critics
(Debenedetti, 2006). The second comprises web
users, often the first persons to purchase a product
(Feick and Price, 1987), people who will offer their
opinions online all the more readily since they consider
themselves experts in a given product category
(Hamilton, 2001; Vernette and Flores, 2004). This
distinction can be found on a general movie website,
allocine.com (cf. Appendix A1), one whose particula-
rity is that it codifies buzz by publishing the average
number of stars attibuted to a film (Thomas, 2004),
thus facilitating and accelerating its diffusion.
The dimensions of the buzz and its operationalization
From a conceptual perspective, buzz can be
defined as a moment t captured in two dimensions
(Neelamegham and Chintagunta, 1999; Liu, 2006). It
has a quantitative dimension volume which repre-
sents the greater or lesser number of individuals revie-
wing and talking about a product. It also has a qualita-
tive dimension valence which represents the more
or less favorable average assessment made of this pro-
duct. The greater the number of individuals involved,
the more widespread the general awareness of a films
existence. This volume effect is the informative
dimension of the buzz. One way of mobilizing it is by
advertising a film during its release week (Zufryden,
2000). The films average review constitutes the
valence effect of the buzz. Whereas an unfavorable
opinion might lead to some denigration of the film, a
favorable opinion will increase the quality expected
from it. This is the persuasive dimension of the buzz.
These two dimensions help to define two kinds of
qualified buzz (cf. Figure 1).
Buzz and recommendations on the Internet. What impacts on box-office success? 45
Figure 1. Buzz dimensions (adapted from Liu, 2006)
Communications
budget

Volume
(Numbers of critics)
Valence
(Critics' opinion)

Knowledge
Attitude
Attitude
}
}
Box
office
Volume
Numbers of Internet users
Valence
Internet users' opinions
Knowledge
Critics' qualified
buzz
Web users'
qualified buzz
03Larceneux 7/02/08 14:06 Page 45
Two outlooks can be suggested to operationalize
this type of buzz. The first is to consider that the afo-
rementioned dimensions are additive. This is the
main approach pursued in studies on the influences
of recommendations in the film sector (Liu, 2006).
The second would be to envisage a single value
capable of measuring the whole of the buzz. After all,
the greater a films popularity, the more widely it will
be recommended (Anderson, 1998). The number of
persons judging a film can therefore be weighted by
the level of recommendation, in which case, our two
dimensions become multiplicative. When conceptua-
lized in this way, the buzz becomes a continuous
single variable, measured at a moment t, taking place
in the early stages of a films life cycle.
The buzz as a predictor or influence:
from correlation to causality
Many empirical studies have focused on identi-
fying factors capable of forecasting or influencing
the box office. Such factors include, for example,
film characteristics like the presence of stars (Ravid,
1999) or critic opinions (Reinstein and Snyder,
2005). Generally, the idea has been that some of the
correlations observed between critic opinions and the
box office can be attributed to causality, even if the
exact relationship is extremely difficult to estimate
(Eliashberg and Shugan, 1997; Basuroy, Chatterjee
and Ravid, 2003). The existence of this kind of causa-
lity derives from the reasonable link between indivi-
dual movie-goer behavior and aggregate economic
data (Eliashberg and Sawhney, 1994). Three perspec-
tives allow us to interpret possible links between
recommendations disseminated online and box-office
success.
A correlation, in the guise of a simple box-
office forecast. Recommendations, together with the
buzz associated with them, constitute indicators that
can be used to approximate a films perceived quality
and anticipate its box-office success (Hennig-Thurau,
Houston and Sridhar, 2006). The opinions of critics
and/or web users are mere reflections of the evalua-
tions made by all audiences, due to the objective
agreement that exists between people (Bourdieu,
1979) representing a particular segment of the general
public or a reference group (Eliashberg and Shugan,
1997). The idea here is that an absence of recommen-
dations, online or in the press, would have no effect on
box-office success.
A mirror effect involving a prediction of
causality. The information featured on allocine.com
echoes a buzz that has already occurred (Dellarocas,
Farag and Zhang, 2007) outside the website. It is a
measurement representing the significance of the
films presence in the press, as well as the more or
less favorable nature of the reviews that critics have
written about it. It also measures the films popula-
rity and the extent of audience word of mouth. Ulti-
mately, it is these magazines analyses and/or real
word of mouth that will influence potential audiences
not the overt indicators found on the site. In sum-
mary, the absence of recommendations on
allocine.com is not likely to modify box-office suc-
cess. On the other hand, the absence of critic film
reviews in magazines and newspapers is likely to
modify it.
A causality that is direct. The recommendations
disseminated via allocine.com represent a criterion
like the films intrinsic characteristics or marketing
pressure that directly influence the choices made
by the potential viewers (Wang, 2005) who visit the
site and can, in turn, influence their own reference
groups (Cakim, 2002). In this case, an absence of
recommendations on allocine.com is apt to modify
box-office results.
From a theoretical perspective, we can say that
this direct causality derives from the recommenda-
tions direct usefulness to potential audiences who,
during their decision-making processes, will have to
cope with informational asymmetries due to ques-
tions about the films quality (Akerlof, 1970; Nelson,
1970); the abundance of films on offer; and the trend
toward a shortening of films life cycles. This context
heightens the need to supply useful information to
ensure the quality of the consumption experience
(Spence, 1974; Rao, Qu and Ruekert, 1999), simpli-
fying the decision-making process and reducing the
perceived risk of a purchase (West and Broniarczyk,
1998). Critics and web users online recommenda-
tions constitute credible sources of information
(Duhan et al., 1997) that enjoy comparable levels of
credibility (Senecal and Nantel, 2004). They are the-
refore apt to generate two complementary types of
buzz (Holbrook, 1999; Dellarocas, Farag and Zhang,
2007). In addition, allocine.com, which receives a
phenomenal number of hits in France, is at least as
Fabrice Larceneux 46
03Larceneux 7/02/08 14:06 Page 46
visible as the countrys more widely distributed culture
magazine, Tlrama (cf. Appendix A1). The recom-
mendations displayed on allocine.com are useful,
credible and widely followed, hence potentially
influential.
HYPOTHESES
The buzz from critics and its potential influence
Film professionals broadly share the opinion that
critics recommendations affect audience decisions
(Blumenfeld and Frodon, 1999). A theoretical pers-
pective might give three explanations for such
influence. First, critics have been institutionalized as
experts by a press that portrays them as spokesper-
sons and attributes their influence to their specialization
in this product category (Feick and Price, 1987;
Debenedetti, 2006). Critics are also frequently tied to
specialist magazines, whose raison dtre and objective
are to provide consumers with impartial information
(Wyatt and Badger, 1990; Bera, 2003). Lastly, a prefe-
rence influence phenomenon exists, insofar as expert
critics can enhance consumers self-image and/or pro-
mote strong consumption by mobilizing a social dis-
tinction effect (Cameron, 1995; Holbrook, 1999).
The intensity of critics influence is not considered
homogeneous over the course of a films life cycle,
however. According to Eliashberg and Shugan
(1997), critic recommendations probably exert a
stronger influence in the short term, when newspa-
pers and magazines featuring their columns are in
circulation. From the second week onwards, movie
world critics are already focusing on the next batch
of releases. This vision should be relativized today
due to the fact that marketing pressures are at their
most powerful when a film is first released, and this
masks other short-term effects (Hennig-Thurau,
Houston and Sridhar, 2006). In addition, critic
reviews can be read online, especially once a film has
been out for a week.
During this first week, the greater the number of
newspapers and magazines that mention the film, the
louder the buzz. We can therefore postulate that the
buzz from critics will have a positive impact on the
box office during a films first week, due to the mirror
effect of the influence conveyed by the films pre-
sence in the press. In other words, this is a predictive
kind of buzz, a forerunner of the press articles future
influence.
Once a movie has been out one week, the press
talks about it less, yet critic recommendations remain
available on allocine.com. There are two possibilities
here: (1) The buzz from critics might be a good
representation of viewer tastes and numbers, and turn
into a good predictor of the post-first week box office
(Eliashberg and Shugan, 1997); and (2) A favorable
review, shared by many magazines identified on
allocine.com, will tend to influence potential viewers
visiting the site. Here, we would either have a predic-
tive buzz, or else an influential buzz directly impacting
the post-first week box-office results.
H1.1a: The buzz generated by critics and featured
on allocine.com has a positive effect on first-week
box-office success.
H1.1b: The buzz generated by critics and featured
on allocine.com has a positive effect on post-first
week box-office success.
Critics volume effect
Certain studies analyze the impact of critic opi-
nions without worrying about the number of articles
written on a film (Eliashberg and Shugan, 1997). Yet,
even if many critics are conveying their opinion
about a film, the fact remains that a greater number of
individuals will be aware of its existence. This simple
informative dimension causes a overall desire to see
the film (Bowman and Narayandas, 2001). During
the first week, the informative dimension should be
stronger than any evaluative dimension (Wyatt and
Badger, 1990; Ravid, 1999) since media coverage is
ultimately the only thing that counts (Hirsh, 1972).
Some studies have already demonstrated this number
of critics effect (Larceneux, 2001; Basuroy, Chatterjee
and Ravid, 2003; Gemser, Van Oostrum and Leen-
ders, 2007). In summary, the box-office impact of the
number of critics mentioned on allocine.com probably
constitutes a mirror effect of the films exposure in
the press.
It remains that during the weeks that follow, the
number of critics automatically has less of an impact
on the box office, since this informative dimension
Buzz and recommendations on the Internet. What impacts on box-office success? 47
03Larceneux 7/02/08 14:06 Page 47
tends to disappear. The extent of a newly released
films exposure in the press should not have any great
effect on audience sizes after the release week (Liu,
2006).
H1.2a: The number of critics featured on
allocine.com has a positive effect on first-week
box-office success.
H1.2b: The number of critics featured on
allocine.com has no effect on post-first week box-
office success.
The valence effect of critic opinions
In previous studies, the number of favorable
reviews seemed to be globally correlated to audience
sizes (Litman, 1983; Litman and Kohl, 1989; Litman
and Ahn, 1998; Prag and Casavant, 1994; Basuroy,
Chatterjee and Ravid, 2003). However, few studies
have highlighted the short-term effect of critics
valence. This can be explained by the fact that during a
films first week, commercial pressure is at its maxi-
mum. This masks any other impact (Lampel and
Shamsie, 2000; Zufryden, 1996).
After the first week, critics average assessment
of films can be considered from two perspectives: (1)
As an assessment that is relatively similar to the
publics (Holbrook and Addis, 2007), it constitutes
an accurate predictor of the films success
(Eliashberg and Shugan, 1997); and (2) As a recom-
mendation of which many people are aware, it can
directly influence the choices made by the indivi-
duals visiting a site.
H1.3a: The opinions of reviews featured on
allocine.com has no effect on first-week box-
office success.
H1.3b: The opinions of reviews featured on
allocine.com has a positive effect on post-first
week box-office success.
The potential influence of the buzz from web users
The influence of the buzz from web users has at
least two explanations. First, people who go to see a
film during its first week will tend to transmit useful
information about their consumption experience
(West and Broniarczyk, 1998; Feick and Price, 1987;
Park, Mothersbaugh and Feick, 1994) especially
since attribution theory (Kelley, 1973) tells us that
consumers perceive the recommendations they
receive as being disinterested (Bickart and Schindler,
2001). In addition, people who check websites are
often characterized by a homophilia principle, i.e., by a
desire to resemble their peers in terms of beliefs and
values (Alpert and Anderson, 1973; Solomon, Tissier
Desbordes and Helbrunn, 2005, p. 359). Thus, they
are more likely to see a film recommended by
someone they consider similar to themselves (Brown
and Reingen, 1987).
However, by definition, this kind of recommenda-
tion takes time to spread, since the web user must
first see the film and then post his/her recommenda-
tions on the site. First-audiences review a film on
allocine.com, enabling a measurement of buzz levels at
the end of the first week. During this period, the box
office is largely influenced by the marketing pressure
that studios deploy (Zufryden, 1996), and visitors to
allocine.com will be subjected to the same influence.
The buzz from web users would therefore logically
be correlated to the box office.
After the first week, two perspectives are pos-
sible: (1) The buzz from web users, as measured
during the first week, will predict the box office
accurately; and/or (2) The more favorable the recom-
mendations from web users, and the greater the num-
ber of individuals contributing to this opinion, the
likelier it is that the buzz will be convincing for
potential allocine.com visitors, hence the greater the
effect on the box office. What we have here is a pre-
dictive type of buzz, and/or a buzz that exerts a direct
kind of influence.
H2.1a: The buzz generated by first-week visitors
to allocine.com correlates positively to first-week
box-office success.
H2.1b: The buzz generated by first-week visitors
to allocine.com has a positive effect on post-first
week box-office success.
Web users volume effect
The number of web users reviewing a film on
allocine.com during its first week derives automati-
cally from producers marketing pressure. Studios
Fabrice Larceneux 48
03Larceneux 7/02/08 14:06 Page 48
will turn the launch of certain films into a gala event,
to be publicised, amongst others avenues, by web
users (Zufryden, 1996; Lampel and Shamsie, 2000;
Duan, Gu and Whinston, 2005).
The number of web users measured during the
first week on allocine.com is supposed to correlate to
the post-first week box office insofar as it predicts
the reach of the buzz (Anderson and Salisbury, 2003;
Liu, 2006). It remains that this indicator does not, at
first glance, constitute a criterion of choice per se.
H2.2a: The number of first-week visitors to
allocine.com correlates positively to first-week
box-office success.
H2.2b: The number of first-week visitors to
allocine.com correlate positively to post-first
week box-office success.
The valence effect of web users opinion
During the release week, actual and electronic
word of mouth normally undergoes a definition pro-
cess. Here, web users opinion of allocine.com at the
time of a films release is not supposed to correlate to
the box office (Liu, 2006). This is because studios
communications actions have the capacity to
influence the box office in the short run, for good but
also for bad films (Basuroy, Desai and Talukdar,
2006).
However, despite such marketing actions, it is
difficult for studios to artificially ensure a films sur-
vival for more than one week. The memory of a parti-
cular film will last longer if people have expressed
favorable opinions about it (Herr, Kardes and Kim,
1991). Now, web user opinions are meant to be
mainly defined during a films first week and remain
stable throughout its life cycle (Liu, 2006). Once
again, there are two possibilities at this level, since
first-week web user opinions are considered as (1)
good predictors of the audiences general opinion
(Holbrook, 1999), hence of the box office; and/or as
(2) decisional criteria for potential viewers visiting
allocine.com (Dean and Biswas, 2001; Wang, 2005).
H2.3a: First-week web user opinions featured on
allocine.com do not correlate to first-week box-
office success.
H2.3b: Web user opinions featured on
allocine.com have a positive effect on post-first
week box-office success.
METHODOLOGY
Experimental laboratory methodology has the
advantage of optimizing tests of critics influence
(Wyatt and Badger, 1990). However, this does not
help us to work on actual box-office takings, a limita-
tion that could seriously weaken the applicability of
any findings. This is the reason why, despite certain
constraints, real field experiments were conducted.
A number of models have been designed to use
traditional variables (like number of stars or number of
copies in circulation) to produce forecasts before a
films release (Litman, 1983; Zufryden, 1996). Other
models that have focused more on predicting post-
release box-office success (Sawhey and Eliashberg,
1996; Neelamengham and Chintagunta, 1999) have
generally been more precise because they included a
larger number of explanatory variables. The present
study will apply this latter approach.
Dependent variables
The split between the first week and subsequent
weeks has been maintained because it is considered
relevant (Elberse and Eliashberg, 2003; Gemser, Van
Oostrum and Leenders, 2007). Different measure-
ments of audience sizes were constructed to increase
the findings internal validity (Campbell, 1969). The
box office was used to represent the volume of first-
week sales per theater average (total box-office
receipts divided by the number of theaters)
1
is a
strong indicator of audience density (Ackman, 2001).
For subsequent weeks, the cumulative box office,
plus the growth rate between the first and the final
week of projection, were used to approximate
audience intensity.
Buzz and recommendations on the Internet. What impacts on box-office success? 49
1. Despite relatively major variations in theater capacities, the lack of
available data means that this information could not be used.
03Larceneux 7/02/08 14:06 Page 49
The data were compiled using information from
the website, cbo-boxoffice.com. The nature of the
variables distribution was such that they had to be
converted to a logarithmic form so that a linear
regression could be carried out (Montgomery, Vining
and Peck, 2001). Normality analyses of the four
dependent variables enabled the usage of parametric
procedures.
Independent variables
A database was built using data found on the
reference website, allocine.com. The opinions
valence was operationalized on a scale of 0 to 4 stars.
This scale corresponded to the one shown to website
visitors. The volume of critics was calculated based
on the number of magazines listed on the site, with
the volume of web users corresponding to the number
of first-week web user reviews posted there. Within
this framework, the two types of buzz were operatio-
nalized by multiplying the volume by the average
grade (1 for 0 star; 5 for 4 stars).
Confusion effects and control variables
A number of control variables were taken into
account as potential antecedents to the buzz and/or as
ways of explaining audience sizes. The number of
copies in circulation during a films first week was
taken to reflect the intensity of its distribution, hence
its accessibility. This variable should have a strong
effect on the box office (Swami, Eliashberg and
Weinberg, 1999; Basuroy, Chatterjee and Ravid,
2003) and also enable an accurate approximation of
the communications budget. The reason is because
advertising spending and certain film attributes ele-
ments that are alleged to have a direct impact on the
box office will, in actual fact, probably have a relati-
vely indirect effect on the box office via screen alloca-
tion decisions (Elberse and Eliashberg, 2003). In line
with Gemser, Van Oostrum and Leenders (2007), the
number of copies in circulation at the time of a films
release was selected, following a logarithmic trans-
formation, to avoid any extreme data effects. Accor-
ding to some estimates, communications spending in
the United States can amount to almost 50% of a
films production budget (Vogel, 2001). To control
the most significant effects, production budget sizes
were coded using available information. Such data is
considered strategic and therefore communicated on
different websites, at best parcimoniously. This
variable, which is something that each studio
assesses in its own way, also has the shortcoming of
not always integrating the same elements in different
films (for example, the producers remuneration may
or may not be included). It would, nevertheless, be
introduced in the analysis whenever available.
The fame of directors and actors, and whether or
not they are considered as stars, can influence a buzz
and have a positive effect on the box office (Wallace,
Seigerman and Holbrook, 1993; Albert, 1998). As
detailed by Bagella and Becchetti (1999), two eva-
luators coded the degree of stardom (unknown stars
versus medium or big stars). A binary variable was
then created. Otherwise, the release date can also
affect a films performance (Krider and Weinberg,
1998; Litman and Ahn, 1998), since more people go to
the movies during school holidays. A binary variable
(vacation vs. school time) was therefore created.
American studies may have paid little attention to
the nationality of a film, but we can estimate that
this information affects potential audience behavior
in France. Hence, the creation of another binary
variable French vs. American (or other kinds of)
films.
The competitive intensity, i.e., the number of
films released at a specific time, also had to be consi-
dered (Liu, 2006). This variable represents, for each
new film, the number of rival films coming out the
same week.
The film genre can affect films success (Elberse
and Eliashberg, 2003). This variable is coded accor-
ding to the French Centre National de la Cinmato-
graphie classification that allocine.com reproduces.
For operational reasons, certain genres were merged
and five binary variables (action, drama, comedy,
dramatic comedy, and miscellaneous) were created.
Type of distributor was not included as a
variable due to the low variance in France, where an
oligopoly of ten comparably sized, large film distribu-
tors accounts for 90% of all takings, with 70 others
sharing the remaining 10% (Leroy, 2005). Several
studies have shown that this variable is not very
significant (Liu 2006; Elberse and Eliashberg 2003;
Litman and Ahn, 1998). Similarly, including a para-
meter like the different prizes (Oscars, Csars, Prix
Fabrice Larceneux 50
03Larceneux 7/02/08 14:06 Page 50
Louis-Delluc, etc.) that are often awarded after a
films release would not have had a significant
impact (Reinstein and Snyder, 2005).
Description of the database
The film database used contained the 534 leading
films released in France between 4 January 2005 and
14 February 2007 (12 films, tantamount to extreme
data, were eliminated from the analysis). Table 1 pre-
sents the variables used.
Previous regression analysis showed that films
genre, competitive intensity, release date and produc-
tion budget did not affect the box office. These fin-
dings were globally in line with results from recent
studies (Gemser, Van Oostrum and Leenders, 2007;
Liu, 2006). Their explanation lies in the fact that
staggering weekly releases tends to diminish competi-
tion among major films, with holiday periods likely
to have the greatest impact on attendances for chil-
drens films. In addition, as we have seen, the expla-
natory power of the number of copies in circulation is
greater than production budgets. Hence, our decision
to exclude these variables from the analysis. On the
other hand, the fact that a film was French (n = 190)
or not (n = 344), whether a big star played in it
(n = 277) or not (n = 257) and the number of
copies in circulation during the first week were signi-
ficant variables for different audience size indicators.
We therefore included these control variables in our
analysis.
FINDINGS
Two regression models were used. The first
assessed the coefficients of the values of each buzz,
with the second determining the coefficients of the
volume and valence dimensions.
Model 1: Buzz from critics and buzz from web users
Significant colinearities existed between the inde-
pendent variables (cf. Appendix A2), in particular
between the buzz from web users and the number of
copies in circulation (r = 0.57). To control potential
induced effects, we implemented a step-by-step pro-
cedure and presented partial correlations in addition to
standardized coefficients (cf. Table 2).
As expected, the buzz from critics and from web
users was significantly correlated to first-week
Buzz and recommendations on the Internet. What impacts on box-office success? 51
Table 1. Summary of sample variables
N Minimum Maximum Average Std. deviation
Total box office
(Number of tickets sold)
534 18,063 7,732,071 624,008 1,008,601.3
First-week box office 534 18,063 3,303,005 248,429.5 364,346.5
Total number of weeks 534 1 18 3.9 2.8
Number of copies in circulation 534 24 956 277.8 196.3
Number of critics 534 2 31 18.5 4.9
Number of web users 534 4 2,136 229,3 278.7
Competitive intensity
(Number of films 534 1 8 5.5 1.3
released at same time)
Production budget
(in million Euros)
270 1 230 27 37
03Larceneux 7/02/08 14:06 Page 51
audience sizes. Thus, H1.1a and H2.1a were validated.
The measurement of the buzz from critics can there-
fore be considered as a meaningful predictor of first-
week box-office results. Moreover, the correlation
between the buzz from web users and the box office
illustrates a films word-of-mouth dynamic. It
remains that the number of copies in circulation
constitutes the best explanation for a films success.
After the first week, the two types of buzz again
had significant effects, on both the cumulative box
office and the films growth rate. This meant that
hypotheses H1.1b and H2.1b were validated. Dimen-
sions analysis would allow us to explain these fin-
dings.
Model 2: Dimensions of the buzz from critics and
web users
Dellarocas and Narayan (2006) have mentioned
the possible existence of threshold effects between
different evaluation levels. To dispel the hypothesis
that a star-based grading scale is linear in nature, a
distributions analysis was implemented, leading to
the creation of three critic grade classes and two web
user grade classes (cf. Appendix A3).
In total, critics assessments could be structured
into three levels: bad/not very good (0 and 1 star);
average (2 stars); and good/very good (3 and 4 stars).
Web user reviews more or less broke down into two
Fabrice Larceneux 52
Table 2. Model 1: Determining coefficients for the buzz from critics and from web users
First week After the first week
Box-office Per Theater Cumulative Growth rate
Average box office after week 1
Ln (BO1) Ln (BO1/copies) Ln (BOT-BO1) Ln ((BOT-BO1)/BO1)
Adjusted R
2
0.847 0.549 0.721 0.259
N 534 534 438 438
B Stand. Partial B Stand. Partial. B Stand. Partial. B Stand. Partial.
(Sig.) corr. (Sig.) corr. (Sig.) corr. (Sig.) corr.
Constant 5.772 5.848 4.286 1.932
(0.000) (0.000) (0.000) (0.000)
Nationality
0.043 0.082 0.071
0.104 0.106 0.131 Ex. Ex
French vs. others
(0.016) (0.015) (0.006)
Presence of stars
0.045 0.090
0.102 0.113 Ex. Ex. Ex. Ex
Stars vs no stars
(0.019) (0.009)
Distribution
0.713 0.655 0.323
0.829 0.727 0.305
Ln (Number of copies)
(0.000) (0.000) (0.000)
0.087 0.159 0.167 0.279
Buzz from critics 0.196 0.198 0.274 0.280
(0.000) (0.000) (0.000) (0.000)
0.297 0.574 0.306 0.172
Buzz from web users 0.515 0.574 0.425 0.162
(0.000) (0.000) (0.000) (0.000)
Ex: variables excluded via a step-by-step procedure.
03Larceneux 7/02/08 14:06 Page 52
classes: not good (between 0 and 2 stars); and good (3
stars and more). These results match those found by
Dellarocas and Narayan (2006) and Dellarocas,
Farag and Zhang (2007). Binary variables were created
and a critics neutral opinion variable, the bench-
mark category, was excluded from the regression.
These variables were then introduced into the model 2
(cf. Table 3). As above, colinearities were managed
via a step-by-step procedure.
Volume and valence of critic recommendations
Correlations analysis (cf. Appendix A4) showed
both that the number of critics is not correlated to the
number of theaters where a movie is shown
(r = 0.04) and also that critic recommendations are
not related to the presence of stars in a film or to its
nationality. They are even negatively correlated to the
number of copies in circulation (r = 0.10) . The
dimensions of the buzz from critics are therefore
relatively independent of commercial pressures, out-
side of what has sometimes been denounced as pro-
motional neo-criticism (Lovell, 1997; Creton, 2000).
During the first week, as expected, the number of
critics, reflecting the importance of media coverage,
had a significant impact on audience sizes. Thus,
H1.2a was validated. On the other hand, whereas a
favorable recommendation from a critic (3 stars) ulti-
mately had a positive impact on audience sizes
during the release week, an unfavorable one (1 star)
seemed not to have any effect. H1.3a, predicting the
absence of an effect, was therefore rejected. A slight
effect could sometimes be observed in the short run
(Basuroy, Chatterjee and Ravid, 2003), resulting
from a negativity bias that may have led to an over-
estimation of the effects of unfavorable recommenda-
tions, hence to potential losses (Ito et al., 1998;
Kahneman and Tversky, 1979). More than this,
however, what was discovered was a relatively positi-
vist effect (Skowronski and Carlston, 1989) in the
sense that critics were expected to act as harsh judges
due to their expert status. Hence, the perception that a
3-star film was an exceptional event and something
that people could not afford to miss. In addition, critics
tended to do more reviews on films they liked. This
could be seen in the fact that there were a greater
number of critics reviewing films that were given a
high grade (m = 21.8) than films with an average
grade (m = 17.9) or a low one (m = 14.8;
F(2.531) = 101.5; p < 0.000). This effect was most
probably linked to the fact that the studios increasingly
choose which critics they will invite to their press
screenings, so as to reduce the likelihood of bad
reviews. Ultimately, the main impact of the buzz
from critics can be explained, in the short term, by an
informative volume effect surrounding films that
people like. It seems reasonable to assert that this
impact should be interpreted in causal terms, either
directly via the website, or more probably via a mirror
effect. Aside from marketing pressure, audiences
choices are influenced by a films visibility in news-
papers and magazines and by the favorable recom-
mendations it receives.
After the first week, the number of critics no longer
correlated to changes in audience sizes. The same
applied to the question of whether or not a film featu-
red a big star. This means that H1.2b was validated. On
the other hand, critic recommendations had a very
significant effect: a favorable one for movies that
people liked (3-star films); and an unfavorable one
otherwise (1-star films). Thus, H1.3b was validated. In
line with Basuroy, Chatterjee and Ravid (2003) and
Liu (2006), these findings can be taken to mean that
the significant effect of the buzz from critics follo-
wing a films release week stems from an evaluative
effect of its quality. At this level, observed correla-
tions can be interpreted as simple predictive tools (all
critics, representing the tastes of all viewers, would
identify good and bad films, hence potential suc-
cesses). However, as we have seen, with the generali-
zation of Internet use, we cannot exclude, at least to
some extent, the fact that these recommendations
influenced the decisions of those movie-goers who
went online, thus, ultimately, box-office results.
Volume and valence of recommendations made by
web users
Correlation analysis (cf. Appendix A4) has shown
that the number of web users posting a recommenda-
tion during the first week correlated to the number of
copies in circulation (r = 0.51). Among other
Buzz and recommendations on the Internet. What impacts on box-office success? 53
03Larceneux 7/02/08 14:06 Page 53
Fabrice Larceneux 54
Tableau 3. Model 2: Determining coefficients for volume and valence dimensions
1 ref.: Neutral opinion, 2 stars.
Ex: variables excluded via a step-by-step procedure.
First week After the first week
Box office Per Theater Cumulative Growth rate
Average box office after week 1
Ln (BO1) Ln (BO1/copies) Ln (BOT-BO1) Ln ((BOT-BO1)/BO1)
Adjusted R
2
0.855 0.577 0.727 0.296
N 534 534 438 438
B Stand. Partial. B Stand. Partial. B Stand. Partial. B Stand. Partial.
(Sig.) corr. (Sig.) corr. (Sig.) corr. (Sig.) corr.
Constant 5.869 5.753 4.225 2.211
(0.000) (0.000) (0.000) (0.000)
Nationality
0.043 0.080 0.082
0.108 0.106 0.153 Ex. Ex
French vs. others
(0.013) (0.014) (0.000)
Presence of stars
0.040 0.068
0.92 0.86 Ex. Ex. Ex. Ex
Stars vs. no stars
(0.034) (0.047)
Diffusion
0.685 0.653 0.373
0.817 0.709 0.338
Ln (Number of copies)
(0.000) (0.000) (0.000)
Number of critics
0.057 0.111
0.128 0.130 Ex. Ex. Ex. Ex
Nb posted on allocine.com
(0.003) (0.003)
Critics unfavorable
0.062 0.131
Ex. Ex. Ex. Ex. 0.107 0.142
opinion 1 star
(0.026) (0.000)
Critics unfavorable
0.041 0.086 0.105 0.197
0.090 0.103 0.168 0.195
opinion 3 stars
(0.038) (0.018) (0.000) (0.000)
Number of web users after
0.324 0.598 0.296 0.093
week 1 0.549 0.601 0.507 0.087
Nb posted on allocine.com
(0.000) (0.000) (0.000) (0.070)
Web users favorable 0.103 0.189
opinions (1
st
week) Ex. Ex. Ex. Ex. 0.178 0.204
4 and 3 stars vs. others (0.000) (0.000)
03Larceneux 7/02/08 14:06 Page 54
factors, this reflected easier access to a film. In addi-
tion, web user opinions were independent of how
widely a film was distributed (r = 0.01), i.e., they
were independent of marketing pressures. Further-
more, web users tended to review films they liked
(m = 295) more than those they did not like
(m = 157, t (532) = 6.0 p < 0.000). It therefore
appears that the buzz generated by web users spreads
via favorable recommendations when a major film is
involved. Lastly, complementary analysis showed
that the number of first-week web users, was strongly
correlated to the total number of web users
(r = 0.96), is a good reflection of the degree of the
buzz throughout a films life cycle.
The number of first-week web users was correla-
ted, as expected, with first-week audience sizes. This
validated H2.2a. Like other movie-goers, web users
were influenced by marketing pressure and media
coverage. An endogeneity problem arose here due to
the possible existence of reverse causality going to
see a film creates a desire to publish an online review
about it, especially if the film is a success. Web user
recommendations, on the other hand, correlated neither
to films success nor to how widely they would be
distributed. Word of mouth was given no time to take
root. This means that H2.3a was validated. The corre-
lation of the first-week buzz from web users may the-
refore be due to its volume dimension, itself a conse-
quence of marketing pressure.
Subsequent to this release week, on the other
hand, the buzz from web users would spread, making
the number of first-week web users a meaningful
predictor for post-first week audience sizes. This pro-
bably translated into a mirror effect, predicting the
effects of actual word-of-mouth activities. In short,
H2.2b was validated. At the same time, contrary to
Lius findings (2006), first-week web user recom-
mendations posted on allocine.com offered a strong
explanation for post-first week audience sizes, i.e.,
H2.3b was also validated. This result can be explained
either by a simple predictive effect (web users have
the same tastes as movie enthusiasts) or else by a
direct causal effect, which is that movie-goers visit
allocine.com and integrate the recommendations
found there as relevant decisional criteria, both for
themselves and for their reference groups. The post-
first week impact of the buzz from web users can be
explained by a dual effect that is both informative
and persuasive.
Dimensions, prediction and causality
The buzz concept improved the different
models explanatory capacity. It remains that the
improved explanation of the post-first week, cumula-
tive box office, obtained through the distinct incor-
proation of different dimensions, turned out to be
slightly higher than the improvement achieved by the
global value that was constructed (dif. r
2
= 0.011;
dif. F(5.427) = 3.5; p = 0.005). Since extreme
reviews are likely to generate more interpersonal
communications than average reviews (Anderson,
1998), a second analysis was conducted on a diffe-
rent operationalization of the buzz measurement.
Weighted by values that would increase as reviews
varied from the average score ( 3 for 0 star; 2 for 1
star; 1 for 2 stars; 2 for 3 stars; and 3 for 4 stars), this
new measurement was strongly correlated to the first
measurement (r = 0.95 for buzz from web users and
r = 0.94 for buzz from critics), with box-office
results remaining similar. The distinct incorporation
of both volume and valence thus turned out to be
more pertinent than the combined value of the two
dimensions.
The dimensions of the buzz improved the quality
of the post-first week box-office forecasts, not only
for traditional variables, but also in terms of the first-
week box office (cf. Table 4). The regression coeffi-
cient of model D, integrating the buzz dimensions
and the first-week box office, turned out to be signifi-
cantly higher than model C, which lacked the buzz
dimensions (dif F(5.428) = 14.9; p < 0.000). This
improved the prediction of the cumulative post-first
week box office in both 2005 (dif F(5.191) = 10.56 ;
p < 0.000) and 2006 (dif F(5.227) = 8.29 ;
p < 0.000)). There were similar outcomes for the
growth rate. In other words, the suggested models
were useful for predicting audience sizes.
Ultimately, out in the field, determining causality
can be a very complex affair (Basuroy, Chatterjee
and Ravid, 2003). Nevertheless, it is possible to iden-
tify causal relations based on associations or series of
events, insofar as knowledge of the past helps to pre-
dict the present. Working from a statistical perspec-
tive, and given the control of potential confusion fac-
tors and the existence of a theoretical mechanism of
influence, Granger (1969) suggested that a variable
X
t
influences a variable Y
t
if the past of X
t
can help to
predict Y
t
given the past of Y
t
. This is tantamount to
Buzz and recommendations on the Internet. What impacts on box-office success? 55
03Larceneux 7/02/08 14:06 Page 55
establishing the model Y
t
= f (Y
t 1
,X
t 1
). If X
t 1
turns out to be significantly different from 0 in the
regression, we will have demonstrated causality in
Grangers sense of this term. The present study has
shown that incorporating the first-week box office
(Y
t 1
) into the post-first week box-office regression
(Y
t
) turns the different confusion factors into some-
thing insignificant. The quality of the model impro-
ved (dif r
2
= 0.14; dif F(1.428) = 493.8; p < 0.000)
and the impact of the first-week box office became
significant (B = 0.96; p < 0.000). Here, only critic
recommendations, whether favorable (B = 0.09;
p < 0.000) or unfavorable (B = 0.06; p = 0.003),
plus web user recommendations (B = 0.09;
p < 0.000), remained significant variables. There
exists therefore a Granger type of causality between
the first-week recommendations disseminated on
allocine.com and post-first week audience sizes.
CONCLUSION AND DISCUSSION
Following on from Gemser, Van Oostrum and
Leenders (2007) in Denmark, or Bagella and
Becchetti (1999) in Italy, the present research project
suggested a box-office prediction model, tested using
non-American data. Due to an informative dimen-
sion, first-week buzz measurements correlated to
audience sizes. However, it is only really after the
first week that our proposed model kicks in. The
buzz from both critics and allocine.com visitors at
the end of the first week ultimately correlated with
post-first week audience sizes as well. More specifi-
cally, it turned out that recommendations from critics
and from web users offered a significant explanation
for films success. Traditional outlooks have suggested
that such opinions constitute a correct representation of
audiences tastes. The studys theoretical and statistical
analyses support the proposition according to which
these recommendations can influence potential
audience decisions. This influence increases in line
with an increase in the number of persons visiting the
reference sites.
Fabrice Larceneux 56
Tableau 4. Improving post-first week box-office forecasts
Cumulative post-first week box office
Ln (BOT-BO1)
Total database
1
2005 2006
N = 534 N = 239 N = 265
Standardised Standardised Standardised
Adjusted R
2
estimation Adjusted R
2
estimation Adjusted R
2
estimation
Models error error error
A Confusion factors
2
0.591 0.892 0.608 0.883 0.565 0.902
B Confusion factors
+ Buzz dimensions
0.727 0.730 0.755 0.699 0.699 0.750
C Confusion factors
+ First-week box office
0.852 0.536 0.877 0.595 0.828 0.568
D Confusion factors
+ First-week box office 0.873 0.598 0.901 0.544 0.853 0.524
+ Buzz dimensions
1. The total database includes more than 30 films released in early 2007.
2. Comprising traditional variables that have been integrated into the models: presence of a star; films nationality; and number of copies in cir-
culation.
03Larceneux 7/02/08 14:06 Page 56
Contributions, limitations and research paths
The concept of buzz turns out to have a greater
explanatory power when its dimensions are operatio-
nalized in an additive as opposed to a multiplicative
manner. It therefore appears appropriate that this
buzz concept be studied through the specific incorpo-
ration of two of its dimensions: volume and valence.
This result can be explained by the fact that the infor-
mative dimension was measured through the number
of web users posting a review on allocine.com. It
would undoubtedly be a better idea to incorporate the
number of hits on the website page featuring a film
during its first week. There is also the possibility that
the extent of the agreement between web users
constitutes another explanatory element for buzz. To
improve the quality of the information received, allo-
cine.com could propose a broader scale enabling the
integration not only of more differentiated recom-
mendation averages but also greater variance in the
recommendations.
In addition, explaining theoretical influence pro-
cesses, on the one hand, and significant correlations
between dependent and independent variables defi-
ned according to their time horizons, on the other,
helps with the maintenance of a causal perspective.
Construed in Grangers sense of the term, some cau-
sality does exist between the recommendations posted
on allocine.com and post-first week box-office
receipts. Nevertheless, caution remains advisable at
this level, due to the limitations created by problems of
endogeneity (the correlation between web users
first-week recommendation and final recommenda-
tions was r = 0.95). Along these lines, critic recom-
mendations have a more verifiable influence than
web user recommendations. This could be demons-
trated more precisely via a laboratory experiment
estimating the respective impact of critic and web
user recommendations on allocine.com visitors
behaviors, and highlighting any interaction effects, as
well as the antecedents that induce individuals to
read or not to read recommendations (DAstout and
Colbert 2002).
From a managerial perspective, incorporating the
volume and valence dimensions found on
allocine.com can be useful for predicting audience
sizes. Such a prediction might be improved by intro-
ducing the films presence on the radio and on TV
into the models. Lastly, these variables should be re-
tested on upcoming films, and more detailed analyses
should be conducted on the different time horizons
involved, week in week out.
The findings offer producers and operators a
chance to improve their screen allocations, as well as
arguments they will be able to use when renegotia-
ting post-first week distribution contracts. The
recommendations posted on allocine.com also enable
professionals to develop new film marketing strate-
gies that are more likely to match public expecta-
tions. For example, this could involve launching
communication campaigns for films that received
favorable first-week reviews. With the hit The Life of
Others, as with many other films, audiences often
have the same tastes as critic and movie buffs
(Holbrook and Addis, 2007).
One of the aims of this predictive model is to sup-
ply forecasts as early as possible, especially before
the end of the first week. Whereas critic recommenda-
tions remain stable over time, web user recommenda-
tions are shaped during the early days of a films life
cycle. The correlation between 3-day web user
reviews and final reviews, for the sub-sample of 80
films for which sufficient data exist, has been evalua-
ted as r = 0.66. In other words, this measurement
remains far from being able to provide an acceptable
representation of web user opinions. However, as
allocine.coms audience develops, and given the total
number of persons who go online, future studies
should be able to estimate whether it will be possible
to come up with an earlier and better quality forecast.
Lastly, another research path could be to explore
the heterogeneous nature of third parties. Different
types of critics are treated similarly here, as if they
constituted a homogeneous population. Yet,
allocine.com is able to identify and review as many
as 31 different critic types. Moreover, different critical
typologies exist (Chang, 1975), with varying effects
(Debenedetti and Larceneux, 2000). Furthermore,
web users can be characterised by different levels of
expertise (as is the case with reviews made by visi-
tors to the financial website, boursorama.com). Corre-
lations with the box office may be more or less pro-
nounced depending on the types of critics or web
users involved. Along these same lines, it should be
possible to assess the impact of embedding systems
in which recommendations center on the profiles of
the experts and/or web users who are making them.
This would enable people visiting a site to gear their
Buzz and recommendations on the Internet. What impacts on box-office success? 57
03Larceneux 7/02/08 14:06 Page 57
searches toward profiles that are similar to their own
personal characteristics and purchasing behavior
(Senecal and Nantel, 2004). In short, the new chal-
lenge for online marketing appears to be personali-
zed recommendations.
REFERENCES
Ackman D. (2001), Movie theaters of the absurd, Forbes,
March, 2.
Akerlof G.A. (1970), The market for lemons: qualitative
uncertainty and the market mechanism, Quarterly
Journal of Economics, 84, 3, 488-500.
Albert S. (1998), Movie stars and the distribution of
financially successful films in the motion picture industry,
Journal of Cultural Economics, 22, 4, 249-270.
Alpert M. and Anderson T. (1973), Optimal heterophily
and communication effectiveness: some empirical
findings, Journal of Communication, 23, 3, 328-343.
Anderson E. (1998), Customer satisfaction and word of
mouth, Journal of Service Research, 1, 1, 5-17.
Anderson E. and Salisbury L. (2003), The formation of
market-level expectations and its covariates, Journal
of Consumer Research, 30, 3, 115-24.
Arndt J. (1967), Role of product related conversations in
the diffusion of a new product, Journal of Marketing
Research, 4, 3, 291-295.
Bagella M. and Becchetti L. (1999), The determinants of
motion pictures box-office performance: evidence
from movies produced in Italy, Journal of Cultural
Economics, 23, 4, 237-256.
Basuroy S., Chatterjee S. and Ravid S.A. (2003), How
critical are critical reviews? The box-office effects of
film critics, star power, and budgets, Journal of
Marketing, 67, 4, 103-117.
Basuroy S., Desai K. and Talukdar D. (2006), An empirical
investigation of signaling in the motion picture industry,
Journal of Marketing Research, 43, 2, 287-295.
Bayus B. (1985), Word of mouth: the indirect effects of
marketing efforts, Journal of Advertising Research, 25,
3, 31-39.
Benghozi P.J. and Paris T. (2001), Cinma et Internet, un
tat des lieux, Conference Proceedings Aimac,
Brisbane, 5, 180-189.
Bera M. (2003), Critique dart et/ou promotion culturelle,
Rseaux, 117, 155-187.
Bickart B. and Schindler R. (2001), Internet forums as
influential sources of consumer information, Journal of
Interactive Marketing, 15, 3, 31-40.
Blumenfeld S. et Frodon J.-M. (1999), Les cinastes franais
se runissent pour discuter du rle de la critique, Le
Monde, November 4, 35.
Bourdieu P. (1979), La distinction, Paris, Les ditions de
Minuit.
Bowman D. and Narayandas D. (2001), Managing
customer-initiated contact with manufactures: the
impact on share of category requirements and word-of-
mouth behavior, Journal of Marketing Research, 38, 3,
281-97.
Bristor J. (1990), Enhanced explanation of word-of-mouth
communications: the power of relationships, Ed. E.C.
Hirschman, Research in Consumer Behavior,
Greenwich, 51-83.
Brown J. and Reingen P. (1987), Social ties and word-
of-mouth referral behavior, Journal of Consumer
Research, 14, 3, 350-62.
Cakim I. (2002), E-fluentials expand viral marketing, 28,
www.imediaconnection.com.
Cameron S. (1995), On the role of critics in the culture
industry, Journal of cultural economics, 19, 4,
321-331.
Campbell D. (1969), Reforms as experiments, American
Psychologist, 24, 409-429.
Chaffee S. (1982), Mass media and interpersonal channels:
competitive, convergent or complementarity?,
Eds. G. Gumpert and R. Cathcart, Intermedia:
interpersonal communication in a media world,
New York, Oxford University Press.
Chang W.H. (1975), A typology study of movie critics,
Journalism Quarterly, 52, 4, 721-725.
Creton L. (2000), Critique et promotion dans lunivers
cinmatographique : distinctions, conjonctions et
dysphories, Entrelacs, Special Issue of Marketing du
cinma, 27-37.
DAstous A. and Colbert F. (2002), Moviegoers consultation
of critical review: psychological antecedents
and consequences, International Journal of Arts
Management, 5, 1, 24-35.
Dean D. and Biswas A. (2001), Third-party organization
endorsement of products ; An advertising cure
affecting consumer pre-purchase evaluation of good
and services, Journal of Advertising, 20, 4, 41-57.
Debenedetti S. (2006), Limpact de la critique de presse sur
la consommation culturelle : un essai de synthse dans le
champ cinmatographique, Recherche et Application
en Marketing, 21, 2, 43-59.
Debenedetti S. et Larceneux F. (2000), Typologie de la
critique cinmatographique et impact de la critique sur le
succs des films, Ed. Toulouse le Mirail University, Le
marketing du cinma, Entrelacs Special Issue, 38-52.
Dellarocas C. (2003), The digitization of word of mouth:
promise and challenges of online feedback mechanisms,
Management Science, 49, 10, 1407-1424.
Dellarocas C. and Narayan R. (2006), Statistical measure
of a populations propensity to engage in post-purchase
on-line word-of-mouth, Statistical Science, 21, 2,
277-285.
Dellarocas C., Farag N. and Zhang X. (2007), Using online
ratings as a proxy of word of mouth in motion picture
revenue forecasting, Under review, Available on
http://web.mit.edu\zxq\www\mit\15575\movieforecast.pdf
Fabrice Larceneux 58
03Larceneux 7/02/08 14:06 Page 58
De Vany A. and Walls D. (1999), Uncertainty in the movie
industry: does star power reduce the terror of the box-
office?, Journal of Cultural Economist, 23, 4, 285-318.
Duan W., Gu B. and Whinston A. (2005), Do online review
matter? An empirical investigation of panel data,
Working paper, Department of Management Science
and Information Systems, University of Texas, Austin.
Duhan D., Johnson S., Wilcox J. and Harell G. (1997),
Influence on consumer use of word of mouth
recommendation sources, Journal of The Academy of
Marketing Science, 25, 4, 283-285.
Dye R. (2000), The buzz on buzz, Harvard Business
Review, 78, 6, 139-146.
Elberse A. and Eliashberg J. (2003), Demand and supply
dynamics for sequentially released product in interna-
tional markets: the case of motion pictures, Marketing
Science, 22, 3, 226-243.
Eliashberg J. and Shugan S.M. (1997), Film critics:
influencers or predictors?, Journal of Marketing, 61, 2,
68-78.
Eliashberg J. and Sawhney M. (1994), Modelling goes to
Hollywood: predicting individual differences in movie
enjoyment, Management Science, 40, 9, 1151-1173.
Faber R.J. and OGuinn T.C. (1984), Effect of media
advertising and other sources on movie selection,
Journalism Quarterly, 61, 2, 371-377.
Feick L. and Price L. (1987), The market maven: a diffuser
of market place information, Journal of Marketing, 51, 1,
83-87.
Frodon J.-M. (2006), Trop de films en salle ?, Les Cahiers du
cinma, 618, 8-17.
Gemser G., Van Oostrum M. and Leenders M. (2007), The
impact of films reviews on the box-office performance of
art house versus mainstream motion pictures, Journal
of Cultural Economics, 31, 1, 43-63.
Godes D. and Mayzlin D. (2004), Using online conversa-
tions to study word-of-mouth communication, Marketing
Science, 23, 4, 545-560.
Granger C.W. (1969), Investigating causal relations in
econometric models and cross-spectral method,
Econometrica, 37, 3, 428-438.
Hamilton J. (2001), A wide web of advice, Business Week,
22, 1, 14-15.
Harrison-Walker L. (2001), The measurement of word-of-
mouth communication and an investigation of service
quality and customer commitment as potential
antecedents, Journal of Service Marketing, 4, 1, 60-75.
Hennig-Thurau T., Houston M. and Sridhar S. (2006), Can
good marketing carry a bad product? Evidence from
the motion picture industry, Marketing Letters, 17, 3,
205-219.
Herr P.M., Kardes F.R. and Kim J. (1991), Effects of word of
mouth on product attribute information on persuasion: an
accessibility diagnostic perspective, Journal of
Consumer Research, 17, 1, 454-462.
Hirsch P.M. (1972), Processing fads and fashions: an orga-
nizational set analysis of cultural industry systems,
American Journal of Sociology, 77, 4, 639-659.
Holbrook M.B. (1999), Popular appeal versus expert
judgments of motion pictures, Journal of Consumer
Research, 26, 2, 144-155.
Holbrook M.B. and Addis M. (2007), Taste versus the market:
an extension of research on the consumption of popular
culture, Journal of Consumer Research, 34, 2, 415-424.
Ito T., Larsen J., Smith K. and Cacioppo J.T. (1998),
Negative information weights more heavily on the
brain: the negativity bias in evaluation categorization,
Journal of Personality and Social Psychology, 75, 4,
887-901.
Jallat F. (2000), Les implications de la nouvelle conomie en
marketing, Confrence Proceedings on Marketing
Trends, Venice University, November 24.
Kahneman D. and Tversky A. (1979), Prospect theory: an
analysis of decision under risk, Econometrica, 47, 2,
263-241.
Kelley H. (1973), The process of causal attribution,
American Psychologist, 28, 2, 107-128.
Krider R. and Weinberg C. (1998), Competitive dynamics
and the introduction of new products: the motion picture
timing game, Journal of Marketing Research, 35, 1, 1-15.
Lampel J. and Shamsie J. (2000), Critical push: strategies for
creating momentum in the motion picture industry,
Journal of Management, 26, 2, 233-257.
Larceneux F. (2001), Critical opinion as a tool in the
marketing of cultural products: the experiential label,
International Journal of Arts Management, 3, 2,
60-71.
Leroy P. (2005), Louverture de la publicit aux spectateurs
interdits, Parliamentary Information Report, n 413.
Litman B. (1983), Predicting success of theatrical movies: an
empirical study, Journal of Popular Culture, 16, 4,
159-175.
Litman B. and Kohl L. (1989), Predicting financial success
of motion pictures: the 80s experience, The Journal of
Media Economics, 2, 35-50.
Litman B. and Ahn H. (1998), Predicting financial success of
motion pictures, Ed. B.R. Litman, The motion pictures
mega industry, Needham Height, MA, Allyn & Bacon,
172-197.
Liu Y. (2006), Word-of-mouth for movies: its dynamics
and impact on box-office revenue, Journal of
Marketing, 70, 3, 74-89.
Lovell G. (1997), Movies and manipulations, how studios
punish critics, Columbia Journalism Review, 35, 5,
9-12.
Lutaud L. (2006), Les Bronzs font leur cinma sur Internet,
Le Figaro, January 21.
Mann C. and Stewart F. (2000), Internet communication
and qualitative research, Thousand Oaks, CA, Sage.
Martilla J. (1971), Word-of-mouth communication in the
industrial adoption process, Journal of Marketing
Research, 8, 2, 173-178.
Montgomery D., Vining G. and Peck E. (2001), Linear
regression analysis, Chichester, Wiley.
Murray K. (1991), A test of service marketing theory:
consumer information acquisition activities, Journal of
Marketing, 55, 1, 10-25.
Buzz and recommendations on the Internet. What impacts on box-office success? 59
03Larceneux 7/02/08 14:06 Page 59
Neelamegham R. and Chintagunta P. (1999), A bayesian
model to forecast new product performance in domestic
and international markets, Marketing Science, 18, 2,
115-36.
Nelson P. (1970), Information and consumer behavior,
Journal of Political Economy, 78, 2, 311-329.
Park C.W., Mothersbaugh D. and Feick L. (1994), Consumer
knowledge assessment, The Journal of Consumer
Research, 21, 1, 71-82.
Prag J. and Casavant J. (1994), An empirical study of the
determinant of revenues and marketing expenditures in
the motion pictures industry, Journal of Cultural
Economics, 18, 3, 217-235.
Price L. and Feick L. (1984), The role of recommendation
source in external search: an informational perspective,
in T. Kinnear Ed., Advances in Consumer Research,
11, 250-255, Provo, UT: Association for Consumer
Research.
Rao A.R., Qu L. and Ruekert R.W. (1999), Signaling
unobservable product quality through a brand ally,
Journal of Marketing Research, 36, 2, 258-268.
Ravid S.A. (1999), Information blockbusters and stars: a
study of the film industry, Journal of Business, 72, 4,
463-492.
Reinstein D. and Snyder C. (2005), The influence of expert
review on consumer demand for experience goods: a
case study of movie critics, Journal of Industrial
Economics, 53, 1, 27-51.
Rosen D. and Olshavsky R. (1987), The dual role of
informational social influence: implications for
marketing management, Journal of Business Research,
15, 2, 123-144.
Sawhey M. and Eliashberg J. (1996), A parsimonious
model for forecasting gross box-office revenues of
motion pictures, Marketing Science, 15, 2, 113-131.
Senecal S. and Nantel J. (2004), The influence of online
product recommendations on consumers online
choices, Journal of Retailing, 80, 159-169.
Skowronski J. and Carlston D. (1989), Negativity and
extremity biases in impressions formation: a review of
explanations, Psychological Bulletin, 105, 1, 17-22.
Solomon M., Tissier Desbordes E. and Helbrunn B. (2005),
Comportement du consommateur, 6
th
Edition, Paris,
Pearson Education.
Spence M. (1974), Market signaling, Cambridge, Harvard
University Press.
Stambouli K. and Briones E. (2002), Buzz marketing,
Paris, ditions dOrganisation.
Swami S., Eliashberg J. and Weinberg C. (1999),
Silverscreener: a modelling approach to movie screens
management, Marketing Science, 18, 3, 352-72.
Thomas G.M. (2004), Building the buzz in the hive mind,
Journal of Consumer Behavior, 4, 1, 64-72.
Vernette E. and Flores L. (2004), Communiquer avec les
leaders dopinion en marketing : comment et dans quel
mdias ? Dcisions Marketing, 35.
Vernette E. (2007), Le leadership dopinion en marketing :
une double force dattraction et de conviction ?,
Proceedings of the 6
th
Conference in Marketing Trends,
ESCP-EAP.
Vogel H. (2001), Entertainment industry economics: a
guide for financial analysis, 5
th
Edition, Great-Britain,
Cambridge University Press.
Wallace W., Seigerman A. and Holbrook M. (1993), The
role of actors and actresses in the success of film,
Journal of Cultural Economics, 17, 1, 1-27.
Wang A. (2005), The effects of expert and consumer
endorsements on audience response, Journal of
Advertising Research, 45, 4, 402-411.
West P.M. and Broniarczyk S.M. (1998), Integrating multiple
opinions: the role of aspiration level on consumer
response to critic consensus, Journal of Consumer
Research, 25, 3, 38-51.
Wyatt O.W. and Badger D.P. (1990), Effects of information
and evaluation in film criticism, Journalism Quarterly,
67, 2, 359-368.
Zufryden F.S. (1996), Linking advertising to box-office
performance of new releases a marketing planning
model, Journal of Advertising Research, 35, 4, 29-41.
Zufryden F.S. (2000), New film website promotion and box-
office performance, Journal of Advertising Research, 40,
1, 55-64.
Fabrice Larceneux 60
03Larceneux 7/02/08 14:06 Page 60
Buzz and recommendations on the Internet. What impacts on box-office success? 61
APPENDIX A1
Frame 1 The allocine.com movie website
Nowadays, many people read up on films online before deciding what they want to see: 25% of all French
movie-goers are said to surf the Net before going to the cinema, with half of these coming to a decision after
they have checked the website, allocine.com (Lutaud 2006). Launched in 1997, allocine.com offers 24/7
information on any and all services that movie buffs or the general public might want: non-stop news, new
releases, showings, movie theater details, French and American awards, press reviews, a host of chat
forums, and recommendations on a scale of 0 to 4 stars. These reviews come from articles written by film cri-
tics, recoded by the website into a number of stars. There are also web user reviews that apply their own star
ratings when describing a film.
According to the Mediametrie Nielsen reference panel (7,000 households), in 2006 the site averaged more than
three million single visitors per month and achieved a reach rate of 17% (meaning that 17 out of 100 web users
went on allocine.com at least once a month). According to Xiti, the company that provides allocine.com
with its real-time information, by June 2006 the site was recording 7.1 million hits a month. As an example, the
UGC Bercy movie theater in Paris receives 92,000 hits a month checking programs, more than the number of
hits on the UGC website itself. One out of every five French web users has visited allocine.com (cf.
Akamai.com). In February 2007, allocine.coms audience was higher than youtube.com (source alexa.com). In
comparison, Frances leading weekly film magazine, Tlrama, has around 2.7 million weekly readers,
mostly the same people subscribing from one issue to the next.
Allocine.coms web users have a young profile matching the movie-goer core target of 18 to 34 (cf. cran Total,
n 624). According to its CEO, the site sections that receive the most hits are web user opinions and
reviews (cf. Le Journal du Net).
APPENDIX A2
Model 1 Independent variables correlation matrix
Films Presence of a star Number of copies Buzz from critics Buzz from web users
nationality in circulation
Films nationality 1 0.238(**) 0.056 0.019 0.106(*)
Presence of a star 1 0.356(**) 0.123(**) 0.248(**)
Number of copies
1 0.061 0.469(**)
in circulation
Buzz from critics 1 0.340(**)
Buzz from web users 1
** p < 0.01 * p < 0.05
03Larceneux 7/02/08 14:06 Page 61
Fabrice Larceneux 62
Model 2 Breakdown of films into classes
Number of films Number of films
in their first week after their first week
Initial classes Created classes Initial classes Created classes
Critics opinion
Unfavorable
0 star 12
123
9
89
1 star 111 80
Neutral 2 stars 247 247 196 196
Favorable
3 stars 160
164
151
154
4 stars 4 4
Web users opinion
Unfavorable 0 star 1 0
and/or 1 star 24 256 15 198
Neutral 2 stars 231 183
Favorable
3 stars 273
278
237
241
4 stars 5 5
APPENDIX A3
APPENDIX A4
Model 2 Independent variables correlation matrix
Firms Presence Number Number Critics Critics Number Web user
nationality of a star of copies of critics unfavorable favorable of first week opinions
in circulation opinions opinions web users (3 vs. 2 stars)
(1 star) (3 stars)
Firms
nationality 1 0.238(**) 0.056 0.035 0.044 0.062 0.093(*) 0.125(**)
Presence
of a star 1 0.356(**) 0.210(**) 0.034 0.016 0.266(**) 0.021
Number
of copies 1 0.036 0.039 0.127(**) 0.514(**) 0.008
in circulation
Number
of critics 1 0.504(**) 0.560(**) 0.294(**) 0.222(**)
Critics
unfavorable 1 0.364(**) 0.161(**) 0.321(**)
opinions
(1 star)
Critics
favorable 1 0.268(**) 0.273(**)
opinions
(3 stars)
Number
of first week 1 0.248(**)
web users
Web user
opinions
(3 vs. 2 stars) 1
** p < 0.01 * p < 0.05
03Larceneux 7/02/08 14:06 Page 62

You might also like