Impact of Online Consumer Reviews on Sales: The Moderating Role of Product and

Consumer Characteristics
Author(s): Feng Zhu and Xiaoquan (Michael) Zhang
Source: Journal of Marketing , Mar., 2010, Vol. 74, No. 2 (Mar., 2010), pp. 133-148
Published by: Sage Publications, Inc. on behalf of American Marketing Association

Stable URL: https://www.jstor.org/stable/20619095

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/20619095?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

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

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

American Marketing Association and Sage Publications, Inc. are collaborating with JSTOR to
digitize, preserve and extend access to Journal of Marketing

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 Feng Zhu & Xiaoquan (Michael) Zhang

Impact of Online Consumer Reviews

on Sales: The Moderating Role of
Product and Consumer
Characteristics
This article examines how product and consumer characteristics moderate the influence of online consumer
reviews on product sales using data from the video game industry. The findings indicate that online reviews are
more influential for less popular games and games whose players have greater Internet experience. The article
shows differential impact of consumer reviews across products in the same product category and suggests that
firms' online marketing strategies should be contingent on product and consumer characteristics. The authors
discuss the implications of these results in light of the increased share of niche products in recent years.

Keywords: Internet marketing, online consumer reviews, word of mouth, video game, long tail

Consumers commonly seek quality information when online consumer reviews are a good proxy for overall word
purchasing new products. With the Internet's growing of mouth (WOM) and can also influence consumers' deci
popularity, online consumer reviews have become sions. Empirical findings support this idea. For example,
an important resource for consumers seeking to discover Godes and Mayzlin (2004) find a positive relationship
product quality. A recent survey by comScore (2007), an between online WOM and television show viewership. Liu
Internet marketing research company, finds that 24% of (2006) studies movie reviews and finds that online movie
Internet users access online reviews before paying for a ser reviews offer significant explanatory power for both aggre
vice delivered offline. Accordingly, many firms are taking gate and weekly box office revenues. Dellarocas, Zhang,
advantage of online consumer reviews as a new marketing and Awad (2007) find that adding online movie ratings to
tool (Dellarocas 2003). Studies show that firms not only their revenue-forecasting model significantly improves the
regularly post their product information and sponsor pro model's predictive power. In general, these studies suggest
motional chats on online forums, such as USENET (May that many consumers make offline purchase decisions based
zlin 2006), but also proactively induce their consumers to on online information and that at least some aspects of
spread the word about their products online (Godes and online WOM are proxies for overall WOM.
Mayzlin 2004). Some firms even strategically manipulate The efficacy of online reviews could nonetheless be
online reviews in an effort to influence consumers' purchase limited. First, online reviews may merely represent con
decisions (Dellarocas 2006; Harmon 2004). sumers' preferences. These reviews may predict product
sales but have little influence on consumers' decisions. In
An underlying belief behind such strategies is that
the terms of Eliashberg and Shugan's (1997) study, online
online consumer reviews can significantly influence con
reviews in this case serve as predictors rather than influ
sumers' purchasing decisions. As we summarize in Table 1,
encers of product sales. Second, reviewers are not a ran
several studies show that professional reviews can signifi
domly drawn sample of the user population. Anderson
cantly influence consumers' decisions. With the prolifera
(1998) finds that extremely satisfied and extremely dissatis
tion of online review systems, many people believe that
fied customers are more likely to initiate WOM transfers. Li
and Hitt (2008) find potential bias in consumer reviews dur
Feng Zhu is Assistant Professor of Strategy, Marshall School of Business, ing early product introduction periods. Finally, interested
University of Southern California (e-mail: [email protected]). parties can easily manipulate online forums. Dellarocas
Xiaoquan (Michael) Zhang is Assistant Professor of ISOM, HKUST Busi
(2006) and Mayzlin (2006) theoretically analyze scenarios
ness School, Hong Kong University of Science and Technology, and is
affiliated faculty at MIT Center for Digital Business (e-mail:[email protected]).
in which firms can anonymously post online reviews to
The authors thank the two anonymous JM reviewers for their generous praise their products or to increase awareness about them.
input into this article that greatly improved its quality. They also thank Hai As a result, potential buyers may heavily discount online
Che, Chris Dellarocas, David Godes, Josh Lerner, John Little, Roland reviews.
Rust, Monic Sun, Chris Westland, and seminar participants at London Several recent studies (for a summary, see Table 2) have
Business School for many insightful comments. All errors or omissions
attempted to identify the relationship between online con
are their own. Both authors contributed equally to this work.
sumer reviews and product sales and have generated mixed

? 2010, American Marketing Association Journal of Marketing

ISSN: 0022-2429 (print), 1547-7185 (electronic) 133 Vol. 74 (March 2010), 133-148

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 TABLE 1
Previous Empirical Research Related to Professional Reviews
Study Method Data Key Findings
Litman (1983) Multiple regression Movies, 1972-1978 Critics' ratings are significant factors in explaining box
office revenue.

Mahajan, M?ller, Diffusion models Movies, 1983 Word of mouth was a significant predictor of
and Kerin (1984) attendance.

Wallace, Seigerman, Multiple regression Movie rental U-shaped relationship between critic ratings and rental
and Holbrook (1993) income.

Sawhney and Forecasting model, Movies, 1990-1991 Critics' reviews are positively significant for the number of
Eliashberg (1996) generalized gamma adopters.
Eliashberg and Correlation analysis Movies, 1991-1992 Critics are predictors rather than influencer; reviews
Shugan (1997) varied across critics.

Reddy, Swaminathan, Multiple regression Broadway shows, Newspaper critics have a significant impact on the
and Motley (1998) 1980-1982 and success of Broadway shows.
1990-1994
Holbrook (1999) Multiple regression Movies, pre-1986 Ordinary consumers and professional critics emphasize
different criteria in the formation of their tastes, but the
correlation between popular appeal and expert
judgments is positive.

Basuroy, Chatterjee, Multiple regression Movies, 1991-1993 Critics can influence and predict box office revenue.
and Ravid (2003)
Elberse and Demand/supply Movies, 1999 Less positive reviews correspond to a higher number of
Eliashberg (2003) model opening screens, but more positive reviews mean more
opening revenue.
Reinstein and Snyder Differences-in Movies, early 1990s Critics' influence on opening weekend box office revenue
(2005) differences is smaller than previous studies would suggest but is still
significant.

Zhang and Dellarocas Multiple regression

Movies, 2003-2004 Critics' influence is more significant than previously sug
(2006) gested, especially on early weeks' box office revenue.

Boatwright, Kamakura, Diffusion model Movies, 1997-2001 Some critics are especially influential in affecting the box
and Basuroy (2007) office revenue.

findings. For example, in an online experiment, Senecal and whose players have more Internet experience. In addition,
Nantel (2004) find that participants who consulted product online reviews are significantly more influential in affecting
recommendations selected these products twice as often as sales of less popular games than sales of more popular
those who did not consult recommendations. Chevalier and games. We also find that the influence of online reviews
Mayzlin (2006) find that online consumer ratings signifi becomes greater after the early, introductory months.
cantly influence product sales in the book market and that Our study is the first to empirically demonstrate the dif
customers actually read review text in addition to the ferential impact of consumer reviews across products in the
reviews' summary statistics. Zhang and Dellarocas (2006) same product category. The results imply that firms' online
obtain similar results in the movie industry. In contrast, marketing strategies may not be effective for all types of
Chen, Wu, and Yoon (2004) and Duan, Gu, and Whinston products, even if they are in the same category. This impli
(2008) find that online reviews do not influence sales and cation contrasts with the extant view that firms need to
serve only as predictors. actively manage online WOM, given the great efficiency of
Different from these studies, which focus on the average the Internet in spreading WOM, and that they should also
effect of online reviews on product sales, in this article, we strategically respond to online consumer reviews (e.g.,
examine contextual factors that moderate the relationship Chen and Xie 2005; Dellarocas 2006).
between the two. We propose a conceptual framework and Our study also suggests that niche producers and pro
hypothesize that product- and consumer-specific character ducers that sell mostly through online channels should be
istics affect consumers' reliance on online consumer more concerned about online consumer reviews and
reviews and thus are important factors governing the effi manipulations of online review systems because online
cacy of online reviews. Using a data set on sales and con reviews could significantly affect their sales. Because the
sumer reviews of video games, we find that online con proliferation of online markets has led to the emergence of
sumer reviews have a greater influence on the sale of games many niche producers, a phenomenon often dubbed the

134 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 TABLE 2
Previous Empirical Research Related to Consumer Reviews
Study Method Data Key Findings
Resnick and Zeck Multiple regression eBay, 1999 Sellers with better reputations are more likely to sell
hauser (2002) their items but they enjoy no boost in price.
Godes and Mayzlin Multiple regression Television shows, Online conversations offer one way to measure word of
(2004) 1999-2000 mouth.

Chen, Wu, and Yoon Multiple regression Amazon.com books, Consumer ratings are not correlated with sales.
(2004) 2003
Senecal and Nantel Generalized esti Online experiment Participants who consulted product recommendations
(2004) mating equations selected recommended products twice as often as those
who did not consult recommendations.

Liu (2006) Multiple regression Movies, 2002 WOM information offers significant explanatory power for
both aggregate and weekly box office revenue, especially
in the early weeks after a movie opens.

Chevalier and Mayzlin Differences-in Books, 2003-2004 Online amateur book ratings affect consumer purchasing
(2006) differences behavior.

Dellarocas, Zhang, Diffusion model Movies, 2002 Online amateur movie ratings can be used as a proxy for
and Awad (2007) word of mouth.

Duan, Gu, and Simultaneous Movies, 2003-2004 The rating of online user reviews has no significant
Whinston (2008) system impact on movies' box office revenues.

"long tail" (Anderson 2006), our results have important experience goods (Nelson 1970), whose product character
implications for their survival. istics are difficult to observe until consumption. Thus,
In the following sections, we develop our conceptual online reviews could be useful in reducing the risk of pur
framework and provide background information about the chasing such products. Figure 1 depicts our conceptual
video game industry and the cross-platform development of framework. Online reviews are expected to influence product
video games. After discussing the data sources, we develop sales only when consumers' reliance on online reviews is
an empirical strategy and present the results. We conclude sufficiently high when they make purchase decisions. In turn,
with a discussion of the implications of our findings. the degree of reliance depends on product- and consumer
specific characteristics. In addition, other factors, such as
competition, business models (e.g., business-to-consumer,
Conceptual Framework consumer-to-consumer), or even the online review system's
We focus on single-purchase products in our study. Infor design (e.g., how ratings are displayed, how easy it is to rate
mation goods, such as books, movies, music, and computer an item), may affect consumers' reliance on reviews.
games, are examples of products purchased only once. The framework is closely related to the psychological
Many of these single-purchase products can be considered choice model in Hansen (1976), in which the effectiveness

FIGURE 1
Conceptual Framework

Product characteristics
(e.g., product popularity)

Consumer characteristics Consumers' reliance on Consumers' purchase

(e.g., Internet experience) online reviews decisions

Other factors
(e.g., competition, business
models, design of online
review systems)

Impact of Online Consumer Reviews on Sales /135

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
of an influencer (online reviews) is moderated by environ feeling and can be interpreted as a preference later (Born
mental and contextual factors (consumer and product char stein 1989; Zajonc 1980). Along this line, Janiszewski
acteristics) and the interactions among these variables even (1993) finds that the mere exposure effect persists even
tually determine the response (purchase decisions). when initial exposure to brand names and product packages
Consistent with this framework, several studies show that is unintended. Because popular products are discussed more
consumers' use of different information sources indeed frequently than less popular products, and thus consumers
varies with product characteristics. Beatty and Smith (1987) are exposed to them repeatedly, the exposure effect could
find that consumers' search effort is influenced by their have a significant impact on consumer purchasing behavior.
product knowledge. Reinstein and Snyder (2005) find that The preceding discussion suggests that both the ratings
professional reviews have a significant effect on opening and the number of reviews could be more salient for popu
weekend box office revenue for narrowly released movies lar products. Because we measure product popularity by its
and for dramas, but not for widely released movies or for sales, we hypothesize the following:
genres such as action movies and comedies. Cheema and
Hla: An increase in online reviews (e.g., online ratings, the
Papatla (2010) find that the relative importance of online
number of online reviews) results in higher incremental
information is higher for utilitarian products than for hedo sales for products that currently have relatively high
nic products. Studies also point to the important effect of sales.
consumer characteristics on the reliance on certain informa
tion sources. For example, Westbrook and Fornell (1979) Conversely, online consumer reviews could be less
show that consumers' background characteristics, such as influential for popular products. For example, consumers
education attainment, affect their need for information may have a lower need to resort to online reviews for popu
related to purchase decisions. Klein and Ford (2003) find lar products. A major reason consumers use online reviews
that consumers' online experience moderates their trust in is to obtain quality information to reduce risk (Bolton,
different information sources. Katok, and Ockenfels 2004; Chen, Xu, and Whinston 2009;
Similar to these studies, we adopt the view that product Clemons, Gao, and Hitt 2006; Forsythe and Shi 2003;
and consumer-specific characteristics can significantly Pavlou and Gefen 2004). Being popular in itself signals
moderate the relationship between online reviews and pur higher quality. Previous studies have shown strong linkages
chase decisions. In our study, we focus on product popular between a product's popularity and its perceived quality.
ity (measured by the products' sales) as the product-specific For example, Caminal and Vives (1996) develop a model
characteristic and consumer Internet experience (measured based on market signaling in the presence of imperfect
by the length of time consumers have been using the Inter information and find that future consumers interpret popu
net) as the consumer-specific characteristic. larity or large market shares as a signal of high quality.
Hellofs and Jacobson (1999) suggest several mechanisms
Product Popularity through which popularity influences perceived quality, such
Online consumer reviews could have a greater impact on as signaling, creation of network externalities, and inclusion
the sales of popular products for several reasons. First, as an attribute in consumers' quality functions.
popular products tend to receive more reviews, and having a Studies have also shown that the purchase of popular
large number of reviews makes such online reviews seem products tends to minimize potential risk. DeSarbo and col
more trustworthy. As Kirby (2000, p. El) explains, a con leagues (2002) argue that consumers prefer popular prod
sumer "may not trust just one nonexpert,... but if 9 out of 10 ucts because popularity represents a type of social cue, and
nonexperts agree, it's probably worth buying." Chen, Wu, following the social cue tends to reduce perceived risk. In a
and Yoon (2004) confirm that an increase in information similar vein, the literature on herding suggests that it is
sources could lead to more trust. They show that as the sometimes optimal for consumers to ignore or not seek pri
number of consumer reviews increases, the overall rating vate information and to follow the crowd (e.g., Banerjee
converges to the true quality. Therefore, reviews of popular 1992; Bikhchandani, Hirshleifer, and Welch 1992). The lit
products could more accurately reflect product quality and erature on consumer decision making (e.g., Josephs et al.
thus could be more influential. 1992; Zeelenberga and Beattieb 1997; Zeelenberga, Van der
Second, given the large number of reviews popular Pligta, and De Vriesa 1996) suggests that consumers take
products receive, consumers may be more confident that greater responsibility for negative outcomes when their
they can find reviews for a popular product online and thus actions deviate from the norm or the default option. In the
are more likely to search for online reviews for popular context of consumer purchase decisions, Simonson (1992)
products. Disproportionately more searches are likely to shows that consumers feel more regret if they choose a
increase the influence of these reviews. In contrast, if con lesser-known brand that turns out to be inferior than if they
sumers believe that reviews of less popular products are rare
choose a well-known brand that turns out not to be better
and difficult to find, they may not search for such reviews at than the lesser-known option. Thus, consumers interested in
all. Reviews of less popular products would then have little a less popular product are likely to search and access more
impact on their purchase decisions. WOM information to shield themselves from possible
Finally, reviews of popular products could have a regret (Chatterjee 2001).
greater effect on consumers' decisions because consumers Finally, consumers use a mix of online (e.g., online
are exposed to these reviews more often. Extant studies sug reviews, blogs) and offline (e.g., family and friends, sales
gest that mere exposure is sufficient to create a favorable people, magazines) WOM information to help structure

136 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 their decisions. Prior research has shown that WOM effec Thus, because consumers with more Internet experience
tiveness depends on the strength of ties or the intensity of are more likely to use the Internet as their primary informa
the relationship among consumers (e.g., Granovetter 1973). tion source and are more likely to have greater confidence
Strong ties are perceived as more influential than weak ties, in the Internet, they are more likely to be influenced by
and they are more likely to be used as sources of informa online reviews. We hypothesize the following:
tion (e.g., Bansal and Voyer 2000; Brown and Reingen H2a: An increase in online reviews (e.g., online ratings, the
1987). Because consumers often cannot determine the number of online reviews) results in higher incremental
source's credibility in an online environment, tie strength sales for products targeting consumers with greater Inter
online is typically weak (e.g., Chatterjee 2001; Mesch and net experience.
Talmud 2006). Indeed, experimental evidence shows that At the same time, however, consumers with greater
when both channels are present, the offline channel is gen Internet experience may find online information to be less
erally preferred over the online channel (Frambach, Roest, credible. Because anyone can provide information online,
and Krishnan 2007). Because popular products are more the quality of such information tends to vary significantly.
likely to be featured in offline channels, such as magazines An experienced online user is more likely to have been
and store demos, and discussed among friends, their con exposed to information sources with lower reliability or to
sumers may not resort to online reviews for quality infor have encountered negative experiences (Cheema and Pap
mation and thus are less likely to be influenced by online ada 2010). As a result, while a novice may easily trust
reviews. online opinions, Internet veterans are not nearly as easily
The foregoing discussion suggests that online reviews influenced. Consistent with this argument, Cheema and
could be more effective in influencing the purchases of less Papatla (2010) analyze data from a telephone survey and
popular products because consumers are more likely to seek show that consumers with greater Internet experience have
quality information to minimize the purchase risk and the diminished interest in online sources. Similarly, through a
likelihood of postpurchase regret, and such quality informa survey conducted with automobile shoppers and purchasers,
tion is likely to be unavailable from offline channels. There Klein and Ford (2003) find that experienced online con
fore, it is an empirical question whether online reviews are sumers rate offline information sources as significantly
more influential for popular or less popular products. We more credible than online sources.
propose the following competing hypothesis: In addition, consumers with greater Internet experience
can easily find many reviews about a product from multiple
Hlb: An increase in online reviews (e.g., online ratings, the
sources. However, assessing the validity of these informa
number of online reviews) results in higher incremental
sales for products that currently have relatively low sales. tion sources imposes significantly higher cognitive costs. To
deal with the information overload problem, consumers are
more selective about the types of information to which they
Consumer Internet Experience
respond (Rust and Chung 2006). As a result, the relation
The Internet significantly reduces search costs (Brynjolfs ship between online reviews and their purchase decisions
son and Smith 2000) and enables the convenient compari could be weaker.
son of various alternatives (Keeney 1999). Consumers with Therefore, because consumers with greater Internet
greater Internet experience are more likely to use online experience lack trust in online information and information
channels to collect product information because their cost overload carries a high cognitive cost for them, they are not
of collecting information from the online channel is likely as easily influenced by online reviews. We hypothesize the
to be lower than that from the offline channel (Cook and following:
Coupey 1998). Several field studies confirm that Internet H2b: An increase in online reviews (e.g., online ratings, the
experience is positively correlated with the frequency of number of online reviews) results in higher incremental
using the Internet to gather information (e.g., Kehoe et al. sales for products targeting consumers with less Internet
1999; Palmquist and Kim 2000; Weiser 2000) and search experience.
performance (e.g., Lazonder, Biemans, and Wopereis 2000). In the context of video games, because games have dif
Similarly, Novotny (2004) studies how users search infor ferent genres and story lines, they may or may not offer an
mation online and finds that a lack of Internet experience online multiplayer mode. In the online mode, a player can
affects user persistence and often leads to quick abandon connect through the Internet with other players and interact
ment of the Internet as an information source. These studies
with them in real time. We subsequently refer to games with
suggest that consumers with greater Internet experience are both an online mode and an offline mode as "online games"
more likely to access online reviews. and those with only an offline mode as "offline games."
Research also shows that a consumer with greater Inter Playing games online requires not only a fast Internet con
net experience is likely to have a different perception of the nection but also a relatively high level of skill in coordinat
attributes of the online channels from that of an Internet
ing and communicating online with other players in real
novice and the consumer may have greater confidence in time. Thus, we expect that, on average, consumers purchas
the Internet (Bart et al. 2005). For an Internet novice, in ing online games will have greater Internet experience.
contrast, using online information may evoke perceptions of Indeed, studies find a positive relationship between the fre
uncertainty and complexity. Therefore, Internet experience quency of playing online games and the frequency and
may moderate the evaluation of online information. length of Internet usage (e.g., Lo, Wang, and Fang 2005).

Impact of Online Consumer Reviews on Sales /137

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
Therefore, we test H2a and H2b by examining the differen ment platform) to program a game in a single language and
tial impact of online reviews on the sales of online and port the game onto several consoles. Many publishers no
offline games. longer view delayed cross-platform development as an
option. Instead, they often mandate that developers release
games on all three major console platforms simultaneously
Video Game Industry and Cross (Reimer 2005).
Platform Game Development We restrict our analysis to games that are developed for
The video game industry is becoming increasingly impor both Sony's PlayStation 2 and Microsoft's Xbox for two
tant, and its growth far outpaces other entertainment indus reasons. First, during the period for which we have review
tries, such as movies and music. From 2003 to 2006, the data, PlayStation 2 and Xbox were the two largest players
video game software industry's annual growth rate in the 128-bit console market and had the largest game
exceeded 17%, in contrast to the U.S. economy's 2% libraries. Second, both consoles target adults between the
growth rate over the same period (The Entertainment Soft ages of 18 and 34, positioning themselves directly against
ware Association 2007). The industry's annual revenue was each other; therefore, we expect the two gaming popula
approximately $17.94 billion in 2007 (Thorsen 2008), tions to be similar. We compare the features of PlayStation
which was almost double the box office revenue in the 2 and Xbox consoles. The only major differences between
motion picture industry. Halo 3, the best-selling game title the two consoles are the clock speed and the amount of
of 2007, took in more revenue ($170 million) in its first day memory.
of sales than the biggest opening weekend ever for a movie
(Spider-Man 3, $150 million). The penetration rate of video
game consoles is also high: Approximately 41% of U.S. Empirical Analysis
households owned video game consoles in 2006 (Arendt Data
2007). Thus, our study not only enriches the literature on
online reviews but also offers insights into this important Data on console sales and game sales come from NPD, a
sector for marketing practitioners. leading market research firm that tracks this industry. NPD
The role of reviews is potentially greater for video collects data from approximately 17 leading retail chains
games than for movies. First, there are more game titles that account for 80% of the U.S. market. From these data,
than movie titles. In 2007, the Entertainment Software Rat NPD formulates estimates of sales figures for the entire
ing Board gave out 1563 ratings to a subset of all games U.S. market. We obtain monthly data for PlayStation 2 and
produced that year. Facing so many choices, a game player Xbox and their associated games from October 2000 to
would need to invest substantial time and energy to identify October 2005. For each game, we compute the average
good games. Second, a video game typically costs more monthly price by dividing the monthly dollar value of sales
than a movie. According to NPD Fun Group (hereinafter, by the volume of units sold.
NPD), the average selling price of a game was $38.36 at the We gather review data from GameSpot.com (also known
end of 2007. Because most gamers are young and have lim as VideoGames.com). According to Alexa.com, a Web site
ited incomes, they frequently use reviews to avoid bad purr providing an online traffic monitoring service, GameSpot,
chases (Bounie et al. 2005). Therefore, it is not surprising com is the 65th most-visited site in the United States and
that Game Informer, a magazine featuring articles, news, the most popular one for video games, reaching more than
and reviews of popular video games, ranks among the most 10 million unique gamers each month (GameZone 2004).
highly circulated magazines, and game review Web sites, GameSpot publishes three kinds of reviews: editors'
such as GameSpot.com, are consistently ranked among the reviews, players' reviews, and reviews from other sources.
top 100 most popular Web sites in the United States. Editors at GameSpot review most games on or around the
Publishers usually fund game development. The cost of day they ship to retail channels. In March 2003, GameSpot
developing a contemporary video game is approximately began publishing player reviews. To ensure the quality of
$6-$ 10 million (Edge 2005). A game can take from one to these reviews, only paid subscribers or users with a suffi
three years to develop depending on the genre, scale, devel cient level of experience (as demonstrated by their partici
opment platform, and amount of assets. In the early days, pation in other parts of the site, such as forums) are allowed
most game titles were developed for a single console, and to post them. A maximum of one review is allowed from the
whenever a game was ported to a new console, a different same log-in name for a given game. These policies mini
team would need to rewrite the entire game. Development mize the potential manipulation of the review system and
teams would use assembly language, a human-readable ensure that reviews are of high quality. For each of five
notation for the machine language, to write most games aspects (game play, graphics, sound, value, and reviewer's
because this language optimized the processing speed and tilt), reviewers use a scale ranging from 1 to 10 for their
required little overhead. Today, because processing speed is reviews, 10 being the best and 1 being the worst. For each
no longer a critical issue, high-level languages, such as C++ review, GameSpot publishes the weighted average of all
and Java, are the most popular game development lan five aspects. We use this weighted average rating of all five
guages (Goodwin 2005). In addition, although code aspects in our analysis. In addition, GameSpot collects crit
libraries for different consoles are not compatible, game ics' reviews from other sources, such as Yahoo! Games and
developers can take advantage of cross-platform middle Hardcore Gamer Magazine, and publishes aggregate scores
ware platforms (e.g., Criterion's RenderWare 3D develop based on these reviews, most of which are published within

138 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 a month after the games are released. The reviews by the cumvent many of the endogeneity problems that typically
editors at GameSpot and from other sources are rarely arise when making causal arguments (Meyer 1995). Our
updated after they are published, and therefore they vary lit empirical analysis hinges on the video games that are
tle over time. Their effects are eliminated in our differences released for two consoles: PlayStation 2 and Xbox. By tak
in-differences estimation. The player reviews vary both ing the differences between the sales of the same game title
across consoles and over time and are the focus of our for the two consoles, we eliminate unobserved common fac
analysis. Even for the same game titles, player reviews tend tors, such as game characteristics, that may affect both
to be different across consoles. reviews and sales on both consoles. By examining the dif
We collect reviews for each game in each month ferences across consoles over time, we control for console
between March 2003 and October 2005. Following previ specific factors, such as the underlying taste difference
ous research on consumer reviews (e.g., Chevalier and between the console-installed bases, which may influence
Mayzlin 2006; Zhang and Dellarocas 2006), we focus on both reviews and sales. A game title often receives different
three review variables: the average rating, the coefficient of reviews on the two consoles. Consider a situation in which
variation of ratings, and the total number of reviews posted. a game title receives better reviews on one console than on
The average rating reflects the level of consumer satisfac the other: The differences-in-differences approach enables
tion and is the focus of most empirical studies on product us to test whether an increase in the game title's sales on
reviews. The coefficient of variation, measured as the ratio one console relative to the same game title's sales on the
of the standard deviation to the mean rating, captures the other console is a result of differences in reviews. As with
degree of disagreement among consumers. High variation other studies using the differences-in-differences approach,
carries both great risk and great reward, while low variation the key underlying assumption here is that the effect of
offers a safe bet. Prior research has shown that for different these unobserved console-specific factors is the same for
products, variation of consumer reviews may be positively games on both consoles in each period.
or negatively associated with product sales (e.g., Martin, Our analysis differs from that of Chevalier and Mayzlin
Barron, and Norton 2008; Sun 2008). We also collect the (2006) in several aspects. First, the two studies focus on dif
total number of reviews as a measure of the volume of dis ferent questions. Chevalier and Mayzlin examine online
cussions. The number of reviews captures the exposure reviews' aggregate influence, while we examine how prod
effect and may signal a game's popularity. uct and consumer characteristics may moderate the influ
Although GameSpot offers a convenient way to mea ence of online reviews. Second, our empirical strategy
sure online WOM, its reviews may not be representative of explicitly controls for competition among games by esti
all online opinions on specific games. Players can also mating a nested logit demand model. The demand for a
obtain review information from other channels, such as game is likely to be affected by the number of competing
online bulletin boards and chat rooms. Therefore, our cur games on the market, and the intensity of this substitution
rent estimate might underestimate the relationship between effect may vary across consoles and over time. Firms' pric
reviews and sales. Had we been able to consider all sources ing strategies may adjust according to the intensity of com
of information, our conclusions would be strengthened. We petition; therefore, it is important to capture this effect to
merge the sales data with the review data to obtain the final obtain unbiased estimates from a demand equation. In
data set. Chevalier and Mayzlin, the demand for individual books is
implicitly assumed to be independent of that of competitors,
Methodology though in reality, the availability of books in the same cate
An inherent problem in measuring the influence of reviews gory is likely to affect the demand for a particular book.
on product demand is that products receiving positive Third, our sales data for games cover the whole U.S. mar
reviews tend to be of high quality. Because quality is often ket, while Chevalier and Mayzlin examine book sales only
unobserved by researchers, it is difficult to determine on two Web sites and approximate book sales using ranks.
whether the review or the quality is responsible for the high Thus, our results capture the effects of online reviews on
demand. Therefore, positive correlations between reviews purchase decisions made both online and offline. Finally,
and product sales might be spurious. our data include all games with positive sales in each
Recent studies propose several methods to circumvent month, while the book data that Chevalier and Mayzlin use
this problem. For example, Einav (2007) and Zhang and are truncated because Amazon.com and BN.com do not
Dellarocas (2006) use fixed-effects specifications to control report rank data for books with low popularity. Therefore,
for unobserved movie quality. Reinstein and Snyder (2005) we cannot fruitfully test hypotheses related to product popu
take advantage of the timing of critics' reviews relative to a larity with book data.
movie's release and find that the measured influence effect We now describe our two-stage nested logit demand
is small but still detectable. Chevalier and Mayzlin (2006) model for games. We assume that there are J games avail
examine book reviews and sales ranks on Amazon.com able for console k and an outside option labeled 0. We place
and BN.com on different dates and use a differences-in the J games in one group, g, and the outside option in
differences approach to eliminate book- and site-specific another group by itself. In the first stage, a player decides
effects.
whether to purchase a game. In the second stage, if the
In this article, similar to Chevalier and Mayzlin (2006), player chooses to purchase a game, he or she then decides
we adopt a differences-in-differences approach. The which game to purchase. For any given game, the player
differences-in-differences approach is widely used to cir has at most unit demand. The perceived utility of player i

Impact of Online Consumer Reviews on Sales /139

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 from purchasing a game j, j e [1, J], for console k at time t, Second, the additive separable functional form yields a lin
Ujjt, is affected by game price, perceived game quality, and ear regression specification, as we discuss subsequently.
other game characteristics. Our conceptual framework sug Therefore, we could use straightforward instrumental
gests that perceived game quality is affected by a combina variable methods to handle endogenous variables, such as
tion of consumer reviews, game popularity, and the player's game prices (Berry 1994).
Internet experience. We employ two measures of popularity. We normalize the utility from the outside good to be
The first measure is a cross-sectional dummy variable that zero. Because game players must have a game console
equals 1 if the game's aggregate sales across the two con before they can play games developed for this console, we
soles are greater than the mean performance of all games in use the size of the installed base of each console as the
the month. The second measure captures the intertemporal potential market. In addition, because the two consoles are
pattern of games' life cycles because the popularity of a incompatible, the potential market for games is console spe
video game often drops rapidly after its release. The aver cific. We denote the share of the potential market captured
age life cycle of all games is approximately 33 months, but by game j of console k in period t as sjt and game j's share
on average, more than 50% of game sales occur within the of the portion of the market that purchases games in period
first four months after a game's release. Therefore, for any t (i.e., the share of game j within group g) as s?|g. Thus, sj^g
game, the first four months after release could be consid can be computed as Sjfy(l - SQt), where SQt is the market
ered the period in which it is popular. The dummy variable share of the outside option in period t for console k. Fol
takes the value of 1 when the game is in the first four lowing Berry (1994) and Cardell (1997), we derive the
months of its life cycle. To operationalize consumer Internet demand equation for the two-stage nested logit model as
experience, we create a dummy variable indicating whether follows:
a game can only be played offline or not. Other variables,
such as market share, prices, and game characteristics, can (2) ta^J-lnfs^?o + ?^ + ?^,.,
be obtained directly from the original data set. Thus, we
express the player's utility, uyt, as a function of price (pj^), +?3(r",-i xpopu]arjt) + ?4(r^_1 xoffline^
lagged review variable (rj?-1)> a dummy indicating whether
a game is popular (popular^), a dummy indicating whether
+?5popularjt + ?6offline + a In(s^g ) + %\.
a game can only be played offline (offlinej), and other game
characteristics (^t): Given the panel structure of data, we decompose the
component ?jft as follows:
(1) 4 = ?o + ?iP* + ?2rit-i + ?3 f?-i x Popularjt )
+ ?4 [rft-\ x ?fflinej ) + ?5P?Pularjt + ?6offlinej where 0jt is a game-specific component that is the same for

+$+4+(l-?H the same game across different platforms but can vary over
time, rij?is the console-specific effect, and e? is an i.i.d. nor
mal error term varying across games and over time. The 9jt
The two dummy variables, popular^ and offlinej, indi component is related to factors such as promotions by game
cate different subgroups among all the video games.
publishers, the brands of the game publishers, and the
Because our conceptual framework suggests that the review
variable's effect is conditional on the type of product, we game's age and quality, and v^ captures the difference in
players' tastes of the consoles and the fit between game j
interact the review variable with these dummy variables.
and console k. Even for the same game title, players' utility
With these interaction terms, we can measure the effects of
may differ because of difference in console characteristics,
online reviews of different types of products (Aiken and
West 1991). For example, ?2 measures the influence of the such as clock speed. Therefore, rj^is time invariant but may
vary across games on the same console.
reviews on games with popular^ = 0 and offlinej = 0 (i.e., Following Equation 2, using superscripts p and x to
less popular and online games). Similarly, ?2 + ?3, ?2 + ?4,
denote PlayStation 2 and Xbox, respectively, we have the
and ?2 + ?3 + ?4 measure the influence of consumer
following:
reviews on popular and online games, less popular and
offline games, and popular and online games, respectively.
We also include two unobservables, ?j^ and vfjt, where ^gt (3) ln(SP)-ln(sPt)=?0 +?lPP +?2rPt_, + ?3(rPt_, xpopularj
represents player utility common to all games of group g
+?4 (rP^! x ofHinej ) + ?5popularjt
and v-jt is an i.i.d. extreme-value distributed error term that
represents player i's idiosyncratic taste for games in group
g. The parameter as [0, 1) measures the correlation of
+?6oiHinej + oln(SP|g) + 9jt+1iP + eP,and
unobserved utility among games in the same group. When
a ?> 1, games within a group are perfect substitutes,
whereas when a = 0, they are independent, and we have the (4) ln(s}t)-ln(s^t)=?0+?1pft+?2rj;t_1 + ?3(rj;t_1xpopularjt)
simple logit model.
We use an additive separable functional form in Equa +?4 (r^.! x offlinej ) + ?5popularjt
tion 1 for two reasons. First, this form enables us to capture
the moderating effects easily using the interaction terms.
+?6offlinej + oln(s*t|g) + 8jt + r\] + e*.

140 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
The within-group market shares (WGSs), ln(sPt|g) and forums, we would expect that most game players read
ln(s*t|g), are, by definition, endogenous and require instru reviews for games on one console. In addition, if reviews
mental variables. Following Einav (2007), we use the num for the two consoles have more or less the same influence
ber of games available for each console at time t as the on game sales, our differences-in-differences approach
instrument for the WGS. A large number of games implies would eliminate their impact, and differences in the review
intense competition and therefore should be negatively variables should not show significant correlations with sales
associated with the WGS. In addition, because the potential difference. Thus, our empirical study serves as a test for this
market size may change sharply across consoles and over assumption.
time, we add the installed base of each console as a control
Summary Statistics
variable in the instrument specification. Furthermore, game
Our final data set consists of 220 game titles that were
prices, PjPt and p*t, could be endogenous in our demand
available for the two consoles between March 2003 and
model. Although we do not have cost-side variables to use
as instruments, as Berry (1994) and Nair, Chintagunta, and
October 2005. Of these, 79 had different release dates for
Dub? (2004) suggest, we could use characteristics of com the two consoles, and we remove them from the sample.
peting games as instruments. For each game, we collected Similar to other empirical studies based on discrete choice
data on its genre (e.g., first-person shooter, party games, models (e.g., Argentesi and Filistrucchi 2007; Einav 2007;
puzzle games) and the Entertainment Software Rating Rysman 2004), a natural concern is the assumption of a
Board rating (e.g., everyone, adult-only, teen). Following single purchase?each consumer purchases, at most, one
Nair, Chintagunta, and Dub?'s approach, we use the number game in each period. This seems to be a reasonable assump
and average age of competing games in the same genre, of tion in the case of video games. According to a recent sur
competing games in the same Entertainment Software Rating vey, more than 80% of consumers purchase one game or
Board group, and of all competing games, as well as their less in each month, on average (Pidgeon and Hu 2003).
squared terms, as instruments for game prices in each month. However, consumers' purchase frequency could exhibit sea
sonal patterns. Figure 2 shows mean revenue and mean
The parameter 0jt contains both observed and unob
served game-specific characteristics. The unobserved char units sold by month for all games over the sampling period.
acteristics are likely to be correlated with price and review As the figure indicates, the monthly game sales exhibit
variables, and omitting their effects would produce biased strong holiday effects: The average monthly units sold are
close to 20 million between January and October, but this
coefficients. Because 0jt is the same across console systems,
we eliminate the game-specific effects by differencing the number increases substantially for November and Decem
data across consoles: ber. Thus, consumers are more likely to purchase multiple
games in November and December. We remove observa
tions in November and December from our data set.1
(5) AMjt=?1Apjt + ?2Arjt_1+?3(Arjt_1xpopularjt)
Several recent articles (e.g., Dub? 2004; Hendel 1999; Hendel
+ ?4 ?Arjt_ j x offline j ) +aAWGSjt + Arij + ?jt, and Nevo 2006) present techniques that can accommodate multiple
where discrete choices. Because these models in general do not yield linear
specifications, it is difficult to combine them with our differences
AMj,t = [ln(sPt) - ln(sgt)] - [ln(s*t) - ln(sSt)], in-differences approach.

Apj,t = PjPt - Pjt.

FIGURE 2
Arjt_! = r^t.! -r^t_!,
Mean Units and Mean Revenue by Month for All
AWGSj,t = ln(SjP|g) - ln(s*|g), and Games
A?lj = ?lJp-ilJx. x10? X109
The variable Ar)j, which captures the differences in console
specific effects, is also unobserved but does not vary over
time. Because console differences may affect differences in
game prices and reviews, we take an additional difference
of Equation 5 between period t and t + 1 and obtain the
following:

(6) AAMjt =?1(AAPjt) + ?2(AArjt_1) + ?3(AApj txpopularjt)

+?4iAApjtxofflinej) + a(AAWGSjt) + 8jt.
Equation 6 is our empirical specification.
An implicit assumption in our empirical methodology is
that the price and reviews of a game on one console have
minimal, if any, influence on the sales of the game with the
same title on another console. Because the two consoles are
Mean units sold
incompatible with each other and most game players only Mean revenue
own one console and participate in its associated online

Impact of Online Consumer Reviews on Sales /141

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 Table 3 provides summary statistics for games in our TABLE 4
sample. A t-test indicates that the monthly unit sales of Summary Statistics for Reviews
games for PlayStation 2 are significantly greater than those
A: Summary Statistics for Reviews of PlayStation 2
for Xbox. The result is consistent with the larger installed Games
base of PlayStation 2 console and the strong indirect net
work effects that Clements and Ohashi (2005) and Zhu and
Variable M SD Minimum Maximum
Iansiti (2009) document. The prices for games for the two Average rating 7.34 1.66 1.40 9.60
consoles are at about the same level, most likely because of Variation of ratings .14 .13 .00 .68
the intense competition among game titles for each console: Number of reviews 9.21 12.43 1.00 63.00
On average, in each month, 475 games on the Xbox console B: Summary Statistics for Reviews of Xbox Games
and the 810 games on PlayStation 2 console have positive
sales. Variable M SD Minimum Maximum
Table 4 presents summary statistics of reviews as of Average rating 7.48 1.49 1.35 9.60
October 2005. The data suggest that reviews are overwhelm Variation of ratings .17 .17 .00 .85
ingly positive for games for both consoles. Researchers
Number of reviews 10.30 16.29 1.00 104.00
have observed similar patterns in other contexts, such as Notes: Panels A and B present summary statistics for reviews of
book reviews on Amazon.com (Chevalier and Mayzlin games on PlayStation 2 and Xbox as of October 2005 in our
sample. The average rating is the arithmetic mean of all rat
2006) and reputation profiles on eBay (Resnick and Zeck ings from March 2003 and October 2005 for each game. We
hauser 2002). On average, there are more than 9 reviews for measured the variation of ratings as the ratio of the standard
each game. The distribution of the number of reviews is deviation to the mean rating. The number of reviews is the
skewed: The number of reviews ranges from 1 to 63 for total number of posted reviews for each game.

PlayStation 2 games and from 1 to 104 for Xbox games. We

find no significant difference in any of the three metrics
Regression Results
across the two consoles, suggesting that the two gaming Table 5 presents the regression results based on the
populations are similar. A concern is that the same reviews differences-in-differences specification in Equation 6. We
may be posted for both consoles, and as a result, reviews use the differences of ln(sp - ln(s0t) across consoles and
from the two gaming populations are artificially similar. We over time as the dependent variables in all models. In
check this possibility and find that only 3.3% of the review Model 1, we use the differences-in-differences measures of
ers write reviews for both consoles.
price, average rating of the reviews, and WGS of the
Figure 3 shows the mean prices, units sold, and ratings games.2 We find that game price negatively affects game
for PlayStation 2 and Xbox games. In all three panels, the demand, while average consumer rating has no effect on it.
patterns for PlayStation 2 and Xbox are similar. Both average The small and nonsignificant coefficient of the WGS sug
price and average units sold decline over time. The average gests that video games are poor substitutes for each other,
price declines almost linearly during the first ten months, which is consistent with Nair's (2007) finding.
and the average units sold also drops significantly for the In Model 2, we add the interactions between the average
first few months. Because many games are not released dur
rating and game popularity and between the average rating
ing the first days in the month, mean units sold during the and game online capability. We first define the popularity of
first month of the release for games for both consoles
a game, popular^, as a dummy variable, which takes the
appear relatively low. Average ratings in the first couple of value of 1 if the sales of game j for both consoles are greater
months are significantly higher than those in later months.
This pattern suggests the existence of a self-selection bias 2We take the natural logarithm of price and review variables
in the reviews and is similar to that reported by Dellarocas, (i.e., average rating, variation of ratings, and number of reviews).
Zhang, and Awad (2007) and Li and Hitt (2008). The vari We use the logarithms of (the number of reviews + 1) and (the var
iation + 1) to handle zero reviews and ratings with no variation.
ance of the mean ratings increases over time because we
have fewer reviews for old pames
TABLE 3
Summary Statistics for Games
A: Summary Statistics for Games on PlayStation 2
Variable Number of Observations M SD Minimum Maximum
Monthly sales (units) 3,330 10,038.84 25,518.40 5 561,540
Monthly price ($) 3,330 21.58 11.23 1.80 54.85
B: Summary Statistics for Games on Xbox
Variable Number of Observations M SD Minimum Maximum
Monthly sales (units) 3,305 5,499.10 15,816.72 7 378,194
Monthly price ($) 3,305 21.32 11.10 1.88 54.79
Notes: Panels A and B present summary statistics for games developed on both PlayStation 2 and Xbox consoles in ou
from March 2003 to October 2005. We calculate the monthly price calculated by dividing the monthly dollar valu
of units sold.

142 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
than the mean performance of all games in month t. Infor sites.3 The coefficient of the rating variable here measures
mation on each game's online capability is collected from the influence of less popular and online games and is posi
GameSpot. GameSpot specifies whether a game has an tive and significant. We also find that the influence is signif
online play mode. We also verify our data with information icantly weaker for popular games or offline games, as evi
from game publishers' and console manufacturers' Web denced by the significant, negative coefficients of the two
interaction variables.
FIGURE 3
In Model 3, we add the differences-in-differences mea
Mean Prices, Units Sold, and Ratings over Time
sures of other review variables, such as the variation of the
for Games on PlayStation 2 and Xbox
30n a more technical note, online play requires online platforms
A: Average Prices
(often provided by either console manufacturers or game publish
ers). Because online platforms designed for one console may not
be usable for a different console, sometimes a game can be played
online on one console before it can be played online on the other.
In addition, not all games can be played online at the time of their
releases. Game players in general are aware of the online capabil
ity of the games (it is often included in the game descriptions).
Even if the online capability of a game is not supported at the time
of purchase, players often anticipate that they will be able to play
it online in the near future. Thus, we define online games or offline
games on the basis of their online capability (instead of whether
the games are actually being played online).

5 10 15 20 25 30 35 40 45
TABLE 5
Measuring the Influence of Reviews on Game
Number of Months After Release
Demand
B: Number of Units Sold Models
12 3 4 5
AA Price -2.12***
(.80) (.80) (1
AA Average ratin
(.21) (.41) (
AA Within-gr
share (.13) (.13) (.13) (.13) (.13)
AA Average -.71** -.68* -.64* -.75*
rating x popular (.36) (.39) (.39) (.39)
AA Average -.73* -.78** -.55 -.61*
rating x offline (.43) (.40) (.42) (.36)
5 10 15 20 25 30 35 40 45 AA Price x popular 2.94** 1.43
Number of Months After Release (1.38) (1.48)
AA Price x offline .89 .67
(1.77) (1.80)
C: Average Score AA Variation of -.83 -.64
rating (.56) (.43)
AA Variation of .26 .45
rating x popular (.52) (.48)
AA Variation of .56 .55
rating x offline (.52) (.53)
AA Number of .55** .50***
reviews (.23) (.18)
AA Number of -.11 -.01
reviews x popular (.15) (.12)
AA Number of -.49** -.49**
reviews x offline (.20) (.20)
Observations 1142 1142 1142 1142 1142
5 10 15 20 25 30 35 40 45 R2 .01 .01 .03 .01 .03
Number of Months After Release
*p<.10.
**p < .05.
PlayStation 2 games ***p<.01.
Xbox games Notes: We use Equation 6 as the regression model. We use the
differences of ln(sft) - In(sot) across consoles and over time
Notes: The plots exclude the observations in November and as the dependent variables in all models. AA indicates that we
December. Because many games are not released during take the differences of the variable across console and over
the first days in the month, mean units sold during the first time. All regressions employ an ordinary least square
month of the release for games on both consoles appear specification. Heteroskedasticity-adjusted standard errors
relatively low. are in parentheses.

Impact of Online Consumer Reviews on Sales /143

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
ratings and the number of reviews, and their interactions TABLE 6
with game popularity and online capability. Because prod Coefficients for Different Types of Games
uct and consumer characteristics may also affect game pub
Popular Less Popular
lishers' pricing decisions, we add the interactions of game
prices with game popularity and online capability. The
Price
Online -.60 -3.54**
results suggest that the demand for popular games is less
sensitive to price. The results on the rating variable are
(1.58) (1.71)
Offline -.29 -2.65**
similar to those in Model 2. In addition, we find that the (1.05) (1.11)
number of reviews has a positive effect on less popular and Average Rating
online games. A possible explanation for this is that having Online .32 1.00*
a large number of reviews signals a game's popularity. This (.33) (.41)
Offline -.45 .22
result is also likely caused by the presence of direct network
effects: For games that can be played online, players are
(.32) (.25)
Variation of Rating
more likely to purchase games that many others have Online -.57 -.68*
bought. The effect of the number of reviews becomes (.43) (.39)
weaker for offline games. Offline -.01 -.27
In Models 4 and 5, we employ an intertemporal measure (.33) (.46)
Number of Reviews
of game popularity. Instead of measuring popularity by
Online .44*** .50***
comparing different games in each month, we define popu
(.14) (.18)
larity for a given game over its life cycle. In these two mod Offline -.05 .06
_?10)_(-12)
els, we consider a game popular if it is less than four
months old and less popular otherwise. We replicate the
*p<.10.
analyses in Models 2 and 3 with this new measure and **p < .05.
obtain similar results. The results suggest that online ***p<.01.
reviews are less influential in the early phases of game life Notes: We use the results in Model 3 of Table 5 to compute the
influence of review variables and price on all four types of
cycles. The result is noteworthy in light of prior research
games (i.e., popular and offline games, popular and online
suggesting that product promotions are more effective in the games, less popular and offline games, and less popular and
early stages of a product's life cycle because uncertainty online games). We use the linear combinations of the
and the need for information tend to be high (Sethuraman estimates and test them against zero to obtain these
coefficients and heteroskedasticity-adjusted standard errors
and Tellis 1991). If we evaluate the possible effects of (in parentheses).
advertising or WOM through the lens of the Bass (1969)
diffusion model, we should expect the effects to be greater Chen, Wu, and Yoon (2004) and Duan, Gu, and Whinston
in the early stages of introduction. A plausible explanation (2008) consider the average rating and the number of
is that in entertainment industries, the heavy use of other
reviews only, and Godes and Mayzlin (2004) focus on the
promotional strategies through offline channels in the early volume of conversations in each newsgroup. In addition,
stages of product life cycles reduces consumers' reliance on many studies find that only one or two aspects of online
online reviews. We also note that R-squares of our models reviews affect product sales. For example, Duan, Gu, and
are relatively small. A possible reason for this is that the Whinston find that the volume of online reviews matters,
variance of the error term in our empirical specification but the average rating does not. In contrast, we show that
could be large because we obtained the error term after dif for products that are less popular and targeted at consumers
ferencing four error terms in our differences-in-differences with great Internet experience, all three aspects could mat
approach. ter. Overall, our regression analysis finds support for Hlb
The coefficients of the price variable and the review and H2a.
variables in Table 5 measure their influence on less popular
and online games. We can use the regression results to com
pute their influence on other types of games (e.g., popular Discussion and Conclusion
and online games, popular and offline games, less popular
and offline games). Table 6 summarizes the results (based
Managerial Implications
on Model 3 in Table 5). The results show that the sales of Understanding how online reviews affect consumers' pur
less popular games are negatively affected by their prices. chase decisions is vitally important to firms that rely on
The coefficients of the average rating and the variation of online WOM to disseminate information about their prod
rating are significant only for less popular and online ucts. We find that for video games, online reviews are more
games. Finally, the coefficient of the number of reviews is influential for less popular and online games. Our empirical
significant for online games. results support the view that the impact of online consumer
These results suggest that all three aspects of the reviews on product sales depends on product and consumer
reviews?the average rating, the variation of rating, and the characteristics. Thus, firms' online marketing strategies
number of reviews?affect the sales of less popular and need to adjust accordingly.
online games. Our results are more comprehensive than The finding that online reviews are more influential for
those in previous studies because most of these studies only less popular games suggests that the informational role of
consider one or two aspects of online reviews. For example, reviews becomes more salient in an environment in which

144 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 alternative means of information acquisition are relatively now shows how many people rated the item with each of
scarce. As such, marketers of less popular products may the 1-5 stars; readers can choose to read reviews for a given
benefit more from allocating resources to managing online star level.
consumer reviews. Because of the scarcity of available
information about niche products, even one negative review Research Implications
can be detrimental. Superior online WOM translates more This research provides a potential positive reconciliation of
easily into sales for niche products, and thus the existence the mixed results from previous studies. For example,
of online review systems gives a great incentive for niche Chevalier and Mayzlin (2006) examine book sales at
market producers to exert efforts to maintain good reputa Amazon.com and find that online reviews influence book
tions. These results are particularly useful in light of niche sales, but, using a similar data set from Amazon.com, Chen,
products' increased market share in recent years, owing to Wu, and Yoon (2004) find the opposite. Similarly, in the
virtually unlimited "shelf space" in online markets. Recent context of the movie industry, Zhang and Dellarocas (2006)
studies (e.g., Anderson 2006; Brynjolfsson, Hu, and find that online reviews influence box office sales, but
Simester 2005) show that as a result of the Internet, the Duan, Gu, and Whinston (2008) find the opposite.
economy is increasingly shifting away from a relatively Researchers have not been able to reconcile the stark differ
small number of mainstream products at the head of the ences in results and instead have attributed them to method
demand curve and toward a huge number of niches in the ological shortcomings. For example, Duan, Gu, and Whin
tail, a phenomenon often dubbed the long tail. For example, ston point out that the mixed finding could be the result of
Brynjolfsson, Hu, and Smith (2006) find that obscure book researchers conducting their analyses in a cross-sectional
titles, which are typically not available in conventional context and not controlling for unobserved differences in
bookstores, account for approximately 40% of Amazon, product quality. Our study suggests that data sets with a dif
corn's book sales in 2000. Although the Internet has ferent mix of product types, even for the same product cate
increased the collective share of niche products, it does not gory, could lead to different conclusions. For example, in
necessarily guarantee the survival of firms producing niche studying online book reviews, two data sets with different
products. Elberse and Oberholzer-Gee (2006) and Elberse proportions of popular and less popular book titles or differ
(2008) find that from 2000 and 2005, though the number of ent proportions of technical books (whose readers presum
video titles selling only a few copies every week increases ably have greater Internet experience) and nontechnical
almost twofold during this period, the number of nonselling books may find that online reviews play different roles.
titles rises rapidly and becomes four times as high as in This work could be extended in several directions. First,
2000. Because many niche products are only sold online further research could take a similar approach to examine
and their buyers are more likely to use online review sys the differential roles of critics' reviews on various types of
tems as the primary source for quality information, our products within the same product category. As an example,
study suggests that online WOM could significantly con Eliashberg and Shugan (1997) find that film critics are lead
tribute to dispersion in the tail. Therefore, it is crucial for ing indicators of a movie's ultimate success but do not influ
niche product producers to devote their marketing efforts to ence its early run at the box office. However, several recent
online review systems when they take advantage of online studies (e.g., Basuroy, Chatterjee, and Ravid 2003; Reinstein
channels to sell their products. and Snyder 2005) find that film critics can influence open
This study also finds evidence to support the notion that ing weekend box office revenues. Heterogeneity across dif
online reviews are more influential when consumers have ferent movies might be a source of these divergent findings.
relatively greater Internet experience. Echoing the discussion Second, this research implies that online consumer
in the conceptual framework about users' Internet experi reviews might significantly affect the diffusion and adop
ence, the empirical results suggest that, at least in the video tion of less popular products that target consumers with
games market, the benefits of reduced search costs (Bryn much Internet experience. Further research could test
jolfsson and Smith 2000) and greater confidence in using whether diffusion models for forecasting the sales of such
the Internet (Bart et al. 2005) seem to dominate concerns products can substantially improve their accuracy after
about the reliability and credibility of online information incorporating online consumer reviews.
sources (Cheema and Papatla 2010; Klein and Ford 2003). Third, further research could investigate firms' online
As the Internet population continues to grow, consumers and offline marketing strategies and compare their effec
will inevitably become more experienced with the Internet. tiveness. Our research indicates that promotions in the
Our study suggests that, over time, marketing managers will offline channel may reduce the efficacy of online reviews.
find online consumer reviews to be increasingly influential Thus, it would be worthwhile to theoretically and empiri
and thus should devote more resources to online channels. cally analyze firms' optimal strategy in allocating marketing
At the same time, firms that rely heavily on using online resources to online and offline channels.
channels to promote their products could also seek ways to Finally, further research could compare the influence of
reduce the search costs for online reviews. After the barrier online reviews among multiple products. While our analysis
to information acquisition becomes lower, even Internet focuses on a single product category, the results are applic
novices could be influenced by online reviews. For exam able to multiple categories. For example, we expect online
ple, to reduce the search costs for reviews, Amazon.com has reviews to have a greater influence on products that are
recently modified the way it reports star levels for items. likely to be purchased or used online (e.g., software) than
While previously it showed only an average star rating, it on those sold or used mostly offline (e.g., apparel).

Impact of Online Consumer Reviews on Sales /145

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 REFERENCES
Aiken,
sion:
Leona -and
Testing
S.
and
Michaeland D. Smith (2000), Stephen
Interpreting
Comparison of Internet
3
"Frictionless Commerce? A
and Conventional Retailers," Manage
G

CA: Sage ment Science, 46 (4), 563-85.

Publications.
Anderson, Chris (2006),
Caminal, Ramon and Xavier Vives (1996), "WhyThe
Market Shares Lon
ness Is Selling Matter: An Information-Based
Less of Theory," More. Rand Journal of Eco New
Anderson, Eugene nomics, 27 (2),
W. 221-39. (1998), "Cu
Mouth," JournalCardell, Scott N.of Service
(1997), "Variance Components Structures for the Res
Arendt, Susan (2007),
Extreme-Value "Game
and Logistic Distributions with Application to
(accessed Models of
August 1,Heterogeneity,"
2008), Econometric Theory, 13 (2),
[availa
185-213.
games/2007/03/game_consoles_i
Argentesi, Chatterjee, Patrali
Elena and (2001), "OnlineLapo
Reviews: Do Consumers Filist
Use
ket Power in aThem?" in Advances in Consumer Research, Mary C. Gilly
Two-Sided Mark
Journal of and Joan Myers-Levy,Econometric
Applied eds. Provo, UT: Association for Con
Banerjee, Abhijit sumer Research,
V. 129-34. (1992), "A S
Quarterly Cheema,
Journal Amar and Purushottam
of PapatlaEconomic
(2010), "Relative Impor
tance of Online Versus Offline Information for Internet Pur
Bansal, Harvir S. and Peter A.
Processes WithinThe a
chases: Effect of Product Category and Internet ExperiPurc
Services
nal of Service ence," Journal of Business Research, forthcoming. 3
Research, (2), 1
Bart, Yakov, Chen, Jianqing, Hong Xu, and AndrewShankar,
Venkatesh B. Whinston (2009),
"Moderated Online Communities and User-Generated Con
(2005), the Drivers "Are and Role
All tent," NET Institute Working
andWeb Sites Paper No. 09-11, (accessed Janu
Consumers?
ary 4, 2010), [available at http://ssrn.com/abstract= 1481772].
Empirical Study," Journal of M
Bass, Frank M.Chen, Pei-Yu, Shin-Yi Wu, and Jungsun Yoon
(1969), "A (2004), "The New
Impact P
of Online Recommendations and Consumer Feedback on
sumer Durables," Management
Basuroy, Suman, Sales," in Proceedings
Subimal of the International ConferenceChat
on
Information Systems, ICIS 2004. Seattle: Association for Infor
(2003), "How Critical Are Criti
mation Systems, 711-24.
Effects of Film Critics, Star Pow
Chen, Yubo and Jinghong Xie (2005), "Third-Party Product
Marketing, 67 (October), 103-1
Review and Firm Marketing Strategy," Marketing Science, 24
Beatty, Sharon E.
(2), 218^0. and Scott M.
Effort: An Investigation Acros
Chevalier, Judith A. and Dina Mayzlin (2006), "The Effect of
Journal of Consumer Research,
Word of Mouth on Sales: Online Book Reviews," Journal of
Berry, Steve (1994), "Estimating
Marketing Research, 43 (August), 345-54.
uct Differentiation," RAND J
242-62. Clements, Matthew T and Hiroshi Ohashi (2005), "Indirect Net
work Effects and the Product Cycle: Video Games in the U.S.,
Bikhchandani, Sushil, David Hirshleifer, and Ivo Welch (1992),
1994-2002," Journal of Industrial Economics, 53 (4), 515-42.
"A Theory of Fads, Fashion, Custom, and Cultural Change as
Clemons, Eric K., Guodong Gao, and Lorin M. Hitt (2006),
Informational Cascades," Journal of Political Economy, 100
"When Online Reviews Meet Hyperdifferentiation: A Study of
(5), 992-1026.
the Craft Beer Industry," Journal of Management Information
Boatwright, Peter H., Wagner Kamakura, and Suman Basuroy Systems, 23 (2), 149-71.
(2007), "Reviewing the Reviewers: The Impact of Individual comScore (2007), "Online Consumer-Generated Reviews Have
Film Critics on Box Office Performance," Quantitative Mar
Significant Impact on Offline Purchase Behavior," press
keting and Economics, 5 (4), 401-425. release, (November 29), (accessed August 2008), [available at
Bolton, Gary E., Elena Katok, and Axel Ockenfels (2004), "How http://www.comscore.com/press/release.asp?press=1928].
Effective Are Electronic Reputation Mechanisms? An Experi Cook, Don Lloyd and Eloise Coupey (1998), "Consumer Behavior
mental Investigation," Management Science, 50 (11), and Unresolved Regulatory Issues in Electronic Marketing,"
1587-1602.
Journal of Business Research, 41 (3), 231-38.
Bornstein, Robert F. (1989), "Exposure and Affect: Overview and Dellarocas, Chrysanthos (2003), "The Digitization of Word of
Meta-Analysis of Research, 1968-1987," Psychological Bul Mouth: Promise and Challenges of Online Feedback Mecha
letin, 106 (2), 265-89. nisms," Management Science, 49 (10), 1401-1424.
Bounie, David, Marc Bourreau, Michel Gensollen, and Patrick - (2006), "Strategic Manipulation of Internet Opinion
Waelbroeck (2005), "Do Online Customer Reviews Matter? Forums: Implications for Consumers and Firms," Management
Evidence from the Video Game Industry," Telecom ParisTech Science, 52 (10), 1577-93.
Working Paper No. ESS-08-02. -, Xiaoquan M. Zhang, and Neveen Awad (2007), "Explor
Brown, Jacqueline Johnson and Peter H. Reingen. (1987), "Social ing the Value of Online Product Reviews in Forecasting Sales:
Ties and Word-of-Mouth Referral Behavior," Journal of Con The Case of Motion Pictures," Journal of Interactive Market
sumer Research, 14 (3), 350-62. ing, 21 (4), 23-45.
Brynjolfsson, Erik, Yu Hu, and Duncan Simester (2005), "Good DeSarbo, Wayne S., Juyoung Kim, S. Chan Choi, and Melinda
bye Pareto Principle, Hello Long Tail: Modeling and Measur Spaulding (2002), "A Gravity-Based Multidimensional Scaling
ing the Effect of Search Cost on Product Sales Distribution," Model for Deriving Spatial Structures Underlying Consumer
workshop on Information Systems and Economics (WISE), Preference/Choice Judgments," Journal of Consumer Research,
University of California, Irvine (December 10-11). 29 (1), 91-100.
-,-, and Michael D. Smith (2006), "From Niches to Duan, Wenjing, Bin Gu, and Andrew B. Whinston (2008), "Do
Riches: The Anatomy of the Long Tail," Sloan Management Online Reviews Matter? An Empirical Investigation of Panel
Review, 47 (4), 67-71. Data," Decision Support Systems, 45 (4), 1007-1016.

146 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
 Dub?, Jean-Pierre (2004), "Multiple Discreteness and Product Dif tive Consequences of Risky Decisions," Journal of Personality
ferentiation: Demand for Carbonated Soft Drinks," Marketing and Social Psychology, 62 (1), 26-37.
Science, 23 (1), 66-81. Keeney, Ralph L. (1999), "The Value of Internet Commerce to the
Edge (2005), "Report: Only 80 Games a Year Will Succeed," Customer," Management Science, 45 (4), 533-42.
(November 16), (accessed November 19, 2009), [available at Kehoe, Colleen, Jim Pitkow, Kate Sutton, Gaurav Aggarwal, and
http://www.edge-online.com/news/report-only-80-games-a Juan D. Rogers (1999), "Results of GVU's Tenth World Wide
year-will-succeed]. Web User Survey," technical report, Graphics Visualization and
Einav, Liran (2007), "Seasonally in the U.S. Motion Picture Usability Center, College of Computing, Georgia Institute of
Industry," RAND Journal of Economics, 38 (1), 127-45. Technology.
Elberse, Anita (2008), "Should You Invest in the Long Tail?" Har Kirby, Carrie (2000), "Everyone's a Critic: Web Sites Hope Online
vard Business Review, 86 (July-August), 88-96. Reviews of Products Lead to Online Buying," San Francisco
- and Jehoshua Eliashberg (2003), "Demand and Supply Chronicle, (January 22), El.
Dynamics for Sequentially Released Products in International Klein, Lisa R. and Gary T. Ford (2003), "Consumer Search for
Markets: The Case of Motion Pictures," Marketing Science, 22 Information in the Digital Age: An Empirical Study of Pre
(3), 329-54. Purchase Search for Automobiles," Journal of Interactive Mar
-and Felix Oberholzer-Gee (2006), "Superstars and Under keting, 17 (3), 29-49.
dogs: An Examination of the Long Tail Phenomenon in Video Lazonder, Ard W, Harm J.A. Biemans, and Iwan G.J.H. Wopereis
Sales," Harvard Business School Working Paper Series, No. (2000), "Differences Between Novice and Experienced Users
07-015. in Searching Information on the World Wide Web," Journal of
Eliashberg, Jehoshua and Steven M. Shugan (1997), "Film Critics: the American Society for Information Science, 51 (6), 576-81.
Influencers or Predictors?" Journal of Marketing, 61 (April), Li, Xinxin and Lorin M. Hitt (2008), "Self Selection and Informa
68-78. tion Role of Online Product Reviews," Information Systems
The Entertainment Software Association (2007), Video Games & Research, 19 (4), 456-74.
the Economy, (accessed December 2009), [available at Litman, Barry R. (1983), "Predicting the Success of Theatrical
http://www.theesa.com/gamesindailylife/economy.asp]. Movies: An Empirical Study," Journal of Popular Culture, 16
Forsyth, Sandra M. and Bo Shi (2003), "Consumer Patronage and (4), 159-75.
Risk Perceptions in Internet Shopping," Journal of Business Liu, Yong (2006), "Word of Mouth for Movies: Its Dynamics and
Research, 56 (11), 86^'-75. Impact on Box Office Revenue," Journal of Marketing, 70
Frambach, Ruud T, Henk C.A. Roest, and Trichy V. Krishnan (July), 74-89.
(2007), "The Impact of Consumer Internet Experience on Lo, Shao-Kang, Chih-Chien Wang, and Wenchang Fang (2005),
Channel Preference and Usage Intentions Across the Different "Physical Interpersonal Relationships and Social Anxiety
Stages of the Buying Process," Journal of Interactive Market Among Online Game Players," CyberPsychology & Behavior,
ing, 21 (2), 26-41. 8 (1), 15-20.
GameZone (2004), "ESPN Videogames Sweeps Gamespot Game Mahajan, Vijay, Eitan M?ller, and Roger A. Kerin (1984), "Intro
Commercial Awards," (February 17), (accessed August 2008), duction Strategy for New Products with Positive and Negative
[available at http://www.gamezone.com/news/02_17_04_12_ Word-of-Mouth," Management Science, 30 (12), 1389-1404.
27PM.htm]. Martin, Jolie M., Gregory M. Barron, and Michael I. Norton
Godes, David and Dina Mayzlin (2004), "Using Online Conversa (2008), "Variance-Seeking for Positive (and Variance-Aversion
tions to Study Word-of-Mouth Communication," Marketing for Negative) Experiences: Risk-Seeking in the Domain of
Science, 23 (4), 545-60. Gains?" Harvard Business School Working Paper Series, No.
Goodwin, Steven (2005), Cross-Platform Game Programming. 08-070.
Boston: Charles River Media. Mayzlin, Dina (2006), "Promotional Chat on the Internet," Mar
Granovetter, Mark (1973), "The Strength of Weak Ties," American keting Science, 25 (2), 155-63.
Journal of Sociology, 78 (6), 1360-80. Mesch, Gustavo and Ilan Talmud (2006), "The Quality of Online
Hansen, Flemming (1976), "Psychological Theories of Consumer and Offline Relationships: The Role of Multiplexity and Dura
Choice," Journal of Consumer Research, 3 (3), Wl-M. tion of Social Relationships," The Information Society, 22 (3),
Harmon, Amy (2004), "Amazon Glitch Unmasks War of Review 137-^8.
ers," The New York Times, (February 14), (accessed November Meyer, Bruce (1995), "Natural and Quasi-Natural Experiments in
12, 2009), [available at http://www.nytimes.com/2004/02/14/ Economics," Journal of Business and Economic Statistics, 13
technology/14AMAZ.html]. (2), 151-62.
Hellofs, Linda L. and Robert Jacobson (1999), "Market Share and Nair, Harikesh (2007), "Intertemporal Price Discrimination with
Customers' Perceptions of Quality: When Can Firms Grow Forward-Looking Consumers: Application to the US Market _
Their Way to Higher Versus Lower Quality?" Journal of Mar for Console Video-Games," Quantitative Marketing and Eco
keting, 63 (January), 16-25. nomics, 5 (3), 239-92.
Hendel, Igal (1999), "Estimating Multiple-Discrete Choice Mod -, Pradeep Chintagunta, and Jean-Pierre Dub? (2004),
els: An Application to Computerization Returns," Review of "Empirical Analysis of Indirect Network Effects in the Market
Economic Studies, 66 (2), 423-46. for Personal Digital Assistants," Quantitative Marketing and
-and Aviv Nevo (2006), "Measuring the Implications of Economics, 2 (1), 23-58.
Sales and Consumer Inventory Behavior," Econometrica, 74 Nelson, Philip (1970), "Information and Consumer Behavior,"
(6), 1637-73. Journal of Political Economy, 78 (2), 311-29.
Holbrook, Morris B. (1999), "Popular Appeal Versus Expert Judg Novotny, Eric (2004), "I Don't Think I Click: A Protocol Analysis
ments of Motion Pictures," Journal of Consumer Research, 26 Study of Use of a Library Online Catalog in the Internet Age,"
(2), 144-55. College & Research Libraries, 65 (6), 525-37.
Janiszewski, Chris (1993), "Preattentive Mere Exposure Effects," Palmquist, Ruth A. and Kyung-Sun Kim (2000), "Cognitive Style
Journal of Consumer Research, 20 (3), 376-92. and Online Database Search Experience as Predictors of Web
Josephs, Robert A., Richard P. Larrick, Claude M. Steele, and Search Performance," Journal of the American Society for
Richard E. Nisbett (1992), "Protecting the Self from the Nega Information Science, 51 (6), 558-66.

Impact of Online Consumer Reviews on Sales /147

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms
Pavlou, Paul A. and David Gefen (2004), "Building Effective Simonson, Itamar (1992), "The Influence of Anticipating Regret
Online Marketplaces with Institution-Based Trust," Informa and Responsibility on Purchase Decisions," Journal of Con
tion Systems Research, 15 (1), 37-59. sumer Research, 19 (1), 105-118.
Pidgeon, Billy and David Hu (2003), "U.S. Console Gamers Sur Sun, Monic (2008), "How Does Variance of Product Ratings Mat
vey: Consumer Interest Justifies Online Initiatives," Zelos ter?" working paper, Graduate School of Business, Stanford
Group report, (accessed November 23, 2009), [available at University.
http://www.zelosgroup.com]. Thorsen, Tor (2008), "NPD: 2007 Game Earnings Nearly $18 Bil
Reddy, Srinivas K., Vanitha Swaminathan, and Carol M. Motley lion, Halo 3 Sells 4.82 Million," (accessed August 1, 2008),
(1998), "Exploring the Determinants of Broadway Show Suc [available at http://www.gamespot.com/news/6184847.html].
cess," Journal of Marketing Research, 35 (August), 370-83. Wallace, W. Timothy, Alan Seigerman, and Morris B. Holbrook
Reimer, Jeremy (2005), "Cross-Platform Game Development and (1993), "The Role of Actors and Actresses in the Success of
the Next Generation of Consoles," (November 7), (accessed Films: How Much Is a Movie Star Worth?" Journal of Cultural
November 23, 2009), [available at http://arstechnica.com/old/ Economics, 17 (1), 1-27.
content/2005/11/crossplatform.ars]. Weiser, Eric B. (2000), "Gender Differences in Internet Use Pat
Reinstein, David A. and Christopher M. Snyder (2005), "The terns and Internet Application Preferences: A Two-Sample
Influence of Expert Reviews on Consumer Demand for Experi Comparison," CyberPsychology & Behavior, 3 (2), 167-78.
ence Goods: A Case Study of Movie Critics," Journal of Indus Westbrook, Robert A. and Claes Fornell (1979), "Patterns of Infor
trial Economics, 53 (1), 27-50. mation Source Usage Among Durable Goods Buyers," Journal
Resnick, Paul and Richard Zeckhauser (2002), "Trust Among of Marketing Research, 16 (August), 303-312.
Strangers in Internet Transactions: Empirical Analysis of Zajonc, Robert B. (1980), "Feeling and Thinking: Preferences
eBay's Reputation System," in The Economics of the Internet Need No Inference," American Psychologist, 35 (2), 151-71.
and E-Commerce, Advances in Applied Microeconomics, Vol. Zeelenberga, Marcel and Jane Beattieb (1997), "Consequences of
11, Michael R. Bay, ed. Amsterdam: Elsevier Science, 127-57. Regret Aversion 2: Additional Evidence for Effects of Feed
Rust, Roland T. and Tuck S. Chung (2006), "Marketing Models of back on Decision Making," Organizational Behavior and
Service and Relationships," Marketing Science, 25 (6), 560-80. Human Decision Processes, 72 (1), 63-78.
Rysman, Marc (2004), "Competition Between Networks: A Study -, -, Joop van der Pligta, and Nanne K. de Vriesa
of the Market for Yellow Pages," Review of Economic Studies, (1996), "Consequences of Regret Aversion: Effects of Expected
71 (2), 483-512. Feedback on Risky Decision Making," Organizational Behav
Sawhney, Mohanbir S. and Joheshua Eliashberg (1996), "A Parsi ior and Human Decision Processes, 65 (2), 148-58.
monious Model for Forecasting Gross Box Office Revenues of Zhang, Xiaoquan M. and Chris Dellarocas (2006), "The Lord of
Motion Pictures," Marketing Science, 15 (2), 321-40. the Ratings: How a Movie's Fate Is Influenced by Reviews," in
Senecal, Sylvain and Jacques Nantel (2004), "The Influence of Proceedings of the 27th International Conference on Informa
Online Product Recommendations on Consumers' Online tion Systems (ICIS). Milwaukee: Association for Information
Choices," Journal of Retailing, 80 (2), 159-69. Systems.
Sethuraman, Raj and Gerard J. Tellis (1991), "An Analysis of the Zhu, Feng and Marco Iansiti (2009), "Entry into Platform-Based
Tradeoff Between Advertising and Pricing," Journal of Market Markets," working paper, Marshall School of Business, Uni
ing Research, 31 (May), 160-74. versity of Southern California, (accessed January 4, 2010),
[available at http://ssrn.com/abstract-1350233].

148 / Journal of Marketing, March 2010

This content downloaded from

154.59.124.141 on Wed, 12 Apr 2023 14:43:10 UTC
All use subject to https://about.jstor.org/terms