2. Theory and hypotheses

(1)

Vox populi?

The moderating role of network externalities on selection systems in the market for video games

Master thesis (05/04/2013) Freek Wiewel (s1494627) University of Groningen

Faculty of Economy and Business

Department of Innovation Management and Strategy Supervisors

Dr. P.M.M de Faria (first supervisor) Dr. T.L.J. Broekhuizen (second supervisor)

Acknowledgements: I wish to express my gratitude to Joost Rietveld for providing the data and his helpful comments.

(2)

2

Abstract

Demand-based research in strategy has indicated that when consumers have difficulties in gauging the use value of a product, they turn to so-called selectors for valuation clarification.

These selectors can either be actors in the market (other consumers), independent experts (e.g.

movie critics), or peers (other producers) who create value by assigning product valuations on which consumers base their decisions. Previous studies have addressed how different weight is attributed to selectors in different product markets. In this paper we test moderating factors under which different selectors have a greater impact on value creation than others within the same product market. While experts are the de facto selectors in markets for experience goods, we hypothesize and test that for products with higher network externalities market selection has greater impact on value creation. We test our hypotheses in the market for video games using a sample of 2,080 current generation console (PlayStation 3, Xbox360 and Nintendo Wii) video games released between 2005 and 2011 in the UK. We find that for games with greater network externalities user reviews have a significant and positive effect on value creation at the expense of expert selectors whose voice becomes less important.

Keywords: Strategy; innovation; demand-side; selection system theory; critic ratings;

consumer ratings; network externalities; video games.

Word count (from introduction to conclusion): 7570 words.

(3)

3

1. Introduction

Demand-side literature looks at how value is created at the consumer end of the value system (Priem, 2007; Priem, Li, & Carr, 2012). Before purchasing a product, consumers engage in a valuation process. During this process consumers attempt to determine if producers offer them enough benefits or use value to justify the exchange price asked (Bowman & Ambrosini, 2000).

The outcome of this valuation process translates into the consumers’ willingness-to-pay for the product. When a consumer’s willingness-to-pay exceeds the price asked by the producer, the consumer engages in exchange. Because value is created via the payments that consumers make to the producers in the value system (Priem, 2007), value is dependent on the judgements of consumers.

The valuation process, however, is not always easy or straightforward. Especially for experience goods such as books, movies or video games, of which the benefits only become clear after consumption, consumers might encounter difficulties in their valuation process prior to purchasing (Nelson, 1970).When consumers experience difficulties in determining the use value of a product, they turn to so-called ‘selectors’ for guidance.

Selection system theory (Wijnberg & Gemser, 2000; Wijnberg, 2004) identifies three types of selectors that guide consumers in their valuation process: market, expert and peer selectors.

While previous studies have shown how one type of selector usually prevails over the others in certain product markets, we know little about why or when one selector is preferred over the other within a single product market. In the entertainment industries for example, reviewing media (experts) are the de facto selector (Basuroy, Chatterjee, & Ravid, 2003; Eliashberg &

Shugan, 1997; Gemser, Van Oostrum, & Leenders, 2007). Yet with the advent of online communication platforms the ‘voice’ of consumers has grown louder and louder, and prospect buyers increasingly turn to other consumers for valuation purposes (De Maeyer, 2012). In this paper we aim to shed light on the conditions under which one type of selector wins the competition from the prevailing selector within a product market.

We argue that distinctive product characteristics modify the weight consumers attribute to different selectors within a product market. Per illustration, we show that the degree to which a product’s value is dependent on network externalities increases the importance of market selection over expert selection in the market for video games. Video games are a fitting example

(5)

5 to illustrate our claim as these entertainment goods are rated, reviewed and discussed in great numbers by both critics and consumers alike. We use stepwise OLS regression with log- transformed dependent variables to test our hypotheses using a sample of 2,080 console video games released in the UK between 2005 and 2011. The results show that for games with higher network externalities, favorable market selection increases its impact on value creation while the effect of expert selection conversely decreases.

2. Theory and hypotheses

2.1 Demand-side literature

In the last decades, literature in the field of management research has had a strong focus on the firm. Porter’s influential work on firm positioning (Porter, 1985) has led the way towards an inward focused view. It laid the foundation for a vast literature on generic strategies, barriers to industry competition and value chains, with the firm as the focal point of interest. The resource based view that became prominent in the 1990’s similarly takes the firm as the starting point. It looks at how firms acquire the necessary valuable, rare, inimitable and non-substitutable resources to ensure continuous revenue streams and sustain a competitive advantage (Barney, 1991). The main focus in this and related literature has been value capture at the expense of the firm’s competitors. Firms engage in a zero-sum game with their competitors and attempt to maximize their share of the value in the value system. By comparison, the origin of value receives less attention. Value is treated as exogenous and mechanisms that drive demand are generally not treated within resource side literature.

More recently, demand-side perspectives have been developed to complement the resource side focused literature (Priem, 2007; Priem, Li, & Carr, 2012). Rather than focusing on increasing the share of an individual firm, demand-side literature takes the creation of value as the starting point. It states that value creation is a precondition for value capture. The origin of value lies in payments made to the value system by consumers in exchange for benefits (Priem, 2007).

Therefore, this strand of literature looks downstream of the focal firm towards consumers and product markets (Priem et al., 2012). The value of the offerings themselves is determined by the willingness-to-pay that consumers exhibit. By increasing the benefits for consumers, additional value is created which in turn gives producers the opportunity to capture some of the created

(6)

6 value. In this light, a focus on value creation by firms is found to be equally important to firms’

success as value capture strategies (Aspara, 2012; Ramirez, 1999; Zander & Zander, 2005).

It is because value is created as consumers engage in exchanges with producers that the consumer plays a vital role as the arbiter of value in demand-side based research. Consumers assign value by engaging in a benefit maximizing valuation process in which they determine if producers offer them enough benefits to justify the exchange price asked (Bowman & Ambrosini, 2000). The value, price, cost (VPC) model (cf. Tirole, 1988: 21-34) offers a framework to illustrate this valuation process. In this model V is labeled as use value, which is the subjective valuation of the benefits a consumer expects to derive from the consumption of the good. P is labeled as the exchange value, which is the price a consumer pays for the good and represents the revenue stream to the value system. C is the production cost the producer suffers. Consumer surplus is then represented by value minus price (V-P) and producer profit is given by price minus cost (P-C).

Value capture occurs when a firm appropriates or retains part of the payments made by consumers, either by denying competitors to obtain these payments or by preventing those payments to be claimed by other actors up- or downstream in the value chain (Bowman &

Ambrosini, 2000). The capture of value revolves around the allocation of exchange value within a value system. Value is created at the consumers’ end when consumers’ willingness-to-pay increases i.e. with an increase in consumer surplus. This occurs when consumers are willing to pay for something new (new V), willing to pay for something better (increased V) or willing to buy more of something at a lower cost (decreased P). Consumers try to maximize their consumer surplus (V minus P) by evaluating the competing value offerings of producers and will proceed to buy the good that they perceive grants them the largest amount of consumer surplus, i.e. the difference between the benefits or use value received and the exchange value paid (Bowman & Ambrosini, 2000; Priem, 2007).

2.2 Experience goods

However, due to information asymmetries between producers and consumers about the use value of a value offering, it is not always easy or straight-forward for consumers to determine the value of a benefit offered by a producer. For certain types of goods it is harder to determine the offered value than for others. Two general ways of obtaining information about the use value of a good

(7)

7 are open to the consumer. Information can be obtained either via search or via experience (Nelson, 1970). The most obvious option to obtain information is via search. However, search is limited in two ways: (1) inspection of the good by the consumer must be possible and (2) the inspection must be done prior to the actual purchase. When inspection prior to purchase is not feasible, consumers defer to the second option: experience. The good is purchased while the consumer is not yet fully certain of its value and only during or after consumption will the consumer be able to ascertain the true value of the good.

In this regard, all goods and services can be placed on a continuum of search and experience attributes (Darby & Karni, 1973; Nelson, 1970; Yi-Ching Hsieh, Hung-Chang Chiu, & Mei-Yi Chiang, 2005). All goods inhibit certain attributes about which information can either be acquired via search or experience. Extending this, goods can consequently be labeled as either search or experience goods, based on the dominant attribute of the good. Search goods then are those goods/services which are dominated by search attributes of which consumers can determine the value prior to purchasing. A typical search attribute is price (Hey & McKenna, 1981) and examples of search goods are clothing and furniture (Siegel & Vitaliano, 2007).

Experience goods are those goods/services of which the value can only be determined during or after consumption, due to their dominating experience attributes. A typical experience attribute is quality (Klein, 1998) and examples of experience goods are books, movies and video games, all part of the entertainment industries (Caves, 2000).

Due to their attributes it follows that consumers have a harder time valuating experience goods.

The cost of experience is generally higher than search and as the value only becomes clear during or after consumption this leads to difficulties in the valuation process. This is even more apparent for cultural experience goods as the standards on which to ascertain the value of cultural goods are seldom clear and rarely obvious (Caves, 2000; Wijnberg & Gemser, 2000). On top of the value only becoming clear during or after consumption, this leads to additional problems in the valuation process. Uncertainty about their value is inherent to the nature of experience goods and even more so for cultural experience goods. This uncertainty is an incentive for consumers to search for credible sources of information to increase their knowledge prior to purchasing (Gemser, Leenders, & Wijnberg, 2008). In this regard, demand-side literature

(8)

8 (Priem, 2007; Priem et al., 2012) points in the direction of selection system theory when consumers have difficulties valuating goods by themselves.

2.3 Selection system theory

Selection system theory (Wijnberg & Gemser, 2000; Wijnberg, 2004) offers a framework to describe how value is determined through the selection of goods by different types of selectors.

It revolves around selection systems in which the characteristics of two types of actors are described, as well as the relationships between them. The first type of actor is called the selected, which are actors that compete with each other for recognition. The second type of actor is called the selector, which are the actors that choose among the offerings, thereby influencing the outcome of the competitive process. By selecting certain goods over others, it is the selector that determines value. Selection itself takes place based on the set of preferences that the selectors inhibit. Goods are more or less valuable proportionally to how much they match these preferences (Mol & Wijnberg, 2007). In essence, selection systems provide a description of the competitive process as selectors choose among the value offerings of producers. It describes “the way in which winners are distinguished from the losers” (Wijnberg & Gemser, 2000: 324).

Three general types of selection are described in selection system theory: market, peer and expert selection (Priem, 2007; Wijnberg & Gemser, 2000; Wijnberg, 2004). In essence, each of these types pertains to the relative weight that consumers attribute to different sources of information when choosing between products (Gemser et al., 2008). The perfectly competitive market is seen as the ideal type of market selection, where consumers are the dominant selectors and determine value. In this selection type consumers base their purchase decisions on their own and other consumers’ evaluation. In peer selection, consumers’ decisions are influenced by the evaluation of selectors that belong to the same group as the selected. The academic world, with its double- blinded peer reviewing process is a typical example of a setting where peer selection is dominant.

Expert selection takes place when value is determined by a group of actors that is neither consumer nor producer/peer. The expert group is ascribed certain specialized knowledge or skills that allows these selectors to form expert opinions and judgements which influence the outcome of the consumers’ purchase decision.

By providing credible information about experience attributes, selectors create value. They reduce the uncertainty regarding the use value of experience goods by assigning value and this

(9)

9 helps consumers make informed choices. Informed consumers are likelier to exhibit a higher willingness-to-pay which leads to the creation of additional value to be captured by producers (Bowman & Ambrosini, 2000). At the same time, informed consumers derive more benefits and satisfaction from the use of a good (De Maeyer, 2012) and in this way selectors also create value for consumers.

The three types of selection systems described here are ideal types, but selection is not by definition performed by a single selector. In reality value is often determined by a combination of multiple selection types. In some instances different selection types operate alongside each other, whereas in other cases the selection is tiered and products will need to be selected by one selector before they are considered for selection by the next set of selectors (Wijnberg, 2011). It is however possible in many markets to distinguish a dominant selection type. Previous studies have shown how certain types of selection system tends to dominate certain markets (Gemser et al., 2007; Reinstein & Snyder, 2005). Academic journals, for example, exhibit peer selection by fellow scholars as the dominant selection system (Wijnberg & Gemser, 2000).

The dominance of one type of selector is however not a given. Producers compete for recognition by selectors, but selectors are equally in competition among themselves to be the dominant value determining actor (Lampel & Shamsie, 2000; Wijnberg & Gemser, 2000). Especially in industries where selectors are the most influential, fierce competition is likely to occur to reap the benefits that come with being a dominant selector (Mol & Wijnberg, 2007). In some cases even the selected can enter into this competition to assist the selector whose preferences best match their value offerings.

The outcome of these competitive processes can invariably change industry dynamics. Wijnberg and Gemser (2000), for example, showed how the rise of Impressionist painters within the market for painted art was facilitated by a change from peer selection to expert selection within that market.

The Impressionists had much to gain from a shift in the dominant selector in this market as the preferences of the new expert selectors matched their product offerings.

In the entertainment industries, which are the focus of this thesis, both expert selection and market selection are present. (De Maeyer, 2012; Eliashberg & Shugan, 1997). Experts have a long history of being part of the entertainment industries. In its modern form, critics have reviewed cultural goods and informed and guided consumers in their purchase decisions from as early as the 18^th century (Thomas, 1978). More recently, the influence of experts on modern entertainment products has

(10)

10 been the focus in multiple studies. Many of these studies find that experts influence the performance of entertainment goods through their reviews (Basuroy et al., 2003; Eliashberg & Shugan, 1997;

Gemser et al., 2007). Experts are consequently seen as the dominant selector by many authors (Gemser et al., 2007; Gemser et al., 2008; Litman, 1983; Reinstein & Snyder, 2005; Sawhney &

Eliashberg, 1996; Zhu & Zhang, 2010).

Next to expert selection, market selection is similarly present in the entertainment industries.

Consumers have always relied on information from other consumers, so called word-of-mouth, when making purchase decisions (De Maeyer, 2012). However with the advent of the internet, the scale on which information from other consumers is available has grown exponentially. Research on the influence of market selectors has come up with mixed results. An influence of market selection on product performance in the form of user ratings has been found in several studies (Chevalier &

Mayzlin, 2006; De Maeyer, 2012; Dellarocas, Zhang, & Awad, 2007). On the other hand, several other studies find either no effect (Chen, Wu, & Yoon, 2004; Duan, Gu, & Whinston, 2008), or find that market selection in the form of user reviews merely serve as predictors of product performance (Dellarocas et al., 2007; Godes & Mayzlin, 2004). Based on previous research we expect that expert selectors are the dominant selectors in the entertainment industries. In this regard we formulate the following hypothesis:

H1: Favorable expert selection has a larger positive impact on value creation than favorable market selection in the entertainment industries.

Differences in selector dominance have been established between markets. However, little theory exists to explain why or when one type of selector is attributed more or less weight within the domain of a single market. The competition between selectors has been described (Mol &

Wijnberg, 2007; Wijnberg & Gemser, 2000), as has the resulting dominance of selectors (Gemser et al., 2007; Mol & Wijnberg, 2007; Reinstein & Snyder, 2005), but little is known about under which conditions one selector wins over the other dominant selector within single product markets. Only one other study by Zhu & Zhang (2010) finds a differential impact of market selectors within the same market based on product popularity and consumers’ internet experience.

An illustration of the competition between selectors within the scope of a single market can be found in the entertainment industries. Expert selection is prevalent in these industries, but with the rise of

(11)

11 online communication platforms, the ‘voice’ of consumers has grown louder and louder (De Maeyer, 2012). Social media, forums and websites facilitate consumers in rating, reviewing and discussing value offerings. Market selection signals are now equally abundant in the entertainment industries and are frequently consulted by prospect buyers prior to engaging in exchange (De Maeyer, 2012).

Consumers can turn to either expert or market selectors for valuation guidance, but it is unclear under which circumstances one is preferred over the other.

To offer a theoretical explanation on this matter, we argue that contingencies embedded in product characteristics modify the weight that consumers attribute to different selectors. More specifically, the degree of network externalities that a value offering inhibits is thought to increase the weight consumers attribute to market selection over expert selection.

Network externalities occur when the utility, or use value, of a product is increased with the number of other consumers of the good (Katz & Shapiro, 1985; Katz & Shapiro, 1986). This value from additional users can manifest itself both in direct and indirect ways. Direct network externalities occur when value is gained from a direct physical effect of other consumers in the network. The telephone is an often cited example where a user of a phone gains direct value from other users in the network (Artle & Averous, 1973; Kim, 2002; Liebowitz & Margolis, 1994). Indirect effects occur when additional users in the network ensure that other goods that are linked with the primary good are sufficiently supplied (Gretz, 2010). When many consumers adopt the same operating system for their personal computer, this will lead to a higher availability of compatible software as producers engage a larger potential consumer base. The availability of post-purchase services may equally be linked to the size of the network and can be seen as an indirect network externality.

Consumers derive more value from a good the larger is the amount of other consumers using the same good. In effect, there are demand-side economies of scale at work (Katz & Shapiro, 1986).

Therefore, there are advantages to simulating the choice of previous consumers; by joining existing networks consumers are able to benefit from network externality advantages (Farrell & Saloner, 1986). It is because of this that consumers include the size of the (future) network in their valuations when engaging in the value maximization process. As a larger network offers more value, the relevant network size influences the consumers’ willingness-to-pay for the product which utilizes this network. For example, when two or more goods are perfectly substitutable and only differ on the size of its network, the good with the superior network size offers the largest consumer surplus.

(12)

12 As part of the value that networked goods offer stems from other users in the network (Katz &

Shapiro, 1985; Katz & Shapiro, 1986), signals from these others are a strong indication of the value that a consumer might derive by entering the network (Caminal & Vives, 1996). Engaging in consumer-to-consumer information exchange positively impacts consumers’ loyalty intentions (Gruen, Osmonbekov, & Czaplewski, 2006), thus firsthand experiences from existing users in the network are an indication of both current and future network strength. The evaluations of other consumers therefore provide potentially richer information on both experience attributes as well as possible (future) network value than experts are able to convey. Therefore we argue that market selection gains in importance for goods with greater network externalities. Conversely we foresee a decrease in the impact of expert selectors for goods with greater network externalities as consumers shift towards market selection. We formulate the following hypotheses:

Hypothesis 2a: For products with greater network externalities within the entertainment industries, the impact of favorable market selection on value creation increases.

Hypothesis 2b: For products with greater network externalities within the entertainment industries, the impact of favorable expert selection on value creation decreases.

3. Methodology

3.1 Setting

We test our hypotheses in the market for video games. This market is well-suited for our analysis because of the specific characteristics of the industry (Broekhuizen, Lampel, & Rietveld, forthcoming). It is part of the entertainment industries and as such inhibits experience goods for which additional information, in the form of reviews prior to purchasing, helps consumers in their valuation process. Furthermore, the role that selectors play in the value determination process is generally more visible in the entertainment industries than in other industries, due to the information asymmetries between producers and consumers that exist in regards to cultural experience goods (Mol & Wijnberg, 2007). The internet offers consumers large amounts of information from both experts and fellow consumers and this allows testing for differences between market and expert selection. Furthermore, video games benefit from network externalities in the form of multiplayer possibilities. Games that offer multiplayer functionality are numerous and can easily be distinguished from single-player only games. Previous research

(13)

13 has shown that network externalities like multiplayer possibilities are an important factor in the market for video games (Venkatraman & Chi-Hyon Lee, 2004; Yang & Mai, 2010). The videogame industry is thus an excellent setting to test the influence of value offering characteristics on the weight attributed to selectors.

3.2 Database

For our analysis we make use of a provided database with secondary data. It contains performance metrics in the form of units sell-through and cumulative revenues on 8.465 videogames that were released between the year 2000 and 2011 in the United Kingdom for the sixth and seventh generation of home-consoles. From this database we draw a sample of 2080 seventh generation home-console games that were released from December 2005 (3/12/2005) up to June 2011 (25/6/2011). The cut-off point was established here as to only include videogames that were released until six months before the data ended. This is to avoid incompleteness of the data.

The sample only includes games for the current, seventh, generation of home-consoles, which are the Microsoft Xbox 360, the Nintendo Wii and the Sony PlayStation 3. Although some online functionality was present during the previous console generation, the seventh generation of home-consoles is the first generation to widely include online gameplay on all competing platforms. This makes the last generation of home-consoles best suited to test our hypotheses, as the inclusion of multiplayer functionality is a deliberate choice instead of being platform dependent. Further limitations that are placed on the sample consist of filtering out certain types of games. Releases that gather multiple earlier released titles, so called “compilations”, are excluded as these titles do not represent clear data on a single game and the earlier releases are already included in the database. Platform endorsed re-releases are also excluded to avoid the possible bias that the endorsement adds to the sales data.

3.3 Data collection

The comprehensive database including performance metrics is extended with hand collected data to facilitate the analysis. Additional data was collected in two rounds. The first round consisted of enriching the entire database by adding the expert and user score for each game, as well as the amount of reviews these scores are based on, from the website www.metacritic.com. This website accumulates reviews and scores on popular media including games, movies, television shows and music and turns them into a weighted average (Alexa.com). For videogames, 139

(14)

14 expert sources are consulted to reach the weighted score on a scale of 0-100. The user score, on a scale of 0-10, is the average of scores given by registered users of the site. The user score variable may not be as consistent as the expert review volume variable as it is the number of users that rated a game at metacritic.com. New ratings can still be added whenever a user wants to, whereas the expert reviews happen mostly within weeks of the release of a title and new ones are almost never added later on. The expert and user scores were downloaded from metacritic.com using an automated process on October 18^th 2012. This resulted in a list of 12777 sixth and seventh generation console games. The output from the script was electronically matched with the database and afterwards manually checked for missing matches.

The first round also included adding variables to the database that are used for control purposes in this study, but are projected to be used in other future research. Variables added consist of the following variables: New IP, a dummy variable which is appointed a value of 1 for games that have no predecessors and are solely based on new video game intellectual property; A sequel dummy variable which is appointed a value of 1 for sequels, prequels, or remakes of a previous game released up to 10 years before the focal release; a Media adaptation dummy variable which is appointed a value of 1 for adaptations of a popular book, movie, or television series; and a Star power dummy variable which is appointed a value of 1 for video games with a celebrity in the title or celebrity on the cover. Information for these variables was retrieved from the English version of Wikipedia.

After merging the existing database with the hand collected data described above, the second round of data collection consisted of adding data to the sample of 2080 video game titles that are used in the current study. For each title in the sample a count variable of the amount of players that each game allows to play simultaneously was added. This variable was recorded for both off- and online modes. Information concerning the amount of players was gathered by referencing the box art of each game using the search engine Google. Consequently this data was manually added to the sample. From the count variables for off-line and online play, three dummy variables were created that indicate if a video game is single-player only (local player count =1 and multiplayer count =0), whether it supports local multiplayer (local player count

=>1) and whether it supports online multiplayer (online player count =>1). When the cover for a

(15)

15 game was unavailable, the information was retrieved via www.gamespot.co.uk. Work on adding this data started on October 25^th and concluded on November 1^st.

The data collection resulted in a unique dataset of 2080 video game titles from which value creation, network externalities, market and expert selection variables are operationalized.

3.4 Dependent and independent variables

Our dependent variable, Value creation, is operationalized as the cumulative number of units sold for a focal game. Unit sell-through numbers are an objective outcome of favorable consumer valuation. Furthermore, the performance metric has considerable operational validity as industry charts are comprised using sell-through numbers rather than profits or revenues (Eliashberg, Elberse, & Leenders, 2006). To further ensure validity of our findings we use cumulative revenues as an alternative dependent variable for robustness purposes.

We operationalize our independent variables as follows. Expert selection: we operationalize favorable expert selection as the aggregated weighted critic score on a scale from 0-100. Expert ratings are used in many studies on the influence of expert selectors on product performance (Eliashberg & Shugan, 1997; Reinstein & Snyder, 2005). We control for the number of expert reviews. Market selection: numerous studies use consumer ratings as an independent variable to test for their influence on product performance (for an overview see: De Maeyer, 2012). Similar to our expert selection variable, we operationalize favorable market selection using an average user score on a scale from 0-10 while controlling for the number of user reviews. Network externalities: The degree of network externalities is represented by two variables. We construct a

‘Single player only’ dummy variable that is appointed a value of 1 for games without any multiplayer optionality. Additionally we create a count variable (local player count) measuring the amount of users that a game allows to locally play simultaneously. We use data on local player count rather than online player count as we argue that the value from playing anonymously with others online is more architectural (inherent to the game’s design) whereas the value from playing with other people locally comes in part from the social benefits of the network.

(16)

16 3.5 Controls

Performance measures of experience goods are influenced by a large number of factors (cf.

Hadida, 2009). We construct multiple control variables where our data allows us to account for additional variables that explain part of the performance measures.

Platform: In the video games market, the platform/console that a focal title is released on may influence the performance of said title (Binken & Stremersch, 2009). We include a Platform control variable in our regression model.

Genre: Genre is found to have both positive (Collins, Hand, & Snell, 2002; Hsu, 2006; Litman, 1983) and negative (Byeng-Hee Chang & Eyun-Jung Ki, 2005) effects on movie performance.

We include a Genre control variable in which games are categorized according to the genre they belong to.

Age rating: In various researches into the film industry, parental guidance ratings were found to have a significant influence on movie performance (Basuroy et al., 2003; De Vany & Walls, 2002; Hsu, 2006; Ravid, 1999). Our dataset contains similar ratings for each video game and we incorporate these ratings in our model as an Age rating control variable.

Selling price: Software price negatively affects sales (Binken & Stremersch, 2009) and we include the average selling price as a control variable in our model.

Competition: The impact of simultaneous releases of competing products is recognized by producers who regularly push back or bring forward the release of titles to avoid stronger competition where possible (Ainslie, Drèze, & Zufryden, 2005; Elberse & Eliashberg, 2003). We use a count variable consisting of the number of games that are released four weeks ahead and four weeks after the focal title is released to control for competition.

Platform age: The age of the platform/console is expected to negatively affect its attractiveness and in turn may affect software sales (Binken & Stremersch, 2009; Venkatraman & Chi-Hyon Lee, 2004). We include a Platform age control variable that consists of the time in days that passed at the time of the release of the focal game.

Seasonality: In markets with a strong seasonal pattern, the release of products often coincide with releases from competitors (Elberse & Eliashberg, 2003; Krider & Weinberg, 1998; Litman

(17)

17

& Kohl, 1989). Releases in the video game industry are frequently aligned with the holiday season, therefore we use a dummy variable to account for December releases.

Platform exclusive: Games that are exclusively released for a single platform/console differ in their performance measures from multi-platform titles (Binken & Stremersch, 2009; Prieger &

Hu, 2012). We control for exclusivity in our model by including a platform exclusive dummy variable.

Sequel: Goods that follow an existing title generally perform better than new franchises (Litman

& Kohl, 1989; Ravid, 1999). We include a dummy variable for sequels to control for this effect.

Star: The presence of celebrities or stars is found to positively affect performance in the movie industry (Basuroy et al., 2003; De Vany & Walls, 2007; Litman & Kohl, 1989). We similarly control for the cooperation of stars by constructing a dummy variable which is appointed a value of 1 if a celebrity is present in the title or on the cover of a focal game.

Media adaptation: The adaptation from previous existing media has been found to both positively (Hennig, Bauhaus, Houston, & Walsh, 2006) and negatively (Byeng-Hee Chang &

Eyun-Jung Ki, 2005) affect the performance of movies. As video games are similar in that they are experience goods, we control for media adaptations via the inclusion of a dummy variable in our model.

After formulating the dependent and independent variables and the proposed relationships between them, the analyses themselves were performed by the provider of the primary database, as full access to the database was restricted for proprietary reasons. The output of the proposed tests is found below.

4. Results

Table 1 presents the descriptive statistics and correlations for our data. The table describes data and correlations for 2080 video game titles in our sample. On average a video game sold 81,762 units and grossed £2,346,786. Critics on average rated video games with a score of 67.37 out of 100. User ratings are very similar with an average rating of 6.88 out of 10. Only 25% of the games in our sample feature no multiplayer whatsoever. This reaffirms the notion that network

(18)

18 externalities are a defining feature in this industry. The local player count (2.33) variable indicates that games which offer multiplayer for 4 players or more are in the minority. We find that the control variables are all significant with the exception of the Star variable.

We perform step wise linear regression analysis to test our hypotheses. We log-transform the dependent variable, the cumulative number of units sold, to control for the skewed distribution in performance usually found in entertainment products (cf. Hadida, 2009). The variables are standardized around their z-score after which we compute the interactions between critic score and local player count (Critic * Local player count), and user score and local player count (User

* Local player count) to reduce the risk of multicollinearity. We test four different models, adding variables in each consecutive step, the results of which are portrayed in table 2.

We enter our control variables in the first step of our regression. Next, in our second model (R- square: 45%, F-change: 155 (p < .01)), we add selection system variables. We find a positive significant effect of critic scores on value creation (p < .01). Notably, market selection has no such direct effect. With this we find support for hypothesis H1, which projected a larger positive effect of favorable expert selection over market selection. In model 3 we enter our network externality variables (R-square: 45%, F-change: 13.59 (p < .01)). While we find a significant negative effect on value creation for single-player only games (p < .01) we find no direct effect on value creation for our local player count variable. Finally, in model 4 we enter our interaction terms to test hypothesis 2a and hypothesis 2b (R-square: 46%, F-change: 9.26 (p < .01)).

Hypothesis 2a is supported as we find a significant positive effect on value creation from favorable market selection for products with greater network externalities (p < .01). Hypothesis 2b is equally supported with a significant attenuated effect on value creation from favorable expert selection for products with greater network externalities (p < .01).

VIF (< 3) and condition index values (< 5) show that multicollinearity is not an issue. The robustness of our results is underlined by running our model using an alternative dependent variable (cumulative revenue generated) which displays similar directions and significance levels for all variables (Table 3).

(19)

19

5. Discussion

Our findings confirm the relative influence of both market and expert selectors on product performance in the market for video games. Our results indicate, in line with previous theory (Priem, 2007; Priem et al., 2012; Wijnberg & Gemser, 2000), that consumers turn to selectors for valuation guidance when confronted with hard to value experience goods. Previous studies find convincing evidence for the influence of experts on product performance (Basuroy et al., 2003;

Eliashberg & Shugan, 1997; Zhang & Dellarocas, 2006) whereas research into the influence of market selectors has come up with mixed results (Chen et al., 2004; Dellarocas et al., 2007; Duan et al., 2008). Our results reaffirm this notion of expert selector dominance. In testing Hypothesis 1 (H1) we find a significant positive influence of favorable selection by critics, and no significant influence of favorable market selection. Based on this and previous research we establish that expert selectors are the de-facto selector in the market for video games.

Although value is in reality often determined via the favorable selection of a mix of selectors, a dominant selector can generally be distinguished within a market (Mol & Wijnberg, 2011). This is apparent from our first hypothesis (H1). Previous research shows that different markets inhibit different dominant selectors. For example, peer selectors are dominant in the world of academics; experts are dominant in the film industries (Gemser et al., 2008). Still, even within single markets, the influence of the dominant selector is not equally strong for all products. Our results show that the relative importance of the dominant selector is modified by defining product characteristics. We test and show that the degree to which a product’s value is dependent on network externalities modifies the weight that consumers attribute to certain selectors in the market for video games. While the degree of network externalities does not affect value creation directly in our regressions, it does influence which selector consumers turn to for valuation guidance. Consumers of goods with a higher degree of network externalities rely more on market selectors (H2a), as part of the value of the good stems from the other consumers in the network.

This shift to market selection is also found in the attenuated effect of expert selection on value creation for goods with a high degree of network externalities (H2b).

Our study is one of the first to examine the relative impact of different selectors within a single market. Only one other study by (Zhu & Zhang, 2010) finds that the importance of market selection in the form of user reviews differs within a single product market based on product

(20)

20 popularity and consumers’ internet experience. However, the relative importance of market and expert selection within a single product market has previously not been included into a single study, as we do here. Furthermore, we contribute to selection system theory. Previous literature deducted the dominant selector from explorative empirical research (Gemser et al., 2007).

Instead, we offer a theoretical explanation for the differential impact of selectors in the form of the modifying role of product characteristics.

Our findings also shed light on some of the mixed results of earlier studies into the influence of market selectors. Previous research has taken a homogenous view of products in the single market. We show that the influence of selectors is modified by defining product characteristics.

Differences on the product level may in part explain the discrepancies in the results between earlier studies.

5.1 Limitations and further research

A limitation in our research is that our data does not allow us to circumvent the problematic causal relationship between ratings and sales. Goods of high quality tend to both be in high demand and receive positive ratings. It is difficult to determine whether it is the review or the quality of the good that is responsible for the high demand. Positive bias may exist as the ratings may be correlated to unobserved product quality (Forman, Ghose, & Wiesenfeld, 2008;

Reinstein & Snyder, 2005). This bias could be removed by adding objective variables on quality;

however quality is hard to measure, especially for experience goods. Previous research has employed proxy variables (Reinstein & Snyder, 2005) or a differences-in-differences approach (Chevalier & Mayzlin, 2006) to eliminate possible bias. In our regressions, we control for the number of ratings that a focal title received; however, future research should employ additional measures for robustness purposes.

We find significant correlations between ratings and product performance, but the exact nature of the relationship could be explored further. In future research, a distinction could be made whether the effect of the ratings on demand is an influencer or predictor effect. This distinction is made by Eliashberg & Shugan (1997), when they describe the different effects that movie critics have on movie box office performance. The influencer effect is present when reviewers voice a leadership opinion on the basis of which consumers decide whether or not to engage in an exchange. The predictor effect is present when reviewers have no influence, but are seen as mere

(21)

21 predictors of product performance. Eliashberg & Shugan (1997) add time variables to review the impact of ratings as time goes by after the release to differentiate between the influence effect (impact in the first four weeks after release) and the prediction effect (impact after the first four weeks). Adding similar time variables to our dataset would be a valuable approach to further refine our results.

Our current study limits its scope to a single moderating product contingency in a single product market, namely network externalities in the market for video games. Network externalities in the form of multiplayer possibilities are especially salient in this market. Additional research could look at whether our theory holds in other markets in which network externalities are prominently present. The market for mobile smartphones is a possible avenue for additional research.

Products in this market are discussed by both experts and consumers alike and network externalities in the form of social connectivity options partly determine value.

Further research could expand upon our work by crafting and testing additional theory to further unravel which conditions explain the attributed weight to certain selectors within single product markets. We advocate a broad research agenda to map the moderating factors of selector dominance.

6. Conclusion

When faced with a choice between different products, consumers attempt to gauge the use value of each alternative to determine which option will offer them the most value (Bowman &

Ambrosini, 2000; Priem, 2007). However, when they encounter difficulties in establishing the value, which is common for experience goods such as books, films or games, consumers will turn to so-called selectors for valuation guidance (Priem, 2007; Priem, Li, & Carr, 2012). There are three types of selectors, market, expert or peer, each type referring to the information source consumers most rely on for their product choice (Gemser et al., 2008). One selector type is usually prevalent over the others, although consumers are often influenced by a mix of selectors (Wijnberg, 2011). These selectors compete with each other to become the dominant selector, especially in markets where selectors have considerable influence (Mol & Wijnberg, 2007).

(22)

22 Different selectors are found to be dominant in different markets (Gemser et al., 2007; Reinstein

& Snyder, 2005; Wijnberg & Gemser, 2000). However it is unclear whether the dominant selector is the most influential for all purchase decisions within a single market. In the entertainment industries, experts are the dominant selector, but the advent of the internet has given market selectors a louder voice and prospect buyers increasingly turn to other consumers for valuation guidance (De Maeyer, 2012). Our study reveals under which circumstances consumers turn away from experts in favor of market selectors by showing the modifying effect of distinctive product characteristics on selector influence in the market for video games.

We test and show that the relative weight that consumers attribute to market and expert selectors in the market for video games is modified by the degree of network externalities that a good inhibits. We find that when goods benefit from a high degree of network externalities, consumers are prone to pay more heed to the opinions of other consumers as they are a strong indication of the possible value that a consumer might derive by entering the network. Conversely, we find that the impact of favorable expert selection decreases for goods with greater network externalities as consumers shift towards market selection.

Our results have some practical implications. We show that the influence of selectors is modified by product characteristics. Tendencies to view markets as a whole, being dominated by a single selector, will therefore lead producers to suboptimal strategies for communicating with selectors.

In our setting producers would do well to analyze the degree of network externalities their product inhibits and identify the relevant selectors accordingly. Strategies for the cooptation of selectors will be more efficient when aimed at the selector which consumers most rely on. For games with greater network externalities this implies that producers would benefit from devising strategies for the cooptation of selectors aimed at consumers, rather than traditional strategies aimed at experts (Lampel & Shamsie, 2000). Possible strategies aimed at market selectors can take the shape of an increased presence on forums, community building and interaction via social media.

Scholars and practitioners alike will be able to benefit from more developed understanding of the circumstances under which consumers turn to different selectors for valuation guidance. Our study is a first step towards this goal of unraveling the conditions under which certain selectors win the competition from other selectors within a set market.

(23)

23

Tables and figures

VariablesMeanStd. Deviation1234567891011121314151617 1. Units¹81762199624 2. Revenues (£)¹23467866408769.989** 3. Avg. selling price (£)23.967.60.453**.578** 4. Competition31.4417.22-.050*-.088**-.261** 5. Platform age (days)961.44499.17-.162**-.170**-.150**.273** 6. Seasonality (December).04.21-.054*-.054*-0.03-0.02-.150** 7. Platform exclusive.28.45-.106**-.116**-.103**0.04-0.020.03 8. Sequel.57.50.139**.149**.133**-0.03-0.030.00-.164** 9. Star.10.31-0.03-0.04-.057**0.04-0.020.00-.230**-.087** 10. Media adaptation.22.42-.113**-.116**-.082**-0.020.000.020.04-.092**.075** 11. Critic score (0-100)67.3713.37.405**.430**.386**-.057**0.00-.063**0.01.093**-.241**-.097** 12. User score (0-10)6.881.22.217**.229**.197**-0.02-0.01-.056*.081**-0.02-.143**0.01.693** 13. Critic count2925.544**.587**.554**-.247**0.04-.099**0.000.03-.141**0.01.472**.296** 14. User count69242.328**.340**.273**-.054*0.02-0.03.067**-.057**-.112**0.04.347**.150**.463** 15. Single player only.25.43.144**.170**.229**0.000.02-0.02-.060**.138**-.219**-.134**.247**.108**.084**.049* 16. Local player count2.331.530.040.03-0.01.088**-0.020.00-0.03.188**-.056*-.501**.052*-0.04-.162**-.068**.136** 17. User * Local player count-.04.96-0.04-0.04-0.010.000.00-.062**-0.010.01.051*0.02-.093**-.080**-.064**-.075**0.03-0.04 18. Critic * Local player count.05.95-.068**-.056*0.03-0.01-0.01-.043*-0.04.104**.053*.057**-.088**-.095**-.056*-.089**.047*.065**.709** ¹ Actual values reported, log-transformed measures are used in analyses. Standardized Pearson correlations reported. **. Correlation is significant at the 0.01 level (2-tailed). 2,080 observations.

TABLE 1 Descriptive statistics and Pearson correlations

2. Theory and hypotheses

Abstract

Table of Contents

1. Introduction

2. Theory and hypotheses

3. Methodology

4. Results

5. Discussion

6. Conclusion

Tables and figures