Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations

IJRM-01091; No of Pages 16 Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Intern. J. of Research in Marketing journal homepage: www.elsevier.com/locate/ijresmar

Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations Michael Steiner a,1, Nico Wiegand b, Andreas Eggert c,d, Klaus Backhaus b,2 a

University of Witten/Herdecke, Alfred-Herrhausen-Straße 50, 58448 Witten, Germany University of Muenster, Marketing Center Muenster, Am Stadtgraben 13–15, 48143 Muenster, Germany University of Paderborn, Warburger Strasse 100, 33098 Paderborn, Germany d Newcastle University Business School, 5 Barrack Road, Newcastle upon Tyne NE1 4SE, UK b c

a r t i c l e

i n f o

Article history: First received on May 26, 2014 and was under review for 6½ months Available online xxxx Keywords: Network effects Platform adoption Systems markets Expectations management Heterogeneity Video gaming

a b s t r a c t Platform-based systems have become the dominant way to market consumer entertainment products. Video games are, for instance, distributed in digital data form, which can only be used on compatible hardware. Network effects drive the diffusion of such systems. This article provides insights into market heterogeneity and the role that expectations of the direct and indirect network effects plays in the game console market. The results of two empirical studies suggest that the console market is strongly fragmented and that the perceptions of network effects differ between the various target segments. The same holds for the importance of consumer expectations: For instance, hardcore gamers make predictions about the future software availability and incorporate these into their current adoption decision, while social gamers care more about the expected potential to interact with others. When introducing novel technologies, platform sponsors can benefit from improved targeting by, for example, providing software selectively, instead of large varieties early on. This study identifies the limits of go-to-market strategies derived from aggregate analyses when dealing with network effects and shows that behavioristic insights should complement them. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Systems provide value through the interplay between a platform and compatible complements (Lee & O'Connor, 2003). Typical examples are most entertainment products based on hardware/software architectures3 (e.g. Katz & Shapiro, 1994): A television set depends on the availability of TV channels, smartphones rely on software applications, e-book readers require digital book content, and game consoles are useless without compatible game titles. Increased digitalization has made these platform-based systems ubiquitous and has altered the market dynamics and consumption behavior globally. Consequently, their underlying economic peculiarities have received much attention (e.g. Eisenmann, Parker, & van Alstyne, 2011; Zhu & Iansiti, 2012). Our target is the home video game console industry, an often-cited canonical example of platform-based systems (Dubé, Hitsch, & Chintagunta, 2010; Corts & Lederman, 2009) and, with the E-mail addresses: [email protected] (M. Steiner), [email protected] (N. Wiegand), [email protected] (A. Eggert), [email protected] (K. Backhaus). 1 Tel.: +49 2302 926 571. 2 Tel.: +49 251 83 22861; fax: +49 251 83 22903. 3 Prior research usually refers to the platform as hardware and to complements as software (Church & Gandal, 1992; Gandal et al., 2000). We follow this convention throughout this paper.

recent release of PlayStation 4 and Xbox One, a highly topical battleground for academics and marketers. System markets have primarily been discussed against the background of network effects, i.e. the dependence of consumer value on the installed base (IB) of users (Chou & Shy, 1990; Church & Gandal, 1992). Direct network effects arise when additional network members instantly enhance a product's possible uses for an individual, as in the case of direct communication technologies (Lee & O'Connor, 2003). The more people use a communication service, the more numerous their options to interact. Indirect network effects, on the other hand, are specific to system markets: A larger IB of platform owners leads to more complementary goods (CG), which in turn positively affect customers' perceived value of the platform (Stango, 2004). Therefore, it has been suggested that indirect network effects are crucial drivers of platform diffusion and eventual success (Schilling, 1999; Shankar & Bayus, 2003). This study addresses important gaps in the literature. First, empirical research on system markets has focused on detecting the indirect network effects of various entertainment products, including DVDs (Inceoglu & Park, 2011), computer hardware and software (Frels, Shervani, & Srivastava, 2003), CDs (Basu, Mazumdar, & Raj, 2003), VCRs (Park, 2004), and home video games (Chintagunta, Nair, & Sukumar, 2009). To date, direct network effects have been rarely considered (e.g. Molina-Castillo, Munuera-Alemán, & Calantone, 2011).

http://dx.doi.org/10.1016/j.ijresmar.2015.05.011 0167-8116/© 2015 Elsevier B.V. All rights reserved.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

2

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

One reason for this is that many platform sponsors have only recently begun to integrate and promote the possibility of a direct interaction with users of the same system (e.g. in-game communication with other game console users). Second, most previous work on indirect effects focuses on aggregate market analyses, using information on hardware sales and software quantities (e.g. Basu et al., 2003; Stremersch, Tellis, Franses, & Binken, 2007). Survey-based studies that enable individual-level data analysis are scarce.4 Consequently, previous research often has two limitations: First, network effects are dealt with in an undifferentiated manner. Consumers are treated as homogeneous entities and the study results reflect the market averages. Heterogeneous preferences and perceptions of network variables are neglected, despite several scholars having identified these as a primary and fruitful research area in the network effects literature (e.g. Dubé et al., 2010; Zhu & Iansiti, 2012). Third, although discussed in theoretical terms in numerous studies (Besen & Farrell, 1994; Katz & Shapiro, 1994; Lee & O'Connor, 2003), consumer expectations have rarely been considered empirically in a network context. The few studies that incorporate expectations in their attempts to model indirect network effects do so under strong assumptions, such as naïve or rational decision making (e.g. Dubé et al., 2010; Park, 2004; Zhu & Iansiti, 2012). Only Frels et al. (2003) use survey data on the expectations of the installed base's future developments and of the availability of complements. However, this study does not explicitly assess the role of expectations, instead considering them a dimension of network strength and focusing on analyzing this aggregate measure. To the best of our knowledge, no previous study has assessed the influence of expectations on system adoption from a consumer perspective. Fourth, the initial product launch stage, which is the most competitive stage in system markets (Andreozzi, 2004; Williams, 2002), has received limited research attention. Arguably, customer expectations may play an important role in this early stage and managers need to understand whether and to what extent they impact customers' adoption intention. For example, managers could invest heavily in the software that is available right from the start. Alternatively, they might stress a console's potential to connect with other gamers. Against this background, we need more insights on how customers' direct and indirect network expectations drive their adoption intention. We contribute to extant literature by addressing these four areas within a framework of two different, yet complementary empirical studies. First, we assess consumer heterogeneity by developing a taxonomy of game console users. Employing choice-based conjoint analysis, we elicit consumer preferences and identify latent market segments. We determine four such segments, which differ substantially in their valuation of network effects variables and personal characteristics. While study 1 clarifies what variables are important for consumers' adoption intention in the console market, study 2 goes a step further and provides a process model that details how these variables work together to form consumers' adoption intention. We integrate customer expectations of the direct and indirect network effects, that could not be included in a conjoint task which requires deterministic information on the given attributes and levels, in a hierarchy-of-effects model based on Zeithaml's (1988) perceived value model. Our analysis shows that expectations of the future network effects, which strongly depend on segment affiliation, affect platform adoption. However, our results do not confirm the unconditional importance of expectations in system markets, which some researchers suggest (e.g. Shapiro & Varian, 1999; Farrell & Klemperer, 2007). Researchers can rely on two different approaches to study platform adoption in system markets: First, they can draw on existing sales data to estimate market response models. Second, they can use mindset 4 To our best knowledge, the only exceptions that use survey or experimental data on an individual level are the studies by Frels et al. (2003), Gupta et al. (1999), MolinaCastillo et al. (2011), and Song et al. (2009).

metrics, such as customer attitudes, expectations and intentions to better understand the adoption process. Both approaches have their pros and cons. While market response models have the advantage of using “hard data,” they can only be estimated when sales figures are available, i.e. after the market launch of the platform system. In contrast, models relying on customer mindset metrics may be employed before the market launch. We purposely conducted our studies before the market introduction at a time when relevant information on consoles, some of the first games, prices, etc. was discussed in the press and in public, but the network effects' influence on the future system value was still unclear. Our research shows how customer mindset metrics can inform managers and provide much needed guidance in the early stages of market introduction. We contribute to research on system markets, and especially on the video game console industry, by (1) jointly examining the direct and indirect network effects at the individual level, (2) by demonstrating that the impact of the installed base and complementary goods differs between various market segments, (3) by demonstrating that expectations play a less prominent role in adoption intentions than many researchers and practitioners believe, and (4) by demonstrating that indirect, rather than direct, network effects still drive game console adoption in approximately 85% of the market. Our findings are important for practitioners and researchers alike, because little is known about the relative importance of these network effects' elements for product adoption, as previous studies only investigate some of them. This lack of knowledge may cause biased interpretation and strategy development. Taking different customer segments into account also enables a better understanding of the market's driving forces. Moreover, analyzing the role of expectations at the customersegment level enables practitioners to better define a new games console's customer groups, who should be targeted as soon as possible, even if the console's software assortment is still limited. The remainder of this paper is organized as follows: First, through an overview of the extant literature, we discuss the direct and indirect network effects, and depict how they potentially influence entertainment platform adoption. We then develop a taxonomy of video game console users (Study 1). Study 2 is based on this taxonomy and integrates the expected direct and indirect network effects into a perceived value model to assess their influence on platform adoption. Finally, we discuss the implications of the results, provide an overview of this study's limitations, and suggest future research avenues. 2. Drivers of network effects and their influence on platform value Video game consoles are proprietary systems; that is, a single firm owns the hardware technology and competing platforms are incompatible with it (Zhu & Iansiti, 2012). Furthermore, the video game console industry does not evolve steadily, but is characterized by different hardware generations, which are also often incompatible with previous generations (Schilling, 2003). The purchase of a proprietary system limits the number of available complements to technically compatible ones. Switching systems requires additional investments in a platform (Farell & Klemperer, 2007) and new complements (here, console games). The incompatibility between platform generations resets the ‘clock’ by introducing new game console generations and starting a new war for market dominance, thus resulting in strong competitive pressure (Zhu & Iansiti, 2012). Competitive pressure is most intense in the initial stage of a new system's market introduction. Platform providers need to reach a critical mass of console adopters (the IB) to stimulate software production and foster further platform adoption (Andreozzi, 2004). Stimulating game console adoption and, subsequently, game production is of the utmost importance, because video game consoles are pure network goods; that is, the hardware itself has no stand-alone value without the complementing games (and vice versa). Consumer value is thus purely realized through network effects, i.e. from the user network (IB) and the

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

complements network (Frels et al., 2003; Lee & O'Connor, 2003; Srinivasan, Lilien, & Rangaswamy, 2004). This makes the video games market an ideal setting to study the influence of network effects on consumer platform adoption (Shankar & Bayus, 2003). System architectures increase choice complexity, because the decision to adopt requires consumers to consider at least two system components: the platform and its complements (Marchand & HennigThurau, 2013). The stronger the network effects, the greater the IB's influence on the consumer value and, ultimately, on the platform adoption (Lee & O'Connor, 2003). Although system markets have mostly been examined in terms of the indirect network effects, the industry is moving towards integrated platforms that not only function together with their complements, but also allow users to interact directly (e.g. smartphones, wearables, game consoles). We consider both types of effects. 2.1. Direct network effects Direct network effects have gained importance in the context of video game consoles, because modern consoles enable consumers to play or communicate with others (Marchand & Hennig-Thurau, 2013). The trend towards social gaming, which the seventh console generation of Nintendo's Wii triggered, is evidence of this development (Entertainment Software Association, 2012; Sun, Xie, & Cao, 2004). To date, most studies have focused on the global IB, defined as the total number of platform adopters (e.g. Economides & Himmelberg, 1995). This research assumes that each additional user adds the same marginal value increase for a platform's users (Birke, 2009). However, recent studies suggest that the direct network effects, which are based on interaction, stem largely from the local IB, i.e. an adopter's friends and acquaintances (Aggarwal & Yu, 2012; Entertainment Software Association, 2012; Suarez, 2005).5 Hence, the local IB is likely to be a major driver of direct network effects. 2.2. Indirect network effects More network participants also increase the market potential for suppliers, who react by enhancing their offers of complements such as games (Cennamo & Santalo, 2010). The supply of such complements for pure network goods is of utmost importance for consumer adoption (Binken & Stremersch, 2009; Gupta, Jain, & Sawhney, 1999) and can constitute a competitive advantage for platform sponsors (Nair, Chintagunta, & Dubé, 2004). Gallagher and Park (2002) even suggest that the platform with the strongest supply of complementary products is likely to become the market winner. The current global IB is therefore a key strategic resource for platform providers (Williams, 2002). Further, software suppliers predict future system development proactively in order to build future demand (a larger IB), and do so by offering complementary goods, although the current IB might not justify their supply. Thus, from a company perspective, the current and expected IBs are relevant when deciding on a platform's software supply (Basu et al., 2003). In other words, the present software assortment is the result of a platform's current and expected market potential. The IB is likely to influence software's variety and quality. An increase in software's sales potential stimulates developers to expand their production and attracts new market entrants. Consequently, more games are available, and the variety increases. From a consumer perspective, a greater complement of variety increases the probability of consumers finding exactly what they want (Desmeules, 2002), which allows individual users to further improve their combination of platform and 5 Data on direct communication in massive multiplayer online games (MMOGs) confirms the importance of the local IB, because communication is limited to an individual's social group. For instance, in games such as Ultima Online, a social group usually has an average of 60 gamers. Also see www.lifewithalacrity.com/2004/03/the_dunbar_numb.html.

3

complements. A higher variety also increases platform owners' flexibility, due to, for instance, dynamic preferences concerning specific games, or variety seeking (Kreps, 1979; Reibstein, Youngblood, & Fromkin, 1975). Previous research confirms that variety has a positive impact on system value and adoption (e.g. Corts & Lederman, 2009; Landsman & Stremersch, 2011). For example, Nair et al. (2004) analyze the effect of software variety on the diffusion of PDAs. Gandal, Kende, and Rob (2000) assess the influence of CD catalogue size on CD player adoption. Compared to other markets (e.g. the music industry), video game consumption lasts much longer, with games consumed in lower quantities (Entertainment Software Association, 2012; West, 2007). Variety might therefore be less influential. Furthermore, a larger market potential is likely to increase software production budgets. Combined with growing competitive pressure, a higher IB is thus also likely to drive the software quality, i.e. the average quality of the game catalogue, which – in turn – will increase the system value from a consumer perspective (Binken & Stremersch, 2009; Gallagher & Park, 2002). Surprisingly, few researchers address the influence that the complement's quality has on platform adoption. Binken and Stremersch (2009) show that, based on sales data, game variety has a significant influence on console adoption. However, once quality is included in the model, the variety effect becomes insignificant, while quality has a major impact. Gallagher and Park (2002) also propose that quality is a major driver of platform success. Other studies indicate that both variety and quality drive platform adoption (Landsman & Stremersch, 2011; Corts & Lederman, 2009). However, because most studies focus on software variety, the findings on the influence of variety and quality on platform adoption are still controversial (Marchand & Hennig-Thurau, 2013). Previous research suggests that high-quality complements show large pay-offs (Rosen, 1981). Other researchers – such as Coughlan (2001) – demonstrate the importance of quality in the games industry, in which about 5% of all video games earn 50% of all software revenues and most titles are unprofitable. Hence, quality seems to be a major driver of game purchases and also increases demand for a compatible platform (Lee, 2009). Finally, the availability of specific single high-quality titles (superstar games) is likely to foster platform adoption (Binken & Stremersch, 2009). Superstar software with disproportionally large payoffs dominate the games industry (Binken & Stremersch, 2009; Rosen, 1981). The availability of a superstar title, or its sequels, might be one of the major reasons for adopting a new game console platform, because such a superstar game is often an individual's favorite title. We thus expect the availability of favorite titles to influence platform adoption. 2.3. The impact of consumer expectations on direct and indirect network effects Most previous research has focused on realized network effects. These studies are based on the assumption that consumers choose the platform that currently provides them with the highest value (e.g. Gandal, 2002; Lee & O'Connor, 2003).6 However, other researchers suggest that the expectations of future network effects also influence adoption decisions. New complementary products constantly influence a system's value. Since software adoption is usually spread out over time (Dubé et al., 2010), consumers may try to predict the future platform evolution. 6 The current market situation is based on network effects realized in the past; that is, the past and current IB. As proposed earlier, firms also try to predict future network effects and offer complements with the expectation of increasing the future IB. However, from a consumer perspective, the reason for complements is irrelevant. Consumers might not care why firms offer a certain assortment of complementary products. Because we focus on assessing network effects from a consumer perspective, we do not consider firms' reasons for providing complementary products at any point.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

4

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Their expectations may thus become an important driver of platform adoption. Consumers will choose the platform they believe will provide them with the highest value in the future (Basu et al., 2003; Dubé et al., 2010; Marchand & Hennig-Thurau, 2013). To date, the few studies that do consider expectations assume that consumers have rational expectations, i.e. that they can perfectly predict future software prices and future software variety. In other words, consumer expectations are consistent with the market's evolution (e.g. Clements & Ohashi, 2005; Dubé et al., 2010; Gallaugher & Wang, 2002; Zhu & Iansiti, 2012). Since these studies are often based on field data, expectations cannot be measured. Instead, consumer choices are interpreted as the result of their expectations about the future system value (Dubé et al., 2010). In a similar vein, Dubé, Hitsch, and Jindal (2013) provide decision makers with expert predictions about future system prices and assume that this information enables respondents to make adoption decisions “with perfect foresight” (p.2). Therefore, they do not survey customer expectations, but assess the tradeoff between buying now or in the future. In summary, the literature is divided about the influence of expectations. Researchers take extreme viewpoints: Some ignore expectations, while others believe that consumers can perfectly predict future system development. To date, only Frels et al. (2003) have explicitly surveyed customer expectations. However, these authors do not analyze their effect on platform adoption, but develop and use an aggregate measure of network strength,7 which summarizes the impact of multiple constructs, among them consumer expectations. Against this background, we conclude that previous studies did not explore the impact of consumers' expectations on platform adoption. Furthermore, previous studies did not explicitly distinguish between the different dimensions of expectations (the expected direct and indirect network effects), whose impact on platform adoption is still unclear.8 3. Study 1: heterogeneous perceptions of network variables While standards wars in other industries have resulted in the emergence of a dominant design (e.g. the Blu-ray standard in the film industry), oligopolistic structures are more common in the games industry, because consumer preferences are heterogeneous (Entertainment Software Association, 2012; Gallagher & Park, 2002; Williams, 2002) and the platforms are positioned differently (Clement & Schollmeyer, 2009; Shankar & Bayus, 2003). Taking consumer heterogeneity into consideration is therefore key to understanding this market. However, consumer heterogeneity is rarely considered in previous research on network effects. Gupta et al. (1999) assess the context of TV format adoption (e.g. HDTV) and focus on the indirect network effects that the variety of films in a specific TV format engender. They conclude that variety of content has a significant influence on platform adoption, whose magnitude differs between market segments. Gupta et al. (1999) thus show that consumer heterogeneity should be explicitly considered when assessing network effects. However, they do not

7 Network strength is defined on the basis of the network's current system size, its expected future size, its compatibility, accessibility, and quality. 8 Extant research suggests that consumers often face difficulties forming reasonable expectations. For instance, multiple studies show that consumers fail to look ahead and are unable to predict their choices' influence on financial investments' outcomes (Benartzi & Thaler, 2007; Beshears, Choi, Laibson, & Madrian, 2011; Choi, Laibson, & Madrian, 2010). Our context and that of financial investments share the fundamental characteristic that risk and high uncertainty influence future development (Chakravarti & Xie, 2006; Landsman & Stremersch, 2011). This might limit expectations' influence on choice making, because consumers often focus on information that is most informative and which risk influences less when evaluating alternative options (Cox, 1967). Furthermore, Erdem et al. (2005) provide an extensive research overview of expectation's influence on choice. They account for expectations' importance in the context of high-technology products, and demonstrate that customers can be forward looking. However, their research overview also indicates that forward looking is limited. To date, no study has assessed the relative impact of expectations in a system products context.

consider the influence of film quality on indirect network effects. Because TV sets (in 1999) did not allow consumers to interact with each other, the authors could not consider the direct network effects. We contribute to research by considering the direct and the indirect network effects and by taking a more differentiated view of the drivers of indirect network effects. In a first step (Study 1), we seek to assess consumer heterogeneity's role in the importance of network effect drivers in a setting focused on realized network effects. In a second step (Study 2), we then analyze consumer expectations' role in game console adoption. Study 1 assesses consumer value perceptions (preferences) based on a choice-based conjoint (CBC) analysis, which is an indirect approach. In other words, the consumer value of each network effects driver is not surveyed directly, but is estimated on the basis of choices between different product alternatives representing these drivers and a no-choice option (Louviere & Woodworth, 1983). We then identify distinct target groups based on the latent class approach (DeSarbo, Wedel, Vriens, & Ramaswamy, 1992; DeSarbo, Ramaswamy, & Cohen, 1995). Latent class analyses have been shown to predict real market behavior well (e.g. Natter & Feurstein, 2002).9 3.1. Stimulus design In a first step, we defined the attributes and their levels for the CBC analysis. The attributes represent the potential drivers of network effects. We assessed the influences of game variety, game quality, and the existence of a favorite title to account for the potential indirect network effects. The direct network effects were considered by assessing the influence of the local IB.10 We also explicitly considered the prices of the hardware and the software, because previous research suggests that hardware and software prices have a major influence on platform adoption (e.g. Chintagunta et al., 2009). We defined the attribute levels on the basis of previous research recommendations. To avoid number-of-level effects, we used the same number of attribute levels for all the attributes (five levels). The only exception was the attribute favorite title available, which is a binary variable (the favorite title is available or unavailable). Thus, only two levels were surveyed. The attribute levels were selected to cover the market levels that are relevant from a consumer perspective (Eggers & Sattler, 2011; Orme, 2002). We used secondary data on product offers and in-depth interviews to define the study attributes. Table 1 provides an overview of the attributes and their levels. We generated choice sets based on a fractional design that takes maximum orthogonality (Huber & Zwerina, 1996; Lusk & Norwood, 2005) and balanced level overlap (Chrzan & Orme, 2000) into account. Each choice set consists of three stimuli and a no-choice option. The attribute order of each stimulus was randomized across the respondents to avoid attribute order effects (Kumar & Gaeth, 1991). Before surveying the respondents' preferences, we conducted a pretest to ensure that the respondents interpret attribute and level descriptions in the intended manner. These attribute and level descriptions were presented to the respondents as a warm-up task within the

9 We also used respondents' answers to estimate consumer preferences based on a hierarchical Bayes approach (Allenby & Rossi, 2006; Sawtooth Software, 2009). We aggregated these estimates of all the respondents to derive the overall value perceptions. 10 Attributes within conjoint analysis should be independent. For example, they should not overlap in meaning but represent distinct features (Orme, 2013). In the context of this study, attributes such as “software variety” and “software quality”, on the one hand, and “global installed base”, on the other hand, overlap in meaning as they represent two options for measuring indirect network effects. The global IB drives software variety and quality. Including the global IB as well as software variety and quality simultaneously, would thus result in inflated importance weights of the indirect network effects and depress the other attributes' importance weights. We included software variety and quality in our study as they have immediate effects on customer preferences while the global IB would only have an indirect effect (by stimulating software developers to offer more and better games for a platform).

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx Table 1 Attributes and levels of the stimuli. Attributes

Levels

Software variety Software quality Favorite title Local installed base20 Software price21 Hardware price22

Very low * Unavailable 0% €70 €300

Low ** Available 15% €55 €250

Medium ***

High ****

Very high *****

30% €40 €200

45% €25 €150

60% €10 €100

20 The local IB is on average 30% in Germany (Institute of Media and Consumer Research 2011). Because the local IB could be 0% for some respondents, we set the minimum to this value. To avoid imbalanced attributes, we defined a maximum value of 60%. 21 Software prices were defined on the basis of the average market price from 2005 to 2011, which was €40 (German Consumer Electronic Market Index, www.gfu.de). The maximum and minimum values were then defined on the basis Amazon prices and consumer interviews. 22 Again, the hardware prices were defined on the basis of the average market price from 2005 to 2011. We derived a value of €200 (German Consumer Electronic Market Index, www.gfu.de). The minimum and maximum values were defined on the basis of data from online comparison shopping websites.

questionnaire.11 Each respondent was asked to evaluate eight choice sets and two additional fixed holdout tasks.12 The first choice set was a holdout task and was only used for warm-up purposes (Bradlow & Rao, 2000). The last choice set represents the second holdout task and was used for validation (Orme, Alpert, & Christensen, 1997). 3.2. The taxonomy of game console users Similar to Bunn (1993), as well as Homburg, Jensen, and Krohmer (2008), we relied on the general idea of a taxonomy development procedure to identify groups of decision makers. This procedure defines clusters based on constructs, which are derived from previous research and which underlie people's buying behavior. In our context, these are the attributes used in the CBC analysis. The respondents' evaluations of these attributes and levels were then used to identify different target groups. Bunn (1993) proposes that one should survey additional variables that are likely to influence respondents' attribute evaluations in order to better understand buying decision making. Our additional variables were identified on the basis of in-depth interviews with a game console journalist, an expert from a major games producer, and 18 game console users. We identified involvement, user expertise, and consumer novelty seeking behavior as factors that are likely to influence network drivers' importance. Furthermore, we identified attitudes towards video games, subjective norms (other peers' social influence), and perceived behavioral self-control (the ability to play video games) as potential influences on preference for the CBC analysis attributes. These three behavioral intentions are the antecedents of the theory of planned behavior (Ajzen, 1991), are commonly used in the field of technology innovations, and can be interpreted as indicators of the intention to use video games platforms. We also surveyed the following demographic variables: age, gender, and occupation. The operationalization of all the measures is presented in Appendix A. 3.3. Sample and data collection We used an online questionnaire to survey our respondents' value perceptions of network effects drivers. Our survey did not consider specific game consoles, nor did we include real brands, because we were interested in assessing value perceptions that are not influenced 11 For example, the attribute “local installed base” was described in line with Aggarwal and Yu (2012) and Suarez (2005)'s definition. The local installed base comprises all of the adopter's friends and acquaintances who are potential console users, i.e. any person a respondent already knows who might play with a game console. 12 Since we focus on a stage before product introduction, there were no products for the new game console generation on the market at the time of the survey.

5

by specific brand characteristics to avoid potential biases due to halo effects. The survey was conducted in mid-2012. We hold that our study benefitted from this timing, due to the media rumors of a new game console generation,13 which meant that our respondents were aware of the topic and the network effects. We recruited game consoles users by posting survey links on game consoles websites and forums. In addition, we recruited respondents by means of Facebook and e-mails. Overall, 1928 people accessed the online questionnaire's starting page; 1030 respondents then started and completed the survey. We eliminated respondents who only selected the no-choice option in the CBC analysis, because they provided no information on their value perceptions, and thus ended up with a sample of 1008 respondents. The structure of our sample corresponds well to the market structure (for a detailed description, see Appendix B). The mean respondent age is 26.14 Most of our respondents belong to the core target group of game console users between 18 and 34 (Zhu & Iansiti, 2012). Less than 50% or our respondents are high school or university students. More males than females responded to our survey (68.9% male respondents), which fits the video game market, because male gamers remain dominant (Entertainment Software Association, 2012). 4. Results and discussion Based on the CBC evaluations, we first computed the individual partworths for each attribute level, using the hierarchical Bayes approach (Allenby & Rossi, 2006, see Appendix C for goodness-of-fit measures and validity measures). These individual estimates were used to compute the individual attribute importance weights that were then aggregated across all the respondents (Table 2). We present these aggregated values to facilitate the interpretation of segment-specific importance weights, i.e. they represent mean importances. Generally, importance weight can be interpreted as the relative influence that an attribute has on a respondent's preferences (Srinivasan, 1988). To account for consumer heterogeneity, we applied a latent class approach (DeSarbo et al., 1992; DeSarbo et al., 1995) to simultaneously identify distinct respondent groups and estimate the group-specific part-worths. We used common information criteria (AIC, CAIC, BIC, ABIC) (Andrews & Currim, 2003) and an entropy measure (EN) (Ramaswamy, DeSarbo, Reibstein, & Robinson, 1993) to identify the optimal number of segments (see Appendix D). As a result, we defined four distinct customer groups. Based on the group-specific part-worth estimates, we again computed the importance weights of our platform adoption drivers (see Table 2). Table 2 summarizes the variables that are deemed to influence the respondents' preferences (Bunn, 1993). To facilitate readability, Table 2 indicates whether or not a group-specific mean differs significantly from the mean across the other groups (Papies, Eggers, & Wlömert, 2011).15 Group 1 respondents form the largest segment (44.3%). The software-related drivers (variety, quality, and price) are the most important network effect drivers, while the existence of a favorite title is less relevant. Compared to the other groups, these respondents' preferences are fairly ‘flat’; that is, no specific network effects driver influences their preferences particularly strongly. Their expertise, interest in gaming, hedonic value derived from games, attitude to gaming, and subjective norm (negative social influence by other peers) are below average, 13 According to Google Trends, rumors about the PlayStation 4 (named Orbis at the time) already started receiving broad media coverage in May 2011, when Sony (incidentally) announced the PlayStation 4. This led to more and more information being made public. For instance, the first details of the game console graphics and processor hardware became public in February and March 2012 and most major news sites covered these. Information on the Xbox One's hardware (at the time called Xbox 720 or Durango) was revealed at about the same time and also broadly covered in the media. 14 The median age is 25 and the ages range between 14 and 77. 15 The average values of each segment are presented in Appendix E.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

6

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Table 2 Latent class results and segment characteristics. Aggregate model (based on HB estimates)

Software variety Software quality Favorite title Local diffusion Software price Hardware price Gender (% of male respondents) Console owners Covariates Age Expertise Involvement Interest Hedonic value Sign value Risk importance Risk probability Theory of planned behavior Attitude Subjective norm Behavioral control Consumer novelty seeking

Latent class analysis Group 1

Group 2

Group 3

Group 4

Casual gamers

Social gamers

Habitual gamers

Hardcore Gamers

(n = 1008; 100%)

(n = 466; 44.23%)

(n = 243; 24.11%)

(n = 147; 14.58%)

(n = 152; 15.08%)

18.27% 24.97% 14.63% 13.61% 16.24% 12.28% 68.9% 62.32%

20.46%2,3,4 23.28%2,3,4 13.32%2,3,4 11.65%2,3,4 18.72%3,4 12.57%2,3,4 64.66% 56.22%

21.49%1,3,4 15.68%1,4 10.82%1,3,4 19.52%1,3,4 19.83%3,4 12.66%1,3,4 65.43% 62.55%

13.55%1,2,4 18.60%1,4 38.55%1,2,4 9.21%1,2,4 9.45%1,2,4 10.64%1,2,4 63.95% 63.95%

12.80%1,2,3 62.19%1,2,3 4.06%1,2,3 7.32%1,2,3 8.2%1,2,3 5.43%1,2,3 88.82% 75.00%

0 −−

−− −

+++ −

0 +++

−−− −−− − +++ 0

0 −− 0 −−− +

0 0 0 0 0

+++ +++ 0 0 −

−−− −−− − −

−−− 0 −−− −−

0 0 0 0

+++ ++ +++ +++

1,2,3,4 : significantly different from the respective segment on (at least) a 5% level (one-tailed chi-squared test according to Fisher (1993) and Yates (1993). −−−/+++: p ≤ .01, −−/++: p ≤ .05, −/+: p ≤ .1, 0: no significant differences from the mean of the other segments.

while their risk is above average. In summary, group 1 respondents do not have a specific preference for any of the attributes and are less interested in gaming. We thus label this group casual gamers. Group 2 respondents pay less attention to software quality, while local diffusion is much more important compared to any of the other groups. In their view, their ability to play games (behavioral control) is weaker than those of the other groups. Risk is also less important to them. Group 2 respondents can be characterized as people who try to stay connected to their peers, and we label them social gamers. Group 3 respondents pay most attention to the availability of their favorite title (and its sequels). These respondents are older than those in the other groups. These are people who once started playing their favorite title and now continue to do so. We label this group habitual gamers. Finally, group 4 respondents' decisions are dominated by the software quality, while all the other network effect drivers are less important. Furthermore, this group of respondents, as well as their peers (subjective norm), has a positive attitude towards gaming and believes that they have the skills needed to play games. These respondents consider themselves experts, are very interested in games, derive much pleasure from gaming, and seek novel games. We therefore label this group expert gamers. Contrary to previous research on platform adoption that does not take customers' preference heterogeneity into account, our study clearly suggests that there are latent segments of consumers with different preference profiles. Identifying the latent segments of customers in the game console and software market is a valuable insight, because it demonstrates that not all customers are equal. Consequently, theoretical models that attempt to understand and describe customer behavior should embrace this complexity and develop segment-specific predictions. In particular, scholars and practitioners should be aware that market response models which ignore customer heterogeneity produce average results that may not adequately reflect the complexity of customers' reality. Finally, from a managerial perspective, Study 1 offers highly practical insights for a segment-specific marketing approach. Marketing managers in the game console and software industry can use our results to take informed targeting and positioning decisions,

and improve the effectiveness of their market communication. For example, our research provides differentiated insights into the question whether variety or quality are more likely to influence adoption (Marchand & Hennig-Thurau, 2013). Based on our results, quality is more important in three of the four segments. Furthermore, the direct network effects that emerge from the local IB are only relevant in one of the four segments. This result is in line with Kleijnen, de Ruyter, and Wetzels (2004), who find similar effects regarding mobile games. Our results also indicate that the benefits that can be derived from the ability to connect with other users are still perceived as low. However, the social gamer group size is substantial (in our study, it represents almost 25%), which might be a sufficient market size for some platform providers. In sum, study 1 provides important theoretical, methodological, and practical insights for managing and researching the game console market.

5. Study 2: a model of platform adoption with direct and indirect network effects Thus far, we have elicited the diverging influence of network effects variables on consumer choice and have defined a taxonomy of video game console users. However, choice-based conjoint analysis is based on the evaluation of current offers and is thus not suited to model consumer expectations of future network effects. Furthermore, CBC is not suitable when aiming to assess a causal chain of effects. As a result, CBC cannot answer the questions of how current marketing decisions influence expected system value, how the expected system value influences the expected direct and indirect effects, and whether consumers incorporate these expectations into their current intention to adopt a game console. Study 2 is designed to shed light on these important õresearch questions. We base our analysis on Zeithaml's (1988) perceived value model, which has been successfully applied in the context of technology adoption and can be easily expanded with elements of network effects theory. The model is based on the perceived value construct, which is central to the study of network effects. Fig. 1 shows our conceptual framework.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

7

Perceived Software Variety Perceived Hardware Price

+

+ Perceived Monetary Sacrifice

Perceived Software Price

Expected Direct Network Effect

+

+

-

+

Perceived System Value

+

+

Adoption Intention

+

+

+

+

Expected Installed Base

+ Expected Indirect Network Effect

Perceived Software Quality

+

hypothesized effect effect controlled for part of the original model by Zeithaml (1988) model extension

Fig. 1. Extended perceived value model.

The model consists of two parts: On the left, we use variables related to study 1's framework to manipulate different perceived system values.16 The value is an aggregate judgment about the current system configuration, including give components (software variety and quality) and get components (hardware prices, software prices, and monetary sacrifice). The directions of the effects are analogous to those of study 1.17 Our analysis emphasizes expectation modeling of the network effects on the right of the framework. Based on the system's value, we show the expected direct (upper path) and indirect network effects (lower path). The expected direct network effect covers the possibilities of interacting with other users of the same game platform, while the expected indirect network effect denotes the future software availability in terms of the assortment's variety and quantity. The proposed causal effects follow the rationales of network effect theory: An increase in the IB yields more possibilities to interact (direct effect), as well as more and better component availability (indirect effect), which ultimately increase the propensity to adopt. To isolate true network effects and avoid combining them with other IB effects, such as bandwagon, snob, or signaling effects (Ceci & Kain, 1982; Leibenstein, 1950; Veblen, 1899), we also include a control path. Focusing on the expectations of the future network effects, we seek to uncover whether the direct and indirect network effects mediate the system value's impact on the platform adoption. We are also interested in whether the strengths of network effects differ between study 1's and 2's taxonomy segments. 5.1. Methodology We start by describing our research design and construct measures. We then provide information on the sample characteristics and the data collection process. Finally, we assess our measurement model's reliability and validity before presenting parameter estimation results. 16 We do not aim to replicate the CBC study, but include a subgroup of factors to create a realistic setting. 17 In addition, the value components may also directly affect the expected size of the IB by serving as signals of the platform's future diffusion. For instance, the software variety may not influence a particular user perception of the system's value (because he does not care much about variety), but leads that user to believe that many people will adopt the system (because he believes they may care more about variety than he does). Therefore, we also include paths from the network effects variables and our monetary sacrifice construct to the expected IB.

5.1.1. Research design Our questionnaire was based on a scenario describing the release of a new game console. In the months prior to the release of Xbox One and PlayStation 4, this was a highly realistic and relevant situation.18 The participants were given an extract of a console test similar to those found in gaming magazines. We combined four system attributes in 16 stimuli in a 2 × 2 × 2 × 2 between-subjects design. Specifically, the console test displayed the game variety (i.e. the number of games available), game quality (approximated by nine game covers of either high-quality or low-quality games), as well as hardware and software prices, which represent the two dimensions constituting the monetary sacrifice in our model. We determined the variable levels by using assessments of the average prices and assortment variety, as well as console game quality ratings from a pre-test (n = 41). The participants were randomly assigned to one of the experimental groups. An exemplary stimulus is presented in Appendix F. Upon inspection of one of the test consoles, the participants were asked to evaluate the system configuration and to provide information on how they believed the system would evolve over time. They then had to indicate whether or not they would adopt the platform. We next discuss the measures used to capture consumers' perceptions and expectations. 5.1.2. Measures All variables were operationalized with reflective multi-item scales (Churchill, 1979; Bergkvist & Rossiter, 2007). Extant adoption literature provided reliable and valid scales for the perceived value model's basic components (i.e. prices, sacrifice, quality, value, adoption intention). We also relied on previous research for the remaining constructs, for instance, for research on direct and indirect network effects.19 We used scales that Suri and Monroe (2003) define for hardware price, software price, and the value construct. Since value was originally operationalized as a quality-to-price ratio, we modified the wording slightly and used an additional item (Purchasing this game console would be a very good deal.) to better cover our variety dimension. To measure variety, we used four items that Kahn and Wansink (2004) propose and which primarily address the various ways of enjoying an assortment's different components (e.g. Given the number of game titles, the assortment will give me a lot of variety to enjoy.). We defined software 18 19

The survey was conducted mid-2013. See Appendices A and G.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

8

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

quality, our model's other get component, as an aggregate judgment of the software catalogue's overall excellence (Zeithaml, 1988), rather than a combination of specific quality features (e.g. graphics, handling). Therefore, reflective measurement is appropriate (Jarvis, MacKenzie, & Podsakoff, 2003). To measure quality, we applied three items that Bearden, Carlson, and Hardesty (2003) propose in the context of retail price advertising. Our network effects variables (expected IB, expected direct network effect, expected indirect network effect) are based on the few individuallevel studies published to date. We built on the scales by Frels et al. (2003), supplementing them with additional items from related studies where necessary. In respect of the expected installed base, our aim was to capture the global diffusion of the platform and the local diffusion within the individual's close social environment (friends and acquaintances). The former is necessary to make assumptions about market mediation, i.e. the strength of the indirect network effects and, therefore, the software availability's future development. On the other hand, local diffusion has proven especially influential for the direct network effects' strength. In line with network effect theory, we model the expected IB as an antecedent of the expected direct network effect and the expected indirect network effect. The expected direct network effect is a fouritem construct that denotes the possibilities to interact with users of the same game console by, for instance, playing online, exchanging information, sharing games, or even interacting in person. For the expected indirect network effect, we applied a measure similar to that of Frels et al. (2003) to cover both dimensions of the software catalogue: variety and quality. We surveyed the answers regarding these constructs' dimensions separately, and modeled the expected indirect network effect as a second-order formative construct. Finally, the adoption intention was measured using three items proposed by Grewal, Monroe, and Krishnan (1998). Thus, in contrast to study 1, we now do not survey choice data, but adoption intentions. Study 2 is based on an experimental approach, i.e. the respondents evaluate one given market offering. In this situation, adoption intentions provide a more differentiated view than binary purchase/no purchase responses. Moreover, Srinivasan, Vanhuele, and Pauwels (2010) have shown that such mindset metrics are useful early signals to explain future sales. All the items were adapted to the present context of video gaming and were measured using seven-point rating scales. 5.2. Sample and data collection We gathered data via an online questionnaire in mid-2013. This timeframe was particularly suitable, because the next generation console battle was about to begin. PlayStation 4 and Xbox One had been officially announced. In our view, game consoles were a very visible and relevant topic at the time. Of the 1491 people who viewed the starting page, 721 completed the survey. We eliminated outliers and sets with a clear answering pattern, leaving us with a total of 709 usable cases and a 47.6% response rate. The sample distribution is comparable to that of the CBC study. We recruited more female participants (48.5%) than in the previous survey. The majority of respondents (55%) currently own a game console, indicating a certain minimum expertise level. More than 57% of these console owners use their platform for offline gaming only, whereas 36% play both online and offline. Finally, over 40% of all the respondents indicated that they were very knowledgeable about video gaming.20 5.3. Analysis of the measurement model We conducted exploratory factor analyses to test the unidimensionality of the proposed measurement structure (Anderson & Gerbing, 20

For a detailed sample description, see Appendix H.

1988). As a result, we dropped two items from our model due to low factor loadings. Table 3 documents the construct correlations. We first evaluated the internal consistency of the remaining 32 indicators based on their factor loadings, composite reliability (CR), and average variance extracted (AVE). The factor loadings range from .86 to .98 and are all highly significant (p b .01). With CR values between .88 and .96, and a minimum AVE of .77, the criteria exceed the commonly postulated thresholds of .6 (CR) and .5 (AVE) (Bagozzi & Yi, 1988; Boomsma, 2000). To assess the discriminant validity, we checked whether each construct's AVE was greater than its shared variance with every other construct (Fornell & Larcker, 1981). Each variable fulfills this requirement in support of discriminant validity. Because we obtained all the data for our model from the same survey, common method variance could pose a problem. We applied Harman's single-factor test (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003), as well as Widaman's (1985) procedure, which compares the original measurement structure to an extended version that includes a first-order method factor. The former reveals that a common factor only explains 35% of the variance between the indicators. To compare the models, we ran confirmatory factor analyses (CFA) of the measurement structure proposed in this study and of the structure that also contains a method factor loading on all 32 items simultaneously. The extended model fits the data well (chi2 = 1078.94, p b .01, chi2/df = 2.58, CFI = .96, TLI = .95, RMSEA = .05, SRMR = .03). However, chisquare difference tests show that adding a method factor does not improve the model fit significantly when compared to the original version (p N .1; with Satorra–Bentler correction).21 We therefore conclude that common method variance is not a serious issue in our data (Carlson & Kacmar, 2000). 5.4. Results We applied an approach similar to that of Bunn (1993) in order to define the customer groups. Based on study 1's taxonomy, we used the proposed covariates as segmentation criteria (see Table 2) and conducted a Ward cluster analysis. The resulting dendrogram suggested a two-segment or four-segment solution as reasonable alternatives. We decided to use the four-segment solution because it reduces heterogeneity within the groups, was easier to interpret, and consistent with the findings from Study 1. Based on the Ward approach, we assigned the respondents to one of the four groups. The proportions of the resulting segments were very similar to those obtained in study 1 (casual gamers: n = 300, 42.3%; social gamers: n = 166, 23.4%; hardcore gamers: n = 128, 18.1%; habitual gamers: n = 115, 16.2%). We then analyzed the measurement models of each segment separately, using the same criteria used in the aggregate sample. All the criteria were fulfilled in the subsamples. Finally, we checked whether our model suffered from collinearity problems. Since the variance inflation factors are well below the proposed threshold of 5 (Hair, Ringle, & Sarstedt, 2011; Menard, 1995), we conclude that multicollinearity is unlikely to influence the parameter estimation. We used partial least squares (PLS) to analyze our structural model (Lohmöller, 1989; Wold, 1985), because the data was moderately nonnormally distributed (Hair, Sarstedt, Ringle, & Mena, 2012). Furthermore, as we assess a rather complex model with 10 latent variables and 15 relationships, and because we extend previous research by assessing novel relationships, we decided to use PLS (Henseler, Ringle, & Sinkovics, 2009; Reinartz, Haenlein, & Henseler, 2009). The R2 of the main endogenous variables ranges from .49 (casual gamers) to .62 21 Owing to the nonnormality of the data, we used the MLR estimator in Mplus (Muthén & Muthén, 2007) to conduct difference tests that are robust against a violation of the normality assumption and against the independence of the observed variables (Satorra & Bentler, 2001).

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

9

Table 3 Correlation matrix of latent construct of the full sample (n = 709).

Software variety Software quality Software price Hardware price Monetary sacrifice System value Expected IB Expected direct effect Expected indirect effect Adoption intention

Software variety

Software quality

Software price

Hardware price

Monetary sacrifice

System value

Expected ib

Expected direct effect

Expected indirect effect

Adoption intention

1 .50 −.08– .08 .03 .40 .38 .42 .60 .40

1 −.18– .03 −.05– .50 .48 .43 .52 .54

1 .07 .66 −.42– −.22– −.13– −.16– −.25–

1 .61 −.30– −.08– .05 .07 −.13–

1 −.50– −.21– −.02– −.02– −.23–

1 .69 .44 .48 .66

1 .54 .52 .65

1 .79 .50

1 .54

1

(habitual gamers) for system value, and from .45 (social gamers) to .76 (hardcore gamers) for adoption intention. These values are in line with studies from related research, for instance, those reported for the technology acceptance model (e.g. Taylor & Todd, 1995; Venkatesh & Davis, 2000; Venkatesh, Morris, Davis, & Davis, 2003), and they confirm the model's good explanatory power. Significance tests were conducted by means of 5000 bootstrap samples (Denham, 1997). Table 4 documents the estimation results of the aggregate and the four-group solution.

The parameter estimates show that casual and social gamers tend to be more price sensitive than the remaining segments, while hardcore gamers consider quality a more important driver of value than monetary sacrifices. Software quality is important for habitual gamers, who only play specific games (likely superstar titles). Since habitual gamers focus on specific games that are important to them, the software price is less influential than the hardware price. In line with study 1, we also find that the software quality's influence on the system value is lower for the social gamers than for the remaining segments. For each

Table 4 Path analysis of the aggregate and four-group solution. Aggregate solution

System features and system value Hardware price → Monetary sacrifice Software price → Monetary sacrifice Monetary sacrifice → Perceived system value Software variety → Perceived system value Software quality → Perceived system value Monetary sacrifice → Software quality Perceived system value → Expected IB Monetary sacrifice → Expected IB Software variety → Expected IB Software quality → Expected IB Perceived system value → Adoption intention

Segmented solution Group 1

Group 2

Group 3

Casual gamers

Social gamers

Habitual gamers

Group 4 Hardcore Gamers

(n = 709; 100%)

(n = 300; 42.3%)

(n = 166; 23.4%)

(n = 128; 18.1%)

(n = 115; 16.2%)

.57***

.58***

.57***

.63***

.47***

.63***

.64***

.56***

.59***

.72***

−.49***–

−.50***–

−.55***–

−.48***–

−.37***–

.24***

.22***

.16***

.23***

.34***

.36***

.34***

.28***

.46***

.42***

−.05**–*

.00***

−.12***–

.02***

−.17***–

.66***

.62***

.63***

.55***

.79***

.12***

.14***

.20***

.03***

.02***

.04***

.03***

.11***

.02***

.02***

.14***

.16***

.02***

.11***

.11***

.36***

.38***

.36***

.36***

.36***

.52***

.50***

.38***

.60***

.72***

.52***

.44***

.44***

.55***

.72***

Direct and indirect network effects Expected IB → Expected direct effect Expected IB → Expected indirect effect Expected Direct Effect → Adoption intention → Expected indirect effect → Adoption intention

.03***

−.02***–

.19***

−.02***–

−.03***–

.19***

.15***

.16***

.22***

.28***

Control path Expected IB → Adoption intention

.29***

.30***

.18***

.37***

.33***

***: p ≤ .01, **: p ≤ .05, *: p ≤ .1

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

10

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Table 5 Qualitative synthesis of the direct adoption drivers.

Perceived system value Expected installed base Expected direct network effect Expected indirect network effect

Aggregate solution

Casual gamers

Social gamers

Habitual gamers

Hardcore gamers

Strong (.36) Medium (.29) Non-significant Modest (.19)

Average Average Average Below average

Average Below average Strongly above average Below average

Average Above average Average Slightly above average

Average Above average Average Above average

of the segments, variety exerts a weaker influence on the system value than on the software quality. Table 5 summarizes the key findings of our latent class structural equation model. It qualitatively compares the direct adoption drivers (i.e., the perceived system value, expected installed base, expected direct and indirect network effects) in the four identified segments versus the aggregate solution. The table shows that the perceived system value has a consistently strong effect on adoption intention across all four customer segments. The expected installed base has a medium effect on adoption intention in the aggregate solution, but its effect is more pronounced for habitual and hardcore gamers, while it is less important for social gamers. While the expected direct network effects have no significant impact on the adoption intention regarding the aggregate solution, they are as an important driver for social gamers. Finally, the expected indirect network effects have a relatively weak impact overall, yet it is an adoption driver of medium importance for hardcore gamers. To shed further light on the dynamic process of network effects, we followed Zhao, Lynch, and Chen's (2010), as well as Hayes, Preacher, and Myers (2010) procedure to examine whether the expected network effects mediate the system value's impact on platform adoption. We used the SPSS macro PROCESS (Hayes, 2013) for our calculations. The results are summarized in Table 6. The estimates show similar results as in our PLS model. The expected direct network effects only serve as a mediator for the social gamer group. The expected indirect network effects are strongest for the hardcore gamers, but are still significant at the 5% level for the casual and habitual gamers. In contrast to previous research's theoretical assumptions (e.g. Dubé et al., 2010), our results show that expectations of network effects are generally only of limited importance for the adoption of game platforms compared to the current system specifications. The direct effect of the system's perceived value is stronger than the expected direct or indirect network effects, and is stable across segments, making it our model's

Table 6 Mediation analysis of the four-group solution. Segment

Path estimate

Confidence interval Lower bound

Upper bound

Direct NE: Perceived System Value → Expected IB → Expected Direct Network Effects → Adoption Intention Casual gamers −.01***– −.047– .031 Social gamers .04*** .006 .191 Habitual gamers −.01***– −.065– .040 Hardcore gamers −.01***– −.085– .059 Indirect NE: Perceived System Value → Expected IB → Expected Indirect Network Effects → Adoption Intention Casual gamers .03***– .003 .072 Social gamers .03***– .005 .081 Habitual gamers .04***– .005 .129 Hardcore gamers .11***– .015 .206 ***: p ≤ .01, **: p ≤ .05, *: p ≤ .1; based on 5000 bootstrap samples path estimates are standardized with mean 0 and variance 1

central construct. Nevertheless, reducing the model to its original version without expectations of network effects leads to the R2 of the endogenous variable decreasing to .43 (Δ .11). Thus, including expectations of network effects in the perceived value model helps explain consumer adoption behavior, in particular in respect of habitual and hardcore gamers, which might reflect their superior video gaming expertise and greater judgment certainty (Chandrashekaran, Rotte, Tax, & Grewal, 2007).

6. Discussion Rapid digitalization has led to systems playing a dominant role in the entertainment industry. The video game console market has reached a considerable size and is very competitive; game console providers therefore need to achieve a critical mass of adopters as soon as possible after market introduction in order to gain a competitive advantage over their rivals (Andreozzi, 2004). Over the past two decades, many researchers have assessed the direct and indirect network effects' impact on system adoption. However, previous research has mainly relied on sales data to verify the roles of the direct and indirect network effects on system adoption at an aggregate level. While sales data is helpful for analyzing consumer decision making ex post, survey research enables managers to better understand, from a consumer perspective, network variables' impact at an individual level. Both perspectives are complementary and necessary for understanding system markets. However, few studies on network effects are based on customer surveys and many researchers have called for more survey research (e.g. Binken & Stremersch, 2009; Dubé et al., 2010; Gandal, Greenstein, & Salant, 1999; Goldenberg, Libai, & Muller, 2010; Landsman & Stremersch, 2011; Lee, 2009; Zhu & Iansiti, 2012). Our research assesses system markets from a behavioristic perspective and thus differs from most studies in the extant literature. We contribute to current knowledge by using individual-level data to better understand the roles of preference heterogeneity and consumer expectations in the process of adopting an entertainment platform (here, a game console) and in integrating the direct and the indirect network effects into the process. Based on two surveys, this study develops a taxonomy of users for the game console industry (study 1) and assesses the expected direct and indirect future network effects on system adoption at the initial stage of a new system's market introduction (study 2). This research is relevant from a theoretical perspective, because previous research takes two extreme positions. Some researcher ignore expectations (e.g. Economides & White, 1993; Gandal, 2002; Lee & O'Connor, 2003), while others suggest that consumers can perfectly predict future system evolution (e.g. Dubé et al., 2010). Our research provides a more differentiated view of the roles of the direct and indirect network effects and customer expectations. We demonstrate that expectations of future network effects differ between market segments, as does their importance for platform adoption. Hence, undifferentiated statements about these effects' strength can lead to incorrect recommendations. For instance, firms need to decide on the games assortment they will offer right from the start of a new platform. This task is of major importance as game development is extremely expensive and, due to the long development times required, a strategic decision (e.g., Lomas, 2014).

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Our findings suggest that, overall, expectations turn out to be less central to consumer decision making than assumed in prior theoretical work (e.g. Besen & Farrell, 1994; Choi, 1994; Lee & O'Connor, 2003; Park, 2004). Expectations may have a weak impact on adoption for two different reasons: (1) The respondents may not trust the reliability of their expectations and (2) they might not consider their (reliable) expectations. This points to an interesting analogy to the judgment uncertainty and magnitude parameters (JUMP) model. The JUMP model assumes that a respondent's assessment of a variable, together with the perception of uncertainty regarding this assessment, determines the stated judgments (Chandrashekaran et al., 2007). Our work cannot disentangle the different reasons for expectations' limited impact, but future studies could pursue this as a promising research avenue. From a managerial perspective, practitioners should ensure that they obtain a competitive advantage immediately on introducing a new system by providing sufficient games, as consumers put less weight on and/or trust their expectations less than many researchers believe. Platform providers that are unable to provide a large assortment of games immediately after the platform's market introduction should first focus on target groups, who (at least to some extent) incorporate expectations about future platform development into their adoption decision. Thus, preannouncing games that will become available for a platform could influence hardcore and habitual gamers' adoption decisions if only a few games are as yet available. Word-of-mouth from their peers about their intended platform adoption could influence social gamers' purchases. Thus, for instance, social media activities could foster platform adoption. However, as noted, platform providers aiming to target the whole market should ensure that they provide sufficient games right from the start. These are important findings regarding implementing marketing strategies more efficiently and fostering platform adoption. Furthermore, researcher and practitioner should distinguish between the different dimensions of expectations, because their impacts on platform adoption differ. In our model, expectations of the future direct network effects have, on average, a much weaker effect on platform adoption than expectations of the future indirect network effects have. As noted, the former are only important for social gamers and the latter only for hardcore gamers. When targeting hardcore gamers, console manufacturers should thus focus on communicating the software evolution rather than the future possibilities of interaction. The opposite is recommended when targeting social gamers. Our research also suggests that merely focusing on certain network effects, or on just one facet of network effects, will lead to incorrect recommendations. For instance, indirect network effects should not be limited to the software variety, but should also consider the perceived quality. The direct network effects should recognize the local effects to capture the growing interaction between the existing peer groups in many platform-based markets. In the present case – video gaming –, our results suggest that software quality exerts a stronger effect on platform adoption than software variety. The lucrative and loyal group of hardcore gamers even values the overall catalogue quality higher than any other network effects variable. Single high-quality games have a disproportionately strong effect on the habitual gamer group. In respect of the direct network effects, we find that the interaction possibilities do not (yet) drive most gamers' adoption decisions. However, the social gamer group appreciates the opportunity to exchange games and information online and personally with their peers. These insights can help marketers target their activities and better position their gaming platforms in the market. Our research also suggests future research fields. For instance, the control path between the expected IB and the adoption intention is fairly strong. Given a large IB of users, the path captures a mixture of effects, including possible bandwagon or trend effects, snob effects, and/or the signaling of trust and a low adoption risk

11

(Farrell & Saloner, 1985; Padmanabhan, Rajiv, & Srinivasan, 1997; Song, Parry, & Kawakami, 2009). Our results make a strong case for controlling for such psychological phenomena when seeking to isolate the direct and indirect network effects, because, if omitted, these phenomena may lead to the overestimation of the network effects. Future research is needed to unravel the different IB effects and to clarify which of the above effects have the greatest influence on platform adoption. Furthermore, our results also indicate that modelers can benefit from this study in various ways. In line with researchers such as Srinivasan et al. (2010), we urge modelers to complement their models based on “hard data” with mindset metrics. Our research indicates which mindset metrics matter and should therefore be included when conducting research on system adoption. Moreover, modelers who do not account for customer heterogeneity will estimate average effects. Our results show that such aggregate analyses might lead to the wrong managerial decisions, as network effects' importance differs between customer groups. In summary, this study provides research and practice with important insights into the role that expectations play in platform adoption. It underlines the importance of the initial games assortment as a strategic decision, because it is a major driver of platform adoption. Moreover, expectations about the future games assortment and the expected adoption intention of consumers' peers are only likely to drive some target groups' adoption decision. Expectations should therefore be managed in a target-group-specific way. Appendix A

Table A1 Operationalization. Item

Wording

Source

Expertise

• I consider myself knowledgeable about game consoles. (1) Interest • The game console I buy is extremely important to me. • I am very interested in game consoles. • Game consoles is something that leaves me quite cold. [reversed] (2) Hedonic value / pleasure • I really enjoy buying a game console. • Whenever I buy a game console, it's like giving myself a gift. • To me, game consoles are a pleasure. (3) Sign value • You can tell a lot about a person from the game console he or she uses. • The game console I buy reflects the sort of person I am. • The game console people buy says something about who they are. (4) Risk importance •It doesn't matter too much if one makes a mistake when buying a game console. •It's very irritating to buy a game console that isn't right. • I would be annoyed with myself if it turned out I had made the wrong choice when buying a game console. (5) Risk probability • When I'm in front of the game console section, I always feel unsure about what to pick. • When you buy a game console, you can never be sure whether or not it was the right choice. • Choosing a game console is pretty difficult. • When you buy a game console, you can never be quite certain about your choice.

Kleiser and Mantel (1994) Kapferer and Laurent (1985, 1993)

Involvement

(continued on next page)

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

12

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Table A1 (continued) Item

Wording

Theory of planned behavior

(1) Attitude • I believe it is a good idea to use a game console. • I think it is a foolish idea for me to use a game console. [reversed] • I like the idea of using a game console. • It would be enjoyable to use a game console. (2) Subjective norm • People who influence my behavior would think that it is advisable to use a game console. • People who influence my behavior would use a console to play games. (3) Perceived behavioral control • I would be able to use a game console to play games. • Using a game console is entirely within my control. • I have the resources, the knowledge, and the ability to use a game console. • I often seek out information about new products and brands. • I like to go to places where I am exposed to information about new products and brands. • I like magazines that introduce new brands. • I frequently look for new products and services.

Innovativeness: Consumer novelty seeking

Source

Manning, Bearden, and Madden (1995)

Note: All items were measured on a 7-point rating scale with 1 = I strongly disagree and 7 = I strongly agree

Segments Log likelihood

Table B1 Demographic sample characteristics — study 1. Frequency and percentage

Occupation

Age

Appendix D

Table D1 Latent class analysis results: number of segments.

Appendix B

Gender

In our study, we had three alternatives and a no-choice option in each choice set. Thus, the probability of randomly selecting the right alternative is ¼ = .25. A value of .704 indicates that the HB model predicts choices almost three times better than a random model. Finally, the average variance and the root mean square (RMS) also measure the model fit. The average variance denotes the average of the part-worths' variances, while the RMS indicates the root mean squares of all the part-worth estimates based on all the decision-makers. Furthermore, we tested the model's validity based on a holdout set whose alternatives were randomly defined. We compared the estimated market shares of these alternatives with observed choice making, and computed a mean absolute error (MAE) and the firstchoice hit rate to evaluate the model's validity. We obtained an MAE of 1.07, which indicates that the observed choice shares differ from the estimated choice shares by only 7%. We achieved a first choice hit rate of 56%, which is a comparable value to other studies (Bradlow & Rao, 2000; Papies et al., 2011). Furthermore, we computed a randomized first choice hit rate (RFC), which was expected to outperform traditional first choice hit rates (Orme & Baker, 2000). The RFC is .83, which indicates that the estimated model exhibits good validity.

Male Female Pupil Apprentice Student Research assistant Unemployed Clerk Self-employed Public servant Pensioner No answer

365 344 72 32 487 45 9 224 37 35 3 24

Mean and SD

51.48% 48.52% 7.4% 3.3% 50.3% 4.6% 0.9% 23.1% 3.8% 3.6% 0.3% 2.5%

1 2 3 4 5 6 7

−8689.140 −8183.278 −8001.285 −7878.063 −7809.844 −7748.936 −7686.412

AIC

CAIC

BIC

ABIC

EN

17,422.280 16,456.556 16,138.569 15,938.125 15,847.688 15,771.872 15,692.824

17,598.160 16,816.311 16,682.198 16,665.629 16,759.066 16,867.125 16,971.951

17,576.160 16,771.311 16,614.198 16,574.629 16,645.066 16,730.125 16,811.951

17,506.429 16,628.309 16,398.107 16,285.448 16,282.796 16,294.765 16,303.502

1 .757 .745 .751 .722 .711 .725

Appendix E

24

4.68

Table E1 Segment characteristics: mean values. Latent class analysis Aggregate model (based on HB estimates)

Appendix C

Table C1 Model fit. HB model Log likelihood Perc. Certainty RLH Avg. variance RMS

−2837.725 .746 .704 3.767 2.600

Percent certainty indicates how much better a model fits the data compared to a random model (Hauser, 1978). A value of .746 indicates that the model performs well above chance. Similarly, the root likelihood (RLH) denotes the predicted probabilities' geometric mean to correctly select an alternative from a set of alternatives (Papies et al., 2011).

(n = 1008; 100%) Age 25.91 Expertise 4.43 Involvement Interest 2.47 Hedonic value 3.37 Sign value 2.10 Risk importance 4.06 Risk probability 2.67 Theory of planned behavior Attitude 5.00 Subjective norm 2.51 Behavioral control 5.91 Consumer novelty 3.73 seeking

Group 1

Group 2

Group 3

Group 4

Casual gamers

Social gamers

Habitual gamers

Hardcore gamers

(n = 466; (n = 243; (n = 147; (n = 152; 44.23%) 24.11%) 14.58%) 15.08%) 25.93 4.26

25.19 4.30

27.34 4.27

25.60 5.15

2.39 3.23 2.04 4.17 2.69

2.49 3.30 2.10 3.88 2.74

2.46 3.44 2.19 4.04 2.65

2.73 3.82 2.20 4.03 2.50

4.86 2.39 5.87 3.67

4.88 2.52 5.67 3.60

5.04 2.53 5.96 3.63

5.59 2.86 6.38 4.23

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

13

Appendix F

Fig. F1. Exemplary stimulus of a console test summary.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

14

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Appendix G

Table G1 Operationalization (part I). Item

Wording

Source

Perceived software variety

•Given the number of game titles, the assortment will give me a lot of variety to enjoy. •Given the size of the game assortment, I will find at least one title I like. •The rich assortment of game titles offers many ways to enjoy it. •How much variety/comprehensiveness do you think there is in this assortment? [7-point rating scale with 1 = very little variety and 7 = very much variety] •Compared to other game consoles of a similar style, how would you rate the average quality of the game catalogue? [7-point rating scale with 1 = much lower and 7 = much higher] •The game catalogue is generally of… [7-point rating scale with 1 = very low quality and 7 = very high quality] •Overall, how would you rate the quality of the game catalogue? [7-point rating scale with 1 = very low quality and 7 = very high quality] •The displayed price of € […] for a game on this console is… [7-point rating scale with 1 = very low and 7 = very high] •I have the feeling that € […] for a game on this console is… [7-point rating scale with 1 = very cheap and 7 = very expensive] •In my opinion, the displayed price of € […] for a game on this console is… [7-point rating scale with 1 = very low and 7 = very high] •The displayed price of € […] for this game console is… [7-point rating scale with 1 = very low and 7 = very high] •I have the feeling that € […] for the game console is [7-point rating scale with 1 = very cheap and 7 = −very expensive] •In my opinion, the displayed price of € […] for this game console is … [7-point rating scale with 1 = very low and 7 = very high] •I think that, given this gaming system's attributes (console and games), it is good value for money. •At the given price, I feel that I am getting a good quality gaming system (console and games) for a reasonable price. •If I bought this game console system (console and games) at the displayed price(s), I feel I would be getting my money's worth. •Purchasing this game console system would be a very good deal.

Kahn and Wansink (2004)

Perceived software quality

Perceived software price

Perceived hardware price

Perceived system value

Bearden et al. (2003)

Suri and Monroe (2003)

Suri and Monroe (2003)

Suri and Monroe (2003)

Note: Unless otherwise indicated, all items were measured on a 7-point rating scale with 1 = I strongly disagree and 7 = I strongly agree

Table G1 Operationalization — part II. Item

Wording

Source

Expected installed base

•I expect this game console will be widely diffused in the future. •I suppose that many people will soon use this game console. •I suppose that a growing number of my friends and acquaintances will use this game console in the future. •I assume that the number of users of this console among my friends and acquaintances will increase strongly. •I expect users of this console will be able to make contact with many other users of the same console in the future. •I expect that users of this console will readily find many others to provide information on and help with the console and its titles. •I suppose that users of this console will be able to play on-line with many other users of the same console. •I suppose that users of this console will have various opportunities to exchange games or play with many other users. •I expect users of this console will be able to choose from a very attractive and large assortment of games in the future. •I expect the assortment of games will offer very good entertainment for all tastes in the future. •The probability that users will be able to choose from a broad variety of very entertaining games is… [7-point rating scale with 1 = very low and 7 = very high] •If I were going to buy a game console, the probability of buying this model is… [7-point rating scale with 1 = very low and 7 = very high] •The probability that I would consider buying this game console is… [7-point rating scale with 1 = very low and 7 = very high] •The likelihood that I would purchase this game console is…[7-point rating scale with 1 = very low and 7 = very high]

Frels et al. (2003)

Expected network strength

Expected nonnetwork strength

Adoption intention

Clements (2004); Frels et al. (2003)

Frels et al. (2003)

Grewal et al. (1998)

Note: Unless otherwise indicated, all items were measured on a 7-point rating scale with 1 = I strongly disagree and 7 = I strongly agree

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Appendix H

Table B1 Demographic sample characteristics: study 2. Frequency and percentage Gender Education

Age

Male Female Pupil Apprentice Student Research Assistant Unemployed Clerk Self-employed Public servant Pensioner No answer

365 344 8 6 530 51 7 88 9 6 2 2

Mean and SD

51.48% 48.52% 1.13% 0.85% 74.75% 7.19% 0.99% 12.41% 1.27% 0.85% 0.28% 0.28% 24

4.68

References Aggarwal, C. C., & Yu, P. S. (2012). On the Network Effect in Web 2.0 Applications. Electronic Commerce Research and Applications, 11(2), 142–151. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. Allenby, G. M., & Rossi, P. E. (2006). Hierarchical Bayes Model. In R. Grove, & M. Vriens (Eds.), The handbook of marketing research: Uses, misuses, and future advances (pp. 418–440). Thousand Oaks: Sage Publications. Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3), 411–423. Andreozzi, L. (2004). A note on critical masses, network externalities and converters. International Journal of Industrial Organization, 22(5), 647–653. Andrews, R. L., & Currim, I. S. (2003). A comparison of segment retention criteria for finite mixture logit models. Journal of Marketing Research, 40(2), 479–487. Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural equation models. Journal of the Academy of Marketing Science, 16(1), 74–94. Basu, A., Mazumdar, T., & Raj, S. P. (2003). Indirect network externality effects on product attributes. Marketing Science, 22(2), 209–221. Bearden, W. O., Carlson, J. P., & Hardesty, D. M. (2003). Using invoice price information to frame advertised prices. Journal of Business Research, 56(5), 355–366. Benartzi, S., & Thaler, R. H. (2007). Heuristics and biases in retirement savings behavior. Journal of Economic Perspectives, 21(3), 81–104. Bergkvist, L., & Rossiter, J. R. (2007). The predictive validity of multiple-item versus singleitem measures of the same constructs. Journal of Marketing Research, 44(2), 175–184. Besen, S. M., & Farrell, J. (1994). Choosing how to compete: Strategies and tactics in standardization. The Journal of Economic Perspectives, 8(2), 117–131. Beshears, J., Choi, J. J., Laibson, D., & Madrian, B. C. (2011). How does simplified disclosure affect individuals' mutual fund choices? In D. A. Wise (Ed.), Explorations in the economics of aging (pp. 75–96). Chicago: University of Chicago Press. Binken, J. L. G., & Stremersch, S. (2009). The effect of superstar software on hardware sales in system markets. Journal of Marketing, 73(2), 88–104. Birke, D. (2009). The economics of networks: A survey of the empirical literature. Journal of Economic Surveys, 23(4), 762–793. Boomsma, A. (2000). Reporting analyses of covariance structures. Structural Equation Modeling, 7(3), 461–483. Bradlow, E. T., & Rao, V. R. (2000). A hierarchical Bayes model for assortment choice. Journal of Marketing Research, 37(2), 259–268. Bunn, M. D. (1993). Taxonomy of buying decision approaches. Journal of Marketing, 57(1), 38–56. Carlson, D. S., & Kacmar, K. M. (2000). Work family conflict in the organization: Do life role values make a difference? Journal of Management, 26(5), 1031–1054. Ceci, S. J., & Kain, E. L. (1982). Jumping on the bandwagon: The impact of attitude polls on polling behaviour. Public Opinion Quarterly, 46(2), 228–242. Cennamo, C., & Santalo, J. (2010). Intraplatform competition, exclusivity, and dissimilarity strategies in the videogame industry. Working Paper. Chakravarti, A., & Xie, J. (2006). The impact of standards competition on consumers effectiveness of product information and advertising formats. Journal of Marketing Research, 43(2), 224–236. Chandrashekaran, M., Rotte, K., Tax, S. S., & Grewal, R. (2007). Satisfaction strength and customer loyalty. Journal of Marketing Research, 44(1), 153–163. Chintagunta, P. K., Nair, H. S., & Sukumar, R. (2009). Measuring marketing-mix effects in the 32/64 Bit video-game console market. Journal of Applied Econometrics, 24(3), 421–445. Choi, J. P. (1994). Network externality, compatibility choice, and planned obsolescence. Journal of Industrial Economics, 42(2), 167–182. Choi, J. J., Laibson, D., & Madrian, B. C. (2010). Why does the law of one price fail? An experiment on index mutual funds. Review of Financial Studies, 23(4), 1405–1432. Chou, C., & Shy, O. (1990). Network effects without network externalities. International Journal of Industrial Organization, 8(2), 259–270.

15

Chrzan, K., & Orme, B. K. (2000). An overview and comparison of design strategies for choice-based conjoint analysis. Sawtooth Software Research Paper Series, Sequim. Church, J., & Gandal, N. (1992). Network effects, software provision, and standardization. The Journal of Industrial Economics, 40(1), 85–103. Churchill, G. A. (1979). A paradigm for developing better measures of marketing constructs. Journal of Marketing Research, 16(1), 64–73. Clement, M., & Schollmeyer, T. (2009). Messung und Wirkung von Netzeffekten in der ökonomischen Forschung. Z. Betriebswirtsch., 58(4), 173–207. Clements, M. T. (2004). Direct and indirect network effects: Are they equivalent? International Journal of Industrial Organization, 22(5), 633–645. Clements, M. T., & Ohashi, H. (2005). Indirect network effects and the product cycle: video games in the U.S., 1994–2002. The Journal of Industrial Economics, 53(4), 515–542. Corts, K. S., & Lederman, M. (2009). Software exclusivity and the scope of indirect network effects in the U.S. home video game market. International Journal of Industrial Organization, 27(2), 121–136. Coughlan, P. J. (2001). Note on home video game technology and industry structure. HBS No. 9-700-107. Boston, MA: Harvard Business School. Cox, D. F. (1967). Risk handling in consumer behavior — An intensive study of two cases. In D. F. Cox (Ed.), Risk taking and information handling in consumer behavior (pp. 34–81). Boston: Harvard University Press. Denham, M. C. (1997). Prediction intervals in partial least squares. Journal of Chemometrics, 11(1), 39–52. DeSarbo, W. S., Ramaswamy, V., & Cohen, S. H. (1995). Market segmentation with choicebased conjoint analysis. Marketing Letters, 6(2), 137–147. DeSarbo, W. S., Wedel, M., Vriens, M., & Ramaswamy, V. (1992). Latent class metric conjoint analysis. Marketing Letters, 3(3), 273–288. Desmeules, R. (2002). The impact of variety on consumer happiness: Marketing and the tyranny of freedom. Academy of Marketing Science Review, 12, 1–18. Dubé, J. P. H., Hitsch, G. J., & Chintagunta, P. K. (2010). Tipping and concentration in markets with indirect network effects. Marketing Science, 29(2), 216–249. Dubé, J. P. H., Hitsch, G. J., & Jindal, P. (2013). The joint identification of utility and discount functions from stated choice data: An application to durable goods adoption. Working Paper. University of Chicago. Economides, N., & Himmelberg, C. (1995). Critical mass and network size with application to the US fax market (No. 95–11). Discussion Paper. New York: Stern School of Business. Economides, N., & White, L. J. (1993). One-way networks, two-way networks, compatibility and anti-trust. Working Paper. New York: Stern School of Business. Eggers, F., & Sattler, H. (2011). Preference measurement with conjoint analysis — Overview of state-of-the-art approaches and recent developments. GFK Market. Intell. Rev., 3(1), 36–47. Eisenmann, T., Parker, G., & van Alstyne, M. (2011). Platform envelopment. Strategic Management Journal, 32(12), 1270–1285. Entertainment Software Association (2012, November). The 2012 essential facts about the computer and video game industry. Retrieved from www.theesa.com Erdem, T., Srinivasan, K., Amaldoss, W., Bajari, P., Che, H., Ho, T., et al. (2005). Theory driven choice models. Marketing Letters, 16(3), 225–237. Farrell, J., & Klemperer, P. (2007). Coordination and lock-in: Competition with switching costs and network effects. In M. Armstrong, & R. Porter (Eds.), Handbook of industrial organization (pp. 1967–2072). Amsterdam: Elsevier. Farrell, J., & Saloner, G. (1985). Standardization, compatibility, and innovation. RAND J. Econ., 16(1), 70–83. Fisher, R. A. (1993). Statistical Methods for Research Workers (5th edition ). Edinburgh: Oliver and Boyd. Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39–50. Frels, J. K., Shervani, T., & Srivastava, R. K. (2003). The integrated networks model: Explaining resource allocations in network markets. Journal of Marketing, 67(1), 29–45. Gallagher, S., & Park, S. H. (2002). Innovation and competition in standard-based industries: A historical analysis of the U.S. home video game market. IEEE Transactions on Engineering Management, 49(1), 67–82. Gallaugher, J. M., & Wang, Y. M. (2002). Understanding network effects in software markets: Evidence from web server pricing. MIS Quarterly, 26(4), 303–327. Gandal, N. (2002). Compatibility, standardization, and network effects: Some policy implications. Oxford Review of Economic Policy, 18(1), 80–91. Gandal, N., Greenstein, S., & Salant, D. (1999). Adoptions and orphans in the early microcomputer market. The Journal of Industrial Economics, 47(1), 87–105. Gandal, N., Kende, M., & Rob, R. (2000). The dynamics of technological adoption in hardware/software systems: The case of compact disc players. The RAND Journal of Economics, 31(1), 43–61. Goldenberg, J., Libai, B., & Muller, E. (2010). The chilling effects of network externalities. International Journal of Research in Marketing, 27(1), 4–15. Grewal, D., Monroe, K. B., & Krishnan, R. (1998). The effects of price comparison advertising on buyers' perceptions of acquisition value, transaction value, and behavioral intentions. Journal of Marketing, 62(2), 46–59. Gupta, S., Jain, D. C., & Sawhney, M. S. (1999). Modeling the evolution of markets with indirect network externalities: An application to digital television. Marketing Science, 18(3), 396–416. Hair, J. F., Ringle, C. M., & Sarstedt, M. (2011). PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice, 19(2), 139–151. Hair, J. F., Sarstedt, M., Ringle, C. M., & Mena, J. A. (2012). An assessment of the use of partial least squares structural equation modeling in marketing research. Journal of the Academy of Marketing Science, 40(3), 414–433. Hauser, J. R. (1978). Testing and Accuracy, Usefulness, and Significance of Probabilistic Choice Models: An Information-Theoretic Approach. Operations Research, 26(3), 406–421.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011

16

M. Steiner et al. / Intern. J. of Research in Marketing xxx (2015) xxx–xxx

Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York: The Guilford Press. Hayes, A. F., Preacher, K. J., & Myers, T. A. (2010). Mediation and the estimation of indirect effects in political communication research. In E. P. Bucy, & R. L. Holbert (Eds.), Sourcebook for political communication research: Methods, measures, and analytical techniques (pp. 434–465). New York: Routledge. Henseler, J., Ringle, C. M., & Sinkovics, R. R. (2009). The use of partial least squares path modeling in international marketing. Advances in International Marketing, 20, 277–320. Homburg, C., Jensen, O., & Krohmer, H. (2008). Configurations of marketing and sales: A taxonomy. Journal of Marketing, 72(2), 133–154. Huber, J., & Zwerina, K. (1996). The importance of utility balance in efficient choice designs. Journal of Marketing Research, 33(3), 307–317. Inceoglu, F., & Park, M. (2011). Diffusion of a new product under network effects: The US DVD market. Applied Economics, 43(30), 4803–4815. Institute of Media and Consumer Research (2011). Communication Networks: Basisdaten für Kommunikationsstrategien in der modernen Wissensgesellschaft. Published online at: www.imuk.de (accessed October 2012) Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model misspecification in marketing and consumer research. Journal of Consumer Research, 30(2), 199–218. Kahn, B. E., & Wansink, B. (2004). The influence of assortment structure on perceived variety and consumption quantities. Journal of Consumer Research, 30(4), 519–533. Kapferer, J. N., & Laurent, G. (1985). Consumer Involvement Profiles: A new practical Approach to Consumer Involvement. Journal of Advertising Research, 25(6), 48–56. Kapferer, J. N., & Laurent, G. (1993). Further Evidence on the Consumer Involvement Profile: Five Antecedents of Involvement. Psychology & Marketing, 19(4), 347–355. Katz, M. L., & Shapiro, C. (1994). Systems competition and network effects. Journal of Economic Perspectives, 8(2), 93–115. Kleijnen, M., de Ruyter, K., & Wetzels, M. (2004). Consumer adoption of wireless service: Discovering the rules while playing the game. Journal of Interactive Marketing, 18(2), 51–61. Kleiser, S. B., & Mantel, S. P. (1994). The Dimensions of Consumer Expertise: A Scale Development. In R. Achrol, & A. Mitchell (Eds.), AMA Summer Educators' Proceedings. Vol. 5. (pp. 20–26). Chicago: American Marketing Association. Kreps, D. M. (1979). A representation theorem for preference for flexibility. Econometrica, 47(3), 565–577. Kumar, V., & Gaeth, G. J. (1991). Attribute order and product familiarity effects in decision tasks using conjoint analysis. International Journal of Research in Marketing, 8(2), 113–124. Landsman, V., & Stremersch, S. (2011). Multihoming in two-sided markets: An empirical inquiry in the video game console industry. Journal of Marketing, 75(6), 39–54. Lee, R. (2009). Blockbuster software titles and indirect network effects in the home video game industry. Working Paper. Lee, Y., & O'Connor, G. C. (2003). New product launch strategy for network effects products. Journal of the Academy of Marketing Science, 31(3), 241–255. Leibenstein, H. (1950). Bandwagon, snob, and veblen effects in the theory of consumers' demand. The Quarterly Journal of Economics, 64(2), 183–207. Lohmöller, J. B. (1989). Latent variable path modeling with partial least squares, Heidelberg. Lomas, Natasha (2014). The console market is in crisis. Retrieved from http://techcrunch. com/2014/03/09/console-crisis/ Louviere, J. J., & Woodworth, G. (1983). Design and analysis of simulated consumer choice or allocation experiments: An approach based on aggregate data. Journal of Marketing Research, 20(4), 350–367. Lusk, J. L., & Norwood, F. B. (2005). Experimental design on choice-based conjoint valuation estimates. American Journal of Agricultural Economics, 87(3), 771–785. Manning, K. C., Bearden, W. o., & Madden, T. J. (1995). Consumer Innovativeness and the Adoption Process. Journal of Consumer Psychology, 4(4), 329–345. Marchand, A., & Hennig-Thurau, T. (2013). Value creation in the video games industry: Industry economics, consumer benefits, and research opportunitites. Journal of Interactive Marketing, 27(3), 141–157. Menard, S. (1995). Applied logistic regression analysis. Paper Nr. 106. Thousand Oaks: Sage-University. Molina-Castillo, F. J., Munuera-Alemán, J. L., & Calantone, R. J. (2011). Product quality and new product performance: The role of network externalities and switching costs. Journal of Product Innovation Management, 28(6), 915–929. Muthén, L. K., & Muthén, B. O. (2007). Mplus User's Guide (5th ed.). Los Angeles, CA: Research Publishers. Nair, H. S., Chintagunta, P. K., & Dubé, J. P. (2004). Empirical analysis of indirect network effects in the market for personal digital assistants. Quantitative Marketing and Economics, 2(1), 23–58. Natter, M., & Feurstein, M. (2002). Real world performance of choice-based conjoint models. European Journal of Operational Research, 137(2), 448–458. Orme, B. (2002). Formulating attributes and levels in conjoint analysis. Sawtooth software research paper series, Sequim. Orme, B. (2013). Getting started with conjoint analysis (2nd ed.). Madison: Research Publishers. Orme, B. K., & Baker, G. C. (2000). Comparing Hierarchical Bayes Draws and Randomized First Choice for Conjoint Simulations. Sequim: Sawtooth Software Research.

Orme, B. K., Alpert, M. I., & Christensen, E. (1997). Assessing the validity of conjoint analysis — continued. Sawtooth software research paper series, Sequim. Padmanabhan, V., Rajiv, S., & Srinivasan, K. (1997). New products, upgrades, and new releases: A rationale for sequential product introduction. Journal of Marketing Research, 34(4), 456–472. Papies, D., Eggers, F., & Wlömert, N. (2011). Music for free? How free ad funded downloads affect consumer choice. Journal of the Academy of Marketing Science, 19(5), 777–794. Park, S. (2004). Quantitative analysis of network externalities in competing technologies: The VCR case. The Review of Economics and Statistics, 86(4), 937–945. Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903. Ramaswamy, V., DeSarbo, W. S., Reibstein, D. J., & Robinson, W. T. (1993). An empirical pooling approach for estimating marketing mix elasticities with PIMS data. Marketing Science, 12(1), 103–124. Reibstein, D. J., Youngblood, S. A., & Fromkin, H. L. (1975). Number of choices and perceived decision freedom as a determinant of satisfactions and consumer behavior. Journal of Applied Psychology, 60(4), 434–437. Reinartz, W., Haenlein, M., & Henseler, J. (2009). An empirical comparison of the efficacy of covariance-based and variance-based SEM. International Journal of Research in Marketing, 26(4), 332–344. Rosen, S. (1981). The economics of superstars. American Economic Review, 71(5), 845–858. Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for moment structure analysis. Psychometrika, 66(4), 507–517. Sawtooth Software (2009). The CBC/HB system for hierarchical Bayes estimation. Sawtooth software technical paper series, Sequim. Schilling, M. A. (1999). Winning the standards race: Building installed base and the availability of complementary goods. European Management Journal, 17(3), 265–274. Schilling, M. A. (2003). Technological leapfrogging: Lessons from the US video game console industry. California Management Review, 45(3), 6–30. Shankar, V., & Bayus, B. L. (2003). Network effects and competition: An empirical analysis of the home video game industry. Strategic Management Journal, 24(4), 375–384. Shapiro, C., & Varian, H. R. (1999). The art of standard wars. California Management Review, 41(2), 8–32. Song, M., Parry, M. E., & Kawakami, T. (2009). Incorporating network externalities into the technology acceptance model. The Journal of Product Innovation Management, 26(3), 291–307. Srinivasan, V. (1988). A conjunctive-compensatory approach to the self-explication of multiattributed preferences. Decision Sciences, 19(2), 295–305. Srinivasan, R., Lilien, G. L., & Rangaswamy, A. (2004). First in, first out? The effects of network externalities on pioneer survival. Journal of Marketing, 68(1), 41–58. Srinivasan, S., Vanhuele, M., & Pauwels, K. (2010). Mind-set metrics in market response models: An integrative approach. Journal of Marketing Research, 47(4), 672–684. Stango, V. (2004). The economics of standards wars. Review of Network Economics, 3(1), 1–19. Stremersch, S., Tellis, G. J., Franses, P. H., & Binken, J. L. G. (2007). Indirect network effects in new product growth. Journal of Marketing, 71(3), 52–74. Suarez, F. F. (2005). Network effects revisited: The role of strong ties in technology selection. Academy of Management Journal, 48(4), 710–720. Sun, B., Xie, J., & Cao, H. H. (2004). Product strategy for innovators in markets with network effects. Marketing Science, 23(2), 243–254. Suri, R., & Monroe, K. B. (2003). The effects of time constraints on consumers judgments of prices and products. Journal of Consumer Research, 30(1), 92–104. Taylor, S., & Todd, P. A. (1995). Understanding information technology usage: A test of competing models. Information Systems Research, 6(2), 144–176. Veblen, T. B. (1899). The theory of the leisure class. London: Oxford University Press. Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2), 186–204. Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478. West, S. (2007). Does size matter: Video game length. Retrieved from www. cinemablend.com Widaman, K. F. (1985). Hierarchically nested covariance structure models for multitraitmultimethod data. Applied Psychological Measurement, 9(1), 1–26. Williams, D. (2002). Structure and competition in the U.S. home video game industry. Int. J. Media Manag., 4(1), 41–54. Wold, H. (1985). Partial least squares. In S. Kotz, & N. L. Johnson (Eds.), Encyclopedia of Statistical Sciences, Vol. 6. (pp. 581–591). New York: Wiley. Yates, F. (1934). Contingency Tables involving small Numbers and the Qui-Squared Test. Supplement to the Journal of the Royal Society, 1(2), 217–235. Zeithaml, V. A. (1988). Consumer perceptions of price, quality, and value: A means–end model and synthesis of evidence. The Journal of Marketing, 52(3), 2–22. Zhao, X., Lynch, J. G., & Chen, Q. (2010). Reconsidering Baron and Kenny myths and truths about mediation analysis. The Journal of Consumer Research, 37(2), 197–206. Zhu, F., & Iansiti, M. (2012). Entry into platform-based markets. Strategic Management Journal, 33(1), 88–106.

Please cite this article as: Steiner, M., et al., Platform adoption in system markets: The roles of preference heterogeneity and consumer expectations, Intern. J. of Research in Marketing (2015), http://dx.doi.org/10.1016/j.ijresmar.2015.05.011