Software multihoming to distal markets: Evidence of cannibalization and complementarity in the video game console industry

Steady software supply is a crucial driver of platform sales. While publishers benefit from releasing software across multiple platforms to tap a greater market, platform manufacturers often seek exclusive release to differentiate from competitors. Research has examined such software multihoming across competing platforms of the same technology generation (i.e., the proximal market); however, publishers increasingly multihome software to platforms in distal markets. In the video game console industry, these include previous-generation consoles, handhelds, or mobile devices. This study investigates multihoming to distal markets in the seventh and eighth game console generations. Whereas multihoming to previous-generation consoles cannibalizes focal console sales, multihoming to mobile devices exerts complementary effects. Software quality and console age moderate these relationships, with negative spillovers from multihoming to previous-generation consoles being rooted in lower-quality games and games released later in the console’s lifecycle. By contrast, multihoming to mobile devices is most beneficial early on.


Introduction
The economic success of digital entertainment platforms, from music streaming to video gaming, hinge on a steady supply of enticing software. New releases of movies and series are vital to attract customers to Netflix and Disney+ and releasing new books is key for the sales of Amazon's Kindle. In the video game console industry, game (i.e., software) releases increase the value of a corresponding console (i.e., platform) to consumers and thereby its sales (Landsman & Stremersch, 2011). Owing to reduced porting costs, publishers increasingly multihome software, that is, they release game titles for more than one platform to tap a greater market (Corts & Lederman, 2009). Today, about 90% of all video games are available for more than one console of the same technology generation (e.g., PlayStation 4 and Xbox One), giving consumers multiple options to play them (Swider et al., 2022). For console manufacturers, whose perspective we adopt, research has shown that such multihoming to direct competitors reduces the benefits of a game release, as it lessens the extent to which the game helps differentiate a focal console (Coughlan, 2004;Lee, 2013).
However, firms not only release software to same-generation game consoles, which we label proximal multihoming, but increasingly also to platforms in other related markets-a phenomenon we term distal multihoming (see Table 1) and that extant research has not examined. For example, many console games are also released for mobile devices, such as smartphones, because this booming market promises additional revenues. Similarly, publishers often multihome games to consoles from the previous technology generation as well as handheld devices. Yet, it is unclear how such distal multihoming, which allows consumers to play a particular game on platforms other than the latest-generation game console, may affect the influence of that game's release on focal, latest-generation console sales. Such insights would be valuable to the video gaming industry and could also inform platform strategy in other markets where software content drives platform adoption (e.g., entertainment streaming).
A scarce empirical literature has examined multihoming limited to the proximal market (e.g., Corts & Lederman, 2009;Landsman & Stremersch, 2011), showing that singlehoming (i.e., the release of software exclusively for one platform) drives platform sales. However, no study has investigated effects of multihoming software to distal markets on focal platform sales, 1 even though understanding such spillovers is crucial for platform launch and lifecycle management as well as negotiations with software publishers. For example, for PlayStation 4 as the focal console, it matters whether multihoming a game to mobile devices inhibits its sales and market share because consumers can play the game on smartphones they already own; or whether it strengthens console sales by giving consumers a taste of the game, inducing them to purchase the console for a superior gaming experience. Similarly, while multihoming to previous-generation consoles may foster consumer inertia to upgrade and thus hurt focal console sales, it might also motivate consumers to purchase the new, technologically superior console.
Current industry practice provides evidence that these questions are far from trivial, as Sony, Microsoft, and Nintendo have pursued remarkably different multihoming strategies in the past. For instance, while Microsoft initially pursued multihoming of all newly released Xbox Series X games to its predecessor Xbox One (Higham, 2020), Sony largely prevented multihoming to other consoles from the previous generation and instead focused on differentiation through singlehoming. Approaches to handheld gaming are another case in point: Contrary to its competitors Sony and Nintendo, Microsoft never entered the handheld device market, thereby actively forgoing potential positive spillovers from software multihoming, but also avoiding potential negative effects. To date, no commonly accepted best practice to multihoming software to distal-market platforms has emerged in the industry.
In this study, we extend the prevalent, narrow conceptualization of multihoming confined to the proximal market to account for software releases on platforms in distal markets. We draw from hedonic consumption theory (Hirschman, 1992;Hirschman & Holbrook, 1982), particularly, immersive hedonic experiences, to explain spillover effects from software releases for distal-market platforms on focal console sales. We propose that substitutability between platforms in the proximal market is high, since all consoles provide a similarly immersive gaming experience (Hennig-Thurau & Houston, 2019). For instance, playing the soccer simulation FIFA on PlayStation 4 provides a similarly immersive experience as playing it on Xbox One, as graphics, audio, and controllers are generally of comparable quality. By contrast, substitutability varies between the focal console and different platforms in distal markets that may offer substantially less immersive gaming experiences. As such, playing FIFA on a mobile device provides much weaker visual and aural effects than the focal console as well as touch-screen gameplay instead of controllers, creating a substantially less immersive gaming experience (see Web Appendix A). To capture such differences, we define the term immersion gap as the disparity in the immersion of the gaming experience that different platforms provide. In this way, the immersion gap is also a measure of substitutability between platforms, as it renders distal-market platforms only imperfect substitutes to the focal console. Prior work on proximal multihoming, where substitutability between platforms is high, thus provides only little guidance to understand how multihoming to distal markets affects focal console sales. In addition, the immersion gap between the focal console and a distal-market platform might be sensitive to software and console characteristics, increasing in software quality and decreasing in console age. We therefore investigate these two characteristics as contingencies to the effects of distal multihoming.
We analyze an extensive dataset of monthly console sales and over 8000 game releases in the seventh and eighth console generations (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) in the United States. Results show that multihoming software to mobile devices strengthens the effect of a game release on focal console sales, giving managers a powerful tool to boost console diffusion. This occurs irrespective of a game's quality and mostly early in a console's lifecycle. By contrast, multihoming to previousgeneration consoles cannibalizes the impact of a new game release. These effects are rooted in lower-quality games and predominately affect the second half of a console's lifecycle. Firms are thus well-advised to prevent prolonged multihoming of games to previous-generation consoles. For multihoming to handhelds cannibalizing spillovers only materialize toward the end of the console's lifecycle. These results underline the importance of adopting a more nuanced view of software multihoming than put forth in prior literature to effectively manage platform sales.

Relevant literature
Our study relates to broader literature on network effects and competition in two-sided markets as well as industry-specific research on hardware-software platforms (Table 2). In twosided markets, an intermediary platform brings together two types of economic agents to facilitate interactions between them (e.g., Armstrong, 2006;Rochet & Tirole, 2003). On either side, agents may choose to connect to more than one platform-that is, to multihome. Research has examined multihoming, for instance, in the context of pricing strategies (e.g., Armstrong, 2006;Caillaud & Jullien, 2003;Hagiu, 2009), advertising markets (Athey et al., 2018), cross-subsidization (Bakos & Halaburda, 2020), or market tipping (Hossain & Morgan, 2013).
Several empirical studies in marketing, information systems, and economics focus on proximal multihoming in the video gaming industry. For example, Corts and Lederman (2009) show that software multihoming to same-generation game consoles induces cross-console indirect network effects, where the combined installed base of all consoles increases supply of multihomed games. Moreover, games multihomed in the proximal market increase focal console sales but less so than singlehomed releases. Lee (2013) shows proximal multihoming to exert expansion effects on software and hardware markets. Cennamo et al. (2018) highlight that multihomed games exhibit lower-quality performance on technologically more complex consoles. Haviv et al. (2020) exploit multihoming to study intertemporal demand spillovers between video games. They find that a 1% increase in game copies sold on a console increases sales of other games on that console by .153%.
Especially related to our research is moreover the work of Landsman and Stremersch (2011) who show that higher sales shares of singlehomed video games, as compared to proximally multihomed games, increase focal console sales. These effects occur mostly for consoles that are nascent and have lower market share. We differ from this work in three key ways. First, while Landsman and Stremersch compare singlehoming to multihoming in the proximal market, we focus on the effects of different forms of distal multihoming. Second, the authors measure singlehoming based on the catalog of all available games for a given console weighted by the games' sales. In this way, they implicitly account for the effects of high-quality games which make up the majority of game sales. By contrast, we apply no such weighting and instead focus on the effects of new game releases to capture the impact of different release regimes irrespective of game quality. Third, we then examine software quality and console age as relevant moderators to the effects of all software release regimes. Using software quality as a moderator, we show that Landsman and Stremersch's results represent a special case to our generalized (i.e., unweighted) model of software multihoming.
Altogether, although extant work provides important insights into the effects of proximal multihoming, we are the first to investigate multihoming to qualitatively different, distal-market platforms (i.e., previous-generation consoles, handheld devices, mobile devices). In addition, we examine game quality and console age as relevant moderators of the effects of such release regimes. We specify a model that uses software release shares to capture the effects of multihoming software to distal-market platforms and distinguish between market stealing and category expansion as their underlying mechanisms. Next, we briefly introduce the game publishing process and the parties involved. We then introduce our conceptual framework.

Software publishing and multihoming in the game console industry
The software development and publishing process involves several parties, most notably platform manufacturers, game developers, and publishers. We focus on console games, as the console market is at the center of this study Market outcome denotes the situation of the involved parties in equilibrium with respect to variables such as revenues, profits, market shares, etc.; Matchmakers denote two-sided markets with direct network effects, which are not characterized by complementary goods as in hardware-software systems. Examples are nightclubs or dating services (we refer to agents involved in matchmaking as parties instead of buyers and sellers); The type of multihoming refers to either customers using different platforms concurrently (user multihoming) or software suppliers offering their products on different platforms (seller multihoming). Studies most closely related to this research are in bold and next provide background information based on expert interviews 2 as well as practitioner publications. Video game consoles source software in three ways: in-house development and publishing (so-called "first-party" releases), contractual agreements with external developers and/or publishers (sometimes referred to as "secondparty" deals), and independent releases (commonly termed "third-party" releases). The console manufacturer's creative and commercial control over the software development and publishing process decreases in this order. While console manufacturers Sony, Microsoft, and Nintendo, are among the largest providers of console games through in-house publishing, second-and thirdparty game development and publishing dominate the market. For example, in 2020, roughly 80% of software revenues on PlayStation 4 came from independent publishers (Mansoor, 2021). These publishers control the decision for which platform(s) to release their games. As a classic example of network effects, hardware and software firms feed off of each other, as games increase console demand, and a larger installed base of consoles drives software sales (Stremersch et al., 2007).
Console manufacturers favor releasing first-party software exclusively for their own platform to differentiate from competitors (Corts & Lederman, 2009). For this, they may even go so far as to acquire game developers such as Microsoft's acquisition of Double Fine Productions and Sony's takeover of Insomniac Games (Lee, 2019). Furthermore, manufacturers enter contracts with publishers to prevent them from multihoming games to other consoles (Mansoor, 2021). These could be outright singlehoming agreements, typically in exchange for a substantial fee, or involve forms of collaboration such as advertising alliances granting exclusive promotion rights to the manufacturer. Such agreements have been limited to the proximal market, with distal multihoming not being a strategic focus of console manufacturers, although many new console games are regularly released for previousgeneration consoles, handhelds, and mobile devices.
Owing to the proprietary nature of platform hardware, a game programmed for one console (e.g., PS4) cannot be played on other platforms, unless it is recoded or "ported." Porting is costly (including acquisition of proprietary development kits, licensing fees, recoding effort) but typically cheaper than the original game development (Williams, 2002). Still, resource restrictions often force smaller third-party developers to release games only for one platform. 3 By contrast, due to a dramatic increase in production costs, large third-party firms often aim to multihome games to several platforms to recoup up-front investments (Edwards, 2006).
Altogether, multihoming console games to proximal-and distal-market platforms is commonplace. While console manufacturers control the software market only partially through first-party publishing, there are multiple ways to incentivize or force third parties to adhere to desired release regimes. Interestingly, practice has largely ignored crossmarket interdependencies between software releases to distal-market platforms, despite their potential importance for console sales. We next propose a conceptual framework that incorporates this aspect and develop a coherent theory of multihoming to distal markets based on the concept of immersion, which is central to the consumption of entertainment products.

Conceptual framework
The availability of software affects the utility that consumers derive from a hardware platform in general (e.g., Gandal et al., 2000;Katz & Shapiro, 1985) and specifically in the video gaming industry (Clements & Ohashi, 2005;Gretz & Basuroy, 2013). Yet, how software releases affect platform sales might also depend on whether and to which platforms games are multihomed. We draw on hedonic consumption theory (Hirschman & Holbrook, 1982) to explain how multihoming software to different distal-market platforms affects focal console sales. In short, video gaming, as a hedonic experience, thrives on a player's immersion in the multisensory act of gaming. The more similar the extent of immersion offered by two gaming platforms, that is, the smaller the immersion gap between them, the more substitutable they are. As a result, multihoming to platforms with a small immersion gap should cannibalize focal console sales while consequences should be less daunting when multihoming to platforms with a large immersion gap. 4 In addition, we investigate two relevant aspects that affect the immersion gap between the focal console and distal-market platforms: game quality (with higher-quality games causing the immersion gap to increase) and console age (with higher age causing the immersion gap to decrease). Figure 1 summarizes our conceptual framework.

Hedonic entertainment experiences, immersion, and substitutability
Hedonic entertainment experiences, such as watching movies or playing video games (Voss et al., 2003) incorporate multisensory, fantasy, and emotive aspects, in line with Hirschman and Holbrook's (1982) original conceptualization. Key to an enjoyable experience, and motivation for consumers to play video games, is the extent to which the experience allows them to escape reality and be transported into a virtual world (Przybylski et al., 2010). This concept, referred to as immersion (Jennett et al., 2008;Lombard & Ditton, 1997), relates to the "cognitive experience wherein the thoughts of the player are wholly absorbed in the action within the game" (Cairns et al., 2013(Cairns et al., , p. 1070. Besides video gaming, immersion is a relevant concept in hedonic contexts where consumers benefit from being drawn into a virtual world like movie consumption (e.g., Visch et al., 2010), virtual reality (e.g., Bowman & McMahan, 2007), reading (e.g., Douglas & Hargadon, 2000), or online shopping (e.g., Visinescu et al., 2015).
Facilitating factors of immersion in video gaming are a platform's technical performance and usability (Burns & Fairclough, 2015), with widescreen high-quality graphics, brilliant sound, and intuitive controls as key ingredients (Grimshaw, 2008;Ryan et al., 2006). As these characteristics vary between game platforms, platforms also vary in the extent to which they can create an immersive experience (see Table 3). This key difference between platforms, which we name the immersion gap, should then define the extent to which platforms are substitutes. In particular, substitutability between two platforms should be inversely related to the immersion gap between them, as similarly hedonic experiences are closer substitutes for one another (Ratneshwar & Shocker, 1991), and consumers are more likely to switch between close substitutes than distant ones (Day et al., 1979;Tversky, 1972). Such substitution phenomena may help explain how distal multihoming affects focal console sales.
In general, the smaller the immersion gap between the focal console and another platform, the higher the substitutability between them and the more multihoming to said platform should cannibalize focal console sales.

Distal multihoming regimes and focal platform sales
Compared to latest-generation consoles, which facilitate similarly immersive gaming experiences (Cennamo, 2018), platforms in distal markets feature inferior technological characteristics and thus provide less immersive gaming experiences (Web Appendix B). The corresponding immersion gap should render them only imperfect substitutes to a focal console. We introduce three types of distal multihoming and briefly discuss how each may affect focal console sales.
Backward multihoming Releasing games to also be playable on previous-generation consoles, which we label backward multihoming, supports the existing installed base of users (Schilling, 2003). Previous-generation consoles offer a gaming experience whose immersive quality should, relative to handhelds and mobile devices, be closest to that of latest-generation consoles as, across generations, console gaming builds on wide-screen television graphics, highquality audio, and similar game controllers. As a result, the immersion gap between these two platform types should be noticeable but comparably low. Backward multihoming should thus cannibalize the positive effects of a new game release on focal console sales. In particular, backward multihoming may foster inertia in owners of previous-generation consoles to purchase a new console to play a specific game if they can also do so with a similarly immersive experience on the console they already own.
However, backward multihoming might also exert complementary effects. As such, it may lead consumers to imagine or experience what they miss out on by not playing the game on the superior latest-generation console, with positive demand effects for the latter. Such comparisons may create a strong pull toward the platform that offers a more immersive experience, in line with findings that adding an inferior option to a choice set increases consumers' choice of the dominant option (Huber et al., 1982;Huber & Puto, 1983;Lehmann & Pan, 1994).
While backward multihoming might therefore affect focal console sales either positively or negatively, the negative effects seem more likely to prevail in our context. Console gamers typically do not face a choice to buy one of two platforms but to either stay with the status quo of their previousgeneration console or upgrade to the focal, latest-generation console. Effects of complementarity may thus be limited unless consumers are at the same time highly enticed by a game but also sufficiently disappointed by the immersion   (Hirschman, 1987), escapism and exploration (Yee, 2006), and social learning through character identification (Green et al., 2004;Hirschman, 1983) Wide-screen TV; superior graphics processing and computing power (e.g., higher frame rates; richer colors) Level of immersion offered on their current console. In addition, the smaller the immersion gap, the less consumers should be willing to engage in the relatively expensive upgrade to the latestgeneration console.
Handheld multihoming Handheld devices are portable gaming platforms with a built-in screen and controls. In general, the immersion gap between them and latest-generation consoles should be noticeable, owing to, for instance, their smaller screen and lower computing power. Consequently, compared to previous-generation consoles, handhelds should be less able to substitute for latest-generation consoles (Day et al., 1979). Handheld multihoming should therefore pose less risk of cannibalizing focal console sales compared to backward multihoming. On the other hand, the high immersion gap between latest-generation consoles and handheld devices might also prevent handheld multihoming from conferring substantial complementary effects on focal console sales. Research shows that complementary effects of adding inferior alternatives decrease if products are less similar (Huber et al., 1982;Mishra et al., 1993). Thus, while backward multihoming might lead consumers to ponder the difference in immersion generated by previous-versus latest-generation consoles, this should occur less for the difference in immersion between handhelds and the latest-generation console. In addition, players of handhelds are typically informed gamers who consciously buy an expensive handheld device specifically for portable gaming, aware of the immersion gap to the focal console. As such, handheld multihoming should not lead to widespread upgrading from a handheld to a latestgeneration console for a more immersive experience.
Mobile multihoming Mobile devices provide a much less immersive gaming experience compared to the wide-screen cinematic visual experiences offered by latest-generation consoles (Barder, 2015;Shafer, 2013). Due to the resulting substantial immersion gap, they should represent only weak substitutes to the focal console (Huber et al., 1982;Mishra et al., 1993) and threats of sales cannibalization should be low. Correspondingly, the eighth console generation has been outselling its predecessor (Anderton, 2018), even though the rise of mobile gameplay was foreseen to start the demise of consoles (Kohler, 2012;Snow, 2012).
Yet, mobile multihoming may exert complementary effects and strengthen the influence of a game release on focal console sales by providing consumers a low-barrier entry into video gaming. Similar to sampling or versioning (Arora et al., 2017;Dey & Lahiri, 2016), mobile multihoming may thus attract new consumers to video gaming by allowing them to try a game free of charge (or for a low price) on their mobile device. In turn, they may purchase the focal console to experience the game in its most immersive version. Thus, mobile multihoming may markedly increase the number of consumers exposed to a game and thus lead to substantial complementary effects (Mishra et al., 1993).

Moderating effects of software quality and console age
Housed in hedonic consumption theory, our prior argumentation builds on the rationale that substitutability, and thus cannibalization, should be inversely related to the immersion gap between the focal console and a distal-market platform. To further investigate this notion, we examine two aspects relevant to the immersion gap. We argue that both affect the extent to which a specific distal-market platform provides a gaming experience whose immersion resembles that of the focal console. First, as a key game characteristic underlying immersion, we investigate the moderating role of software quality (e.g., Binken & Stremersch, 2009). Second, regarding the focal console, we examine the age of the console in its product lifecycle (e.g., Gretz et al., 2019).
Software quality Video games differ in the extent to which they exploit a platform's immersive potential. While some titles take full advantage and bring out the best of a platform's graphics and sound capabilities, others only scratch on the surface of what the platform has to offer. Software quality may thus reflect the ability of a game to unearth a platform's full immersive capacity. In this regard, highquality games typically operate at the frontier of the technically possible and may thus be better able to fully exploit a console's immersion potential than lower-quality games (Hood et al., 2020). Therefore, game quality should alter the immersion gap, and thus the substitutability, between two platforms for a specific game release.
We argue that backward multihoming might cannibalize focal-console sales due to the small immersion gap between the focal console and previous-generation consoles. However, since the full immersion potential of the focal console may only emerge with high-quality games, the immersion gap should be largest for them. By contrast, for low-quality games, the difference in immersion might be much less significant. Owing to the smaller immersion gap, cannibalizing effects of backward multihoming should be stronger for lower-quality games.
By contrast, for platforms that per se exhibit a larger immersion gap to latest-generation consoles, software quality might not play a meaningful role. Thus, for handheld or mobile devices which, due to their smaller screens as well as inferior graphics and audio, provide rather low levels of immersion, software quality should not substantially affect the immersion gap to latest-generation consoles. It should thus neither affect the relevance of handheld nor mobile multihoming for focal console sales.
Console age Game consoles are marketed by generations in which they follow a product lifecycle (e.g., Clements & Ohashi, 2005). Console age then refers to the "aging process that the hardware experiences during its useable life" (Gretz et al., 2019, p. 395). Adding to our prior theorization, we argue that console age affects the relevance and extent of the immersion gap between a focal console and distal-market platforms. As a result, cannibalization from distal multihoming should intensify in console age.
Prior research shows that early adopters of innovations are especially interested in the new features a product has to offer and tend to be more enthusiastic about new features rather than comparisons with previous products (Rietveld & Eggers, 2018). In video gaming, early adopters seek out the best possible hedonic experience and thus immersion when playing a game (Rogers, 2003), which the latest-generation console provides. Thus, the ability to play a new game on their previous-generation console, as afforded by backward multihoming, should exert a weaker influence early in the focal console's lifetime, that is, during the buying phase of early adopters. By contrast, such substitution should become more relevant later in the console's lifecycle where prospective late adopters may increasingly settle with the lower immersion provided by the previous-generation console instead of upgrading to the latest-generation console. Toward the end of a console generation, backward multihoming might also lead consumers to defer upgrading until the release of the next console generation. The immersion gap should thus become more important later in a console's lifecycle so that cannibalization from backward multihoming should strengthen in console age.
Next, the lifecycles of consoles and handheld devices are typically not aligned (Preceden, 2021) so that new and improved handheld devices become available during a console's lifecycle. As the immersion offered by the focal console remains practically unchanged over its lifecycle, 5 the immersion gap between it and handheld devices decreases as the console ages, rendering them more substitutable. Cannibalization from handheld multihoming should therefore become more severe toward the end of a console's lifecycle when the immersion gap is smallest (Day et al., 1979). Similar effects should occur for mobile multihoming. New mobile devices are continuously released throughout a console's lifecycle, decreasing the immersion gap between the latest-generation console and mobile devices as the console ages. Cannibalization due to mobile multihoming might thus increase with console age.

Description
Our data cover all six game consoles (Wii, PS3, Xbox 360, Wii U, PS4, Xbox One) in the US home video game industry broken out monthly from November 2005 through March 2017 (148 consecutive months). These consoles constitute the seventh and eighth generations of the stationary console gaming market. We use a subset of these data starting from the month in which more than one console was available to consumers (November 2006 for the seventh generation and November 2013 for the eighth generation) to obtain a balanced panel per generation, a necessary condition for proximal multihoming to occur. The data comprise information on console hardware and approximately 8000 corresponding game releases.
We collected the number of games, their release dates on latest-generation consoles, previous-generation consoles, handhelds, and mobile devices, and more than 1.2 M quality ratings from multiple well-known gaming websites (e.g., Metacritic, IGN, Gamespot). Because game release dates differ between regions, we limited data collection to the US market. We included both expert and user ratings to avoid potential bias (Marchand & Hennig-Thurau, 2013).
Data for console sales originate from vgcha rtz. com, a source used in previous analyses of the gaming industry (e.g., Healey & Moe, 2016). We obtained console prices over their lifecycles from the manufacturers' websites and dynamically adjusted them for inflation using the American Consumer Price Index (CPI). Finally, Kantar Media provided information on console and game ad spending for each of the six consoles throughout the entire analysis period and on ad spending for the sixth console generation (prior to our observation period), which we use as instruments (see section on endogeneity). The advertising data include television, print, radio, out-of-home, cinema, and digital display, search, and video advertising. Table 4 summarizes the variables, their data sources, and the operationalization of our software release measures.

Measures
We measure software multihoming as the share of new games released on console c in month t under each distal multihoming regime as well as the share of singlehomed games to account for releases in the proximal market. 6 We also control for the total number of games released for console c in month t. This approach is preferable to using absolute release counts of each multihoming regime as these could increase or decrease over time and thus introduce bias. We use new game releases because our focus is on the impact of monthly releases on console sales, not software stock as a cumulated game catalog (which we, however, also control for). Furthermore, revenues in the video gaming industry are highly skewed and predominately generated at, or shortly after, a game's introduction (Hennig-Thurau & Houston, 2019). Console sales are monthly US-market unit sales of a specific video game console. Figures  WC1 and WC2 in Web Appendix C show game releases and console sales over time for each console. We include prices of each console's basic version in month t and the console's age counted from the first date that all consoles were available for purchase. Console advertising is the monthly gross spending (in US dollars) on media communications, including advertising on accessories such as controls and component parts.
We operationalize the software-related control variables as follows. Game catalog size denotes the number of games available for a given console at time t-1, which equals the cumulative number of game releases up to the month of interest. Game advertising is the monthly gross spending for all new games available on the focal console. We also include the number of very high-quality game introductions (superstar games) to control for further potential endogeneity threats as superstar games can have disproportionate effects on console sales (Binken & Stremersch, 2009). Systematic introduction of superstars under a specific release regime (e.g., singlehoming) could thus create a confound. In line with previous literature (e.g., Gretz et al., 2019), we define superstar games as games with an average quality rating of 90 (on a scale from zero to 100) or above. Table 4 includes the descriptive statistics of our variables; we also provide a correlation table in Web Appendix C.

Model
We formulate an error-correction model (ECM) adjusted for panel data to estimate the effects of software multihoming on console sales (Van Heerde et al., 2013). Because effects might differ across consoles, we employ a hierarchical model structure to allow for heterogeneity in the parameter estimates (Fok et al., 2006). We next present the general model and detail our control variables and moderating variables (i.e., software quality and console age). Furthermore, we discuss how we correct for possible endogeneity bias.

Model specification
Our model builds on network effects theory according to which a higher number of complements (here, video games) increases customers' valuation of the compatible platform (here, the game console), fueling platform sales (e.g., Katz & Shapiro, 1985). This mechanism suggests a simple count model including the absolute monthly number of new game releases under each multihoming regime to distal-market platforms (previous-generation consoles, handhelds, mobile devices). However, if releases under one regime (e.g., mobile multihoming) increase or decrease over time, effects could be spuriously driven by the number of games rather than the specific multihoming regime (e.g., Gretz & Basuroy, 2013;Nair et al., 2004). We therefore specify a model that retains the general structure and variables of such a count model but includes shares of each software release regime (MultihomeShare ict and SinglehomeShare ct ), while controling for the overall number of game releases (TotalRelease ct ).
Changes from one period to the next are indicated by Δ (ΔY t = Y t -Y t-1 ) and ln is the natural logarithm; the superscripts CML and IM denote cumulative and immediate effects, respectively, estimated directly from the ECM specification (Fok et al., 2006). The cumulative effect includes the immediate effects (in month t) as well as longer-term effects (beyond month t). CSales ct is the sales volume of console c in month t. Our model implicitly sets up the moderation of game releases by software release regimes, using total monthly release counts (TotalRelease) to yield the overall software release effects β CML 5 and β IM 5 , and including the shares of games according to each release regime to establish the moderation effects. MultihomeShare ict is the share of games released in month t under distal multihoming regime i. SinglehomeShare ct is the share of singlehomed games released in month t with respect to the proximal market, which we include to correctly identify all effects. We include the share of proximally singlehomed (≙ 1-share of proximally multihomed) games to keep the model comparable to prior studies. As we explain in more detail below, we also distinguish between k different controls that have both an immediate and a cumulative effect as well as j different controls that correct for trending patterns.
The error term ε ct follows a normal distribution with a block structure covariance for console generations. Specifically, we (1) ln SinglehomeShare ct−1 − β CML 5c TotalRelease ct −δ CML kc Control kct−1 � + α jc TrendControl jct + ε ct allow for correlations between the equations of one generation (e.g., Wii, PS3, and Xbox360 as seventh-generation consoles), but not across generations. The parameter γ c represents the speed of adjustment according to which explained and unexplained shocks converge toward the long-term equilibrium (Fok et al., 2006). The console-specific parameter μ c captures time-invariant unobserved effects on console sales. For all other parameters, we combine information across consoles in pooling equations: where θ c are the parameters from Eq. (1) except for the intercept and the error term. In particular, we denote , γ c , δ IM kc , δ CML kc , α jc , with superscripts IM and CML indicating the immediate and cumulative effects, respectively, and subscripts i, j, and k indicating the distal multihoming variables, controls, and trend controls. This simplified description of the pooling equations indicates that each variable in in our model has its own hypermean (i.e., θ ) and error term (i.e., ν c ). The errorterms are uncorrelated across equations.

Control variables
As control variables we include inflation-adjusted console prices, console advertising, and competitor console advertising, all as natural logarithms. We also include variables that are specific to the software-platform dynamics we intend to capture, that is, the number of superstar software releases and software advertising spending. As trend control, we include the size of the compatible games catalog up to month t-1 (Binken & Stremersch, 2009). We also include the console's age in months as well as November and December holiday dummy variables, in line with previous studies. We rescale console age to values between zero and one, following prior error-correction specifications (e.g., Van Heerde et al., 2013).

Moderation analyses by software quality and console age
To test whether software quality influences the effectiveness of the different software release regimes, we computed the mean quality of all games and added one standard deviation to distinguish between high-quality games and average-tolower-quality games (lower-quality games, in short). We then calculated the share of high-quality and lower-quality game releases for each multihoming regime. We performed estimations for each regime separately to avoid overburdening the model (Van Heerde et al., 2013). For example, the quality moderation of mobile multihoming includes the first difference (immediate effect) and lag (cumulative effect) of (2) θ c = θ + ν c ,

MultihomeShare
, while retaining the other variables. The high-and lowerquality share variables represent implicit moderations because multiplication with the total number of mobilemultihomed software releases yields the count of high-and lower-quality releases, respectively. To test whether the effects of multihoming regimes change over a console's lifecycle, we mean-centered the console age variable and multiplied it with the share of each release regime. We also mean-centered the release regime variables to facilitate comparison of the effects of different release regimes to those in our main model.

Endogeneity
Console manufacturers may foster the exclusive release of specific titles on their console via first-party publishing or singlehoming deals to support console sales, similar to advertising (Papies et al., 2017). This strategic behavior could occur in the proximal market for games that are particularly important for the sales of the focal console (Coughlan, 2004). Consequently, our estimates could be biased, which we address by using instrumental variables as described below.
Our efforts focus on the release decisions of first-party publishers in the proximal market, because multihoming to distal markets or by third-party publishers are unlikely to cause bias. As discussed, distal multihoming has not been a strategic consideration for console manufacturers as part of their efforts to influence console sales. This leaves us with third-party publishers' multihoming decisions that might be endogenously determined and that could affect results, if publishers made release decisions for distal-market platforms based on (expected) console demand. However, since our model includes console fixed effects, 7 bias would only occur if publishers also timed game releases based on expected console demand in month t. Given that thirdparty publishers do not have a strong incentive to coordinate releases on distal markets (e.g., mobile release) to console sales in the proximal market, and considering long software development cycles as well as high costs of changing prefixed release schedules, strategic release timing for distalmarket platforms seems extremely unlikely. This view is backed by our interviews, which reveal that publishers do not consider console sales when releasing games to distal markets. We thus focus our endogeneity correction approach on the proximal market.
To assure unbiased estimation of software release effects in the proximal market, we took to markets related to the seventh and eight generation US game consoles for suitable instruments, following Papies et al. (2017). As a related but geographically distinct market, we collected information on singlehomed game releases in Europe. The resulting instruments have high validity because despite several (exogenous) commonalities, substantial differences in platform dynamics exist. Specifically, we observe that, to some extent, first-party publishers implement global release strategies, which should not be driven by US console sales and thus be exogenous. Furthermore, third-party publishers typically do not optimize their game releases based on console sales, while being likely to release games on both European and US markets. These commonalities give us the exogenous variation we need for a strong and valid instrument.
On the other hand, substantial differences in platform dynamics exist, which would induce publishers to vary their release timing decisions between Europe and the US. For example, Sony executive Jim Ryan stated that in "…Europe, it's really been fortress PlayStation by at least 3-to-1 in unit sales, compared to 2-to-1 in the US" (Peckham, 2017). European game releases should thus capture common exogenous patterns (e.g., simultaneous release of games that follow a global release strategy), while not affecting manufacturers' release decisions in the US market.
Console advertising is another potentially endogenous variable. Following prior literature, we expect a negative bias, because console manufacturers might boost advertising when they expect demand to decrease. As an instrument from related markets, we propose a geographically overlapping but temporally distinct measure: console advertising spending for previous console generations. Specifically, we use monthly advertising spending in the seventh (sixth) console generation as an instrument for advertising spending in the eight (seventh) generation, owing to their similar patterns along the consoles' lifecycles. Irrespective of the generation, console manufacturers plan advertising seasonally and based on the console's age, while demand across generations is clearly independent. Since we control for a time trend and seasonal patterns, the instrument also does not capture common demand patterns across generations, which is necessary for it to be exogenous.
To employ our instrumental-variable approach, we use a procedure common in Bayesian estimation where we correlate the endogenous equations with the relevant first-stage equations (e.g., Ataman et al., 2008), for a system of nine equations per console generation (one first-stage equation and two endogenous equations per console) in the main model and in the quality and age moderations. Because we have two endogenous variables, we can test for the strength of the instruments with a pooled Angrist-Pischke F-test (Angrist & Pischke, 2009). The results indicate that the instruments are sufficiently strong (p < .05). Separate tests for all sets of control variables demonstrate the strength of our instruments across all specifications. Web Appendix D provides further details on our endogeneity corrections and results of the firststage equations.

Results and discussion
We obtain posterior densities of all parameters using 100,000 MCMC draws, discarding the first 50,000 draws to calibrate the samplers (see Web Appendix E for the sampling procedure). Table 5 shows the hyperparameters of our hierarchical model specification. These reflect the average effects across all six game consoles which we focus on in our discussion. 8 Due to the log-log formulation, we can interpret the coefficients of the different release regimes as elasticities and compare them directly. We focus our discussion on the cumulative effects (over several months) since prior research has documented a lasting impact of software on hardware sales (e.g., Binken & Stremersch, 2009), and because cumulative effects include the immediate effects (within the release month), which represents a strength of our ECM specification.

How multihoming affects console sales
Previous-generation consoles are the closest distal-market substitute to the focal console. We find a negative impact of backward multihoming on focal-console sales ( β CML 1 = − .236), indicating cannibalization. Because our estimates are elasticities, an increase of 1% in the share of backward multihomed games leads to a .236% decrease in console sales, holding all other variables constant, including the total number of software releases. 9 Based on the average number of 402,000 consoles sold per month, this would translate to 948 fewer console units sold. In line with our theorization, consoles manufacturers should thus be aware that backward multihoming may foster inertia in gamers and make them feel no need to purchase the new console so long as the old console, which provides a similarly immersive gaming experience, is still supported with fresh titles. We do not find a significant effect of handheld multihoming on console sales ( β CML 2 = − .109). This finding speaks to the independence of the two markets and shows that, in general, consumers tend not to trade off playing a game on the latest-generation console with playing it on a handheld device that provides lower immersion. The availability of a game on a handheld does not increase latest-generation console sales either, rendering handheld multihoming neither a threat nor an opportunity for console manufacturers. This lack of permeability might thus be a chance for manufacturers and publishers to extend their activities from the console market to the handheld market without fear of repercussions.
Mobile multihoming has a positive impact on focal-console sales ( β CML 3 = .260) and thus complements rather than cannibalizes. Here, a 1% increase in the share of mobile multihomed games leads to a .260% increase in console sales, which amounts to 1045 units for the monthly sales average. This finding aligns with our argumentation and the results of our pre-study that mobile devices are only weak substitutes to latest-generation consoles due to the considerably less immersive gaming experiences they offer. This finding is interesting, as industry observers have predicted the demise of consoles at the hand of mobile gaming for years (e.g., Forbes, 2018). Our results indicate instead that the mobile market generates positive spillovers to consoles. Console manufacturers could thus leverage mobile multihoming to allow consumers to try out a low-immersion version of a game to then purchase the focal console to experience the game with much higher immersion. Mobile versions of console games thus seem to provide easy access to new titles and may, in this way, draw new players to the gaming market who may then purchase the focal console for the "full" gaming experience.
Although we focus on sales effects of distal multihoming, the effects of singlehoming on the proximal market are also noteworthy as they qualify insights from prior research. Specifically, we find no effect of singlehoming ( β CML 4 = − .092), implying that, on average, whether a game is released on one or several consoles in the proximal market does not affect focal console sales. This finding challenges the widely-held belief in singlehoming as the superior release regime over multihoming to differentiate a console from its direct competitors. One corresponding explanation could be that game titles are not equally suitable to foster console differentiation, so that singlehoming is not unequivocally beneficial. Instead, some games may be more useful than others to help a console stand out from its competitors (Lee, 2013). We revisit this aspect in our discussion of moderating effects. Table 6 shows results of the moderation analyses by software quality and console age. For backward multihoming, lower-quality games have a strong negative effect on focal-console sales (β CML : LQ = −.284, Table 6, column 1). Figure 2 illustrates the differences in elasticities between lower-and high-quality games, which based on the average number of 402,000 consoles sold per month, would translate to 1311 fewer console units sold for lower-quality games under a backward multihoming regime. This result suggests that gamers are less willing to upgrade to a new console as long as lower-quality games keep being released on the outdated platform (Gretz et al., 2019). Technologically, lowerquality games translate more readily to previous-generation consoles, while high-quality games are more likely to fully exploit the performance advantage of the latest console generation. The gaming experience immersion thus differs most vividly between previous-and latest-generation consoles for high-quality games, dampening consumers' inertia to purchase the latest-generation console.

Moderating effects of software quality and console age
The negative effect of backward multihoming further strengthens with console age (β CML : AGE = −.677, Table 6, column 5) and thus hurts the latest-generation console especially later in its lifecycle. 10 As illustrated in Fig. 3, panel A, the negative effect of backward multihoming becomes 9 Consider the following example for interpretation: If 50 games are released in a given month, with the number of backward-multihomed games increasing from five to six (changing their share from 10% to 12%, which corresponds to a 20% increase), this would lead to a 4.7% decrease in console sales (or 18,894 units). 10 The magnitude of this effect does not translate directly into an adjustment of the main effect (i.e., share elasticity) for an extra month along the console's lifecycle because age is scaled between zero and one. As a rough estimate, dividing the coefficient by 69 (average number of months per console) yields the change in elasticity of the main effect for each month the console ages.

Table 6
Moderation by software quality and console age Moderation by software quality

Moderation by console age
Coef.
(1) Distal: backward (2)   significant around a console's lifecycle midpoint. This finding aligns with our rationale that it is especially late adopters, caring relatively less for the advanced graphics and innovative gameplay offered by the focal console (Gretz et al., 2019;Rogers, 2003), who experience increased inertia to upgrade from their outdated console in light of ongoing support with fresh titles. Furthermore, consumers could increasingly defer a console purchase towards the end of the console's lifecycle, when the next generation is about to be released. Exemplary for backward multihoming, we find that the significant mid-lifecycle elasticity ( β CML∶AGE mid = -.192) and end of lifecycle elasticity ( β CML∶AGE end = -.528) would translate to 806 and 2123 fewer monthly console units sold, respectively.
For handheld multihoming, we find no significant effect on focal console sales, irrespective of software quality (β CML : LQ = -.120, Table 6, column 2), as also illustrated in Fig. 2. This result reflects our argument that software quality does not meaningfully alter the immersion gap between the focal console and a platform with per se substantially lower immersion. Yet, as the focal console ages, the negative effect of handheld multihoming strengthens (β CML : AGE = -.926, Table 6, column 6). As shown in Fig. 3, panel B, this effect becomes significant toward the end of the focal console's lifecycle, possibly resulting from the decreasing immersion gap between the platforms, as new handheld devices are released during the console's tenure. Consumers might thus be more reluctant to buy the focal console, leading to accelerating cannibalization toward the end of a console's lifecycle.
We find a marginally significant, positive effect for multihoming lower-quality games to mobile devices (β CML : LQ = .230, Table 6, column 3) that does, however, not differ from the effect of high-quality games (β CML : HQ = .170), as also visualized in Fig. 2. Thus, mobile multihoming complements focal console sales irrespective of software quality. This finding also aligns with our argument that mobile devices provide substantially lower immersion than latestgeneration consoles in general so that software quality does not play a major role for its effect.
The positive effect of mobile multihoming on console sales moreover diminishes with console age (β CML : AGE = -.695, Table 6, column 7). Since new and improved mobile devices are released throughout the focal console's lifecycle, the immersion gap between these platforms decreases. As such, consumers might be increasingly less likely to purchase the focal console after having experienced a game on a mobile device. Our results thus support the notion that mobile multihoming is a suitable promotion tool especially during a console's nascency, nudging consumers who learned about a new game on their mobile devices to then purchase the console for the most immersive gaming experience.
We define high-quality software as games with an average rating higher than the mean quality across all games +1 SD and lower-quality software as games with an average rating lower than the mean quality across all games +1 SD. Console age is measured continuously, denoting the months elapsed since the focal console was launched Table   6 (continued) Moderation by software quality

Moderation by console age
Coef.
(1) Distal: backward (2) Distal: handheld (3) Distal: mobile (4)  .032 .021 .020 .018 .018 Game advertising (Δ, ln)  With respect to the proximal market, we notably find that singlehoming high-quality games exerts a strong positive effect on focal console sales (β CML : HQ = .202, column 4), while the effect of singlehoming lower-quality games is insignificant (β CML : LQ = -.096). These results indicate that only high-quality titles foster platform brand differentiation (Gretz et al., 2019). They also further detail findings by Landsman and Stremersch (2011), who examine the effects of proximal singlehoming (vs. multihoming) by weighting the catalog of available games by the games' respective sales, thus implicitly focusing on the effects of bestselling titles. Since game quality can be interpreted as a proxy for game sales, our quality-moderation model closely resembles their main model and results. This finding is crucial for platform manufacturers that aim to secure singlehoming agreements with software publishers or engage in own software development, hoping to help their platform succeed. Finally, although we find that console age negatively moderates the effects of singlehoming (β CML : AGE = -.764, column 8), Fig. 3, panel D, shows that, for an average game, the effect remains insignificant along the console's entire lifecycle.

Additional analysis: Market stealing or category expansion/detraction?
The estimates in our model represent net effects, driven by (1) gains or losses in market share within the proximal console market (market stealing) and (2) overall increases or decreases in the number of consumers in the market (category expansion or detraction). To disentangle these effects, we specify a Market Share Attraction (MSA; Cooper & Nakanishi, 1988) version of our model (see Sotgiu & Gielens, 2015 for a similar setup). We construct console market shares from sales to analyze the effects of multihoming on focal-console market share. In Web Appendix G, we provide a detailed explanation of the MSA model development as a Hierarchical Bayesian ECM and summarize its results.
We find that the effects of backward multihoming are significantly driven by market stealing ( β CML 1 = -.294), indicating that backward multihoming not only hurts absolute console sales but also affects the console's relative position vis-à-vis other consoles as its direct competitors. The results also suggest a marginally significant, positive effect of mobile multihoming on the focal console's market share of similar magnitude ( β CML 3 = .305). Mobile multihoming can thus support a console's relative growth trajectory when competing in the proximal market.
In addition, comparing the sales and MSA estimates sheds light on which proportion of the sales effect stems from market stealing and category expansion/detraction, respectively. In general, we expect predominately market stealing effects from other latest-generation consoles because the overall number of consumers in the console market has remained steady in past decades (Statista, 2019). Overall sales show a similar growth pattern for each generation where the manufacturers compete for their share. However, especially multihoming to distal markets may inhibit consumers to upgrade to the latest-generation console, such as in the case of backward Fig. 2 Software quality (lower versus higher quality) moderation for different software release regimes. Notes. MH = multihoming; lower quality (grey) versus high quality (black); significance between low and high is based on the draws of the estimates; ** significant at 5% level, and *** significant at 1% level based on both tails of the highest posterior density interval (HPI). Elasticities denote the effect of a game release under a specific release regime on focal console sales for high-quality (defined as games with an average rating higher than the mean quality across all games +1 SD) and lower-quality games (defined as games with an average rating lower than the mean quality across all games +1 SD) multihoming, or draw in new consumers, such as in the case of mobile multihoming, which the relative proportions of category expansion/detraction effects should reflect.
We can interpret the estimates of our sales model as elasticities due to the log-log relation between the dependent and multihoming variables. In the MSA model, we use the log-log equivalent, that is, multiplicative competitive interaction. Here, elasticities equal estimates scaled by (1-market share/100). For comparison, we use a market share of 33% which is each console's average market share in a market with three key players. For backward multihoming, approximately 83% of the sales effect reflects stealing market share from direct competitors in the proximal market. For mobile multihoming, this proportion is only 76%. Thus, mobile multihoming has a relatively stronger effect on total sales in the proximal market, from which all latest-generation consoles benefit. In other words, backward multihoming seems to hurt especially the focal console, while mobile multihoming helps the focal console but also more strongly fosters growth of the entire video game console category.

Theoretical implications
This study advances prior research on multihoming in important theoretical directions. We provide a comprehensive conceptualization of multihoming, expanding its limited scope in the literature. Extant work in marketing (e.g., Landsman & Stremersch, 2011), management (e.g., Corts & Lederman, 2009), and economics (e.g., Choi, 2010 has treated the release of platform complements as a simple either/or-decision between singlehoming and multihoming in the proximal market. Yet, it has become common industry practice to also multihome games to related platforms in distal markets. We therefore contribute to previous work by introducing a more nuanced view on platform competition, proposing that multihoming software beyond narrow market boundaries may exert differential effects on focal platform sales. Thus, instead of investigating whether software multihoming affects platform performance, as put forth in previous literature (Hagiu & Lee, 2011;Mantena et al., 2010), we propose to focus on when and which type of multihoming helps or hurts a focal platform.
This study also contributes to hedonic consumption theory (Hirschman & Holbrook, 1982) by proposing immersion as a key concept to analyze substitution effects between proximal-and distal-market entertainment platforms. Specifically, we extend the theory to a digital context, introducing immersion as an overarching construct to differentiate between experience qualities of platform technologies in hardware-software markets. We show that immersion, as a psychological state, is driven by objectively measurable technological features (e.g., screen size and graphics), conceptualizing it as an important antecedent to consumers' digitalized "multisensory, fantasy, and emotive responses to product usage" (Hirschman & Holbrook, 1982, p. 92)cornerstones of hedonic consumption. Future research may incorporate this link between objective performance features and consumer response to analyze the role of technology in fostering hedonic experiences.
With respect to our study focus, extant literature has argued in favor of releasing software exclusively on one platform to drive its differentiation (Binken & Stremersch, 2009;Landsman & Stremersch, 2011). While this view is reasonable for hedonic products that provide similar levels of immersion (i.e., platforms within the proximal market), it is less suitable for explaining spillover effects between products that differ in immersion (i.e., platforms within the distal market). Our conceptualization built around immersion addresses this issue and provides a more complete theoretical foundation to assess substitutability and complementarity between digital entertainment products. In this way, it also speaks specifically to findings on inter-generational technological progress as a determinant of substitution effects between hardware generations. For example, Kretschmer and Claussen (2016) find that backward compatibility (i.e., making latest-generation technologies compatible with previousgeneration software) slows diffusion of the new technology. However, this effect does not unfold if technology differences between generations are large, a result in line with immersion being a key driver of substitutability between platforms. Our extended framework may thus also serve as a blueprint to map competitive landscapes beyond video game platforms, providing a fresh conceptual lens on substitutability between experience products that ties in with literature on competitive spillovers and new technology introductions (e.g., Bhattacharjee et al., 2007;Ghose et al., 2006).

Managerial implications
Console manufacturers have ample opportunities to boost hardware sales through strategic software releases. Although third parties are important to ensure a steady supply of fresh titles, manufacturers continue to expand their in-house software development capacities, acquire independent developers, and strike contractual release agreements with publishers. Doing so, they often seek to secure exclusive software releases only on their console to differentiate from competitors in the proximal market. However, we find that software multihoming to specific distal markets can indeed foster console sales. Thus, managers need to broaden their view on software multihoming and realize the strategic potential of a wider range of release regimes. We next discuss concrete implications which we summarize in Table 7.
First, managers face a difficult decision whether to facilitate multihoming to previous-generation consoles, trading off additional game revenues against the sales of the new console. Our results show how releasing fresh software titles to outdated platforms cannibalizes latest-generation console sales and that associated negative effects can largely be attributed to a loss of market share. However, since this occurs predominately toward the second half of a console's lifecycle and for lower-quality games, manufacturers are well-advised to pursue backward multihoming mostly when the console is still nascent and for high-quality games. To this end, they may create strategic incentives for third-party publishers, for example, by linking advertising support for new games to their unavailability for previous-generation consoles. Manufacturers may further consider unilateral actions to lower consumers' barriers to upgrade and thus mitigate losses from backward multihoming at later stages. For instance, they could flaunt the immersive gaming experience of the latest-generation console, provide price discounts or trade-in bonuses, or use word-of-mouth campaigns to leverage peer effects from other owners of the latest console.
Second, our results offer promising news for console manufacturers that also market handheld devices. We find no negative spillovers from handheld multihoming to focal console sales, except for the very end of a console's lifecycle. As such, cannibalization emerges only as the focal console increasingly loses its immersive edge over new, technologically advanced handhelds that enter the market during the console's tenure. Two implications follow. First, handhelds constitute a market that is largely disconnected from the console market so that firms might benefit from treating them as independent. Hybrid gaming systems like the Nintendo Switch may thus represent a problematic compromise as they only allow manufacturers to monetize one platform and thus prohibit additional revenues. This may explain why Nintendo also released the Switch Lite, a designated handheld. Second, manufacturers should closely (+) positive effect for high-quality software (−) less important as console matures • Focus on singlehoming to boost platform sales only for high-quality software and assess what the title will be worth for the platform • Pursue predominately early in a console's lifecycle • Consider the profit impact of singlehoming through expensive third-party agreements and avoid costly bidding wars with competitors coordinate their multihoming efforts with the release of handhelds and consoles. Once the focal console ceases to sufficiently outperform current handhelds in terms of immersion, managers should work to stall multihoming to the latter for the remainder of the console's lifecycle. Third, we find positive spillovers from mobile multihoming to console sales, especially early in a console's lifecycle. Thus, console manufacturers should use their in-house studios and encourage third parties to develop game versions for mobile devices. They should also make greater strategic use of mobile games as samples for console games, especially during the console's nascency. As Corts and Lederman (2009) show, fostering these positive spillovers could even spur on cross-platform network effects where the large mobile installed base induces third parties to increase game development for both markets.
Console manufacturers are slowly realizing the synergetic potential between mobile and console gaming. For example, Bethesda Game Studios launched the free mobile game Fallout Shelter six months before its blockbuster Fallout 4 to promote the forthcoming console game (The Noobist, 2015). Our market shares analysis further highlights the potential to foster category expansion through mobile multihoming so that competing console manufacturers could consider teaming up to jointly exploit such effects.
Interestingly, and beyond our focus on distal multihoming, we also find that singlehomed games within the proximal market may not be the "shining beacons that attract players to an expensive new console" (MacDonald, 2020). While singlehoming can indeed propel focal console sales, this occurs only for high-quality software and especially early in the console lifecycle. This result qualifies work by Landsman and Stremersch (2011) as well as Corts and Lederman (2009), suggesting that their findings on singlehoming may partly be attributable to specific types and timings of software releases. Manufacturers need to carefully assess which singlehoming deals will indeed drive console sales, especially given their often-extensive costs (Hartup, 2013). To unlock the full potential of singlehoming, they should focus their first-party development and singlehoming agreements on high-quality software and, in line with Landsman and Stremersch (2011), incentivize such releases mainly during a console's nascency.
Although the empirical context of this study is the video game industry, our findings also have important broader implications. For example, in entertainment streaming, firms can use our conceptualization of immersion-based multihoming to attract consumers to superior products. A case in point for such efforts to realize spillovers from distal markets is Sky's mobile platform SkyGo that extends its TV-based subscription model. SkyGo restricts content access to smartphones and tablets, while more immersive widescreen TV use requires a higher-priced, full subscription.
Another example for distal-multihoming effects is the movie industry, where content producers struggle with multihoming titles for the cinematic screen to less immersive entertainment streaming services like Netflix and Disney+ in fear of cannibalization. Industry evidence, however, suggests that cannibalization is much weaker than expected and that many consumers prefer watching (at least some) films in theatres (Katz, 2021). This corresponds to our theory of multihoming to distal markets, with the immersion gap between theatres and home entertainment being sufficiently large to mitigate cannibalization.

Limitations and future research
A few limitations of this study may motivate future research. First, we investigate the effects of software release regimes across various platforms but not PCs, as almost every console game in our data is also available for PCs. In addition, measuring PC immersion is not trivial due to heterogeneity across models and consumers' idiosyncratic configurations, unobservable without individual-level data. Yet, since PC gaming is an important market in the video gaming industry, future research could obtain such data to also investigate the effects of PC multihoming.
Second, we focus on multihoming effects on focal, latestgeneration game console sales from the viewpoint of the console manufacturer. Future research could broaden this scope and examine cross-platform effects on the sales of all proximal and distal-market platforms. In addition, it would be valuable to investigate these effects from a software perspective, analyzing the multiple tradeoffs publishers face when releasing a game on any number of proximal and distal platforms.
Third, we apply a static classification of software release regimes and neglect that software may initially be singlehomed to only be multihomed later in a console's lifecycle. In our data we observe this for a small share of games (249 games, or approximately 7% of all releases). However, examining the effectiveness of this approach with a more suitable dataset might unearth additional relevant insights.
Fourth, while we use an extensive set of covariates in our empirical models, including additional game characteristics, such as game genres or ESRB ratings, could help further tease out the effects of specific multihoming regimes on focal platform sales. In addition, some platforms might be more suitable than others for certain game genres or audiences so that the effects of multihoming to them could affect focal console sales differently. Future research could thus incorporate these and other software content characteristics to further explore how they affect the influence of software multihoming on platform sales.
Lastly, new forms of gaming will influence market dynamics and the role of software in platform diffusion. We examine mobile gaming as the latest mass phenomenon, but game streaming services aiming to become the "Netflix of gaming" (Hecking, 2019) may introduce an additional type of distal multihoming. As new competitors enter the market and cross-industry spillovers multiply, the latest-generation console wars may have some surprises in store.