INTRODUCTION

The multicultural teams of multinational enterprises (MNEs) are often difficult to coordinate and manage, but at the same time they can be wells of creativity that produce high-quality outcomes (Kotabe & Murray, 1990; Taylor & Greve, 2006; Wang, Cheng, Chen, & Leung, 2019). Optimization of multicultural team performance has long been an important topic in both team research (Ren, Gray, & Harrison, 2015; van Knippenberg, Dawson, West, & Homan, 2011) and international business research (de Jong & van Houten, 2014; Stahl, Maznevski, Voigt, & Jonsen, 2010). As MNEs adopt evermore-advanced digital technologies in an effort to build competitive advantage, multicultural teams are increasingly operating in semi-virtual environments in which members interact online and face to face (Griffith & Meader, 2004; Maynard, Mathieu, Gilson, Sanchez, & Dean, 2019), their interaction complicated by advanced communication tools and diversified task environments.

The strengths of multicultural teams are at the same time at the root of their drawbacks. Cultural diversity ensures wide-ranging sources of information, knowledge, and perspectives (Blau, 1977), and at the same time hampers communication (Ren et al., 2015), and provokes conflict and social disintegration (van Knippenberg, De Dreu, & Homan, 2004). Inevitably, human beings will categorize others based on their perceived sociocultural identities and favor those they see as being similar to themselves (Tajfel, 1982; Tajfel & Turner, 1986). It is widely accepted that digitalization reduces the cost of communication (Brouthers, Geisser, & Rothlauf, 2018), and we know that engaging in technologically mediated virtual environments greatly impacts team dynamics as it changes the way people interact, the scope of their interaction, and even their conceptualization of it (Gabarnet, Montesano, & Feixas, 2022). This raises the following question: How does cultural diversity influence the performance of semi-virtual teams?

Esports offers an ideal context for investigating the impact of cultural diversity on semi-virtual team outcomes. Gamers have built a well-defined worldwide community that transcends the virtual world and the physical one (Taylor, 2012; Wagner, 2007). That community shares a common body of knowledge, beliefs, social norms, and values (Alvarez & Sachs, 2019; Clark, 1996) that are not subject to national borders (Seo, 2015; Taylor, 2012), rather, in-game-out-of-game self-identification (Grooten & Kowert, 2015:77) gives rise to an esports gamer identity, which is semi-virtual and relatively free of physical world constraints (Eklund, 2015). When gamer identity is more salient, the focus is more likely to be on what is held in common, reducing the impact of biases and stereotyping (Kane, 2010; Rink & Jehn, 2010), and facilitating effective coordination and social integration in multicultural teams. The result is improved team performance. With disembodiment and anonymity, gamer identity is more likely to be salient in the virtual world (Kukshinov, 2015), although in the physical one there are still sociocultural challenges with which to contend, e.g., cultural identity and categorization by national origin (Eklund, 2015).

We hypothesize three conditions that will lead to gamer identity becoming more salient. One, unilateral disturbance in the virtual world caused by changes in the rules of the game directs gamer attention away from the physical world toward the game world, increasing the salience of gamer identity. Two, to play with numerous virtual characters, as opposed to a small number of specialized ones, is likely to make gamer identity salient, as when there is a wide range of characters, it is unlikely that many of them will be connected to a player’s cultural background in the physical world. Three, to play at home court means being in a physical environment where gamer identity is more salient than any other sociocultural identity, the strength of which enables a multicultural team to develop high-quality outcomes. We empirically examined these three moderating effects on the relationship between cultural diversity and team performance using 8070 team-game observations involving 102 professional League of Legends (LoL) teams worldwide from 2017 to 2020, and find support for all of them.

Leveraging the insights of virtual identity research and social identity theory, we make two important contributions. First, we extend the body of knowledge on the internationalization of the digital economy by highlighting the impact of digitalization on social identity and further on the outcome of digitalized social interaction. One does not have to erase the cultural identities formed in the physical world; rather, these identities can be dominated by culturally irrelevant ones formed at least in part in the virtual world, given appropriate managerial attention and tactics in the virtual and physical worlds. Second, our findings enrich cultural diversity research in the era of digitalization in showing that non-national and semi-virtual identities that emerge in semi-virtual groups can serve as superordinate identities that mitigate the categorization effect and enhance the advantages of multicultural teams.

CONCEPTUAL FRAMEWORK

Cultural Diversity, Information Processing, and Social Categorization

Some have come to view cultural diversity as a double-edged sword (Minbaeva, Fitzsimmons, & Brewster, 2021) as different quantitative studies have found positive, negative, and nonlinear impacts on team performance (de Jong & van Houten, 2014; Stahl et al., 2010). Previous studies that show that cultural diversity has a positive effect focus primarily on information processing and knowledge recombination benefits. Specifically, cultural diversity can provide a team with more diverse sources of information, a richer knowledge base, and a heterogeneous mindset, which together can mitigate group-think (De Dreu & Weingart, 2003; Miron-Spektor, Gino, & Argote, 2011) and lead to novel recombination (Taylor & Greve, 2006), yielding more creative and better-quality decisions (Kotabe & Murray, 1990; Wang et al., 2019).

Studies that have shown cultural diversity to have a negative effect on team outcomes have primarily been based on theories of social identity (Tajfel & Turner, 1986) and social categorization (Tajfel, 1982). Social identity is our perception of the social category to which we belong (Hogg & Abrams, 1988). We tend to categorize both ourselves and others based on cultural identities, and to be favorably biased toward in-group members and negatively toward out-group ones (Vahtera, Buckley, Aliyev, Clegg, & Cross, 2017). Divergent cultural identities can engender barriers to communication, conflict, dissatisfaction, and social disintegration (Ren et al., 2015). These effects decrease the efficiency of knowledge development and information sharing among culturally different group members and can lead to a lack of motivation and unwillingness to cooperate (Minbaeva et al., 2021; Stahl & Maznevski, 2021). The result is suboptimal team performance. However, a number of ways to mitigate the negative effect of social categorization have been suggested including diversity training, collective trust development, and global leadership, that is, managers appreciative of local cultures and able to work across many different ones (Maznevski & Chui, 2018; van der Kamp, Tjemkes, & Jehn, 2015; Zander, Mockaitis, & Butler, 2012). Multicultural teams can also achieve superior performance when there is a salient superordinate team identity (Crisp, Stone, & Hall, 2006; van Knippenberg et al., 2011).

We are all multidimensional beings, in the sense that we have more than one social identity and belong to more than one social category. We wear different hats in different situations according to the relative salience of aspects of our identity (Ashforth & Johnson, 2001). An identity is salient “when an individual is prompted to categorize himself or herself along (the) identity-oriented criteria” (Forehand, Deshpande, & Reed, 2002: 1087). There is a tendency to conform, at least to some extent, to the stereotype of the salient identity (Forehand et al., 2002), and thus that identity is likely to affect social interaction processes (Oakes, 1987). Likewise, when a superordinate team identity is salient, team members have a propensity to identify with the team as a whole, sharing values, beliefs, and norms about what the team is and what it does (Kane, 2010). A strong focus on what members have in common as opposed to their dissimilarities can reduce conflict and increase social cohesion (Rink & Jehn, 2010).

Esports as a Source of Superordinate Identity

In addition to identifying with the team, members may share other superordinate identities, such as the esports gamer one we consider in this paper, the salience of which can mitigate the social categorization effect. Wagner (2007: 182) has described esports as “…an area of sport activities in which people develop and train mental or physical abilities in the use of information and communication technologies.” Esports has given rise to a young, digitally based, global industry with a sizable market that has grown dramatically (Karhulahti, 2017). Unlike some other sports, which in many cases have taken decades – if not hundreds of years – to gain worldwide popularity, esports can almost be seen as a born-global. The development of information technologies and the evolution of business models have together substantially reduced the marginal cost of esports entering new foreign markets (Hennart, 2014). Marketing and retail distribution take place mainly on the Internet, which makes it possible for a new game, or a new version of an old game, to be released almost simultaneously worldwide (Karhulahti, 2017). Moreover, clearly defined terms and signals that are readily translatable minimize differences in the gaming experience regardless of the national origin of the players (Furley, 2021).

Esports gamers share common ground free of extraneous social value systems and norms (Clark, 1996), and are also free of the stricture of physical borders (Karhulahti, 2017). Professional and amateur gamers, clubs, fans, game developers, tournament organizers, sponsors, traditional and new media, and other self-identified stakeholders constitute a new community that shares in the same knowledge, and that has in common a language, beliefs, and values which facilitates stakeholder communication and cooperation (Alvarez & Sachs, 2019; Fiol & Romanelli, 2012; Taylor, 2012). In addition, stakeholders build trust in one another by jointly forming social rules, based on which they make commitments and share resources that solidify trust (Fiol & Romanelli, 2012). Gamers have been very successful in their collective efforts to promote esports and to have it recognized by the general public as a valid sporting activity (Wagner, 2007).

In contrast to players of what might be thought of as traditional sports, the professional identity of an esports gamer is semi-virtual. The esports world provides a space where gamers can ‘disembody’ and so escape from the sociocultural constraints of the physical world, and as there are no interpersonal physical bonds in virtual interaction, gamers are not bounded by their physical referential or any behavioral rules based upon it. Game world anonymity conceals any and all physical identifiers from gender, to age, to race, to cultural background (Kukshinov, 2015). Disembodiment and anonymity create a safe place for gamers to explore and form new identities that need only comply with the rules set for globally standardized virtual characters (Ecenbarger, 2014; Wang, Yang & Shen, 2014). As interaction between gamers takes place in cyberspace and in the physical world, their virtual and physical identities merge to some extent in the way of representation or projection (Kim & Kim, 2016). This yields a semi-virtual gamer identity that follows the online and offline rules of the “in-game-out-of-game” community which are held and monitored worldwide.

Cultural Diversity and the League of Legends Team Strategy

LoL is a multiplayer online battle arena game that requires players to control virtual characters and get tasks done together as a team (Hebbel-Seeger, 2012). Each character has unique abilities and relative strengths. The combination of characters defines team strategy, and the abilities and strengths of characters in each position defines the quality of the strategy. Thus, LoL team performance depends on the skill of each individual player and also on the quality of team strategy. This is why we focus on team strategy, that is, virtual character configuration, as the outcome of a team process. The dynamism and competitiveness of LoL games make devising good strategy a creative task and its implementation complex. Developers update games frequently, with subsequent versions having different rules. Usually some of the virtual characters become stronger and others weaker. When a new version is released, professional gamers throughout the world quickly reach a consensus on the new best strategy. Implementing that strategy is not easily done, first because mastering a character made stronger than in a previous version of the game may be beyond the expertise of all team members, and second because a strong character that team members have already mastered may be banned or effectively countered by a rival team.1 This means that for a professional LoL team to be successful, its members need to have knowledge about a large number of virtual characters and to have mastery over them.

Whereas LoL gamers from different countries share common ground, usually they are skilled at playing different sets of virtual characters. Esports takes place in a highly competitive, stressful, and fiercely aggressive virtual environment. Rapid decision-making and lightning-quick responses are crucial, and so a fast network connection is essential. Professional gamers spend most of their time on domestic games, even though international games are easier to play online than offline.2 International differences in mastery of virtual characters have emerged over time, most likely because of national cultures and social environments (Mathur & Pillania, 2014). For example, Chinese gamers have better attack skills, with significantly higher kill, death, and assist rates than gamers from other countries,3 while Korean gamers are better at defense. Indeed, Korean domestic games frequently call for a firm, steady style of fighting.

The literature on information processing and knowledge combination (Adler, 1986; Cohen & Levinthal, 1990) suggests that a culturally diverse team can benefit from its mastery of various virtual characters and its knowledge of play under different rules, thus getting close to the optimal strategy. Players from different cultural backgrounds may not be able to quickly agree on how to play the game, or be able to collaborate smoothly (Eklund, 2015), but such handicaps may be overcome by the jointly-held language and shared mindset of gamer identity. At the same time, the social rules of the field will guide gamers in structuring their own behavior and in interpreting the behavior of others in ways that are considered appropriate and acceptable, thereby reducing the chances of intra-team conflict. In other words, the more salient the gamer identity, the less likely it is that culturally different gamers will categorize one another based on cultural background.

The Moderating Effect of a Salient Gamer Identity

The salience of social identity in a particular situation depends largely on the social meaning of that situation (Blanz & Aufderheide, 1999). We all activate the social categories that are most meaningful to us or that we see as appropriate in a particular social situation according to our motives and goals (Ellemers & van Knippenberge, 1997). Eklund (2015) found that even though disembodiment and anonymity allow gamers to play without regard to any physical world sociocultural rules, there are nonetheless some benefits when playing with others who have culturally similar backgrounds. In essence, cultural similarity facilitates offline interaction and that can make up for the limited social interaction of high-paced game sessions. This suggests that although gamer identity is vital to in-game interaction, it alone is usually not sufficient to maintain a stable out-of-game community, but with the support of other offline social identities, culturally based ones for instance, such a community can thrive (Eklund, 2015). This is why gamer identity salience is of such importance to make multicultural team deliver in semi-virtual settings. Only when gamers see themselves – and fellow gamers – first and foremost as part of the esports community rather than different cultural subgroups, can they effectively communicate, coordinate, and integrate with each other, leading to positive team outcomes.

The salience of gamer identity can be achieved in both the virtual and the physical world, and this can be done in three ways. The first is to increase the intensity of exposure to the virtual world. When this happens, team member national cultural differences become less relevant. As Oakes (1987) suggests, social identity is more likely to be salient when the stimulus is unavoidable. We see this when changes to the rules of the game are made shortly before a tournament commences causing a mind-focusing disruption in the virtual world but having a negligible effect in the real world. When there are last-minute rule changes, professional gamers are compelled to spend considerable time online concentrating on the new rules, strategizing, and practicing. This increases the salience of gamer identity as it pushes gamers to use the game’s communication style, mindset, and social rules, thus minimizing the downsides of team cultural diversity. Therefore, we put forward the following hypothesis:

Hypothesis 1:

The relationship between cultural diversity and the strategy quality of a semi-virtual team is more likely to be positive when the game rules change shortly before a game commences.

The second way to increase salience of gamer identity is to increase the number of characters each gamer plays in the virtual world. Team tasks are completed online via virtual characters. For social categorization based on cultural identities to have an impact, team members would need to attach cultural stereotypes to the virtual characters of other members. This is unlikely to happen when they play with many virtual characters. Indeed, categorizing gamers by their offline cultural identities could lead to cognitive dissonance. In such situations, we tend to cope with dissonance by deferring to the identity exerting the most pressure (Ashforth, Harrison, & Corley, 2008). In the context of professional gaming tournaments, that identity is most likely to be gamer identity. Playing with multiple virtual characters also provides more choices and with that a higher level of perceived control in the virtual world. These factors are essential for identification development (Kim & Kim, 2016), and thus create a strong association between a sense of self and of the virtual side of gamer identity. Gamer identity on the whole is more likely to be salient as a result (Ashforth et al., 2008). In summary, LoL gamers playing with many different virtual characters are less likely to self-categorize, categorize others, or to be categorized by others. Hence, multicultural teams in which gamers in general are playing with many different virtual characters will suffer less from social categorization, leading to the design and implementation of higher-quality strategies.

Hypothesis 2:

The relationship between cultural diversity and the strategy quality of a semi-virtual team is more likely to be positive when team members on average control more virtual characters.

A third way to increase the salience of the gamer identity is to increase gamer identity in the physical world. A team playing at home benefits from audience support. Being cheered on increases the importance of winning, and so also the desire to win (Furley, Schweizer, & Memmert, 2018). In studies of the impact of playing at home, Neave and Wolfson (2003) found physiological markers of aggressiveness and also found that players become territorial. In both cases, teams playing at home are more motivated to win. In addition, playing on home turf is an advantage in that it eliminates the distraction of being in an unfamiliar place (Jones, 2018). All of these elements activate gamer identity in the physical world and facilitate teamwork. Hence, we suggest the following hypothesis:

Hypothesis 3:

The relationship between cultural diversity and the strategy quality of a semi-virtual team is more likely to be positive when a multicultural team is playing at home court.

METHODS

Data

LoL was first released online in 2009 and rapidly gained popularity, making it today the world’s most popular game of its kind. There were some 32.5 million registered players by the time the first global tournament was held in 2011.4 As is the case with other esports, the characteristics and past performance of professional LoL teams are recorded and provided by digital game platforms such as Wanplus, China’s largest esports game-level database. Leaguepedia, the LoL esports Wiki, is another important source of data, which provides demographic information on gamers and teams. We test our hypotheses using data on 15,553 professional-level LoL games played between 2017 and 2020. We dropped records with missing data and the first game of each team in each season as they had no predecessor. The remaining 11,218 games provide a sample of 22,436 team-game observations. Restricting the analysis to teams with at least one foreign player resulted in a final sample of 102 teams, 4035 games, and 8070 team-game observations.

Measures

Dependent variable

An effective team strategy requires an appropriate combination of strong virtual characters. Because there is always a consensus about the optimal roster, the frequency of each virtual character being picked reflects perceived character strength given the rules in force. A character’s seasonal pick rate is the number of games that character was picked during the season divided by the season’s total number of games. Strategy quality is the sum of the seasonal pick rates of a team’s five characters. A higher value indicates that a team managed to assemble a better roster in a particular game. It is one of the most powerful predictors of multicultural LoL team seasonal ranking.

Independent variable

Between 2015 and 2020 roughly 15% of formally registered LoL gamers were not native to the formally registered country of their team. We refer to them as foreign gamers, and over 70% of professional LoL teams had at least one. Following previous research (Hoever, Zhou, & Knippenberg, 2017), we used Blau’s heterogeneity index to measure a team’s cultural diversity (Blau, 1977):

$$Bla{u}'s\,\,Index=1-\sum_{i=1}^{n}{P}_{i}^{2},$$

The number of nationalities among the members on a team is indicated by n, and Pi is the share of team members who are nationals of the ith country.

Moderators

Although LoL developers update the game every 15 days, competition organizers do not change the rules that frequently. The version of the game chosen is at the discretion of the organizers. We identify the version used in each game, and define the variable rule change time as the time between the release date of the version used and the date of the game, which was from 1 to 115 days in our sample. According to the rules of the game, gamers select one virtual character per game. They can keep the same character in the next game or choose another one. We counted the number of characters each player controlled before the focal game, and call the average number of characters used by each five-member team the number of virtual characters. A smaller value for this variable means that team members chose from a smaller pool of virtual characters. The variable ranged from 16 to 56 characters in our sample. A home game is when a team plays in the country where it is registered. A dummy variable named home court was set to 1 in this case, and 0 otherwise. In an actual game, it is possible that neither team is playing at the home court.

Control variables

We included several control variables. First, team harmony, the within-team variance in the amount of gold won. In LoL, each team member chooses a character that is specialized in a particular task that requires a different level of gold to complete. In a well-coordinated team, members leave more gold-earning opportunities to the character carrying the most important task. A small within-team variance in gold usually indicates that the functions of the players have overlapped to a certain extent or that some players did not play their own role but tried to usurp that of another, which may indicate conflict among the players.5 We captured a team’s record by the dummy variable losing game, which is equal to 1 if the team has lost its last game, and 0 otherwise. We use two dummy variables, BO1 if the competition consisted of a single game, BO3 if it was best-of-three, the omitted category being best-of-five. The variable targeted foreign gamer (TFG) is the number of virtual characters mastered by foreign players and banned by the opponents in the team’s most recent game. A higher value for TFG indicates that the foreign player is countered more heavily. We used the World Bank estimate of the percentage of a country’s Internet servers that are secure (SIS) to measure the level of Internet development in the countries where players are based and entered the average for each team (team average SIS). To control for player experience, we calculated professional age, which is the sum of the number of years that the five players on a team have played professionally. Member turnover is the number of team members that have left the team, divided by team age. The larger that number, the higher the turnover. Team familiarity is the number of days between the day on which the last member joined the team and the focal game. Coaching experience is the number of months from when a coach joined the team to the month of the focal game. Same-country coach is a dummy variable equal to 1 if the nationality of the coach is that of the team’s registered country, and 0 otherwise. We also included year dummies to control for year fixed effects, with 2020 as the base year.

Model Specification

The strategies of opponents are not independent because a team cannot use the characters picked or banned by the other team. That lack of independence mandates that we use generalized estimating equations (GEEs) (Liang & Zeger, 1986), because GEEs allow a correlation structure for nested observations of the dependent variable (Zorn, 2001). We used team IDs to connect all observations for a given team, and game IDs to link the two teams playing against one another. The sample contains 102 team IDs and 4035 game IDs. We calculated standardized effect size estimates using robust standard errors (Grijalva, Maynes, Badura, & Whiting, 2020; Gupta & Wowak, 2017). The working correlation matrix was set to be first-order autoregressive, and we used a Gaussian distribution and an identity link function in the modeling. As some factors can simultaneously determine the presence of foreign players on a team and the strategy adopted – the affluence of a game club for example – we used Heckman two-stage modeling to address selection bias and to control for endogeneity (Heckman, 1979). Exclusion restrictions in the selection equation used to predict whether a team has foreign players are the number of team members, the team win rate in the last season, and the GDP growth rate of the country where the team is registered.

RESULTS

Table 1 presents summary statistics and a correlation matrix describing the key variables. The maximum and mean variance inflation factor (VIF) are 3.976 and 1.750, respectively (for the year 2017 dummy), both of which fall well below the rule-of-thumb cutoff of 10 (Ryan, 1997). Thus, multicollinearity is not likely to be an important issue. Table 2 reports the coefficients of second-stage GEEs models. The inverse Mills ratio (IMR) calculated from the first-stage regression is significant in all second-stage models (β = - 0.017, p ≤ 0.050), indicating that we are addressing the potential for selection bias.

Table 1 Descriptive statistics and correlation matrix
Full size table
Table 2 Coefficients of second-stage Heckman GEEs regressions predicting strategy quality
Full size table

Model 1 is the baseline model, model 2 adds the independent variable, and model 3 the interaction variables. In model 3, the coefficient of cultural diversity x rule change time is negative and significant (β = - 0.003, p ≤ 0.010). We calculate effect size, and find that when rule change time is long (above the 90th percentile), a change in cultural diversity from one standard deviation (s.d.) below the mean to 1 s.d. above it leads to a 2.36% decrease in strategy quality, and when rule change time is short (below the 10th percentile), a change in cultural diversity from 1 s.d. below the mean to 1 s.d. above it leads to a 3.63% increase in strategy quality. Hence, Hypothesis 1 is supported. The coefficient of cultural diversity x number of virtual characters is positive and significant (β = 0.008, p ≤ 0.050). When the number of virtual characters is high (above the 90th percentile), an increase in cultural diversity from 1 s.d. below the mean to 1 s.d. above it leads to a 2.56% increase in strategy quality, but when the number of characters is low – below the 10th percentile – the marginal effect is almost zero (- 0.82%). Hypothesis 2 is supported. The coefficient of cultural diversity x home court is positive and significant (β = 0.153, p ≤ 0.050). When a team plays at their home court, a change in cultural diversity from 1 s.d. below the mean to 1 s.d. above it leads to a 1.66% increase in strategy quality, and when the team plays away from home, a change in cultural diversity from 1 s.d. below the mean to 1 s.d. above it leads to a 2.65% decrease in strategy quality. Therefore, Hypothesis 3 is supported.

Supplementary Analysis and Robustness Tests

Due to the limitations of GEEs models, we are not able to establish a formal mediation model. Nevertheless, the data allow us to analyze the relationship between strategy quality and team performance. The ANOVA tests show that the strategy of the winners and top three teams in each season was of significantly better quality than that of other teams (p ≤ 0.001). Further, we can predict a team’s seasonal average winning rate with its seasonal average strategy quality employing IV-Tobit models to account for endogeneity. The results show that a better configuration of characters leads to a higher winning rate (β = 5.460, p ≤ 0.010). Finally, we examine how strategy quality influences the seasonal ranking of multicultural LoL teams, and find a significant positive impact (p ≤ 0.050).

We test the robustness of these findings with several additional analyses. We include an endogeneity control (Chatterjee & Hambrick, 2007). We regress the cultural diversity of a team on number of members, its winning rate in the last season, and the GDP growth rate of the country where the team is registered in the previous year, and include this estimate in regression models as an endogeneity control. In a second test, we change the measurement of our independent variable. Using the Kogut and Singh (1988) index, we calculate the cultural distance between the country of each player and the registered country of the team, and use the average as our measure of the independent variable. As an alternative, we try the maximum cultural distance. For our third test, we control for the potential influence of player age or experience by restricting the number of virtual characters to that used in the previous three games. In a fourth test, we restrict the sample to non-worldwide, regular-season, and post-season competitions. For our fifth and final test, we change the specifications of the correlation structure of GEEs. The results are not affected by any of the above changes.

DISCUSSION AND CONCLUSION

France banned the use of English gaming jargon by government workers in June 2022. The Ministry of Culture found the sector to be rife with language that “could act as ‘a barrier to understanding’ for non-gamers” (The Guardian, 2022).6 We have shown in this study that shared common ground allows gamers from different cultural backgrounds to communicate and collaborate. Although previous research on the relationship between diversity and team outcomes has found that a superordinate identity can be beneficial, the focus to date has been on team identity. Insufficient attention has been paid to newly emerging identities relatively free from national cultural restrictions in this globalized and digitalized era (Sobol, Cleveland, & Laroche, 2018; Wang, Gu, Glinow Von, & Hirsch, 2020). Globalization of organizations, advances in communication technologies, and most recently the isolation of the COVID-19 pandemic have led to an increasing number of semi-virtual teams. To the best of our knowledge, this study is among the first to investigate how a semi-virtual identity may influence the relationship between cultural diversity and team performance.

We look at 4035 LoL games played by 102 teams from 2017 to 2020 to determine how cultural background diversity influences the quality of team strategy. Addressing potential endogeneity and nonindependence of observations, we find that cultural diversity improves the quality of team strategy when the esports gamer identity of a player is more salient. In particular, the relationship between diversity and strategy quality is more positive when there are last-minute game rule changes, when team members play on average with more virtual characters, and when a team plays at home court. The findings of this paper suggest that a new, semi-virtual superordinate identity can help mitigate social categorization effects in multicultural teams.

Notably, the semi-virtual identity discussed in this study is different from the digital identity (Friedman & Wagoner, 2015) or virtual identities that are limited to cyberspace (Wagner, 2007). It is a social identity that participants build in the esports community, one straddling the border between the real world and cyberspace. Unlike mainstream participants in the digital economy (e.g., engineers and artificial intelligence researchers), gamers are a definite minority and in a relatively closed social group – one that has been subject to prejudice on the part of the general public for years. Their social distinctiveness makes their shared identity highly salient (Forehand et al., 2002). The generalizability of our findings to other born-digital, born-global sectors remains to be demonstrated.

We contribute to the body of research on the relationship between cultural diversity and team performance by introducing as a contingency shared non-country-based and semi-virtual superordinate identities. Whereas common ground can also be found in other professional situations (e.g., doctors, teachers), the existing literature leaves unexplored the mechanisms by which interaction between the virtual and physical worlds affects the salience of the semi-virtual gamer identity. We have also clarified how the salience of a semi-virtual identity can be influenced by both online and offline factors. Our findings also contribute to a better understanding of global convergence in international business (Wang et al., 2020) by showing that cultural identities do not have to be eliminated in order for multinational teams to resolve conflict. A better alternative is to develop a country-neutral but task-related identity shared by all members. Recent advances in information technologies have reduced the cost of internationalization, giving rise to more born-global and born-digital firms (Hennart, 2014), making it possible for MNEs to better exploit the benefits of multicultural teams by carefully managing their shared semi-virtual identities.

NOTES

1There were 159 virtual characters in LoL at the end of 2020. Before a game starts, the five gamers in each team can pick five virtual characters with no overlap based on their own expertise and the opponents’ choices. Each gamer is also given an opportunity to ban a character from the entire game; thus, ten characters are dropped out of the character set in each game. Each team has 27 seconds to adjust its character choices to counter its opponents’ choices before the game starts. The ban and adjustment decisions are mainly based on knowledge about the opponents’ expertise and preferences.

2LoL teams play on local servers in their home countries for 7 months of the year. Only the top teams in each country are eligible to play in international competitions.

3In an analysis of variance, Chinese LoL gamers had a within-group kill rate variance significantly smaller (5.63) than between-group variance with other nationalities (1032.18, p ≤ 0.001). Other indicators showed similar differences.

4Data were gathered and arranged by the authors from https://www.riotgames.com.

5Take traditional sports as an example. If a striker and a goalkeeper on a soccer team run a similar distance in a match, there must be something wrong with the team.

6France bans English gaming tech jargon in push to preserve language purity | France | The Guardian, accessible at www.theguardian.com/world/2022/may/31/france-bans-english-gaming-tech-jargon-in-push-to-preserve-language-purity.