Размещено на http://www.allbest.ru/

Team diversity and performance: evidence from international sport leagues

Panchishina Alena Alekseevna

Contents

Introduction

1. Cultural diversity and performance. Theoretical foundations and review of previous studies

1.1 The definition of diversity

1.2 Different approaches applied to evaluation the impact of diversity on performance

2. Empirical Framework

2.1 The data

2.2 The descriptive statistics

2.3 Estimation procedure

3. Empirical results

3.1 The regression model results for the regular season in hockey

3.2 The regression model results for the regular season in basketball

Concluding discussion

References

Annotation

The current research aims at evaluating the impact of cultural diversity of teams of the international sports leagues (the KHL, the VTB United League) on the team performance. The research is based on the data from 8 regular seasons of hockey and 6 regular seasons of basketball since these sports require the highest interpersonal interaction between team players. This study fills the gap regarding the diversity-performance nexus in Russia. In the empirical test regression models are estimated using OLS with fixed effects for teams and seasons; the basic diversity metric is HHI. The main finding is that it is proved that cultural team diversity is an essential factor of boosting team performance both in hockey and basketball. sport cultural hockey

Целью данного исследования является оценка влияния культурной разнородности команд международных спортивных лиг (КХЛ, Единой лиги ВТБ) на результативность команд. Исследование основано на данных 8 регулярных сезонов хоккея и 6 регулярных сезонов баскетбола, так как эти виды спорта требуют наиболее высокого взаимодействия между игроками команды. Это исследование заполняет пробел в отношении исследования взаимосвязи разнородности и результата команды в России. Регрессионные модели оцениваются МНК с фиксированными эффектами для команд и сезонов; основной метрикой разнородности является индекс Херфиндаля-Хиршмана. Наиболее важный вывод состоит в том, что доказано, что культурная разнородность команд является существенным фактором повышения эффективности команды как в хоккее, так и в баскетболе.

Introduction

According to the current trend, supporting and advancing international business is regarded as the key driver of the country's economic growth. The assumption concerning the higher profitability of international companies compared with national ones is based on the following facts: firstly, they are independent from different cultural peculiarities of the country and obtain the highest level of corporate culture; secondly, the global entities possess one of the most valuable asset such as diverse in different dimensions human resource; finally, MNEs are aimed at economic efficiency (Khanna; 2016). Thus, team diversity is assumed beneficial for the financial state of the entity.

Moreover, according to the research conducted by McKinsey in 2015 on 366 public companies, the companies which are ethnically and racially diverse in management were 35% more likely to make profits above the industry average. So, team diversity might be an advantage either for the staff or management.

Global business is the most economically viable since it covers the worldwide spread information, the global financial market, and the global structure of innovation. One of the most substantial advantages of the global business is that companies have an opportunity of hiring employees who belong to dissimilar cultural groups. In other words, the working teams in international companies can be built by combining representatives of the national population with foreign employees who possess diverse professional skills and knowledge, that make culturally heterogeneous teams more productive in comparison with the homogeneous staff in national business.

Despite the fact that hiring culturally nonhomogeneous staff have some positive effects on financial state of the entity, specifically: highly qualified experts with diverse and highly worked skills in a related field are expected to provide the team with both innovative and suitable approaches to performing tasks and thus enhance the team's productivity and company's profits; such incorporation might entail additional costs for integrating new staff and then eliminating misunderstandings between the employees due to different languages, cultures and attitudes (Lazear,1999).

The problem of building the pool of competent, versatile, multi-skilled team is faced by both commodity producers and sports industry insofar as manufacturers and sportsmen equally strive to surpass competitors in order to earn more and to rise in the rankings. Therefore, managers must cope with the problem stated above, specifically whether foreign team members should be integrated to the working team. The managers tend to solve the problem by comparing the potential profits with losses caused by this kind of incorporation and as a result they conclude with the ideal team formula. (Bishop Smith, Yuan Hou; 2014).

The aforesaid issue is especially crucial for team sports in which interaction between team members plays a major role in the success of the entire team.

According to the fact, that in hockey close relationship, communication, strategy development are intrinsic parts of training and competing, besides, hockey players strive to achieve the common goal to beat the opponent; the effect of diversity on team performance is investigated in hockey, particularly, in the Continental Hockey League. In addition to hockey, the diversity-performance nexus is studied in basketball on the basis of the VTB United League for the same reason: there is a high level of interpersonal interaction between basketball players which should be made efficient and proactive to succeed.

Moreover, sports industry adheres to the policy of transparency of the team's characteristics and performance statistics, so the data on sport is more available and fuller than in any other industry.

In the present paper we aim at contributing to the relatively recently started discussion on whether cultural heterogeneity of team members is an essential factor of boosting team performance in international business or this statement is just a coincidence.

We base our research on the data from the KHL. The KHL was chosen for the following reasons. We believe that the KHL provides us with a better setting for an empirical test. First, the proportion of foreign players in the KHL is higher than in the NHL. Second, those players come from a wider set of counties (for instance, 29 nationalities are presented during 8 latest seasons in the league), which makes our inferences more relevant. Third, our main contribution is that we analyze diverse teams which perform in different countries due to the peculiarity of the KHL, while the previous studies investigated the teams consisted of players who come from different countries and work together in the same county. Hence, we research the relevance of diversity in conditions of co-working (performing) in diverse countries.

In order to investigate the issue under study from the point of view of the sport with other specifics of players' interaction another dataset is included and tested. The effect of team diversity is also evaluated in basketball. Firstly, the league adheres to the policy of financial transparency (team budgets are disclosed annually); secondly, the variety of nationalities presented in the league is even greater than in hockey (43 nationalities in the league during the 6 latest seasons versus 29 ones); which will lead to more reliable outcomes.

The main reason for examining the KHL and the VTB United League is that, in general, various aspects of the issue of interdependence of team diversity and performance have already been studied on the basis of the different industries and countries. However, the largest and one of the most multinational countries such as Russia left unattended. In this research I fill the gap regarding cultural diversity and performance in Russia using the database from the KHL and the VTB United League.

The aim of the research is to evaluate the impact of cultural diversity of teams of the international sports leagues on the team performance and then compare the obtained output.

The current paper is the extension of the previous paper, which has been aimed at measuring the effect of cultural diversity of the hockey team on its productivity. Thus, we base the current research on the previous one including the following upgrades: firstly, the basketball database is added in order to compare the effects of both chosen leagues; secondly, the issues encountered previously have been tackled; finally, some new approaches to the measurement of team diversity and performance has been suggested. The continuation is supposed to make the research more comprehensive and provide us with more solid inferences regarding the research question.

According to the results obtained from the vast number of papers on the diversity-performance nexus, we put forward the following hypotheses.

H1: Cultural team diversity affects team performance.

H2: The effect of cultural team diversity varies depending on the sport.

The first hypothesis can be explained the following way: the increase in number of cultures presented in the sports team provide performance gain in the efficiency of the team since foreign players obtain the different skillset and can introduce innovative strategies. In addition, a foreign player can carry out functions and have personality traits, which are necessary for the team's success, but these skills are not represented in the native population team (Lazear,1999). Meanwhile, the impact of adding foreign hockey players to the team on their output might be either adverse as integration and transaction costs are relatively high in comparison with the benefits obtained from hiring an additional foreign player (Lazear,1999).

The second hypothesis derives from the difference of sports and the specificity of team members' interaction during training and games.

The paper consists of five sections. There is the literature review after the introduction part. Methodology and data applied in the research are presented in the third section. The obtained results regarding the performance-diversity nexus are analyzed in Section 4. The last section discusses the outcomes and points the limitations.

1. Cultural diversity and performance. Theoretical foundations and review of previous studies

1.1 The definition of diversity

Team diversity can be defined as the dissimilarities between individual members of a team on various dimensions such as physical (weight, height), demographic (age, gender), cultural which includes nationality, religious background, language; educational differences and labour diversity (functional background, task skills). In this paper we focus on cultural, in particular, linguistic nonhomogeneity.

The pros and cons of diverse teams

When composing a working team, managers should compare inputs of foreign employees such as transaction costs with outputs which refer to results of co-working (Lazear, 1999). So companies can benefit in many ways from hiring diverse employees. Firstly, adding employees who differ in some dimensions from the native workers may challenge their brain to overcome its stale ways of thinking and sharpen its performance. Secondly, diverse teams are more likely to conduct a permanent reevaluation of facts and as a result these teams remain objective. Besides, by breaking up workplace homogeneity, native employees are given an opportunity to become more aware of their own potential biases - ingrained ways of thinking that can otherwise distract them from major information and even lead them to make serious mistakes in decision-making processes.

Another important contribution of culturally diverse teams is that they can be a boon to innovativeness. For instance, according to the London Annual Business Survey, in which 7,615 firms participated, businesses run by culturally diverse leadership teams were more likely to develop new products than those with culturally homogeneous management.

Thus, expanding the employee pool with representatives of different genders, races, and nationalities is crucial for stimulating joint intellectual potential of the company (Rock, Grant, 2016).

According to some theoretical papers, team diversity might be beneficial for firm productivity in the case of complementarity stimulation between highly- and lowly-skilled workers; skill and knowledge spillover between novice and highly-experienced workers; the probability of creating more comfortable workplace (educational or skill diversity might be appreciated by employees); or demand stimulating and investment attracting (diverse workforce may be more preferable for customers and investors). Notwithstanding aforementioned benefits, some drawbacks of team diversity are revealed. For instance, it will cause misunderstandings, communication difficulties and interpersonal conflicts between team members - all that issues subside performance (Akerlof, Kranton, 2000).

If we consider the issue from another angle, specifically from the employees' viewpoint, team diversity might be advantageous as well as detrimental. Team diversity might lead to the inefficacy due to an unpleasant working atmosphere. Yet employees can overcome this obstacle through earning higher wages. As the competitive labour market theory goes, employees get the salary according to their marginal revenue products. Consequently, if team diversity can influence productivity, it may also have an impact on the wages team members earn. However, it is not the case of the current research since the paper is aimed at estimating the diversity-performance nexus. But this finding should be taken into consideration, so when conducting the empirical test we should control for wages.

Another suggestion is put forward by Akerlof and Kranton (2000) concerning benefits and shortcomings of team nonhomogeneity implies that the productivity of employees who join a team depends positively or negatively on the proportion of team member who belong to the same (or different) social category. The researchers have focused on gender team diversity so they conclude according their research question, however their findings could be applied to any dimension of diversity. It is inferred that the presence of the dominant group adversely influence the whole team efficiency. It should be mentioned that not only the similar in some kind members but also different members can comprise a dominating group. In particular, it is anticipated that growing gender nonhomogeneity might have an adverse impact on the firm performance, especially if men comprise a socially `dominating' group (Haile, 2012). Thus, team diversity can cause some obstacles despite the fact that the firm might benefit from hiring diverse employees. The reasonable question arises how the optimal proportion of diverse team members can be evaluated in order to maximize team performance thereby boosting the firm profit.

1.2 Different approaches applied to evaluation the impact of diversity on performance

As it have already been mentioned the question why some groups are especially productive in accomplishing the goals while others are unsuccessful performing the same tasks has become crucial currently. Thus companies are concerned about discovering new ways which lead to outperforming their competitors. Diversity is considered to be one of the factors of increasing team performance (Stura, Lepadatu, 2014).

The issue of interdependence of heterogeneity of a group of collaborating people and the efficiency of so called the team has been researched insufficiently so far due to the novelty of this problem. Nevertheless, some studies have already been conducted on this field. They are devoted to investigating empirically the differentiated teams by various criteria and their results.

The first study "Globalization and the Market for the Team-Mates" was carried out in 1999 by Lazear; so this paper can be considered as the beginning of studying the problem. The paper provides us with the seminal theoretical model. The most important contribution of this paper is that the author proposes the notion of a global firm, which has become critical due to the globalization, as a group of teammates who come from different countries and adhere to various cultures. Also it is supposed that forming multicultural team imposes significant costs on communication, transaction; consequently, the management should compare possible costs with potential benefits from hiring multinational staff. Lazear concludes that the diverse team is able to generate productivity gains only if the following three factors are present. Firstly, team members must have different skills, abilities, or information. Secondly, the different skills, abilities, or information of team members must be relevant to one another. Thirdly, communication is vital for team members to perform the relevant joint tasks and engage in knowledge transfer to amplify effectiveness. Lazear's argument that productive teams should be diverse along the skills, ability, and information criteria, but homogeneous in other dimensions, such as demographic characteristics, that minimize communication costs or what he calls `costs of cross-cultural dealing' seems to be quite reasonable.

However, assuming that diversity has positive impact, why do most companies not implement this strategy? According to Lazear, it is connected with the excess the costs over the benefits, but are there any other reasons for that? Thus, this study requires further empirical evidence.

The theoretical model suggested by Lazear received the subsequent developing through more practical model which was devised by Ottawiano and Peri in their paper "The economic value of cultural diversity: Evidence from U.S. cities" (2006). This study focuses on the potential benefits received from hiring immigrants and thereby creating diverse workforce. The authors investigate whether cultural, in particular, linguistic diversity across U.S. cities affects the wage of the average U.S.-born staff, and they also qualify the effect on particular sub-groups of the U.S. citizens (e.g. black workers versus white workers, more educated versus less educated). Similar to Lazear's paper, this study put forward the hypothesis according to which the heterogeneity of cultures implies diversity of production skills, abilities and professions that enhances the productive performance of a city. The research reaches the conclusion that the performance at work of native U.S.-born citizens is higher in those U.S. cities in which a great percentage of foreign-born citizens resides.

In general, this study provides us with the empirical model based on the average wage of the homogeneous in age over 1,200,000 staff of metropolitan in 3 periods, however, taking into account only average wage as the result of collaborating diverse workers is imprecise due to the fact that many other factors can influence higher salaries in some cities or lower ones in others in the same industry. Another limitation of this work is that the interdependence of diversity and performance is considered on statistical level; hence the results cannot be applied on less-scale level, such as firm level.

The next research conducted by Hamilton, Nickerson, Owan in 2004 eliminates the limitation of the previous study. This paper provides another framework and concentrates on age and ethnical diversity and regards 9 nationalities. The data for the research is the personnel records of workers gathered from the California apparel sewing factory; thereby the findings of this study can be applied on intrafirm level. The authors observe individual productivity data in order to be able to econometrically distinguish between the impacts of diversity in worker abilities and demographic diversity. The key findings of this paper are: firstly, working teams with more heterogeneous worker abilities are more efficient; secondly, remaining the distribution of team ability constant, teams with greater diversity in age are less productive, and those composed only of one ethnicity (Hispanic workers in the paper) are more productive. Nevertheless, this paper includes some limitations concerning the issue whether the findings are applicable to the firms where teams are engaged in more complex problem-solving tasks, since in the study they investigate team collaboration in accomplishing tasks during the relatively simple production technology. Besides, the metric of ethnic diversity seems to be imprecise inasmuch as there are plenty of nationalities with its specific features beside Hispanic and non-Hispanic. Thus, these results are not robust to alternative model specifications.

In another research conducted by Leonard and Levine (2004) the issue of interdependence of diversity and performance is investigated from another angle. They add gender diversity to ethnical and age heterogeneity. The study is based on the information from undisclosed firm's 800 stores in the retail industry. As performance indicator, sales and sales growth are employed. This approach controlling for individual FE leads to the following insights: estimated coefficients turn out to be insignificant for the gender and ethnical heterogeneity in their impact on the dependent variables (sales and sales growth). In contrast, being in a numerical minority, especially it concerns the ethnicity, increases the turnover. Basically, this study excludes inaccuracy connected with disclosure of the notion `diversity' through considering 3 types of heterogeneity: age, gender and ethnicity. Nevertheless, one of the limitations of the research is that gender turns to be unimportant indicator due to the specification of the industry under study. Consequently, some more research should be carried out on this field

The most of aforementioned papers include significant limitation concerning the evaluation of company's performance. Therefore a new approach is applied: they examine sport teams instead of companies from the productive sector. The approach assumes using data from sport industry, particularly team sports, and measuring team performance by their game statistics since they are considered to be objective and comparable.

Kahane et.al. paper (2013) investigates the issue on the basis of the data from the NHL (National Hockey League) teams and players' data. For gauging the team performance the following indicators are considered: team skill level, relative payroll, coaching input. Team diversity is measured using Herfindahl-Hirschman index which is based on the shares of a team's players belonging to the country of their birth (Europeans and North Americans). The study reveals the following findings. On the one hand, the NHL teams that employ a higher proportion of European players perform better. On the other hand, the results also indicate that teams perform better when their European players come from the same country rather than being spread across many European countries. This assumption is probably based on the excess of integration costs, implying language and cultural differences between teammates, over any gains from diversity, as it is suggested in the theoretical model developed by Lazear (1999). The authors of this paper take into consideration the limitations of the previous studies, moreover, the results seem to be plausible, hence we can base our research on this paper. Kahane examines the NHL, but what can be inferred investigating the data from the KHL (Continental Hockey League) which is less prestigious and include sportsmen from the large spread of countries? This issue turns to be unexplored so far.

As for studying some other team sports, baseball and basketball are examined by Timmerman (2005). The data was gathered from 1950 to 1997 on the individual level and then aggregated to the team level. The diversity metric is Blau's (1977) index (diversity=1-?pi2, where p is the share of the members of a racial category; i is a number of different categories in a team). Performance metric is the percentage of winning. The results reveal that racial diversity is negatively associated with basketball team performance. However, the same dimension of diversity turns to be unrelated to team productivity in baseball. The author explained the inferences that basketball and baseball require task with strong and low level of interdependence respectively. So, the obtained results are controversial, and the effect of diversity should be measured at the more disaggregated level (race replace with nationality or language group).

The findings from the study based on the data from 5-season period of the professional soccer reveal that there is not empirical evidence proving that national diversity between team members significantly influences team performance. However, it is noteworthy that with the increase in number of cultures presented in the defensive block team productivity deteriorates ceteris paribus (Brandes et al., 2009).

Some alternative metrics for measuring cultural diversity are employed in the paper focused on estimating diversity-performance nexus on the basis of the German soccer (Ben-Ner at al., 2013). Specifically, the diversity is gauged by ethnic, national and linguistic background of football players. Performance is measured through game scores on the team level and objective performance ratings on the individual level. The authors include the following control variables: players' talent which is measured as average performance rating in all the Bundesliga games; position (two classifications: (1) forward, midfield and defense, and (2) offensive players, and defensive players); team fixed effects; the nationality, experience, number of substitutions and tenure of manager/coach; average players' age; average team tenure (the duration of the average contract).

Beside searching the answer to a question on the effects of team diversity on team productivity, the study concentrates on measuring the effects of diversity in the case of varied tasks. Moreover, the issue under study is explored in the terms of the joint tenure on a team.

In general, the obtained results indicate that the effect of diversity on performance is not large. The significant effects of team nonhomogeneity can be revealed in the case of disaggregated into subgroups teams. Team diversity become significant when the teams are subdivided on the certain basis such as the place of birth; position the player take or the task the player carries out according to the strategy applied; the total time the player spent together with the certain team; and controlling for team fixed effects and average team `talent'.

A prolonged stay of German players in conjunction with team diversity enhances the team productivity as well as individual performance, while the opposite has the adverse impact, and with more force, for non-German players.

Besides, the outcomes indicate the positive impact of diversity among defensive players; this effect is amplified by longer cooperation of defenders, while the opposite is untrue for the offensive block.

Hence the aforementioned studies based on sport teams data shows that the effects of the cultural diversity on team performance depends on the nature of the fundamental tasks the players should accomplish and some other factors which should be concerned in the further research in this field.

The issue of the team's diversity and performance is studied not only on the basis of the common sports, but also, they analyze it basing the empirical test on exotic sports. For example, an unusual approach is suggested in Rose and Frick's study (2017). The subject of their research is the diversity-performance nexus of Himalayan expedition team. The inferences include that the probability of a team success is positively influenced by cultural heterogeneity. Despite the extraordinary field under study, this paper suggests that the expedition finishes successfully if all the climbers survive; besides in the industry there is no competitiveness between teams, so it makes it difficult to assess the effect of diversity correctly.

The diversity-performance nexus has been examined on the basis of eSports data (Parshakov et.al. 2018). The authors discuss some dimensions of non-

homogeneity: diversity of culture, diversity of language and diversity of skill. The key findings include the approval of fruitfulness of cultural diversity for the team's performance. For instance, the absence of diversity decreases the productivity by 30%. It is also found out that linguistic and experience (the proxy for it is the team's average age) diversity adversely affect performance. Due to the ambivalence of the obtained results, the authors advise companies to sensibly increase team diversity.

To sum up, a sufficient number of various industries including sport have been examined to evaluate the impact of national diversity on team productivity. The closest subject of the research to the one of the present paper turns to be in the Kahane et al. paper (2013), thereby we base it on this study; however, due to specification of the KHL we suggest different empirical models. In addition, some metrics for diversity and estimation strategy can be integrated from the Ben-Ner's paper, which analyzes 10 seasons of the German football league.

While exploring various dimensions of diversity, it is worth remembering about quotas - the restrictions imposed on the number of diverse team members - since these quotas influence the diversity metrics, thereby the restrictions affect the results of the research.

So, there is a study which is aimed at revealing the impact of quota rules on the group cooperation (Dorroug et al., 2016). Group cooperation is considered to be one of the most essential aspects of team performance. The authors test the hypothesis about the positive effect of gender diversity. Nevertheless, it is a controversial question whether team contributes from growing team diversity in the terms of quota rules. The researchers conduct two experiments which involve a real-effort task. The study considers the impact of quotas on group collaboration as compared to promotions based on productivity. Thus, participants are classified either by gender or by artificial category, which has been assigned randomly. The findings indicate that collaboration within groups reduces when promotion is based on quota compared to the promotion based on the results of the work, regardless of the criteria of categorization. Further analysis shows that this negative impact of quota rules on cooperation occurs not due to procedural fairness or anticipations about the efficacy of the promoted team member.

Thus, as the estimated effect is independent from categorization criteria (gender vs. artificial), the results can be applied to the sport industry. Put different, quota rules might cause a decrease in team performance.

However, the paper includes significant limitations such as an unrepresentative sample, concentration on the psychological aspects of the diversity-performance nexus in the terms of quotas instead of focusing on economic issues. Also due to designing artificial experiments it is a complicated task to apply the results to the real-life cases. Moreover, in the real life co-workers are not aware of candidates' professional qualifications and they are often uninformed about the involvement in quota rules. Such lack of information may cause lower anticipations regarding the performance of the incoming person promoted by quota (Heilman et al., 1992).

2. Empirical Framework

2.1 The data

The database consists of two parts: the data on basketball and on hockey. Namely, the data on hockey was gathered manually using the official website of the KHL http://www.khl.ru. We got the data from 8 seasons from 2009/10 to 2016/17. Initially, the database includes the KHL players' individual characteristics, specifically, demographic ones: such as the nationality, according to the players' citizenship, and age; and physical characteristics: current weight and height. So, the data on individual level comprises 2292 observations for hockey. Then we aggregated players' data to team level as we aim at estimating the effect of team diversity on team productivity. After aggregating the database on hockey consists of 209 cross-section observations.

As for basketball, the data was also gathered manually from the official website of the league http://www.vtb-league.com. The data obtained from 6 latest seasons excluding the current one (since the season 2018/19 is still in progress at the moment of writing the paper) from 2012/13 to 2017/18. Similarly to hockey, firstly we got the league's players' individual information and then aggregated it to the team level. So, the unaggregated data comprises 1480 observations, aggregated data consists of 98 cross-section observations.

The information about team budgets of both leagues under study which is required as the control variable was searched in the news section of the website https://www.sports.ru and in the news portal https://rsport.ria.ru. According to the KHL policy, the league does not require its teams to disclose the budgets. Thus, the budget information on hockey included missing data. However, we found out the team budgets' data for the Russian teams for seasons 2012/13 and 2013/14 on the reliable Internet sources. The remaining part of missed data was filled according to the methodology which will be described below.

In contrast, VTB United League adheres to the policy of the financial report transparency, therefore all the league's teams must reveal their budgets before the oncoming season. The budgets were actually publicly disclosed for the Russian teams; the missing data for foreign clubs was filled according to the methodology presented below.

Another control variable requires the info about the main coach, in particular, his nationality according to the citizenship.

Also, we collected the teams' game statistics (such as number of games played per each season, number of gained scores and number of playoffs per season) from the official websites of leagues mentioned above.

After description the sources of the data for the empirical research, it's logical to move to the definition of what is implied under diversity and performance and how these variables are measured.

In the previous studies they consider and assess different diversity metrics: the share of North Americans and the share of Europeans in the team (Kahane et al., 2013); the number of representatives from different countries (Brandes et al., 2009), Herfindahl-Hirschman concentration index (HHI) (Ben-Ner et al., 2013).

In this paper linguistic diversity is the proxy for the cultural diversity since most misunderstandings arise due to the language barriers. Language is most closely connected with thinking, which cannot be separated from speech activity: in other words, while thinking, person says mentally. However, the level of detailing and preciseness of the information reported in the conversation varies between different languages, so it might cause misunderstandings and interpersonal conflicts in a team (Nickerson and Goby, 2018).

As it has been stated earlier, cultural diversity is a multidimensional notion that's why together with the language, the effect of some other observable but too complicated to be identified, grouped, and measured separately unique features which are inherent to the culture is evaluated in its impact to team performance (for example, mentality, mindset, the style of communicating and some other).

Language diversity is considered from two dimensions: disaggregated and aggregated one. The disaggregated level of diversity implies the nationality of the player as it is given in the player's bio on the official website; whereas, the aggregated diversity is based on the methodology suggested by Parshakov et al. (2018). According to the authors, they group together players who come from the post-Soviet countries, as they are supposed to speak Russian fluently. Also, the Americans, the Canadians, the Australians and the English are supposed to be fluent in English that's why they are considered to be similar at the aggregated level. For example, when distinguish players on the disaggregated level the Russian and the Belarussian are as different from each other as the American and the Australian. When the players are aggregated, the both mentioned pairs are regarded as being belonged the same nationality.

Nevertheless, we apply two-dimension approach to diversity to basketball, and identify diversity solely at the disaggregated level for hockey players, inasmuch as the number of nationalities presented in the VTB United League at the considered period is much higher than in hockey, however, the sample for hockey is larger.

Among the metrics suggested by the previous papers, HHI is chosen as the diversity metric. In addition to it, we calculate number of nationalities per team. Aggregated and disaggregated HHI and number of nationalities are calculated for basketball, while for hockey we calculate both metrics at the disaggregated level.

Generally, there are two basic explanatory variables: HHI_nat and N_nat; and two additional ones for robustness check: HHI_lan, N_lan.

· HHI is the market concentration index which is used to measure the degree of monopolization of markets, to determine the structure of export, import, domestic trade and other indicators where it is necessary to estimate the level of competition. In the general case, it is calculated as the sum of the squares of market shares (sales shares) of firms in the industry. HHI varies from 0 to 10 000; the smaller the number of firms in the industry the larger the share of them, the higher the concentration of the market, the closer the market structure to monopoly.

The same idea underlies the calculation of HHI in the present research. However, in order to consider the actual diversity, we include the total time played by each player to the formula. So, the applied formula is as follows:

(1)

where n is the number of nationalities presented in the team in the particular season.

It should be also stated that if HHI=1, it means that there is no diversity and the team consists of one nationality solely; consequently, as the value approaches 0 diversity is increasing.

Another diversity metric is number of nationalities (N_nat).

· Number of nationalities implies the number of unique nationalities whose representatives play for the certain team in the particular season.

, (2)

where n is the number of nationalities in the team per season.

It should be stated that the aforementioned variables are calculated for each team in every season under consideration.

Similarly to the notion of diversity, the definition of performance is multifaceted. In the previous studies authors applied different proxies for determining team performance, for instance, profit (Leonard, Levine, 2004), matches won (Kahane, 2013), rank (Brandes et al., 2009), game scores (Ben-Ner et al., 2013).

In the present study we use scores per game in the regular season as proxy for performance. As a result, we estimate the effect of diversity in its impact to the team's success in the single game, but not success in the whole regular season.

(3)

It should be pointed out that we measure the effect of diversity in the regular season and do not consider play-off matches which follow up the regular ones, because of insufficiency of the number of games and the differences in the gaming scheme (in the regular season there is a round-robin system; but in the play-offs there is a "knockout" system).

As for the control variables applied in the current paper, basically, we build on the economic model suggested by Kahane et al. (2013). According to the authors, the team's success depends on the players' characteristics, the coach's characteristics, the size of the team's budget.

So, in the research we control for the following variables: physical characteristics such as average team's weight, height and age; coach/manager features (coach nationality); and financial metrics (team budget).

Some comments about the methodology of estimating each applied control.

1) Average team's weight is the sum of all team's player's weight divided by the number of players in the team.

2) Average team's height: similarly to 1).

3) Average team's age: similarly to 1).

4) Coach's nationality: it is a dummy variable, if the coach of the same nationality as the team, the value is 1, otherwise the value is 0.

Similarly to the calculation of HHI, we use two-dimension approach basing on the same logic. So, there is dummy for coach's nationality at aggregated and disaggregated level.

5) Team budget: as it was stated above, there was some missing data both in the sample for basketball and hockey. So, we managed to cope with this problem by introducing our own methodology.

The data on basketball had not been disclosed only for the foreign teams. So, we got the information about the capacity of the home arenas of the Russian teams and correlated it with team's budget. The correlation turns to be high for all team except for CSKA (Moscow) and Khimki (Moscow region), which are officially sponsored by the largest Russian corporations such as "Nornikel" and "Gazprom" respectively. We exclude these two team and calculated the average coefficient as:

(4)

Then we applied the average coefficient to the calculation of foreign teams' budget.

(5)

As for the hockey sample, the aforedescribed method is not applicable to it since the arena capacity and budget are not correlated. Moreover, the data was missed for 6 seasons out of 8.

According to the KHL's strategy, the Russian clubs are financed by the region's budgets. That's why we gathered the data on budgets of those regions where the clubs are located. Then we calculated the share of team budget in the regional budget for two disclosed seasons, found the average, and filled the missing data. As for the foreign hockey club, the methodology that is similar to basketball one was used. As a result, we filled all the missed data on team budgets, which was the big problem in the previous study.

To sum up, HHI and N_nat are explanatory variables. Dependent variable is scores per game in the regular season. The variable description is presented in the Table 2.

Table 2

Variable description

Type of variable	Variable	Description
Dependent	Scores per game	The sum of scores gained by the team during the regular season divided on the number of games during the regular season.
Control	Weight	Average player's weight (based on all players of the team's roster).
	Height	Average player's height (based on all players of the team's roster).
	D_coach_nat	1, if the main coach belongs to the same nationality as the team 0, otherwise.
	Budget_mln	The sum of players' and coaches' salary costs.
	Age	Average player's age (based on all players of the team's roster).
Explanatory	HHI_nat
	N_nat	,

Before estimating the equations we have checked the sample for statistical outliers through making boxplots. So, after conducting preliminary data analysis the sample does not contain outliers thereby it is considered to be homogeneous. In addition, there are not values those contradict common sense (for example, HHI is higher than 1, the share of scores is an adverse figure, player's age, height and weight are not corresponded to the common expectations).

2.2 The descriptive statistics

The sample for hockey comprises 209 observations for regular season (including cross-section data is 34 teams and 8 periods). The data for the research is gathered from the eight latest regular seasons of the KHL from 2008/09 to 2016/17. For estimating the equations we employ panel data.

The descriptive statistics of the non-dummy variables are presented in the Table 3.

Table 3

Descriptive statistics of variables for hockey

Variable	Obs.	Mean	Std. Dev.	Min	Max
Score_per_game	209	1.5017	.35828	0.6667	2.3167
Age	209	27.1989	1.7661	23.6119	34.7279
Height	209	182.1057	3.3130	165.0495	186.6557
Weight	209	85.2848	2.7392	75.4963	91.2870
N_nat	209	5.4162	1.6910	1.0000	13.0000
HHI_nat	209	0.52934	0.1397	0.1430	1
Budget_mln	209	559.9007	307.682	127.1098	1540.099

Ncountries is related to measurement of diversity, so according to the statistics from the Table 3, number of countries from which players come varies from the minimum 1 (it means that only native players are presented in the team roster) to the maximum 13 countries in one team.

The sample for basketball comprises 98 observations for regular season (including cross-section data of 26 teams and 6 periods). The data for the research is gathered from the six latest regular seasons (excluding the current one) of the VTB United League from 2012/13 to 2017/18. For estimating the equations, we employ panel data.

The descriptive statistics of the non-dummy variables are presented in the Table 4.

Table 4

Descriptive statistics of variables for basketball

Variable	Obs.	Mean	Std. Dev.	Min	Max
Score_per_game	98	80.6110	6.2979	66.7778	95.5333
Age	98	25.9559	1.4010	22.1818	29.5000
Height	98	198.4029	2.6424	186.6429	202.4000
Weight	98	95.7637	3.1266	89.000	107.4375
N_nat	98	3.8469	1.1429	1	7
N_lan	98	3.2449	1.0058	1	6
HHI_nat	98	0.4099	0.1416	0.2008	1
HHI_lan	98	0.4591	0.1449	0.2319	1
Budget_mln	98	601.0404	563.4794	60	2759

The more detailed descriptive statistics is presented below.

Figure 1. The dynamics of the hhi_nat during 2009/10 to 2016/17 season (hockey)

Figure 2. The dynamics of the hhi_nat during 2009/10 to 2016/17 season (basketball)

As it can be evident from the Figures 1-2, in hockey diversity tends to fall (as hhi grows); in contrast, in basketball diversity has a tendency to grow (hhi decreases).

Figure 3. HHI_nat distribution (hockey)

Figure 4. HHI_nat distribution (basketball)

According to the Figures 3-4, hhi_nat is approximately normally distributed for both hockey and basketball.

Table 5

Top-5 clubs in the KHL with the highest diversity

Club	HHI_nat
Donbass (Donetsk)	0,144064
Kunlun Red Star (Beijing)	0,192156
Medveљиak (Zagreb)	0,327385
Lev (Prague)	0,350452
Barys (Astana)	0,35774

Table 6

Top-5 clubs in the VTB United League with the highest diversity

Club	HHI_nat
Uniks (Kazan)	0,271224
Spartak (Moscow)	0,294904
Khimki (Moscow region)	0,328023
Lokomotiv (Krasnodar)	0,336217
PGE-Turov (Poland)	0,337981

2.3 Estimation procedure

In the present study the diversity-performance nexus is empirically tested separately for hockey and basketball, then the obtained effects are compared.

In the empirical test, which is aimed at measuring the impact of cultural diversity measured through 2 metrics on team performance, we estimate regression models using OLS with fixed effects for teams and seasons.

Diversity metrics are correlated with each other noticeably and moderately according to the Chaddock's scale, thus, they are put into regression equation separately in order to avoid multicollinearity. The correlations between explanatory variables and dependent variables are presented in the Tables 7-8.

Table 7

Correlation analysis results for hockey

	Scores per game	hhi_nat	n_nat
Scores per game	1.0000
hhi_nat	-0.0970	1.000
n_nat	0.1220	-0.6225	1.0000

Table 8

Correlation analysis results for basketball

	Scores per game	hhi_nat	n_nat
Scores per game	1.0000
hhi_nat	-0.2141	1.000
n_nat	0.1654	-0.7332	1.0000

Before estimating the regression coefficients five Gauss-Markov conditions have been verified, so the estimated coefficients are BLUE (best linear unbiased estimate). In order to obtain the most accurate estimations we calculate robust standard errors by employing White covariances and standard errors.

The following regression models are aimed at evaluating the effects of diversity on the team performance measured through scores per game in the regular season.

The models (6) and (7) are designed for estimating diversity-performance nexus, where diversity is considered at the disaggregated level. The models (6) and (7) are applicable for both hockey and basketball.

(6)

(7)

Robustness check

Some procedures aimed at ensuring robustness of the applied regression models had been conducted. Specifically, models (8) and (9) can be regarded as the robustness check, since we replace the basic explanatory variables (hhi_nat and n_nat) with additional one (hhi_lan and n_lan, respectively); we also replace the coach dummy with aggregated one (d_coach_lan), the other components of the equations are not changed.

(8)

(9)

It is necessary to estimate the correlation between them and dependent variable. Regression models output are reported in the Section 3.

Table 9

Correlation analysis results for additional metrics (basketball)

	Scores per game	hhi_lan	n_lan
Scores per game	1.0000
hhi_lan	-0.2748	1.000
n_lan	0.2069	-0.7525	1.0000

Moreover, we tested the nonlinear specifications of the suggested models (in particular, squared term is added). The results are presented in the next section.

3. Empirical results

3.1 The regression model results for the regular season in hockey

According to the Table 10, the regression model (6) indicates that hhi_nat is statistically significant on 5 % significance level. It means that diversity positively influences performance (the coefficient is an adverse figure). In particular, if hhi_nat grows by 0.01 ceteris paribus, scores per game will fall by (-0.3781).

The model (7) shows that another diversity metric n_nat is also significant on the 5% significance level. It confirms the results obtained in model (6). The interpretation is similar to the results described above: if n_nat grows by 0.01 ceteris paribus, scores per game will increase by 0.0324.

...