Roses with or without thorns? A theoretical model of workplace friendship

Abstract Previous studies on the effects of workplace friendship on team effectiveness have yielded somewhat inconsistent findings. Despite the inconsistent findings, our understanding of why those inconsistent findings exist are still in a nascent state. Furthermore, beyond group-level task performance, it remains unclear how workplace friendship affects group-level contextual performance, such as organizational citizenship behaviors and workplace deviance behaviors. On the basis of a qualitative review of studies in the existing literature, the authors build a theoretical model highlighting the roles of three contextual factors (i.e., group-level emotional labor demands, team-based rewards, and group members’ average age) in explaining how workplace friendship is related to three aspects of team effectiveness. This proposed model provides theoretical insights to facilitate future empirical studies on workplace friendship.


PUBLIC INTEREST STATEMENT
Social relationships with others are integral parts of our lives, and this is no less true for our organizational lives, given the significant amount of time that we spend at work. For example, according to the OECD (Organization for Economic Cooperation and Development), in 2017, an employed person in the United States spent an average of 1,780 hours per year at work. Employees are social animals who seek friendships in their organizational lives. However, past studies have shown somewhat mixed findings on the relationship between friendship and performance. In the present paper, we review the literature on workplace friendship and develop a theoretical model in a team context, proposing when and how to facilitate the positive effects on team effectiveness of workplace friendship. This paper aims to facilitate future studies on workplace friendship by providing fruitful venues for empirical testing.

Introduction
As Maslow's famous hierarchy of needs delineated, humans have a strong need for friendship (Maslow, 1954). According to the OECD (Organization for Economic Cooperation and Development), in 2017, an employed person in the United States spent an average of 1,780 hours per year working (OECD, 2019). Given that they spend a great many hours of their lives in the workplace, employees, as social animals, are always seeking friendship ties with others (Lincoln & Miller, 1979). Thus, it is natural that friendships are formed not only outside, but also inside organizations. Rath (2006) suggests that forming friendships at work is important in increasing job engagement, job satisfaction, and even life satisfaction. However, previous empirical studies on workplace friendship have shown somewhat inconsistent findings on the relationship between friendship and performance (Jehn & Shah, 1997). Moreover, even though some studies examined contextual factors affecting when and why friendship matters in the workplace (e.g., Jehn & Shah, 1997;Shah, 1993), it is still less clear for organizations whether they should facilitate or even condone workplace friendship. 1 The statistics in the United States also support the importance of workplace friendship. For instance, in 2004, about 25 percent of Americans reported that they had no close confidants at work. In addition to that, the average number of confidants was 2.08 in , down from 2.94 in 1985(McPherson et al., 2006 suggesting that employees are becoming more isolated and, consequently, would have even stronger needs for friendship at work. Workplace friendship has been studied in communication and management as well. On the one hand, communication research focuses on workplace friendship per se. For instance, a large body of research has demonstrated relatively consistent impacts of workplace friendship on an individual's attitudinal or behavioral outcomes, such as low turnover intention (e.g., Morrison, 2004;Nielsen et al., 2000), career success and high job satisfaction (Markiewicz et al., 2000), employee engagement (Khaleel et al., 2016), and lower "burnout" (Kruger et al., 1995). On the other hand, management research tends to direct its attention toward the friendship network. For instance, Morrison (2002) demonstrated that, in the context of newcomers' socialization, friendship network characteristics are more strongly related to non-work performance outcomes (e.g., social integration and organizational commitment), rather than to work-related (or work performance) outcomes (e.g., organizational knowledge, task mastery, and role clarity). In sum, previous research into workplace friendship shows relatively consistent findings on non-work performance, especially affect-based outcomes, but not on work performance (Jehn & Shah, 1997).
Although several studies have attempted to address the reasons behind the inconsistent research findings on the relationship between workplace friendship and group performance (e.g., Jehn & Shah, 1997;Shah, 1993), there is still a scarcity of empirical studies examining whether and when workplace friendship is beneficial. Especially given that studies of workplace friendship in management literature have mainly focused on the effects of an individual's friendship network on his or her attitudinal or behavioral outcomes or choices, it is still less known whether friendship per se in a work group leads to beneficial outcomes. In addition, most previous empirical investigations of the relationships between workplace friendship and group performance have been conducted in laboratory settings, which raises the question of whether workplace friendship truly has such significant effects. Lastly, when it comes to performance outcomes, previous literature tends to focus heavily on task performance, rather than contextual performance such as organizational citizenship behaviors (OCBs). Therefore, the present paper aimed to develop a theoretical model that: (a) focuses on the role of three group-level contextual factors (i.e., group-level emotional-labor demands, team-based rewards, and group members' average age) as contextual moderators in the relationships between workplace friendship and group performance, and (b) incorporates three dimensions of performance (i.e., task, citizenship, and counterproductive) as outcomes of workplace friendship. The model thus suggests theory-based methods of supplementing previously reported findings (e.g., Harrison et al., 2006;Rotundo & Sackett, 2002), and by providing fruitful venues for empirical testing, fosters future studies on workplace friendship. The proposed theoretical model of workplace friendship is depicted in Figure 1.

Theoretical model development
The following sections provide a qualitative review of previous studies on workplace friendship in the existing literature. Furthermore, building upon the qualitative review, this paper develops a theoretical model highlighting the roles of group-level contextual factors in the relationship between workplace friendship and team effectiveness.

Workplace friendship
In the workplace, employees form a variety of relationships, such as leader-subordinate or mentorprotégé. In addition to those formal workplace relationships, employees also have informal and personal relationships with organizational insiders, namely friendships. Workplace friendships are defined as "nonexclusive workplace relations that involve mutual trust, commitment, reciprocal liking and shared interests or values" (Berman et al., 2002, p. 218). They are distinguishable from other work relationships in that they are voluntary and have a personalistic focus (Sias & Cahill, 1998;Sias et al., 2004). In a similar vein, social network theory also conceptually distinguishes friendship networks from information or advice networks because friendship networks are fundamentally grounded in intimacy, trust, and social liking (i.e., affect), rather than being based around a specific task (e.g., Brass et al., 2004;Gibbons, 2004). In addition, they are statistically different from other informal social networks (e.g., Brass, 1984;Ibarra, 1995;Morrison, 2002).
Previous research (Brass, 1984;Ibarra, 1995;Morrison, 2002) shows some noteworthy findings with respect to whether friendship groups perform better than acquaintance groups, or vice versa. The two groups are different in that, as network theory suggests (e.g., Granovetter, 1973), acquaintanceship can be defined as weak ties, whereas friendship can be defined as strong ties, because friendship imposes more constraints on individuals. In their laboratory study, Shah (1993) showed that friendship groups perform significantly better than acquaintance groups on both decision-making and motor tasks. In another laboratory study, Jehn and Shah (1997) further demonstrated that friendship groups outperform the acquaintance group via higher commitment and cooperation for both cognitive tasks and motor tasks. Additionally, their results showed that friendship groups are better than acquaintance groups in terms of information sharing and morale-building communication. The two studies not only provide relatively clear findings with respect to the positive relationships between workplace friendship and group task performance, but they also explain why and how friendship groups can perform better than acquaintance groups, regardless of task types.
Even though most literature on workplace friendship focuses on the positive aspects, some warn that there are potential downsides. For instance, workplace friendships can blur boundaries and sometimes serve as distractions from work (Morrison & Nolan, 2007). Sias and Cahill (1998) also argue that workplace friendships often keep employees locked into the workplace even it is unhealthy and dysfunctional. Sias et al. (2004) provide some clues to why workplace friendships may deteriorate. Their explanatory study revealed five causes of workplace friendship deterioration: (1) problem personality, (2) distracting life events, (3) conflicting expectations, (4) betrayal, and (5) promotion. Consequently, they argue that workplace friendship deterioration can cause emotional stress, reduced ability to perform tasks, and increased turnover. In this regard, Morrison (2009) called workplace friendships a "double-edged sword" (p. 122).
Despite the prevalence of workplace friendship and somewhat conflicting views on its outcomes, researchers have tended to direct their attention away from friendship per se. Especially in management literature, research tends to focus its attention on the individual's friendship network and its outcomes. For instance, Krackhardt and Stern (1988) found that friendship ties across groups serve as resource-sharing during a crisis by providing coordination in response to crises. Brass (1984) demonstrated that an individual's relative position in the friendship network is related to his or her promotion. With respect to performance, using a sample of MBA students, Baldwin et al. (1997) showed that centrality in friendship networks is positively related to students' grades.
In sum, although research shows that an employee's friendship network plays a significant role in his or her career decisions and performance as well, there is a dearth of research examining workplace friendships per se at the group level and their effects on group-level outcomes (exceptions are Jehn & Shah, 1997;Shah, 1993).

Team 2 effectiveness
As previous research shows (e.g., Jehn & Shah, 1997;Shah, 1993), friendship groups perform better than acquaintance groups regardless of task types (i.e., whether highly cognition-demanding or not). Thus, following the previous findings, we can predict that workplace friendship in a group would be positively related to group task performance.
Beyond task performance, Rotundo and Sackett (2002) argue that there are three dimensions of job performance: task, citizenship, and counterproductive. They define task performance as "behaviors that contribute to the production of a good or the provision of a service" (p. 67). The other two are different from the task performance in that they are only indirectly related to task performance or, more generally, to effectiveness. Specifically, Rotundo and Sackett (2002) define citizenship performance as "behavior that contributes to the goals of the organization by contributing to its social and psychological environment" (pp. 68-69) and counterproductive performance as "voluntary behavior that harms the well-being of the organization" (p. 69). Thus, in extending the traditional criterion of job performance (i.e., task performance), given the high inter-correlation among the three performance domains, it is meaningful to examine the other two criteria.
Organizational citizenship behavior is defined as "an individual behavior that is discretionary, not directly or explicitly recognized by the formal reward system, and that in the aggregate promotes the effective functioning of the organization" (Organ, 1988, p. 4), and it comprises two targets: an organization (OCBO) and other individuals (OCBI) (Williams & Anderson, 1991). The latter directly reflects the former regarding behaviors (e.g., helps others who have been absent or who have heavy workloads). Since friendship at the workplace provides support, a sense of belonging, and identity (Morrison, 2002), the group workplace friendship would be positively related to group OCBI (G-OCBI). A meta-analysis of OCB (Podsakoff et al., 2000) also supports this argument by showing the positive correlation between group cohesiveness and all five dimensions of OCB: altruism, courtesy, conscientiousness, sportsmanship, and civic virtue.
On the other hand, workplace deviance behavior (WDB) is defined as "voluntary behavior of organizational members that violates significant organizational norms, and in so doing, threatens the well-being of the organization and/or its members" (Robinson & Bennett, 1995, p. 556). Similar to OCB, a meta-analysis (Berry et al., 2007) demonstrated that the target can be distinguished as either an organization (WDBO) or an individual (WDBI). As the meta-analysis has shown the negative correlation between OCB and WDB (Dalal, 2005), WDBI can be considered the flip side of OCBI. Hence, contrary to the prediction about the positive relationship between group friendship and OCBI, group friendship would be negatively associated with group WDBI (G-WDBI). Supporting this argument, much previous research has shown the negative correlations between group cohesion and social loafing (e.g., Druskat & Wolff, 1999;Duffy et al., 2006). In sum, workplace friendship would facilitate group-level helping behaviors (i.e., G-OCBI) and mitigate group-level destructive behaviors directed toward individuals (i.e., G-WDBI).
Proposition 1a: Workplace friendship is positively related to group task performance.
Proposition 1b: Workplace friendship is positively related to G-OCBI.
Proposition 1c: Workplace friendship is negatively related to G-WDBI.

Emotional labor demand
Although friendship research has shown somewhat mixed results with regard to its effects on work performance (Jehn & Shah, 1997), a large body of research on workplace friendship has demonstrated its relatively consistent impacts on non-performance-related outcomes such as low turnover intention (e.g., Morrison, 2004;Nielsen et al., 2000), career success and high job satisfaction (Markiewicz et al., 2000), and low burnout (Kruger et al., 1995). It is worth noting that most of the positive effects of workplace friendship fall into a category that concerns employees' psychological well-being (especially job satisfaction or burnout). Friendship network research also shows similar patterns (e.g., Morrison, 2002). For instance, in the context of newcomers' socialization, friendship network characteristics are more strongly related to non-work performance outcomes (e.g., social integration and organizational commitment) rather than work-related (or work performance) outcomes (e.g., organizational knowledge, task mastery, and role clarity). Therefore, we can infer that the importance of workplace friendship would be more salient in certain occupations or work groups where emotional labor requirements or stress levels are high.
Workplace friends can be targets with whom an employee discusses sensitive issues, either personal or work-related, since friendship fosters more open and intimate communication (Jehn & Shah, 1997;Sias & Cahill, 1998). In other words, workplace friends are sources of psychological or emotional support. This is consistent with the findings of Kram and Isabella (1985), which indicated that peer relationships can reduce or serve as a buffer against job-related stress. Thus, it is highly likely that members of teams in which friendship is prevalent will be better able to manage stress and emotional distress than members of teams where such friendship is lacking. That is, since they have informal and private communication channels for sensitive issues that might be important to them, but which they are reluctant to discuss in public, they are much better at managing their emotions. By serving as a sort of a pressure-release valve for stress, these friendships would yield greater benefits in workplace environments that are particularly stressful. In other words, the positive effects of workplace friendship on team effectiveness would be much more salient for the groups with high emotional labor demands.
Proposition 2: The relationship between workplace friendship and team effectiveness is moderated by emotional labor demands such that the relationship is stronger (for group task performance (2a) and G-OCBI (2b)) and weaker (for G-WDBI (2 c)) when emotional labor demands are higher.

Team-based rewards
Friendship and cohesion-which can be defined as the strength of group members' desires to remain a part of their life-are distinct constructs, but they also have much in common in that "[they] both focus on interpersonal affinity among group members" (Jehn & Shah, 1997, p. 776). Thus, Jehn and Shah (1997) suggest that "important insight into friendship groups can be gained by further investigating past research on cohesion" (p. 776).
Several previous studies consistently demonstrated that the relationship between group cohesion and performance can be contingent upon many factors, such as goal acceptance (Podsakoff et al., 1997). A meta-analysis (Beal et al., 2003) demonstrated that there are several factors influencing the strength of the correlation between group cohesion and performance. In particular, the meta-analysis emphasized the roles of behaviorally defined (as opposed to the outcome) and efficiency-based (as opposed to effectiveness) performance measures as moderators of the relationships. In sum, they all put emphasis on the roles of group performance norms in determining the effects of group cohesiveness on productivity.
Since workplace friendship is based on mutual trust and commitment, a shared goal would facilitate its positive performance impacts. In a similar study, Podsakoff et al. (1997) also pointed out that, based on goal-setting theory (Locke, 1968;Locke & Latham, 1990), goal acceptance matters in the relationship between group cohesiveness and task performance. In addition, although the extent to which experience conflicts differ by task characteristics (decision-making task vs. motor task), group research shows that conflict is better managed in a friendship group than in an acquaintance group (Shah, 1993).
Team-based rewards can be considered as shared goals for group members. In other words, the main argument behind group-based rewards is that outcomes being contingent on group performance provides incentives to work cooperatively (Deutsch, 1949). In addition, the group's goals are accepted by its members more readily when group cohesion is high (Forsyth, 1983). Hence, the positive effects of workplace friendship on team effectiveness would be much more salient when reward structures are team-based.
Proposition 3: The relationship between workplace friendship and team effectiveness is moderated by team-based rewards such that the relationship is stronger (for group task performance (3a) and G-OCBI (3b)) and weaker (for G-WDBI (3 c)) when the rewards structures are team-based.

Team members' average age
The younger generation (i.e., Generation Y) in the workplace has shown somewhat different perspectives and attitudes compared to the previous generations. For instance, it has been changed from "It's just business, it's not personal" to "it is all personal, and then there is business." (Lipkin & Perrymore, 2009, p. 150). In other words, the younger workforce is blurring the lines between their work and personal lives, at least in terms of relationships. For instance, in 2008, there was a special issue on "generational differences at work" in the Journal of Managerial Psychology. It showed that younger generations at the workplace are more likely to show higher self-esteem, narcissism, anxiety, and depression; a lower need for social approval; and a more external locus of control (Twenge & Campbell, 2008). In addition, younger workers tend to put more importance on status and freedom work values (Dries et al., 2008). Thus, teams with relatively younger employees would be more comfortable with workplace friendship and, consequently, better at coping with group tasks shared with those who are defined as workplace friends. Hence, the younger the team members are, the stronger the relationship between workplace friendship and team effectiveness.
Proposition 4: The relationship between workplace friendship and team effectiveness is moderated by team-based rewards such that the relationship is stronger (for group task performance (4a) and G-OCBI (4b)) and weaker (for G-WDBI (4 c)) when the average team members are younger.

Discussion
The proposed theoretical model from this review offers several theoretical contributions to workplace friendship literature. First, this paper sheds light on the importance of workplace friendship in predicting team effectiveness. Specifically, it aims at providing future research venues to address the previous inconsistent findings on the effects of workplace friendship on team effectiveness by proposing the three contextual factors (emotional labor demands, teambased rewards, and group members' average age). Second, this paper extends the previous workplace friendship literature by incorporating the more diverse aspects of team performance. That is, whereas previous literature has focused on the effects of workplace friendship on group task performance, this paper extends the previous findings by incorporating the other two performance domains, G-OCBI and G-WDBI. Specifically, the model suggests that workplace friendship is not only beneficial for group task performance, but is also conducive to increased helping behaviors (OCBI) and decreased detrimental behaviors (WDBI). Ultimately, it suggests that the benefits of workplace friendship should not be overlooked due to its potential downsides when it is deteriorated (e.g., Sias et al., 2004). Finally, whereas most previous literature on workplace friendship is based on experimental settings, this paper offers future empirical testing avenues to test the effects of workplace friendship in field settings.
The paper also provides important practical implications for organizations. Organizations are not eligible to inhibit the workplace friendship formations, but may facilitate them if they are conducive to increased group performance. This proposed model in this paper suggests when organizations should put their efforts toward facilitating workplace friendship in groups. Particularly, as the two moderators imply, organizations should seek to encourage workplace friendship when the occupations are highly emotionally demanding, or when the average age in the groups is relatively young.
Although this present paper focused on the positive sides of workplace friendship, it is important to note that we do not argue that workplace friendship is always beneficial. As previous literature on workplace friendship demonstrated, there are definite potential downsides. However, as this review paper implies, sometimes the benefits outweigh the costs. Roses are still beautiful, even though they have thorns.