Trust Consensus Within Culturally Diverse Teams: A Multistudy Investigation

Conceptualization of Team Trust Consensus

Trust is defined as a psychological state comprising the willingness to be vulnerable to the actions of another party based upon positive expectations of the intentions or behavior of that party (Mayer et al., 1995: 712). By extension, team trust can be broadly defined as aggregate perceptions of trust among team members (Langfred, 2004), with trust residing at the team level of analysis and pertaining to the team as a whole as the referent of trust (Fulmer & Gelfand, 2012). To date, scholars have predominantly conceptualized team trust as a shared group construct (Coultas et al., 2014; Waller, Okhuysen, & Saghafian, 2016), which assumes that trust emerges to the team level through compositional processes that lead team members to converge and develop shared perceptions of trust in their team (Kozlowski & Klein, 2000). Under this approach, trust magnitude is assumed to be the most relevant property of team trust, while dispersion in members’ perceptions of trust is assumed negligible and/or irrelevant to understanding how trust manifests and operates at the team level. Despite the dominance of this approach, multilevel theorists (Kozlowski & Klein, 2000) and increasingly trust scholars (De Jong & Dirks, 2012) have argued that this is not the only nor necessarily the best way to conceptualize group-level phenomena like team trust. Instead, team trust can also be conceptualized as a configural group construct that emerges to the team level via compilational processes and is characterized by meaningful dispersion in perceptions of trust among team members (De Jong & Dirks, 2012; Kozlowski & Klein, 2000).

In support of this alternative conceptualization, there is mounting evidence across studies suggesting that dispersion in trust is pervasive and persistent in team settings. A field study of R&D teams by Gillespie (2005), for instance, revealed that over a quarter of the teams sampled had substantial variation in members’ trust in the team. Likewise, in their study of fire-fighting teams, Colquitt et al. (2011) reported severely insufficient levels of interrater agreement (sample-average r_WG = .26) across constructs, including trust, indicating the presence of considerable within-team dispersion in their sample. Consistent with these findings, a growing number of studies (e.g., Braun et al., 2013; Geister et al., 2006; Mach & Baruch, 2015) have similarly reported insufficient agreement in team members’ levels of trust across the teams in their sample (as indicated by sample averages of r_WG < .70). In addition to prevalence, these studies also evidence the persistence of trust dispersion, revealing that it manifests not only in short-term teams (e.g., Geister et al., 2006; Mach & Baruch, 2015) but also in ongoing teams whose members have already worked together extensively (e.g., Braun et al., 2013; Colquitt et al., 2011). Consistent with this notion, Jarvenpaa and Keating’s (2012) case study of global project teams shows that dispersion in trust often persists and even worsens over time, as initial differences in trust and dependence give rise to increasingly divergent perspectives. Taken together, these findings suggest that trust dispersion can be an enduring property of team trust and is not necessarily resolved over time, as some scholars assume.

Consistent with the latter conceptualization and empirical evidence, our focus here is on team trust consensus, defined as the extent to which team members share their levels of trust in the team (Fulmer, 2018). Team trust consensus is a dispersion-based construct (Chan, 1998) that can range from low sharedness—reflecting high dispersion in members’ trust perceptions—to high sharedness—reflecting low dispersion in trust perceptions. Our notion of trust consensus is consistent with prior conceptualizations (Carter & Mossholder, 2015; Korsgaard, Brower, & Lester, 2015; Tomlinson, Dineen, & Lewicki, 2009) in that it captures the emergent property of trust dispersion. However, it differs in that our focus is on consensus (or: dispersion) at the team level and in the team referent rather than at the dyadic level and in individual referents.

Team trust consensus is an emergent state, “a property of the team that is dynamic in nature and varies as a function of team context, inputs, processes, and outcomes” (Marks, Mathieu, & Zaccaro, 2001: 357). As an emergent state, trust consensus has four defining characteristics (Waller et al., 2016). First, it presumes the presence of different levels of existence, in that its elemental content—trust perceptions—originates from and resides at the level of individual team members, which through compilational emergence processes and microlevel interactions gives rise to the presence of meaningful differences in trust at the team level of analysis. Second, it assumes an inherent dynamism-stability duality. On the one hand, trust consensus is inherently dynamic as it emerges through, and is continuously shaped by, ongoing team interactions. At the same time, rather than being in a constant state of flux, it also has a stable element that is sustained and endures for a period of time. Indeed, owing to its relative endurance, the third characteristic of trust consensus as an emergent state is that it can be sensed and experienced by team members, such that it influences their behavior. Fourth, by emerging through a complex interplay between elemental content and team interactions, trust consensus represents a unique team-level manifestation that is irreducible to the elements and interactions from which it emerged. As such, the state of trust consensus is distinct from the dynamic process of emergence that gave rise to it, in that it captures the extent of sharedness of trust at a given point in time (Waller et al., 2016). While we acknowledge the importance of understanding emergence, given the nascent stage of trust consensus research, our focus in this paper is on trust consensus as an emergent state and understanding its determinants and performance implications.

National Culture Diversity as an Antecedent

Due to the nascent state of the literature on team trust dispersion, there has been little theorizing or empirical examination of the antecedents of team trust dispersion. However, one common antecedent emerging from the broader trust dispersion literature (which includes but extends beyond team contexts) is compositional diversity (Korsgaard et al., 2015), particularly cultural diversity. Van Knippenberg and Mell (2016) proposed that compositional diversity could be an important positive predictor of trust dispersion in teams. In her study of military teams, Fulmer (2012) proposed cultural diversity as a positive predictor of team trust dispersion toward leader referents. Similarly, Zaheer and Zaheer (2006) proposed that differences in national cultures could serve as a social basis for the development of trust asymmetry in interorganizational relationships. Likewise, Zaheer and Fudge Kamal (2011) theorized how national cultural context is critical for understanding asymmetric trust relationships. Following these propositions, we focus on national culture diversity as our key antecedent factor.

Cultural background refers to the set of attitudes, values, and norms that are learned by individuals socialized in a particular cultural setting (Gibson et al., 2014). Culture guides the meaning people attach to their experiences and their perceptions of appropriate conduct in social interactions (Hofstede, 1984). While the cultural settings individuals are socialized in are commonly defined in terms of national boundaries (Gibson et al., 2014), they often span such boundaries to form cultural clusters of countries that share a similar set of attitudes, values, and norms (e.g., “Anglo,” “Confucian Asia,” “Latin America”; Ronen & Shenkar, 2013). Team diversity in cultural background can range from “low” (i.e., all members have a similar cultural background) to “high” (i.e., each member has a different background; Harrison & Klein, 2007).

Research on within-team dispersion proposes compositional diversity functions as a source of perceptual divergence (Roberson & Colquitt, 2005). Similarly, the broader team literature theorizes that cultural diversity increases perceptual divergence within teams (see Stahl et al., 2010; Van Knippenberg & Schippers, 2007). Culture functions as a perceptual filter and interpretative lens, “a collective programming of the mind” (Hofstede, 1984: 21), through which people experience and perceive their team. Hence, cultural diversity “brings very different sources and means of information–processing to a team” and is a key source of divergence in members’ perceptions of the team (Stahl et al., 2010: 692). Incorporating these insights with research on culture and trust, we propose national culture diversity leads to divergence in members’ perceptions of trust in the team through the causal mechanism of differential cognitive processing of trust cues.

National culture background has been shown to influence what is important for establishing trust, that is, the bases of trust and their relative weighting (e.g., Branzei, Vertinsky, & Camp, 2007; Yuki, Maddux, Brewer, & Takemura, 2005) and expectations of how trust should be enacted in work contexts, with evidence of culturally specific meanings and manifestations of trust (e.g., Wasti, Tan, & Erdil, 2011). In reviewing this literature, Ferrin and Gillespie (2010) conclude that trust operates as a variform universal: The general principle of trust holds across national cultures, but the dimensions and enactment of trust can differ across cultures.

Branzei and colleagues (2007) proposed and found support for a culture-contingent model of trust formation in emergent relationships. Drawing on work showing that culture influences trustors’ implicit conceptions of what causes behavior, they found culture influences the trust cues that are expected and relied upon to build trust. For example, culture influences the importance of dispositional (i.e., individual attributes) versus situational (i.e., nature, scope, and depth of the relationship) trust signals. A conclusion from this work is culture influences the way trust signals are interpreted and processed, resulting in different trust beliefs. This supports Doney, Cannon, and Mullen’s (1998) assertion that national culture influences how trust-relevant cues [i.e., information that signals (un)trustworthiness] are perceived and interpreted.

Applying this theorizing to the context of teams, we argue that the national culture background of team members will influence their cognitive processing (i.e., expectations, perceptions, attributions, and interpretations) of trust-relevant cues about the team. That is, culture acts as a filter through which members interpret the behavior of the team. Culturally diverse team members will each apply their own unique cultural lens to interpret, process, and make sense of trust-relevant cues in the team, resulting in divergent perceptions of trust in the team (low trust consensus). In contrast, culturally homogeneous team members share a common cultural filter through which to interpret and make sense of trust-relevant cues about the team. Hence, they will develop shared perceptions of trust in the team (high trust consensus). In this way, national culture diversity acts as a source of divergence in members’ trust in the team:

Team Performance Implications

While prior research has mostly focused on the moderator role of trust dispersion (e.g., De Jong & Dirks, 2012; Mach & Baruch, 2015), some studies have empirically examined its direct impact on team performance. Bergman, Small, Bergman, and Rentsch (2010) found a negative relationship between asymmetry in trustworthiness perceptions and group performance in an experiment involving ad hoc teams. Consistent with these findings, Carter and Mossholder (2015) found a positive effect of cognition-based trust congruence between teams and their supervisor on task performance in a field study of hospital teams. In her study of military teams, however, Fulmer (2012) failed to find support for a positive impact of trust consensus among members toward their leader on team performance. As empirical evidence is equivocal, there is a need for greater theoretical and empirical understanding of the direct impact of trust consensus on performance.

Consistent with the notion that trust operates as a social exchange mechanism and uncertainty reducer (Colquitt, LePine, Piccolo, Zapata, & Rich, 2012) and building on the broader literature on trust dispersion (e.g., Carter & Mossholder, 2015), we propose that team trust consensus benefits team performance in two ways. First, team trust consensus enhances team performance by promoting reciprocity of social exchange. Trust consensus increases team members’ willingness to share resources, as knowing they are equally trusted by their team signals that they are likely to receive resources in return (De Jong & Dirks, 2012). Conversely, trust dispersion (i.e., a lack of consensus) inhibits social exchange reciprocity as low-trusting team members feel they are putting themselves at risk by sharing their resources. Furthermore, despite receiving resources from teammates who trust the team, low-trusting members may be unwilling to utilize these resources to benefit the team, based on doubts about their teammates’ motives or the quality of the resources received from their team. At the same time, high-trusting team members, by willingly sharing their resources with their team, run the risk of being taken advantage of by low-trusting teammates, receiving nothing in return and leaving them with fewer resources to perform their own tasks. As a result, trust dispersion leads to imbalanced patterns of resource exchange and underutilization of the team’s resources, thereby hampering performance.

Second, trust consensus enables team members to effectively work together by reducing relational uncertainty. That is, trust consensus facilitates the development of a shared understanding of relationships members have with the team, which helps them to understand each other (even if it is from a place of low trust) and provides a common basis for working with each other in ways that align with this shared understanding (Tomlinson et al., 2009). By contrast, dispersion in team trust creates ambiguity about the nature of the relationship, which prevents members from establishing effective approaches to interacting and working together. Low-trusting members may, for instance, have a strong preference to work independently (to avoid depending on and sharing resources with the team) and formalize work agreements (based on skepticism about whether others will do their part), while high-trusting team members prefer to work interdependently and be flexible regarding who does what and when (Langfred, 2007). Their inability to agree on effective work arrangements and to understand their teammates’ position is likely to lead to mutual frustration and conflicts (Bergman et al., 2010) that send team members “spiralling into opposition” (Carter & Mossholder, 2015). These disruptive dynamics undermine the team’s ability to perform well. Hence, we propose:

Sample and Procedures

Data for this study were collected in four annual waves (from 2013 to 2016) from teams drawn from a variety of organizations, primarily located in the Netherlands. We used student-recruited sampling, involving master’s students drawing on their contacts to collect cross-sectional survey data (on cultural background and team trust) from team members and (on team performance) from their supervisors. This sampling strategy has been shown to yield samples that are equally representative compared to non-student-recruited sampling strategies (Wheeler, Shanine, Leon, & Whitman, 2014). This strategy has the benefit of yielding more organizationally heterogeneous samples, resulting in greater external validity than can typically be achieved within a single organization (Shen, Kiger, Davies, Rasch, Simon, & Ones, 2011). Students were provided with clear instructions about data collection procedures and set specific goals regarding the amount of data to collect. To ensure data quality, manually entered data were checked by a trained research assistant.

In line with extant work team definitions (e.g., Kozlowski & Ilgen, 2006), teams were eligible for data collection when (a) embedded within a larger organization, (b) comprised of three team members or more, (c) its members worked interdependently toward a common goal, and (d) the team was psychologically meaningful to respondents (i.e., it was clear who was and was not part of their team). We employed several strategies to maximize response rates, including advance notices, personalization, reminders, and ensuring confidentiality. Consistent with prior trust dispersion research (e.g., De Jong & Dirks, 2012; Mach & Baruch, 2015), we focused on recently formed teams and selected those teams in which members had worked together for 2 years or less (cf., Joshi & Roh, 2009) from our larger team sample. To be included in the final sample, teams needed to have matched team–supervisor data (necessary to examine the trust consensus–team performance relationship) and complete responses from all team members (to avoid distortion in national culture diversity and trust consensus variables; Allen, Stanley, Williams, & Ross, 2007). Applying these three criteria yielded a final sample of 120 teams with an average life span of 1.11 years (SD = .49). The average size for these teams was 3.63 (range: 3-7), participants were 30.03 years old on average (SD = 9.35), and 58% of them were male.

Measures

Statistical Procedures and Interpretation of Results

The means, standard deviations, and correlations among our team-level variables are presented in Table 1. We examined our hypotheses using OLS regression analysis. Recognizing the limitations of Null Hypothesis Significance Testing, we followed recent recommendations and assessed our hypothesis tests through substantive interpretations of the magnitude and precision of our point estimates (Cumming, 2012). Point estimate magnitudes are expressed as regression coefficients, while their precision is expressed through confidence intervals (CIs). A CI represents an estimation of the range of plausible values of the true population parameter, and a 95% CI indicates that we can be 95% confident that the CI’s lower and upper bounds contain the population parameter (Cumming, 2012). We also report standard errors (SE) and exact p values in our results tables for comprehensiveness.

OPEN IN VIEWER

Table 1

Study 1: Descriptive Statistics and Correlations

	Variables	M	SD	1	2	3	4	5	6	7	8	9	10
1.	National culture diversity (NCD)	0.29	0.37		0.105	0.270	0.643	0.006	0.116	0.015	0.054	0.000	0.533
2.	Functional diversity	0.54	0.39	0.15		0.744	0.415	0.359	0.573	0.294	0.829	0.069	0.022
3.	Team tenure (TT)	1.11	0.49	0.10	−0.03		0.461	0.254	0.370	0.783	0.402	0.485	0.535
4.	NCD × TT	0.02	0.18	−0.04	0.08	−0.07		0.388	0.669	0.704	0.800	0.768	0.213
5.	Team trust consensus (TTC)	−0.73	0.50	−0.25	−0.08	−0.10	0.08		0.000	0.000	0.000	0.000	0.004
6.	Team trust magnitude (TTM)	5.63	0.57	−0.14	−0.05	−0.08	−0.04	0.45		0.000	0.000	0.000	0.202
7.	TTC × TTM	0.13	0.44	0.22	0.10	0.03	−0.03	−0.45	−0.42		0.000	0.000	0.692
8.	TTC × TTC	0.25	0.50	0.18	0.02	0.08	0.02	−0.71	−0.35	0.59		0.000	0.100
9.	TTM × TTM	0.32	0.63	0.34	0.17	0.06	−0.03	−0.32	−0.48	0.69	0.39		0.508
10.	Team performance	5.41	0.80	0.06	0.21	−0.06	0.11	0.26	0.12	−0.04	−0.15	−0.06

Note: N = 120 teams. Below-diagonal values represent correlation coefficients, and above-diagonal values represent exact p values. Cross-product terms are based on centralized component variables.

In line with growing recognition that effect sizes should be interpreted in reference to those previously reported in similar research, we relied on the average “demographic variables-teams/groups” main effect size estimate across team-level meta-analyses (observed mean correlation: | r | = .05), reported by Paterson, Harms, Steel, and Crede (2016). We set this as our benchmark for interpreting the magnitude of the national culture diversity–trust consensus relationship and their average “teams/groups-performance” effect size estimate (observed mean correlation: | r | = .22) for interpreting the magnitude of the trust consensus–team performance relationship. Based on this benchmark, we qualify our effect sizes as “below average,” “(near) average,” or “above average” in magnitude. Since Paterson et al.’s (2016) benchmarks are expressed as uncorrected observed correlations, we use observed correlations (rather than regression coefficients) to interpret effect size magnitudes.

Discussion Study 1

Our results offer empirical support for prior theorizing of the relevance of national cultural diversity for trust dispersion (Fulmer, 2012; Van Knippenberg & Mell, 2016), indicating that achieving consensus in trust perceptions is more challenging for culturally diverse than culturally homogenous teams. More broadly, these results support Stahl et al.’s (2010) assertion that cultural diversity is a key source of divergence in members’ perceptions of their team. Our finding that trust consensus directly impacts team performance, independent of trust magnitude, corroborates a prior study by Bergman et al. (2010) and suggests that rather than representing a variable of secondary importance (Buvik & Tvedt, 2016; Mach & Baruch, 2015), trust consensus can be considered a core variable of substantive interest. The multiorganizational nature of our sample further strengthens generalizability of our findings across organizational settings (Shen et al., 2011). This is noteworthy since many trust dispersion studies to date have relied on student teams (e.g., Bergman et al., 2010) or nontraditional work teams, like professional sports or military teams (e.g., Fulmer, 2012; Mach & Lvina, 2017).

Our finding that trust consensus promotes team performance further suggests that trust consensus is indeed sensed and perceived by team members and influences their behavior as they work toward team goals. Whereas prior research found that trust magnitude can quickly emerge and be perceived by team members (Carter, Carter, & DeChurch, 2018), our supplementary analysis provides initial evidence that the same holds true for trust consensus by showing that newly formed teams have comparable trust consensus levels as more established teams. Together with prior evidence for its presence in both short-term and ongoing teams (Colquitt et al., 2011; Mach & Baruch, 2015) and its endurance over time (Jarvenpaa & Keating, 2012), these findings underscore the validity of conceptualizing trust consensus as an emergent state at the team level (Waller et al., 2016).

Towards a Contingency Model

Encouraged by the Study 1 findings, we conducted a second study with two aims: first, to test the replicability of Hypotheses 1 and 2 in a different team setting and using different measures of our focal constructs, and second, to extend understanding of the determinants of trust consensus by advancing a contingency model of national culture diversity. The basic tenet of this contingency perspective is that the impact of cultural diversity on teams can only be properly understood by accounting for contextual factors (Gibson et al., 2014; Stahl et al., 2010). Indeed, this is the dominant perspective in team diversity research and is widely recognized in both conceptual and meta-analytic work (Joshi & Roh, 2009; Van Knippenberg & Schippers, 2007).

Our choice of contingency variables was guided by conceptual work from the within-team dispersion literature. While using different labels, conceptual frameworks on within-team dispersion (e.g., DeRue et al., 2010; Roberson & Colquitt, 2005; Schneider & Reichers, 1983) consistently identify compositional (surface- and deep-level individual characteristics), interactional (behavioral and verbal acts involving communication among group members), and structural (externally imposed work structures and internally emerging network structures) as three distinct categories of antecedents that broadly explain dispersion in group members’ perceptions. Furthermore, this literature identifies unique theoretical mechanisms through which antecedent factors operate to affect within-team dispersion (Fulmer & Ostroff, 2016). For example, as we argue in Hypothesis 1, national culture diversity (a compositional antecedent) is theorized to create perceptual divergence among team members through the mechanism of differential processing of relevant cues. By contrast, two theoretical mechanisms identified as producing perceptual convergence among members are social influence and shared exposure (e.g., Mason, 2006; Roberson & Williamson, 2012). Finally, consistent with our contingency perspective, this literature further theorizes that antecedent factors are likely to operate simultaneously and interactively, such that factors promoting divergence and convergence offset each other’s impact in shaping within-team dispersion (Roberson & Colquitt, 2005).

These theoretical insights led us to focus on team virtuality and collective leadership as specific contingency factors that align with conceptually distinct antecedent categories and unique theoretical mechanisms and that have also been previously identified as relevant moderators of cultural diversity in teams (for virtuality, see Gibson et al., 2014; for team leadership, see Klein, Knight, Ziegert, Lim, & Saltz, 2011). Specifically, we chose team virtuality as our interactional antecedent, because of its potential to mitigate the diverging effects of national culture diversity by promoting convergence through shared exposure. Similarly, we chose collective leadership (density) as our structural antecedent, because of its potential to mitigate diverging effects of national culture diversity by promoting convergence through shared social influence. We theorize about their moderating impact in more detail below.

The Moderating Role of Team Virtuality

While team virtuality has been studied in many different ways (Gilson, Maynard, Young, Vartiainen, & Hakonen, 2015), there is growing scholarly agreement that team virtuality should be conceptualized as a multidimensional construct. A prominent example is Kirkman and Mathieu (2005: 702), who conceptualize team virtuality as encompassing three dimensions: “(a) the extent to which team members use virtual tools to coordinate and execute team processes, (b) the amount of informational value provided by such tools, and (c) the synchronicity of team member virtual interaction.” This conceptualization implies that team virtuality is a function of both the intensity with which teams rely on virtual tools and the type of virtual tools on which they rely, such that teams can be considered more virtual to the extent that they rely more (rather than less) on virtual tools and to the extent that the tools they rely on create more asynchronous (rather than synchronous) interaction among members and provide low (rather than high) information value (Gibson et al., 2014). As such, nonaudio, nonvisual tools such as e-mail are considered more virtual than audio-visual tools such as Skype. Paralleling Kirkman and Mathieu’s (2005) conceptualization, other scholars have proposed similar dimensions as defining elements of a team’s overall level of virtuality (Schmidt, Temple, McCready, Newman, & Kinzler, 2008). Whereas the use of virtual tools is a necessity for geographically dispersed teams, co-located teams, such as the ones we examine in Study 2, can and often do choose to communicate in a virtual manner, either as their predominate mode of interaction or as a complement to face-to-face communication (Gibson et al., 2014; Kirkman & Mathieu, 2005).

While scholars have traditionally assumed that virtuality creates challenges and obstacles for teams (Raghuram, Hill, Gibbs, & Maruping, 2019), meta-analytic evidence suggests that virtuality can actually be beneficial for teams (Fjermestad, 2004; Rains, 2005), particularly culturally diverse teams, by helping to avoid or alleviate some of the difficulties associated with cultural diversity (Gibson et al., 2014; Stahl et al., 2010). For instance, research shows that culturally diverse teams performed better, and were more satisfied, when team members worked virtually as opposed to face-to-face (Staples & Zhao, 2006; Takeuchi, Kass, Schneider, & VanWormer, 2013). Consistent with this evidence, we propose that virtuality can be beneficial for teams by mitigating the negative impact of national culture diversity on trust consensus. Specifically, we argue that virtuality mitigates the differential processing of information in culturally diverse teams by enhancing members’ shared exposure to the team.

Culturally diverse teams commonly experience language barriers and uneven communication patterns, such as unequal participation in group discussions and ignoring perspectives from those who are in the cultural minority (Bhappu, Griffith, & Northcraft, 1997; Earley & Mosakowski, 2000). This is partly due to differences in language proficiency and communication styles among culturally diverse members, who often differ in the speed with which they process and contribute to face-to-face discussions. However, virtual tools enable asynchronous interactions, removing constraints on turn-taking and allowing members from culturally diverse backgrounds to process, reflect on, and respond to input at their own pace (Carte & Chidambaram, 2004; Fayard & Metiu, 2014). The written nature of low media rich virtual tools enables more objective and specific communication, creates a searchable repository of the team’s knowledge and data sharing, and provides an audit trail of team agreements, expectations, and member contributions, thereby making them more salient, explicit, and accessible across the team (Fayard & Metiu, 2014; Kirkman & Mathieu, 2005). Hence, the use of asynchronous, written virtual tools enables more even and shared exposure and interaction in culturally diverse teams, which in turn reduces divergence in members’ processing and perceptions of their team (Carte & Chidambaram, 2004; Roberson & Colquitt, 2005).

In support of our theorizing, Rains (2005) found teams that used computer-mediated communication had more equal and frequent participation in conversations among team members than teams relying on face-to-face communication. A more recent study showed that the negative effects of differences in language proficiency among members (which included native speakers dominating conversations and difficulty in accurately assessing the team’s competence) were reduced when teams relied on text-based, computer-mediated rather than face-to-face interaction (Li, Yuan, Bazarova, & Bell, 2019).

Furthermore, the salience of surface-level cultural attributes—such as ethnicity, accents, and nonverbal cues—depends on the medium through which the team interacts. When interacting virtually, through asynchronous tools with low media richness (e.g., e-mail), the visual and vocal cues about ethnicity are reduced (Carte & Chidambaram, 2004). Nonverbal communication and the norms governing face-to-face social interactions are heavily culturally based; hence, by reducing these cultural cues, low media rich virtual communication limits the extent to which members culturally filter team information, reducing the culturally differentiated processing of trust-relevant information.

The Moderating Role of Collective Leadership

Collective leadership is defined as “an emergent team property of mutual influence and shared responsibility among team members, whereby they lead each other toward goal achievement” (Wang, Waldman, & Zhang, 2014: 181). Collective leadership originates at the level of individual members enacting one or more leadership roles and emerges to the team level through reciprocal, recursive influence processes among team members that come to define the structural pattern of leadership within the team (Carson, Tesluk, & Marrone, 2007; D’Innocenzo, Mathieu, & Kukenberger, 2016). We propose that collective leadership moderates the divergence in team trust caused by national culture diversity by mitigating the differential cognitive processing of trust-relevant information. Power and role differentiation have been shown to influence trustor’s cognitive processing of trust-relevant information (Graebner, 2009). Kramer’s (1996: 224-225) work showed that parties in subordinate roles tend to engage in “paranoid cognition,” using “more elaborate and differentiated mental accounting systems for tracking trust-related transactions” and paying more attention to trust violations than confirmations, compared to those in leadership roles. This aligns with research suggesting that power and role differentiation bring divergence in the way parties perceive and interpret trust-relevant information, resulting in different experiences and levels of trust (Graebner, 2009; Korsgaard et al., 2015). It also aligns with theorizing that through processes of structural equivalence, those occupying similar roles within the team develop convergent perceptions of the team (Roberson & Colquitt, 2005). This occurs through being able to perspective take with others in similar leadership roles, having access to the same sources of information about the team, and increased social influence among team members as they share their views and opinions about how to lead the team toward goal achievement.

Hence, by sharing leadership within the team, power and social influence are equalized between team members (Aime, Humphrey, DeRue, & Paul, 2014), mitigating the differential processing of trust-related information in the team. In contrast, when leadership and consequent power and influence are differentiated in the team, this compounds and exacerbates the differential processing of trust cues in culturally diverse teams.

Sample and Procedures

The data for this study were obtained from a sample of 550 students in a large Australian university enrolled in the gateway course of the MBA program who worked together in 95 project teams (average team size = 5.79). The MBA teams worked on a 12-week project that required them to (1) identify an organizational behavior problem in a real organization, (2) analyze the problem by doing research in the organization applying relevant academic literature, (3) write a group term paper on the teams’ analysis and proposed solutions, and (4) deliver a class presentation on their findings. Rather than self-selecting into a team, students were centrally assigned to teams by student services to achieve balanced compositions in terms of gender and functional background (see Langfred, 2004, 2007, for a similar approach). The average age and work experience of respondents was 31 and 8.3 years, respectively, and 31% were female. This research context was ideally suited to our research purposes as it afforded considerable variance in the antecedent factors. Specifically, the MBA program attracted international students from diverse cultural backgrounds, with 60.2% of participants in our sample (N = 328) born outside Australia and collectively representing 59 countries. Furthermore, being self-managed meant teams had considerable control and discretion over the frequency and virtual tools through which they communicated and how they organized their leadership structure. While the MBA course involved face-to-face class lectures, these lectures provided limited opportunities for students to meet as a team. As such, face-to-face contact was not situationally induced or imposed but rather occurred at the team’s discretion. Being enrolled in the same course meant that teams were performing similar tasks and were subject to the same evaluation standards, thereby minimizing distorting effects that could otherwise result from collecting data on different types of teams across different organizational contexts (as in Study 1).

We collected data on our variables at three different time points: Cultural background was measured at Time 1 (Week 1), team virtuality and collective leadership at Time 2 (Week 6), and team trust both at Times 2 and 3 (Week 12), with the latter (operationalized as consensus) representing our key variable of interest. Temporally separating our variable measurements allowed us to better account for the causal directionality implied in our model and mitigate common method variance. We intentionally chose to measure team trust consensus at Time 3. Given recent studies show trust magnitude can emerge to the team level in as quickly as 80 minutes or just a couple of weeks (Carter et al., 2018), the same should be the case for team trust consensus. This notion is supported by our Study 1 supplementary analyses, which show newly formed teams having comparable trust consensus levels as more established teams. These findings suggest that 12 weeks (i.e., Time 3) should provide sufficient time for trust consensus to manifest at the team level as an emergent (as opposed to emerging) state. Moreover, since the teams disbanded after the project, 12 weeks was the maximum possible time for trust consensus to emerge in these teams. While course instructor ratings of team performance were also collected at Time 3, the teams were not informed of their performance until after the team trust data were collected, to prevent performance feedback from influencing trust ratings. We obtained near-perfect response rates across all three waves (T1: 100%; T2: 96.7%; T3: 98%).

Measures

Statistical Procedures and Interpretation of Results

The means, standard deviations, and correlations among our team-level variables are presented in Table 4. Consistent with Study 1, we again tested our hypotheses using OLS regression analysis. Following best practice recommendations (Aguinis, Edwards, & Bradley, 2017), we tested our hypothesized national culture diversity main effect (Hypothesis 1) using the full model that also includes both cross-product terms. In this model, the national culture diversity effect size estimate can be interpreted as the average effect size across the full range of the moderators. We mean-centered the lower order terms prior to creating the higher order terms to facilitate the interpretation of our interaction effects. To interpret the magnitudes for our main effects, we again relied on Paterson et al.’s (2016) benchmark. For interaction effect size estimates, we relied on the average moderation effect size across team-level meta-analyses (semipartial correlation: | sr | = .20) reported by O’Boyle, Banks, Walter, Carter, and Weisenberger (2015). To compare our effect sizes against this benchmark, we converted our unstandardized regression coefficients into semipartial correlations using Cohen and Cohen’s (1983) formula. We followed the same procedures regarding control variables as in Study 1.

OPEN IN VIEWER

Table 4

Study 2: Descriptive Statistics and Correlations

	Variables	M	SD	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
1.	National culture diversity (NCD)	0.70	0.26		0.299	0.611	0.004	0.115	0.473	0.542	0.000	0.852	0.026	0.043	0.394	0.528	0.166	0.495	0.827	0.121
2.	Functional background diversity	0.86	0.10	−0.11		0.659	0.868	0.181	0.074	0.003	0.235	0.903	0.668	0.806	0.091	0.654	0.419	0.989	0.071	0.527
3.	Team virtuality (TV)	11.19	2.16	0.05	0.05		0.000	0.068	0.467	0.341	0.132	0.178	0.104	0.104	0.174	0.149	0.970	0.541	0.243	0.984
4.	Overall communication frequency	6.61	1.25	0.29	−0.02	0.78		0.097	0.062	0.365	0.000	0.325	0.009	0.740	0.209	0.044	0.733	0.733	0.303	0.473
5.	Collective leadership density (CLD)	0.62	0.06	−0.16	0.14	0.19	0.17		0.403	0.812	0.091	0.064	0.500	0.010	0.462	0.034	0.586	0.386	0.919	0.901
6.	Collective leadership decentralization	0.80	0.09	0.07	0.18	0.08	0.19	−0.09		0.087	0.051	0.143	0.024	0.674	0.891	0.595	0.291	0.883	0.794	0.726
7.	Prior familiarity	0.13	0.13	−0.06	0.30	0.10	0.09	0.02	0.18		0.785	0.433	0.399	0.861	0.661	0.198	0.932	0.651	0.616	0.647
8.	Work experience	8.40	2.70	−0.44	0.12	−0.16	−0.39	0.17	−0.20	−0.03		0.279	0.027	0.120	0.596	0.729	0.695	0.371	0.866	0.993
9.	NCD × TV	0.03	0.43	0.02	0.01	0.14	0.10	0.19	−0.15	−0.08	−0.11		0.567	0.000	0.718	0.007	0.552	0.260	0.296	0.016
10.	NCD × CLD	0.00	0.01	0.23	−0.04	0.17	0.27	0.07	0.23	−0.09	−0.23	0.06		0.076	0.848	0.659	0.159	0.693	0.823	0.704
11.	Team trust consensus (TTC), Time 3	−0.88	0.35	−0.21	0.03	0.17	0.03	0.26	−0.04	−0.02	0.16	0.36	0.18		0.039	0.000	0.009	0.001	0.993	0.002
12.	Team trust consensus, Time 2	−0.91	0.32	−0.09	−0.17	0.14	0.13	0.08	−0.01	0.05	−0.06	0.04	−0.02	0.21		0.598	0.324	0.778	0.337	0.246
13.	Team trust magnitude (TTM), Time 3	0.00	0.57	0.07	0.05	0.15	0.21	0.22	−0.06	0.13	−0.04	0.28	0.05	0.41	0.05		0.991	0.088	0.011	0.014
14.	TTC × TTM	0.08	0.19	0.14	0.08	0.00	0.04	0.06	−0.11	−0.01	0.04	−0.06	−0.15	−0.27	−0.10	0.00		0.000	0.002	0.287
15.	TTC × TTC	0.12	0.17	0.07	0.00	−0.06	−0.04	−0.09	−0.02	0.05	−0.09	−0.12	−0.04	−0.35	−0.03	−0.18	0.44		0.613	0.272
16.	TTM × TTM	0.32	0.38	0.02	0.19	0.12	0.11	0.01	0.03	−0.05	0.02	0.11	0.02	0.00	0.10	−0.26	0.32	0.05		0.408
17.	Team performance	76.38	4.37	−0.16	0.07	0.00	−0.07	−0.01	−0.04	0.05	0.00	0.25	0.04	0.32	−0.12	0.25	−0.11	−0.11	−0.09

Note: N = 95 teams. Below-diagonal values represent correlation coefficients; above-diagonal values represent exact p values. Cross-product terms are based on centralized component variables.

Results

Hypothesis 1 proposed national culture diversity is negatively related to team trust consensus. Consistent with this hypothesis, the CI and point estimate (B), shown in Table 5 (Model 2), indicate that the culture diversity–team trust consensus relationship is negative on average across moderators, with plausible population parameter values ranging from –.58 to –.07, with –.32 the most plausible estimate. The correlation associated with this point estimate is above average in magnitude (r = –.21, Table 4). All values within the CI are negative and do not include zero, indicating the presence of a meaningful relationship between national culture diversity and team trust consensus. Consequently, Hypothesis 1 is supported.

OPEN IN VIEWER

Table 5

Study 2: Regression Results for Antecedent Factors Impacting Trust Consensus

Independent Variables	Dependent Variable: Team Trust Consensus, Time 3
	Model 1					Model 2						Model 3
	B	SE	CIL	CIU	p	B	SE	CIL	CIU	p	sr	B	SE	CIL	CIU	p	sr
National culture diversity (NCD)	−0.24	0.13	−0.51	0.02	0.074	−0.32	0.13	−0.58	−0.07	0.014		−0.17	0.15	−0.46	0.12	0.252
Team virtuality (TV)	0.02	0.02	−0.01	0.05	0.172	0.01	0.02	−0.02	0.04	0.412		0.03	0.03	−0.02	0.08	0.231
Collective leadership density (CLD)	1.27	0.63	0.03	2.52	0.045	0.83	0.60	−0.35	2.02	0.167		0.71	0.62	−0.52	1.95	0.255
NCD × TV						0.25	0.08	0.10	0.40	0.001	0.31	0.26	0.08	0.11	0.42	0.001	0.31
NCD × CLD						5.32	2.60	0.16	10.48	0.043	0.19	6.21	2.70	0.84	11.58	0.024	0.21
Team trust consensus, Time 2												0.20	0.10	0.00	0.41	0.054
Functional diversity												−0.01	0.37	−0.74	0.72	0.982
Overall communication frequency												−0.04	0.05	−0.14	0.05	0.360
Collective leadership decentralization												0.09	0.41	−0.72	0.90	0.828
Prior familiarity												0.02	0.26	−0.51	0.54	0.945
Work experience												0.02	0.01	−0.01	0.05	0.175
R 2	0.12					0.25						0.31
ΔR2, p						0.13	0.001					0.06	0.324
F	3.95					5.83						3.33

Note: N = 95 teams. B = unstandardized regression coefficient; SE = standard error; CI_L = 95% confidence interval lower limit; CI_U = 95% confidence interval upper limit; p = exact p value; sr = semi-partial correlation.

Hypothesis 2 proposed that team trust consensus is positively related to team performance. The results reported in Table 6 (Model 1) show a positive direct effect of team trust consensus on performance, with plausible population estimates ranging between 1.58 and 6.50, and 4.04 as the most plausible value. The correlation coefficient for trust consensus (r = .32) is above average in magnitude. All population estimates within the CI are positive and do not include zero. Therefore, Hypothesis 2 is supported.

OPEN IN VIEWER

Table 6

Study 2: Regression Results for Trust Consensus and Team Performance

Independent Variables	Dependent Variable: Team Performance
	Model 1					Model 2					Model 3					Model 4
	B	SE	CIL	CIU	p	B	SE	CIL	CIU	p	B	SE	CIL	CIU	p	B	SE	CIL	CIU	p
Team trust consensus (TTC), Time 3	4.04	1.24	1.58	6.50	0.002	3.84	1.36	0.80	6.47	0.013	3.64	1.43	0.80	6.47	0.013	3.77	1.50	0.78	6.76	0.014
Team trust magnitude (TTM), Time 3						1.05	0.82	−0.54	2.74	0.187	1.10	0.83	−0.54	2.74	0.187	1.10	0.90	−0.69	2.88	0.227
Team trust consensus, Time 2						−2.64	1.35	−5.37	0.03	0.053	−2.67	1.36	−5.37	0.03	0.053	−2.69	1.39	−5.45	0.08	0.057
TTC × TTM											−1.23	2.36	−5.93	3.46	0.603	−1.48	2.85	−7.15	4.19	0.605
TTC × TTC																0.95	2.88	−4.78	6.67	0.743
TTM × TTM																−0.12	1.29	−2.69	2.45	0.924
R 2	0.10					0.15					0.16					0.16
ΔR2, p						0.05	0.065				0.00	0.603				0.00	0.938
F	10.62					5.56					4.20					2.76

Note: N = 95 teams. B = unstandardized regression coefficient; SE = standard error; CI_L = 95% confidence interval lower limit; CI_U = 95% confidence interval upper limit; p = exact p value.

With respect to Hypothesis 3, the CI and point estimate show a positive interaction between national culture diversity and team virtuality, with parameter estimates ranging between .10 and .40 (Table 5, Model 2). The most plausible estimate is .25, and the semipartial correlation (sr = .31) indicates an effect that is above average in magnitude. Despite the range of parameter estimates, all estimates are positive and do not include zero, indicating a meaningful interaction effect between the two independent variables. The direction of the effect size estimates—negative for national culture diversity (B = –.32) and positive for team virtuality (B = .01) and the cross-product term (B = .25)—is consistent with the pattern of interaction we hypothesized (Gardner, Harris, Li, Kirkman, & Mathieu, 2017). Plotting the regression results confirms that the negative relationship of culture diversity with trust consensus is weaker for teams with high than with low levels of virtuality (see Figure 1). We complemented these interaction plots by estimating the simple slopes of national culture diversity across the full range of team virtuality values in our data using Preacher, Curran, and Bauer’s (2006) statistical routines. This procedure yields a single line representing how the magnitude of the simple slopes (on the y axis) changes as a function of continuous values of the moderator (on the x axis), along with a 95% CI. These analyses (see Figure 1) indicate that with 95% confidence, the true simple slopes of national culture diversity are likely to be negative for teams with low levels of team virtuality and that the simple slopes gradually become weaker in magnitude as virtuality increases, to the point where national culture diversity–team trust consensus relationship can no longer be considered meaningful (evidenced by the CIs including zero). Together, these findings support Hypothesis 3.

OPEN IN VIEWER

Figure 1

Interaction and Regions of Significance Plots: National Culture Diversity by Team Virtuality on Trust Consensus

Regarding Hypothesis 4, the CI and point estimate (Table 5, Model 2) show a positive interaction between national culture diversity and collective leadership density on trust consensus, with population parameter estimates ranging between .16 and 10.48, and 5.32 being the most plausible estimate. The semipartial correlation associated with this point estimate (sr = .19) should be interpreted as near average in magnitude. All plausible parameter estimates are positive and do not include zero, indicating the presence of a meaningful interaction effect between the two variables. Table A5 (Model 2) furthermore shows that the direction of the effect size estimates is consistent with the pattern of interaction we hypothesized, with negative estimates for national culture diversity (B = –.32) and positive estimates for both collective leadership density (B = .83) and the cross-product term (B = 5.32) (Gardner et al., 2017). Plotting this interaction (see Figure 2) further confirms that the relationship between national culture diversity and trust consensus is weaker for high levels than for low levels of collective leadership density. Subsequent analysis across the full range of moderator values (see Figure 2) shows that, with 95% confidence, the true simple slopes of national culture diversity are likely to be negative for low levels of collective leadership and that the national culture diversity simple slopes estimates gradually become weaker in magnitude as collective leadership increases, up to the point where national culture diversity–team trust consensus relationship can no longer be considered meaningful (evidenced by the CIs including zero). Together, these results support Hypothesis 4.

OPEN IN VIEWER

Figure 2

Interaction and Regions of Significance Plots: National Culture Diversity by Collective Leadership on Trust Consensus

To assess the incremental validity of our focal predictor variables, we reran our analyses and added control variables to the models: functional diversity, overall communication frequency, collective leadership decentralization, prior familiarity, and work experience for the Hypotheses 1, 3, and 4 tests, and trust consensus at Time 2, team trust magnitude (Time 3), squared terms for trust magnitude and consensus (Time 3), and a magnitude-consensus cross-product term (Time 3) for the Hypothesis 2 test. In addition, we also reran the model using an unweighted virtuality measure to assess the impact of our weightings (Evans, 1991) and thus robustness of our results for this variable. The details of both sets of analyses can be found in the Online Supplement, Appendix B and Table 5 Model 3, Table 6 Models 2-4, and Table B1. In short, none of the results altered our substantive conclusions.

The Importance and Impact of Team Trust Consensus

Our paper challenges the dominant assumption that team trust is inherently characterized by shared perceptions among members, and hence dispersion in team trust is negligible and/or unimportant, by providing consistent evidence across two studies that trust consensus enhances team performance. In demonstrating its independent impact, above and beyond trust magnitude, our findings suggest that research focusing exclusively on magnitude (Breuer et al., 2016; Costa et al., 2018) provides an incomplete understanding of the ways in which team trust impacts team performance. Our insights do not necessarily dismiss or contradict the importance of trust magnitude but rather indicate that examining trust consensus complements and enriches understanding of team trust (De Jong & Dirks, 2012; Fulmer, 2018). Hence, rather than treating trust dispersion as a construct of secondary importance—by focusing on its moderator role or as relevant only when aggregation statistics fail to support a mean-level operationalization—our results show that trust consensus is an important and meaningful variable in its own right. As such, we suggest trust consensus be routinely incorporated and accounted for to provide a comprehensive understanding of how team trust manifests and operates in teams. Our findings also have implications for conceptualizations of team trust. Specifically, rather than defining team trust in terms of shared perceptions (e.g., Breuer et al., 2016; Costa et al., 2018), team trust may be more appropriately defined simply as aggregate perceptions of trust among team members (e.g., Langfred, 2004). The latter definition has the advantage of making no assumptions about how members’ perceptions aggregate to the team level and hence allow team trust to manifest as either shared or dispersed perceptions among members.

Our findings are consistent with our theorizing that team trust consensus positively and directly impacts team performance by promoting reciprocity of social exchange among team members and reducing relational uncertainty within the team. Our results do not support an interactive effect between trust consensus and trust magnitude on team performance, with the former operating as a moderator of the latter. This is not necessarily surprising given the inconsistency of prior evidence, with some studies finding support for this moderating effect (De Jong & Dirks, 2012), others finding support only for a three-way but not a two-way interaction (Mach & Baruch, 2015), others failing to find support (Buvik & Tvedt, 2016; Miles, 2016), and still others finding the opposite pattern of interaction, with trust dispersion strengthening rather than weakening the impact of team trust magnitude (Fulmer, 2012). Given that many of these studies rely on situational strength theorizing (Mischel, 1973) to argue for a moderating role, it is interesting to note that a recent review on situation strength similarly found mixed evidence, with a majority of the results failing to support this theory or even contradicting it (Keeler, Kong, Dalal, & Cortina, 2019). While our results do not support the moderating role of trust consensus, we do not see our finding of its direct impact on team performance as necessarily negating the possibility of it having a moderating effect as well. In fact, different causal roles have similarly been found for team trust magnitude, which has been proposed as a moderator in some instances and a main effect in others (Dirks & Ferrin, 2001). However, our findings do raise the question of when exactly trust dispersion is likely to exert a moderating effect, a main effect, or perhaps both at the same time. We see this as an important avenue for future research.

Determinants of Team Trust Consensus

Our study is one of the first to offer empirical insight into the antecedent factors influencing trust consensus in teams and the ways in which they do so. Our findings provide empirical support for prior scholarly propositions that cultural diversity is a key predictor of trust dispersion within teams (Fulmer, 2012). Our findings further corroborate the contingency perspective on cultural diversity (Stahl et al., 2010), including the moderating roles of virtuality and team leadership (Gibson et al., 2014; Klein et al., 2011). Our finding complements prior work, suggesting alternative ways in which leadership can benefit team trust. That is, whereas prior work shows that team leadership impacts the development of team trust magnitude directly (Braun et al., 2013), our study shows that, in the context of culturally diverse teams, collective leadership benefits team trust indirectly, by preventing cultural diversity from bringing about divergent perceptions of team trust among members.

Furthermore, by identifying team virtuality as a moderator of national culture diversity, our study answers recent calls to examine the intersection between cultural diversity and virtuality (Gibson et al., 2014). Our finding that virtuality mitigates the negative impact of national culture diversity challenges the dominant assumption in the literature that virtuality poses difficulties and obstacles for teams (Raghuram et al., 2019) and joins other scholars in suggesting that, in the context of culturally diverse teams, virtuality can actually be beneficial (Gibson et al., 2014; Stahl et al., 2010). Specifically, our study highlights virtuality’s previously unrecognized benefit of helping to prevent cultural diversity from producing harmful trust dispersion among members. In doing so, our study also complements prior research, which showed that teams relying on virtual technology can still establish high levels of trust (i.e., trust magnitude, Jarvenpaa & Leidner, 1999), by showing how teams relying on such technology are able to develop high consensus in trust despite the culturally diverse composition of its members.

The study of trust dispersion in teams is an emerging area of research that is still in a nascent stage of development. As such, our study represents an important step toward opening up this field of research and guiding future investigations into this topic. For instance, while our aim was not to test a comprehensive integrative framework, we do note that our focal factors—national culture diversity, team virtuality, and collective leadership—align with theoretical frameworks distinguishing between team composition, interactions, and structure as critical antecedent categories of emergent within-team dispersion (DeRue et al., 2010). Furthermore, our theorizing on their underlying mechanisms—differential processing of cues, shared exposure, and social influence—is consistent with theoretical mechanisms identified within this literature (Fulmer & Ostroff, 2016; Mason, 2006). Finally, our findings on the interplay between these factors are consistent with Roberson and Colquitt’s (2005) theoretical proposition that factors may act as either sources of divergence or sources of convergence and are likely to interact and offset each other’s impact. As such, the conceptual model we test in our paper represents one specific way in which these theoretical building blocks can be applied to research on trust dispersion. Clearly, these building blocks are broader in scope and applicability than the specific variables we examined in this paper and thus may serve as a foundation for identifying and exploring additional determinants of team trust consensus.

Beyond developing cumulative insight into the antecedents of team trust consensus, another logical extension of our research is to examine the generalizability of our findings to other levels of analysis and trust referents (Fulmer & Gelfand, 2012). While our focus was on trust dispersion at the team level, with the team as a whole as the referent, research within the broader literature on trust dispersion has focused on other levels of analysis (e.g., dyadic, Korsgaard et al., 2015) and/or in other referents (e.g., team leader, Carter & Mossholder, 2015). Fulmer and Gelfand (2012) caution against assuming antecedents of trust are functionally isomorphic across levels of analyses or referents in absence of empirical evidence. Likewise, Korsgaard et al. (2015) propose that some antecedents of trust dispersion may be referent-specific. We agree that whether our findings generalize to other levels and referents is ultimately an empirical question. The factors identified in our study may serve as a promising starting point for advancing understanding of trust dispersion at other levels and in other referents.

Another natural extension of our work is to shift from a focus on team trust consensus as an emergent state at a given point in time to team trust emergence as a dynamic process. While our Study 2 data do not allow for statistically modeling emergence, inspection of the within-team differences in trust consensus scores between Times 2 and 3 reveal an intriguing pattern: Half (50.5%) the teams increased in consensus over time, while the other half (49.5%) decreased. This pattern aligns with multilevel theory and evidence, which recognizes that compilational (or dissensus) emergence is an equally viable trajectory as compositional (or consensus) emergence and that teams do not necessarily all follow the same emergence trajectory (Kozlowski & Klein, 2000). Understanding the antecedent factors that explain team trust emergence and between-team variability in emergence trajectories over time is an important direction for future research. Our model and findings may be a useful starting point to think through how our proposed antecedents and underlying mechanisms might change when shifting from trust dispersion states to dynamic trajectories over time. We recommend that future research into trust emergence engage with process theories of emergence, adopt high-density repeated measurement designs, and take advantage of recently developed modeling techniques.

Limitations and Future Directions

Our studies have potential limitations, and addressing these opens up promising areas for future research. First, while we proposed theoretical explanations for why we expect our focal variables to be related in a certain way (e.g., differential cognitive processing of trust cues as the theoretical mechanism explaining why national culture diversity impacts trust consensus), we did not explicitly test the causal mechanisms underlying these theoretical explanations. As such, we were neither able to empirically ascertain whether our theorizing accurately describes the actual processes through which our focal variables affect each other, nor were we able to exclude alternative theoretical explanations and mechanisms. Explicit examination of these causal mechanisms is an important direction for future research.

Second, given the relationship between team means and standard deviations should in principle be an inverted U shape, the linear relationships we find between trust consensus and magnitude in both studies (Study 1: r = .45; Study 2: r = .41) indicate systematic range restriction in our team trust data (Lindell & Brandt, 2000), with low trust teams being underrepresented. Since range restriction attenuates effect size estimates (Cole et al., 2011), the effect sizes we report for trust consensus are likely to be conservative estimates. We encourage future research to employ strategies aimed at maximizing the full range of values on the trust measure in order to obtain more accurate estimates.

Third, while our conceptualization and operationalization of trust consensus are consistent with prior research and best practices (Harrison & Klein, 2007; Mach & Baruch, 2015), they only capture the level of trust dispersion but do not capture distributional forms or patterns of dispersion (DeRue et al., 2010; Kozlowski & Klein, 2000). Investigation into the latter would be a natural extension of our current study. Echoing other scholars (Coultas et al., 2014; Fulmer, 2018), we see configurational and social network approaches to team trust as a promising direction for moving the field of trust dispersion forward. These approaches allow scholars to capture complex patterns and configurations of trust within the team, enabling insight into what antecedent factors give rise to distinct patterns of trust and how these patterns, in turn, differentially impact outcomes across levels of analysis.

Finally, our operationalization of team virtuality attempted to capture all three dimensions of Kirkman and Mathieu’s (2005) conceptualization and as such is an improvement on many prior measures (Raghuram et al., 2019). However, some scholars argue that information value levels are not fixed or inherently associated with particular virtual tools but should be understood as a subjective assessment by those who use these tools (Maynard et al., 2019). Furthermore, while aggregation statistics justified our mean-level operationalization, we also observed within-team variability in virtuality (sample-mean within-team SD = .41). This observation aligns with the view that team members may differ in their levels of virtuality for a variety of reasons, including individual differences in their preference for or competence in working with particular virtual tools (Kirkman & Mathieu, 2005). However, our mean-aggregated measure does not capture or model this individual-level variance. We recommend future research examine the influence that alternative operationalizations of team virtuality have on team trust consensus.

Practical Implications

Our results provide practitioners with unique insights and actionable knowledge on how to manage trust in culturally diverse teams. First, our finding that team trust consensus predicts team performance, combined with prior work indicating that dispersion in trust is pervasive and persistent, clearly indicates that trust building in teams needs to be proactively managed in order to reap its performance benefits. Our results show that it is not only the magnitude of trust in the team that is important but also the extent to which members share levels of trust in the team. Second, our results show that increasingly common features of contemporary teams—namely, being composed of individuals from diverse backgrounds (national culture diversity), with members interacting through virtual means (team virtuality), and structuring their activities through self-management (collective leadership)—not only pose challenges for the development of shared team trust but also offer opportunities to overcome these challenges. Specifically, our findings indicate that the negative affect of cultural diversity on shared team trust can be neutralized by (a) sharing leadership roles within the team and (b) complementing face-to-face interaction with virtual communication using asynchronous, low information value technologies. As such, our findings provide practitioners with actionable knowledge on ways to counteract the negative effects of cultural diversity in teams.