Abstract
Previous research has demonstrated the critical role communication plays in a group’s ability to recognize its expert members. This study looks broadly at the different forms of communication that might influence expertise recognition and considers how structural, relational, and communicative factors are related to individuals’ success in having their expertise recognized by other group members. In addition, we advance a view of expertise recognition in terms of expertise sharing and consider the circumstances under which an individual’s self-perceived expertise is likely to match the perceptions of other group members. Drawing on survey data from 99 employees at a financial services company, we find that it is communication practices, and not structural influences, that primarily relate to group members having their expertise recognized by coworkers. The findings extend theory that views attributions of individuals’ expertise in organizations as a communicative phenomenon that emerges through work practices.
Research over the past several decades on expertise recognition has found that groups that are more effective at identifying their expert members outperform comparable groups in terms of task effectiveness (Austin, 2003; Libby, Trotman, & Zimmer, 1987; Littlepage, Hollingshead, Drake, & Littlepage, 2008). However, this work has also found that groups generally do a poor job identifying experts (Littlepage, Robison, & Reddington, 1997) and that assessments of expertise may vary among members (Yuan, Fulk, & Monge, 2007). Recently, scholars have examined the specific reasons that people make errors in evaluating the expertise of other group members and argued that communication among individuals over time may be both the cause of, and solution to, failures in expertise recognition (Bunderson & Barton, 2011; Hollingshead, Brandon, Yoon, & Gupta, 2010).
To better understand the role of communication in expertise recognition, this study inverts the question of how workers identify the expertise of peers and instead asks what might lead individuals to have their expertise recognized by others in an organizational group. By simultaneously testing the potential influence of structural, relational, and communicative cues that may influence expertise recognition, we contribute to the growing body of scholarship on the relationship between communication and expertise recognition in organizational groups (Bunderson, 2003; Treem, 2012; Yuan, Bazarova, Fulk, & Zhang, 2013). Given our focus, this study considers a single broad research question:
The study presented here builds on previous research and extends our understanding of the mechanisms underlying the recognition of workers’ self-perceived expertise in two interrelated ways. First, by focusing on the communicative cues provided in relation to self-perceived experts, our approach allows for the possibility that workers may play an active role—intentionally or not—in shaping how other group members view their expertise. Theories of expertise recognition have been previously criticized for failing to account for the various motivations and biases that workers may have in organizations (Wittenbaum, Hollingshead, & Botero, 2004). Although researchers have examined various motivations and predictors regarding decisions of group members to retrieve, allocate, or exchange expertise (Hollingshead et al., 2010; Yuan, Fulk, & Monge, 2010), less consideration has been given to opportunities workers might have to influence expertise recognition, which is an important precondition to expertise utilization (Austin, 2003). Examining different predictors of expertise recognition from the perspective of the self-perceived expert is important because workers may have conflicting individual and group goals for communicating and sharing expertise and therefore may act in self-serving ways (Cabrera & Cabrera, 2002; Jarvenpaa & Majchrzak, 2008).
Second, by looking at expertise recognition among workers in large, ongoing, non-collocated work groups, this study responds to calls for researchers to look at how expertise recognition might operate in organizational contexts (Lewis & Herndon, 2011; Moreland, 1999). Researchers have sought to extend concepts of transactive memory theory (Wegner, 1987)—originally intended to apply to small, collaborative groups—to larger organizational contexts (Jackson & Klobas, 2008) or settings in which individuals move among teams and face different tasks over time (Lewis, Lange, & Gillis, 2005). Similarly, research incorporating principles of expectation states theory (Berger, 1974) found that status characteristics influencing assessments of expertise can emerge at the organizational level (Bianchi, Kang, & Stewart, 2012) and among groups lacking a collective task focus (Treem, 2013).
Investigating how the actions of self-perceived experts and the diverse structures of work groups may influence expertise recognition is important to better understand what predicts expertise recognition in organizational settings (Lewis & Herndon, 2011). Because contemporary workers are increasingly members of different organizational groups simultaneously and move fluidly among tasks (O’Leary, Mortensen, & Woolley, 2011), it is important to reconsider the scope and applicability of theories of communication originally formulated based on more stable contexts. To explore predictors of whether workers will have their self-perceived expertise recognized by group members, we draw upon survey data regarding the perceptions, attributes, and relations of employees at a financial services company.
Theoretical Background
Although research has demonstrated the integral role communication can play in expertise recognition in organizational groups, there is uncertainty regarding which specific forms of communication occurring in work contexts matter most in the assessments workers make about each other (Bunderson, 2003; Su, 2012; Yuan et al., 2013). For instance, research on group information sharing has demonstrated that individuals are more likely to attribute expertise to others on the basis of behaviors displayed in discussions such as talkativeness (Bottger, 1984; Littlepage, Schmidt, Whisler, & Frost, 1995), or attributes such as role assignment (Stewart & Stasser, 1995). Similarly, research adopting a transactive memory perspective has found that working groups attribute expertise to other members on the basis of a variety of behaviors such as shared training (Liang, Moreland, & Argote, 1995; Moreland & Myaskovsky, 2000) or attributes such as gender (Hollingshead & Fraidin, 2003) or cultural stereotypes (Yoon & Hollingshead, 2010). The same pattern is found in research applying expectation states theory, which has shown that expertise recognition and associated influence in groups can be related to attributes such as gender or technical certifications (Bunderson, 2003) or behaviors of workers such as the use of technologies (Treem, 2013). Although each approach makes distinct predictions about how group members recognize the expertise of others, together they advance a framework suggesting that individuals provide communicative signals that other group members use to form expectations about each individual’s knowledge, abilities, and likely action (Bunderson & Barton, 2011; Driskell & Mullen, 1990). To help evaluate the nature and scope of the relationship between communication and expertise recognition in organizations, we draw on the construct of expertise sharing developed by information systems researchers studying knowledge management in organizations (Ackerman, Dachtera, Pipek, & Wulf, 2013; Ackerman, Pipek, & Wulf, 2003; Pipek, Hinrichs, & Wulf, 2003). Within this literature, expertise comes to be shared in organizations through a cooperative process in which individuals identify experts and those experts are willing to provide expertise. As McDonald and Ackerman (1998) noted, The term expertise assumes the embodiment of knowledge within individuals. Our definition distinguishes expertise, which is a range, from expert. An individual may have different levels of expertise about different topics. Expertise can be topical or procedural and is arranged and valued within social and institutional settings. (p. 315)
In other words, expertise is substantive in that individuals may possess expertise, but expertise is only manifest relationally in organizations through forms of action and interaction that can be assessed by others. For instance, a worker at an engineering company may be an expert in computer-aided design, but if others in her organization do not perceive her as having this expertise, she is not likely to be assigned related tasks, and cannot meaningfully act as an expert in that particular context—Her expertise would not be recognized. Similarly, if others see her as an expert in this domain, but she does not, she is not likely to assume responsibility for this expertise. In this latter case, the expertise has been located but not recognized in an operational sense that would benefit the group.
Importantly, focusing on expertise sharing as a precondition for expertise recognition does not mean that perceptions of expertise will necessarily differ substantially from more normative approaches to expertise recognition. For instance, Austin (2003) measured the accuracy of expertise recognition by first comparing the matches of workers’ self-perceived rating on 11 knowledge domains to the coworkers’ ratings of the respective workers, and second based on workers’ performance on problem-solving scenarios. Results revealed a high correlation (.82) between the self-report measures and the objective measure and the study proceeded with the self-report measure under the rationale the approach was more easily replicable in future work. Similarly, McDonald (2001) asked workers at a software company to guess the scores of coworkers on an instrument evaluating domain knowledge and found that the perceptions of coworkers’ performances were highly correlated with actual results (.88). Reliance on perceived measures of expertise (both self-perceived and assessments of others) to evaluate expertise recognition does not necessarily mean that assessments will vary significantly from approaches that use more decontextualized or normative standards of expertise. Furthermore, the perspective on expertise recognition advanced in this study is agnostic to the motives, intentions, or purposefulness with which expertise is shared within organizations. It is not known or assumed a priori that individuals’ actions or perceptions are shaped by any specific group or individual interests and workers may, or may not, engage in a variety of behaviors designed to communicate or conceal knowledge in hopes of obtaining or avoiding attributions of expertise (i.e., Leonardi & Treem, 2012). We now turn our attention to how various forms of expertise sharing might influence expertise recognition in organizational work groups.
Three Influences on Expertise Sharing
Communication that may inform assessments of workers’ expertise by coworkers is influenced by a variety of factors that may be structural (e.g., where work physically takes place), relational (e.g., whether others will interact with an individual), or individual (e.g., whether one chooses to share knowledge). These influences vary in the extent to which they may constrain the abilities of group members to actively shape how their expertise is communicated to others. Viewing these influences in terms of what workers can and cannot control provides insight into the differential roles of active communicative behaviors in facilitating the recognition of individuals’ expertise.
Group Structural Influences
Workers often have little control over the composition of organizational groups and, in turn, often have limited input into the experience, knowledge, or roles of other group members. Yet despite this limitation, the structure of the group in terms of the diversity and salience of the attributes of members can influence the ability of group members to communicate their expertise. For instance, groups that have higher variability in expertise among members are better able to identify expert members (Libby et al., 1987), and differentiation in knowledge among group members provides individuals with a form of “diagnostic information” that can allow for easier recognition of expertise (Baumann & Bonner, 2004, p. 90). If group members develop distinctive or specialized forms of expertise, it may be easier for other group members to identify and agree upon the different types of expertise for each individual (Austin, 2003). In addition, unique forms of expertise may be more salient and therefore more likely to be used by others in developing expectations of individuals (Fişek, Berger, & Norman, 2005). However, research on information sharing in groups has also demonstrated that individuals are likely to pay more attention to, and value as expertise, information that is shared between themselves and other group members, exhibiting what is termed the mutual enhancement effect (Wittenbaum, Hubbell, & Zuckerman, 1999). Because theory suggests that expertise recognition could be predicted by either the presence of unique or shared self-perceived expertise, we do not make any specific predictions regarding the relationship and instead present competing hypotheses.
Another element that may play a role in the success of individuals communicating their expertise is the extent to which that expertise is seen as meaningful to others in a group. When a form of knowledge is seen as relevant to organizational tasks facing group members, it activates a process of assessing who is knowledgeable in a respective domain and who is not (Cohen & Zhou, 1991). In addition, if individuals lack the knowledge needed to complete a task, it may increase their motivation to seek knowledge from others (Moreland, 1999). The need for a particular form of expertise will make the communication of signals associated with that knowledge more salient. Therefore, having expertise that other people find valuable may be associated with greater expertise recognition.
Relational Structural Influences
Even if one’s knowledge is both unique and needed within a group, obstacles may exist that prevent coworkers from interacting with, or learning more about, a group member who has this knowledge. Coworkers in different wings of a building or separated by floors may communicate infrequently and have little knowledge of each other. Merely sitting more than 30 m from another group member greatly reduces the frequency of communication and the level of informal communication among coworkers (Allen, 1977), and coworkers on distributed teams who are separated geographically or working across time zones are less likely to communicate compared with collocated workers (O’Leary & Cummings, 2007). In a study of active work groups, Su (2012) found that employees who worked remotely from each other had significantly worse assessments of the expertise of colleagues than collocated employees. He noted when a large portion of group members’ work is done remotely, “it creates a hurdle for them to closely observe each other’s behaviors and discern expertise cues” (p. 13). Therefore, we would expect the location of workers relative to each other to relate to expertise recognition.
Although individuals cannot easily influence the knowledge held or needed by other group members, or where they are physically assigned to work, they have some control over with whom they interact in the course of work. Group members develop a number of relationships, both formal and informal, that provide them with knowledge needed to complete work, and affective resources that aid in social support (Ibarra & Andrews, 1993). As an aggregate, these relationships create networks that inform how knowledge moves through an organization (Monge & Contractor, 2003).
One form of relationship that may be associated with an individual communicating expertise is the extent to which a worker is a friend of others in an organizational group. Friends commonly communicate more often, and in more meaningful ways, than compared with interaction among acquaintances (Granovetter, 1973), and workers use friends to gather information that is useful in making organizational decisions (Kilduff, 1992). These friend relationships may result in greater familiarity regarding each other’s strengths and weaknesses. In a study of MBA consulting teams, Lewis (2004) found that groups where members noted greater familiarity with others also had a better sense of the division of expertise among group members. This previous research leads to the following hypothesis:
Communication Practices
Workers often interact with other group members to obtain information that they need to complete tasks or find answers to problems, and in doing so, constitute a network of workers seeking and providing advice (Bonaccio & Dalal, 2006; Cross & Sproull, 2004). Not surprisingly, research indicates that individuals commonly seek task-related advice from people they see as experts in a domain of interest to the advice seeker (Sniezek & Van Swol, 2001). Because the assessment, or recognition, of the expertise of the advice provider typically precedes someone’s decision to seek advice from someone else, it is not likely to affect expertise recognition as the act of advice seeking would be a consequence of, not an antecedent to, expertise recognition (Leonardi, 2007). But when a person decides to seek task-related advice from others, the advice seeker often discusses areas in which he or she has expertise. In typical advice consultations, advice seekers tell their would-be advisers about the things they do know and the things they do not know, which consequently, can lead the adviser to form an impression about the advice seeker’s expertise (Blau, 1955). The result is that while advice seekers gain knowledge from individuals they already considered to be experts, those expert advisers are also forming impressions of the advice seekers that they would have not otherwise been able to form had the advice seeker not approached them (Yaniv & Kleinberger, 2000). Because workers can initiate advice seeking, asking questions may serve as a means for workers to increase communication with other group members and, in doing so, provide signals that promote the sharing of expertise.
Not all advice networks are centered on task-specific information or how to approach assignments. Often, individuals reach out to coworkers to gain critical information about organizational norms or acceptable practices within a particular group context. Workers assigned specialized tasks may have limited direct insight into each other’s assignments, and interaction about norms may be an effective way to make one’s capabilities known to others. Thus, when an individual seeks advice from someone about broad norm-based issues within the organization, they also inadvertently learn about areas in which that source has task-based expertise (Cross, Borgatti, & Parker, 2003). When individuals seek advice around norms, it may provide a more complete picture regarding the advice giver’s expertise on important work-related topics (Podolny & Baron, 1997).
Finally, the volume and form of individuals’ active communication with coworkers through workplace technologies is another option for signaling expertise that is largely under each worker’s control (Treem, 2013). Direct communication among work members is posited as the most effective way knowledge of who knows what develops in organizational groups (Hollingshead & Brandon, 2003), particularly when communication is task oriented (Yuan et al., 2013). In addition, because communication continues over time among group members, individuals can update their directories of who knows what in an organization and possibly alter assessments made based on initial stereotypes (Hollingshead & Fraidin, 2003).
Communication in contemporary organizations is commonly mediated through the use of technologies, and communication media vary greatly in the ways they allow people to present knowledge to others in an organization (Hinds & Bailey, 2003). Media such as the telephone or instant messaging are primarily used to support interpersonal, or one-to-one, communication. Other media, such as file repositories or social networking sites, can be characterized as communal in that they provide information that is simultaneously accessible to all organizational members (Fulk, Flanagin, Kalman, Monge, & Ryan, 1996). Interpersonal communication may be an effective way to signal expertise because it is more focused and directed. Alternatively, communal repositories may be useful because they both store individuals’ knowledge for access over a period of time and because they provide a visible directory of who knows what (Choi, Lee, & Yoo, 2010; Hollingshead, Fulk, & Monge, 2002; Nevo & Wand, 2005). The potential role of all of these different forms of communication in sharing expertise leads to the following set of hypotheses:
Method
Site and Data Collection
This study was based on survey responses from employees at a large financial services organization in the Midwestern United States. The company, American Financial (a pseudonym), employs more than 15,000 individuals located across the globe and offers a variety of products and services including credit cards, banking, and loans. The individuals participating in this study were from two groups in American Financial’s leadership development program (marketing and operations) and rotated through positions in their respective groups. Specifically, the leadership program offered employees in these groups the opportunity to move among different working teams in the organization every 6 months in an effort to expose them to different parts of the organization. There are three reasons this population of employees is particularly useful for our goal of exploring issues related to expertise recognition in organizational settings beyond small teams collaborating on set tasks. First, these employees were not hired for a particular role and are frequently moving among different project teams, meaning their areas of expertise cannot be easily determined by role or position. In addition, because members of the leadership program groups often worked in different rooms, floors, or buildings, and work was primarily conducted in front of computers, there were limited opportunities for members to observe each other engaged in tasks. Second, members in the leadership program were interdependent and collectively oriented in the sense that they commonly faced organizational tasks in which they relied on each other for knowledge and expertise. However, because of the rotational system, these employees were also members of other project-based teams meaning both that the leadership program groups operate as cross-functional groups and that individuals had dual responsibilities to their leadership program group and whatever project team they currently worked with. Third, because different cohorts of workers entered into the program each year, employees in this group had different levels of tenure, organizational rank, and were likely to have developed varied relationships. Therefore the structure of this group served as an opportunity to extend the scope conditions of studies of expertise recognition to more naturalistic organizational settings where individuals may have varied motives for sharing expertise, and group members may draw upon different experiences in evaluating the expertise of others (Bunderson & Barton, 2011; Majchrzak, Jarvenpaa, & Hollingshead, 2007). Both the variety of interactions available to the leadership program employees, as well as the diversity of tasks they encounter as they operate across project teams, makes this group of workers an interesting population with which to study expertise recognition.
Data for this study were collected using a survey that was sent to all members of the leadership program in both the marketing and operations business groups (N = 99), supplemented by data provided by American Financial management related to organizational rank and seating location. The survey included questions that captured demographic data, communication practices, relationships with coworkers, and perceptions of the expertise of coworkers in the group. Specifically, data were collected on what respondents believed to be areas in which they were experts and what they believed to be the expertise of coworkers in their respective leadership program group. In total, 89 fully completed surveys were returned (response rate of 89.9%), and partial data were present for all 99 individuals. Rather than exclude additional material and lose valuable data, and because several variables are based on the assumption of a full network of leadership program group members, we imputed averages from the sample for the missing values. (Less than 7.5% of all possible entries were missing. We also ran analyses that excluded single missing data entries and any respondents with missing data, and found no significant differences in the models.) Thus, our full sample is N = 99.
Dependent Variable
The dependent variable in this study is the extent to which others in the leadership program accurately identified the self-perceived expertise of each other member in their respective group, which we refer to as expertise recognition. In the survey, each respondent was asked to provide three pieces of task-related knowledge about which they considered themselves an expert. They were not required to list three areas of expertise; however, of those responding, only five individuals only listed two areas, one individual only listed one area of expertise, and all other respondents listed three areas of expertise. Respondents were then presented with a list of everyone in their respective leadership program group (48 individuals in Marketing, 51 individuals in Operations) and asked to indicate, for each of those people, three pieces of task-related knowledge about which they believed that individual had expertise. Again, they were not required to list three areas and, in this section, were not required to list any areas of expertise for workers they noted they had not previously worked with (this was the case for more than half of all individuals evaluated; however, workers included three areas of expertise for all group members they listed as knowing). Two independent raters (one with familiarity with American Financial) then took the answers provided by each respondent about what types of task-related knowledge others possessed and looked to see whether these perceptions matched with the self-reports provided by their coworkers. Following the guidance provided by Krackhardt (1987) for studying shared perceptions among group members, we examined the full Cognitive Social Structure (CSS) of perceptions of group members’ expertise and considered the intersection of each Locally Aggregated Structure (LAS) regarding an individual’s self-perceived expertise and the perception of his or her expertise from each other group member. This means that we looked at the expertise listed by Worker A and compared this with what Worker B listed as Worker A’s expertise, and then repeated this until we examined how each other group member evaluated Worker A. To establish the expertise recognition score for each individual, we took the aggregate number of times another group member listed that a worker had a form of expertise that that group member had listed as a form of expertise for him or herself, and divided that by the size of the respective leadership program group. The equation below describes the procedure used:
For example, if John listed that he was an expert in fraud prevention, and Debra listed that she thought John was an expert in fraud prevention, that would be one instance of expertise recognition. The higher the expertise recognition score, the more successful the worker was in having his or her expertise known by others. Cohen’s kappa for inter-rater reliability for the coding of matches in perceptions of workers’ expertise was .83. The expertise recognition score was an average of .13, and ranged from 0 to .625 (out a possible score of 1).
This definition of expertise recognition was adopted because it most appropriately represents the ways that expertise would relate to task assignments or knowledge sharing in this organizational context. Using the example above, if John considers himself an expert in fraud protection but no one else in the group holds this view, it is unlikely John would be given tasks related to fraud protection or be seen as a valued resource for knowledge in the area. Alternatively, if others consider John an expert on fraud protection but he does not share this view of himself, John is less likely to seek tasks or offer knowledge related to this domain. It is also important that expertise recognition was calculated agnostic to the domain or exclusivity of the expertise. Our goal was to examine expertise recognition within a specific organizational content and focus on expertise as viewed by workers with different roles and experiences. A result of this is that areas of expertise varied in their specificity and applicability. Some forms of expertise were relevant to a specialized piece of software used by a particular project team while others were related to tasks that cut across groups such as organizing data. Because we were interested in the recognition of expertise, and not examining the specific consequences of this recognition, the relevance of differences in forms of expertise is not captured by our operationalization of expertise (for discussion of how the specificity of expertise can influence group performance, see Littlepage et al., 2008). Coding of expertise recognition did not require the exact same words (i.e., spreadsheets and Excel were coded as a match; fraud investigations and fraud protection coded as a match), and coders were directed to match domains regardless of their scope or task relevance.
Using the congruence of perceived expertise from the potential expert and the assessor as a measure of expertise recognition is critical given the cross-sectional nature of these data. Because this study deals with judgments made at a single point in time, our analysis provides limited insight into how or why these perceptions developed. Asking workers to indicate the expertise they viewed as present or needed related to their respective work—rather than examining gaps between perceptions and some objective measure—increased the likelihood that workers provided responses reflective of situated perceptions and behaviors at the time of data collection. There are several limitations and trade-offs in this approach that will be addressed in the Discussion section of this article.
Independent Variables
Unique expertise
Using the same pool of responses used to determine expertise recognition, a measure was constructed for each group member based on the number of times a form of self-perceived expertise he or she provided was also listed by another coworker in the member’s respective leadership program group. This was conducted using the same approach to matching as was used for the expertise recognition score (a result of 0 for an individual would mean that no other workers listed any of the self-perceived forms of expertise provided by that respective worker). Therefore, this variable actually represents the frequency, or commonality of expertise, and a negative relationship with expertise recognition would reflect an influence for unique expertise.
Needed expertise
On the survey, respondents were also asked to list three forms of expertise they needed to complete their job. Again, they were not required to list three forms of expertise but only five respondents listed two forms of expertise. The variable for needed expertise represents the number of times, for each worker, that a listed form of self-perceived expertise was mentioned by another group member as something that was needed. Here again, the same approach to coding and analysis was used as with the expertise recognition and unique expertise measures, including accounting for the size of each individual’s leadership program group.
Physical proximity
A variable was constructed to serve as a representation of the number of members, within each respective leadership program group, an individual would likely encounter or generally be exposed to during the course of regular work. Following Allen’s (1977) work on the relationship between proximity and the likelihood of communicating with or overhearing communications by coworkers, a cutoff point of 30 m was set as the perimeter within which one might expect exposure to others. A scaled seating chart diagram provided by the company was used to determine the number of group members (from that person’s respective leadership program group) who sat within a 30 m radius of the focal individual. Each individual received a score indicating the number of coworkers falling within this radius (range = 0-15).
Friendship network centrality
This variable was calculated by asking respondents to look down a list of coworkers from their respective leadership program group and indicate the names of people considered to be personal friends. The friendship data were arranged into a square matrix for each division with cell entries of 0 or 1. Because friendships are reciprocal relationships, they were only counted as an existing relation if both workers listed each other (i.e., Kilduff, 1992). For this reason, the matrix was symmetrized using the rule that if both members of a pair nominated the other, the pair was considered to be a friendship relation. Using this symmetrized matrix, normalized degree centrality scores were calculated for each individual in the network. The degree centrality score gauges the quantity and strength of direct ties that a member has with others in the network (Freeman, 1979).
Task advice network centrality and norm advice centrality
Similar to the method used to measure friendship relations, workers were asked to list the individuals they went to for advice. In the case of tasks, workers were asked, “Who would you go to for information about how to complete your specific work tasks?” And for norms, they were asked, “Who would you go to for information about work-related norms?” Unlike for friendship, these relations did not need to be symmetric. As Carley and Krackhardt (1996) argued, asymmetric ties in advice networks should be considered as structural characteristics of social interaction, not as errors to be corrected. Following from H5, out-degree was used as a measure of degree centrality in the task advice network. An actor’s out-degree centrality is defined as the number of ties emanating from him or her; a person’s out-degree score is the number of people that person goes to for advice (Casciaro, 1998). Following from H6, in-degree was used as a measure of degree centrality in the norm advice network.
Communication variables
Each of these variables was calculated based on self-reported measures of how frequently employees used various communication technologies available to them at American Financial. Employees were asked to list on a Likert-type scale of 1 to 5 (1 = never, 5 = always) how frequently they used seven different communication technologies at work. The measure for interpersonal technologies was calculated using the average of responses for email, phone, and the instant message application Sametime. The measure for communal technology used the average of the reported use of wikis, blogs, and a shared file repository. The measure of overall communication frequency factored in all of six of the mentioned technologies, and use of the corporate directory, which contained personnel information of other workers.
Control Variables
Demographic variables
Survey responses for gender and age (in years) were used. For analysis of gender, females were used as the reference group and males were the comparison group. The survey also asked about the highest level of education reached for each worker, but this variable was not included in analysis because of a lack of variability among respondents (98% of respondents had received a bachelor’s degree).
Organizational rank
Respondents were assigned one of eight organizational levels within American Financial based on different positions they could attain in the organization. This information was provided by American Financial and not self-reported.
Work interaction
This variable represents that number of other individuals in a person’s leadership program group that the employee worked with (at the time of the survey or previously) on a project team. As discussed earlier, though all respondents were members of a collective leadership program group, they also rotated positions of project teams and, as a consequence, some had more independent roles and some had more collaborative assignments with these teams during their tenure. To construct this variable, respondents were presented with a roster of everyone in their leadership program group and asked to indicate those individuals with whom they were currently working or had worked with in the past. We used the matched responses to determine each individual’s work interaction. Because individuals could have worked on a number of project teams during their employment, but the analysis consists of cross-sectional data, this is the most appropriate way to control for the influence of project team interaction on within-group expertise recognition at the leadership program group level. To control for the influence of workers’ current project teams, but not past teams, would unduly privilege the current teams while neglecting the influence of previous team interactions.
Tenure
This variable was measured using the number of months of employment at American Financial. These figures were provided by the company’s human resources department.
Analysis
We tested the effects of the independent variables on the success of expertise recognition using ordinary least squares (OLS) regression analyses. Because the population consisted of individuals from two different leadership program groups, we first determined whether significant differences existed among individuals across these groups—both in terms of the dependent variable and other variables of potential influence. We compared the groups in two ways. First, to test whether leadership program group membership influenced expertise recognition, we ran a full regression model including a dummy variable for leadership program group membership. Although the full model significantly predicted expertise recognition (F = 4.58, p < .05), leadership program group membership was non-significant (p = .72). Next, we conducted a Chow test (Chow, 1960) to determine whether significant differences were present between the coefficients of each independent variable when separate regressions were run for each group. The Chow test was non-significant (F = 0.87, p = .61). Based on these results, and the fact that we standardized the expertise measures based on group size, we determined it was appropriate to pool the members of the two leadership program groups for analysis.
Because we were interested in the variables both as individual influences on expertise recognition and as potentially different types of influence based on the level of individual agency possible, we used hierarchical regression models. First, we estimated a baseline model that included the demographic variables (age, gender, work interaction, tenure, and organizational rank) as controls. This allowed us to both determine whether these variables alone exercised a significant influence on expertise recognition and set a standard upon which we could measure the additional influences of the other variables. Subsequently, we added blocks of predictor variables to determine whether the addition of these variables produced a significant change in explaining the dependent variable of expertise recognition. Because, as we discussed earlier, potential influences were grouped together based on the amount of agency individual workers had in manipulating them, we adopted a hierarchical approach to determine if a particular type of variable influenced expertise recognition. Recognizing the differential agency of workers in relation to the included variables helps reveal the extent to which workers may or may not have the capacity to actively change influences on expertise recognition.
Results
Table 1 displays the descriptive statistics and the zero-order correlations among all of the variables examined. Several variables were significantly correlated with expertise recognition, which presented a need for tests of multicollinearlity. The variance inflation factor (VIF) for each variable was less than 3.5, which is considered an acceptable, conservative number with which to proceed with model testing (Hair, Black, & Babin, 2009).
Means, Standard Deviations, and Zero-Order Correlations (N = 99).
p < .05. **p < .01.
The effects of the independent variables on individual expertise recognition were then tested using multiple regression analyses, and the results are displayed in Table 2. First, a baseline test was run including the control variables to establish the amount of variance in expertise recognition explained by these factors (see Table 2, Model 1). Although organizational rank was significantly correlated with expertise recognition, and the model as a whole explained a statistically significant amount of variance in expertise recognition (F = 3.02, p < .05), the overall amount of variance explained is low (R2 = .14) when considering only the controls. Then, keeping these controls in the model, we tested the subsequent blocks of predictor variables described earlier, proceeding from those least influenced by group members to those workers can directly influence through communicative action (Table 2, Models 2-4). Using these results, we can examine each grouping and variable individually to investigate the hypotheses.
Predictors of Success at Advertising Expertise (N = 99).
Note. Standardized coefficients presented.
p < .05. **p < .01.
H1a, H1b, and H2 tested how the scarcity, commonality, and expressed value of the workers’ self-perceived forms of expertise were related to the likelihood of that expertise being recognized among group members. H1a and H1b examined the relationship between the uniqueness of an individual’s expertise and the recognition of that expertise by coworkers. However, given the variable used measured a lack of uniqueness, a negative relationship would indicate that uniqueness predicted expertise recognition (H1a) and a positive relationship would indicate that having common expertise predicted expertise recognition (H1b). H2 speculated that the expressed need for a form of expertise would lead to greater recognition of that expertise. Adding both of these variables to the model did not significantly improve the explanatory power of the model (see Table 2, Model 2) and neither variable was significant in subsequent models (Models 3 and 4). Therefore, we found a lack of support for H1a, H1b, and H2.
H3 and H4 predicted that the extent that one was exposed to, or had the structural potential to interact with, others at work would influence the ability of a worker to have his or her expertise recognized within the group. When these variables were entered into the regression (see Table 2, Model 3), the model demonstrated significant improvement over the previous model with the demographic variables and the group structural variables (F test model improvement = 7.99 p < .01). Although the physical proximity of workers to peers did not significantly explain success in advertising individual expertise, centrality in the organizational group friendship network was significant in Model 3 (β = .31, p < .01), suggesting an influence on recognition of an individual’s expertise. However, when additional variables related to communication practices were added in Model 4, the significant effect of friendship centrality disappeared, indicating that this influence was better explained by the alternative variables. Therefore H3 and H4 were not supported.
The last group of hypotheses predicted that the communication practices of workers would influence whether a worker had his or her expertise recognized. H5 speculated that greater centrality in seeking task advice would result in greater recognition of one’s expertise within the group, and H6 explored the relationship between centrality in being sought out for advice on organizational norms and success sharing one’s expertise. H7a to H7c posited that various forms of active individual communication—frequency of communal technology use, frequency of interpersonal technology use, and overall communication frequency—would each have a positive influence on expertise recognition. When these five variables were added to the regression, the resulting model represented a significant improvement from the previous model in the amount of variance of expertise recognition explained (F test model improvement = 5.52, p < .01). Looking at each variable individually, task advice centrality (β = .29 p < .01; see Model 4), and communal media use (β = .21 p < .05; see Model 4), both had a significant influence on having one’s expertise recognized. In this model, organizational rank also remained a significant influence (β = .32, p < .01; see Model 4) as it had in the previous models. The effects of personal technology use and overall communication frequency were not significant. Therefore H5 and H7a are supported and H6, H7b, and H7c are not supported. We examined the residuals for each variable in Model 4, and although they largely exhibited an even distribution, there was some indication of heteroscedasticity. Therefore, we estimated an OLS using heteroscedasticity-consistent standard errors (Hayes & Cai, 2007), and the results replicated what was found in our original OLS models.
Finally, because of the number of predictors included in the model and the finding that most variables were non-significant, we tested a more parsimonious regression model with only the variables that were significant in the model including all the predictors (see Model 5) in an effort to determine if the non-significant variables explained a meaningful level of variance beyond the significant predictors. At this stage, our goal was to use the post hoc analysis to provide insight into the relative strength and fit of a model based only on these three predictors. Model 5 not only explained significantly more variance in expertise recognition than Model 3, but the explained variance was not substantially different than the results of Model 4. Therefore, we concluded that Model 5 is the most appropriate model for predicting expertise recognition.
The presence of significant zero-order correlations among several of the predictors (see Table 1)—specifically the significant communication practice variables—made it difficult to specify the influence of each individual variable on expertise recognition. To further investigate the relationship between each variable and expertise recognition, a commonality analysis (Kraha, Turner, Nimon, Zientek, & Henson, 2012; McPhee & Seibold, 1979) was conducted to determine the unique effect of each variable in Model 5, and the first-order common effect of each pair of variables together. The results, presented in Table 3, indicated that task advice centrality had the greatest unique effect in explaining the variance in expertise recognition (accounting for 31.12% of the total variance explained). The analysis also revealed that when organizational rank was examined along with task advice centrality, the common explained variance was negative (−5.45% of variance). Because organizational rank cannot be altered directly through individual actions (i.e., we cannot interpret this as indicating an increase in a worker’s task advice centrality measure would cause a movement down in organizational rank), we cannot conclude that either variable suppresses the effect of the other on expertise recognition. Instead, this provides evidence that though organizational rank and task advice centrality are significantly related to expertise recognition, they appear to represent two different and relatively distinct paths to expertise recognition.
Commonality Analysis of Significant Predictors of Expertise.
Discussion
The primary contribution of this work is the finding that the most influential path explaining whether a group member’s self-perceived expertise was recognized by others was through communication. Our findings revealed that when expertise recognition was operationalized as expertise sharing, neither structural factors related to the supply and demand for particular forms of knowledge nor the frequency with which people might interact with others predicted whether a worker had his or her expertise recognized. In other words, the attributes over which self-perceived experts have little or no control—such as whether or not their expertise is needed or unique within the group or where their desk is situated in the office—did not play a significant role in determining whether someone will be recognized as an expert.
Self-perceived experts who were most likely to have their expertise recognized by group members engaged in more active advice seeking related to tasks and more communal communication. We do not make any claims regarding the intent or strategic nature by which this expertise was communicated, only that it was primarily through means that can be shaped by individual workers. Broadly, these findings support the idea that perceptions coworkers develop about the expertise of other group members are shaped by a variety of communicative signals that individuals provide in the ongoing practice of work (Bunderson, 2003; Treem, 2012; Yuan et al., 2013). As Su (2012) noted in his study of expertise identification, “expertise recognition is fundamentally a communicative process” (p. 631). More specifically, by looking across structural, relational, and communicative factors that might influence expertise recognition, this study responds to calls for research examining how features of groups and worker interactions influence perceptions of who knows what in organizational groups (Bunderson & Barton, 2011; Lewis & Herndon, 2011)
The findings of this study are in line with some, but not all, of the influences on expertise recognition suggested by previous theories of how expertise is recognized among group members. For instance, we did not find any significant relationship between the uniqueness or commonality of individuals’ self-perceived expertise in the group and a greater likelihood they would have their expertise recognized (H1a and H1b). This null result does not align with findings suggesting that distinctiveness of expertise will improve expertise recognition (Austin, 2003), or research on the mutual enhancement effect, which would predict greater recognition for common and shared areas of expertise (Wittenbaum et al., 1999). Though the data do not indicate the specific mechanism underlying this result, it is possible that because only four group members noted completely unique expertise, and there was a small range of variance in the level of unique expertise among group members, the findings are a function of the distribution of expertise in these groups. Future research should consider this question in more detail by comparing groups characterized by high differentiation of members’ expertise and settings where all members share expertise. In addition, although previous research has demonstrated that working remote from a group negatively affects the ability of an individual to recognize the expertise of group members (Su, 2012), our results did not find that working near a larger number of other team members influenced the recognition of a worker’s self-perceived expertise (H3). However, individuals did not work near many other group members (an average of only slightly more than five other group members sat within 30 m); therefore, an inclusion of some measure of whether workers shared digital spaces (i.e., were on common Listservs or accessed shared information repositories) would have been useful in capturing the likelihood of interaction among group members. Future work should explore the different influences of the physical and technologically mediated distance of group members on expertise recognition.
The significant influence of out-degree centrality of individuals in the task advice network (H5) indicates that workers might effectively communicate what they do know by asking for help about things they do not know and supports the notion that expertise is communicated through the regular practice of work. However, here we see the notion of judging expertise inverted. Studies of knowledge exchange in organizations commonly examine how the advice seeker views the expertise of the advice giver (i.e., Borgatti & Cross, 2003). However, because the focal interest is the leveraging of knowledge from pre-identified experts, this approach discounts the knowledge provided by the advice seeker. The results of this research bolster findings that advice exchanges lead to judgments of knowledge from both the advice seeker and the advice giver and that signals of knowledge are actively conveyed in this communicative process (Bonaccio & Dalal, 2006; Sniezek & Van Swol, 2001).
The finding that communication through communal communication technologies was related to expertise sharing and expertise recognition (H7a), but that interpersonal communication technology use (H7b) and overall communication technology use (H7c) was not, contributes to a growing body of literature on the role of communication and technology use in supporting knowledge sharing and perceptions of coworker expertise (Heinz & Rice, 2009; Wasko & Faraj, 2005). Previous research on the relationship between communication among individuals and expertise recognition has largely examined dyadic communication consisting of allocation or retrieval of information (i.e., Jarvenpaa & Majchrzak, 2008; Palazzolo, Serb, She, Su, & Contractor, 2006). However, the findings in this study indicated that, when specifically dealing with technologically mediated communication, interpersonal communication might be less effective in sharing meta-knowledge about expertise compared with communal technologies. Additional research is needed to see if the use of communal technologies by workers has similar effects for other measures of expertise recognition.
The use of communal communication technologies, such as social media, may afford workers the opportunity to more effectively share expertise because these media facilitate communication that is potentially visible to everyone in an organization (Treem & Leonardi, 2012). This visibility may increase the likelihood of assessors forming shared perceptions of a communicators’ expertise in two ways. First, when making decisions, workers are more likely to rely upon information that they know is shared by other group members (Stasser & Titus, 1985). By communicating in a shared space, a worker is able to send the same signal of knowledge to everyone. Second, because the individual has communicated in such a public, visible manner via this technology, it may be seen as a more reliable communicative signal of what that person knows or who he or she is, and therefore may be more influential in shaping impressions (Donath, 2007). Both the visibility of signals of expertise in communal technology and the symbolic value of one’s willingness to make information visible make this information useful for making attributions of a worker’s expertise.
The influence of organizational rank is interesting because it indicates a potential relationship between perceptions of expertise and organizational power though the nature of that relationship is not immediately clear and cannot be evaluated with these data. Those in higher organizational positions may be visible due to their status and having their expertise recognized could be a function of subordinates paying more attention to what superiors do at work, or it could be that the ability of these individuals to advertise their expertise helped them rise in the organization. Alternatively, organizational rank may be a heuristic tool whereby workers assume people at a certain level in the organization have particular abilities. In addition, the different types of expertise exhibited at different organizational levels could influence this relationship. Managerial expertise may be either more visible or identifiable than other more specialized forms of expertise carried out at different levels. More research is needed to explore the relationship between organizational rank and expertise.
It is important to consider that expertise in this study was measured based on the matching of perceptions between the expert and his or her coworker—expertise was shared—and was not based on any objective measure or achieved status. Moreover, the forms of expertise investigated were not predetermined by the researchers but rather were generated by the respondents with the goal of retaining ecological validity regarding what they considered expertise. Within this design, some employees were very successful in having their self-professed expertise recognized and others were not. This raises questions about the nature of expertise in organizational settings. Organizational expertise is often associated with the idea of esoteric knowledge, or the idea that experts possess skills and abilities that few others have or understand (Starbuck, 1992). Yet the esoteric nature of expertise also means it can be difficult for organizations, workers, and even scholars to describe the forms of expertise that may be present in a particular context (Alvesson, 2001), and this makes it difficult to compare findings related to expertise recognition and utilization in groups (Lewis & Herndon, 2011). Our findings regarding the relationship between active communication and expertise sharing support the view that expertise is contextually defined in organizations (Agnew, Ford, & Hayes, 1997).
Caution should be exercised when applying the conclusions of this work more broadly. First, by design, our study departed from earlier work on expertise recognition in groups by looking at ongoing groups of larger sizes and with a more fluid team structure than those previously examined. In doing so, we purposely violate the posited scope conditions of these theories. This allows us to look at what relations are consistent across settings, but we have limited ability to determine whether the null results for many hypotheses are a function of the group size or structure. Overall, because this study examines perceptions of workers’ expertise among individuals in a single organization, the findings may not be easily generalizable to other work contexts. There are other aspects of our data that limit the conclusions we can draw from this analysis. A number of variables relied on self-reported measures by employees and may not accurately capture the relations or communication practices among workers. Also, although the workers operated with both collective goals and individual task goals, we did not have a specific measure of task interdependence, which is seen as an important precondition for expertise recognition in small task-based groups. Future research can address expertise recognition across groups where members have different forms of task interdependence. In addition, it is important to recognize that use of a survey and cross-sectional data meant that our analysis does not examine the specific communication processes that comprise the variables included. Therefore, although our results are in line with research that treats expertise recognition as a co-constructed process between the expert and the assessor, we do not present that process of co-construction in this work. However, overall, the goal of this work was primarily to demonstrate the validity and value of viewing organizational expertise as something that can be shared by, and among, individuals in work settings. Future research can examine the varied ways that workers share their expertise in diverse organizational settings and the concurrent consequences this has for knowledge sharing, task assignments, and group performance.
Finally, the findings of this work have practical implications, and even offer encouragement, for workers seeking to have their self-perceived expertise recognized by coworkers. The relationship between communication practices and expertise recognition indicates that workers can potentially advertise their expertise to other group members. This means that despite the initial distribution of expertise in a group, or assumptions made about group members based on attributes or stereotypes, active communication with coworkers may increase the likelihood of a worker having his or her expertise recognized.
Footnotes
Acknowledgements
The authors thank Casey Pierce and Stephanie Dailey for their assistance with data collection and Lindsay Young for her help with data analysis. In addition, they are grateful to Michael Roloff, William Barley, David Seibold, Ronald Rice, and three anonymous reviewers for their comments and feedback on this work.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Generous funding for this study was provided by three grants from the National Science Foundation (SES-1331492, SES-1057148, and ACI-1322103).