1 INDIVIDUALS' FORMAL POWER AND THEIR ...

25 downloads 0 Views 471KB Size Report
AND THEIR SOCIAL NETWORK ACCURACY: A SITUATED COGNITION PERSPECTIVE. Joshua E. Marineau. North Dakota State University. Daniel J. Brass.
1 INDIVIDUALS’ FORMAL POWER AND THEIR SOCIAL NETWORK ACCURACY: A SITUATED COGNITION PERSPECTIVE

Joshua E. Marineau North Dakota State University Daniel J. Brass Stephen P. Borgatti University of Kentucky And Patrizia Vecchi University of Modena

1 INDIVIDUALS’ FORMAL POWER AND THEIR SOCIAL NETWORK ACCURACY: A SITUATED COGNITION PERSPECTIVE

ABSTRACT Individuals differ in the accuracy of their perceptions of the social environment, but research and theory provide conflicting predictions on whether those with power are more or less accurate than others. Drawing on social network theory and the situated focus theory of power, we examine the relationship between individuals’ formal power and their perceptual accuracy of social network relationships. We propose that individuals’ perceptual accuracy is affected by: 1) their formal power in the organization 2) the type of relationship being perceived (expressive/instrumental, positive/negative) and its relevance to task and goal completion, and 3) the dependence relationship with the target of perception (whether the perceiver is dependent on the perceived to get their work done). Predictions were tested using cognitive social network data collected from a technical call center within a division of a large corporation in the US. Results showed that formal power was linked to increased accuracy for some relationship content (negative expressive relationships), and managers tended to be more accurate than non-managers when perceiving their own personal incoming relationships.

KEYWORDS: Formal Power, Social Network Analysis, Network Perception and Accuracy, Instrumental and Expressive Ties, Positive and Negative Ties, Situated Cognition

1

Social network theory views organizations as a web of formal and informal social relationships where each individual occupies a position based on their ties or relationships to others (Borgatti, Mehra, Brass, & Labianca, 2009). Research based on this view suggests that these informal relational ties are key sources of benefits for individuals, including unique information, social support, and status, and also liabilities, including negative gossip and undermining, all of which can influence individual outcomes (e.g., Borgatti et al., 2009; Brass, Galaskiewicz, Greve, & Tsai, 2004; Duffy, Ganster, & Pagon, 2002; Ellwardt, Labianca, & Wittek, 2011; Kilduff & Brass, 2010; Labianca & Brass, 2006; Mehra, Kilduff, & Brass, 2001). Most network research is based on the assumption that ties have their effects regardless of whether third parties are aware of those ties. For example, in his theory of structural holes, Burt (1992) argues that a person with ties to, say, five unconnected others has an advantage over someone who has the same number of contacts who are all tied to each other. The first individual is more likely to receive independent, non-redundant information. The information benefits to this individual accrue regardless of whether third parties are aware of the individual’s ties. However, other mechanisms suggested to underlie the benefits or liabilities of networks presume that the other members of the network perceive these relations. For example, an individual can enhance their status by having ties to high-status players, but only if others are aware of these ties. Thus, the cognitive or perceptual school in the study of social networks holds that to fully appreciate the effects of networks we must take account of third-party perceptions of ties. This cognitive perspective assumes that each individual may hold a unique view or perception of the informal ties and relationships around him or her in the social network.

2 These perceptions of informal relationships form individuals’ mental maps of the social world, or cognitive social structure (CSS) (Krackhardt, 1987). Cognitive social structures are the relationships that a person perceives to exist among actors (including one’s own ties) in the network. For example, individuals hold perceptions of who interacts with whom and who are friends with whom, and these beliefs and perceptions can be unique for each individual in the network (e.g., Freeman, Romney, & Freeman, 1987). These structures are integral to how individuals interpret, experience, and interact with the social environment, influencing their decisions and subsequent behaviors (e.g., Baldwin, 1992). Given the enormous amount of information individuals are exposed to in their social environment, it is reasonable to believe that not all social information is processed, or perceived equally accurately (Fiske & Taylor, 1991; Kenny & West, 2010). Indeed, research has shown that individuals’ cognitive maps are prone to large amounts of error (e.g., Bondonio, 1998; Casciaro, Carley, & Krackhardt, 1999), which is not surprising when considering that even a small network of only 25 individuals has 600 (N*(N-1)) dyadic relationships among them. Thus, individuals’ knowledge of the social relationships around them can vary, with some people having considerably more knowledge than others. The degree of social knowledge an individual possesses can have significant consequences; indeed, accurate knowledge of the network an individual is embedded in can be seen as a vital resource in itself, providing the individual with a source of competitive advantage. For example, Krackhardt (1990) showed that individuals with more accurate knowledge of the social relationships around them had more power in their organizations, which would enable to them to get things done and perform well in their jobs. This may be particularly important for those in formal power positions in the organization, individuals who can determine the rewards

Commented [sb1]: Ties are potential; dyads are always there.

3 and punishments of subordinates and affect the behavior of others by controlling valued resources (Fiske & Berdahl, 2007; French & Raven, 1959; Keltner, Gruenfeld, & Anderson, 2001), such as managers. For example, if a manager needs to assign employees to a work team, knowing which employees have conflict or negative relationships could allow her to create a more effective team with less performance-destroying relationship conflict (Jehn & Mannix, 2001; Jehn, Greer, Levine, & Szulanski, 2008; Shaw et al., 2011). While accurate perceptions of relationships may be particularly important for those in power, having a position of formal power in an organization may in itself affect the accuracy of one’s cognitive social structure. Indeed, power has been linked to behavioral and cognitive consequences for those who wield it (Galinsky, Rus, & Lammers, 2011). In her summation of the current research on the effects of power on individuals, Guinote noted that “power has a fundamental impact on how individuals perceive the world around them and how they behave” (2010: 141). Despite the importance of understanding this relationship, the effects of power on cognitive social structure have received little research attention. Broad predictions of how power affects social structure perceptions have been tested (Casciaro, 1998; Simpson & Borch, 2005; Simpson, Markovsky, & Steketee, 2011). The few results have been consistent, but limited in scope—power tends to be related to decreased accuracy of the social network. Casciaro (1998) tested a number of variables including hierarchical level and personality in relation to friendship and advice network accuracy among a group of 24 Italian scientists. She predicted that formal power would be positively related to advice network accuracy, arguing that people in power positions had greater access to work-flow related information. She also predicted that they would be inaccurate with regard to the friendship network because those in

4 formal power positions are likely to avoid friendships to subordinates and subordinates would be reticent about befriending those in power. This would limit managers’ access to knowledge about the friendship network. Results supported the predictions about friendship accuracy, but not for advice tie accuracy—both advice and friendship tie accuracy were negatively related to formal power. Simpson et al., (2011) used a lab experiment to test the relationship between feeling powerful and social structure perception. They found that feeling powerful was related to less accurate recall of relationships in an influence network, suggesting that high-power individuals perceived more relationships than actually existed. However, feeling powerful did not significantly relate to learning the relationships that did exist (Simpson et al., 2011). While both these studies indicate that power is associated with decreased accuracy of the social network, recent theorizing in the social psychology of power suggests that the relationship might not be so clear. Our contribution to social network theory is to incorporate the situated cognition approach to power from social psychology (Guinote, 2010) into understanding how formal power might relate to network accuracy. New theoretical insights and research suggests that power is related to increased flexibility in, attention to, and cognition of social information (Guinote, 2007a) such that the powerful are able to block out peripheral information that is not task- or goal-oriented, while the powerless do not possess the same level of “situated” cognition (Guinote, 2010). Another relevant perspective is the power-as-control model advanced by Fiske (Fiske, 1993a; Fiske, 1993b), who argues that the powerful view the less powerful differently, relying on stereotypes in thinking about other people because their power positions allow for less cognitive effort in thinking about the social environment (Fiske, 1998). Similarly, Overbeck and Park (2006) argue that the manner in which individuals depend on each other is key to understanding the effects of power on social cognition. This body of work suggests that power

5 leads to increased accuracy in certain task- and goal-oriented contexts and for role-activated social content, as well as awakening the attention of the powerful toward the less powerful. Thus, in this study, we will argue that those in formal power positions tend to have increased accuracy with respect to perceiving two kinds of network ties: information-sharing/advice ties, and conflict ties. In addition, those high in formal power will also be accurate in perceiving the network connections of specific others on whom they depend to get their work done. Finally, we will also argue that these managers are less accurate about positive expressive relations, which are less directly tied to task- and goal-oriented outcomes. We begin by articulating a view of managers as differentially motivated in how they perceive their social environment in comparison to non-managers. In particular, we outline the differences this perspective implies for managers’ and non-managers’ accuracy for the entire constellation of network relationships within the workplace – the whole network. We then consider how accurate managers and non-managers are about each other’s personal social networks, and finally we compare managers’ and non-managers’ accuracy for their own incoming social network ties. Thus, we move from perceptions of the whole network to perceptions of the ties of certain target individuals (supervisors or subordinates) and end with a person’s perceptions of their own incoming ties.

Theory and Hypotheses The predominant view of how power affects perception suggests that having power results in increased use of stereotypes and more heuristic processing of social information, while powerlessness is related to increased systematic or controlled processing of social information (Depret & Fiske, 1993; Fiske, 1993a; Goodwin, Gubin, Fiske, & Yzerbyt, 2000; Goodwin,

Commented [sb2]: The level of analysis is always the individual

6 Operario, & Fiske, 1998). Powerful individuals are seen as cognitive misers, very limitedly expending cognitive energy on the less powerful , and generally more lackadaisical in attending to social information, which results in less perceptual attention and accuracy (Fiske, 1993a). Powerful individuals are less attentive than the powerless because they have more control over their own outcomes and so can afford to ignore others, while those at the bottom of the organizational hierarchy are dependent on the powerful and must attend to more closely in order to ensure good outcomes for themselves (Keltner et al., 2001). Also, because of the pyramidal nature of most organizational hierarchies, the more powerful are generally outnumbered by those lacking formal power, which means that there are usually more low-power individuals to attend to in the social environment than there are high-power individuals. This creates a higher cognitive load for those in formal power positions, who would need to keep track of many more powerless individuals and their network ties, as compared to the powerless, who can focus their attention on a much small number of powerful others. These factors suggest that “attention follows power” and predict that on average, individuals with formal power will be less accurate about the social world than the powerless (Fiske, 1993a). A few social network studies have attempted to relate formal power with accuracy of the whole network of relationships in an organization. Krackhardt (1990) controlled for formal power in his field study of network accuracy, but results showed no significant relationship between hierarchical level and mental map accuracy of advice and friendship networks. Casciaro (1998) argued that individuals in higher hierarchical positions would have a greater interest in and access to work-related ties, such as instrumental advice ties, than expressive, friendship ties. However, in her field study she found that hierarchical level was negatively related to both friendship and advice network accuracy. Simpson, et al., (2011) theorized that

7 low power individuals would make fewer errors when learning social networks than high power individuals. High- and low-power positions did not differ statistically when only considering ties that existed (present), but low-power subjects were more accurate on absent ties. Thus, although the evidence is somewhat mixed, these studies suggest that power has some influence on how individuals perceive social information generally, and social networks specifically. However, there has been a movement in social psychology away from viewing power as leading to more automatic processing and powerlessness as leading to more controlled processing (Galinsky et al., 2011), and towards a more nuanced view of the powerful. This situated cognition perspective suggests that the powerful are more flexible and have more selective attention in where and when controlled processing of social stimuli is activated as compared to the powerless (Guinote, 2007a). While power might relate to cognitive processing that leads broadly to inaccuracy, powerful individuals’ motivated cognitions will affect cognitive priorities, filtering out peripheral or non-essential information thus increasing accuracy of pertinent goal or task-oriented (social) information. This approach views cognition as situated, in that individuals’ social cognitive processes are adaptive and not static or insulated from context (Smith & Semin, 2004). Cognitive processing operates selectively, as it adapts to the “primary factors” of the situation, such as environmental cues, expectancies, and goals (Guinote, 2010: 147-148). Indeed, recent work has shown that powerful individuals are not cognitive automatons, but have flexibility and focus in where and how they prioritize their attention and cognitive energy (e.g., Overbeck & Park, 2006). This view is best articulated by the situated focus theory of power (Guinote, 2010) , which accounts for how the powerful can inhibit peripheral information and instead focus on information related to relevant constructs such as goals, needs, or tasks (Guinote, 2007b).

8 The situated focus theory combines a cognitive bias view and an objective view to explain the various effects of power on individuals’ judgment and behavior variability. The bias approach relates to mindsets, or higher cognitive level biases that are “strategic informationprocessing orientations” (Guinote, 2010) such as approach and inhibition, selectivity, and a use of a wider range of mental processes. The objective approach relates to basic cognition, such as flexibility in attention and focus. The reasoning is that individuals in positions of power have fewer constraints and are therefore able to selectively aim their attention at goals and needs. The powerless are more constrained, have less control over their destinies, and therefore must attend to a multitude of information sources in order to reduce uncertainty and increase predictability and control (Guinote, 2010). Individuals with power are able to focus attention toward goalrelevant cues, expectancies and information which aid goal achievement, and block out peripheral information irrelevant to the activated constructs (Guinote, 2007a; Overbeck & Park, 2006). This model reconciles findings that suggest power produces both focused, as well as less focused attention for information processing. The key component of this model, as mentioned above, is the notion that cognition is situated; that is, context or situation plays a key role in how power affects cognitive and behavioral processes. Individuals navigate the social world with limited cognitive capacity, and they are guided by social psychological processes which “guide social attention and depth of processing”—and interdependence, or social power is a particularly important factor affecting these differences (Guinote & Vescio, 2010: 2). Consistent with this situated focus perspective, we argue that the situation (work-related goals and priorities activated by the formal power inherent in manager status) will differentially affect perceptions of various types of workplace social relationships, which we organize along

Commented [J3]: Need page number

9 two axes. The first axis distinguishes instrumental ties from expressive ties (Fombrun, 1982; Ibarra, 1995; Lincoln & Miller, 1979; Podolny & Baron, 1997; Tichy, Tushman, & Fombrun, 1979). Instrumental ties involve gathering resources necessary to accomplish a task. Expressive ties involve expressions of interpersonal affect. The second axis distinguishes the perceived valence of the relation (positive or negative). Combining the two axes, we study four canonical types of workplace ties: task trust, task distrust, friendship, and dislike. Task trust relates to “judgment based on evidence of another’s competence and reliability” (Chua, Ingram, & Morris, 2008: 437). Following Chua (2008), we define task trust ties as when a person relies on another to “complete a task they agreed to do […] and have the knowledge and competence needed to get tasks done” (442). Task distrust ties arise when individuals do not believe that others are competent with a task or able to complete it.

--------Insert Figure 1 about here-------

Power affords individuals the freedom to focus on features of the environment which are important and to pay less attention to information that isn’t. Managers are likely to prioritize social information pertinent to their primary function, which is the management of individuals in their achievement of tasks and goals. An important aspect of task and goal achievement in a work setting is how individuals interrelate instrumentally, in terms of getting tasks done they said they would complete and having the competence to do the task correctly (Chua et al., 2008). We argue that managers will be especially interested in understanding whether individuals have positive or negative relationships with respect to these instrumental task-related ties, because it will often be important to ensure that individuals trust each other to complete tasks assigned to

10 them, and where distrust reigns, to either resolve it or find some manner of managing it. Thus, individuals with formal power (i.e., those in managerial roles) are more motivated than subordinates to perceive, focus on, and accurately describe relationships related to goal and task completion – task trust and task distrust ties.

H1: Managerial status will be positively related to accuracy in perceiving the instrumental task trust and task distrust networks.

The second axis represents expressive ties, which involve expressions of interpersonal affect and convey a sense of belonging. As with the instrumental ties described above, we view these expressive ties as being valenced as well (e.g., Labianca & Brass, 2006). Positive expressive ties (i.e., friendships) in organizations are sources of social support, provide a sense of identity and personal belonging, and serve to transmit normative expectation (Coleman, 1988, 1990). Negative expressive ties denote interpersonal dislike, can be a threat to one’s security and identity, and can serve as counter-role models. Friendship ties are often the focus of studies of network perceptual accuracy (e.g., Krackhardt, 1990), but our situated cognition perspective suggests that those in formal power positions will be less motivated to attend to these relations. We would, therefore, expect that managers would be less accurate about friendships in the workplace network than would nonmanagers, which would replicate Casciaro’s (1998) empirical findings. However, it is important to recognize that expressive ties can also be negatively valenced (e.g., Umphress, Labianca, Brass, Kass, & Scholten, 2003). Negative expressive ties denote ongoing and stable interpersonal dislike (Labianca & Brass, 2006; Labianca, Brass, & Gray,

Commented [sb4]: If something is ongoing, it doesn’t recur.

11 1998). As such, they can be a threat to one’s security and identity, and can serve as counter-role models—an employee might strive to be dissimilar to a person he or she dislikes (French & Raven 1959). These negative ties should garner greater cognitive activity and attention than would positive expressive ties (see Labianca & Brass, 2006). One explanation for this negative asymmetry is that negative ties are more threatening because they increase the likelihood of goal frustration. Indeed, recent research in workplace networks demonstrates that negative ties lead to direct attempts to harm the other (e.g., Venkataramani & Dalal, 2007), as well as indirect attempts through such mechanisms as negative gossip (e.g., Ellwardt et al., 2011; Grosser, Lopez-Kidwell, & Labianca, 2010). Thus, we argue that managers will be strongly motivated to perceive these negative expressive relationships because of their heightened potential to impede goal completion. On a more mundane level, it is likely also true that managers have a greater tendency to be called in to be mediators of interpersonal disputes in organizational settings as compared to non-managers because bureaucratic structures and logics often dictate that conflicts are to be mediated or arbitrated by those with formal authority (Perrow, 1986). Managers’ greater awareness, sensitivity to, and exposure to these negative expressive relationships will relate to greater accuracy of expressive negative ties in the network relative to non-managers.

H2: Managerial status will be negatively related to accuracy in perceiving the positive expressive (friendship) network. H3: Managerial status will be positively related to accuracy in perceiving the negative expressive (dislike) network.

12 The general approach taken to understanding network accuracy has been to compare how accurate individuals’ mental maps of the entire network relative to the actual whole network reported by the network members (e.g., Krackhardt, 1990; Casciaro, 1998). However, our situated cognition perspective emphasizes that attention and, hence, accuracy is motivated. As such, the social environment should not be treated simply as a whole, but rather theoretical predictions about where in the network individuals are more likely to be motivated to perceive accurately can be made at a dyadic level. One such example that is critical to account for is examining the network accuracy of two individuals entwined by a reporting relationship (Fayol & Coubrough, 1930). We would expect that, in general, individuals bound by a reporting relationship would be more motivated to attend to each other’s networks than would two organizational members that are free of this bureaucratic bond. In early accounts of the consequences of power on cognition, researchers tended to predict that attention follows power (Fiske, 1993a), which would lead to the conclusion that managers will not attend to subordinates as much as vice versa. However, more recent accounts based on the situated cognition perspective show that managers do not necessarily ignore subordinates and are not always lazy observers of the social world. For example, Overbeck and Park (2001) argued that high-power individuals will be more likely to focus on the lower-power individuals over which they had outcome control than low-power individuals would focus on those who had power over them. In a series of experimental studies, they found that high-power individuals were better able to recall information and behavior of subordinates than subordinates were able to recall about their superiors (Overbeck & Park, 2001). Thus, we argue that because managers are responsible for their direct reports, and the performance of subordinates reflects upon the manager, managers will prioritize their own subordinates’ relationships over the

13 relationships of others in the network. This will result in increased attention and perceptual focus on their subordinates’ network relations, resulting in greater accuracy for those subordinates’ ties as compared to other organizational members generally.

H4: Managers will be more accurate about their subordinates’ social networks than about their non-subordinates in the organization.

Of course, the reporting relationship has another party – the subordinate – who should also be motivated to attend to their managers, and their managers’ network relations. Subordinates will be more cognizant of their own managers’ networks compared to the networks of others in the organization, such as managers from other departments or groups to whom they do not report, as well as other subordinates. This is because individuals in positions of low power are motivated to attend to those who have the power to control their outcomes (Fiske & Dépret, 1996; Keltner et al., 2001).

14 In addition, there should also be cognitive energy devoted toward the source of the subordinates’ potential rewards and punishments in relation to others the subordinate might have a dependence relationship. While subordinates may be dependent in terms of their goal and task outcomes on their peers within their group, they should be more intensely motivated to attend to their manager than their fellow subordinates. Thus, subordinates should be accurate about their manager’s social network ties, and they should be more accurate for their managers’ network ties than the social network ties of peers even within their own group.

H5a: Subordinates will be more accurate about their managers’ social networks than about others’ social networks in their organization. H5b: Subordinates will be more accurate about their manager’s social network than for each other’s networks within their department.

Having generated hypotheses about the effects of power on accuracy of the entire network and accuracy toward particular other individuals (subordinates), we now turn to one’s accuracy about one’s personal network— that is, how accurate an individual is about the relational ties he or she receives from others. Just as theories of power bring special attention to managers in their perceptual accuracy of the broader network (H1-H3) and how power plays a role in network accuracy in a dyadic framework (H4-H5), power can also influence an individuals’ perception of their own incoming ties in the social network. Based on the previous arguments, we would expect that managers will be more accurate than subordinates about their own incoming instrumental ties and expressive negative ties. However, self-regulatory theories of power, suggest that the powerful are unconcerned about some social information, such as how

15 others view them. For example, Keltner and colleagues (Keltner et al., 2001) argue that having social power will activate an individuals’ approach system, which is related to goal pursuit, rewards and avoidance of punishment and experience of positive affect. Powerlessness activates the inhibition system, which is sensitive to punishment, non-rewards and is related to negative affect, effectively limiting goal pursuit. Thus, the power that a person holds can affect their behavior, cognition and emotions affecting their goal pursuit, feelings about others and processing of social information (Keltner et al., 2001). Recent studies have supported the approach/inhibition theory of power. Anderson and Berdahl (2002) found high power individuals expressed their opinion and attitudes more openly, experienced more positive emotions, and were more likely to perceive that their partner liked them (i.e., rewards) than when their partner was angry with them (i.e., punishments). The less powerful, in contrast, are more likely to attend to threats in their immediate environment, in an effort to predict outcomes and gain some sense of control (Fiske, 1993a; Fiske & Dépret, 1996). The powerful are less concerned with threats because their power allows them to control their own outcomes, where powerless individuals cannot. Thus, managers with formal power, will direct less attention and focus to incoming negative ties, resulting in less accuracy compared to those without formal power. There is some irony in this hypothesis – while those in formal power positions are more motivated to see accurately negative ties among others, they are also more likely to be blind to others’ negative ties directed at them.

H6: Managerial status will be negatively related to incoming negative (dislike and task distrust) tie accuracy.

16 METHODS Sample and Setting We conducted this research in a Southern U.S. division of a multinational focused on design and manufacturing of advanced automatic espresso machines. The sample came from the firm’s technical call center which serviced customers (e.g., restaurants) and service providers (e.g., on-site service techs) with service support. The employees were expert in the sophisticated technical aspects of equipment operation and repair, and received regular training on new equipment and features. The call center employees were housed in one large open room with only partial cubicle barriers between them and were organized into four functional groups. Groups 1 and 2 were the largest groups and were responsible for day-to-day service calls (n = 25 & 17, respectively). Group 3 was responsible for scheduling regular maintenance with new and existing customers (n = 6), and Group 4 consisted of administrative personnel (n = 4). There were 52 individuals in the call center, including 12 managers. One individual was a non-respondent and 3 had incomplete data (n= 48; 92% response rate). All 12 managers provided data (100% response rate). The sample was 41% female and 62% minority (two individuals did not disclose their ethnicity); 29.2% were college graduates, and 62.5% had some college or post high school education. Organizational tenure ranged from 5 to 160 months (M = 34.4).

Data We collected data in three stages. We began with 25 semi-structured interviews to develop our survey. We then conducted an online survey sent to all four groups (n=52) which included demographic and personal social network questions. Finally, we collected cognitive social structure (CSS) data from Groups 1 and 2 (n=42) where each individual provided a

17 complete map of their perceptions of the ties among everyone else in their respective groups using a matrix design (see Appendix 1 for a detailed description of CSS data collection).

Dependent variable: Accuracy of network perception We created accuracy measures for all three levels of perception: whole network accuracy (H1-3), target in-out accuracy (H4-5), and self incoming tie accuracy (H6). Each measure used the Jaccard index in UCINET (Borgatti, Everett, & Freeman, 2002) to calculate accuracy. This measure captures the agreement among two sets of ties (perceived and criterion) as a proportion of the set of all ties. Thus, a person’s accuracy was based on the proportion of agreement their perceived network had with the criterion network compared to the total set of ties reported in either networks. Mathematically, if A is the set of perceived ties and B is the set of criterion ties, the Jaccard coefficient is the size of the intersection of A and B divided by the size of their union: J(A,B)= |A∩B|/|A∪B|. 1 The analyses for H1-5 required CSS data, and were conducted on the smaller sample, while H6 could be conducted on the overall sample of 52 (see table 1 for a description of the levels of analysis and the portion of the sample used). The criterion network (the “true” network) was calculated as a locally aggregated structure (LAS; Krackhardt, 1987) for the analyses presented, though we also calculated an alternative criterion matrix using a Bayesian Network Accuracy Model (BNAM: Butts, 2003). Results were consistent across both methods of calculating the criterion networks. Individuals’ perception of the target’s incoming and outgoing ties compared to the target’s actual incoming and outgoing ties was used for the target in-out analyses (H4-H5), and the individual’s report of his or her own network ties was 1

The more common method of evaluating network accuracy using a Pearson (e.g., Casciaro, 1998) was problematic in our setting because of the large number of zeros in the adjacency matrices. This was especially true for the negative tie networks, which tend to be exceptionally sparse. Indeed, nine of our CSS dislike matrices and eight task distrust CSS matrices had no ties whatsoever, rendering the Pearson correlation coefficient, which requires both variables to have variance, incalculable.

Commented [sb5]: They are all dyadic.

18 used to build the criterion ties for H6 (see Appendix 2 for an in-depth description of how the criterion networks were constructed).

-----Insert Table 1 about here-------

Independent Variables Formal Power. Formal power was measured as managerial status, as defined by the companies’ formal hierarchy, and consistent with prior work (e.g., Casciaro, 1998). There were 12 managers and 40 subordinates (manager = 1; non-manager = 0). The managers in this setting held strong outcome control over their subordinates. Raises and bonuses were wholly based on goal achievement, which was set for both individuals and departments by the direct supervisor and department head, respectively. Managers were also key in determining which subordinates could be promoted out of the call center and into positions elsewhere in the organization (receiving a “negative write-up” by a manager made an employee ineligible for a promotion or new position for a six month period). This variable was used for H1-3 and H6. To test the accuracy of managers’ perceptions about their subordinates and subordinates’ perceptions about managers (H4-5), two 42x42 matrices were created based on the formal reporting roles provided by the organization. The “manager of” variable was a matrix which indicated if person i was the manager of person j. The “subordinate of” (or reports to) variable was a matrix which indicated if person i was the subordinate of j. If the relationship between individuals i and j was present, that cell received a “1” and a “0” otherwise.

19 Control Variables Workflow Centrality. 2 Employees were required to work with others both within and outside their workgroup. Using a roster method, each individual was asked to select those people from whom they either received inputs or sent outputs to. These “required” workflow ties might serve as additional information about network ties, so we controlled for individuals’ indegree centrality in this required workflow network for H1-3 and H6. For the dyadic level hypotheses (H4-5), we symmetrized the required workflow network on minimum and entered the 42x42 matrix in the model. Network Centrality. Because central individuals might have more access to social information in the network (e.g., Bondonio, 1998; Casciaro, 1998; Casciaro et al., 1999; Krackhardt, 1990), we controlled for individuals’ indegree centrality in all four of the networks (task trust, task distrust, friendships, and dislikes) in H1-3 and H6. For the target person analyses (H4-5), we used the outgoing ties from the row dominated LAS for each network and transposed those matrices to determine the incoming ties. This resulted in eight matrices: “Task Trust Out”, “Task Trust In”, “Task Distrust Out”, “Task Distrust In”, “Friend Out”, “Friend In”, “Dislike Out”, and “Dislike In”. CSS Density. The number of ties that a person perceives in their CSS can vary significantly, so we included each person’s CSS density (the number of ties recorded over the total number of possible network ties) for each type of tie in the appropriate regression model. For H6, these perceiver differences were controlled for by including each individual’s count of perceived incoming ties in the appropriate regression model (e.g., Casciaro et al., 1999). Dyadic Similarities. We controlled for dyadic similarities between individuals because homophily might impact the sharing of network information and relate to individual’s accuracy 2

For H1-3, all centrality measures were normalized to control for the different group sizes.

20 of others’ networks (e.g., McPherson, Smith-Lovin, & Cook, 2001). We created attribute similarity matrices for ethnicity, gender, and sharing the same manager (1 = match, 0 = none) using the “Attribute to matrix” function in UCINET (Borgatti et al., 2002). This resulted in three 42x42 control matrices: “Same Gender”, “Same Ethnicity”, and “Same Manager”. Sender and Receiver Effects. To control for the possibility that some attributes might be related to either paying more attention to network ties (sender or perceiver) or attracting more attention (receiver or target), we created three matrices for use in the dyadic analysis (H4-5): “Gender,” which was the sender’s gender, “Gender (target)”, and “Manager (target)” (manager as sender was already included as an independent variable in those analyses). Embedded (Simmelian) Ties. Individuals who are strongly tied to others through reciprocated friendships with third parties can have increased similarities in network perceptions (Krackhardt & Kilduff, 2002; Krackhardt, 1999). The “Simmelian/Embedded Tie” algorithm in UCINET (Borgatti, 2012) was used on the friendship network to create the 42x42 matrix variable “Simmelian Tie”.

Analytical Approach We used linear regression with robust standard errors to model the relationship between formal power and network accuracy for H1-3 and H6 (e.g., Hansen, Mors, & Løvås, 2005; Moulton, 1986; Rogers, 1994). This was because the association between the dependent and independent variables was expected to be linear and the data were cross-sectional. To control for potential supervisor effects, standard errors were pooled within supervisor in the regression models testing H6. These models were run in STATA v12 (StataCorp, 2011). Hypotheses 4-5 focused on examining one person’s accuracy in perceiving another

Commented [sb6]: You mean errors were pooled by supervisor?

21 person’s ties as a function of those individuals’ relationships to each other (manager or subordinate). These dyadic relational data are non-independent and thus, we used Multiple Regression Quadratic Assignment Procedure (MRQAP; Dekker, Krackhardt, & Snijders, 2007; Krackhardt, 1988). All data in square 42x42 matrix form. 3

RESULTS Whole Network Accuracy (H1-3) Hypothesis 1 proposed a significantly positive relationship between manager status and perceptual accuracy for instrumental ties (task trust and distrust). Results are presented in Table 4, and they do not support H1. Managerial status was not significantly related to either task trust accuracy (b = -.04, t = -0.54, ns) or task distrust accuracy or (b = -0.08, t = -0.76, ns). Hypothesis 2 proposed that manager status would be significantly negatively related to accuracy with respect to the positive expressive (friendship) network. However, manager status was not significantly related to friendship accuracy (b = 0.01, t= 0.08, ns), rejecting H2. Hypothesis 3 proposed that manager status would be significantly positively related to perceptual accuracy in the negative expressive (dislike) network. Results found a positive, significant coefficient for dislike network accuracy (b = .30, t =2.06, p < .05), indicating that managers were more accurate than non-managers about dislike ties in the organization. Thus H3 received support. 4

3

Network responses from three individuals were missing, and these were left blank as missing data. Additionally, because the Groups 1 and 2 were different sizes, and the CSS data were collected within groups (not across) the dependent variable matrices have, by definition, missing data between groups. We thus used the “Double-Dekker Semi-Partialling MRQAP” (Dekker et al., 2007) algorithm in UCINET. This allowed for partitioning the variable for groups to permute only within groups or “node classes” (Borgatti, S. 2012. UCINET Software Versions 1 Jan 2012 - 31 Jan 2012, Vol. 2012. https://sites.google.com/site/ucinetsoftware/versions/versions1jan2012-31jan2012. ) 4 Despite a lack of support for some our formal hypotheses, we do find significant positive correlation between manager status and task trust perceptual accuracy (r = .352, p < .05) in Table 2. The correlation between manager status and perceptual accuracy of task distrust (r = .175) and dislike (r = .218) were also positive, but not significant. Counter to our prediction that positive instrumental ties would have a negative relationship to manger status, friendship accuracy was also positively related to manger status (r = .308). These correlations suggest that

22

------Insert Table 2 & 3 about here--------

Target Person In-Out Accuracy (H4-5) Hypothesis 4 argued managers would be more accurate about their subordinates’ ties than the ties of others in the organization. MRQAP results (see Table 5) showed that managers were only marginally more accurate for their subordinates’ incoming dislike ties (b= .08, p