Page - (000233) - in Knowledge and Networks

228 not be officially designated as leaders by…management.” We did not specify what we meant by the term “leader” because we were interested in capturing respon- dents’ implicit theories of leadership (Lord & Maher, 1991/1993, p. 11). Advice Centrality We assessed centrality in the advice network by counting the number of times an individual was identified by others as someone they tended to turn to for advice on work related matters. Control Variables Rank This variable was coded as 1 for supervisors and as 0 for non-supervisors. The data for coding this variable came directly from company records. We con- trolled for rank because high-ranking individuals, due to the resources they control and the prestige they enjoy, are likely to garner informal leadership influence on the basis of their reward power (French & Raven, 1959). Tenure Taken from company records, this was the number of months the person had been employed by the company. Job Performance In the research on work organizations, the majority of perfor- mance ratings tend to come from supervisors (Bretz, Milkovich, & Read, 1992, p. 331). Meta- analytic evidence suggests that supervisory evaluations represent valid assessments of true performance (Arvey & Murphy, 1998, p. 163). Our three- item measure of job performance was based on confidential ratings provided by each individual’s direct supervisor. Performance ratings that are collected for research purposes tend to more reliable and valid than those obtained for adminis- trative purposes (Wherry & Bartlett, 1982). The three items that made up our mea- sure were selected on the basis of multiple discussions with a panel comprising the firm’s human resources director, a long tenured member with broad knowledge of the company, and four employees who represented a range of different job types at the firm. The items asked supervisors to rate subordinates in terms of (a) their “over- all job performance” (1 = poor, 5 = excellent); (b) the likelihood that the subordinate would “achieve future career related success (such as promotions, awards, bonuses, and involvement in high profile projects)” at the company; and (c) would be some- one the supervisor would pick as a successor for their job (1 = highly unlikely; 5 = highly likely). The reliability of the scale as measured by Cronbach’s alpha was .83. We considered and then rejected the inclusion of individual sex (male/female) as a control variable. This variable was insignificant in all the analyses reported below, but its inclusion produced poorer fitting models. M. Kilduff et al.