This content downloaded from 60.240.29.164 on Fri, 10 Jul 2020 10:51:45 UTC All use subject to
96 Academy of Management Journal February ments is required to form higher-level constructs. Thus, within-group agreement and between-group variability need to be demonstrated to justify data aggregation. We assessed within-group agreement on group-focused leadership using the i?wg statistic(James, Demaree, & Wolf, 1984). The mean and median #wg's for the 70 groups were .91 and .94, respectively, indicating a high level of within group agreement (Lance, Butts, and Michels [2006],citing James [1988]). We performed a one-way anal ysis of variance (ANOVA) to examine between group variability. The ICC(l) value was .22 (F = 3.36, p < .01), within the normal range found in organizational research (Bliese, 2000; Kirkman etal., 2009; Klein et al., 2000). The ICC(2) value was .70, reaching the desirable level suggested by Klein et al. (2000). These results supported aggregating the individual scores to the group level. Differentiated individual-focused transforma tional leadership. This construct was measured with two indicators: differentiated individualized consideration and differentiated intellectual stim ulation. These differentiated leadership indicators were based on eight items from the appropriate subscale of the MLQ5x (Bass & Avolio, 1995). Each indicator consists of four items (aggregated to the group level). The same response scale described above was used. The original MLQ wording was kept, with "individual" as the referent (e.g., "My group leader helps me to develop my strengths" and "My group leader challenges me to re-examine critical assumptions to question whether they are appropriate"). As a configural group property, dif ferentiated leadership fits Chan's (1998) dispersion composition model (in which within-group varia tion conveys the substantive meaning of a con struct) and represents the disparity type of diver sity within a group (Harrison & Klein, 2007). The within-group differentiation measure was opera tionalized as a coefficient of variation (Allison, 1978), a statistic that demography researchers use as a scale-invariant measure of dispersion (Tsui & Gutek, 1999). We calculated it by dividing the with in-group standard deviation of the individual-fo cused leadership measure by the within-group mean score of the same variable. The larger the value of this coefficient, the more dispersion there is in the group members' perceptions of leader be havior, given adjustment for mean differences be tween groups. Group identification. This construct was as sessed with two manifest indicators based on four items (e.g., "I identify myself as a member of my group" and "I identify with other members of my group") from Doosje, Ellemers, and Spears (1995). Each indicator was the average of two items (aggregated to the group level). The response scale ranged from 1, "totally disagree," to 5, "totally agree." As a shared group property, group identification follows a direct consensus model (Chan, 1998). Within group agreement (mean Rwg = .90, median i?wg = .95) and between-group variability (ICC[1] = .07; F = 1.64, p < .01) were examined before the aggre gation. Though the ICC(2) was less than satisfac tory, (.39) partly because the presence of some small groups in the sample, high within-group con sensus (demonstrated by the jRwg values), and suf ficient between-group differences (the significant F-test) suggested that data aggregation was justifi able (Bliese, 2000; Kirkman et al., 2009; Klein et al., 2000).
Want to read all 18 pages?
Previewing 9 of 18 pages Upload your study docs or become a member.
Want to read all 18 pages?
Previewing 9 of 18 pages Upload your study docs or become a member.
End of preview
Want to read all 18 pages? Upload your study docs or become a member.