Leader-Member Exchange and Citizenship Behaviors: A Meta-Analysis
Illies, et al.
Journal of Applied Psychology
January 2007 Vol. 92, No. 1, 269-277
As interest in both LMX and citizenship behaviors at work has increased exponentially in the past 2 decades (see Podsakoff et al., 2000, and Schriesheim, Castro, & Cogliser, 1999), we believe that a comprehensive meta-analysis that addresses multiple aspects of the relationship between LMX and citizenship behavior via moderator analyses is needed. Such a comprehensive meta-analysis is needed for three primary reasons. First, there appears to be a high degree of variability in empirical estimates from individual studies for the strength of the relationship. For example, Wayne, Shore, Bommer, and Tetrick (2002) found a relatively weak relationship between LMX and citizenship behavior (r = .20) whereas Tekleab and Taylor (2003) found a much stronger relationship (r = .52). Hackett et al. (2003) suggested that this variability can be explained by sampling error and differential reliability. However, these authors included only a small set of studies in their analyses (as we explain in this article, we identified and coded 33 additional studies) and based their conclusion with respect to homogeneity on analyses that did not include the 3 effect sizes (out of 18) that were most different from the mean.
Second, different authors related LMX to different types of citizenship behaviors (e.g., helping, altruism, job dedication). Because LMX may differentially influence specific citizenship behaviors, the type of citizenship behavior examined in particular studies may have influenced the nature and magnitude of the observed relationship with LMX. It is entirely possible that such differential effects explain the variability in empirical estimates of the LMX-citizenship behaviors correlation.
Third, different studies have used various methodologies, including ratings of the relationship between LMX and organizational citizenship behaviors from different rating perspectives (Tierney, Bauer & Potter, 2002; Townsend, Phillips & Elkins, 2000) and different measures with different reliabilities (Janssen & Van Yperen, 2004; Settoon et al., 1996), and thus the estimates are not directly comparable. Furthermore, it is useful to investigate whether methodological factors, such as the rating source, further moderate the relationship between LMX and individual- and organization-targeted citizenship behaviors.
