Letters to the Editor 1169 Comments on the article by Tracey Trottier, Montgomery Van Wart, and XiaoHu Wang, “Examining the Nature and Significance of Leadership in Government Organizations,” PAR 68(2) Aleksey Tikhomirov University of Binghamton Studying Leadership in the Government Sector: Issues and Opportunities Certain issues relevant to exploring the nature of leadership in the governmental sector received little attention in Tracey Trottier, Montgomery Van Wart, and XiaoHu Wang’s study of transformational and transactional leadership in government settings, pub- lished in the March/April 2008 issue of PAR. Re- viewed here, these issues mark opportunities for advancing the understanding of leadership processes in the government context. Levels of Analysis Te need for clearly specifying the levels of analysis for leadership constructs and theorized interactions is long argued (Dansereau, Alutto, and Yammarino 1984; Rousseau 1985). By stating the level at which a leadership phenomenon is asserted to theoretically exist, one ensures that the measurement of constructs/ relationships and the data analysis are consistent with the specified levels of analysis (Yammarino et al. 2005). Otherwise, one may not be certain about the empirical testing of appropriate levels of analysis, as well as the properness of inferences about leadership drawn from such analysis. Multiple levels may be expected in the data set used by Trottier, Van Wart, and Wang. Items in their table 1 appear to associate leadership with the be- havior of individual leaders, but also of a collective of individuals in leadership positions. Under the assumption that the data set allows authors to match multiple subordinate reports to specific lead- ers, the use of multilevel data-analytic techniques might have been appropriate. One of the following three procedures can be considered when approach- ing analysis of multilevel data, including multilevel random coefficient modeling (such as hierarchical linear models; see Bryk and Raudenbush 1992), within and between analysis (Dansereau, Alutto, and Yammarino 1984), or cross-level operator anal- yses in regression and analyses of variance ( James and Williams 2000). Tese procedures test for the level-specific leadership effects and outcomes and allow for drawing level-specific conclusions about them.