This study empirically investigates the effectiveness of Distributed Leadership Teacher Training (DLT) program on improving student's academic achievement. In addition, it both tests the assumption that the year 1 impacts are stable across calendar years and examines the importance of properly accounting for the fact that the standard error of the first-year impact estimate, which becomes the measurement error of the variable (the first-year impact point estimate) used for adjustment in the two-year impact analysis, needs to be taken into account in the analyses. It discusses the extent to which the findings of this examination of the Bell and Bradley adjustment approach have general applicability. The Distributed Leadership Teacher Training Program (DLT) was implemented in a large urban school district. A total of 26 elementary and middle schools participated in the five-year demonstration project. An early analysis of first year impacts for the first cohort of schools and students showed found some evidence of impacts on teacher-leadership team behavior, but no impacts on student outcomes (Cole, 2008). The reasons that no significant impacts on student outcomes for this early sample could be because the intervention really will not impact student outcomes; but, it also could be because it takes time for the impacts to move from teacher leaders to student outcomes; but it also could be due to the small sample size (not enough power). This second study of the program includes seven more elementary schools and one more wave of data in the impact analysis. The two-year impacts potentially would be larger than the one-year impacts and the larger sample size is has more power to detect the program effects (albeit still not as much as desired). This study will obtain unbiased estimates of one- and two-year impacts of DLT on student's math and reading achievement. Furthermore, Bell & Bradley's (2008) assumption will be empirically examined. (Contains 2 figures.)