Although there has been sizable growth in the number of empirical studies of shared forms of leadership over the past decade, the bulk of this research has been descriptive. Relatively few published studies have investigated the impact of shared leadership on school improvement. This longitudinal study examines the effects of distributed leadership on school improvement and growth in student math achievement in 195 elementary schools in one state over a 4-year period. Using multilevel latent change analysis, the research found significant direct effects of distributed leadership on change in the schools' academic capacity and indirect effects on student growth rates in math. The study supports a perspective on distributed leadership that aims at building the academic capacity of schools as a means of improving student learning outcomes. (Contains 2 tables, 2 figures and 8 notes.)