For too long, research on retrospective voting has fixated on how economic trends affect incumbents' electoral prospects in national and state elections. Hundreds of thousands of elections in the United States occur at the local level and have little to do with unemployment or inflation rates. This paper focuses on the most prevalent: school boards. Specifically, it examines whether voters hold school board members accountable for the performance of their schools. The 2000 elections reveal considerable evidence that voters evaluate school board members on the basis of student learning trends. During the 2002 and 2004 school board elections, however, when media (and by extension public) attention to testing and accountability systems drifted, measures of achievement did not influence incumbents' electoral fortunes. These findings, we suggest, raise important questions about both the scope conditions of retrospective voting models and the information voters rely upon when evaluating incumbents.