The research literature on principals’ effects on student achievement has employed value-added models that have only limited usefulness for actual performance evaluation systems. Principals’ effects are typically identified through school leadership changes, meaning that effectiveness data cannot be generated for all principals and that effects are comparable only within small groups of schools connected by principal transfers. In contrast, school value-added estimates, which are more commonly used in practice to evaluate principals, can be generated for all principals but include multiple influences beyond principal quality. We examine the extent to which school value-added reflects elements of principals’ effectiveness that persist over time and impact multiple grades. Using longitudinal data on fourth through eighth grade Pennsylvania students and their math and reading outcomes, we calculate the percentage of total variation in school value-added at a given grade that is explained by principal quality estimated from a different grade and time period. Our main findings indicate that the general leadership skills of principals who are in their first three years at their current position explain at most 14 percent of the variation in school value-added. We conclude that school value-added data do not provide substantial information useful for evaluating these principals.