Value-added models are now a mainstay of research on teachers and are used by some states and school systems for teacher evaluation or accountability. Linear analysis of covariance (ANCOVA) models are widely used in the estimation of teacher value added. In this research memorandum, we use the potential outcomes framework to define teacher value-added effects as causal effects of the contributions of teachers and classroom peers, and we establish conditions for ANCOVA models to yield consistent estimates of those effects.