School districts across the United States increasingly use value-added models (VAMs) to evaluate teachers. In practice, VAMs typically rely on lagged test scores from the previous academic year, which necessarily conflate summer with school-year learning and potentially bias estimates of teacher effectiveness. We investigate the practical implications of this problem by comparing estimates from "cross-year" VAMs with those from arguably more valid "within-year" VAMs using fall and spring test scores from the nationally representative Early Childhood Longitudinal Study–Kindergarten Cohort (ECLS-K). "Cross-year" and "within-year" VAMs frequently yield significant differences that remain even after conditioning on participation in summer activities.