Drawing on data from the Early Childhood Longitudinal Study-Kindergarten Class 1998-1999 of the United States, this article evaluates teacher expectancy effects on achievement growth in kindergarten. We attempt to disentangle teacher expectancy effects from omitted variable bias or predictive validity by exploiting counterfactual predictions in the summer months. If the estimates of academic achievement during the academic session (fall and spring) are corrupted by omitted variable bias or predictive validity, similar estimates predicting achievement growth during the summer break should be positive precisely because the interactions between teachers and students are absent but omitted variables still work during the summer months. The results for math test scores as well as reading test scores support the existence of remarkable teacher expectancy effects during the academic session, which do not appear to be contaminated by unobserved confounding variables or predictive validity.