There has been an active debate in the literature over the validity of value-added models. In this study, the author tests the central assumption of value-added models that school assignment is random relative to expected test scores conditional on prior test scores, demographic variables, and other controls. He uses a Chicago charter school's lottery to identify school effects, and then compares this "experimental" estimate to that of a school value-added model, which is estimated from all students in a large district. An innovative approach is used to generate a value-added-based estimator of the lottery effect that is unbiased under the null hypothesis that the value-added model is correctly specified. The results imply small but potentially important upward bias in the math, and very minimal bias in reading. Tables are appended.