| This paper studies the sensitivity of estimates on various assumptions about aggregation in modeling the school's effect in child educational production. By controlling for the endogeneity of school qualities in the production function, we evaluate the performance of a "correct" aggregation educational production model versus simple aggregation educational production model in predicting the school resources' effect on academic outcome. Monte Carlo simulations on different modeling specifications shows that simple aggregation of school resources over a geographic area causes serious specification errors, and thus generates biased estimates for the marginal contribution of the school resources to test scores. The two aggregation models are empirically estimated, and we find that having heterogeneity control in the production function reduces the estimated effect of school characteristics on test score. We also find that the "Correct" Aggregation model and Simple Aggregation Model perform differently in the empirical study. |
