Working Paper: NBER ID: w19937
Authors: Weili Ding; Steven F. Lehrer
Abstract: Unobserved ability heterogeneity has long been postulated to play a key role in human capital development. Traditional strategies to estimate education production functions do not allow for varying role or development of unobserved ability as a child ages. Such restrictions are highly inconsistent with a growing body of scientific evidence; moreover, in order to obtain unbiased parameter estimates of observed educational inputs, researchers must properly account for unobserved skills that may be correlated with other inputs to the production process. To illustrate our empirical strategy we use experimental data from Tennessee's Student/Teacher Achievement Ratio experiment, known as Project STAR. We find that unobserved ability is endogenously developed over time and its impact on cognitive achievement varies significantly between grades in all subject areas. Moreover, we present evidence that accounting for time-varying unobserved ability across individuals and a more general depreciating pattern of observed inputs are both important when estimating education production functions.
Keywords: education production functions; unobserved ability; cognitive achievement; Project STAR
JEL Codes: C23; I21
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| unobserved ability heterogeneity (D29) | educational achievement (I24) |
| unobserved ability heterogeneity is developed over time (D29) | influence on cognitive achievement (I24) |
| influence on cognitive achievement varies significantly across grades (I24) | effects diminish from grade 1 to grade 3 (A23) |
| unobserved ability heterogeneity must be accounted for (D29) | accurate estimates of education production functions (C51) |
| ignoring unobserved ability heterogeneity (D29) | misleading conclusions about educational inputs (I26) |
| impacts of educational inputs (I21) | sensitive to how unobserved ability is treated (C92) |