Cleaning up the kitchen sink: Specification tests and average derivative estimators for growth econometrics

Theory and case-study evidence suggest that non-linearities are pervasive in the growth process. Growth empirics have attempted to characterize these non-linearities with regression trees, additively separable non-parametric estimates, or simple interaction terms. Each method requires specific assumptions about functional form which we demonstrate may not be defensible. We provide two alternate mechanisms for making inference about the growth effects of production–function shifters that do not make a priori assumptions about functional form: monotonicity tests and average derivative estimation. Our results suggest that the growth effects of policies are country-specific while the effects of institutions are more robustly monotonic.