Policy Evaluation and Empirical Growth Research

This paper provides a critique of the use of growth regressions to derive policy implications. The author challenges the conventional interpretation of empirical results, arguing that current econometric practice has yielded a body of evidence that is not policy relevant. Extending his own previous work, the author raises two issues of critical importance for policy purposes. First, policy recommendations arising from growth regressions are usually based on the statistical significance of some regression coefficients, which does not necessarily constitute a valid evaluation of alternative policy trajectories. Moreover, the statistical significance of a parameter does not provide information on the relative merit of it for policymakers’ objectives. Second, growth regressions as conventionally constructed do not provide credible evidence of economic structure. Consequently, policymakers are unable to make better decisions based only on regression results. The author proposes an alternative approach to the interpretation of growth regression based on Bayesian averaging techniques, which allow the weighing of different growth determinants relative to the pay-off function of the policymaker and in the context of model uncertainty (because the modeler does not know what growth determinants must be included in a model or what forms of country-level heterogeneity need to be accounted for in the model).