Nonstructural analysis of productivity growth for the industrialized countries: a jackknife model averaging approach

Various nonstructural models of productivity growth have been proposed in the literature. In either class of models, predictive measurements of productivity and efficiency are obtained. The different approaches to productivity and efficiency measurement all have their merits. However, they generate different results and it is in principle impossible to know which of them produces ”true” estimates. We argue that instead of choosing between approaches, the analyst can average across the results generated by them. Unfortunately, this begs the question which averaging method to use? This paper examines the model averaging approaches of Hansen and Racine (2012), which can provide a vehicle to weight predictions (in the form of productivity and efficiency measurements) from different nonstructural methods. We first describe the jackknife model averaging estimator proposed by Hansen and Racine (2012) and illustrate how to apply the technique to a set of competing stochastic frontier estimators. The derived method is then used to analyze productivity and efficiency dynamics in 25 high-industrialized countries over the period 1990 to 2014. Through the empirical application, we show that the model averaging method provides relatively stable estimates, in comparison to standard model selection methods that simply select one model with the highest measure of goodness of fit.