The shape of aggregate production functions: evidence from estimates of the World Technology Frontier

The article provides multifaceted evidence on the shape of the aggregate country-level production function, derived from the World Technology Frontier, estimated on the basis of annual data on inputs and output in 19 highly developed OECD countries in the period 1970–2004. A comparison of its estimates based on Data Envelopment Analysis and Bayesian Stochastic Frontier Analysis uncovers a number of significant discrepancies between the nonparametric estimates of the frontier and the Cobb–Douglas and translog production functions in terms of implied efficiency levels, partial elasticities, and returns-to-scale properties. Furthermore, the two latter characteristics as well as elasticities of substitution are found to differ markedly across countries and time, providing strong evidence against the constant-returns-to-scale (CRS) Cobb–Douglas specification, frequently used in related literature. We also find notable departures from perfect substitutability between unskilled and skilled labor, consistent with the hypotheses of skill-biased technical change and capital–skill complementarity. In the Appendix, as a corollary from our results, we have also conducted a series of development accounting and growth accounting exercises.