Another Look at Panel Estimates of the Elasticity of Substitution between Capital and Labor

Abstract

We construct a broad panel with aggregate data for 94 countries over the period 1970-2014 from the latest version of the Penn World Tables (PWT) dataset, version 9.0. We estimate the elasticity of substitution between capital and labor based on the linearization of a Constant Elasticity of Substitution (CES) production function. In our preferred econometric specification, we find estimates of the elasticity of substitution that are below one. We obtain the same result when we correct the variables for cross-country differences in human capital. Additionally, estimation of a linearized CES for alternative panels constructed with data from older versions of the PWT corroborate our initial estimates. The finding that the elasticity of substitution is below one casts doubts on the Cobb-Douglas assumption commonly adopted in development accounting exercises, as well as on the recent literature on declining labor shares which assumes that the elasticity of substitution is above one.