In dubio pro CES

Capital-labor substitution and total factor productivity (TFP) estimates are essential features of growth and income distribution models. In the context of a Monte Carlo exercise embodying balanced and near balanced growth, we demonstrate that the estimation of the substitution elasticity can be substantially biased if the form of technical progress is misspecified. For some parameter values, when factor shares are relatively constant, there could be an inherent bias towards Cobb-Douglas. The implied estimates of TFP growth also yield substantially different results depending on the specification of technical progress. A Constant Elasticity of Substitution production function is then estimated within a “normalized” system approach for the US economy over 1960:1–2004:4. Results show that the estimated substitution elasticity tends to be significantly lower using a factor augmenting specification (well below one). We are able to reject Hicks-, Harrod- and Solow-neutral specifications in favor of general factor augmentation with a non-negligible capital-augmenting component. Finally, we draw some important lessons for production and supply-side estimation.