CES Technology and Business Cycle Fluctuations

This paper contributes to an emerging literature that brings the constant elasticity of substitution (CES) specification of the production function into the analysis of business cycle fluctuations. Using US data, we estimate by Bayesian methods a medium-sized DSGE model with a CES rather than Cobb-Douglas (CD) technology. The main empirical result is to confirm decisively the superiority of CES rather than CD production functions in terms of model fit. We estimate a elasticity of substitution of elasticity well below unity at 0.15-0.18 and in a marginal likelihood race assuming equal prior model probabilities, CES beats the CD production decisively. The marginal likelihood improvement is matched by the ability of the CES model to fit the data in terms of second moments and a comparison with a DSGE-VAR further confirms the ability of the CES model to reduce model misspecification. We find that the CES model performance is further improved when the estimation is carried out under the imperfect information assumption. The principle reason for our result is that the CES specification captures movements of factor shares at the business cycle frequency. Hence the main message for DSGE models is that we should dismiss once and for all the use of CD for business cycle analysis.