How Large Are Returns to Scale in the U.S.? A View Across the Boundary

There is considerable disagreement in the empirical macro literature as to the degree of returns to scale in U.S. production. While many studies find evidence of a small degree of increasing returns, standard errors are typically large. This issue is of importance for assessing the possibility of equilibrium indeterminacy and sunspot-driven business cycles. The theoretical literature has shown that even small degrees of increasing returns can lead to non-uniqueness. Whereas earlier studies are based on production function regressions and single equation methods, this paper takes a structural and general equilibrium perspective. I argue that the question of indeterminacy is a property of a system and cannot be conclusively answered by single equation methods. Moreover, estimates of returns to scale will be biased when the possibility of indeterminacy and sunspot driven fluctuations is not properly accounted for. In this paper, I estimate a canonical business cycle model for the U.S. economy including variable factor utilization. Based on the methodology developed by Lubik and Schorfheide (2004) I use Bayesian methods to estimate the model over the entire parameter space, allowing for sunspot equilibria generated by increasing returns to scale in production. I find that returns to scale are increasing, but not considerably so. However, I do not find evidence of indeterminacy based on Bayesian posterior odds tests. When abstracting from variable capital utilization, estimates of the scale parameter increase, but again indeterminacy can be rejected. The results of this paper therefore suggest that increasing returns to scale are not the source for sunspot fluctuations in U.S. business cycles