A Monte Carlo Study on the Fit of a Cost Function by a Fourier Flexible Functional Form

Abstract

We use the simulation model of Chalfallt and Gallant (1985) to investigate the fit of a Fourier flexible form to a cost function when interest is to estimate elasticities of substitution. The data is subject to errors in variables and the estimation method is seemingly unrelated regressions. We use two approximating Fourier functional forms with 13 (FFF13) and 22 (FFF22) parameters. The biases in point estimation are negligible for FFF13. The classical t statistics do not follow, in general, the Student's distribution. Even when estimates are properly centered and scaled there are cases where the normal approximation does not hold.