|M.Sc Student||Allon Gad|
|Subject||Convex Entropic Nonparametric Estimation of Production|
|Department||Department of Industrial Engineering and Management||Supervisor||Professor Ury Passy|
The econometric analysis of production functions has a long history, dating back to the pioneering work of Cobb and Douglas. A constant theme to this history has been the search forever more flexible functional forms. This literature is entirely parametric. In the last 20 years nonparametric econometrics has grown in popularity. In the context of production functions, a nonparametric estimator is necessarily more flexible than the most flexible parametric methods. The disadvantage to nonparametric estimation is that the fact the size of the problem grows with the number of factors of production.
More than 30 years ago Hanoch and Rothschild suggested a nonparametric methodology for testing the predictions of production theory. They saw that their methodology could, in principle, be used to provide nonparametric estimates of the production function. However, they did not envisage that this was practically feasible.
Our main contribution is to show how the Hanoch - Rothschild methodology can be turned into a nonparametric estimator. Our contribution is therefore entirely practical; we have found a practical solution to an old problem.
In the paper we suggest a nonparametric methodology for estimating production functions. We make no parametric assumptions about the distribution of the disturbances, and only the weakest of assumptions about functional form. We assume that the production function is nonnegative, non-decreasing, and concave (diminishing marginal returns). The suggested approach is based on the maximum likelihood method, entropic distance and uses convex programming techniques.