Nonparametric and Semi-Parametric Estimation of Efficient Frontier

This paper has presented a nonparametric approach to estimate the deterministic frontier model with unknown function form. Single index modeling and partially linear modeling techniques have been modified for estimation of a semi-parametric deterministic frontier model. Those approaches are free from functional form specification and error term distribution assumption. The methods are very attractive for computational simplicity and straightforward interpretation. They can be extended to estimate the stochastic frontier model with more parametric assumption. At the center is the kernel estimation of the continuous density and conditional expectation. These kernel estimations are used as the instruments in estimating expectation of efficiency level, production frontier, and firm's specific inefficiency