Frontier estimation in nonparametric location-scale models

Conditional efficiency captures efficiency of firms facing heterogeneous environmental conditions. Traditional approaches estimate nonparametrically conditional distribution requiring smoothing techniques. We rather use a flexible nonparametric location-scale model to eliminate the dependence of inputs/outputs on these factors. These “pre-whitened” inputs/outputs define the optimal frontier function and a “pure” measure of efficiency more reliable to produce rankings, since the influence of external factors has been eliminated. Both full and order-m frontiers are used. The asymptotic properties are established. We can also derive the frontiers in the original units with their asymptotic properties. The approach is illustrated with some simulated and real data.