A Bi-Extremal Principle for Estimating Efficiency Frontier Parameter Values.

摘要

A new approach is supplied for evaluating the efficiency of decision making units, locating efficiency frontiers and estimating parameters from observational data. This is accomplished by means of a nonlinear-nonconvex bi-extremal principle which is subsequently shown to be essentially reducible to a finite sequence of linear programming problems. The development is illustrated by means of multiple output functions which are piecewise of Cobb-Douglas or general log linear type and which also allow for increasing, decreasing and constant returns to scale. The reduction of the bi-extremal principle to linear programming equivalence is also accomplished for much more general classes of functions. (Author)