Data Envelopment Analysis is superb in the identification of the Pareto-Koopmans efficient set, but its greatest strength is also its greatest weakness in that it tends to result in facets on the frontier which are less than fully dimensional or to put it another way in zero-prices.

The original Charnes Cooper Rhodes DEA Model (CCR) was quickly amended to include the non-archimedian infinitesimal, “epsilon”, to address this issue, but does so only in a strictly theoretical way since epsilon is infinitesimally small!

Researchers have replaced the original radial measures with non-radial measures of efficiency intended to address the consequences of zero prices in DEA and to take better account of the slacks and surpluses arising in the DEA solutions.

Abstract:

“Modeling the data telecommunications industry as a Spatial Oligopoly Variational Inequalities Model reveals an industry whose structure, in the absence of any meaningful capacity constraints, will result in zero or near zero pricing on each link.

This structure is economically unsustainable and has led to accountancy fraud and to so-called ‘Drive Through Chapter 11’ arrangements.

These bankruptcies do not remove capacity from the industry and leave the competitor firms at a disadvantage.

The industry is an example of the failure of markets. This in turn raises the question have policy makers applied the right regulatory or deregulatory model?”

The Structure of the Data Telecommunications Industry: An Application of Nagurney’s Spatial Oligopoly Variational Inequalities Model.pdf.