Monitoring the performance of soviet cotton-refining enterprises: Sensitivity of findings to estimation techniques

Summary and Conclusions If planners and policymakers are to raise the overall level of performance in an industry by bringing the performance of the least productive enterprises up to that of the most productive enterprises, then obviously both subgroups of enterprises must first be identified. Scholars at TsEMI in Moscow have been using a single technique to measure and rank enterprise productive efficiency, that technique being the estimation of a stochastic CES production frontier. Their technique is not very discriminating about the most efficient enterprises and generates a non-unique ranking of enterprises by productive efficiency scores. Confidence in their analysis could be increased if additional analytical approaches generated similar rankings with, perhaps, greater discriminatory power among the most efficient enterprises.The purpose of this paper has been to compare the Afanas"ev-Skokov results, obtained by estimating a stochastic CES production frontier, with analogous results obtained by estimating a semi-stochastic ray-homothetic production frontier and by using linear programming techniques to calculate a deterministic nonparametric production possibilities set. The rankings and the identities of the most efficient and least efficient enterprises are very similar across the three techniques. This inspires confidence in subsequent efforts to bring the performance of poor performers up toward the level of performance of the good performers.Earlier versions of this paper were presented at the 1988 Atlantic Economic Society Meetings in Philadelphia and the 1988 Allied Social Science Association Meetings in New York. The authors are grateful to M. Afanas"ev for sharing the data, to Johnathan Leightner for his research assistance, to Joe Nowakowski, Bill Pfouts, and Cliff Huang for their helpful comments, and to IREX and the National Science Foundation for their support.