Data Envelope Analysis Accounting for a Stochastic Component in the Data
Eric Severance-Lossin · 1995
Data Envelope Analysis, DEA, has become a standard tool in applied economics for estimating production frontiers and for assessing firms' technical efficiency. While the analysis is based on solid economic principles it does not take into account any stochastic component in the data. An alternative nonparametric method for estimating a production frontier and assessing technical effciency which accounts for a stochastic component in the data is proposed here. The method uses nonparametric kernel regression to smooth the data, and then uses the results obtained from the regression to test the hypothesis that all firms are technically efficient. Unfortunately, this test is only powerful against a limited set of alternatives. An additional method is proposed to examine the relationship between the portion of variance accounted for by measurement error and the portion of inefficient firms in the sample.