HETEROSCEDASTICITY CORRECTION MEASURES IN STOCHASTIC FRONTIER ANALYSIS

Abstract

he stochastic frontier analysis (SFA) model, designed to assess technical efficiency in production models, operates under the assumption of homoscedasticity. However, in practical scenarios, either the random error or the technical efficiency error, or both, can exhibit non-homoscedasticity. This research proposes heteroscedasticity correction measures for the random error (HCRE), technical efficiency error (HCTE), and both (HCRTE) within the SFA model. The study aims to determine which correction measure yields the most efficient parameter estimates when heteroskedasticity is present. The comparison involves evaluating the mean squared error (MSE) across different forms of heteroscedasticity and sample sizes through Monte Carlo simulations comprising 5000 replications. The findings indicate that attempting to correct for heteroskedasticity in the absence of such issues can adversely affect the parameter estimates of the SFA Model. Conversely, the HCRTE measure consistently produces the most efficient estimates when dealing with heteroskedasticity in terms of both random error and technical efficiency. Moreover, in cases where heteroskedasticity exists, applying the HCRTE measure not only enhances parameter estimates but also improves the technical efficiency measure of the SFA model.

Keywords

• stochastic frontier production
• heteroscedasticity correction
• technical efficiency
• monte carlo simulation