Estimating Technical Efficiency: Does the Choice of Estimation Methodology Matter?

In this paper we compare estimates of technical efficiency for a cross-section of Australian dairy farms using various frontier methodologies; Bayesian and Classical stochastic frontiers, and Data Envelopment Analysis. We find that our Bayesian stochastic frontier estimates, when we impose monotonicity and curvature restrictions at all data points, significantly impact on our parameter estimates. Our results indicate no technical inefficiency is present in the sample data. However, all other specifications indicate the presence of technical inefficiency in our sample data. For these models we identify statistical differences between the point estimates of technical efficiency. But, the rank of farm level technical efficiency is statistically invariant to the estimation technique employed. Unlike any existing comparative studies in the literature we also compare interval estimates of technical efficiency for methods used. We find significant overlap for many of the farms’ interval estimates irrespective of frontier method employed. Therefore, our results allow us to conclude that the choice of estimation methodology may matter, depending on the purpose of the analysis.