Estimating technical efficiency under unobserved heterogeneity with stochastic frontier models: application to northern German dairy farms

This article uses panel data estimation techniques to examine technical efficiency of individual dairy farms in northern Germany. To the extent that agricultural production is characterised by heterogeneous production conditions, estimation techniques that do not account for unobserved heterogeneity produce biased efficiency estimates. We therefore estimate a number of conventional panel data models and Greene's recently proposed 'true' random-effects model, as well as an extension of the model to ascertain the effects of different specification on the production function and efficiency estimates. Our results appear to support theoretical expectations and previous findings according to which a specification that is both time-variant and also controls correlations between unobserved heterogeneity and the explanatory variables avoids heterogeneity bias and thus ensures consistent efficiency estimates.