SEMINAR: How efficient are rice farms in Ghana? A comparison of alternative Bayesian stochastic metafrontier approaches

The subject of technological heterogeneity has been a major issue in the application of production function approach to efficiency estimation. Battese et al. (2004) introduced a metafrontier approach that addresses technological heterogeneity in efficiency estimation using linear programming (LP) methods. In this paper, we propose two alternative methods for constructing the metafrontier and establishing its statistical properties. Our first proposal relates to constructing the metafrontier using upper confidence bounds for the group frontiers and estimating the metafrontier using Bayesian stochastic metafrontier methods. The second alternative integrates the group and higher-level frontier estimations into a single Bayesian estimation procedure by constructing LP based metafrontier for each Markov Chain Monte Carlo (MCMC) group frontier draws. Using rice production data from two agro-ecological zones of Ghana, we find that farmers in the Guinea savannah zone are more productive and efficient compared with those in the Sudan savannah zone. However, rice production in both agro-ecological zones is characterised by substantial levels of inefficiency. Methodologically, we find that alternative metafrontier models generate slightly variable technology gaps and that is a function of the method of constructing the metafrontier and its statistical properties.