A Poisson Stochastic Frontier Model with Finite Mixture Structure

Standard stochastic frontier models estimate log-linear specifications of production technology, represented mostly by production, cost, profit, revenue, and distance frontiers. We develop a methodology for stochastic frontier models of count data allowing for technological and inefficiency induced heterogeneity in the data and endogenous regressors. We derive the corresponding log-likelihood function and conditional mean of inefficiency to estimate technology regime-specific inefficiency. We further provide empirical evidence that demonstrates the applicability of the proposed model.