Constrained and averaged methods in stochastic frontier analysis

The field of stochastic frontier analysis has seen many recent developments embracing the power and flexibility of semi and nonparametric methods. Yet, with application of these methods also comes likely violations of economic theory, such as monotonicity and concavity. Methods to impose shape constraints in classical regression settings abound, but have yet to appear with as much force in frontier analysis. At the same time, model averaging has steadily gained traction as a viable means to estimate and combine models in the presence of model uncertainty. This is clearly the case in frontier analysis where it is commonly debated the necessary distributional assumptions and the variables which influence inefficiency. This work reviews both of these methods and their viability within the frontier analysis community.