Nonparametric Frontier Estimation

Resumo: In this paper we analyze a nonparametric model for estimating production frontiers with multi-dimensional inputs, with some advantages over traditional frontier estimators as Data Envelopment Analysis (DEA) and Free Disposal Hull (FDH). In comparison to Martins-Filho and Yao (2007) approach, our method is simpler, more general and requires less restrictive assumptions. We also show consistency and asymptotic distributional theory of our proposed estimator under standard assumptions in the multi-dimensional input setting. The finite-sample performance of our proposed estimator is illustrated in a simulation study where we compare it with Martins-Filho and Yao´s, DEA and FDH.