A0544
Title: A semiparametric stochastic frontier model with two-way fixed effects and nonparametric inefficiency function
Authors: Taining Wang - Capital University of Economics and Business (China) [presenting]
Kai Sun - Shanghai University (China)
Abstract: A semiparametric stochastic frontier panel model is proposed, relaxing conventional parametric assumptions on both inefficiency and frontier. First, distributional assumptions are not imposed on the inefficiency for identification, but only the existence of its mean conditioning on observables is assumed. Unlike existing models, the level of inefficiency mean function is identified by estimating it with the frontier in a sequential step, achieved by applying conventional two-way within the transformation. Second, to combat the curse of dimensionality, a single-index structure in the inefficiency mean and the inputs elasticity is introduced, which can vary with contextual environmental variables in a nonlinear fashion. Third, the inefficiency is disentangled from latent heterogeneities in firm and time dimensions. A three-step estimation procedure that combines the use of series and kernel estimator is employed, and their appealing finite-sample performance through simulation studies is demonstrated. Our model's applicability is showcased by performing an efficiency analysis in the banking industry.
