关键词: 半参数空间ZISF, 零技术无效率项, 极大似然估计, 反馈分类, 蒙特卡罗模拟

Abstract: Incorporate spatial effects, nonparametric function and technical inefficiencies with non-continuous distribution, and get semi-parametric spatial zero inefficiency stochastic frontier models (short for semi-parametric spatial ZISF). These models are more applicable and can effectively avoid specification error and biased and inconsistent estimators. Use B-splines to approximate nonparametric function, then use maximum likelihood method and JLMS method to estimate parameters, nonparametric function and technical efficiencies. Propose a classification method with feedback to judge the categories of technical inefficiencies based on Bernoulli's law of large numbers. Monte Carlo simulations show that: (i) the estimation accuracies of objective method is high. With sample size increasing, the accuracies become higher. Neglecting any effect will get low accuracies. The objective models and estimation method are necessary. (ii) The classification thresholds are with large span. Judging the classification by Bayesian posterior probabilities is subjective and unreliable. The classification with feedback is with high accuracies.

Key words: Semi-parametric Spatial ZISF, Zero Technical Inefficiencies, Maximum Likelihood Estimation, Classification with Feedback, Monte Carlo Simulations