| Summary: | 博士 === 國立中央大學 === 經濟學系 === 101 === This dissertation consists of two independent yet related essays which study the topic of the stochastic frontier model. The stochastic frontier model is widely used in assessing the efficiency or performance in banking, firms, the government and medical systems. Recently, cross-sectional data has been supplanted by the use of panel data that has the advantage of a large sample size. In addition, the accompanying econometric issues including fixed effects and the specification of efficiency have been introduced and discussed at length. Despite these issues, many studies have started to use the factor structure to describe the heterogeneous impacts caused by common shocks, which model involves this factor structure also accommodating the property of cross-sectional dependence. However, there is still less discussion in the literature on stochastic frontier analysis. Compared with the existing literature, we treat the factor structure and efficiency as separate components instead of treating all factor structure as efficiency. We further discuss the reasons why the factor structure and efficiency should be separated. A transformation has been proposed in this dissertation to obtain a consistent estimate via the maximum likelihood based approach.
The first chapter of this dissertation introduces the main issues in the stochastic frontier model in regard to the panel data. The two essays are organized as chapter 2 and chapter 3. I provide brief summaries of these two chapters as follows.
Chapter 2 develops a panel stochastic frontier model with unobserved common shocks to capture cross-sectional dependence among individual firms. The novel feature of our model is to separate technical inefficiency from the effects induced by unobserved common shocks and individual heterogeneity. We propose a modified maximum likelihood method that does not require estimating unobserved common correlated effects and discuss the asymptotic properties of the proposed estimation procedure. The basic idea of our approach is similar to that in Pesaran (2006) for the linear panel regression. We show that the proposed method can control the common correlated effects and obtain consistent estimates of parameters for the panel stochastic frontier model. Our Monte Carlo simulations show that the modified MLE has satisfactory finite sample properties under a significant degree of cross-section dependence for relatively small T. The proposed method is also illustrated in applications based on a comparison of the efficiency of savings and the commercial banking industry in the US.
In chapter 3, we consider a linear model with time-invariant fixed effects to represent heterogeneity and the cross-sectional dependence by introducing common correlated effects, and the time-variant technical inefficiency and idiosyncratic errors jointly characterized by a multivariate skew normal distribution. To consistently estimate the slope coefficients and variances in the above model, we propose a transformation which is similar to that introduced in chapter 2 to eliminate fixed effects and common correlated effects. Based on the transformed likelihood function, we then introduce an EM Algorithm to robustly estimate these parameters. Our Monte Carlo simulation shows that the proposed method is quite accurate in the presence of common correlated effects, while conventional models that do not take these effects into account can result in severely biased parameter estimates.
|
|---|
