2009-08-012024-05-17https://scholars.lib.ntu.edu.tw/handle/123456789/693978摘要：此研究計畫將針對近來在隨機邊界模型文獻中受到重視的長期追蹤資料模型，提出新的估計方法。該方法能夠解決目前文獻遇到的困難，提供更理想的估計結果。長期以來，Schmidt and Sickles (1984)提出的估計方法，一直是隨機邊界文獻中，估計長期追蹤資料模型的最主要方法。然而該模型假設廠商無效率不會隨時間改變，且模型中無法同時將無效率效果及個體異值效果（individual heterogeneity）同時包含在內，因此在使用上受到許多限制。針對這些缺點，Greene（2005）提出了較理想的模型，但其提出的估計方法，有嚴重的次要參數問題incidental parameters problem），將影響估計參數的一致性。Wang and Ho（2008）提出的模型及估計方法，雖可處理次要參數問題，但該模型要求模型的無效率具有特定的分配型態。在此研究計畫中，我們利用hamberlain（1982）提出的方法，解決相關問題。該方法不僅可以處理次要參數問題，而且不需對無效率項的分配做特定假設，因此更具一般性，適用範圍更廣。<br> Abstract: This research proposal is to propose a new estimation method of panel stochastic frontier (SF) models. Until recently, the distribution-free estimator of Schmidt and Sickles (1984) had been the main trust of empirical estimation of SF panel data model. Although pioneering in its own right, the model’s time-invariant assumption on inefficiency and the inseparability of inefficiency and individual heterogeneity have drawn criticisms in the literature. Greene (2005) proposed a “true fixed-effect model” to address the problems, but the estimation method suffers from incidental parameters problem. Wang and Ho (2008) provided a solution to avoid the incidental parameters problem, but the model depends on a particular distribution assumption on the model. In this research, we propose a different estimator for the true fixed-effect model. The estimator uses the method first proposed by Chamberlain (1982). The estimator circumvents the incidental parameters problem and does not rely on distribution assumptions on the model’s inefficiency term. Therefore, it is applicable to all stochastic frontier models that have different distribution assumptions on the inefficiency terms. We demonstrate its use using a panel data set of Taiwan’s hotel industry.隨機邊界模型長期追蹤調查資料固定效果stochastic frontier modelpanel datafixed effects長期追蹤資料隨機邊界模型：另一種估計法