具衡量誤差的追蹤資料隨機邊界模型之估計方法

機構典藏 > College of Business and Management > Graduate Institute & Department of Economics > Research Paper > Item 987654321/101162

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/101162

Title:	具衡量誤差的追蹤資料隨機邊界模型之估計方法
Other Titles:	Estimation of Panel Stochastic Frontier Models with Errors in Variables
Authors:	陳怡宜
Contributors:	淡江大學經濟學系
Keywords:	衡量誤差;隨機邊界模型;長期追蹤資料;一般化動差估計法;measurement errors;stochastic frontier model;panel data;GMM
Date:	2013-08
Issue Date:	2015-04-15 10:27:53 (UTC+8)
Abstract:	此研究的目的, 在於針對具有變數衡量誤差的長期追蹤資料隨機邊界模型, 提出具有一致性的一般化動差估計式。對於許多實證研究而言, 迴歸變數常具有明顯的衡量誤差。變數產生誤差的原因很多, 包括數據加總、自我填寫、或以代理變數作為替代等等, 皆是可能的原因。在古典衡量誤差的假設下, 衡量誤差將使其變數的係數將被低估, 而 Griliches and Hausman (1986) 說明了此低估的情形在長期追蹤資料模型中更為嚴重, 因為該類模型估計過程中採用的如 within- transformation 的方法, 將使衡量誤差的影響被放大。文獻中雖已有針對具衡量誤差的隨機邊界模型提出研究, 但其研究對象皆是橫斷面的模型。鑑於現今文獻中, 長期追蹤資料隨機邊界模型的使用愈來愈被重視, 提出一個能夠修正衡量誤差的估計方法, 有相當的重要性。我們提出的估計方法, 主要是根據 Wansbeek (2001), 利用模型殘差與衡量誤差的結構, 得到適當的動差方程式, 作為估計模型係數的基礎。接著, 我們利用 Chen and Wang (2012) 提出的動差, 估計模型殘差項的參數。 In this research we propose a GMM estimator for a panel stochastic frontier model where one of the explanatory variables is measured with errors. The problem of mismeasured variables are prevalent in empirical studies. The mismeasurement may arise from data aggregation, the use of self-reporting data, or the use of proxy variables when the actual data are not available. Under the assumption of classical measurement errors, the errors in variables would bias the coefficient estimate downward. The situation is particularly serious for fixed effect panel data models because the bias is magnified in the estimation process (Griliches and Hausman 1986). Although a few studies have devoted to the problem of errors in variables for stochas- tic frontier models, they are all tailored to cross-sectional models. Therefore, this will be the first study to provide a solution to panel SF models with errors in variables. Our solution strategy is based on Wansbeek (2001) and Chen and Wang (2012). We follow Wansbeek’s suggestion and obtain moment conditions to estimate the slope coefficients of the model under the condition that one of the variables is measured with errors. We then use the results from Chen and Wang to obtain the estimates of variance parameters. Monte Carlo experiments will be provided to show the performance of the model
Appears in Collections:	[Graduate Institute & Department of Economics] Research Paper

Files in This Item:

There are no files associated with this item.

Loading...