發刊日期/Published Date |
2000年12月
|
---|---|
中英文篇名/Title | 廠商勞動需求的推估與樣本自我選擇性偏誤—台灣製造業廠商之實證研究 Labor Demand Estimation and Sample Selection Bias: The Empirical Studey of Taiwan Manufacturing |
論文屬性/Type | 研究論文 Research Article |
作者/Author | |
頁碼/Pagination | 563-596 |
摘要/Abstract | 本文係利用台灣工商普查資料,推估台灣製造業廠商的勞動需求函數,以期對台灣製造業廠商的勞動需求的工資及產出彈性大小,有一基本的認識,並在衡量勞動雇用政策之就業效果時,作爲衡量廠商反應程度之依據。文中利用製造業普查的個別廠商資料,估計製造業 20 個二分位產業內廠商的勞動需求函數。由於使用的是廠商個體資料(micro data),文中將探討在使用此種資料估計勞動需求時之樣本自我選擇偏誤問題,以及解決的方式。此一實證研究有別於以產業或製造業加總的資料爲主的方法,其估計結果可幫助我們了解台灣各產業內不同特性生產者的勞動需求決策,亦可用以檢視產業間勞動需求的差異。在衡量勞動政策對勞動市場的衝擊時,亦可提供較正確的就業效果預測。而根據本文的實證結果顯示,考慮樣本自我選擇誤差 (sample selection bias) 而以Heckman 的二階段估計法加以修正後,20個被估計的產業中有 10 個產業有顯著的樣本選擇偏誤。因此,若不考慮因市場選擇而退出的樣本廠商,將高估廠商的產出及工資彈性。在本文之估計過程中,同時亦考慮了廠商異質性 (firm heterogeneity) 與產出變數之衡量誤差 (measurement error) 所可能造成的估計偏誤,文中亦以差分法 (differecing) 及工具變數 (instrumental variable) 加以處理,以獲得不偏及具一致性之估計結果。綜合而言,台灣由於製造業廠商以中小企業居多,其勞動需求之產出彈性多小於 1,顯示勞動使用仍處於規模報酬遞增之生產階段。而工資彈性雖較產出彈性小但亦頗爲顯著,尤其以勞力密集產業為最。 The development of the Taiwanese manufacturing sector has induced a series of fundamental changes in the labor market, such as employment conditions and wage structures. Demand-side policies, including the health insurance system and the pension or social security system, have a direct impact on the production costs of individual producers. To assess the effects of these policies on factory as well as industry's demand for labor, it is necessary to examine the hiring decisions of the heterogeneous producers more closely. The purpose of this study is to build a labor demand model to estimate the long-run labor demand using Taiwanese Manufacturing survey data of individual plants. The focus is on estimating the plant-level wage and output elasticities. In doing so, the sample selection bias, endogeneity, and measurement error problems within the estimation are discussed. Different estimation models and estimation methods are chosen to solve the problems. In particular, a differencing model is used to correct for plant specific characteristics, an instrumental variable estimator is used to correct for the measurement error in the output variable, and the Heckman's two stage estimation method is used to correct for a self selection bias that results from plant failures. |
關鍵字/Keyword | |
學科分類/Subject | |
主題分類/Theme | |
DOI | |
檔案下載/Download |