現今女性受教育的普及程度與勞動參與程度愈來愈高,因此女性在勞動市場中扮演著不可或缺的角色。勞動經濟學中,教育報酬率之估計一直都是重要議題之一。估計女性教育報酬率存在選擇偏誤、內生問題與異質特性,若使用傳統OLS估計法則會產生估計偏誤。故本文之重點為探討如何準確估計出女性教育報酬率,並以台灣的資料作為試驗。本文將使用Heckman二階段估計法、二階段工具變數法與傾向分數配對法,試圖修正自我選擇與內生問題。另外,針對教育報酬率的異質特性,本文採用分量迴歸模型,期望能在不同薪資水準下,反映真實的教育報酬率。本研究的實證結果發現,不同的估計方式所得到的高等教育報酬率約介於3%-10%,其中OLS估計法之結果最低,其次依序為Heckman二階段估計法、二階段工具變數法、傾向分數配對法,證明OLS估計有嚴重低估問題。而在分量迴歸模型估計結果則發現,忽略選擇偏誤與內生問題,將使女性教育報酬率之薪資分配右尾嚴重低估,教育報酬率與薪資分配呈現顯著正向關係。由本文研究可得知,教育與個人能力會互為強化,因此教育會擴大女性間薪資的差距。
The role of female in the labor market becomes indispensable as a result of the increasing popularization of female education, as well as the expanding labor force participation of women. Furthermore, estimating the return to education is always an important issue in labor economics. Three potential problems exist in estimating the return to education of females: selection bias, endogeneity and heterogeneous characteristics. Because of these problems, the traditional OLS estimation will lead to biased results. The goal of this thesis is to discuss how to obtain the unbiased estimate of the return to education of females using Taiwan’s data. This study applies the Heckman’s two-step estimation approach, the two-stage least squares estimation, and the propensity score matching approach to deal with the selection bias and the endogeneity problems. In addition, this study adopts the quantile regression model to take into account the heterogeneous characteristics issue in order to obtain better estimates of the return to education at different wage levels. The estimation results show that the return to higher education from different econometrics methods is between 3%-10% approximately. The estimated return from the OLS estimation method is the smallest, and followed by the Heckman’s two-step estimation approach, the two-stage least squares estimation, and the propensity score matching approach. This result suggests that the OLS estimation has largely underestimated the return to education. The quantile regression model’s results indicate that ignoring the selection bias and the endogeneity problems will underestimate the return to education of females on the right tail of the wages distribution. The return to education and the wages distribution show a significantly positive relationship. This study shows that education and personal skills are mutually reinforcing, hence females wage gap will be enlarged by education.