在臺灣高等教育普及化的背景下，學歷勞動市場帶來的辨識效果日益下降，教育體系內部的分化現象則逐漸凸顯，學科選擇成為影響社會流動的重要機制。尤其在臺灣，高等教育擴張與勞動市場需求並行，使不同高教就學類型的報酬與流動機會產生顯著差異。本研究關注STEM、非STEM與未上大學者三種教育選擇，探討三者在不同高教擴張世代中的社會流動效果。

本文使用臺灣社會變遷基本調查2007、2012、2017與2022年四筆資料，建構高教擴張世代與高教就學類型選擇的量化模型，將原生家庭社經地位與樣本社經地位差異作為流動指標，並採用傾向分數加權法處理高教選擇的內生性問題，另於檢視不同性別樣本中的效果異質性。

研究結果顯示，大學學歷對社會流動具有顯著正向效果，但其幅度隨高教擴張而遞減。整體樣本中，STEM主修者的流動程度迴歸係數顯著低於非STEM主修者，但效果量低，顯示兩者的實際差距有限。若分世代觀察，STEM與非STEM的差異並僅在擴張前世代顯現，且差異因效果量偏小而意義有限。性別分析則揭示異質性：女性在所有世代中，STEM主修流動程度皆低於非STEM；男性則僅在擴張發展期出現STEM優勢，但幅度有限，且效果在擴張後世代消失。

論文外文摘要： Against the backdrop of the massification of higher education in Taiwan, the signaling effect of educational credentials in the labor market has diminished, while internal differentiation within the education system has become more prominent. Consequently, field of study has emerged as a crucial mechanism shaping social mobility. In Taiwan, the simultaneous expansion of higher education and shifts in labor market demand have led to significant disparities in returns and mobility opportunities among different educational pathways. This study focuses on three educational tracks—STEM, non-STEM, and no university education—to examine their heterogeneous effects on social mobility across different higher education expansion cohorts.

Using four waves of the Taiwan Social Change Survey (2007, 2012, 2017, and 2022), this research constructs quantitative models linking cohorts of higher education expansion and educational choices. Social mobility is measured as the difference in socioeconomic status between respondents and their family of origin. To address endogeneity in educational tracking, this study adopts a propensity score weighting method. Furthermore, it investigates effect heterogeneity by gender.

The results indicate that a university degree has a significant positive effect on social mobility, but its magnitude diminishes across higher education expansion cohorts. In the overall sample, the social mobility of STEM majors is found to be significantly lower than that of non-STEM majors, though the small effect size suggests a limited practical difference between the two. When analyzed by cohort, the difference between STEM and non-STEM majors only appears in the pre-expansion cohort, and this difference is of limited practical significance due to the small effect size. Gender-stratified analyses reveal significant heterogeneity: for women, social mobility among STEM majors is consistently lower than that of non-STEM majors across all cohorts. For men, a STEM advantage appears only in the "expansion development" cohort, with a limited magnitude, and is not observed in the post-expansion cohort.