廣義之潛在變數模型之探討及其在選舉研究之應用(第2年)

Abstract

本期末報告之內容為「廣義之潛在變數模型之探討及其在選舉研究中之應用」兩年期計畫中，第二年計畫執行之進度報告。第一年計畫之重點在介紹廣義之潛在變數模型，並且將該模型應用於測量台灣民眾的統獨立場。此研究成果已以〈如何評估潛在變數的影響效果？以2013年TEDS統獨立場測量為例〉為題發表於學術研討會，且在修改後已以〈潛在變數的測量及其影響？2013年TEDS台灣民眾統獨立場的分析〉一文發表於《選舉研究》在2015年出版的第22卷第1期。延續第一年計畫的研究成果，此第二年計畫之重點則在於將廣義之潛在變數模型應用於解釋選舉制度對於選民投票行為的影響，更具體而言，此部分在探討並立式單一選區兩票制如何影響選民在投票時，是以候選人特性為依據，抑或是以政黨標籤為主要考量。此研究已經撰寫成一篇英文學術論文，其主要論點是此兩票制中的計票方式雖分別為「單一選區相對多數決」與「封閉式比例代表制」，但都是傾向於以政黨標籤為焦點的制度，所以政黨標籤應是影響選民在此兩票投票選擇的主要因素。然而，由於此兩票制結合裡種選舉制度，因此會使得選民的投票抉擇受到兩種制度的交互影響，特別是會強化候選人因素對於選民投票行為的影響。也因為如此，該文認為候選人因素會影響選民在「單一選區相對多數決」這一票的投票選擇，也可能影響選民的政黨票。針對上述論點，該文以台灣為例進行經驗分析，而分析結果支持該文論點。其中有些發現更提供未來進一步研究的可能性。此研究成果在國際研討會發表後，已經投稿至國際期刊，目前仍在審查中。

This research report presents the academic achievements accomplished in the second year of the two-year project “A General Framework for Latent Variable Models and Its Applications to Survey Data Analysis.” In the first year of the project, I introduced generalized latent variable models and applied these models to analyzing Taiwan politics in a conference paper, which was written in Chinese, and has been published on an academic journal with minor revision, titled as “The Measurement of Latent\nVariables and Its Effects: An Analysis of Taiwanese Attitudes on the Independence-Unification Issue in 2013.”In the second year of the project, I investigate the effects of mixed-member electoral systems that combine single-member district plurality (SMDP) and closed-list proportional representation (CLPR) on voting behavior. Building on the literature of electoral institutions, this article provides an explanation to how mixed-member systems structure voter behavior and achieve a balance between candidate- and party-centric representation. Using Taiwan as a case of MMS, this article tests hypotheses using survey data and investigates the determinants of voting decisions for the two ballots. By employing a Bayesian\nbivariate probit model, this article shows that, first,\npersonal reputation influences voters’ choices of the\nnominal vote. Second, partisan factors affect voter behavior in both nominal and list ballots. But it is\naffective rather than rational considerations for political parties that play the major role. Finally, there is a\nmoderately positive correlation between the two ballots, which potentially results from affective, partisan\nconsiderations.