本文主要目的在探討候選人評價此一因素在2004年總統選舉中可能發揮的純粹選舉效應。本研究採預測增進法的分析策略,並運用「多階段區塊因果順序投票模型」,來分析候選人評價因素對於選民投票行為與選舉結果的影響。在使用TEDS2004P的選後調查獨立樣本並暫不考慮未表態樣本的研究脈絡下,本研究發現,整體來說,陳水扁確實較連戰擁有較佳的「候選人評價」優勢,但此一優勢所發揮的選票利益並不如一般想像中的大。分析結果顯示,候選人評價的純粹選舉效應雖是可見但卻是相當有限,只幫助陳水扁多增加約五千張的選票(0.04%)。本研究分析結果只是一個暫時的結論,分析模型的假定、選後調查資料的使用、以及未表態受訪者的排除等因素,都影響了本研究結論的確定程度。
This paper aims to investigate the unique effects of party candidate evaluations in the 2004 presidential election in Taiwan, both on individual vote choice and on overall election outcome. With the application of Multi-Stage Block Causal Recursive Model of voting developed by Miller & Shanks in The New American Voter, the research indicated two major findings as follows. First of all, in general, Chen, Shui-bian (DPP party candidate) enjoyed more advantages on the comparative evaluations of candidates than did his competitor Lien Chan (KMT party candidate), and had a visible but small effect on vote choice. Second, the favorable evaluations of Chen, however, were less contribution to the aggregate election outcome due to the estimate that only about additional 5,000 votes or 0.04% will be incurred as a result of the candidate effects. The electoral effects of party candidate evaluations, therefore, may not be as strong as usually supposed. The research results are not decisive but confined in the voting model assumptions, in the use of post-election data, and the omission of the undecided voters' potential voting preferences. All the factors will vibrate the degree of certainty of our research conclusions.