投票行為之空間觀點與空間分析－以臺灣2004年總統選舉為例

Abstract

過去對於臺灣選舉的投票行為研究，較少由空間觀點切入。然而投票行為的地域差異，早已被人注意，且勢將支配單一選區制度下的選區劃分與選舉結果。本文的目的，在於發掘並揭示選舉研究中的空間問題。首先，本文將空間效應界定為空間異質性與空間相依性，並論述其影響投票行為的過程。其次，本文介紹空間自相關與空間迴歸模式的分析方法，用以探索與檢驗空間效應。本文的後半部分，藉由臺灣2004年總統選舉的實證研究成果，支持三個假設：1.投票行為具有空間聚集的現象，亦即在距離愈近的村里中，投票行為愈為相似；2.傳統迴歸模式的誤差具有地域差異，顯現空間異質性的存在；3.控制了年齡、教育、行業、所得、族群等社經因素後，鄰近村里的投票行為仍然相關，證明空間相依性的支配力。本文的結論是：投票者在哪袒，是徊必須被考慮的問題。

Spatial perspectives were seldom applied on the study of voting behavior in Taiwan in the past. However, the regional differentiation of voting behavior had been generally attended. Moreover, it would certainly dominate the division of Single-Member Districts and the eventual outcome. The authors attempt to explore spatial issues in electoral studies in this article. First, by defining spatial effects as spatial heterogeneity and spatial dependency, it could be then revealed how it influenced voting behavior. Secondly, indicators of spatial autocorrelation and the spatial regression model are introduced to explore and examine spatial effects. In the later half of the article, the 2004 Taiwan presidential election is taken as an empirical example to support three hypotheses: 1) similar voting are spatially clustered, in other words, voting are similar among neighboring communities; 2) residuals in the classical regression model exhibited regional differentiation that reveals spatial heterogeneity; 3) contiguous voters can still be affected even the social factors such as age, education, industry, income and ethnicity have been controlled (a revelation of the dominance of spatial dependence). The authors thus conclude that it shall always be taken into consideration regarding where the voters live.