本文根據投票計算理論(the calculus of voting),以2004年第一次施行的公投選舉為背景,探究選民的選票機率、效用落差、投票成本、及公民責任感等投票計算變項,究竟對他們的公投行為產生多大的影響。根據多項式常機模型(Multinomial Probit, MNP)的估算結果,本文最重要的是發現台灣選民的公民責任意識與民主價值理念並未對參與公投產生作用。也就是說公民責任感越強者,並不必然較積極參與三二○公投表達意見。尤甚者,與公民意識高度連結的教育程度與參與公投呈現負相關,也就是教育程度越高的選民,越不會去投票所正式或非正式的表達他們的看法,他們寧願選擇待在家中不去投票。總括來說,本文發現公民投票所應代表直接民主與公民教育的意義,在首次舉辦的公投選舉中並未彰顯出來。 除了公民責任感與教育程度的影響之外,本文也發現選票機率、效用落差及投票成本大致都符合投票計算理論的基本假設。簡單來說,選民認為個人選票作用越大者,參與選舉的可能性越高;選民投票成本越高的話,特地跑去投票所的可能性越低;投票結果的效用落差越大,越有動機去投票表達意見。此外,本文也發現選民的政黨認同會影響到公投選舉;認同泛綠的選民明顯傾向去投票,但若是讓他們選擇投廢票或不領票,則不如不出門投票;認同泛藍的群眾則剛好相反,他們比較不會去投票,但若一定要去投票所,會選擇拒領或投廢票,顯見泛藍某些團體倡議的拒領公投選票運動有達一定程度的作用。
This paper, based on Riker and Ordeshook's calculus of voting theory, explores the first referendum, which took place in 2004. Variables such as probability, utility differential, cost, civic duty are considered to detect their influence on voting behavior. Most surprisingly, the result of Multinomial Probit analysis shows that voters' civic duty or democratic values did not improve the incentives of participation. Moreover, it is found education level was negatively correlated with the turnout, meaning intellectuals were more unlikely to participate in the referendum. Except for the influence of civic duty and education, the other voting calculus variables (such as probability, utility differential, and cost) basically matched the research hypothesis. In addition, this paper also found that party identification had a very important impact on voters' decisions in the referendum. Pan-green voters were very likely to participate in the referendum, and they preferred to stay home and abstain from both presidential and referendum elections rather than to cast invalid ballots. Pan-blue voters, on the contrary, were unlikely to participate, and those who did were likely to intentionally cast invalid ballots.