解讀政黨票源與分裂投票：布蘭德──奧特曼差異圖和政治相對發展指標的新應用

Understanding Party Vote Share and Split Voting: An Application of Bland-Altman Difference Plot and Political Relative Development Index

我國近年來持續推動選舉制度改革，如2008 年將立委選舉改為「單一選區兩票制」，2012 年首次將總統與立委選舉合併辦理，選民可同時投下總統、不分區立委與區域立委三張選票，此一制度變革提供了研究全國性選舉的政黨票源與分裂投票的機會。過去有關分裂投票的實證研究，或是透過民意調查的個體資料、或是利用選舉得票的總體資料、也有的是同時結合個體和總體的資料進行分析。不過，這些方法皆未考量各地區的選舉人口規模、政黨或候選人在特定地區的得票數相對全國得票數的比例，故難以綜合判斷分裂投票行為與地域性的票倉大小、政黨票源之間的關係。本研究嘗試提出新觀點，檢視全國性選舉的政黨票源與分裂投票現象，並以2012 年的選舉結果為例，說明與驗證新測度方法的效果。在分析流程上，首先，以368 個鄉鎮市區為分析單位，計算各政黨和藍綠陣營的得票率，並以傳統的分裂投票測量方式計算得票率差距；第二，以布蘭德──奧特曼差異圖呈現各政黨在不同得票實力地區的分裂投票樣態；第三，討論布蘭德──奧特曼差異圖的特性及限制，進而建構一套政治相對發展指標－－BDI 和CDI，提出可同時解讀政黨票源與測度分裂投票的新框架，衡量選民支持特定政黨或藍綠陣營的相對強度；最後，將此測度結果和傳統方法所得結果進行比較，並說明未來研究方向。

Taiwan has been under electoral reforms in recent years: Legislator election in 2008 started to adopt “single-district two votes system”. Combined legislative and presidential elections in 2012 allowed voters to vote for president, at-large legislators and local legislators concurrently for the first time, offering a great opportunity to study party vote share and split voting in general elections. There were practical researches on split voting, some of which analyse micro information from polls, and some are based on the macro data of votes in elections. However, these studies took into consideration neither the size of eligible voters in districts nor the ratio of national votes to regional votes a party or candidate get, which thus makes it hard to determine the correlation between split voting, the size of electoral districts and a party’s real political power strength. This study attempts to examine party vote share and split voting from a novel perspective on proposing a new measurement and exemplifying with the general election in Taiwan in 2012. The research started with calculating the vote share of the Pan-Blue Coalition, the Pan-Green Coalition and each party in 368 administrative districts and gauging the voting gap with a traditional inspection method. Secondly, we tried to use a Bland-Altman difference plot to show the pattern of split voting by districts since every party’s ability to gain votes varies with areas. Thirdly, we discussed the strengths and inadequacies of the application to Bland-Altman difference. And then, the political relative development index, BDI and CDI, are introduced as the new framework for ascertaining party vote share and measuring split voting to decide the relative level of voters’ support for some party, the Pan-Blue Coalition or the Pan-Green Coalition. Finally, the result from the new measurement is compared with that from the traditional method to validate the performance of the new approach, as well as to indicate future research direction.