並立式混合選制下兩票之連動效果：日本眾議員選舉政黨重複提名策略與成效

The Contamination Effect in a Mixed-Member Majoritarian System: the Impact of Party's Dual Candidacy Strategies in Japanese House Elections

郭銘峰(Ming-Feng Kuo)

指導教授 ： 黃紀

共同指導教授 ： 王鼎銘

國立臺灣大學/社會科學院/政治學研究所/博士(2011年)

https://doi.org/10.6342/NTU.2011.03320

摘要

混合式選制融合了多數決與比例代表兩種選制的精神，不過關於此種選制兩票架構對投票的影響，有兩種截然不同的主張：一是認為兩票各自獨立運作，其影響可以獨立估算；另一派則主張兩票抉擇會相互影響，亦即具有所謂連動效果或感染效果的存在，因此必須綜合考量兩票架構的互動關係。而日本自1996年眾議員選舉開始施行的「小選舉區比例代表並立制」，又因允許候選人在兩種選票間重複提名，使得其兩票間的連動關係益形複雜。
本文主軸鎖定於日本改採新制後的候選人重複提名規定，對於日本政壇主要政黨（包括自民黨、民主黨、社民黨）提名參選策略與其實質成效的影響。有別既有文獻分析兩票連動效果經常未顧及變項間內因性與非隨機選組機制因素，導致參數值估算偏誤的因果推論問題，本文首先爰引James J. Heckman建構之效應模型進行校估。實證結果顯示：日本各主要政黨在歷屆眾議員選舉對單一選區候選人採重複提名參選的策略，其對政黨比例代表得票率的效應會因黨、因年而異；當中包括自民黨在2009年選舉、民主黨在2003年選舉，社民黨在2000年、2005年、2009年選舉，均以重複提名策略有效提升該黨的政黨比例代表得票率。
為更準確驗證選區層次資料分析結果的效度，本文進一步根據日本選舉研究(JES)的民調資料，以日本自民黨為背景，採William H. Greene建構之雙變數機率單元模型深入解析新制下歷屆眾議員選舉中個體選民的兩票效應。研究結果不僅發現自民黨採重複提名策略確實對於選民兩票架構的決策互動造成影響，原本重複提名對於自民黨在PR得票上的負向作用，更有逐年轉為正向提升自民黨PR選票的趨勢，此一結果與選區資料的分析發現相當一致；此外，傳統日本選舉文獻所強調的政黨認同、保革意識型態與都市化程度等因素，仍在新制下的幾屆眾議員選舉中對選民投票行為產生影響。總括本文的分析，不僅佐證既有日本選舉投票與日本政黨重複提名參選成效等相關文獻的預期，也吻合混合選制學理兩票連動效果的論點。

關鍵字

日本眾議員選舉 ； 小選舉區比例代表並立制 ； 重複提名制度 ； 獨立效果 ； 感染效果 ； 效應模型 ； 雙變數機率單元模型

並列摘要

The Mixed electoral system combines the spirits of Plurality and Proportional Representation. The way two-ballot structure influencing the electoral behavior however is not settled. Of the two distinctive schools, one believed two ballots work separately and therefore, their influences should be estimated independently. The other group of people advocated the existing of interaction effects (or the so-called “contamination effects”), and believed it is necessary to consider the interaction between two ballots. As for the Mixed-Member Majoritarian (MMM) Electoral System which was realized in Japanese House Election since 1996, it is actually more complicated because of the addition of dual candidacy.
The purpose of this study is trying to evaluate the impact of this new electoral system in Japan. Specifically, this study is especially interested in the interaction effects of two-vote structure that caused by parties’ dual candidacy strategies. To avoid the causal biased estimation and capture the contamination effect more precisely based on district-level data, this study applies Heckman’s treatment-effects model to remove the endogeneity and self-selection problem. The result not only shows the dual candidacy strategy influences the interaction between two ballots, it also indicates the impact is dissimilar from different parties and different periods of time. More specifically, the dual candidacy strategy improves the PR votes obviously for the LDP (Liberal Democratic Party) in 2009, the DPJ (Democratic Party of Japan) in 2003, and the SDP (Social Democratic Party) in 2000, 2005, and 2009.
In order to make sure the validity of the findings of district-level data and get further information, this study takes advantage of the survey collected by Japanese Election Studies (JES) since 1996. Specifically, this study uses the Liberal Democratic Party (LDP) as an example, and applies Greene’s bivariate probit model to analyze the interaction between the voters’ two ballots. The result shows that the dual candidacy in the new electoral system did influence the interaction between the two ballots; the LDP’s dual candidacy strategies indeed helped to attract more PR votes gradually. The result is consistent with the findings of district-level data. In addition, the bivariate probit result also shows that the party identification, political ideology, and the urbanization of districts still influence the voters’ voting behavior. In conclusion, this study supports existing literatures and theories, such as the Japanese voting behavior, the impact of parties’ dual candidacy strategies, and the contamination theory of mixed-member electoral systems.

並列關鍵字

Japanese House Election ； Mixed-Member Majoritarian Electoral Systems (MMM) ； Dual Candidacy ； Independent Effects ； Contamination Effects ； Treatment-effects model ； Bivariate Probit Model

參考文獻

許介鱗，1991，《日本現代史》，台北：三民書局。

黃紀，2008c，〈因果推論與觀察研究：「反事實模型」之思考〉，《社會科學論叢》，2 (1): 1-21。

楊鈞池，2005，〈一九九０年代日本選舉制度改革及其影響的分析〉，《高大法學論叢》，創刊號: 167-216。

盛治仁，2006，〈單一選區兩票制對未來台灣政黨政治發展之可能影響探討〉，《台灣民主季刊》，3 (2): 63-86。

盛杏湲、周應龍，2008，〈選樣偏誤模型在調查研究中項目無反應問題的應用〉， 《台灣政治學刊》，12 (1): 147-183。