日本自民黨之選票穩定度研究：1993、1996及2000年眾議院選舉之定群追蹤分析

Electoral Stability and Change in Japan: A Panel Study of the House Elections in 1993, 1996 and 2000

王鼎銘(Ding-Ming Wang) ； 蘇俊斌(Chun-Pin Su) ； 黃紀(Chi Huang) ； 郭銘峰(Ming-Feng Kuo)

《選舉研究》 11卷2期 (2004/11) Pp. 81-109

https://doi.org/10.6612/tjes.2004.11.02.81-109

摘要

1993年是日本政黨政治的一大轉捩點，長期居「五五體制」一黨獨大的自民黨，首度在眾議院大選失去了執政的地位。但三年後自民黨再度贏得國會最多席次、主導聯合政府的成立，並於2000年的選舉後繼續保持執政。本文是以一筆日本定群追蹤的資料為基礎，探討1993、1996及2000年三屆眾議院選舉中，自民黨票源的穩定與變遷程度，希望能一窺這段日本政黨政治的轉型時期，選民政黨投票的動態過程。
本文以馬可夫鍊模型（Markov chain models）探索定群追蹤樣本對自民黨的投票穩定度，發現去留模型（mover-stayer model）中的特例－Converse所提出的黑白模型（black-and-white model) ，最適合用來說明定群追蹤樣本在這三屆選舉中支持自民黨的動態變遷情形。也此是說，這組樣本可分為異質的兩群人：一群人自始至終堅定不移（約佔48.28%)，另一群人則無既定偏好、隨機變動（約佔51.72%）。在1993年時，隨機變動群中有40.46%支持自民黨，而堅定不移群也有37%是對自民黨效死忠的。到了1996年，原先堅定不移群之中支持自民黨的37%繼續對自民黨效死忠；而隨機變動群之中原支持自民黨的40.46%有一半外流，原未投給自民黨的59.54%則有一半轉而支持自民黨，結果自民黨在1996年的支持者淨增4.93%。依黑白模型推算，到了2000年自民黨在隨機變動群之中支持者的比例已經趨於穩定，故淨變量甚微。

關鍵字

定群追蹤資料 ； 投票穩定與變遷 ； 馬可夫鏈 ； 投票慣性 ； 黑白模型

並列摘要

Japanese party politics witnessed a dramatic change in 1993. The then ruling Liberal Democratic Party (LDP), which had dominated the Japanese politics since 1955, lost its majority seats in the 1993 House of Representatives election. Three years later, however, the LDP staged a comeback and played the major role in forming the new coalition government. It managed to maintain this control in the 2000 House election. We are curious about how Japanese voters responded to the turmoil of Japanese party system during this period. Taking advantage of a panel data of the Japanese Election Study (JES), we explore the dynamics of electoral stability and change of the LDP supporters during the 1993, 1996 and 2000 House elections.
Among several Markov chain models analyzed in this paper, we find that the Black and White model, a special case of the Mover-Stayer model, fits the JES panel data best. Specifically, this model splits subjects into two latent subgroups with maximum intragroup homogeneity and intergroup heterogeneity, each of which has its own estimated transition probabilities. The first subgroup (“stayers”), which consists of about 48.28% of the panel, always maintains its LDP or non-LDP party preference without any change. The second subgroup, which consists of about 51.72% of the panel, however, moves randomly. In 1993, 37% of “stayers” supported the LDP while 40.46% of “random movers” supported the LDP. In 1996 election, half of this 40.46% of “random movers” shifted away from the LDP while the other 59.54% non-LDP supporters shifted to the LDP. As a result, the LDP enjoyed a net gain of 4.93% in total votes in 1996. Since the Black and White model has reached its steady state after the 1996 transition, it indicates little net change in year 2000’s election.

並列關鍵字

panel data ； electoral stability and change ； Markov chain ； habitual voting ； Black and Withe model

參考文獻

陳陸輝(1999)。『固定樣本連續訪談法』(panel studies)樣本流失問題的探討。選舉研究。6(1)，175-206。

Abramson, Paul R.,John H. Aldrich,David W. Rohde(1998).(Change and Continuity in the 1996 Elections).

Achen, Christopher H.(1975).Mass Political Attitudes and the Survey Response.American Political Science Review.69,1218-1231.

Achen, Christopher H.,W. Philips Shively(1995).(Cross-Level Inference).

Anderson, T. W.(1954).Probability Models for Analyzing Time Change in Attitudes.(Mathematical Thinking in the Social Sciences).