本文分析易淹水地區水患治理計畫對高雄地區選舉結果的影響，以討論臺 灣是否有回顧型投票的現象。筆者先由個體的選擇行為建立村里的得票率模 型，再整合高雄地區各村里歷屆選舉、治水工程投入、水患災情以及所得的長 期追蹤資料，檢視高雄地區縣市首長執政期間各村里治水計畫的變動，和 2005 年到 2014 年間縣市長選舉各村里得票率變動的關聯。本研究發現，政府在各 村里所投入的治水資源，在統計上能顯著地拉開現任地方首長和一般候選人的 選情；然而，筆者卻沒有足夠的證據能主張，治水資源的投入也能為代表中央 執政陣營的候選人帶來選舉利益。

This study analyzes the impact of the Regulation Project of Flood-prone Areas on electoral results in Kaohsiung, and discusses whether Taiwan has the phenomenon of retrospective voting. The present study constructs a regression model for aggregate data based on representative voters, and combines information regarding the voting share of each candidate in each village for several elections, the allocation of flood control works, flooding records, and median income in each village. It examines the relationship between the change in flood control works allocated in each village during different mayoral terms, and the change in voting shares for each candidate in each village among elections. This study finds that flood control works can create a significant difference in voting shares between the incumbent mayor (or county magistrate) and an ordinary candidate; however, there is insufficient evidence to conclude that these constructions can also make a difference in voting share between an ordinary candidate and the candidate from the national incumbent pary.

回顧型投票; 災害管理; 易淹水地區水患治理計畫; 巢狀 勝算對數模型; 追蹤資料

retrospective voting; disaster management; Regulation Project of Flood-prone Areas; nested logit model; panel data