本研究乃參著行銷學中產品核心需求、動機、消費者行為、情境等理論為依據， 並將行銷觀念應用於選民投票行為之預測。其中由行銷「產品屬性」觀念建構的「選民需求 指標」選舉預測模式，應用於 1997 年臺南市市長「多人參選當選單一席次」選舉選前得票 率預測，獲得相當良好之效果；此次再以 1998 年臺南市安南區市議員選舉為例，加以驗證 「選民需求指標」模式於「多人參選當選多席次」之選舉預測上。雖因現有電信系統限制而 無法按鄰、里分層比例抽樣，但本研究以模式調查後預估之結果和與選舉專家深度訪談之資 料相比對，再根據實況予以修正，最後亦能有效預測當選結果。因此，本研究擬探討的重點 包括： 一、探討行銷學理論是否亦適用於「多人參選當選多席次」之「市議員選舉」選民投票行為 之預測？ 二、基層市議員選舉，因血緣、地緣、人緣等因素影響，致選民投票抉擇時是否仍以「候選 人因素」為首要考量？ 三、模式操作時是否與「多人參選當選單一席次」選舉，可達簡易快速、確切反應狀況之目 的？ 四、選前三天，選民之策略性投票影響選情至鉅，其中又以「多人參選當選多席次」選舉為 甚。從事選舉預測時如何因應？因應策略為何？ 五、模式未來發展之方向？ 研究結果發現本研究所建構之「選民需求指標」模型確能準確預測選民之投票行為。

　　　　　Based on the theories of core demand, motivation, consumer behavior, and strategic contingencies from marketing management, this paper focused on applying marketing concepts to the prediction of voter's voting behavior. In this research, the idea of using "product attribute" in marketing management, has constituted a good prediction model with "Voter's Request Indicator". Before the 1997 "Many candidates participate in an election and there is only one elected" Tainan Mayoral Election, the said concept had been applied and indeed brought about effective voting rate prediction. Furthermore, the 1998 Municipal Election in An-Nan District in Tainan City has again proved the model of "Voter's Request Indicator" can be applied to predict the results of "Many candidates participate in an election and many are by ballot" election. Although being limited by the current telecommunication system, the sampling can hardly be divided between Lin (a small community unit which consists of a number of families in the same neighborhood) or Li (a neighborhood). With some correction, according to some status quo, I compare the model with the data collected by experts, this paper has shown strong merits. Hence, in this research the following points will be explored: 1. Can marketing management theory also be applied well in the prediction of voter's voting behavior in terms of "Many candidates participate in an election and many are by ballot" municipal election? 2. Is "Candidates' Attributes" the premise or blood relationship, place connection, popularity can be one important cause while voting? 3. Can using the model be a true reaction of status quo, with proficiency, compared with "many candidates participate in an election and there is only one elected" election? 4. Three days before the election, voters' strategic voting might effect the results greatly, especially "Many candidates participate in an election and many are by ballot" How do election predictions adjust to the situation? 5. What will be the future development of the model? This paper has come to the conclusion that using the model of "Voter's Request Indicator" indeed can predict voter's voting behavior correctly.