「選舉預測」最富挑戰的地方，就是投票結束即有了答案；因此，準確的預測方法影響選戰至巨。在實務上，以「直接詢問法」預估得票率，尚有大量的「拒答」、「不知道」之選民，其結果與實際開票結果比較，顯然低估每位候選人的得票率；因此，要能準確估算候選人之得票率實為不易。1997年台南市市長選舉，參選之七位候選人各個皆為一時之選，競爭情況非常激烈；本研究以此為研究對象，除探討「候選人取向」為最重要之投票抉擇因素外；並參著行銷學「產品屬性」觀念，建構「選民需求指標」選舉預測模式。其內容設計主要以候選人屬性（條件、特質）為根據，經初步調查求出選民需求量後，再經調整、修正及推估求出候選人真正的預估得票率。此模式經多次驗證，結果確能達到操作過程簡易、反應事實準確快速之目的；再者將「選舉預測」界定為非營利行銷之領域，更能達成行銷理論與選戰結合之目標。

The most challenging in "Election Prediction" is that the an-swer is right there after voting; hence, the way of precise prediction affects an election greatly. In practical experience, with the way of "Direct Inquiry" to predict the rate of ballot, there is still a large amount of "refusal to answers," "having no ideas" voters. Appar-ently, in comparison with the prediction and the result, the way un-derestimates the rate of ballot of each candidate. Hence, it is not easy to precisely predict the rate of ballot of each candidate. In 1997 Tainan Mayoral Election, seven candidates are all first rate, so that the campaign was keenly competitive. The study is based on the above-mentioned election to explore that voter's be-havior-Candidate orientation is the most crucial factor for the elec-tion, and to construct "Voter's Request Indicator" in "Election Pre-diction Model" referred to the idea of "Product Attribute" in Mar-keting Management. The design of "Voter's Request Indicator" is mainly contributed by the candidates' attributes (condition, charac-teristics). After the preliminary investigation, with the amount of voter's request, we acquire the result of the rate of ballot of each candidate in accordance with adjustment, amendment and estima-tion. The model is verified several times, the result shows the goals of the easy process of operation and the fast precise reflection of the reality. Further, "Election Prediction" is defined in the unprof-itable sphere, so it can achieve the integration of the theories of Marketing Management and the election.