選舉民意調查可以發現，即使到選前幾天往往仍會有近三成的受訪者「無意見」、「不知道」或者「拒答」，但這群未表態者的投票動向卻是選戰勝負的關鍵所在。因此，在擬定選戰策略時，所發展的選舉預測模型是否能有效推算出現有及潛在支持者的比率，就顯得格外重要。基於此，本研究擬以行銷「品牌知名度」理念建構選舉預測模型，並以民意調查中「表態者」資料為基礎，加上對「未表態者」的研判及推估而形成最後預測值。品牌知名度是指消費者是否容易想到與所認識品牌的某些特性。品牌知名度是協助消費者簡化產品資訊，方便購買決策的一項有利工具。如果品牌知名度很高，則消費者在進行購買決策時，該品牌進入消費者「喚起集合」的可能性就會提高，亦即該品牌會進入「考慮購買」的名單中，而被購買的機會也會增加。本研究即根據上述原則，並參酌李錦河與溫敏杰（1998）運用行銷「產品屬性」理念所建構的「選民需求
指標法選舉預測模型」中「流出與流入」的概念與公式，共同建構新的選舉預測模型。本研究藉由2005 年中國國民黨黨主席選舉、2008 年台南市第二選區立委選舉等地區性調查，以及2008 年總統選舉全國性調查，驗證預測模型之有效性。研究結果發現，以「品牌知名度」理念所建構的選舉預測模型，確能充分掌握選民對候選人之投票意願，及因外在與內在環境變化所造成投結果的影響。因此，本模型確能達到操作簡易、精確、快速反映事實之目的，為一良好的選舉預測模型。

Most survey institutions in Taiwan take directly inquiry method to get
the candidate supporting ratio these days, such as “If the election was held tomorrow, which candidate would you vote for?” However, nearly 30 percent of respondents won’t show their opinions even before Election Day, and unfortunately, these respondents’ may decide the election result. Therefore,the election prediction model which policy makers use to estimate the present and potential support ratio is regarded as important. Brand awareness is the extent that consumers are easy to remind or know some characteristics of the brand. It’s a tool helps consumers to simplify the product information and to make the decision quickly. Hence, based on the brand awareness theory of marketing and information from polls, this paper develops an election prediction model which includes estimation of the unknown opinion response.This paper takes three examples to investigate the effective of a new model as follows: the 2005 Kuomintang party chairman election, 2008 legislative election of the second electoral district in Tainan, and 2008 president election. The findings reveal that an election prediction model constructed using brand awareness theory and voter’s request indicator can
fully reflect the voting intention and the election result even though it is influenced by changing internal and external environment. Therefore, this model performs well on both the accuracy of the model and the simplicity of the computation process.