針對選舉預測研究，本文嘗試提供一項新的選舉預測分析工具─空間訊息法（spatial signals approach）以進行選舉預測分析。此項選舉預測方法為KLR訊息法的延伸，KLR訊息法原屬於貨幣危機預警的重要預測方法，本文則將此項預測方法納入空間性質的地區因素，探討鄰近地區預測因素的重要性，進而發展出空間訊息法。本文主要的研究目的，在於提出結合地理脈絡因素的空間訊息法，並對於臺灣的總統選舉結果進行檢驗，主要驗證以下兩點：第一、嘗試探討臺灣歷年來總統選舉的空間脈絡性（或稱為鄰近效果）；第二、希望驗證空間訊息法於臺灣選舉預測分析上的有效性。實證分析上，本文使用2000年至2012年臺灣總統選舉資料，進行空間訊息分析，探討鄰近村里得票率對於該村里得票率的預測能力。分析結果指出，無論是空間訊息配適方法還是預測方法，皆顯示前期鄰近村里的得票率，對於當期該村里的得票率，在空間分布上有著相當穩定的解釋與預測性。本文介紹並衡量空間訊息法對於臺灣選舉結果的預測能力，並據此驗證臺灣總統選舉村里得票率的地理脈絡性與鄰近效果。

This paper aims to introduce a new election prediction method which is called the ＂spatial signals approach＂. The spatial signals approach is a direct extension of the KLR signals approach of Kaminsky et al . (1998), which was developed as an early warning system of crisis prediction. There are two research purposes of this paper; the first is to examine the neighborhood effect of presidential elections in Taiwan, and the second is to verify the effectiveness of prediction of the presidential elections in Taiwan by the spatial signals approach. Our empirical data comprised of the results of the 10th, 11th, 12th and 13th presidential elections from 2000 to 2012. The spatial units of the spatial signals approach are villages. We find that the spatial signals approach is a better fit for the outcome of the presidential election. The findings also reveal that the best forecast accuracy is almost up to 90 percent, and the other forecast accuracy is 70 percent on average. This study concludes that the prediction accuracy of the spatial signals approach is good for predicting the outcome of presidential elections in Taiwan, and shows that the presidential elections in Taiwan indeed exhibit the neighborhood effect.