台灣選舉預測：預測市場的運用與實證分析

Abstract

透過預測市場的機制來預測選舉，在國外已經有相當豐碩的研究成果，也被證明具有很高的準確度。因此，本論文要回答的問題是：預測市場是否可以準確預測台灣選舉結果？本論文先介紹預測市場的機制，進而運用預測市場的機制來預測台灣的選舉。根據未來事件交易所的資料顯示，我們發現預測市場機制運用於2006 年北高市長選舉以及2008年台灣總統與立委選舉同樣可以準確預測選舉結果。在這三次的選舉預測中，預測市場對選舉預測的準確度均超過同一時間所進行（但尚未加權）的民意調查。對此，本論文也比較了預測市場和民意調查的差異。

Prediction Markets have been widely used abroad to forecast election results for many years. It has also been proved that such a mechanism can accurately predict election outcomes. Therefore, this paper would like to understand whether prediction markets can accurately predict Taiwan`s election outcomes. First, this paper introduces the mechanism of prediction markets and then applies the prediction market mechanism to predict Taiwan`s elections. According to the data provided by the Exchange for Future Events, we find that the prediction markets mechanism has accurately predicted the outcomes of Taiwan`s Taipei and Kaohsiung mayoral elections in 2006, and Taiwan`s presidential election and legislator election in 2008. For these three elections, the predictive accuracy of prediction markets is higher than (unweighted) opinion poll surveys conducted during the same period. This paper then discusses the differences between prediction market and opinion survey.