政黨傾向及投票行為: 以代理人基方式建模

Abstract

綜觀歷屆選戰易知，同樣的競選方式可能在某次選舉中輕易獲得選民支持並獲得勝利，但是亦可能在不同的社經情況中招致挫敗。故對於參與選戰的政黨人士而言，若能建構模型模擬: 一: 選民意見的形成 二: 不同政黨間的選戰策略，將會具有相當的價值。 本研究針對上述問題利用代理人基建模，使用並延伸一意見動態模型 Bounded Confidence model (BC model )。本模型可用來研究個別的代理人間相互的關係以及它們的集體行為。模型中亦加入了了兩項延伸: 一: 候選人自選民中抽離開來，而以一種特別的代理人類型進行模擬，以構造其不同的選戰策略。二: 選民除了遵循 BC model 之要領外，亦增加了一政黨認同的屬性。 本研究中利用了不同的組合來探討候選人和選民的關係。本模型可被作為一實驗選戰策略之工具，亦可被用作實證選戰結果的分析工具。

Throughout the course of history, it is easily observable that certain methods could easily persuade voters and results a victory, while the very same methods might fail under a different socio-economic circumstance. A simulation model that could take the following criteria into considerations would be valuable to stakeholders of campaign planners of political parties: (a) the opinion formations of the constituents (b) the campaign strategies of the competing political groups. This study employs the paradigm of agent-based modeling, and adopts and extends an opinion dynamics model, Bounded Confidence model. The model is used to investigate the influence between individual agents and their collective behaviors. The model is extended in this study by introducing the following extensions: (a) the election candidates are modelled as a special class of agents with different power and campaign tactics (b) the voters are modelled as agents with an extra attribute of the affinity toward each political party. Different simulations of this model are conducted to investigate the relationships between voters and candidates under different configurations. The model provides a testbed of election campaign strategies and could serve as an analysis tool for empirical field studies.