候選人臉書粉絲專頁追蹤網絡的面貌、形成與結果：以2016年區域立委選舉為例的社群網絡分析

Abstract

社群媒體已成為政治候選人開展政治競選活動的舞臺。本研究旨在探討政治候選人透過Facebook粉絲專頁與其他候選人建立追蹤關係的樣貌、原因（包括哪些重要的個人特徵和選舉背景）和效果（最終的得票率）。本研究主要使用社會網絡分析的方法，來分析196位採用臉書助選的候選人，形成了怎樣的同儕追蹤網絡，這個網絡具有怎樣的特徵（包括密度、大小、相互性和傳遞性）；探討哪些候選人的個人特徵和同儕追蹤網絡特徵會影響他／她跟同黨或者非同黨的候選人建立臉書追蹤關係；考察候選人在同儕追蹤網絡中的位置，包括閉合性（closure）、跨黨派性（cross-partisanship）和中介性（brokerage）是否可以影響他們最終的投票結果。這項研究的背景為2016年臺灣的區域立法委員選舉。該研究包含了多種資料來源作為預測變項和控制變項，包括政治候選人的Facebook粉絲專頁的追蹤資料、其他網路選舉活動，新聞媒體的曝光率、個人特徵、選舉特徵、粉絲數、發文數以及最終的選舉結果。分析共分為三部分。首先，使用社會網絡整體網絡和點的位置指標分析候選人的同儕追蹤網絡的樣貌。其次，是使用ERGM模型（Exponential Random Graph Model，指數隨機圖模型）預測候選人建立追蹤關係的可能性，自變項包括候選人和選舉的特徵、被追蹤者的政黨屬性和網絡本身的相互性。本研究使用階層多元迴歸分析預測候選人的得票率。結果顯示，在同儕間具有高閉合性的候選人會得到較多的選票，但跨黨派性和高中介性對得票率卻沒有顯著關聯。最後本研究會討論這些結果在政黨對立的台灣具有哪些現實的意涵。

Social media has become an increasingly popular arena for political candidates to run political campaigns. The current study aims to examine how political candidates formed online social networks with each other on Facebook. The context of the study is the 2016 legislative election (candidate sample size n=196) in Taiwan, which is the only Chinese society that has a democratic voting system. First, social network analysis is used to explore size, density, reciprocity, and transitivity of their online networks. Then, ERGM is used to analyze the factors underlying the formation of social ties in their online networks, including personal, election, and structural characteristics. Finally, hierarchical multiple regressions are performed to examine the effects of candidates' structural positions (closure, cross-partisanship, and brokerage) in the peer network on the election outcome while controlling for candidates' personal characteristics, key election features, and the amount of news coverage in traditional media. Overall, our results suggest that there is a statistically significant positive relationship between the closeness of peer social networks and election results. However, we did not find significant impacts of the cross-partisanship and the brokerage related to the election results. Implications were also discussed.