| 摘要: | Facebook在2004年成立後，不僅成為使用者分享生活的主流平台，也成為商業、政治吸引關注的新興管道。Facebook在2006年的美國期中選舉時，為候選人建立個人頁面(Profile)，提供候選人自行更新資訊、政見與選民互動。社群媒體與選舉結果之研究濫觴由此開始。2008年美國總統大選，民主黨陣營使用微定位選舉策略，利用線上社群媒體與線下人口普查等資料，推估選民的政治立場進而量身訂做宣傳策略。在此之後，社群媒體與選舉結果之研究蓬勃發展，但仍然面臨許多問題與挑戰。
因此，本研究提出兩個研究問題企圖解決此議題上的困難，其一為社群媒體是否能提供預測選舉結果簡單且有效的方法。其二為社群媒體預測選舉結果是否受到選舉屬性影響。
為回答上述研究問題，本研究蒐集候選人投票日前一日在Facebook貼文之按讚、留言及分享數與中央選舉委員會公告的選舉資料，整理3項佔比率及15項指標。使用相關分析、迴歸分析、敘述統計、Wilcoxon等級和檢定及Kruskal-Wallis檢定檢驗。
研究結果有幾下幾點發現，第一，平均按讚佔比率、平均分享佔比率與得票率之關係具有統計意義。第二，最大按讚數、平均按讚數及最小分享數3項指標與選舉結果之關係具有實務意義。第三，選舉屬性確實對社群媒體預測選舉結果之能力造成好壞差異。第四，社群媒體能提供預測選舉結果簡單且有效的方法，但相較其他觀測期間較長之研究，也相對犧牲了一定程度的正確率。;After Facebook′s establishment in 2004, it soon became the main platform to share our daily life, as well as business and political point of views. In 2006 United States midterm election, Facebook provides the "Profile" for the candidates, enabling them to update their information, manifesto, and interact with the voters. This starts the research between social media and the political voting results. In 2008 United States presidential election, The Democratic Party employ the "Micro-positioning" strategy, using online social media and offline census data to evaluate voters′ political standpoint, and tailor-made their advertising strategy.
From that time, research concerning the relationship between social media and the voting result increase drastically. Yet, there still lays some questions and difficulties needed to confront. This research particularly aims to resolve the debates of these issues, which is: 1. Whether social media can provide a simple and effective method to predict election results. 2. The election attributes whether impact on the predicting ability of social media.
To answer the questions, this research calculated the amount of the Likes, Comments and Shares on Facebook for each candidates′ posts one day prior to the election day. We compared the collected data with data from Central Election Commission (CEC, Taiwan), and sorted out 3 shares and 15 indexes. We use correlation analysis, regression analysis, narrative statistics and Wilcoxon signed-rank test to examine.
The result indicated that, firstly, AvgLikeShares and AvgShareShares have statistical significance between the percentage of votes obtained. Secondly, three indexes, Maximum Likes, Average Likes and Minimum Shares, have practical significance between the voting result. Thirdly, electoral attributes did have impacts on the effectiveness of social media′s predicting ability. Lastly, social media could provide a simple and effective predict to the vote, by giving up some accuracy compared to other long-term research. |
