台灣政治中的臉書廣告：2024年區域立委候選人的內容策略與受眾規模分析

Abstract

有別於電視廣告或是大型廣告看板，臉書廣告提供了候選人精準投放廣告的功能。候選人得使用低廉的價格，精準地將廣告內容投放給目標受眾。然而，政治廣告技術的進步不僅帶來便利，也帶來了隱憂。候選人可以利用這項技術，讓假訊息或是不利於對手的資訊僅存在於少數選民的動態牆（newsfeed）上。臉書廣告有可能成為不實資訊的溫床，或降低公眾討論的可能性。在這樣的背景下，研究自Meta廣告資料庫（Meta Ad Library）下載了2024年區域立委候選人的廣告資料，從候選人臉書廣告中的主題、風格、功能，以及這些廣告的花費、曝光次數及預估受眾規模，了解台灣競選政治廣告的面貌，以及分析「隱藏貼文（dark posts）」的可能性。 研究分成三部分。第一部分採用迴歸分析，了解候選人之年紀、政黨、選區、執政狀態、選情對於投入廣告預算的關係。同時，也會帶入平等化與均值化假設之辯論。第二部分先採用內容分析法，將廣告編碼。再從議題所有權理論的角度，使用迴歸分析了解不同的候選人型態是否投放了不同的廣告內容。第三部分專注於討論廣告的「預估受眾規模（estimated audience size）」，嘗試歸納候選人投放的策略為何，以及隱藏貼文是否存在。 研究發現，作為挑戰者以及處在激烈選區的候選人顯著花費較多廣告預算。且臉書廣告給予小黨更有效觸及選民的機會。在內容方面，大部分的廣告內容以正面的個人資訊為主；負面攻擊廣告、談論議題或政策的廣告皆為少數。另外，負面廣告及議題相關討論少，加上國民黨、民進黨投放的內容與政黨呈現高度連結，顯示大黨傾向於使用臉書廣告動員群眾，而非透過議題倡議吸引中間選民。在廣告功能方面，逾半數的廣告以傳遞資訊為目標，而僅有以動員為目標的廣告呈現明顯時間趨勢，隨選舉接近而增加。最後，大多數廣告以較為廣泛的大眾為目標受眾，預估受眾規模的數值在本研究中也僅能以策略差異作結，隱藏貼文的疑慮在本次選舉並不明顯。

Unlike television ads or large billboards, Facebook ads enable candidates to microtarget their advertisements cheaply and ensure that content reaches the intended audience. However, this advancement in political ad technology raises concerns about disseminating misinformation and reducing public discourse opportunities. This study examines the 2024 regional legislative candidates' ad data from the Meta Ad Library, analyzing the topics, styles, functions, spending, exposure, and estimated audience size of candidates' Facebook ads to understand the landscape of political campaign advertising in Taiwan and the potential for dark posts. The research is divided into three parts. The first part uses regression analysis to explore the relationship between candidates' parties, ages, constituencies, incumbency status, election competitiveness, and advertising budgets, addressing the debate on normalization and equalization hypotheses. The second part employs content analysis and regression analysis from the perspective of issue ownership theory to determine if different types of candidates are associated with different ad content. The third part focuses on the "estimated audience size" of the ads to deduce the candidates' targeting strategies and the presence of dark posts. The study finds that challengers and candidates in highly competitive constituencies spend significantly more on advertising. Also, Meta ads provide small parties with compelling opportunities to reach voters. Regarding the contents, most ads focus on positive personal information, with fewer negative attack ads and issue-based ads. Additionally, the scarcity of negative advertisements and issue-centered discussions, combined with the highly partisan content from KMT and DPP, indicates that significant parties use Facebook advertisements to mobilize their base rather than attract swing voters with issue advocacies. In terms of advertising functions, more than half of the advertisements aim to convey information. Only ads targeting mobilization show a clear temporal trend, increasing as the election approaches. Lastly, most ads target a relatively wide range of voters, and differences in estimated audience size indicate strategic variations rather than a prominent concern for dark posts in this election.