川普總統社群媒體發文對股市報酬之影響

Abstract

本研究主要目的探討川普總統在社群媒體發文對S&P 500指數報酬之影響，主要透過實證模型檢視政策主題之差異性，進一步檢驗影響來源為政策內容之資訊管道 （Information channel），抑或文本語氣之情緒管道 （Sentiment channel）；同時比較第一任與第二任發文之差異性。本研究聚焦於三個與市場相關的核心政策主題：外交與地緣政治、經濟與財政，以及政府與司法。樣本資料之文本主要來自X（原Twitter）與Truth Social兩平台，時間涵蓋川普總統第一任當選2016年11月至任期結束2021年1月19日，以及第二任當選2024年11月至2026年6月4日。實證模型中納入星期固定效果，以控制不同時間的共同市場環境。 實證結果顯示，川普總統不同政策主題的發文對市場具有異質性影響。外交與地緣政治類發文比重對當日 S&P 500 指數報酬呈邊際顯著負向效果，顯示此類文本較容易引發市場即時反應；經濟與財政議題類發文比重則對隔日報酬呈邊際顯著負向效果，表示市場對此類文本的消化較偏向延遲反應。整體而言，市場並不會將川普總統所有發文視為同質資訊，而是會依不同政策主題採取不同的解讀方式。進一步分析顯示，市場對文本語氣的反應較政策資訊量更為明顯。特別是在資訊量與情緒之聯合模型中，政策資訊效果整體較弱，而外交類文本的情緒對隔日影響管道最為顯著，顯示投資人較可能根據特定政策主題中的語氣線索，而非單純根據政策資訊含量本身調整市場預期。此一結果意味著，政治人物社群媒體文本對市場的影響，並不只來自其談論的內容，也與其表達方式密切相關。 此外，本研究亦發現第一任與第二任之間的市場作用機制存在差異。整體而言，本研究指出，川普總統社群媒體發文對股市報酬的影響，不僅具有政策主題異質性，也具有文本作用機制與任期異質性。本研究發現提醒投資人，在解讀政治人物社群媒體文本時，應將其放回當時的政治與市場環境中判斷。相同的正向語氣或政策表述，在不同任期、不同平台與不同市場預期下，可能代表不同訊號。

This thesis examines the impact of President Trump’s social media posts on S&P 500 index returns. Using empirical models, this study first investigates whether market reactions differ across policy topics, and further examines whether the impact comes from the information channel of policy content or the sentiment channel of textual tone. It also compares the differences between Trump’s first and second terms. The analysis focuses on three market-related policy themes: Foreign Policy and Geopolitics, Economy and Fiscal Issues, and Government and Judiciary. The text data are collected from X, formerly Twitter, and Truth Social, covering Trump’s first term from November 2016 to January 19, 2021, and his second term from November 2024 to June 4, 2026. The empirical models include weekday fixed effects to control for weekday patterns and common market conditions across time. The empirical results show that Trump’s posts across different policy topics have heterogeneous effects on the market. Posts related to Foreign Policy and Geopolitics have a marginally significant negative effect on same-day S&P 500 returns, suggesting a more immediate market response. Posts related to Economy and Fiscal Issues have a marginally significant negative effect on next-day returns, indicating a more delayed market reaction. Overall, the market does not treat all of Trump’s posts as homogeneous information, but responds differently depending on the policy topic. Further analysis shows that the market reacts more clearly to textual tone than to policy information density. In the joint model, the effect of policy information is generally weak, while the sentiment of foreign-policy-related posts remains the clearest next-day channel. This suggests that investors may adjust market expectations based more on tone signals within specific policy topics than on policy information density alone. The results indicate that the market impact of political figures’ social media texts depends not only on what is discussed, but also on how it is expressed. In addition, this study finds differences in the market mechanisms between Trump’s first and second terms. Overall, the findings suggest that the impact of Trump’s social media posts on stock market returns involves policy-topic heterogeneity, differences in textual channels, and term heterogeneity. These results remind investors that political social media texts should be interpreted within their political and market context. The same positive tone or policy statement may represent different signals across different terms, platforms, and market expectations.