社群媒體對股市交易量之影響

Abstract

社群媒體在你我生活中之佔比日益增加，對各領域影響也逐漸提升，對股 票市場之影響更是不容小覷，美國總統川普或是特斯拉執行長馬斯克的一篇推 特文章，就可以對美國股市帶來巨大影響，雖然台灣並無如他們那般呼風喚雨 的人物，但從之前散戶大戰華爾街之 GAMESTOP 案例，即可了解社群媒體平 台及平台上凝聚之散戶已擁有不可忽視之影響力，且台灣也陸續出現許多投資 類型的社群媒體平台。

本研究使用台灣最大股市社群 ─ CMONEY 股市爆料同學會之發文量作為 研究資料。本研究以台灣上市櫃公司為研究對象，研究期間為 2020 年 1 月至 2022 年 12 月，共計三年。本研究採用固定效果模型，實證結果發現社群媒體 發文量確實對交易量具正向影響。接著使用子樣本法及交乘項法兩方法檢驗該 影響之異質性。研究結果顯示該影響於下列情境下較為顯著:機構法人持股 高、熱門度高、投機性高、波動度高、成長性高、資訊不對稱程度高之標的或 大盤指數處於高點之時期，期望本研究可作為未來欲以社群媒體資料預測股價 或交易量模型之參考。

然本研究仍受限於許多現實面因素，如研究資料僅可取得發文量，無法獲得文章內容、讚數及留言數、平台用戶間之追蹤關係等資料，因此無法進行更深入之研究，如:對文章及留言進行情緒分析、探討社群網路架構如何影響情緒傳播等。此外，由於投機、熱門、成長等概念較為抽象，在尋找代理變數去衡量這些項目時必定會出現誤差，進而影響結果。隨未來資料量提升，可以選用更適合之模型並尋找更貼切之變數以衡量這些抽象概念，期望更進一步探索社群媒體對股市之影響，對業界及學界帶來更多貢獻。

The impact of social media on our daily life increases profoundly. For instance, a tweet by Donald Trump or Elon Musk can bring huge volatility on US stock market. Though nobody in Taiwan could be as influential as them, the investment social media and the small investors on those platforms are a lot more powerful than before. From the case of GameStop, a group of small investors on Reddit could still give Wall Street a wild week. Recently, there are more and more investment social media platforms in Taiwan.

This research uses the data from CMoney, the largest investment social media in Taiwan, and focuses on companies listed on TWSE and the OTC market. The research period starts from January 2020 to December 2022. The model used in the research is fixed effect model, including the entity fixed effect, time fixed effect, and the trading volume lag 1 period to 3 periods respectively as control variables.

The empirical results show that social media discussion does affect the trading volume in the stock market. Then, we use two methods, subsample approach and interaction term approach, to examine the heterogeneity of the social media effect. The empirical results reveal that the effect is stronger on the popular stocks, speculative stocks, volatile stocks, high-growth stocks, stocks which are majorly held by institutional investors and stocks with higher information asymmetry.The effect is also stronger when the TAIEX is relatively high. I hope these results could enhance the accuracy of the model which uses social media content to predict stock price or trading volume.

However, this research is limited due to practical reasons. The only data I could get from CMoney is the amount of post. The content of the post, how many likes and comments the post has, and the following relationship among those platform users are not available. Therefore, many in-depth research could not be done, such as sentiment analysis on posts and comments, or how would the social media structure affect the flow of information and sentiment. In addition, the abstract concepts, such as popularity and high-growth potential, cannot be evaluated accurately. Thus, the proxy variables used in the research might be biased. I hope that in the future, as the data amount surges, we could choose a more suitable model and better variables to further explore the impact of social media on the stock market. By doing so, we could benefit the industry and the academy.