證券分析師報告之資訊內涵 – 文字探勘技術之應用

Read Between the Lines: A Textual Analysis of Equity Research Reports

黃信翔(HSIN-HSIANG HUANG)

指導教授 ： 陳坤志

國立臺灣大學/管理學院/會計學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202200915

摘要

本研究使用字典法與Latent Dirichlet Allocation（LDA）等文字探勘（Text-mining）技術，對美國某間知名國際金融服務公司所發布之證券分析師報告內文進行情感分析與主題辨識，定義撰寫語氣（Tone）與探討主題（Topic），並測試該二項資訊與投資評等、市場報酬變化之相關性。

實證結果顯示：當分析師在報告中使用了更多的負面語氣詞彙，將傾向下調投資評等；投資者對這些資訊的認知，亦將反映於下降的市場報酬上。反之，若分析師使用了更多的正面語氣，則會有完全相反的結果。除了較基礎的正負語氣外，其他語氣類型對投資評等與市場報酬變化亦有影響。此外，透過語氣與主題變數之交互作用，本研究亦發現，對於分析師與市場中之投資人而言，那些議題之研究觀點可能是重要的。

綜上所述，本研究驗證了分析師報告之資訊價值，為報告使用者對投資評等之解讀與學者對報告資訊衡量之研究提供了參考。

關鍵字

分析師報告 ； 文字探勘 ； 情感分析 ； Latent Dirichlet Allocation ； 資訊價值

並列摘要

This paper uses text-mining techniques such as Latent Dirichlet Allocation (LDA) to conduct sentiment analysis and topic discovery to define the tones and topics in the analyst reports, and examines whether these two qualitative attributes are significant in explaining changes in stock rating and market response.

The empirical results show that when analysts use more negative tone terms in their reports, they tend to downgrade the ratings; lower market returns also reflect investors' perception of this information. Conversely, if analysts use a more positive tone, the opposite results will occur. Other tones also have an impact on stock ratings and market return.

In addition, through the interaction of tones and topics, I figure out what research perspectives may be important to analysts and investors in the market.

In conclusion, this paper validates the information value of analyst reports and provides a reference for the interpretation of stock ratings and the measurement of information in reports.

並列關鍵字

Analyst Reports ； Text-mining ； Sentiment Analysis ； Latent Dirichlet Allocation ； Information Value

參考文獻

Agrawal, S., Chen, V. Y. S., Zhang, W. (2016) The information value of credit rating action reports: A textual analysis. Management Science, 62(8), 2218-2240.

Atkins, D. C., Rubin T. N., Steyvers M., Doeden M. A., Baucom B. R., Christensen A. (2012). Topic models: A novel method for modeling couple and family text data. Journal of Family Psychology, 26(5), 816–827.

Bao, Y., Datta, A. (2014). Simultaneously discovering and quantifying risk types from textual risk disclosures. Management Science, 60(6), 1371-1391.

Bellstam, G., Bhagat, S., Cookson J. A. (2021). A text-based analysis of corporate innovation. Management Science, 67(7), 4004-4031.

Blei, D. M., Ng, A. Y., Jordan, M. I. (2003). Latent Dirichlet allocation. Journal of Machine Learning Research, 3, 993-1022.