Paper: arXiv 2411.13813

Authors: Linying Lv

Abstract

I examine the value of information from sell-side analysts by analyzing a large corpus of their written reports. Using embeddings from state-of-the-art large language models, I show that qualitative information in analyst reports explains above 10% of contemporaneous stock returns out-of-sample, a value that is more economically significant than quantitative forecasts. I then perform a Shapley value decomposition to assess how much each topic within the reports contributes to explaining stock returns. The results show that analysts’ income statement analyses account for more than half of the reports’ explanatory power. Expressing these findings in economic terms, I estimate that early acquisition of analyst reports can yield significant profits. Analyst information value peaks in the first week following earnings announcements, highlighting their vital role in interpreting new financial data.

Complexity vs Empirical Score

  • Math Complexity: 5.5/10
  • Empirical Rigor: 8.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper provides a strong empirical analysis evaluating the qualitative value of sell-side analyst reports using state-of-the-art LLM embeddings. It introduces a novel framework using Shapley value decomposition to break down textual explanatory power into specific financial topics, demonstrating both statistical and economic significance via a multi-agent model.

Research Flowchart

  flowchart TD
    A[Research Goal: Evaluate Information Value of Sell-Side Analyst Reports] --> B[Data Inputs: Corpus of Written Analyst Reports & Stock Returns]
    B --> C[Methodology 1: Extract Text Embeddings via State-of-the-Art LLMs]
    C --> D[Methodology 2: Perform Out-of-Sample Regression & Shapley Value Decomposition]
    D --> E[Outcome 1: Qualitative Text Explains >10% of Stock Returns]
    D --> F[Outcome 2: Income Statement Analysis Accounts for >50% of Explanatory Power]
    E & F --> G[Key Conclusion: Early Report Access Yields Profits, Peaking 1 Week Post-Earnings]

Cited References (123)

  1. Empirical asset pricing via machine learning (2020)
  2. Financial asset returns, direction-of-change forecasting, and volatility dynamics (2006)
  3. The characteristics that provide independent information about average US monthly stock returns (2017)
  4. A unified approach to interpreting model predictions (2017)
  5. A five-factor asset pricing model (2015)
  6. Momentum crashes (2016)
  7. Risk adjustment and trading strategies (2003)
  8. Momentum has its moments (2015)
and 115 more
  1. Risk, return, and equilibrium: Empirical tests (1973)
  2. Forecasting stock indices: a comparison of classification and level estimation models (2000)
  3. How costly is it to ignore breaks when forecasting the direction of a time series? (2004)
  4. Time series momentum (2012)
  5. Market expectations in the cross-section of present values (2013)
  6. Investor sentiment aligned: A powerful predictor of stock returns (2015)
  7. A value for n-person games (1953)
  8. Noise (1986)
  9. Regime changes and financial markets (2012)
  10. Deep learning in asset pricing (2023)
  11. Autoencoder asset pricing models (2021)
  12. Cross-Sectional Expected Returns: New Fama-MacBeth Regressions in the Era of Machine Learning (2023)
  13. Characteristics are covariances: A unified model of risk and return (2019)
  14. Economic significance of predictable variations in stock index returns (1989)
  15. Predictability of stock returns: Robustness and economic significance (1995)
  16. Predicting corporate bond returns: Merton meets machine learning (2020)
  17. Option return predictability with machine learning and big data (2023)
  18. Exchange rate prediction with machine learning and a smart carry trade portfolio (2020)
  19. Machine-learning the skill of mutual fund managers (2023)
  20. A cross-sectional machine learning approach for hedge fund return prediction and selection (2021)
  21. Return sign forecasts based on conditional risk: Evidence from the UK stock market index (2013)
  22. Return signal momentum (2021)
  23. Forecasting cryptocurrencies under model and parameter instability (2019)
  24. Replicating anomalies (2020)
  25. Commonality in the determinants of expected stock returns (1996)
  26. The cross section of expected stock returns (2014)
  27. Aggregation of information about the cross section of stock returns: A latent variable approach (2017)
  28. Sparse signals in the cross-section of returns (2019)
  29. Taming the factor zoo: A test of new factors (2020)
  30. Dissecting characteristics nonparametrically (2020)
  31. Shrinking the cross-section (2020)
  32. Thousands of alpha tests (2021)
  33. AlphaPortfolio: Direct construction through deep reinforcement learning and interpretable AI (2021)
  34. Anomalies and the expected market return (2022)
  35. Integrating factor models (2023)
  36. Comparing predictive accuracy (2002)
  37. Information content of equity analyst reports (2005)
  38. Evidence on the information content of text in analyst reports (2014)
  39. Analyst information discovery and interpretation roles: A topic modeling approach (2018)
  40. Analyst report readability (2015)
  41. Dissecting corporate culture using generative AI–Insights from analyst reports (2024)
  42. Attention is all you need (2017)
  43. Determinants of the informativeness of analyst research (2006)
  44. Use of R2 in accounting research: measuring changes in value relevance over the last four decades (1999)
  45. Unusual Financial Communication-Evidence from ChatGPT, Earnings Calls, and the Stock Market (2024)
  46. Expected returns and large language models (2022)
  47. On the pricing of corporate debt: The risk structure of interest rates (1974)
  48. Product market threats, payouts, and financial flexibility (2014)
  49. From man vs. machine to man+ machine: The art and AI of stock analyses (2024)
  50. The Value of Data: Analyst Vs. Machine (2023)
  51. The association between revisions of financial analysts’ earnings forecasts and security-price changes (1990)
  52. The relative informativeness of analysts’ stock recommendations and earnings forecast revisions (1997)
  53. Do sell-side analysts exhibit differential target price forecasting ability? (2013)
  54. What drives sell-side analyst compensation at high-status investment banks? (2011)
  55. Financial analysts and their contribution to well-functioning capital markets (2017)
  56. Disclosure processing costs, investors’ information choice, and equity market outcomes: A review (2020)
  57. On the relationship between analyst reports and corporate disclosures: Exploring the roles of information discovery and interpretation (2010)
  58. PEAD. txt: Post-Earnings-Announcement Drift Using Text (2023)
  59. The virtue of complexity everywhere (2022)
  60. Analysts’ forecasts and asset pricing: A survey (2016)
  61. An empirical analysis of analysts’ target prices: Short-term informativeness and long-term dynamics (2003)
  62. Can investors profit from the prophets? Security analyst recommendations and stock returns (2001)
  63. Reading between the lines: An empirical examination of qualitative attributes of financial analysts’ reports (2012)
  64. Analyst forecast revisions and market price discovery (2003)
  65. The timing and value of forecast and recommendation revisions (2004)
  66. Bridging the gap between value relevance and information content (2000)
  67. Beyond the numbers: Measuring the information content of earnings press release language (2012)
  68. Analyst information production and the timing of annual earnings forecasts (2014)
  69. Information interpretation or information discovery: Which role of analysts do investors value more? (2012)
  70. Management earnings forecasts and value of analyst forecast revisions (2015)
  71. The effect of analyst forecasts during earnings announcements on investor responses to reported earnings (2017)
  72. Earnings announcement disclosures and changes in analysts’ information (2017)
  73. A content analysis of sell-side financial analyst company reports (1994)
  74. The effect of disclosures by management, analysts, and business press on cost of capital, return volatility, and analyst forecasts: A study using content analysis (2009)
  75. The “dartboard” column: Second-hand information and price pressure (1993)
  76. Conflict of interest and the credibility of underwriter analyst recommendations (1999)
  77. Do brokerage analysts’ recommendations have investment value? (1996)
  78. The cross section of analyst recommendations (2006)
  79. ChatGPT and corporate policies — → 2409.17933
  80. Soft information in the financial press and analyst revisions (2021)
  81. When are analyst recommendation changes influential? (2011)
  82. Are analysts’ recommendations informative? Intraday evidence on the impact of time stamp delays (2014)
  83. The use of target prices to justify sell-side analysts’ stock recommendations (2002)
  84. Analyzing the analysts: When do recommendations add value? (2004)
  85. Lookahead bias in pretrained language models (2024)
  86. Continuous auctions and insider trading (1985)
  87. Liquidity and the strategic value of information (2025)
  88. Imperfect competition among informed traders (2000)
  89. Insider trading in continuous time (1992)
  90. When the underwriter is the market maker: An examination of trading in the IPO aftermarket (2000)
  91. Trade classification algorithms for electronic communications network trades (2007)
  92. Inferring trade direction from intraday data (1991)
  93. Firm size and the informational content of financial statements (1984)
  94. The anatomy of the performance of buy and sell recommendations (1995)
  95. Financial analyst characteristics and herding behavior in forecasting (2005)
  96. Expertise in forecasting performance of security analysts (1999)
  97. Learning from prices and the dispersion in beliefs (2011)
  98. Tipping (2007)
  99. Can chatgpt forecast stock price movements? return predictability and large language models — → 2304.07619
  100. News and asset pricing: A high-frequency anatomy of the SDF (2025)
  101. The news in earnings announcement disclosures: Capturing word context using LLM methods (2025)
  102. Informed trading before analyst downgrades: Evidence from short sellers (2010)
  103. The value of client access to analyst recommendations (2006)
  104. Whose money is smart? Individual and institutional investors’ trades based on analyst recommendations (2021)
  105. Trading in the presence of short-lived private information: Evidence from analyst recommendation changes (2018)
  106. What is the value of sell-side analysts? Evidence from coverage initiations and terminations (2015)
  107. Revisiting the Cross-Section of Expected Stock Returns: Evidence from a Textual Analysis of Buy Recommendations (2023)
  108. The use and usefulness of big data in finance: Evidence from financial analysts (2024)
  109. Do analysts say anything about earnings without revising their earnings forecasts? (2019)
  110. Analyzing the analysts: Career concerns and biased earnings forecasts (2003)
  111. Before an analyst becomes an analyst: Does industry experience matter? (2017)
  112. Expectations management and beatable targets: How do analysts react to explicit earnings guidance? (2006)
  113. Do analysts gain an informational advantage by visiting listed companies? (2018)
  114. Manager-analyst conversations in earnings conference calls (2018)
  115. Do analyst stock recommendations piggyback on recent corporate news? An analysis of regular-hour and after-hours revisions (2015)

2 of 123 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.