LLMs for Earnings Prediction: Outperforming Analysts

Here’s a summary of recent developments in using Large Language Models (LLMs) for company earnings prediction:

Key Developments:

• LLMs Outperforming Analysts: Recent research indicates that LLMs can outperform some financial analysts in predicting the direction of future earnings. A study from the University of Chicago Booth School of Business suggests LLMs can achieve an accuracy of 60.4% using “chain-of-thought prompting,” surpassing the average analyst prediction by 7 percentage points. This is particularly impressive because the LLMs were provided only with balance sheet and income statement data, without additional narratives or context.
• Active Role in Financial Decision-Making: LLMs are capable of playing a more active role in financial decision-making. The forecasts generated by LLMs added more value when human biases or inefficiencies were present.
• Superior Forecasting Accuracy: ChatGPT 4.0 has demonstrated superior directional earnings forecasts compared to human analysts, after examining anonymized historical balance sheet and income statement data. This AI-based analysis could potentially lead to stock market outperformance.
• Alpha Generation: AI models can outperform the broader stock market if annual portfolios are formed based on their predictions, with performance measured monthly.
• Fine-Tuning LLMs: A novel approach involves fine-tuning LLMs to analyze earnings reports alongside external factors like market index performance and analyst grades, with the goal of providing more reliable predictions of next-day stock performance.
• QLoRA-Enhanced LLM Approach: One paper introduces an advanced approach employing LLMs instruction fine-tuned with instruction-based techniques and quantized low-rank adaptation (QLoRA) compression. This methodology integrates ‘base factors’ (financial metric growth, earnings transcripts) with ‘external factors’ (market indices, analyst grades) to create a comprehensive dataset.
• llama-3-8b-Instruct-4bit Model: The llama-3-8b-Instruct-4bit model showed significant improvements over baseline models, achieving 16% higher accuracy and 10% better Weighted F1 compared to GPT-4.
• Limitations: LLMs are more likely to be inaccurate for smaller firms, those with higher leverage, or those reporting losses. Analysts may be better at dealing with complex financial circumstances by factoring in “soft information” and context outside of financial statements.
• Core Earnings Measurement: LLMs can estimate core earnings from annual 10-K filings by scraping filings, extracting text, and using the GPT-4o API. The models can generate valid, neutral, scalable measures of core earnings.
• Sentiment Analysis and Return Prediction: LLMs outperform traditional word-based models in both sentiment analysis and return prediction. Returns respond slowly to news, and LLMs can capture this.

Commentary:

The developments in employing LLMs for company earnings prediction are promising. The ability of these models to process financial data and generate predictions that rival or even surpass human analysts highlights their potential to revolutionize financial analysis. However, it’s important to acknowledge the limitations. LLMs are not perfect and can be less accurate for certain types of companies or in complex financial situations. Human analysts still have a role to play, particularly in incorporating qualitative information and contextual understanding.

The ongoing research and development in this area, including fine-tuning LLMs with external factors and using techniques like QLoRA, are paving the way for more accurate and reliable financial forecasting tools. As LLMs continue to evolve, their role in financial decision-making is likely to expand, offering investors and regulators valuable insights and potentially democratizing access to sophisticated financial analysis.

Disclaimer: above content was searched, summarized, synthesized and commented by AI, which might make mistakes.

Offered by Creator: Company Recommender is a leading-edge platform dedicated to democratizing access to professional company knowledge and insights. By leveraging advanced artificial intelligence and intuitive design, Company Recommender empowers every individual to discover, evaluate, and understand companies with unprecedented depth and clarity through the technologies of recommender systems, statistical machine learning and large language models, e.g., AI forecasted company earnings and forecast explanations.

Try Company Recommender today!