Constructing Equity Investment Strategies Using Analyst Reports and Regime Switching Models

Area of Science:

• Computational finance
• Natural language processing
• Machine learning

Background:

• Financial analysts' reports contain valuable information for investment strategies.
• Quantifying analyst sentiment from text is challenging but potentially rewarding.
• Traditional investment strategies may not fully leverage textual data.

Purpose of the Study:

• To determine the utility of analyst sentiment for stock investment strategies.
• To develop a quantifiable sentiment index from analyst reports.
• To construct and evaluate a data-driven investment portfolio.

Main Methods:

• Natural language processing (NLP) was used to create a stock-specific polarity index from analyst reports.
• Deep learning models performed time series forecasting on the polarity index.
• A regime switching model clustered forecasted values by volatility, identifying change points.
• A portfolio was constructed and rebalanced based on identified regime change points.

Main Results:

• The developed polarity index successfully captured analyst sentiment.
• Deep learning and regime switching models effectively forecasted and clustered the index.
• The proposed investment strategy, rebalanced at regime change points, significantly outperformed a benchmark portfolio.
• The study confirmed the practical utility of the polarity index in investment decision-making.

Conclusions:

• Analyst sentiment, quantified via a NLP-derived polarity index, is a valuable component for effective stock investment strategies.
• Integrating deep learning for forecasting and regime switching for portfolio rebalancing enhances investment performance.
• This research provides a novel framework for leveraging textual financial data in quantitative investing.