摩根大通：投资的AI收益电话分析 – 解决理性注意力缺失问题

Investment Rating - The report does not explicitly state an investment rating for the industry or companies analyzed [2]. Core Insights - The report investigates the use of deep sentence-level analysis of earnings calls to extract insights that can serve as investment signals, addressing the issue of "rational inattention" [2]. - The SmartBuzz on Transcript signal shows promising performance with a Sharpe Ratio of up to 0.8, indicating a balance of low alpha and low volatility [2]. - Weekly to monthly frequencies are recommended for analyzing earnings call transcripts, favoring average scores to avoid rewarding longer calls [2][4]. - The Corporate branch of the SmartBuzz taxonomy is identified as the strongest performer, followed by ESG and Macro themes, while External Events are less effective for stock-specific insights [2][32]. Summary by Sections SmartBuzz Introduction - SmartBuzz is an NLP framework that tracks market themes and their exposure to stocks, sectors, and countries, utilizing a taxonomy of over 3,000 financial terms [4][5]. SmartBuzz Backtesting - The backtesting results indicate a strong correlation between SmartBuzz signals and subsequent returns, particularly for the Corporate and ESG themes [8][20]. Results – SmartBuzz from Transcripts - Weekly backtests from 2010 to 2024 show that portfolios using SmartBuzz Aggregate scores with a decay of 0.9 achieved a Sharpe Ratio of 0.85 and returns of 6.87% [21][24]. - Monthly backtests indicate that a decay of 0.5 with a 6-month forward fill yields a Sharpe Ratio of 0.70 and annual returns of 5.01% [26][29]. SmartBuzz Transcripts at Lower Theme Levels - At Level 2, the Corporate branch is the most significant, followed by Macro and ESG, with External Events being less impactful [32][41]. Performance Metrics - The report provides various performance metrics for different themes, indicating that Corporate themes generally outperform others in terms of returns and Sharpe Ratios [41][44].