Quantitative Models and Construction Methods Model Name: Net Liquidity - **Construction Idea**: Net liquidity is derived from the Federal Reserve's balance sheet, focusing on the core components of cash in circulation and bank reserves[2] - **Construction Process**: - Calculate net liquidity as total assets minus Treasury General Account (TGA) and reverse repos - Formula: $ \text{Net Liquidity} = \text{Total Assets} - \text{TGA} - \text{Reverse Repos} $ - This represents the base money supply under the money multiplier effect, directly determining the amount of money available for transactions and credit activities in the market[2][21] - **Evaluation**: Net liquidity effectively reflects the real available funds in the market, providing a clearer signal than total assets[31] Model Name: Federal Reserve Credit Support - **Construction Idea**: Federal Reserve credit support is based on the quality of collateral purchased by the Fed, aiming to enhance credit by buying lower-grade collateral[2] - **Construction Process**: - Construct the credit support indicator as the ratio of long-term government bonds, federal agency bonds, and mortgage-backed securities (MBS) to cash in circulation, reserves, and reverse repos - Formula: $ \text{Credit Support} = \frac{\text{Long-term Government Bonds} + \text{Federal Agency Bonds} + \text{MBS}}{\text{Cash in Circulation} + \text{Reserves} + \text{Reverse Repos}} $ - This indicator is smoothed and compared year-over-year to identify the direction of credit support changes[2][42] - **Evaluation**: The credit support indicator is significantly negatively correlated with credit spreads, indicating its effectiveness in reducing default risk in the economy[42] Model Name: Fed Sentiment Index - **Construction Idea**: The Fed Sentiment Index captures the sentiment of Federal Reserve officials' public statements to predict policy tendencies[3] - **Construction Process**: - Use Natural Language Processing (NLP) to analyze the sentiment of Fed officials' speeches, interviews, tweets, etc. - Assign scores ranging from extremely dovish to extremely hawkish - Calculate the total sentiment score daily to provide timely and comprehensive interpretations of Fed communication[57][59] - **Evaluation**: The Fed Sentiment Index improves the accuracy of predicting federal funds rates and bond yields, offering better differentiation for the S&P 500 compared to low-frequency document signals[59] Model Name: Market Implied Rate - **Construction Idea**: The market implied rate tracks the market's expectations of future interest rate changes based on federal funds rate futures contracts[3] - **Construction Process**: - Calculate the implied rate as $ 100 - \text{futures price} $ - Focus on the price difference between futures contracts maturing in the next month and those maturing in the month of the upcoming FOMC meeting - Smooth the quarterly differences to identify marginal changes in market expectations[68][72] - **Evaluation**: The market implied rate indicator leads actual policy rate adjustments, providing early signals of policy shifts[72] Model Name: Announcement Surprise - **Construction Idea**: Announcement surprise captures the unexpected impact of FOMC meeting decisions on market expectations[3] - **Construction Process**: - Use the price changes of federal funds rate futures contracts maturing three months after the meeting to calculate the difference between actual and implied rate changes - Sample high-frequency data 10 minutes before and 20 minutes after the meeting to precisely capture the policy expectation gap[74][75] - **Evaluation**: Announcement surprise effectively identifies the unexpected tightening or easing of Fed policies, with significant impacts on bond yields[74] Model Backtest Results Net Liquidity - **Annualized Excess Return**: 5.1% relative to S&P 500 equal-weight benchmark[92] - **Annualized Excess Return**: 7.2% relative to Nasdaq 100 equal-weight benchmark[92] - **Maximum Drawdown Reduction**: 15% for S&P 500, 31% for Nasdaq 100[92] Federal Reserve Credit Support - **Annualized Sharpe Ratio**: Enhanced for most assets during periods of increased credit support[48] Fed Sentiment Index - **Annualized Excess Return**: Significant differentiation for S&P 500 returns in hawkish vs. dovish sentiment periods[61] Market Implied Rate - **Annualized Excess Return**: Effective in predicting policy shifts, leading actual rate adjustments[72] Announcement Surprise - **Bond Yield Impact**: Higher future bond yields in unexpected easing scenarios compared to unexpected tightening scenarios[76] Quantitative Factors and Construction Methods Factor Name: Net Liquidity - **Construction Idea**: Derived from the Federal Reserve's balance sheet, focusing on cash in circulation and bank reserves[2] - **Construction Process**: - Calculate net liquidity as total assets minus TGA and reverse repos - Formula: $ \text{Net Liquidity} = \text{Total Assets} - \text{TGA} - \text{Reverse Repos} $ - This represents the base money supply under the money multiplier effect, directly determining the amount of money available for transactions and credit activities in the market[2][21] - **Evaluation**: Net liquidity effectively reflects the real available funds in the market, providing a clearer signal than total assets[31] Factor Name: Federal Reserve Credit Support - **Construction Idea**: Based on the quality of collateral purchased by the Fed, aiming to enhance credit by buying lower-grade collateral[2] - **Construction Process**: - Construct the credit support indicator as the ratio of long-term government bonds, federal agency bonds, and MBS to cash in circulation, reserves, and reverse repos - Formula: $ \text{Credit Support} = \frac{\text{Long-term Government Bonds} + \text{Federal Agency Bonds} + \text{MBS}}{\text{Cash in Circulation} + \text{Reserves} + \text{Reverse Repos}} $ - This indicator is smoothed and compared year-over-year to identify the direction of credit support changes[2][42] - **Evaluation**: The credit support indicator is significantly negatively correlated with credit spreads, indicating its effectiveness in reducing default risk in the economy[42] Factor Name: Fed Sentiment Index - **Construction Idea**: Captures the sentiment of Federal Reserve officials' public statements to predict policy tendencies[3] - **Construction Process**: - Use NLP to analyze the sentiment of Fed officials' speeches, interviews, tweets, etc. - Assign scores ranging from extremely dovish to extremely hawkish - Calculate the total sentiment score daily to provide timely and comprehensive interpretations of Fed communication[57][59] - **Evaluation**: Improves the accuracy of predicting federal funds rates and bond yields, offering better differentiation for the S&P 500 compared to low-frequency document signals[59] Factor Name: Market Implied Rate - **Construction Idea**: Tracks the market's expectations of future interest rate changes based on federal funds rate futures contracts[3] - **Construction Process**: - Calculate the implied rate as $ 100 - \text{futures price} $ - Focus on the price difference between futures contracts maturing in the next month and those maturing in the month of the upcoming FOMC meeting - Smooth the quarterly differences to identify marginal changes in market expectations[68][72] - **Evaluation**: Leads actual policy rate adjustments, providing early signals of policy shifts[72] Factor Name: Announcement Surprise - **Construction Idea**: Captures the unexpected impact of FOMC meeting decisions on market expectations[3] - **Construction Process**: - Use the price changes of federal funds rate futures contracts maturing three months after the meeting to calculate the difference between actual and implied rate changes - Sample high-frequency data 10 minutes before and 20 minutes after the meeting to precisely capture the policy expectation gap[74][75] - **Evaluation**: Effectively identifies the unexpected tightening or easing of Fed policies, with significant impacts on bond yields[74] Factor Backtest Results Net Liquidity - **Annualized Excess Return**: 5.1% relative to S&P 500 equal-weight benchmark[92] - **Annualized Excess Return**: 7.2% relative to Nasdaq 100 equal-weight benchmark[92] - **Maximum Drawdown Reduction**: 15% for S&P 500, 31% for Nasdaq 100[92] Federal Reserve Credit Support - **Annualized Sharpe Ratio**: Enhanced for most assets during periods of increased credit support[48] Fed Sentiment Index - **Annualized Excess Return**: Significant differentiation for S&P 500 returns in hawkish vs. dovish sentiment periods[61] Market Implied Rate - **Annualized Excess Return**: Effective in predicting policy shifts, leading actual rate adjustments[72] Announcement Surprise - **Bond Yield Impact**: Higher future bond yields in unexpected easing scenarios compared to unexpected tightening scenarios[76]