Quantitative Models and Construction Methods Model: Duration Timing Model - **Construction Idea**: Predict the yield curve and map the expected returns of bonds with different durations[20] - **Construction Process**: - Use the improved Diebold2006 model to predict the instantaneous yield curve - Predict level, slope, and curvature factors - Level factor prediction based on macro variables and policy rate following - Slope and curvature factors prediction based on AR(1) model[20] - **Evaluation**: The model effectively predicts the yield curve and provides actionable insights for bond duration management[20] - **Test Results**: - July return: 6.6bp - Benchmark return: -25.8bp - Strategy excess return: 32.4bp[21] Model: Gold Timing Model - **Construction Idea**: Relate the forward real returns of gold and US TIPS to construct the expected return model for gold[32] - **Construction Process**: - Use the formula: $E[Real\_Return^{gold}]=k\times E[Real\_Return^{Tips}]$ - Estimate parameter k using OLS with an extended window - Use the Fed's long-term inflation target of 2% as a proxy[32] - **Evaluation**: The model provides a robust framework for predicting gold returns based on TIPS yields[32] - **Test Results**: - Expected return for the next year: 22.4% - Past year absolute return: 39.77%[33][35] Model: Active Risk Budget Model - **Construction Idea**: Combine the risk parity model with active signals to construct an active risk budget model for optimal stock and bond allocation[37] - **Construction Process**: - Use the Fed model to define equity risk premium (ERP): $ERP={\frac{1}{PE_{ttm}}}-YTM_{TB}^{10Y}$ - Adjust asset weights dynamically based on ERP, stock valuation percentiles, and market liquidity (M2-M1 spread) - Convert equity asset signal scores into risk budget weights using the softmax function: $softmax(x)={\frac{\exp(\lambda x)}{\exp(\lambda x)+\exp(-\lambda x)}}$[39][47] - **Evaluation**: The model dynamically adjusts asset weights based on multiple dimensions, providing a balanced risk-return profile[37] - **Test Results**: - July stock position: 18.72% - Bond position: 81.28% - July portfolio return: 0.84% - August stock position: 7.44% - Bond position: 92.56%[51] Model Backtest Results 1. **Duration Timing Model** - July return: 6.6bp - Benchmark return: -25.8bp - Strategy excess return: 32.4bp[21] 2. **Gold Timing Model** - Expected return for the next year: 22.4% - Past year absolute return: 39.77%[33][35] 3. **Active Risk Budget Model** - July stock position: 18.72% - Bond position: 81.28% - July portfolio return: 0.84% - August stock position: 7.44% - Bond position: 92.56%[51] Quantitative Factors and Construction Methods Factor: High-Frequency Macroeconomic Factors - **Construction Idea**: Use asset portfolio simulation to construct a high-frequency macro factor system to observe market macro expectations[12] - **Construction Process**: - Combine real macro indicators to form low-frequency macro factors - Select assets leading low-frequency macro factors - Use rolling multiple leading regression to determine asset weights and simulate macro factor trends[12] - **Evaluation**: High-frequency macro factors provide leading indicators for market expectations, offering valuable insights for asset allocation[12] Factor: Convertible Bond Valuation Factors - **Construction Idea**: Compare the relative valuation of convertible bonds and stocks, and between convertible bonds and credit bonds[25] - **Construction Process**: - Construct the "100-yuan conversion premium rate" to compare the valuation of convertible bonds and stocks - Use the "modified YTM - credit bond YTM" median to compare the valuation of debt-biased convertible bonds and credit bonds - Construct style rotation portfolios based on market sentiment indicators like 20-day momentum and volatility deviation[25][27] - **Evaluation**: The factors effectively capture the relative valuation and style characteristics of convertible bonds, aiding in portfolio construction[25][27] - **Test Results**: - "100-yuan conversion premium rate": 33.71% - "Modified YTM - credit bond YTM" median: -2.06% - Style rotation annualized return: 24.54% - Maximum drawdown: 15.89% - IR: 1.47 - Monthly win rate: 65.17% - 2025 return: 35.17%[26][29] Factor Backtest Results 1. **High-Frequency Macroeconomic Factors** - High-frequency economic growth: Upward trend - High-frequency consumer inflation: Downward trend - High-frequency producer inflation: Upward trend[17] 2. **Convertible Bond Valuation Factors** - "100-yuan conversion premium rate": 33.71% - "Modified YTM - credit bond YTM" median: -2.06% - Style rotation annualized return: 24.54% - Maximum drawdown: 15.89% - IR: 1.47 - Monthly win rate: 65.17% - 2025 return: 35.17%[26][29]