金融工程专题：融合股指贴水的四因子择时策略

Quantitative Models and Construction Methods 1. Model Name: Nonlinear Equity Allocation Signal Model - Model Construction Idea: The model integrates microstructure factors for independent prediction and combines macro and meso factors to divide the market into eight operational states, forming a nonlinear equity allocation signal framework [3] - Model Construction Process: 1. Microstructure factors are used for independent prediction 2. Macro and meso factors are domain-synthesized to classify the market into eight operational states 3. A nonlinear equity allocation signal is constructed based on these states [3] - Model Evaluation: This model provides a comprehensive framework by combining multiple dimensions of market signals [3] 2. Model Name: Four-Factor Timing Signal - Model Construction Idea: Combines three-dimensional timing signals (macro, meso, and micro) with stock index futures timing signals to address the limitations of delayed predictions at market turning points [5][61] - Model Construction Process: 1. Construct three-dimensional timing signals based on macroeconomic state, meso-level prosperity index, and microstructure risk 2. Develop stock index futures timing signals using the correlation between futures basis and index trends 3. Synthesize the two signals to form the four-factor timing signal [5][61] - Model Evaluation: The synthesized signal effectively complements the shortcomings of three-dimensional signals in predicting market inflection points [5][61] 3. Model Name: Stock Index Futures Timing Signal - Model Construction Idea: Utilizes the correlation between stock index futures basis and index trends to reflect market sentiment and construct daily timing signals [4][52] - Model Construction Process: 1. Define annualized basis rate: $ \text{Annualized Basis Rate} = \frac{\text{Next Month Futures - Current Month Futures}}{\text{Spot Price}} \times 12 $ 2. Use the CSI 500 basis time series to reflect market-wide changes 3. Construct a high-frequency basis indicator to express panic sentiment 4. Perform IC matrix testing and calculate stratified returns based on normalized signal values [52][53][55] - Model Evaluation: The signal captures market sentiment effectively and provides a quantitative basis for timing decisions [52][53] --- Model Backtesting Results 1. Nonlinear Equity Allocation Signal Model - Excess Return: Not explicitly provided in the report 2. Four-Factor Timing Signal - Excess Return: - CSI 300: Cumulative excess return of 73.61%, annualized excess return of 9.86% [62] - CSI 500: Cumulative excess return of 69.90%, annualized excess return of 9.36% [63] - CSI 1000: Cumulative excess return of 61.61%, annualized excess return of 8.25% [65] - CSI All Share: Cumulative excess return of 65.00%, annualized excess return of 8.70% [69] 3. Stock Index Futures Timing Signal - IC Matrix Testing: The stratified returns distribution is relatively linear when holding for 20 trading days [56] --- Quantitative Factors and Construction Methods 1. Factor Name: Macro State Factor - Factor Construction Idea: Reflects the favorability of the macro environment on asset performance using logistic regression [20][21] - Factor Construction Process: 1. Stabilize macro variable sequences by lagging one period and calculating DIF values (12-period EMA minus 26-period EMA) 2. Define multi-long/short states based on the 60-day future return of A-shares 3. Perform rolling logistic regression using historical data ([t-1000, t-60]) to predict the favorability of the macro environment [20][21] - Factor Evaluation: High logical correlation with future index trends but limited predictive power for trend intensity [24] 2. Factor Name: Meso-Level Prosperity Index - Factor Construction Idea: Uses high-frequency meso-level data to predict A-share profitability trends [26][27] - Factor Construction Process: 1. Select high-frequency meso-level variables (e.g., industrial output, profits, economic activity) 2. Use PCA to synthesize major factors and regress against the net profit growth rate of the Shanghai Composite Index 3. Increase frequency to daily and predict using the trained model [26][27] - Factor Evaluation: Demonstrates strong predictive power for A-share profitability trends and leading indicators [27] 3. Factor Name: Microstructure Risk Factor - Factor Construction Idea: Measures valuation, risk premium, volatility, and liquidity to assess market risk [31][32][34][36] - Factor Construction Process: 1. Valuation: Average percentile of PE and PB over the past five years 2. Risk Premium: Percentile of ERP over the past five years 3. Volatility: Average percentile of rolling 20-day volatility over the past five years 4. Liquidity: Percentile of free-float turnover rate over the past five years 5. Combine the above metrics equally to form a left-side prediction indicator [31][32][34][36][38] - Factor Evaluation: Provides a comprehensive measure of market risk, with current A-share structural risk at a low level (0.27) [38] --- Factor Backtesting Results 1. Macro State Factor - Correlation with Index Trends: - CSI All Share: 0.9628 (up/down), -0.0132 (trend intensity) [25] - CSI 300: 0.9659 (up/down), -0.1271 (trend intensity) [25] - CSI 500: 0.9765 (up/down), 0.1211 (trend intensity) [25] - CSI 1000: 0.9677 (up/down), 0.0396 (trend intensity) [25] 2. Meso-Level Prosperity Index - Predictive Performance: - Leading A-share profitability trends with significant accuracy [27] 3. Microstructure Risk Factor - Current Risk Levels (as of June 28, 2024): - CSI All Share: Valuation (0.122), Risk Premium (0.034), Volatility (0.427), Liquidity (0.105), Structural Risk (0.353) [39] - CSI 300: Valuation (0.196), Risk Premium (0.151), Volatility (0.078), Liquidity (0.263), Structural Risk (0.294) [39] - CSI 500: Valuation (0.182), Risk Premium (0.102), Volatility (0.479), Liquidity (0.062), Structural Risk (0.372) [39] - CSI 1000: Valuation (0.147), Risk Premium (0.050), Volatility (0.638), Liquidity (0.055), Structural Risk (0.396) [39]