Quantitative Models and Construction Methods - **Model Name**: Dynamic Macro Event Factor-based CSI Hong Kong Stock Connect Consumer Index Timing Strategy **Model Construction Idea**: The model explores the impact of China's macroeconomic factors on the overall performance and trends of Hong Kong-listed consumer companies, using dynamic macro event factors to construct a timing strategy framework [2][3][20] **Model Construction Process**: 1. **Macro Data Selection**: Select 20+ macroeconomic indicators across four dimensions: economy, inflation, currency, and credit, such as PMI, PPI, M1, etc [21][23] 2. **Data Preprocessing**: - Align data frequency to monthly frequency by either taking the last trading day of the month or calculating the monthly average for daily data - Fill missing values using the median of the first-order difference of the past 12 months added to the previous value $ X_{t}=X_{t-1}+Median_{diff12} $ [27] - Apply filtering using one-sided HP filter to avoid future data leakage $ \hat{t}_{t|t,\lambda}=\sum\nolimits_{s=1}^{t}\omega_{t|t,s,\lambda}\cdot y_{s}=W_{t|t,\lambda}(L)\cdot y_{t} $ [28] - Derive factors using transformations such as year-on-year, month-on-month, and moving averages [29] 3. **Macro Event Factor Construction**: - Determine event breakthrough direction by calculating the correlation between data and next-period asset returns - Identify leading or lagging relationships by deriving lagged event factors (0-4 periods) and selecting the most suitable lag period - Generate event factors using three types: data breaking through moving average, data breaking through median, and data moving in the same direction, with different parameters (e.g., moving average length: 2-12, rolling window: 2-12, same direction period: 1-5) [30][32] 4. **Event Factor Evaluation and Screening**: - Use two metrics: win rate of returns and volatility-adjusted returns during opening positions - Initial screening criteria: t-test significance at 95% confidence level, win rate >55%, occurrence frequency > rolling window period/6 [31][32] 5. **Combining Event Factors**: Select the highest win rate event factor as the base factor, then combine it with the second-highest win rate factor with a correlation <0.85. If the combined factor improves the win rate, it is selected; otherwise, the base factor is used [33] 6. **Dynamic Exclusion**: If no event factor passes the screening, the macro indicator is marked as empty for the period and excluded from scoring [33] 7. **Optimal Rolling Window Determination**: Test rolling windows of 48, 60, 72, 84, and 96 months to find the most suitable parameter for each macro indicator based on volatility-adjusted returns during opening positions [33] 8. **Final Macro Indicators**: Five macro factors were selected based on their performance in the sample period: - PMI: Raw Material Prices (96-month rolling window) - US-China 10Y Bond Spread (72-month rolling window) - Financial Institutions: Medium-Long Term Loan Balance: Monthly New Additions: Rolling 12M Sum: YoY (48-month rolling window) - M1: YoY (48-month rolling window) - New Social Financing: Rolling 12M Sum: YoY (96-month rolling window) [34][35] 9. **Timing Strategy Construction**: - If >2/3 of factors signal bullishness, the category factor signal is marked as 1 - If <1/3 of factors signal bullishness, the category factor signal is marked as 0 - If the proportion of bullish signals falls between these ranges, the category factor is marked with the specific proportion - The score of each category factor is used as the timing position signal for the period [3][35] **Model Evaluation**: The strategy effectively captures systematic opportunities and avoids systematic risks, demonstrating superior performance compared to the benchmark in terms of annualized returns, maximum drawdown, Sharpe ratio, and return-drawdown ratio [2][3][20] --- Model Backtesting Results - **Dynamic Macro Event Factor-based CSI Hong Kong Stock Connect Consumer Index Timing Strategy** - **Annualized Return**: 10.44% - **Annualized Volatility**: 18.47% - **Maximum Drawdown**: -29.72% - **Sharpe Ratio**: 0.59 - **Return-Drawdown Ratio**: 0.35 [2][11][22] --- Quantitative Factors and Construction Methods - **Factor Name**: PMI: Raw Material Prices **Factor Construction Idea**: Use raw data to capture macroeconomic trends affecting asset returns [35] **Factor Construction Process**: Utilize raw data with a 96-month rolling window [35] - **Factor Name**: US-China 10Y Bond Spread **Factor Construction Idea**: Reflect the impact of interest rate differentials on asset returns [35] **Factor Construction Process**: Utilize raw data with a 72-month rolling window [35] - **Factor Name**: Financial Institutions: Medium-Long Term Loan Balance: Monthly New Additions: Rolling 12M Sum: YoY **Factor Construction Idea**: Measure credit expansion and its influence on asset returns [35] **Factor Construction Process**: Utilize raw data with a 48-month rolling window [35] - **Factor Name**: M1: YoY **Factor Construction Idea**: Capture monetary supply changes and their impact on asset returns [35] **Factor Construction Process**: Utilize raw data with a 48-month rolling window [35] - **Factor Name**: New Social Financing: Rolling 12M Sum: YoY **Factor Construction Idea**: Reflect credit growth and its effect on asset returns [35] **Factor Construction Process**: Utilize raw data with a 96-month rolling window [35] **Factor Evaluation**: The selected factors demonstrated strong performance in the sample period, with high win rates and volatility-adjusted returns during opening positions [34][35] --- Factor Backtesting Results - **PMI: Raw Material Prices** - **Rolling Window**: 96 months [35] - **US-China 10Y Bond Spread** - **Rolling Window**: 72 months [35] - **Financial Institutions: Medium-Long Term Loan Balance: Monthly New Additions: Rolling 12M Sum: YoY** - **Rolling Window**: 48 months [35] - **M1: YoY** - **Rolling Window**: 48 months [35] - **New Social Financing: Rolling 12M Sum: YoY** - **Rolling Window**: 96 months [35]