Workflow
风格及行业观点月报:风格轮动模型持续得到验证,行业轮动两模型均推荐配置非银-20250605
GUOTAI HAITONG SECURITIES·2025-06-05 11:16

Quantitative Models and Construction 1. Model Name: Macro + Volume-Price Dual-Driver Large-Cap and Small-Cap Rotation Strategy - Model Construction Idea: This model integrates macroeconomic factors and micro-level volume-price factors to predict the rotation between large-cap and small-cap styles[6][7] - Model Construction Process: - The model uses multiple single-factor signals, including PMI seasonal average difference, social financing growth rate, monetary liquidity, US-China interest rate spread, macro adjustment momentum, and style crowding indicators[7] - Each factor is assigned a signal value: 1 for large-cap signals, -1 for small-cap signals, and 0 for no effective signal[7] - The comprehensive score is calculated by summing the signals of all factors. If the score > 0, the portfolio is fully allocated to the CSI 300 Index; if the score < 0, it is fully allocated to the CSI 1000 Index; if the score = 0, the portfolio is equally weighted between the two indices[7] - Model Evaluation: The model demonstrates a high backtest win rate of 82.22% as of Q1 2025, indicating strong predictive power[6] 2. Model Name: Macro + Volume-Price Dual-Driver Value-Growth Rotation Strategy - Model Construction Idea: This model integrates macroeconomic factors and micro-level volume-price factors to predict the rotation between value and growth styles[12][13] - Model Construction Process: - The model uses multiple single-factor signals, including PMI new orders seasonal average difference, PPI-CPI growth rate, 1-year government bond yield, 3-month US bond yield, macro adjustment momentum, style crowding indicators, and market sentiment[13] - Each factor is assigned a signal value: 1 for value signals, -1 for growth signals, and 0 for no effective signal[13] - The comprehensive score is calculated by summing the signals of all factors. If the score > 0, the portfolio is fully allocated to the CSI Value Index; if the score < 0, it is fully allocated to the CSI Growth Index; if the score = 0, the portfolio is equally weighted between the two indices[13] - Model Evaluation: The model demonstrates a backtest win rate of 77.78% as of Q1 2025, showcasing its effectiveness in predicting style rotations[12] 3. Model Name: Industry Rotation Model (Single-Factor Multi-Strategy and Composite Factor Strategy) - Model Construction Idea: This model evaluates industry rotation using factors from historical fundamentals, expected fundamentals, sentiment, volume-price technicals, and macroeconomics[18][19] - Model Construction Process: - Single-factor multi-strategy: Constructs portfolios based on individual factors and evaluates their performance[18] - Composite factor strategy: Combines multiple factors into a composite score to rank industries and construct portfolios[18] - Both strategies select the top 5 industries from the 30 first-level industries in the CITIC classification and construct equal-weighted long portfolios[18] - Model Evaluation: The single-factor multi-strategy outperformed the composite factor strategy in May 2025, with higher monthly absolute and excess returns[20] --- Backtest Results of Models 1. Macro + Volume-Price Dual-Driver Large-Cap and Small-Cap Rotation Strategy - YTD Return: -2.41%[11] - Annualized Return: -5.83%[11] - Annualized Volatility: 17.17%[11] - Maximum Drawdown: 10.49%[11] - Sharpe Ratio: -0.34[11] - Calmar Ratio: -0.56[11] 2. Macro + Volume-Price Dual-Driver Value-Growth Rotation Strategy - YTD Return: 1.79%[17] - Annualized Return: 4.48%[17] - Annualized Volatility: 18.06%[17] - Maximum Drawdown: 10.36%[17] - Sharpe Ratio: 0.25[17] - Calmar Ratio: 0.43[17] 3. Industry Rotation Model - Composite Factor Strategy: - Monthly Absolute Return: 2.43%[20] - Monthly Excess Return: -0.64%[20] - YTD Absolute Return: 4.81%[20] - YTD Excess Return: 3.98%[20] - Single-Factor Multi-Strategy: - Monthly Absolute Return: 3.31%[20] - Monthly Excess Return: 0.33%[20] - YTD Absolute Return: 4.56%[20] - YTD Excess Return: 3.83%[20]