红利风格投资价值跟踪（2025W13）：外部不确定性加剧，资金重新聚焦红利

Quantitative Models and Construction Methods 1. Model Name: Dividend Timing Model - Model Construction Idea: This model aims to predict the relative excess return of the CSI Dividend Index compared to the Wind All A Index, based on macroeconomic indicators such as global liquidity (10-year U.S. Treasury yield), domestic liquidity (M2 YoY), and domestic economic expectations (M1-M2 YoY spread) [8] - Model Construction Process: - The model uses historical data to establish the relationship between the above macroeconomic indicators and the relative excess return of the CSI Dividend Index - The timing strategy has achieved an annualized excess return of 8.25% since 2010, though it recorded -5.28% in 2025 YTD [8] - Model Evaluation: The model demonstrates strong predictive power for excess returns under certain macroeconomic conditions, but its performance can be volatile depending on market dynamics [8] 2. Model Name: Regression Model for PETTM - Model Construction Idea: This model evaluates the relationship between the CSI Dividend Index's PETTM valuation and its future returns, both in absolute and relative terms [18][19] - Model Construction Process: - Absolute PETTM: - The absolute PETTM valuation is calculated using a weighted factor adjustment to align with the dividend yield-weighted characteristics of the index - Historical data shows a correlation coefficient of -29.42% between the three-year PETTM percentile and future one-year absolute returns - Regression equation: $ y = -0.2791x + 0.2641 $ Here, $y$ represents the future one-year absolute return, and $x$ represents the three-year PETTM percentile [18][22] - Relative PETTM: - The relative PETTM is calculated as the ratio of the CSI Dividend Index's PETTM to the Wind All A Index's PETTM - Historical data shows a correlation coefficient of -31.87% between the three-year relative PETTM percentile and future one-year relative excess returns - Regression equation: $ y = -0.1149x + 0.0974 $ Here, $y$ represents the future one-year relative excess return, and $x$ represents the three-year relative PETTM percentile [19][28] - Model Evaluation: The model effectively identifies valuation extremes, providing insights into potential risks and opportunities based on historical valuation levels [18][19] 3. Model Name: Price and Volume Regression Model - Model Construction Idea: This model explores the relationship between price/volume metrics and future returns of the CSI Dividend Index [23][29] - Model Construction Process: - Price Dimension: - Calculates the weight of index constituents above their 120-day moving average - Historical data shows a correlation coefficient of -43.73% between this weight and future one-year absolute returns - Regression equation: $ y = -0.2387x + 0.2133 $ Here, $y$ represents the future one-year absolute return, and $x$ represents the weight of constituents above the 120-day moving average [23][25] - Volume Dimension: - Analyzes the percentile of absolute and relative trading volume - Absolute trading volume: Correlation coefficient of -39.66% with future one-year absolute returns - Regression equation: $ y = -0.3852x + 0.3443 $ Here, $y$ represents the future one-year absolute return, and $x$ represents the absolute trading volume percentile [29][34] - Relative trading volume: Correlation coefficient of -12.94% with future one-month relative excess returns - Regression equation: $ y = -0.0163x + 0.0092 $ Here, $y$ represents the future one-month relative excess return, and $x$ represents the relative trading volume percentile [30][34] - Model Evaluation: The model highlights the predictive power of price and volume metrics, particularly in identifying reversal risks at extreme levels [23][29] 4. Model Name: Dividend 50 Optimized Portfolio - Model Construction Idea: This portfolio combines high dividend yield stocks with a linear multi-factor model to enhance capital gains while maintaining a stable dividend style [44] - Model Construction Process: - Selects high dividend yield stocks as the base - Applies a linear multi-factor model to optimize capital gains - Incorporates Barra style factor constraints to ensure style consistency - Adjusts dividend style exposure based on the three-dimensional dividend timing model [44] - Model Evaluation: The portfolio demonstrates strong performance in both absolute and relative terms, with consistent excess returns over the CSI Dividend Index [44] --- Model Backtesting Results 1. Dividend Timing Model - Annualized excess return since 2010: 8.25% - 2025 YTD excess return: -5.28% [8] 2. Regression Model for PETTM - Absolute PETTM: - Current PETTM: 8.95x - Three-year percentile: 96.93% - Estimated future one-year absolute return: -0.65% [18][22] - Relative PETTM: - Current relative PETTM: 0.48x - Three-year percentile: 76.50% - Estimated future one-year relative excess return: 0.95% [19][28] 3. Price and Volume Regression Model - Price Dimension: - Weight above 120-day moving average: 58.44% - Estimated future one-year absolute return: 7.38% [23][25] - Volume Dimension: - Absolute trading volume percentile: 76.64% - Estimated future one-year absolute return: 4.91% [29][34] - Relative trading volume percentile: 6.54% - Estimated future one-month relative excess return: 0.78% [30][34] 4. Dividend 50 Optimized Portfolio - Performance Metrics: - 2022-2025 cumulative absolute return: 52.04% - 2022-2025 cumulative excess return: 22.83% - Recent one-year absolute return: 12.03% - Recent one-year excess return: 3.31% [45]