多因子选股周报：成长因子表现出色，四大指增组合本周均跑赢基准-20250719

Quantitative Models and Factor Construction Quantitative Models and Construction Methods - Model Name: Maximized Factor Exposure (MFE) Portfolio Model Construction Idea: The MFE portfolio is designed to test the effectiveness of individual factors under realistic constraints, such as industry exposure, style exposure, stock weight limits, and turnover constraints. This approach ensures that the factors deemed "effective" can genuinely contribute to return prediction in the final portfolio[41][42]. Model Construction Process: The MFE portfolio is constructed using the following optimization model: $ \begin{array}{ll} max & f^{T} w \ s.t. & s_{l} \leq X(w-w_{b}) \leq s_{h} \ & h_{l} \leq H(w-w_{b}) \leq h_{h} \ & w_{l} \leq w-w_{b} \leq w_{h} \ & b_{l} \leq B_{b}w \leq b_{h} \ & \mathbf{0} \leq w \leq l \ & \mathbf{1}^{T} w = 1 \end{array} $ - Objective Function: Maximize single-factor exposure, where $f^{T} w$ represents the weighted exposure of the portfolio to the factor $f$, and $w$ is the stock weight vector. - Constraints: 1. Style Exposure: $X$ represents the factor exposure matrix for stocks, $w_b$ is the benchmark weight vector, and $s_l, s_h$ are the lower and upper bounds for style factor exposure[42]. 2. Industry Exposure: $H$ is the industry exposure matrix, and $h_l, h_h$ are the lower and upper bounds for industry deviations[42]. 3. Stock Weight Deviation: $w_l, w_h$ are the lower and upper bounds for stock weight deviations relative to the benchmark[42]. 4. Constituent Weight: $B_b$ is a binary vector indicating whether a stock is part of the benchmark, and $b_l, b_h$ are the lower and upper bounds for constituent weights[42]. 5. No Short Selling: Ensures non-negative weights and limits individual stock weights to $l$[42]. 6. Full Investment: Ensures the portfolio is fully invested with $\mathbf{1}^{T} w = 1$[43]. - Implementation: 1. Define constraints for style, industry, and stock weights. For example, for CSI 500 and CSI 300 indices, industry exposure is neutralized, and stock weight deviations are capped at 1%[45]. 2. Construct the MFE portfolio at the end of each month based on the constraints[45]. 3. Backtest the portfolio, accounting for transaction costs (0.3% per side), and calculate performance metrics relative to the benchmark[45]. Model Evaluation: The MFE portfolio effectively tests factor performance under realistic constraints, making it a robust tool for evaluating factor predictability in practical scenarios[41][42]. Quantitative Factors and Construction Methods - Factor Name: DELTAROA Factor Construction Idea: Measures the change in return on assets (ROA) compared to the same quarter in the previous year, capturing improvements in asset utilization efficiency[16]. Factor Construction Process: $ DELTAROA = ROA_{current\ quarter} - ROA_{same\ quarter\ last\ year} $ Where $ROA = \frac{Net\ Income}{Total\ Assets}$[16]. Factor Evaluation: DELTAROA is a growth-oriented factor that has shown strong performance in multiple sample spaces, particularly in the CSI A500 index[19][25]. - Factor Name: Standardized Unexpected Earnings (SUE) Factor Construction Idea: Measures the deviation of actual earnings from expected earnings, standardized by the standard deviation of expected earnings, to capture earnings surprises[16]. Factor Construction Process: $ SUE = \frac{Actual\ Earnings - Expected\ Earnings}{Standard\ Deviation\ of\ Expected\ Earnings} $[16]. Factor Evaluation: SUE is a profitability factor that performs well in growth-oriented indices like CSI 1000 and CSI A500[19][23][25]. - Factor Name: One-Year Momentum Factor Construction Idea: Captures the trend-following behavior of stocks by measuring price momentum over the past year, excluding the most recent month[16]. Factor Construction Process: $ Momentum = \frac{Price_{t-12} - Price_{t-1}}{Price_{t-1}} $ Where $t-12$ and $t-1$ represent the stock price 12 months and 1 month ago, respectively[16]. Factor Evaluation: Momentum is a widely used factor that has shown consistent performance in large-cap indices like CSI 300 and CSI 500[19][21]. Factor Backtesting Results - CSI 300 Sample Space: - Best-Performing Factors (1 Week): Single-quarter revenue growth, DELTAROA, single-quarter ROE[19]. - Worst-Performing Factors (1 Week): Three-month volatility, one-month volatility, three-month turnover[19]. - CSI 500 Sample Space: - Best-Performing Factors (1 Week): One-year momentum, standardized unexpected revenue, standardized unexpected earnings[21]. - Worst-Performing Factors (1 Week): SPTTM, single-quarter SP, dividend yield[21]. - CSI 1000 Sample Space: - Best-Performing Factors (1 Week): Three-month reversal, standardized unexpected revenue, single-quarter surprise magnitude[23]. - Worst-Performing Factors (1 Week): Dividend yield, one-month volatility, BP[23]. - CSI A500 Sample Space: - Best-Performing Factors (1 Week): DELTAROA, standardized unexpected earnings, single-quarter ROA[25]. - Worst-Performing Factors (1 Week): Three-month volatility, one-month turnover, one-month volatility[25]. - Public Fund Heavyweight Index Sample Space: - Best-Performing Factors (1 Week): One-year momentum, standardized unexpected revenue, expected net profit QoQ[27]. - Worst-Performing Factors (1 Week): Dividend yield, one-month volatility, three-month volatility[27].