Quantitative Models and Construction Methods - **Model Name**: Fund Position Estimation Model **Model Construction Idea**: The model aims to estimate the complete stock holdings of public funds based on disclosed top 10 heavy holdings and industry allocation information, addressing the time lag in fund reports[2][8][9] **Model Construction Process**: 1. Use quarterly fund reports to obtain the top 10 heavy holdings and industry allocation data[9] 2. Combine this with the top 10 shareholders' information from listed companies[8] 3. Assume that non-heavy holdings remain unchanged in terms of stock types across two periods, with only proportional weight adjustments[10] 4. Scale the weights of non-heavy holdings from the previous period (T-1) to simulate the current period (T) non-heavy holdings[10] **Model Evaluation**: The model provides a practical approach to estimate fund positions with limited disclosed data, though it relies on assumptions about non-heavy holdings' stability[8][10] - **Model Name**: Fund Style Monitoring Model **Model Construction Idea**: This model monitors the daily style exposure of public funds by decomposing their stock holdings into large-cap and small-cap components[3][11] **Model Construction Process**: 1. Use simulated fund holdings to estimate stock positions[11] 2. Decompose stock holdings into large-cap and small-cap components[3] 3. Apply weighted least squares regression using fund daily returns and large-cap/small-cap index daily returns to estimate fund allocations to large-cap and small-cap portfolios[11] 4. Monitor fund positions daily and analyze style changes across dimensions such as size exposure and industry allocation[3][11] **Model Evaluation**: The model effectively captures dynamic changes in fund styles, providing insights into size and industry exposures[3][11] Model Backtesting Results - **Fund Position Estimation Model**: - Median stock position for equity funds: 87.39% as of 2025-05-30, up 0.10% from two weeks prior[4][16] - Median stock position for hybrid funds: 83.76% as of 2025-05-30, down 0.18% from two weeks prior[4][16] - Large-cap allocation changes: +2.48% for equity funds, +1.34% for hybrid funds[4][16] - Small-cap allocation changes: -2.37% for equity funds, -1.52% for hybrid funds[4][16] - **Fund Style Monitoring Model**: - Equity funds' large-cap allocation: 44.2%, small-cap allocation: 43.2% as of 2025-05-30[15] - Hybrid funds' large-cap allocation: 36.8%, small-cap allocation: 47.0% as of 2025-05-30[15] - Current position percentiles (2016 onwards): 34.43% for equity funds, 35.04% for hybrid funds[16] Quantitative Factors and Construction Methods - **Factor Name**: Size Allocation Factor **Factor Construction Idea**: This factor quantifies the allocation of fund holdings to large-cap and small-cap stocks[3][11] **Factor Construction Process**: 1. Decompose fund holdings into large-cap and small-cap components based on simulated positions[11] 2. Use weighted least squares regression to estimate allocation weights for large-cap and small-cap portfolios[11] **Factor Evaluation**: The factor provides a clear measure of size exposure, aiding in understanding fund style dynamics[3][11] - **Factor Name**: Industry Allocation Factor **Factor Construction Idea**: This factor measures the weight of fund holdings in different industries based on simulated positions[3][24] **Factor Construction Process**: 1. Classify stocks into industries using the CITIC Level-1 industry classification[24] 2. Calculate the weight of each industry in the fund's simulated holdings[24] **Factor Evaluation**: The factor effectively captures industry preferences and shifts in fund allocations[24] Factor Backtesting Results - **Size Allocation Factor**: - Equity funds: Large-cap 44.2%, Small-cap 43.2% as of 2025-05-30[15] - Hybrid funds: Large-cap 36.8%, Small-cap 47.0% as of 2025-05-30[15] - **Industry Allocation Factor**: - Top industries for equity funds: Electronics (14.32%), Medicine (10.99%), Electrical Equipment (9.29%), Machinery (6.51%), Food & Beverage (8.46%)[20] - Top industries for hybrid funds: Electronics (16.46%), Medicine (9.29%), Electrical Equipment (8.36%), Machinery (8.00%), Food & Beverage (6.10%)[20]

金融工程 | 金工定期报告 金融工程 证券研究报告 2025 年 05 月 05 日 基金风格配置监控周报： 上周偏股混合型基金显著下调小盘股票仓位 公募基金风格监控 基金季报中披露有基金前十大重仓股，同时企业季度报告中披露有企业的 前十大股东信息（主要使用上市公司 1、3 季度十大股东信息），结合这两 部分信息我们可以得到基金持仓更加详细的重仓股信息。因此模拟全部股 票持仓只需要估计基金剩余的非重仓持股明细。 基金季报（T 期）披露有股票持仓的证监会行业配置信息，基金最近（T-1 期）半年报、年报会披露全部股票持仓，可得基金半年报、年报末非重仓 股持仓信息。假设基金非重仓股持仓在两期各行业下持股种类不变，仅持 股权重按比例变化，那么根据 T-1 期非重仓股持仓按比例放缩权重可得到 T 期非重仓股模拟持仓。 在每年 1、4、8、10 月末我们能得到公募基金模拟持仓（其中 8 月末为基 金半年报真实持仓），基于模拟持仓我们首先估计公募基金股票仓位，进而 刻画公募基金的风格表现。基于公募基金模拟持仓，本文将基金股票持仓 拆分成大、小盘两部分，利用基金日收益率以及基金大、小盘指数日收益 率，本文用加权最小二乘法估计 ...