Guosen Securities(SZ:002736) Guoxin Securities·2025-09-02 14:31Quantitative Models and Construction Methods - Model Name: ETF Institutional Holding Penetration Algorithm Construction Idea: The algorithm aims to accurately calculate the actual proportion of institutional investors holding ETF shares by correcting the overestimation caused by treating ETF-linked funds entirely as institutional holdings[1][13][14] Detailed Construction Process: 1. Institutional holdings are derived from two components: institutional portions within ETF-linked funds and direct institutional holdings of ETF shares[15][17] 2. Formula for corrected institutional holding proportion: P_{\rm inst}=P_{\rm inst}-P_{\rm inst}\ast\left(1-S_{\rm inst}\right)\tag{1} - : Published institutional holding proportion - : Institutional proportion within ETF-linked funds[17] 3. For missing data, reverse calculation is applied using the formula: P_{\mathrm{link}}=A_{\mathrm{link}}*P_{\mathrm{fund}}/A_{\mathrm{eff}}\tag{1} - P_{\mathrm{link}: Proportion of ETF-linked fund holdings - A_{\mathrm{link}: Net asset value of ETF-linked funds - P_{\mathrm{fund}: Published fund holding proportion - A_{\mathrm{eff}: Effective ETF net asset value[18] Evaluation: Provides a more accurate representation of institutional investor dynamics by correcting biases in reported data[13][17][19] Model Backtesting Results - ETF Institutional Holding Penetration Algorithm: - Institutional holding proportion for stock ETFs increased to 64.64% in 2025H1[21] - Institutional holding proportion for bond ETFs rose to 91.43% in 2025H1[22] - Institutional holding proportion for commodity ETFs slightly recovered to 19.06% in 2025H1[26] - Institutional holding proportion for cross-border ETFs climbed to 46.77% in 2025H1[23] Quantitative Factors and Construction Methods - Factor Name: Institutional Holding Proportion by Asset Class Construction Idea: Categorize ETFs by asset class and analyze institutional holding proportions[21][22][23] Detailed Construction Process: 1. Aggregate ETF data by asset class (e.g., stock, bond, commodity, cross-border)[21][22][23] 2. Calculate institutional and individual holding proportions using corrected data from the penetration algorithm[19][21][22] Evaluation: Highlights trends in institutional preferences across different asset classes[21][22][23] - Factor Name: Institutional Holding Proportion by Sector Construction Idea: Analyze institutional holding proportions within stock ETFs categorized by sectors (e.g., cyclical, financial, technology, consumer)[29][33][38] Detailed Construction Process: 1. Categorize stock ETFs into sectors based on the ETF classification system[29][33] 2. Calculate institutional holding proportions using corrected data from the penetration algorithm[19][33][38] Evaluation: Provides insights into sector-specific institutional investment trends[29][33][38] - Factor Name: Institutional Holding Proportion by Theme Construction Idea: Analyze institutional holding proportions within stock ETFs categorized by themes (e.g., 5G, chips, military, medicine, dividends, securities, new energy vehicles)[40][43][44] Detailed Construction Process: 1. Categorize stock ETFs into themes based on the ETF classification system[40][43] 2. Calculate institutional holding proportions using corrected data from the penetration algorithm[19][40][44] Evaluation: Identifies thematic investment preferences among institutional investors[40][43][44] - Factor Name: Institutional Holding Proportion by Broad-Based Index Construction Idea: Analyze institutional holding proportions within ETFs tracking broad-based indices (e.g., SSE 50, CSI 300, CSI 500, CSI 1000, A500, ChiNext, STAR Market)[51][53][56] Detailed Construction Process: 1. Categorize ETFs into broad-based indices based on the ETF classification system[51][53] 2. Calculate institutional holding proportions using corrected data from the penetration algorithm[19][51][56] Evaluation: Highlights institutional preferences for broad-based index ETFs[51][53][56] Factor Backtesting Results - Institutional Holding Proportion by Asset Class: - Stock ETFs: 64.64% in 2025H1[21] - Bond ETFs: 91.43% in 2025H1[22] - Commodity ETFs: 19.06% in 2025H1[26] - Cross-border ETFs: 46.77% in 2025H1[23] - Institutional Holding Proportion by Sector: - Cyclical: 37.95% in 2025H1[38] - Financial: 35.98% in 2025H1[38] - Consumer: 20.30% in 2025H1[38] - Technology: 25.56% in 2025H1[38] - Institutional Holding Proportion by Theme: - Military: 56.40% in 2025H1[47] - Securities: 23.57% in 2025H1[47] - Chips: 23.73% in 2025H1[48] - Dividends: 42.08% in 2025H1[44] - Institutional Holding Proportion by Broad-Based Index: - SSE 50: 86.76% in 2025H1[56] - CSI 300: 89.49% in 2025H1[56] - CSI 500: 77.98% in 2025H1[56] - CSI 1000: 94.27% in 2025H1[56] - A500: 63.41% in 2025H1[60] - ChiNext: 43.19% in 2025H1[60] - STAR Market: 34.87% in 2025H1[60]