ETF资金流向视角下的行业轮动配置

Quantitative Models and Construction Methods 1. Model Name: Industry Allocation Model Based on ETF Fund Flows - Model Construction Idea: The model leverages ETF fund flow data to identify industry rotation opportunities. It incorporates short-term fund inflows/outflows, style-adjusted holding levels, marginal changes in holdings, and the divergence between ETF and active equity fund holdings to construct an industry allocation strategy[3][69][72] - Model Construction Process: 1. Short-term Fund Flows: Calculate the first-order difference of weekly ETF holdings to identify industries with significant inflows or outflows[40][44] 2. Style-Adjusted Holdings: Adjust industry holdings based on market style (e.g., large-cap vs. small-cap, growth vs. value) using a single-sided HP filter and factor momentum to determine style trends[49][50][57] 3. Marginal Changes in Holdings: Analyze the marginal changes in ETF holdings by ranking industries into five groups based on their monthly holding changes[22][25] 4. Divergence with Active Equity Funds: Compare ETF holdings with active equity fund holdings to identify industries with higher or lower relative allocations. Use regression-based methods to estimate active fund holdings when real data is unavailable[27][28][31] 5. Final Strategy: Combine the above factors equally, select the top six industries, and rebalance the portfolio bi-weekly[72] - Model Evaluation: The model effectively captures industry rotation opportunities by integrating multiple dimensions of ETF fund flow data and market style trends[72] --- Model Backtesting Results 1. Industry Allocation Model Based on ETF Fund Flows - Annualized Return: 15.57% - Excess Annualized Return: 7.56% (compared to equal-weighted industry benchmark) - Information Ratio (IR): 0.93 - Maximum Drawdown: 8.30% - Monthly Excess Win Rate: 64% - Payoff Ratio: 1.38x[72] --- Quantitative Factors and Construction Methods 1. Factor Name: Short-term Fund Flows - Factor Construction Idea: Identify industries with significant short-term fund inflows or outflows to capture immediate price impacts[40][44] - Factor Construction Process: 1. Calculate the first-order difference of weekly ETF holdings 2. Rank industries based on the magnitude of fund flow changes[40][44] - Factor Evaluation: Demonstrates strong monotonicity in short-term returns, with industries experiencing inflows showing higher returns[44] 2. Factor Name: Style-Adjusted Holdings - Factor Construction Idea: Adjust industry holdings based on prevailing market styles (e.g., large-cap vs. small-cap, growth vs. value)[46][49] - Factor Construction Process: 1. Use a single-sided HP filter to smooth market style data (e.g., CSI 300/CSI 1000 index ratios) 2. Define factor momentum as the difference between the current value and the average of the previous two periods 3. Classify industries into five groups based on their adjusted holdings[49][50][57] - Factor Evaluation: Captures the relationship between industry holdings and market style trends, effectively identifying style-driven opportunities[47][57] 3. Factor Name: Marginal Changes in Holdings - Factor Construction Idea: Analyze the marginal changes in ETF holdings to identify industries with increasing or decreasing allocations[22][25] - Factor Construction Process: 1. Calculate the monthly difference in ETF holdings for each industry 2. Rank industries into five groups based on the magnitude of changes[22][25] - Factor Evaluation: Demonstrates a strong correlation with growth and value style trends, providing insights into industry rotation opportunities[47] 4. Factor Name: Divergence with Active Equity Fund Holdings - Factor Construction Idea: Compare ETF holdings with active equity fund holdings to identify industries with higher or lower relative allocations[27][28] - Factor Construction Process: 1. Use regression-based methods to estimate active fund holdings when real data is unavailable 2. Calculate the difference between ETF and active fund holdings and rank industries into three groups based on the magnitude of divergence[27][28][31] - Factor Evaluation: Highlights the pricing power of ETF flows relative to active funds, especially post-2021[31][65] --- Factor Backtesting Results 1. Short-term Fund Flows - Absolute Return: 6.17% (highest group) - Annualized Volatility: 21.22% - Sharpe Ratio: 0.29 - Maximum Drawdown: -37.61%[42] 2. Style-Adjusted Holdings - Annualized Return: 9.66% - Excess Annualized Return: 5.82% - Information Ratio (IR): 0.75 - Maximum Drawdown: -29.11%[55] 3. Marginal Changes in Holdings - Annualized Return: 7.80% (highest group) - Excess Annualized Return: 6.91% - Information Ratio (IR): 1.13 - Maximum Drawdown: -16.10%[71] 4. Divergence with Active Equity Fund Holdings - Annualized Return: 14.01% - Excess Annualized Return: 6.11% - Information Ratio (IR): 0.76 - Maximum Drawdown: -28.80%[64][65]