大类资产配置之流动性周期指数构建

Quantitative Models and Construction Methods - Model Name: Liquidity Cycle Index Construction Idea: Dynamically adapt to China's monetary policy evolution by integrating signals from policy tools, market rates, and transmission efficiency to capture long-term trends in monetary policy[2][3][4] Construction Process: 1. Policy Tools Dimension: - Four-layer signal processing framework: data cleaning → signal extraction → time-varying weight synthesis → turning point enhancement output[2] - Dynamic weight allocation: Price-based tools' weight increased from 20% to 48%, while quantity-based tools' weight decreased from 55% to 25%[76][83][84] - Noise suppression techniques: 6th-order Butterworth low-pass filter and Savitzky-Golay filter for high-frequency signal stability[76][77] - Formula: $Enhanced_t = 0.80 \cdot Signal_t + 0.10 \cdot ShortTermSignal_t + 0.08 \cdot MidTermSignal_t + 0.02 \cdot AccelerationSignal_t$ Parameters: Short-term (20 days), mid-term (40 days), acceleration signal (policy turning speed)[77] 2. Market Rates Dimension: - Weighted integration of money market (60%) and bond market (40%) signals[4][100] - Money market signals: DR007 (35%), R007 (30%), CD rates (20%), SHIBOR term spread (15%)[103][104] - Bond market signals: Yield curve analysis (10Y-1Y spread) and yield change analysis (30/60-day changes)[118][119] 3. Transmission Efficiency Dimension: - Metrics: M2-social financing growth gap, M2-M1 growth gap, financing cost (MLF-LPR, OMO-DR007 spreads), market segmentation (R007-DR007, interbank-exchange spreads), and financing pressure (real interest rate percentile)[5][123][125][137][142] - Dynamic weight adjustment based on policy stage (e.g., quantity-based tools dominate earlier periods, price-based tools dominate later)[124][130][136] - Evaluation: The model effectively captures monetary policy trends and provides high explanatory power for policy turning points[2][3][76] Model Backtesting Results - Liquidity Cycle Index: - Annualized return: 5.53% (monthly rebalancing), 4.88% (quarterly rebalancing)[161] - Sharpe ratio: 0.974 (monthly), 0.876 (quarterly)[161] - Maximum drawdown: 15.652% (monthly), 13.805% (quarterly)[161] - Win rate: 64.762% (monthly), 64.286% (quarterly)[161] Quantitative Factors and Construction Methods - Factor Name: Policy Tools Dimension Construction Idea: Reflect the evolution of China's monetary policy from quantity-based to price-based tools[2][76] Construction Process: - Dynamic weight allocation: Price-based tools (e.g., OMO, MLF) gradually dominate over quantity-based tools (e.g., reserve ratio)[76][83][84] - Signal enhancement: Use filters and acceleration metrics to highlight policy turning points[76][77] - Formula: $M_t = \alpha \cdot M_{t-1} + \beta \cdot \Delta R_t \times \Gamma(\Delta R_t, \Delta R_{t-1})$ Parameters: α=0.98 (memory decay), β=25 (impact coefficient), ΔR_t (policy change)[83][86] - Factor Name: Market Rates Dimension Construction Idea: Capture short-term and long-term liquidity changes through money and bond market signals[4][100] Construction Process: - Money market: Weighted integration of DR007, R007, CD rates, and SHIBOR term spread signals[103][104] - Bond market: Analyze yield curve dynamics and yield changes using rolling averages and standard deviations[118][119] - Factor Name: Transmission Efficiency Dimension Construction Idea: Quantify the effectiveness of monetary policy transmission to the real economy[5][123] Construction Process: - Metrics: M2-social financing growth gap, M2-M1 growth gap, financing cost spreads, market segmentation, and real interest rate percentile[125][130][137][142] - Dynamic weight adjustment based on policy stage[124][130][136] Factor Backtesting Results - Policy Tools Dimension: Enhanced signal improves turning point recognition compared to basic signals[78][79] - Market Rates Dimension: Money market signals show high sensitivity to liquidity changes; bond market signals effectively capture long-term trends[103][118] - Transmission Efficiency Dimension: Metrics like financing cost and market segmentation provide nuanced insights into policy effectiveness[123][137][142] Liquidity Cycle Index Application - Dynamic Asset Allocation Strategy: - Five liquidity cycle states: Strong easing (>0.25), mild easing ([0.05, 0.25]), neutral ([-0.15, 0.05]), mild tightening ([-0.35, -0.15]), strong tightening (<-0.35)[152][153] - Asset allocation weights: Stocks (0-30%), bonds (50-80%), commodities (0-15%), gold (5-20%) based on cycle state[154][155] - Performance: - Monthly rebalancing achieves annualized return of 5.53%, Sharpe ratio of 0.974, and maximum drawdown of 15.652%[161] - Strategy outperforms benchmark (40% equity, 60% fixed income) in all metrics[161]