CMS·2026-01-18 14:36Quantitative Models and Construction Methods 1. Model Name: Deposit Migration Signal - Model Construction Idea: The model is designed to identify the phenomenon of "deposit migration," where household deposits decrease while non-bank deposits increase, using monthly deposit data[5][13][14] - Model Construction Process: 1. Calculate the proportion of newly added household deposits and non-bank deposits to total deposits for a given month: - $ \text{Household Deposit Share} = \frac{\text{Cumulative New Household Deposits (12 months)}}{\text{Cumulative New Total Deposits (12 months)}} $ - $ \text{Non-Bank Deposit Share} = \frac{\text{Cumulative New Non-Bank Deposits (12 months)}}{\text{Cumulative New Total Deposits (12 months)}} $ 2. Compare the current month's values with the average of the previous three months: - If the current month's value is greater than the average, it is considered "rising"; otherwise, it is "falling"[13][14] 3. Define the "deposit migration" signal as a scenario where the household deposit share decreases while the non-bank deposit share increases[14] 4. Evaluate the performance of the signal by analyzing the average return and win rate of the All-A Index one month after the signal is triggered[14][15] - Model Evaluation: The model effectively identifies periods of increased equity market activity driven by deposit migration, but its effectiveness may diminish when funds migrate to non-equity assets, such as bonds[14] 2. Model Name: Short-Term Timing Strategy - Model Construction Idea: This model integrates macroeconomic, valuation, sentiment, and liquidity indicators to generate weekly timing signals for the equity market[19][20] - Model Construction Process: 1. Macroeconomic Indicators: - Manufacturing PMI > 50 indicates economic expansion, providing a positive signal - Credit impulse and M1 growth rates are compared to historical percentiles to assess economic strength[19][22] 2. Valuation Indicators: - PE and PB ratios are compared to their historical percentiles; high percentiles indicate overvaluation, providing cautious signals[19][22] 3. Sentiment Indicators: - Beta dispersion, volume sentiment scores, and market volatility are analyzed to gauge market sentiment[20][22] 4. Liquidity Indicators: - Money market rates, exchange rate expectations, and leverage financing trends are used to assess liquidity conditions[20][22] 5. Combine the signals from the above indicators to generate an overall timing signal (optimistic, neutral, or cautious)[19][20] - Model Evaluation: The model demonstrates strong performance, with an annualized return of 16.65% and a maximum drawdown of 15.05%, significantly outperforming the benchmark strategy[21][24] 3. Model Name: Growth-Value Style Rotation Model - Model Construction Idea: This model identifies periods to overweight growth or value styles based on macroeconomic, valuation, and sentiment factors[28][29] - Model Construction Process: 1. Macroeconomic Factors: - Growth is favored when the earnings cycle slope is steep and the credit cycle is strengthening - Value is favored when the interest rate cycle is high[28][30] 2. Valuation Factors: - Growth is favored when the PE valuation gap is narrowing - Value is favored when the PB valuation gap is widening[28][30] 3. Sentiment Factors: - Growth is favored when turnover and volatility differences between growth and value are high[29][30] 4. Combine the signals from the above factors to determine the allocation between growth and value styles[28][30] - Model Evaluation: The model achieves an annualized return of 13.30% with a maximum drawdown of 43.07%, outperforming the benchmark strategy[29][31] 4. Model Name: Small-Cap vs. Large-Cap Rotation Model - Model Construction Idea: This model uses 11 effective rotation indicators to determine the relative attractiveness of small-cap and large-cap stocks[32] - Model Construction Process: 1. Key indicators include financing purchase balance changes, thematic investment sentiment, PB dispersion, and trading volume of small-cap indices[32][34] 2. Signals are aggregated to generate a composite rotation signal, which determines the allocation between small-cap and large-cap stocks[32][34] - Model Evaluation: The model consistently generates positive annual excess returns, with a 2026 year-to-date excess return of 2.88%[33][34] --- Model Backtesting Results 1. Deposit Migration Signal - Average return of All-A Index one month after signal: 1.72% - Win rate: 64.9% - Median return: 1.88%[15] 2. Short-Term Timing Strategy - Annualized return: 16.65% - Annualized volatility: 14.80% - Maximum drawdown: 15.05% - Sharpe ratio: 0.9802 - Monthly win rate: 66.46%[21][24] 3. Growth-Value Style Rotation Model - Annualized return: 13.30% - Annualized volatility: 20.76% - Maximum drawdown: 43.07% - Sharpe ratio: 0.6098 - Monthly win rate: 58.60%[29][31] 4. Small-Cap vs. Large-Cap Rotation Model - Annualized return: 20.60% (composite signal) - Annualized excess return: 12.95% - Maximum drawdown: 40.70% - Monthly win rate: 50.11%[34][36]