金融工程定期：1月转债配置：转债估值偏贵，看好偏股低估风格

Quantitative Models and Construction Methods 1. Model Name: "百元转股溢价率" (Premium Rate per 100 Yuan Conversion) - Model Construction Idea: This model compares the valuation of convertible bonds and their underlying stocks by calculating a time-series comparable valuation metric, "百元转股溢价率" (Premium Rate per 100 Yuan Conversion), and evaluates the relative allocation value using rolling historical percentiles[3][14] - Model Construction Process: - Fit the relationship curve between the conversion premium rate and conversion value in the cross-sectional space at each time point - Substitute a conversion value of 100 into the fitted formula to obtain the "百元转股溢价率" - Formula: $y_{i} = \alpha_{0} + \alpha_{1} \cdot \frac{1}{x_{i}} + \epsilon_{i}$ where $y_{i}$ is the conversion premium rate of the $i$-th bond, $x_{i}$ is the conversion value of the $i$-th bond, and $\epsilon_{i}$ is the error term[46][47] - Model Evaluation: The rolling three-year and five-year percentiles of this metric are at 99.30% and 99.60%, respectively, indicating that convertible bonds are relatively expensive compared to their underlying stocks[3][14] 2. Model Name: "修正 YTM – 信用债 YTM" (Adjusted YTM Minus Credit Bond YTM) - Model Construction Idea: This model evaluates the relative allocation value between debt-heavy convertible bonds and credit bonds by isolating the impact of conversion terms on the convertible bond's yield-to-maturity (YTM)[4][14] - Model Construction Process: - Adjust the YTM of debt-heavy convertible bonds using the following formula: $\text{Adjusted YTM} = \text{Convertible Bond YTM} \times (1 - \text{Conversion Probability}) + \text{Expected Annualized Return from Conversion} \times \text{Conversion Probability}$ - The conversion probability is calculated using the Black-Scholes (BS) model, incorporating stock price, strike price, stock volatility, remaining term, and discount rate - The difference between the adjusted YTM and the YTM of credit bonds of the same rating and maturity is calculated for each bond, and the median value is taken as the metric: $\text{"修正 YTM – 信用债 YTM" Median} = \text{median}\{X_1, X_2, ..., X_n\}$ where $X_i$ represents the difference for the $i$-th bond[48] - Model Evaluation: The current median value of this metric is -5.00%, indicating that the overall allocation cost-effectiveness of debt-heavy convertible bonds is relatively low[4][14] --- Model Backtesting Results 1. "百元转股溢价率" Model - Rolling three-year percentile: 99.30%[3][14] - Rolling five-year percentile: 99.60%[3][14] 2. "修正 YTM – 信用债 YTM" Model - Median value: -5.00%[4][14] --- Quantitative Factors and Construction Methods 1. Factor Name: 转股溢价率偏离度 (Conversion Premium Deviation) - Factor Construction Idea: Measures the deviation of the conversion premium rate from its fitted value, enabling comparability across different parities[20] - Factor Construction Process: $\text{Conversion Premium Deviation} = \text{Conversion Premium Rate} - \text{Fitted Conversion Premium Rate}$ The fitted value is determined by the relationship curve between conversion premium rate and conversion value, as described in the "百元转股溢价率" model[20][46] - Factor Evaluation: The quality of the fit depends on the number of convertible bonds, and this factor is effective in identifying valuation deviations[20] 2. Factor Name: 理论价值偏离度 (Theoretical Value Deviation, Monte Carlo Model) - Factor Construction Idea: Measures the price expectation difference by comparing the closing price of a convertible bond to its theoretical value, which is calculated using Monte Carlo simulation[20] - Factor Construction Process: $\text{Theoretical Value Deviation} = \frac{\text{Convertible Bond Closing Price}}{\text{Theoretical Value}} - 1$ The theoretical value is derived by simulating 10,000 paths for each time point, considering conversion, redemption, downward revision, and resale terms, and using the discount rate of bonds with the same credit rating and maturity[20] - Factor Evaluation: This factor fully accounts for the complex terms of convertible bonds and is particularly effective in identifying valuation discrepancies[20] 3. Factor Name: 转债综合估值因子 (Comprehensive Convertible Bond Valuation Factor) - Factor Construction Idea: Combines the rankings of the above two factors to create a comprehensive valuation metric for convertible bonds[20] - Factor Construction Process: $\text{Comprehensive Convertible Bond Valuation Factor} = \text{Rank}(\text{Conversion Premium Deviation}) + \text{Rank}(\text{Theoretical Value Deviation})$ This factor is used to construct low-valuation indices for different convertible bond styles (equity-heavy, balanced, and debt-heavy)[20][21] - Factor Evaluation: The comprehensive factor performs well across all styles, while the theoretical value deviation factor is particularly effective for equity-heavy convertible bonds[19][20] --- Factor Backtesting Results 1. Conversion Premium Deviation Factor - No specific backtesting results provided 2. Theoretical Value Deviation Factor - No specific backtesting results provided 3. Comprehensive Convertible Bond Valuation Factor - Equity-heavy Convertible Bond Low-Valuation Index: - Annualized return: 26.97% - Annualized volatility: 20.65% - Maximum drawdown: 22.94% - IR: 1.31 - Calmar ratio: 1.18[23] - Balanced Convertible Bond Low-Valuation Index: - Annualized return: 16.04% - Annualized volatility: 11.99% - Maximum drawdown: 15.95% - IR: 1.34 - Calmar ratio: 1.01[23] - Debt-heavy Convertible Bond Low-Valuation Index: - Annualized return: 12.43% - Annualized volatility: 9.80% - Maximum drawdown: 17.78% - IR: 1.27 - Calmar ratio: 0.70[23] --- Style Rotation Model and Construction Methods 1. Model Name: 转债风格轮动 (Convertible Bond Style Rotation) - Model Construction Idea: Captures market sentiment using momentum and volatility deviation factors to rotate among low-valuation style indices (equity-heavy, balanced, and debt-heavy)[27] - Model Construction Process: - Calculate the following sentiment capture metric: $\text{Sentiment Capture Metric} = \text{Rank}(\text{20-day Momentum}) + \text{Rank}(\text{Volatility Deviation})$ - Rank the indices based on this metric and allocate weights accordingly. If all three styles are selected, allocate 100% to the balanced style[27][28] - Rebalance every two weeks[27] - Model Evaluation: The style rotation model effectively captures market sentiment and enhances returns compared to equal-weight indices[27][32] --- Style Rotation Model Backtesting Results 1. Convertible Bond Style Rotation Model - Annualized return: 25.65% - Annualized volatility: 16.82% - Maximum drawdown: 15.89% - IR: 1.52 - Calmar ratio: 1.61[32]