中金公司在固收加量化策略方面有哪些积累和研究？ 中金公司固收团队在固收加量化策略方面的研究主要从两个维度展开。首先是 从股债组合被动配置的角度，分析典型的固收加资产配置模型，包括国家恒定、 目标风险、目标日期、风险平价等避险策略，以及动态量化配置模型。这些策 略通过历史回测和参数优化，提供了更为清晰的组合拆解方法，降低主观判断 对择时层面的影响。 其次是从细分资产角度进行深入研究。股票方面主要集中 于风格轮动和行业轮动，同时也包括因子策略，如红利板块增强策略。纯债方 面则开发了多因子策略，通过平衡期限溢价和信用溢价，为投资者提供择券思 路。此外，中金团队还探索将大模型和深度学习应用于固收加资产组合构建， CPPI 避险策略通过无风险资产票息形成安全垫，但传统方法缺乏精确计 量。边缘 CPPI 策略通过债券和股票资本利得调配权重，满足回撤要求， 更符合当前市场需求。 合成期权策略通过模拟股债配置实现类似期权效果，可调整至不完全保 本，作为一种灵活有效的避险工具。通过调整参数，可以使整个组合的 年化收益率超过 CPI。 宏观择时模型基于广义和狭义流动性指标进行股债轮动配置，通过控制 股债比例和杠杆，提升组合年化收 ...

Summary of Conference Call Records Industry or Company Involved - The discussion revolves around the Fixed Income + Intelligent Agent (固收+智能体) and the Black-Litterman (BL) model, focusing on asset allocation and investment strategies in the financial sector. Core Points and Arguments - The BL model addresses the sensitivity of traditional asset allocation models to input data, enhancing the stability and accuracy of return predictions [1] - The model calculates market implied asset returns by constructing a market portfolio, using CAPM to compute expected returns, and assuming no Arrow part to derive excess returns [2] - The BL model reflects market risk preferences by translating required return compensation for unit risk exposure into expected returns for each asset [3] - Investor views are integrated into the BL model through absolute and relative perspectives, adjusting posterior returns based on asset correlations and confidence intervals [4] - In China, the BL model requires the use of benchmark portfolios instead of market portfolios, considering contractual constraints and controlling turnover rates [5] - The Fixed Income + Intelligent Agent utilizes a segmented asset model (e.g., GBR model) to predict returns, achieving volatility reduction in portfolios despite limited accuracy for individual assets [6] - Introducing confidence intervals and segmented asset indicators significantly enhances the predictive accuracy of the BL model, indicating substantial development potential for active fixed income strategies [7] Other Important but Possibly Overlooked Content - The Fixed Income + Intelligent Agent consists of two main components: client-facing applications using large models for visualizing investor expectations and a research segment that includes asset selection, performance understanding, return and risk forecasting, combination selection, and trade enhancement [8] - The BL model's application in China necessitates specific adjustments, such as using benchmark combinations and considering contractual constraints on fixed income portfolios [9][10] - The GBR model, while simple, uses price and volume data to predict future returns, achieving an average accuracy of less than 50% for individual assets, but showing reduced volatility in portfolios [11] - Future development of active fixed income strategies relies on stronger and more accurate views of segmented assets, which can lead to better model performance and superior outcomes compared to passive index products [12] - Resources for further research and application of the BL model are available, including Python libraries and pre-existing code for practical implementation [13]