全球固收量化：四大流派、五大局限未来已来系列之一

1. Report Industry Investment Rating No information provided in the content. 2. Core Viewpoints of the Report - The bond market is at a transformative moment, and the "Future is Here" series of reports focuses on exploring cutting - edge technologies affecting the bond market to empower investment research [3]. - Fixed - income quantification is an inevitable product of financial industrialization and the answer of the bond market in the AI - empowered era. It has evolved from subjective to systematic, model - based, and data - driven, and has moved from the edge to the core of the trading desk [3]. - Compared with relatively mature equity quantification, fixed - income quantification has unique characteristics, including more complex tools and market structures, stronger policy and institutional factors, and more prominent liquidity and data quality issues [3]. - The report aims to answer four questions: the main schools of fixed - income quantification and their basic logics, the applicable market environments for these quantification technologies, the problems that quantification technologies cannot solve or may amplify risks, and the future prospects and optimization spaces of fixed - income quantification [3]. 3. Summary According to the Table of Contents 3.1 Global Fixed - Income Quantification: Four Schools and Basic Logics 3.1.1 Fundamental Quant - Focuses on using economic logic, macro data, and fundamental factors to predict market directions or asset values. It tries to "model" the logic that traditional macro research relies on analysts' experience for [8]. - The process includes data input (such as GDP, CPI, and PMI), building models (e.g., a two - factor model of "growth" and "inflation" or a multi - dimensional macro - factor system), and formulating trading logics (e.g., going long on interest - rate bonds in the "loose money + tight credit" cycle) [8]. - With the development of data technology, it uses high - frequency data for nowcasting to capture economic temperature changes. However, it faces challenges such as "overfitting" risk, structural breaks, and the risk of "fundamental desensitization" and model failure [8][9][10]. 3.1.2 Technical Quant - Focuses on using market volume and price data to capture trading opportunities from trends, reversals, or micro - structures without relying on macro - economic explanations [11]. - Trend - tracking and CTA fixed - income strategies use time - series momentum trading on interest rates and bond prices, which has significant long - term trend premiums and is important in multi - asset CTA strategies. The strategies are applied through unified momentum/trend rules on multiple products and can be part of a cross - asset trend strategy [11][12][15]. - Market - making and micro - structure quantification focuses on using quantification technology to improve pricing and inventory management in aspects such as order - book modeling, quoting strategies, and execution algorithms [18][20]. 3.1.3 Relative Value Quant - Focuses on cross - sectional comparison or finding pricing deviations to earn mean - reversion returns or risk premiums, often involving long - short hedging or factor - based bond selection [18]. - Interest - rate term structure and curve trading use various interest - rate term structure models to factorize the yield curve and conduct relative - value trading based on the deviation between the theoretical and actual curves [18][23]. - Carry/Roll - down strategies aim to earn the "time value" of the interest - rate curve and bonds. It is effective in stable or downward - trending interest - rate environments and is often incorporated into the factor - investment framework [26][28]. - Credit and spread factor strategies map bond characteristics into a series of credit and style factors to construct long - short or over -/under - weighted portfolios to earn factor premiums [33]. - Relative value and basis arbitrage focus on price "dislocation" between different tools with the same or similar risk exposures and use methods like PCA, mean - reversion modeling, and high -/medium - frequency data mining to construct statistical arbitrage strategies [38]. 3.1.4 Multi - Factor Models - Aims to systematically integrate excess returns from different sources. The core logic is to decompose the expected return of bonds into a linear combination of several risk factors to build a portfolio with a higher Sharpe ratio and smaller drawdowns [39]. - It is related to the three previous schools. It uses a large amount of fundamental data, includes momentum factors from the technical school, and is mainly used for "bond selection" similar to the relative - value school [43][45]. 3.2 Market Environments Suitable for Quantitative Technologies - Liquidity and Trading Systems: High - liquidity, low - transaction - cost markets are suitable for curve trading, relative - value, CTA trend, market - making, and high - frequency strategies; medium - liquidity markets are suitable for term - structure models, carry, and some relative - value and factor strategies; low - liquidity, OTC - dominated credit markets are suitable for medium - to - low - frequency factor strategies, duration/barbell allocation, and some structured - product pricing [48][49]. - Interest - Rate Levels and Volatility Environments: When the interest - rate center is declining with mild fluctuations, carry and roll - down strategies perform well, and term - structure strategies can profit from the "loose - neutral" switch. When interest rates rise rapidly or policies change suddenly, term - structure and carry strategies are prone to net - value drawdowns, while CTA trend and duration - hedging strategies can provide some protection [50]. - Credit Environments and Macroeconomic Cycles: In a low - default - rate, credit - expansion period, credit factors and credit - sinking strategies have high "tailwind returns". In a credit - contraction and high - default period, quantitative models may underestimate tail risks and are difficult to capture sudden "black - swan events" [51][53]. 3.3 Five Limitations of Fixed - Income Quantification - Policy and institutional inflection points are "unquantifiable" because central - bank monetary policies and regulatory reforms often show "discrete" and "mutant" characteristics, and historical - data - trained models may fail when regime shifts occur [55]. - "Liquidity black holes" and "out - of - model" risks exist because most models assume a "frictionless market", but the credit - bond market often faces liquidity shortages, which can lead to the failure of traditional strategies [56]. - Credit defaults have "small - sample" and jump risks. Bond defaults are sparse events, resulting in model overfitting or non - convergence, and the non - standardized information in the default process is difficult to cover [57]. - Complex terms and game behaviors are non - modelable. Many fixed - income products have complex option terms, and their triggering depends on issuers' subjective will, causing the deviation between the theoretical option value calculated by quantitative models and the market price [58]. - Crowded trading and endogenous instability occur when quantitative strategies are highly homogeneous. Once the market fluctuates in the opposite direction, the concentrated stop - loss orders can cause a stampede and more severe fluctuations [59][60]. 3.4 Outlook: The Future Landscape of Fixed - Income Quantification - Quantamental (Quantitative + Fundamental): Quantitative analysis will empower fundamental analysis. Future mainstream models include "quantitative support with subjective decision - making" or "subjective logic with quantitative verification", applicable in macro - asset allocation and credit screening [62]. - In - Depth Penetration of Alternative Data and AI Technologies: With the development of large - language models, non - structured data can be processed, providing new sources of alpha. Applications include semantic and sentiment analysis of text data and using satellite and geographical data for investment analysis [63]. - Algorithmic and Automated Trade Execution: The increasing proportion of electronic trading in the Chinese bond market provides a foundation for algorithmic trading. Intelligent order - splitting algorithms can reduce impact costs, and machine - learning - based market - making strategies can adjust quotes and control inventory risks [64][66].