走在债市曲线之前系列报告（十一）：信用债流动性评估进阶指南（上）

1. Report Industry Investment Rating No relevant content provided. 2. Core View of the Report - The report focuses on the existing credit bond liquidity scoring systems, analyzes the construction logic and operation methods of the "scoring - ranking system" and the "ranking - scoring system", and elaborates on the three - step scoring process of the principal component analysis method in the former and the ranking and scoring rules of the simple linear regression method in the latter. It also analyzes the influence laws of multiple factors on liquidity scores from cross - sectional and time - series perspectives, and summarizes the scoring characteristics and fluctuations under different systems to provide a reference for understanding the logic of credit bond liquidity scoring [3]. 3. Summary by Relevant Catalogs 3.1 Credit Bond Liquidity Scoring System and Full - Process Scoring Analysis - Investors tend to refer to existing systems' liquidity scores for bonds to simplify the decision - making process as it's difficult to obtain and analyze all indicator data independently, and a single indicator can't fully reflect real liquidity [17]. 3.2 "Scoring - Ranking System" vs "Ranking - Scoring System" 3.2.1 "Scoring - Ranking System" - It uses the principal component analysis method, calculating the absolute liquidity score of a single bond by transforming the absolute values of multi - dimensional factors into quantiles. The scoring process includes "quantile standardization", "orthogonal dimensionality reduction", and "principal component weighting". There are no full - score or zero - score bonds, and the scatter plot of the liquidity score and ranking shows an irregular non - smooth curve [22][26]. - Fourteen factors such as trading days, average daily turnover rate, etc., are selected to construct fourteen principal components. The factors with greater influence on the score include average daily quote volume, average daily trading volume, trading days, and specified - day bond balance [31][39]. 3.2.2 "Ranking - Scoring System" - It is based on the idea that liquidity reflects the relative performance in market trading behavior. It ranks bonds according to the performance of liquidity factors in the recent 30 trading days, standardizes the ranking quantiles, linearly weights them to get a comprehensive score, and then assigns scores according to the ranking. There are full - score and zero - score bonds. The core driving factor is the number of broker transactions [25][40]. 3.3 "Principal Component Analysis" vs "Simple Linear Regression" - The principal component analysis method can automatically reduce dimensions, avoid multi - collinearity of indicators, and focus on core information through orthogonal transformation. The evaluation result is more comprehensive, but the factor weights are dynamically adjusted with orthogonalization and have weak interpretability. The simple linear regression method is easy to operate, with artificially assigned weights, but it highly depends on trading - related factors, is easily affected by outliers, and focuses more on short - term trading activity [8]. 3.4 Statistical Analysis of Historical Liquidity Score Data of the Two Systems 3.4.1 Cross - sectional Perspective - Bond type dimension: Both systems show that financial bonds have the highest liquidity score, followed by industrial bonds, and then urban investment bonds. The "scoring - ranking system" has relatively conservative scores, while the "ranking - scoring system" has generally higher scores [55]. - Remaining maturity dimension: In the "scoring - ranking system", bonds with a remaining maturity of 0 - 1 year usually have the lowest score, while in the "ranking - scoring system", they often have the highest score [58]. - Subject rating dimension: The liquidity score is positively correlated with the issuer's credit rating. High - rating financial bonds have the highest score, and low - rating financial bonds have the lowest score. The score contraction of financial bonds is more significant than that of urban investment bonds and industrial bonds when the rating drops [64]. - Bond issuance characteristics dimension: Unsecured bonds, public - offering bonds, and perpetual bonds have higher liquidity scores than secured bonds, private - offering bonds, and non - perpetual bonds in both systems [69][70]. - Bond balance dimension: The liquidity score is positively correlated with the bond balance. Bonds with a balance of over 10 billion have the highest score, while those with a balance of less than 1 billion have the lowest score [72]. - Urban investment bond administrative level dimension: Provincial urban investment bonds have the highest liquidity score, followed by municipal and district - county - level bonds. The difference is more significant in the "ranking - scoring system" [77]. - Industrial bond enterprise attribute dimension: Central state - owned enterprises' industrial bonds have the highest score, followed by local state - owned enterprises and other enterprises, and private enterprises have the lowest score [81]. - Industrial bond industry attribute dimension: The liquidity score is related to the bond market scale. Industries with large bond balances, such as banks and non - bank finance, have high scores, while industries with small bond balances, such as light manufacturing, have low scores [85]. - Provincial and administrative rating dimension: There are regional differences in the liquidity scores of urban investment bonds at different administrative levels, which are highly related to the local bond market scale [88]. 3.4.2 Time - series Perspective - Urban investment bonds by province: There are differences in the liquidity stability among provinces in both systems. The stability has little correlation with the bond balance and the scoring rankings of the two systems are not highly related [91]. - Industrial bonds by industry: The liquidity stratification among industries in the "scoring - ranking system" is more stable than that in the "ranking - scoring system". Industries with large bond balances generally have higher score stability, and the rankings of the two systems are somewhat correlated [96].