| 钱一蕾,蔡冬美,钟宁桦,赵珈露.专项债券项目收益风险定价有效性研究——兼论优化专项债券管理机制的路径[J].数量经济技术经济研究,2026,(2):14-126 | | 专项债券项目收益风险定价有效性研究——兼论优化专项债券管理机制的路径 | | The Pricing Efficiency of Project Return Risk in Special Bonds: With a Discussion on Pathways for Optimizing Management Mechanisms | | | | DOI： | | 中文关键词: 专项债券 项目收益风险 债券定价 地方政府债务管理 | | 英文关键词: Special Bonds Project Return Risk Bond Pricing Local Government Debt Management | | 基金项目: | | | 中文摘要: | | 专项债券是积极财政政策的重要抓手，但其项目收益不达预期的问题日益突出，优化专项债券管理机制迫在眉睫。本文使用机器学习模型测算了专项债券利差中反映的项目收益风险。结果显示，2019~2023年，在发行主体和债券特征均一致的情况下，专项债券利差仅比一般债券高出0.55BP，这表明债券市场几乎无法识别出项目收益风险。为了探究上述现象的成因，本文拆解了6302只专项债券对应的8.5万余个项目，研究发现,当前专项债券发行管理中普遍存在项目打包发债现象，模糊了债券与项目的对应关系，且单个债券包内各项目的收益水平参差不齐，致使专项债券定价扭曲。本文进一步分析揭示了优劣项目打包发债的制度性成因,在项目申报阶段，为了争取更多限额，部分地方政府存在“重数量、轻质量”的动机，导致入库项目良莠不齐；在债券发行阶段，为了实现收益和融资的平衡，则倾向于将优劣项目打包发行。新增债务融资需求越强的省份，专项债券项目的质量参差越大，优劣项目打包发债的程度越高。本文基于项目申报和债券发行两个环节构建专项债券管理矩阵，因地制宜提出针对性的优化路径，助力实现更合理的财政资源配置和更有效的债务风险管理。 | | 英文摘要: | | Local government special bonds serve as a key instrument of positive fiscal policy; however, risks associated with under-performance in project returns are increasingly evident, underscoring the urgent need to refine and optimize the bond management framework. This study employs general bonds as a benchmark to measure the project return risk of special bonds. General and special bonds share identical issuers, with their core difference lying in their funding sources for debt repayment. General bonds are used to finance nonrevenue-generating projects and are repaid solely from fiscal funds, while special bonds finance revenue-generating projects, with repayment coming from both project revenues and fiscal funds. Thus, the difference in repayment risk between these two bond types primarily stems from project revenue performance. This study utilizes an XGBoost model to estimate the yield spread between special and general bonds, which represents the project return risk. Our estimates reveal that from 2019 to 2023, after controlling for issuer and bond characteristics, the yield spread of special bonds was merely 0.55 basis points higher than that of general bonds. This implies that the bond market fails to price the additional risk associated with project returns of special bonds.To elucidate the underlying causes of this phenomenon, this study conducts an analysis of over 85,000 projects, corresponding to 6,302 special bonds issued from 2019 to 2023. The findings reveal that each special bond finances an average of 14 projects, with these projects exhibiting significant heterogeneity in expected returns. The mean coefficient of variation for expected returns across projects within individual special bonds is 0.28. The bundling of high- and low-return projects obscures the one-to-one correspondence between special bonds and their underlying investment projects, hiding market-available information and making it difficult for bond markets to accurately identify project-specific revenue special risk.Further investigation uncovers deep-rooted institutional drivers of this phenomenon. During the project application stage, some local governments-motivated by the pursuit of a larger debt limit-exhibit a bias toward quantity over quality, resulting in significant variation in the quality of projects admitted to the reserve pool. In the bond issuance stage, to balance expected revenues with financing costs, local governments tend to bundle high- and low-quality projects together for issuance. Provinces with stronger demand for new debt financing display greater variation in the quality of their special bond projects and a higher degree of such bundling. Finally, this study constructs a management matrix for special bonds based on the two key stages of project application and bond issuance and proposes targeted optimization strategies to promote more rational fiscal resource allocation and more effective debt risk management.This study makes three contributions to the literature. First, it incorporates project-level revenue risk into the analysis, thereby extending the literature on local government bond pricing. Existing research has primarily focused on regional economic and fiscal characteristics, macroeconomic conditions, and policy considerations. In contrast, this study focuses on special bonds-officially mandated to achieve self-balancing between project revenues and financing costs and examines the bond market’s capacity to price project-level return risk, offering a novel perspective on the pricing mechanisms of local government bonds.Second, it uncovers key institutional drivers behind the pricing distortions of special-purpose bonds, providing actionable insights for reforming their issuance and management framework. Drawing on a dataset that links 6,302 special bonds to over 85,000 underlying projects, we find that the widespread practice of bundling high- and low-return projects in a single bond issue is a primary cause of pricing inefficiencies. This not only dilutes the risk-signaling function of bond yields but also enables low-quality projects to free-ride, thereby potentially increasing ex post repayment risks. Reforms should prioritize curbing the bundling of heterogeneous projects and enhancing the transparency and identifiability of project-level revenue risk in bond pricing.Third, the study identifies inefficiencies in the project application and debt limit allocation process for special bonds and proposes specific recommendations to improve the local government debt limit management system. As of the end of 2024, the total special debt limit had reached RMB 35.5 trillion, making it one of the most significant fiscal resources. Our analysis reveals that during the application stage, local governments’ strong incentives to secure debt limit, coupled with information asymmetry between central and local governments, led to substantial heterogeneity in project quality. Therefore, we recommend strengthening ex post performance evaluations and tying future debt limits to actual project outcomes. This will promote more precise and efficient fiscal resource allocation while curbing the accumulation of local debt risk. | | 查看全文 相关附件： 下载数据代码附录 |
| | |
