Abstract:

The persistently high financing cost of urban investment companies has become a key constraint in resolving local government debt risks. Based on issuance data of Urban Investment Bonds (UIBs) from China’s interbank bond market between 2020 and 2024, this paper empirically analyzes the impact of a bond issuer’s valuation errors in the secondary market on the pricing of its new bond issuances in the primary market from a micro-market structure perspective. The findings indicate that: (1) A significant positive correlation exists between the issuer’s average valuation error of outstanding bonds and the issuance of new bonds, suggesting that increased valuation errors elevate UIB issuance costs; (2) this correlation is most pronounced for privately placed UIBs and those with credit ratings below AA+; (3) the observed regional heterogeneity is primarily driven by the issuer’s administrative level. This indicates that optimizing valuation methods and enhancing pricing accuracy constitute an effective pathway to controlling the financing costs of UIBs. Therefore, this paper further develops an XGBoost valuation model optimized with time-series frameworks, and empirical results demonstrate its effectiveness in significantly reducing bond valuation errors. This research provides a feasible pathway for applying artificial intelligence technologies to mitigate bond valuation biases and improve the bond market valuation system, while also offering empirical evidence and decision-making references for regulators to enhance financial market infrastructure and reduce local government debt financing costs.

Key words: Urban Investment Bonds, valuation error, issuance interest rate, XGBoost Machine Learning Model, urban development investment companies, financing cost management