Sovereign-Fiscal Contagion Without Exit: Withdrawing Implicit Guarantees on China’s Local Government Debt

Abstract

Existing DSGE models of China's land finance system treat local government borrowing costs as exogenous, a simplification that cannot analyze the fiscal-financial feedback set in motion by the 2021 guarantee withdrawal. I introduce an endogenous risk premium on local government debt into a Gertler and Kiyotaki (2010) financial accelerator framework. A regime switch governs the risk premium, which is dormant under guarantees and active after withdrawal. Once guarantees withdraw, repriced default risk raises local government borrowing costs, higher debt service and falling land revenues weaken fiscal capacity, and the deterioration feeds back into still-higher spreads. Because commercial banks hold large portfolios of local government bonds, the fiscal stress simultaneously erodes bank net worth and tightens credit supply. I estimate this doom loop on nine quarterly observables over 2014--2025. Bayesian subsample estimation identifies an active doom loop in the post-2021 data, with a 90% highest posterior density interval that excludes zero. Four findings emerge. (i) The doom loop is a \textit{recovery trap} that barely deepens the initial downturn but extends the return to trend by nine quarters. (ii) The post-2021 contraction was housing-led, not capital-quality-led, and the doom loop is identified as a fiscal channel independent of the standard banking crisis mechanism. (iii) Systematic elimination of policy instruments shows that monetary easing, fiscal rules, land supply policy, and macroprudential tools each leave the trap intact; only debt swaps that directly compress LGFV borrowing costs shorten the recovery, but even these decay before the underlying revenue shortfall runs its course. (iv) Banks' preference for amortized-cost accounting is individually rational since fair-value recognition would deepen the trap, and yet they face a difficult tradeoff as the same accounting shield blocks the price signals needed to reprice the underlying exposure. Key Words: DSGE, financial accelerator, local government debt, LGFV, implicit guarantees, doom loop, regime switching, Bayesian estimation, China JEL Codes: E44, E58, G21, H63, H74