Assessing Macrofinancial Linkages in China Using a Machine-Learned Parsimonious VAR Model

This paper examines macrofinancial linkages between property developers, financial institutions, and macroeconomic outcomes in China. Using a parsimonious vector autoregressive (VAR) model enabled by a machine learning algorithm, it quantifies how idiosyncratic shocks can propagate and be amplified across sectors, with potential implications for financial stability. Stress originating from privately owned developers and regionally focused financial institutions—though relatively limited in scale—can generate persistent spillovers through lending relationships, common exposures, shared markets, and changes in market sentiment. A decline in property prices may undermine investment, weaken consumer confidence, and adversely affect the health of both the property and financial sectors, thereby disrupting financial intermediation and weighing on broader economic growth. Policy considerations should take into account these feedback loops. Market- and exposure-based tools can be helpful for monitoring macrofinancial linkages and assessing the transmission of shocks.