Upstream and Downstream Credit Exposure Modeling Around Core Industrial Enterprises

Abstract

Core industrial enterprises often anchor large supplier and distributor networks, and their operational or financial changes may affect upstream procurement, downstream sales, trade credit, receivable turnover, and bank credit exposure. Credit risk assessment focused only on a single borrower may miss hidden risk accumulated through dependence on a core enterprise. This study proposes an upstream and downstream credit exposure modeling method around core industrial enterprises. The model constructs a multi-relational graph linking core firms, suppliers, distributors, receivables, purchase orders, bank credit lines, guarantee contracts, logistics records, ownership ties, and payment outcomes. A graph neural inference module captures credit exposure transmission across upstream and downstream relations, while knowledge rules identify dependency risks from delayed payment, customer concentration, order contraction, shared guarantors, and regional industry shocks. Experiments are conducted on a core-enterprise credit dataset containing 8,240 core industrial firms, 162,000 upstream suppliers, 96,000 downstream distributors, 2.18 million purchase-order records, 1.34 million receivable entries, 74,000 bank credit contracts, and 9,860 confirmed credit-risk cases over 40 months. The proposed method shortens median early-warning time from 76 days to 28 days compared with a borrower-only credit model. It identifies 10,420 upstream exposure chains and 6,780 downstream repayment-risk paths linked to core-enterprise financial stress. Graph aggregation reduces 21,600 borrower-level alerts to 4,950 supply-chain exposure review cases. Portfolio simulation shows that graph-informed monitoring reduces expected overdue credit exposure by 156 million RMB during the validation period. The results demonstrate that upstream and downstream graph reasoning can improve credit risk management around core industrial enterprises by revealing dependency-based financial risk transmission.