Trade credit insurance (TCI) is a specialized line of property and casualty insurance that protects seller businesses against financial losses due to buyer insolvency. It plays a critical role in firm-to-firm trade, where accounts receivable often constitute a substantial portion of firm assets, and supports firm viability by mitigating counterparty risk. Moreover, TCI helps dampen the propagation of shocks through supply chains. Despite this importance, research on TCI remains limited, largely due to the lack of granular data. Consequently, both predictive risk modeling for TCI claims, which is complicated by the data’s complexity, and the role of TCI in facilitating trade and mitigating supply chain risks remain underexplored in the literature.

Leveraging six years of detailed TCI data from an Asian TCI insurer, extended to seven years in the last essay, this dissertation studies these issues through three essays: (1) Statistical Learning of Trade Credit Insurance Network Data with Applications to Ratemaking and Reserving, (2) Network Auto-Logistic Regression Model for Longitudinal Trade Credit Insurance Claim Data, and (3) Trade Credit Insurance in Firm-to-Firm Trade Network.

The first essay develops a bivariate, network-augmented Generalized Linear Mixed Model (GLMM) for ratemaking and reserving in TCI. The model jointly captures claim occurrence and reporting delays, incorporates business- and policy-level random effects, and uses network information to improve predictive performance. The second essay develops an extended Network Auto-Logistic Regression (NAR) model for longitudinal TCI claim data. By encoding dependence through operators constructed from the directed buyer-seller network and estimating the model via maximum pseudo-likelihood, it provides a computationally scalable framework for edge-level claim prediction. The third essay examines the role of TCI coverage in firms’ trade-network adjustments. Using difference-in-differences (DD) and triple-difference (DDD) frameworks around a large external shock, namely the U.S.-China trade war, it studies whether firms in more exposed industries reallocate trading relationships and insured amounts differently from firms in less exposed industries, and whether these adjustments vary with pre-shock TCI coverage.

Taken together, the three essays contribute to the literature in several ways. First, they develop modeling frameworks tailored to the longitudinal and directed-network structure of TCI data. Second, they show that network information materially improves risk classification, ratemaking, and reserving in TCI portfolios. Third, they provide new empirical evidence that TCI functions not merely as a passive risk-transfer device, but also as an active mechanism that interacts with firm behavior, network adjustment, and shock propagation in trade relationships. While the empirical setting is TCI, the methodological insights extend more broadly to problems involving edge-level outcomes on directed networks.