Identifiability of Contagion Components amid Environmental Fluctuations in Aggregated Default Counts

Can contagion components be identified in aggregated default counts when default probabilities fluctuate with the economic environment? We study this question as an identifiability problem for coarse-grained default-count distributions. We adopt a binomial mixture with a Probit--Normal default probability as a parsimonious null model for environmental fluctuations and test whether contagion provides additional identifiable structure. Against this null, we consider two contagion mechanisms: cumulative contagion in the Davis--Lo model and threshold-type contagion in the Torri model. The Davis--Lo extension substantially improves the likelihood and information criteria for the ALL and speculative-grade classes, whereas the Torri component is largely absorbed by environmental heterogeneity; evidence is weaker for investment-grade defaults. As a robustness test, we replace the Gaussian environmental distribution by a two-component Gaussian mixture, and the Davis--Lo advantage persists for ALL and speculative-grade defaults. These results indicate that cumulative-contagion signatures remain distinguishable from the environmental null models considered here in the ALL and speculative-grade classes, even under annual aggregation.