To examine what physical risk information remains identifiable for credit in Caribbean Small Island Developing States (SIDS), where hazard data are rich but loss records scarce.
Using storm tracks and four damage-function calibrations, we assess Caribbean risk rankings through Monte Carlo simulation, stochastic dominance and independent records, and compare Caribbean and United States wind-loss relationships using wild cluster bootstrap inference.
Territory tiering is identifiable and broadly stable across damage-function choice, whereas point magnitudes remain imprecise under sparse Caribbean loss records.
We identify an ordinal-cardinal boundary in Caribbean SIDS: risk tiering supports screening today, while precise loss estimates for pricing require better records that artificial intelligence (AI)-assisted methods may help develop.
Territory tiering can support risk screening, while data limitations motivate further research on AI-assisted loss-data generation for finer pricing.
Jia, X. and Capodagli, S. (2026): Ranking, Not Point Estimates: Risk Tiering and the Case for AI-Assisted Loss Data in Caribbean SIDS Credit. World Journal of Entrepreneurship, Management and Sustainable Development (WJEMSD), Vol. 22, No. 6, pp. 585-600.
