Macro-Financial Conditions and Sponsor and Underwriter Heterogeneity in Primary Market CAT Bond Spread Formation (Cronfa (Swansea University))

Summary

• The study uses a hand-collected dataset of 702 issuances by 101 sponsors (1997–2020).p.1
• Selected sponsor and underwriter identities are associated with spread discounts of approximately 0.7–1.7 percentage points.p.1
• Europe's natural catastrophe protection gap still exceeds 70% of total losses.p.4
• XGBoost and CatBoost reduce MAPE by 33.93% and 30.66% relative to the linear benchmark in the 75/25 random train-test split.p.7
• Sponsor and underwriter discounts are substantial relative to an average spread of roughly 7 percentage points.p.7
• The linear model's RMSE falls from 0.0221 (bond-specific only) to 0.0212 (bond + sponsor), 0.0167 (bond + macro) and 0.0164 in the full set.p.17

Key insights

1. Credit market conditions and the reinsurance pricing cycle are associated with primary market spreads and give the strongest improvements in predictive accuracy.p.1
2. The first European CAT bond ETF was launched in December 2025.p.4
3. Boosted trees do not keep their advantage under rolling out-of-time evaluation.p.7
4. Investment-grade bonds have spreads about 2.7 percentage points lower than non-investment-grade issues.p.15
5. Bonds exposed to peak multiperil structures command 1.8–2 percentage points higher spreads.p.15
6. Macro-financial variables deliver the largest improvement in predictive accuracy, while sponsor-related information adds only modest gains.p.17
7. CAT bond pricing assessments should compare several model classes and information sets rather than rely on a single specification.p.20