Choosing a Caribbean property catastrophe reinsurance structure

The current structure costs 8% of $100m, or $8m, and retains losses up to $25m plus any amount above $125m. The alternative costs 12% of $150m, or $18m, and retains losses up to $40m plus any amount above $190m. The extra $10m of premium buys $50m of additional limit but also increases the retention by $15m. The marginal rate on line for the extra $50m is therefore 20%, which is expensive relative to the current 8%. On expected loss cost alone, the layer from $125m to $190m is remote: the 1-in-100 loss is $160m, so the layer is only partly exposed at the 1-in-100 return period and is more relevant at the 1-in-200 level. The capital benefit comes from reducing the net 1-in-200 loss. Under the current structure, the insurer retains $110m of a $210m loss. Under the alternative, it retains $60m. That is $50m less tail risk. At a 6% cost of capital, holding $50m less capital is worth about $3m a year, before any risk-margin or rating-agency benefit. That does not cover the extra $10m premium. The alternative therefore looks expensive unless the insurer places a high value on tail protection or the model understates the tail.

The modelled 1-in-100, 1-in-200 and 1-in-250 losses are not precise. Catastrophe models can be off by 20%–30% or more, especially for Caribbean hurricane risk, where landfall probability, demand surge and claims inflation are uncertain. The model has not been independently validated, so the board should not treat the 1-in-200 loss of $210m as a hard number. Flood and storm surge are only partially modelled, which means the true tail could be higher than the model suggests. Non-modelled perils include business interruption, contingent business interruption, litigation and post-event demand surge. The layer from $125m to $190m is exactly the part of the tail where model uncertainty is greatest. A stress test using the 1-in-250 loss of $280m shows that the alternative still leaves $90m retained, while the current structure leaves $155m retained. That difference may matter to rating agencies even if the expected-cost calculation is unfavourable. The insurer should also consider reinsurer credit risk: a cheap layer from a weak reinsurer is not the same as a cheap layer from a strong one.

On pure expected cost and cost of capital, the current $100m excess of $25m structure is better value. The alternative costs an extra $10m for $50m of additional limit, and the capital relief is worth only about $3m a year. I would recommend retaining the current structure unless the board places a strategic value on reducing the 1-in-200 net loss or the rating agency explicitly requires more tail cover. If the board wants more protection, a better option may be to negotiate a lower rate on line, buy a smaller additional layer, or increase the retention less. The recommendation would change if independent validation shows the tail is materially higher than modelled, if flood or storm surge exposure is larger than assumed, or if the reinsurer offers the alternative at a materially lower rate. It would also change if the insurer's capital model is recalibrated to a 1-in-250 basis or if the cost of capital rises. Before renewal, the insurer should stress test both structures against the 1-in-250 loss, confirm the reinsurer's security rating, and document why any additional spend is justified by capital or rating benefits rather than by premium alone.