In the nine preceding lessons the cedant was an insurer or a reinsurer. It can be a government, and the need then changes in kind. What a state protects is not a book, it is a budget: after a disaster it must fund the emergency before damage has been assessed, at the moment its receipts fall and its spending jumps. The hurricanes of the 2000s showed that small island developing states were the most exposed and the least able to fund a rapid response. The mismatch is not one of scale, it is one of timing, and it is that mismatch which is bought.
The parametric trigger is the answer, and here it is all but compulsory. It conditions payment on a measurable physical parameter crossing a threshold, wind speed, an earthquake's magnitude and depth, water height, rainfall deficit, rather than on any assessment of the damage actually suffered. Compensation follows a scale agreed in advance, with no loss adjustment. The obvious advantage is speed, but it is not the main one: the main one is that nobody needs to agree on the amount of the damage in the middle of a crisis. An indemnity trigger presupposes an administration able to survey, to argue and to settle, which is precisely the capacity the disaster has just disorganized.
The regional facility adds a second calculation, and it is what makes the arrangement affordable. The Caribbean Catastrophe Risk Insurance Facility, CCRIF, was launched in 2006 at the World Bank's initiative for the governments of Caricom member states. Each state subscribes its own policy, with its own thresholds and level of cover, and pays an annual premium proportional to its exposure. What is pooled is the carrying: the facility brings together geographically dispersed countries whose losses are weakly correlated, and that decorrelation lowers the cost of cover compared with individual policies. The concept has since been extended to other regions through sovereign reinsurance programs.
The most instructive case is dated and quantified, and it reads two ways that must be held together. After Hurricane Maria, in September 2017, Dominica received a parametric payout of 19.3 million dollars from CCRIF in under two weeks, while total damage on the island was estimated at 1.3 billion dollars. As an indemnity the payout is one and a half percent of the damage, which would be a failure if that were its function. As liquidity available in days rather than months, it funded the emergency without waiting for assessments or international aid. The instrument was not sized to rebuild, but to pay for the first two weeks.
Basis risk, named in an earlier lesson, takes a political form here. A parameter staying below its threshold while the damage is real leaves a government having paid a premium and received nothing, and it has to explain that to its population on the day that population has the least patience. The reference case is dated: the African Risk Capacity, a pan-African mutual backed by the African Union, made its first payouts in January 2015 to Senegal, Mauritania and Niger after the poor 2014 rainy season; in 2016 its model at first triggered no payout for Malawi despite an observed drought, and the institution paid roughly eight million dollars after recalibration. The calibration of the threshold and the choice of index therefore settle the balance between responsiveness and fidelity, and they are settled at drafting, never at the claim.
One technical condition underlies all of this, and it holds well beyond the sovereign case: a geography everyone reads the same way is required. The CRESTA system divides each country into zones that are homogeneous from a natural catastrophe exposure standpoint, at a granularity following the country's weight for the perils concerned: Germany has 130 zones, Japan 68, France 57. Cedants declare their accumulations by zone when negotiating catastrophe treaties, and the same zoning feeds the catastrophe models. Without a shared reference each insurer would use its own carve-up, and aggregating exposures at market level, hence detecting dangerous concentrations, would become impossible.
What remains is to say what these arrangements are not, because that is where the misunderstandings form. They are not aid, they do not replace reconstruction finance, and they do not make the risk go away. What they do is precise and limited: convert an uncertain and badly timed budget charge into a known annual premium and a fast, capped payment. The limit is structural rather than fixable: a parametric cover pays little against the damage, fast, and against a parameter. A state that reads it as an indemnity will be disappointed by the amount; a state that reads it as emergency cash will find it does exactly what it was bought for.
An island state joins a regional parametric facility in 2026. Its cyclone policy pays on a scale indexed to wind speed recorded at three stations: 6 million dollars at the first tier, crossed at 150 kph, 18 million at the second, crossed at 185 kph, and a cap of 30 million. The annual premium is 2.1 million. In October 2026 a cyclone tracks along the coast: the stations record 148, 142 and 151 kph, and damage is estimated at 240 million dollars. The finance minister states that "the insurance did not work". What must be corrected, and what must be compared before renewal?
The first thing to do is not to judge the arrangement but to read the contract, because the three readings give different answers depending on the indexing rule. If the scale reads the maximum of the three stations, 151 kph crosses the first tier and 6 million is due. If it reads the average, 147 kph crosses nothing and the payout is nil. If it reads a weighting by zone or by population, the answer depends on a calculation set out in a schedule nobody has reread since inception. Until that rule is quoted the minister's sentence is not established, and asserting it without the rule would dress a reading error as a failure of the instrument. The second correction concerns expectation. The cap is 30 million, that is 12.5 percent of this episode's damage: even a direct hit above 185 kph would have paid 18 million, and no combination of parameters could have indemnified the loss. What remains to be compared before renewal is therefore neither the amount nor the premium, both known, but the calibration. Three comparisons measure it: the distance between the stations and where the population and the assets actually are, since a wind recorded forty kilometers from the inhabited shore does not describe what that shore went through; the correspondence between the 150 kph tier and the speed at which damage begins on this island's building stock, which is not that of a generic one; and what this scale would have paid on cyclones already observed in the archipelago, the only test that says whether the index describes this territory or an average one. Naming the gap does not remove it: it is chosen, and the open question is whether the gap observed in October is the one that had been chosen.
- 01When the cedant is a state, what is protected is a budget and not a book: the mismatch bought is one of timing, between emergency spending and damage assessment.
- 02The parametric trigger pays on a measurable physical parameter and a scale agreed in advance, with no adjustment: its first virtue is to require no agreement on the amount of damage during the crisis.
- 03CCRIF, launched in 2006 at the World Bank's initiative for Caricom states, pools the carrying of weakly correlated countries, which lowers the cost compared with individual policies.
- 04Dominica, Hurricane Maria, September 2017: 19.3 million dollars paid in under two weeks against 1.3 billion of estimated damage. Derisory as an indemnity, decisive as emergency cash.
- 05Basis risk takes a political form here. Malawi 2016 is the reference: ARC's model at first did not trigger despite an observed drought, and roughly eight million dollars was paid after recalibration.
- 06Every index presupposes a shared geography. CRESTA divides each country into homogeneous zones, 130 in Germany, 68 in Japan, 57 in France, and without a common reference no accumulation aggregates at market level.