Retrocession is the floor above cession, and nothing more: a reinsurer that has accepted risks passes part of them on to another reinsurer, called a retrocessionaire. The chain then continues, a retrocessionaire being able to retrocede in turn. That cascading mechanism is what spreads very large risks across the world market, so that no single participant carries a catastrophic exposure alone.
The principle is sound and does exactly what is expected of it, as long as the circle stays open. A risk starting with a Japanese insurer and spreading among European and North American carriers and specialist funds is genuinely diffused: each carrier holds a fraction, and nobody holds it twice. That is what is called diversification, and it is the economic reason the whole floor exists.
The defect appears when the circle closes, that is, when the same participants retrocede to one another. A risk ceded several times inside a small group can come back to strike its original carrier, repeatedly and without anyone noticing. Each believes it has transferred part of its exposure when it has kept it by an indirect route. Closed-circle retrocession does not spread risk, it circulates it, and a circulating risk accumulates at the nodes of the circle instead of diluting there.
The London market lived through this at the turn of the 1990s, and the episode remains the common reference on the subject. Major losses working back up the chain revealed that carriers who believed themselves protected found the same loss several times over in their accounts, each turn of the spiral consuming a further layer. What had been bought as protection had become a multiplier.
The gravest consequence of that mechanism is nevertheless not an accounting one, it is informational, and that is what makes it hard to correct. Once retrocessions are unwound across several floors, nobody holds a consolidated view of who really carries what. A carrier knows its direct counterparties; it does not know its counterparties' counterparties, still less the floor above. Everyone's net exposure becomes an estimate, which is the exact opposite of what a risk transfer is meant to produce.
That opacity has a prudential consequence that must be named. A cession to a third party outside the circle eases the capital requirement, legitimately, since the risk has really left. A risk retroceded in a closed circle has not left: it is recycled, and it should therefore not ease the requirement as a real cession would. A group counting as transferred an exposure that returns to it by another route would believe itself stronger than it is, and would believe it at the moment that matters least, that is, before the event.
What the market took from it comes down to a few practices, all of the same family: know who you are dealing with. Know your counterparties, and so far as possible theirs. Track exposures by modeled event rather than by contract, which reveals an accumulation the list of treaties does not show. Be wary of capacity coming back cheaply on a risk you have just ceded, a classic sign of a route that loops. And prefer, on the highest layers, collateralized capacity whose source of capital is identifiable, which is one of the reasons alternative transfer has taken the place it holds on retrocession.
A mid-sized reinsurer places its 2027 catastrophe retrocession with four counterparties. Two are collateralized funds whose capital is identified and posted. The other two are traditional reinsurers from which it has itself accepted, during the year, substantial cessions on catastrophe portfolios in the same territory. Its chief risk officer finds the placement satisfactory: four counterparties, no apparent concentration. What deserves a closer look?
The count of four counterparties is accurate and it does not measure what the chief risk officer believes it measures, because it counts names and not exposures. The signal sits in the second half of the sentence: the two traditional reinsurers are participants from which this house has itself accepted cessions on the same territory and the same peril in the same year. There are therefore at least two routes between it and them, and the risk it retrocedes to them can come back through the underwriting door, which is the very definition of the spiral the London market went through at the turn of the 1990s. The check to run is not contractual but event-based: model a major event on the territory concerned and compute the house's net exposure taking account both of what it retrocedes and of what it has accepted from those same counterparties. The list of treaties will never reveal that accumulation, since each contract is regular taken on its own. Two consequences follow. The first is prudential: the share of retrocession returning by another route should not ease the capital requirement as a real exit of the risk would, and counting it so would make the house believe itself stronger than it is. The second concerns the conduct of the placement: the two collateralized funds, whose capital is identified and posted, do not present that defect, and that is precisely one of the reasons alternative transfer has taken the place it holds on retrocession.
- 01Retrocession is the floor above cession: a reinsurer passes on part of the risks it has accepted.
- 02As long as the circle stays open, it genuinely spreads large risks and that is its economic reason for being.
- 03When the circle closes, it no longer spreads risk, it circulates it, and a circulating risk accumulates at the nodes instead of diluting there.
- 04The London market lived through this at the turn of the 1990s: carriers who believed themselves protected found the same loss several times in their accounts.
- 05The gravest consequence is informational: once retrocessions are unwound across several floors, nobody holds a consolidated view of who carries what.
- 06A risk recycled in a closed circle has not left, so it should not ease the capital requirement as a real cession to an outside third party would.