A treaty does not end at expiry, it ends when its last loss is settled. On a short-tail class the distance between the two is measured in months. On a liability account it is measured in decades, and throughout that time two parties with no further business together keep exchanging accounts, reserving, tying up capital and carrying credit risk on one another. Four exits exist, and they do not cost the same: letting it run off, commuting, novating, or having capped the notification window in advance. The first three are decided after the event, the fourth is negotiated at inception, which makes it the only one of the four that cannot be retrieved later.
Run-off is the default exit, the one obtained by deciding nothing. The portfolio stops writing new business but keeps settling the losses arising from contracts already concluded until they are extinguished. Older underwriting years keep being notified and keep developing for ten to fifteen years on long-tail classes. A specialist market has grown around transferring these portfolios, dedicated players buying run-off liabilities on the bet that the reserves transferred will suffice to cover the residual losses while leaving a margin. The central risk in run-off is reserving: any under-estimation of losses to come turns an apparently profitable transaction into a loss, particularly on late-reported claims.
Commutation is the negotiated exit. The reinsurer pays a lump sum, and the cedant takes back the whole of the future losses, notified or not. Each side has a distinct interest in it, and that is what makes the deal possible. The reinsurer closes an underwriting year, releases reserves and the capital that goes with them, and removes a long-dated uncertainty. The cedant receives cash immediately, eliminates credit risk on a signature it judges fragile, and simplifies its balance sheet. What both sides are buying is the end of the cost of the tail: on a long-tail class, keeping a relationship open for twenty years costs more in administration, capital and attention than the residual uncertainty is worth.
Everything turns on the price, and the price rests on a discounted estimate of future losses. There lies the asymmetry of a commutation, and it is not a matter of bad faith: two estimates confront each other, the reinsurer's wanting out and the cedant's wanting to be indemnified for what it takes back, and nothing says which is right since the event to be estimated has not happened. The discount rate and the duration adopted are enough to move the deal by millions, before any discussion of the loss experience itself. A badly valued commutation silently transfers a liability, and regulators examine its assumptions for that reason: the operation is irreversible and it moves the liability outside the supervision that bore on the reinsurer.
Novation substitutes one reinsurer for another on a treaty in force, the new one assuming the entirety of the rights and obligations and the old one being definitively released. It requires the consent of all three parties, the cedant included, which distinguishes it from a retrocession where the original reinsurer remains sole debtor towards its cedant and covers itself upstream. That distinction is the key: retrocession leaves the liability in place and merely adds an asset against it, only novation removes the liability. For the cedant, consenting is a credit decision in its own right and not a formality: it swaps one signature for another, and its consent extinguishes all recourse against the outgoer.
That leaves the case where each party's consent cannot be obtained, and it is a common one. Take a reinsurer withdrawing from a market and transferring 340 million euros of reserves spread over 87 treaties: 61 cedants accept the novation after examining the transferee, 26 refuse, representing 94 million, and the outgoer must keep those liabilities until extinction. This is the blocking minority problem, and the English scheme of arrangement answers it: the company proposes a collective settlement to its creditors, sanctioned by the court and binding on all once a qualified majority consents. Its distinctive feature is that it applies to a solvent company too, which makes it a tool of exit rather than of failure. The criticism bears on the fate of the minority, compelled to accept a valuation it disputes.
The sunset clause is the fourth exit, and the only one the cedant pays for alone. It fixes a date beyond which a loss can no longer be presented to the reinsurer, even one that occurred during the period of cover and even one covered on the merits, with a period commonly running from three to ten years after expiry. What it solves is the indefinite tail that prevents closing an underwriting year. Its counterpart bears entirely on the cedant, which alone keeps the losses notified out of time, including those whose lateness is not attributable to it. Regulators and auditors watch it for that reason: they check the cedant reserves for that implicit retention rather than treating it as a nil risk.
A cedant considers in 2026 commuting a liability treaty written in 1998, whose remaining ceded reserves stand at 31 million euros. The reinsurer, discounting at 3.4 percent over a nine-year duration, offers 23.5 million. The deal closes at 26.2 million. Four years later the losses actually incurred total 29.8 million. What did each party buy, and what should be concluded from the outcome?
The gap between the 31 million of reserves and the 23.5 million offered is not an arbitrary haircut: it is almost entirely the effect of discounting. Discounting 31 million at 3.4 percent over nine years gives roughly 23 million, which means the opening offer accepted the cedant's reserves without contesting them, paying only their present value. Negotiation then moved the 23.5 to 26.2 million, that is 2.7 million more, and that gap is the real subject of the discussion: it corresponds to a lower discount rate or a shorter duration, not to a different view of the loss experience. The outcome four years later is that the cedant received 26.2 million for a liability that cost 29.8 million, and therefore bears 3.6 million no recourse can reopen. The easy conclusion should be resisted. Those 3.6 million do not prove the cedant negotiated badly, and showing why requires reasoning as it had to at the time. It bought three things on signing: immediate cash of 26.2 million, the disappearance of twenty-eight years of credit risk on a 1998 signature, and the end of the administration and capital costs of a treaty nobody was managing any more. The reinsurer bought the definitive closure of an underwriting year. Each got what it was after, and a commutation is not a bet on loss experience, it is the exchange of an uncertainty for an amount. What the case establishes is narrower and more useful: an exit that settles definitively is paid for with an estimate that can no longer be reopened, and the gap is always observed afterwards. That is why regulators look at the assumptions and not at the outcome.
- 01A treaty ends when its last loss is settled, not at expiry: on a long-tail class the distance between the two is measured in decades.
- 02Run-off is the default exit, the one obtained by deciding nothing, and its central risk is the reserving of late-reported claims.
- 03In a commutation the discount rate and the duration move the deal by millions before any discussion of the loss experience itself.
- 04A commutation exchanges an uncertainty for an amount: a gap observed afterwards does not prove a bad negotiation, which is why regulators examine the assumptions and not the outcome.
- 05Only novation removes the outgoer's liability; a retrocession leaves it in place and merely adds an asset. Consenting to a novation extinguishes all recourse against the outgoer.
- 06The sunset clause is the only exit the cedant pays for alone, and it is negotiated at inception: the only one of the four that cannot be retrieved after the event.