Which treaty year answers, and what clean cut changes

Before computing how much a reinsurer owes, one has to know which reinsurer owes it. The question looks trivial as long as a single year is in view, and stops being trivial as soon as a policy straddles two treaties, which is the case for every annual policy written other than on January 1. Two answers exist, they are called risk attaching and losses occurring, and they point to different years for the same loss. The sharing studied so far, quota share, surplus, excess of loss, applies identically under both: what changes is not the computation of the recovery, it is the identity of the debtor. Getting the basis wrong is not an error of amount, it is calling on the wrong panel.

A risk attaching treaty covers every policy issued or renewed during its period, and follows them to their own expiry, including after the treaty itself has expired. A 2026 treaty therefore protects a policy written in November 2026 until November 2027, while the 2027 treaty has been in force for eleven months. This is the normal basis for proportional treaties, and the reason is economic: it makes the cession of premium and the cession of risk coincide. The reinsurer receives a fraction of the policy premium and carries the matching fraction of losses, for the whole of its term. A losses occurring basis would break that coherence by making one reinsurer carry the losses of a premium another one collected.

A losses occurring treaty covers the losses that occur during its period, regardless of when the underlying policies were issued. An annual policy written in October is therefore covered by two successive treaties: the current year's until December 31, the following year's thereafter. This is the normal basis for non-proportional covers and for the whole of catastrophe reinsurance, because it matches the period of cover to the risk season, which is precisely what a treaty underwriter wants to control. It knows exactly which events concern it, the ones dated inside its window, without having to track the cedant's policy portfolio. The trade-off is that it inherits risks written before it came in, on terms it never saw.

Each basis therefore has a defect, and that defect explains a market practice. Risk attaching spreads: a treaty stays exposed for up to twenty-four months after inception, which makes it impossible to say what a year cost until long afterwards and forces tracking by underwriting year rather than by accounting year. Losses occurring exposes to year-end anti-selection: the incoming reinsurer picks up business the cedant wrote in December, on terms it saw nothing of, which is what explains the clauses restricting changes of underwriting policy mid-treaty. Neither basis is better. They answer two different questions, and one does not mix them inside a single program without knowing it.

The defect of risk attaching produced a remedy with a name of its own, clean cut. It closes a proportional treaty definitively at expiry, with no residual tail: unearned premiums and outstanding loss reserves are transferred, against payment, to the following year's treaty and to its reinsurers. Two symmetrical entries perform the operation. On the way out, the expiring treaty pays the incoming reinsurers the value of the liabilities it passes to them. On the way in, the new treaty receives those sums and assumes the liabilities. What clean cut buys is precise: the ability to close each year and to reason by accounting year.

What clean cut costs is a valuation risk, and it is asymmetric. The outgoing reinsurer is paid on the basis of the reserves as estimated on the day of transfer. The incoming reinsurer inherits a liability whose estimate is not its own, and which it often had no time to audit. Any under-reserving at the cut-off date is therefore borne entirely by the incomer, with no recourse against the outgoer, which makes that valuation the main object of the negotiation. The transfer rate applied to unearned premiums and the method used to value the reserves are the two parameters that decide who gains from the operation.

A simple case gives the order of magnitude. Take a property quota share closed by clean cut on December 31: the outgoing portfolio transfers 18.4 million euros of unearned premium and 26.1 million of loss reserves to the following year's reinsurers, against equivalent payment. An audit conducted two years later reveals an under-reserving of 3.7 million at the cut-off date. It is borne entirely by the incomers. Measured against the 26.1 million transferred, the error is only 14 percent, which is well within the ordinary range of a reserve estimate: clean cut therefore does not create an exceptional risk, it moves an ordinary risk onto a party that did not make the estimate and can no longer contest it.