The past rather than the future: ceding losses that have already occurred

Everything above covers the future. A treaty, a facultative cession, a cat bond, a sidecar, an industry loss warranty all bear on losses that have not occurred and may never occur. Retrospective reinsurance turns the question around: it covers liabilities already born. The event has happened, it is incurred, it is reserved in the cedant's accounts. What remains uncertain is no longer whether it occurs, but what it will finally cost, that is, its development. On long tail lines, the gap between the reserve carried and the ultimate cost is an insurer's principal risk, and it sits on years it believed were behind it.

Two instruments do the work, and they do not do the same thing. The loss portfolio transfer, known by its initials LPT, hands a reinsurer a set of losses that have already occurred together with the reserves attached, in return for a premium: the portfolio leaves the cedant's balance sheet in economic terms, and the reinsurer takes on its run-off. The adverse development cover, or ADC, transfers nothing: it leaves the portfolio where it is and protects only against reserves overrunning beyond an agreed attachment. The first is an exit, the second is tail protection. Confusing them is expensive, since they have neither the same balance sheet effect, nor the same counterparty profile, nor the same price.

What a cedant seeks in them is plain and legitimate. It stabilizes its result, since the volatility of the run-off passes to a third party. It frees the capital tied up to carry those reserves, which becomes available again to write new business. And it closes an inherited exposure, which has a value of its own when the activity concerned is no longer its own: a line it has withdrawn from, a portfolio acquired with a company, an old year nobody in the house remembers underwriting. These transactions are the central tool of the legacy market and of run-off situations, where they turn a page that natural extinction of the claims alone would take fifteen years to turn.

What the transaction does not do must now be said, and it is the ambiguity at the heart of the product. It does not extinguish the cedant's obligation toward its policyholders. In law the policy remains its own, the insureds remain its insureds, and it is the one that must pay. What it has done is exchange one risk for another: a claims risk, known and capable of being reserved, against a counterparty risk. One might judge the exchange a good one, since counterparty risk can be measured and secured. That would hold if it were independent of the first, and it is not.

This is the technical point of the lesson. The counterparty risk taken on is correlated with the peril transferred, because the acquirer fails precisely when the liabilities develop badly. A reinsurer specialized in legacy that has taken on portfolios of the same nature from several cedants sees all its acquisitions deteriorate together if the common factor materializes: a court decision widening a liability, claims inflation, a serial peril emerging. The protection therefore weakens at the exact moment it must respond, and that shape is the retrocession spiral seen earlier. An untariffed counterparty risk correlated with the peril is no better than the one ceded, it is merely somewhere else on the balance sheet and harder to see.

One drift in the product should be known to anyone reading one. These transactions tend to bear less on dead lines than on the prior years of very much living books, and the intention changes nature there. Removing a year's volatility while keeping most of its profit, through a results commission or a share of the surplus, describes an accounting effect more than a risk transfer. The question then becomes one of significant transfer: is enough genuine uncertainty actually ceded for the transaction to be reinsurance, or has nothing been bought but a presentation of the accounts and a spreading over time? Neither the auditor nor the supervisor conducts that test with any constancy, which leaves it to the cedant's board, which is rarely the body that asks it.

There remains where these transactions matter most. Long tail lines are their natural ground, certain liability classes first among them, and cyber in time, for a structural reason: the longer the interval between occurrence and final settlement, the more the reserve is an estimate and the less it is an observation. Development risk lives in that interval, and it is what these contracts buy and sell. Reading one therefore comes down to a single question in its various forms: does the price paid compensate a real probability of drift beyond the attachment, or does it mostly compensate time, that is, the discounting of reserves that will be paid ten years from now?