Stop loss is the third non-proportional form and the broadest of the three. Where a per-risk excess of loss slices a loss and a per-event one slices an event, stop loss slices a whole year. It triggers when the cedant's annual loss burden, all losses taken together, exceeds a threshold usually expressed as a percentage of earned premium, and it pays above that up to a ceiling also expressed in ratio points.
The notation looks like a layer's but the unit changes, and that is what makes it treacherous. A stop loss described as "30 points excess of 80%" does not talk about millions: it says the cedant carries its loss ratio up to 80% of earned premium, and that the reinsurer then takes what sits between 80% and 110%. Translating into euros requires knowing the year's premium base, which is not known when the cover is bought, and that alone is a notable difference from an ordinary layer.
What stop loss protects, no other structure protects: an accumulation of frequency. A year can be bad with no severe loss at all, simply because there were too many. A year of repeated hail, a motor account drifting a few points, a health book meeting an ordinary epidemic: none of that pierces a per-risk attachment or a per-event one, and all of it ruins an account. Stop loss is the only treaty that looks at the result rather than at the losses.
That is precisely why it is hard to place, and the difficulty is not commercial, it is structural. By covering the result, it covers indistinguishably what belongs to chance and what belongs to the conduct of the business: inadequate rating, unmanaged growth, slackening underwriting, a reserve set too late. The reinsurer cannot separate the two from outside, and it knows the cedant can tell them apart perfectly well. Moral hazard is not a theoretical risk here, it is the central subject of the negotiation.
Three answers to that recur in nearly every placement, and you recognize them by how uncomfortable they make the contract. The first is a threshold set high, well above break-even, so that the cover responds only in a genuinely abnormal year and never in a merely disappointing one. The second is a share retained by the cedant above the threshold, often ten or twenty percent, keeping it interested in containing the burden even while the treaty pays. The third is a ceiling in points, bounding the reinsurer's commitment whatever happens.
To that are added exclusions that look technical and are in fact the heart of the contract. A stop loss frequently excludes perils already covered elsewhere in the program, so as not to pay for the same loss twice, and it often excludes the drift attributable to a change of scope: a new class, an acquisition, a change in underwriting policy. Reading a stop loss without reading its exclusions is like reading a ceiling without its base.
One must finally beware of a reading that turns it into result insurance, because that is wrong in both directions. Stop loss does not guarantee a combined ratio: it looks only at the loss burden, and leaves overheads, acquisition costs and the cost of the other reinsurance treaties entirely outside. A cedant can therefore pierce its stop loss threshold and still run at a loss, or stay below it and lose money on expenses. What is bought is a bound on one precisely named component of the result, and nothing more.
A health mutual buys for 2026 a stop loss of "25 points excess of 82%", with a 10% share retained by the cedant above the threshold, on earned premium estimated at 120 million euros. The year closes with a loss burden of 118 million euros against actual earned premium of 124 million, and overheads and acquisition costs of 19% of premium. The chief executive asks whether the treaty brings the mutual back to break-even.
The first step is to recompute on the actual base rather than on the estimate, since the threshold is expressed as a percentage of earned premium as booked. With 118 million of burden against 124 million of premium, the loss ratio comes out at 95.2%. The 82% threshold corresponds to 101.7 million, and the ceiling 25 points above, that is 107%, corresponds to 132.7 million: the 118 million burden therefore falls inside the layer and does not reach its top. The excess above the threshold is 118 less 101.7, that is 16.3 million. The 10% retained share leaves 1.63 million with the mutual, and the reinsurer pays 14.67 million. After the treaty the net burden comes back to 103.3 million, a net loss ratio of 83.3%. The answer to the question asked is nevertheless no, and that is the point to explain to the board: with 19% of expenses, the net combined ratio comes out at 102.3%, so the mutual still runs at a loss despite a recovery of nearly fifteen million. The stop loss did exactly what it promises, bounding the loss component, and none of what is attributed to it, guaranteeing a result. It remains to check before notifying that none of the drift comes from a change of scope during the year, most contracts excluding that part.
- 01Stop loss slices a whole year and triggers on the annual loss burden, expressed as a percentage of earned premium.
- 02It is the only treaty that looks at the result rather than at the losses, therefore the only one that protects against an accumulation of frequency.
- 03By covering the result it covers chance and the conduct of the business indistinguishably: moral hazard is the central subject of its negotiation.
- 04The three usual answers are a threshold set high, a share retained by the cedant above it, and a ceiling in points.
- 05Its exclusions are the heart of the contract: perils already covered elsewhere, and drift attributable to a change of scope.
- 06It guarantees no combined ratio: overheads, acquisition costs and the cost of the other treaties all stay outside.