With excess of loss, both the family and the logic change. Proportional treaties share a risk along a fraction fixed in advance; non-proportional treaties share nothing at all, they slice the amount of the loss into tranches and assign each tranche to someone. The cedant pays everything up to a threshold, the reinsurer pays above it up to a ceiling, and the cedant takes over again beyond that if nobody has bought the next tranche.
Two numbers describe a layer and they must be kept apart, because the market notation runs them together. A layer written "10 excess of 5" carries an attachment of 5 and a limit of 10: it begins at 5 and stops at 15. The attachment, also called the attachment point, is what the cedant keeps on every loss; the limit is what the reinsurer agrees to pay above it. A loss of 12 pierces the layer for 7. A loss of 4 never touches it. A loss of 30 exhausts it at 10 and leaves 15 above, which fall on the cedant if no higher layer exists.
A program is therefore built by stacking layers, and that stacking is not an accounting convenience: each layer is priced, placed and renewed separately, often with different reinsurers. Low layers see losses every year, sometimes several a year; high layers almost never do. That difference in frequency is so structural that it produced two trade names. The working layer is the one hit regularly; the cat layer is the one reached only by the rare event.
This vocabulary is defined by what a layer sees, never by its rank in the program. A first layer placed far above a heavy retention can perfectly well be a cat layer, and the third layer of a program with tight tranches can still be a working layer. The distinction drives how you price: a working layer is priced on experience, because there are losses to count; a cat layer is priced on an exposure model, because there are not.
A layer exhausts, and that is the point descriptions most often leave out. If the 10 limit has been consumed by a first loss, it is no longer available for the second, which will fall entirely on the cedant. The reinstatement clause exists for that: it makes the limit available again after a loss, a set number of times. A treaty described as carrying two reinstatements therefore offers the limit three times over the year, one original and two reinstated, and no more.
Reinstatement is paid for, and its formula is a constant source of error. It is almost always computed pro rata to the amount consumed, and frequently pro rata to the time left to run as well. Consuming half a limit then costs half a reinstatement premium, itself often equal to the layer's original premium; and if the loss occurs mid-year, half again. A cedant that budgets its reinsurance premiums without providing for reinstatements discovers its real spend on the day it can least afford it.
What is left is what an excess of loss does not do, and it mirrors quota share. It does not reduce volume, it does not release capital in the way a proportional cession does, and it commissions nothing, since there is no pro rata ceded premium but a premium of the layer's own. What it buys is exclusively protection against severity, and it buys it expensively per unit of capital protected, because the reinsurer is selling a rare exposure whose cost it cannot offset with volume.
A cedant buys for 2026 a layer of "8 million euros excess of 2 million", with one reinstatement, payable at 100% pro rata to the amount consumed and with no time pro rata. The layer premium is 600,000 euros. On February 20, a 6 million euro loss hits it. On September 14, a second loss of 9 million euros occurs. What does the cedant recover on each, and what does the layer cost it over the year?
The first loss of 6 million exceeds the 2 million attachment, so the layer pays 4 million and the cedant keeps its first 2 million. The 8 million limit is eaten into by 4, leaving 4 available immediately. The reinstatement triggers on the amount consumed: 4 million out of an 8 million limit, that is one half, therefore half a reinstatement premium equal to 100% of the original premium, that is 300,000 euros. The limit returns to a full 8 million. The second loss of 9 million exceeds the 2 million attachment, so the part exposed to the layer is 7 million, below the 8 available: the layer pays 7 million and the cedant keeps 2 million. That second consumption would in turn call for a reinstatement, but the treaty's single reinstatement was already used in February: the limit cannot be reinstated again and the layer is now down to 1 million for the rest of the year. Over the year the cedant recovered 11 million euros, retained 4 million through the two attachments, and the layer cost it 900,000 euros, being the 600,000 premium plus the 300,000 reinstatement. The point to carry forward is not the recovery, it is what is left: a third serious loss would find an almost empty layer, and the cedant carries that exposure with no warning other than its own reading of the treaty.
- 01A non-proportional treaty does not share a risk: it slices the loss amount into tranches and assigns each one to someone.
- 02"10 excess of 5" means an attachment of 5 and a limit of 10, so a layer running from 5 to 15.
- 03Working layer and cat layer are defined by what the layer sees, never by its rank in the program.
- 04A limit exhausts: with no reinstatement, the next loss falls entirely on the cedant.
- 05Two reinstatements mean the limit three times over the year, no more, and each is paid pro rata to the amount consumed.
- 06Excess of loss buys protection against severity alone: it does not reduce volume and it commissions nothing.