Non-proportional reinsurance
Per-risk excess of loss, stop loss and mass lapse reinsurance
What is non-proportional reinsurance?
In a non-proportional treaty, the reinsurer pays only the amount above an agreed point, up to a limit. A per-risk excess of loss applies this to each single claim. A stop loss applies it to the total claims for the year. Mass lapse reinsurance works like a stop loss on lapses instead of claims.
Per-risk excess of loss
The reinsurer pays the part of each single claim above a retention, up to a limit.
How it works
- In the example, the expected claims of 100 are three claims of 10, 30 and 60. A layer of 20 above 20, meaning a retention of 20 and a limit of 20, pays 0, 10 and 20, so 30 in total
- That is the same 30 as the quota share, but all of it comes from the two larger claims
More details
| Item | Detail |
|---|---|
| Covers | Each single claim between the retention and the retention + limit |
| Recovery | Depends on the size of each claim |
| Premium | A rate on the subject premium of the cedant, such as its gross net premium income (GNPI), paid in advance |
| Commission or allowance | Usually none |
| Term | Usually one year |
Cash flows
Large claims take time to settle. In the example, 40% of the recoveries are paid in year 1 and 60% in year 2. The premium of 33 is paid in advance.
| Year | Premium paid | Recoveries received | Net cash flow to the cedant |
|---|---|---|---|
| 1 | 33 | 40% × 30 | 12 − 33 |
| 2 | – | 60% × 30 | 18 |
| Total | 33 | 30 | −3 |
Pricing
The reinsurer estimates the expected loss in the layer from restated past claims (experience rating), from the risk profile (exposure rating), or both. It adds the time value of money, expenses and a risk load (Clark 1996). The premium is usually a rate on the subject premium, which is the premium the cedant collects on the covered business. In the example, the premium of 33 is the expected recovery of 30 plus a loading of 3, which is 33 ÷ 90 = 36.7% of the subject premium of 90.
IFRS 17
| Item | Treatment |
|---|---|
| Scope | In scope if the risk transfer is significantIFRS 17.B18–B19 |
| Measurement model | General model, or PAA if eligibleIFRS 17.69. One-year cover is eligible, but the PAA is optional |
| Commission or allowance | Usually none |
| Contract boundary | Usually one year |
| Watch out for | Large claims settle slowly, so discount the recoveriesIFRS 17.36 |
How is the recovery % calculated?
The recovery % is the share of the underlying claims that the reinsurer is expected to pay. The LoRC is the underlying loss × the recovery %.
In the example, the recovery % = 30 ÷ (10 + 30 + 60) = 30%, so the LoRC = 20 × 30% = 6.
Stop loss
The reinsurer pays the part of the total claims for a year above an attachment point, up to a limit. The attachment point works like the retention of an excess of loss, but on the total for the year.
How it works
- In the example, the total claims are 60, 100 or 160, with probabilities of 30%, 50% and 20%, so the expected claims are 100. A layer of 50 above 110 pays only in the bad year, so the expected recovery = 20% × 50 = 10
- Apply the layer in each scenario, then weight the results by probability. Applied to the expected claims of 100, the layer would pay 0
| Scenario | Probability | Total claims | Recovery | Probability × recovery |
|---|---|---|---|---|
| Good year | 30% | 60 | 0 | 0 |
| Normal year | 50% | 100 | 0 | 0 |
| Bad year | 20% | 160 | 50 | 10 |
| Expected | 100% | 100 | 10 |
More details
| Item | Detail |
|---|---|
| Covers | The total claims for a year between the attachment point and the attachment point + limit |
| Recovery | The part of the total above the attachment point |
| Premium | A fixed amount, paid in advance |
| Commission or allowance | None |
| Term | Usually one year |
Here s is a claims scenario. A stop loss pays only above the attachment point, so apply the formula in each scenario and weight the results by probability.
Cash flows
The total claims for the year are known only after the year end, so the recovery is paid in year 2. The premium of 12 is paid at the start of the year.
| Year | Premium paid | Expected recoveries | Net cash flow to the cedant |
|---|---|---|---|
| 1 | 12 | – | −12 |
| 2 | – | 10 | 10 |
| Total | 12 | 10 | −2 |
Pricing
The reinsurer models total claims with an aggregate model and adds loadings for expenses and risk (Clark 1996). In the example, the premium of 12 is the expected recovery of 10 plus a loading of 2.
IFRS 17
| Item | Treatment |
|---|---|
| Scope | In scope if the risk transfer is significantIFRS 17.B18–B19 |
| Measurement model | General model, or PAA if eligibleIFRS 17.69. One-year cover is eligible, but the PAA is optional |
| Commission or allowance | None |
| Contract boundary | Usually one year |
| Watch out for | Weight the recovery in each scenario by its probabilityIFRS 17.B38–B39. The LoRC cap can apply after claims increase (see Cap and release) |
How is the recovery % calculated?
The recovery % is the share of the underlying claims that the reinsurer is expected to pay. The LoRC is the underlying loss × the recovery %.
In the example, the recovery % = 10 ÷ 100 = 10%, so the LoRC = 20 × 10% = 2.
What happens at the next reporting date?
At each reporting date, the LoRC can be at most the underlying loss component × the current recovery %. When the estimates change, this cap can fall below the LoRC, and the cedant then reduces the LoRC to the cap.
- The bad year becomes more likely, 30% instead of 20%. The expected claims rise to 110, so the loss component rises to 30, and the expected recoveries rise to 15. The increase of 5 goes to profit or loss and takes the LoRC from 2 to 7IFRS 17.66(c)(i)
- The cap = loss component × new recovery % = 30 × (15 ÷ 110) = 30 × 13.6% = 4.09IFRS 17.B119F. So the cedant takes 2.91 of the 7 off the LoRC against the held CSMIFRS 17.66(bb)
Opening LoRC 2 rises by 5 to 7 before the cap. A reduction of 2.91 brings closing LoRC to 4.09.
- OpeningLoRC2
- Recoveriesup 10 → 15+5
- Beforethe cap7
- Cap(66(bb))−2.91
- ClosingLoRC4.09
Example: The stop loss LoRC
The underlying contract group has a loss component of 20.
- Recovery % = expected recoveries ÷ expected claims = 10 ÷ 100 = 10%
- Loss-recovery income = 20 × 10% = 2
- Applying the recovery formula to the expected claims of 100 instead would give 0
Example: The stop loss after one period
The bad year becomes more likely: 30% instead of 20%, and the good year 20% instead of 30%.
| Scenario | Probability | Total claims | Recovery | Probability × claims | Probability × recovery |
|---|---|---|---|---|---|
| Good year | 20% | 60 | 0 | 12 | 0 |
| Normal year | 50% | 100 | 0 | 50 | 0 |
| Bad year | 30% | 160 | 50 | 48 | 15 |
| Expected | 100% | 110 | 15 |
- New recovery % = 15 ÷ 110 = 13.6%
- Increase in expected claims = 110 − 100 = 10
- The whole increase in expected claims goes to the underlying loss component, so it becomes 20 + 10 = 30
- Increase in expected recoveries = 15 − 10 = 5, in profit or lossIFRS 17.66(c)(i)
- LoRC before the cap = 2 + 5 = 7
- Cap = 30 × 15 ÷ 110 = 4.09IFRS 17.B119F
- LoRC = 4.09. The excess of 2.91 is an expense in profit or loss, and the net-cost CSM decreases by 2.91IFRS 17.66(bb)
- Net income in profit or loss = 5 − 2.91 = 2.09
After the premium of 12 is paid, the held asset is the expected recoveries of 10 plus the net-cost CSM of 2 + 2 = 4, so 14. The LoRC of 2 is part of it.
| Movement | ARC excluding LoRC | LoRC | Total |
|---|---|---|---|
| Opening balance | 12.00 | 2.00 | 14.00 |
| Recovery of losses: recoveries 10 → 15IFRS 17.66(c)(i) | – | 5.00 | 5.00 |
| Cap of 30 × 15 ÷ 110 = 4.09: CSM adjustmentIFRS 17.66(bb) | – | (2.91) | (2.91) |
| Closing balance | 12.00 | 4.09 | 16.09 |
Estimates for a real portfolio
For a real portfolio, three methods give an estimate without a full stochastic modelIFRS 17.B38–B39:
- Fitted distribution: Calculate the mean and variance of total claims from mortality and sums at risk. Fit a normal or lognormal distribution and calculate the expected layer payment
- Burning cost: Restate past years of total claims to today's exposure. Apply the treaty to each year and take the average
- Reinsurer's pricing: Use the expected loss cost of the layer from the reinsurer as a cross-check
Mass lapse reinsurance
A lapse is a policyholder canceling a policy before it ends. A mass lapse is when many policyholders do this in the same year, and the cedant loses the future profit on those policies. Mass lapse reinsurance works like a stop loss on lapses. When the lapse rate in a year exceeds an attachment point, the reinsurer pays for the lost profit.
How it works
In the example, the cover is on 10'000 policies and pays 0.2 for each lapse between a lapse rate of 20% and 50%. For most other treaties, the premium is close to the expected recoveries. Cedants buy mass lapse cover mainly for capital relief, so its premium of 15 is almost double the expected recoveries of 8. Most of the difference pays for the reinsurer's capital.
Premium 15 consists of expected recoveries 8, cost of capital 6, and expenses 1.
- Expectedrecoveries8
- Cost ofcapital+6
- Expenses+1
- Premium15
| Part of the premium | Calculation | Amount |
|---|---|---|
| Expected recoveries | Weighted over the lapse scenarios | 8.0 |
| Cost of capital | Capital 150 × 4% | 6.0 |
| Expenses | Assumption | 1.0 |
| Premium | 2.5% rate on line × limit of 600 | 15.0 |
More details
| Item | Detail |
|---|---|
| Covers | The lapses in a treaty year between the attachment rate and the limit rate |
| Recovery | Excess lapses × an agreed loss per lapse. The loss per lapse is a table in the treaty, often by policy year, that approximates the lost profit or the unrecovered acquisition cost |
| Premium | A rate on line, which is a percentage of the limit, paid in advance |
| Commission or allowance | None |
| Term | Short, often one or two years |
Here s is a lapse scenario. As for a stop loss, apply the formula in each scenario and weight the results by probability.
Scenarios and cash flows
The cover is on 10'000 policies. The attachment point is a lapse rate of 20%, the limit is a lapse rate of 50% and the loss per lapse is 0.2.
| Scenario | Probability | Lapse rate | Excess lapses | Recovery | Probability × recovery |
|---|---|---|---|---|---|
| Normal | 97.0% | 8% | 0 | 0 | 0 |
| Stress | 2.5% | 30% | 10% × 10'000 | 1'000 × 0.2 | 2.5% × 200 |
| Extreme | 0.5% | 60% | (50% − 20%) × 10'000 | 3'000 × 0.2 | 0.5% × 600 |
| Expected | 100% | 8.0 |
The premium of 15.0 is paid at the start of the year. The lapse rate for the year is known only after the year end, so the recovery is paid in year 2.
| Year | Premium paid | Expected recoveries | Net cash flow to the cedant |
|---|---|---|---|
| 1 | 15.0 | – | −15.0 |
| 2 | – | 8.0 | 8.0 |
| Total | 15.0 | 8.0 | −7.0 |
Pricing
When the treaty is written, a mass lapse is unlikely, so the expected recoveries are small (Milliman 2021). The price is a rate on line (Gallagher Re) and rises sharply as the attachment point falls. The Solvency II mass lapse stress is 40% (EIOPA 2025), and treaties attach at or below that level. The example attaches at 20%. Cedants compare the price with the capital that the cover releases. One presentation gives an equivalent cost of capital of 0.7% to 2.5% (Gallagher Re). The 2.5% in the example is the rate on line, which measures something else.
The reinsurer's premium covers the expected recoveries, its expenses and a return on capital for the tail event (Clark 1996, section 6). Illustrative assumptions:
- Capital: 150, after diversification
- Cost of capital: 4% a year. The Solvency II risk margin uses 6% (Solvency II Delegated Regulation, Article 39), which the 2025 review lowers
- Expenses: 1.0 a year
The limit is the extreme-scenario recovery of 600, so the rate on line is 15.0 ÷ 600 = 2.5%. Only 8.0 of the premium pays for expected recoveries. The rest pays for capital and expenses.
IFRS 17
| Item | Treatment |
|---|---|
| Scope | Usually reinsurance heldIFRS 17.B15. A cover that pays a fixed amount per lapse, with no link to the actual loss, can be a derivative under IFRS 9 |
| Measurement model | General model, or PAA if eligible (69) |
| Commission or allowance | None |
| Contract boundary | Usually the treaty term (34) |
| Watch out for | The risk adjustment is large compared with the expected recoveries, because the value of the cover lies in rare extreme years. Project the premiums per scenario if they depend on the business in force |
How is the recovery % calculated?
The recovery % is the share of the underlying claims that the reinsurer is expected to pay. The LoRC is the underlying loss × the recovery %.
In the example, the treaty recovers no claims, so the recovery % = 0 ÷ 100 = 0% and the contract group gets no LoRC from itIFRS 17.B119D.
A change in the expected recoveries is treated the same way even without an initial LoRCIFRS 17.66(c)(i). If a change in lapse assumptions goes to the underlying loss component and also changes the expected recoveries, the change on the reinsurance held goes to profit or loss too, which reduces volatility. The standard then requires a LoRCIFRS 17.66B within the capIFRS 17.B119F. It is unresolved how that cap works with a recovery % of 0%.
Comparison
The three treaties differ in what triggers a payment, and so in the recovery % for the LoRC.
| Item | Per-risk excess of loss | Stop loss | Mass lapse reinsurance |
|---|---|---|---|
| Pays for | Each single claim above a retention | Total claims above an attachment point | Excess lapses × loss per lapse |
| Expected recovery from | Claim sizes and numbers | Claims scenarios | Lapse scenarios |
| Scope | Reinsurance held, if the risk transfer is significant | Reinsurance held, if the risk transfer is significant | Usually reinsurance held |
| Recovery %IFRS 17.B119D | 30% in the example | 10% in the example | 0%, because the cover recovers no claims |
| Initial LoRC on an onerous contract group | Loss component × recovery %: 6 | Loss component × recovery %: 2 | None |
| Later changesIFRS 17.66(c)(i) | Apply | Apply | Apply, but the LoRCIFRS 17.66B is unresolved |