How Much Catastrophe Reinsurance to Buy: PML and Ratings

How Much Catastrophe Reinsurance Should an Insurer Buy? Return Periods, PML and Rating-Agency Tests

Last reviewed: November 2026

How much catastrophe reinsurance to buy depends on one question: which capital test binds first, and at what return period. For most insurers the answer sits between the 1-in-100 and the 1-in-250 modelled loss, adjusted for model uncertainty and reinstatement costs.

The limit is not a single number pulled from a cat model. It is a board decision that balances earnings volatility, regulatory capital, rating-agency capital and the price of the top layers.

This guide is for CFOs, CROs, chief actuaries and heads of outwards reinsurance. It explains return periods, OEP and AEP, the main capital tests, and a worked example that compares buying to 1-in-100, 1-in-200 and 1-in-250.

Key Takeaways

• The top limit of a cat programme should match the return period of the capital test that binds, plus a buffer for model error.
• Solvency II is calibrated to a 99.5% one-year Value-at-Risk, which is broadly a 1-in-200 annual loss.
• AM Best's BCAR uses net per-occurrence PMLs at 1-in-100, 1-in-200 and 1-in-250 for its higher confidence levels.
• Net PML should include reinstatement premiums, not just the retained loss.
• OEP sizes the limit, but AEP shows whether several retained events can hurt more than one big one.
• High layers are usually cheap per unit of capital protected, so the last 1-in-200 to 1-in-250 slice is often the best value in the programme.

What Determines How Much Catastrophe Reinsurance to Buy?

Three things set the amount: the retention you can absorb, the return period your binding capital test uses, and the margin you add for model uncertainty. Price matters, but it usually decides layer structure more than the top limit.

The retention is an earnings question. The top limit is a capital and solvency question.

Many programmes go wrong because one person owns the retention and another owns the limit. The CFO, CRO and head of outwards should agree both against the same modelled loss curve.

What Do Return Periods, PML, OEP and AEP Mean?

A return period expresses how rare a loss is, so a 1-in-200 loss has a 0.5% chance of being exceeded in any year. The PML at that return period is the modelled loss at that point on the curve.

Occurrence Exceedance Probability (OEP)

OEP measures the largest single event loss in a year. Cat excess of loss is bought per occurrence, so the OEP curve is the natural basis for setting retentions and limits.

Aggregate Exceedance Probability (AEP)

AEP measures the total of all event losses in a year. It is always equal to or higher than the OEP at the same return period, and it matters wherever capital is measured on an annual basis.

Gross, Net and "Net Including Reinstatements"

The gross PML is the loss before reinsurance. The net PML is the retained loss after recoveries, and the most useful version also adds the reinstatement premiums triggered by the event.

Which Capital Tests Set the Top of the Programme?

The top of the programme should be set by whichever test binds first for your company. The table below summarises the main tests and how they treat catastrophe risk.

Solvency II

Article 101(3) of the Solvency II Directive calibrates the SCR to the Value-at-Risk of basic own funds at a 99.5% confidence level over one year. Article 101(5) lets firms reflect risk-mitigation techniques, provided the credit and other risks they create are also captured.

Standard formula firms feel cat reinsurance through the natural catastrophe sub-module scenarios. Internal model firms see it directly in their net annual loss distribution, which is an AEP view.

AM Best BCAR

AM Best's Understanding Universal BCAR criteria procedure adds modelled catastrophe losses to required capital at each confidence level. The net PML is taken from the per-occurrence, all-perils combined curve, net of reinsurance and including reinstatement premiums.

The criteria map return periods to confidence levels: 1-in-20 at 95%, 1-in-100 at 99%, 1-in-200 at 99.5% and 1-in-250 at 99.6%. The "Strongest" BCAR assessment requires a score above 25 at the 99.6% level, so the 1-in-250 net PML carries real rating weight.

AM Best publishes jurisdiction-specific versions of its BCAR criteria and updates them periodically. Confirm the version that applies to your rating unit before relying on the detail.

Other Rating Agencies and Regulators

Other agencies also test capital at several confidence levels, with catastrophe charges that rise with the severity of the stress. Because criteria change, read the current published methodology for your target rating rather than relying on older rules of thumb.

How Should the Retention Be Set?

The retention should equal the event loss the insurer can absorb without breaching its earnings tolerance. A common expression is a share of planned annual profit or a small percentage of equity.

Three checks keep the retention honest:

1. Frequency test. Model how many events in a bad year could each land just under the retention.
2. Net-of-inuring test. Calibrate the retention after per risk and quota share recoveries, not on gross losses.
3. Reinstatement test. Add the reinstatement premiums triggered by a retention-level event to the earnings impact.

How exposure data and models feed this decision is covered in our guide to catastrophe modelling infrastructure.

Worked Example: Buying to 1-in-100, 1-in-200 or 1-in-250

The example below is illustrative and simplified. All figures are in USD millions, and the cedent has equity of 400.

Gross per-occurrence (OEP) PML curve, all perils:

Layers available, each with one reinstatement at 100% pro rata as to amount:

The cedent compares three options. Option A buys L1 and L2 to the 1-in-100, Option B adds L3 to the 1-in-200, and Option C adds L4 to the 1-in-250.

Reinstatement premiums equal each layer's premium multiplied by the share of its limit used. In the 1-in-250 event every layer below the top is fully exhausted, so the cedent pays a full reinstatement on each.

What each extra slice buys:

Moving from A to B costs 2.45 a year and removes 67.55 from the net 1-in-200 and 1-in-250 losses. Moving from B to C costs 0.75 and removes a further 29.25 from the 1-in-250, which is the point AM Best uses for its 99.6% level.

Why the Annual Aggregate View Still Matters

Now test Option B against a year with two events of 150 and 120. Neither event comes close to the 1-in-200 OEP, yet the year is expensive.

The two-event year costs 95.70 net, against 60.95 for a single 1-in-200 event. An internal model built on annual losses will see this, even though the per-occurrence limit looks ample.

The lesson for cedents: size the top limit on the OEP curve, then test the retention and reinstatement provisions on the AEP curve. If the aggregate view binds, a lower retention, an aggregate cover or an extra reinstatement may protect capital better than more top limit.

The lesson for reinsurers: a cedent that buys to the 1-in-250 but holds a high retention may still show capital strain in a frequency year. Ask for both curves before you quote.

When Should You Buy to 1-in-100, 1-in-200 or 1-in-250?

The right top limit depends on which test binds and how much you trust the model. The table maps common situations to the usual choice.

These are starting points. The final choice should come from testing net results at every relevant return period, not from habit.

Which Mistakes Lead Insurers to Buy the Wrong Amount?

Most cat buying errors come from measuring the wrong loss, not from buying the wrong layer. Each of the mistakes below is visible before renewal if you look for it.

What Should Decision-Makers Ask Before Fixing the Limit?

Fix the limit only after the finance, risk and outwards teams have the same data and agree which test binds. The data request and questions below make that practical.

What Data Should the Buying Team Assemble?

Which Questions Should the Board Ask?

• Which test binds today: Solvency II, BCAR, another rating agency or the local regulator?
• What is the net PML at that return period, including reinstatement premiums?
• How much did the limit change because of model updates rather than exposure growth?
• What does the worst plausible frequency year look like under the current retention?
• What does each extra slice of limit cost per unit of capital it protects?

The rating-agency side of these questions is covered in more depth in what the chief actuary should challenge about rating-agency capital.

How Does the Purchase Affect Capital and Reporting?

Cat reinsurance reduces the capital needed for catastrophe risk but adds counterparty exposure and a reinsurance cost. Both effects should be modelled together.

Under Solvency II, the counterparty default risk module charges for the reinsurance recoverables the programme creates. Rating agencies also charge for recoverables according to each reinsurer's credit quality, so panel security affects how much capital the purchase truly saves.

Under IFRS 17, reinsurance contracts held are measured separately from the underlying contracts. Reinstatement premiums and recoveries therefore need to be captured accurately in both the treaty calculation and the reporting.

For a broader view of how reinsurance moves regulatory capital, see our guide to solvency relief through reinsurance.

What Does the Market Look Like for 2027 Renewals?

Property catastrophe capacity has been plentiful through 2026, and pricing has softened from the peaks that followed the 2023 reset. That gives buyers more room to extend limit or add aggregate protection at the 2027 renewals.

Softer pricing changes the economics of the top layers most. When the last slice to 1-in-250 is cheap, the case for stopping at 1-in-200 becomes harder to defend.

Buyers should still separate price from structure. A cheaper programme that leaves the binding test exposed is not a better programme.

Checklist for CFOs, CROs and Heads of Outwards

Use this checklist before you fix the limit and retention for the next renewal.

• The binding capital test and its return period are identified and agreed
• The net PML at that return period includes reinstatement premiums
• OEP and AEP curves are both reviewed, gross and net of inuring covers
• An explicit uncertainty load is applied for secondary perils and data quality
• The retention is tested against a frequency year, not only a single event
• The cost of each extra slice of limit is compared with the capital it protects
• Panel security is reflected in the capital benefit of the purchase
• Layers, reinstatements and panel shares sit in one controlled treaty system

For the last point, a reinsurance treaty management platform holds layer terms, reinstatement provisions and premium calculations in one place.

Key Terms

Return period: The average interval between losses of a given size, so a 1-in-200 loss has a 0.5% annual chance of being exceeded.

PML (probable maximum loss): The modelled loss at a chosen return period. It can be stated gross, net, or net including reinstatement premiums.

OEP (occurrence exceedance probability): The probability that the largest single event in a year exceeds a given loss. It is the usual basis for sizing cat XoL limits.

AEP (aggregate exceedance probability): The probability that the total of all event losses in a year exceeds a given loss. It matters for annual capital measures.

Exhaustion point: The top of the cat programme, equal to the retention plus the total limit purchased.

Reinstatement premium: The additional premium paid to restore a layer's limit after a loss, usually pro rata as to amount.

BCAR: AM Best's Capital Adequacy Ratio, which compares available capital with net required capital at several confidence levels.

Value-at-Risk (VaR): The loss that will not be exceeded at a given confidence level over a set period, such as 99.5% over one year.

Sources

• Directive 2009/138/EC (Solvency II), EUR-Lex
• AM Best: Understanding Universal BCAR (criteria procedure)
• IFRS 17 Insurance Contracts, IFRS Foundation
• EIOPA: Solvency II resources