Home / Wiki / C / Catastrophe Probable Maximum Loss (PML)

Catastrophe Probable Maximum Loss (PML)

Actuarial and Risk SciencesUpdated 17 Nov 2026

Catastrophe probable maximum loss (PML) is the loss an insurance or reinsurance portfolio is expected to suffer from catastrophe events at a specified level of rarity, expressed as a return period such as 1-in-100 or 1-in-250 years. It is read from the output of a catastrophe model, which simulates many thousands of possible years of hurricanes, earthquakes, floods or other perils against the portfolio's insured exposures and ranks the resulting losses.

The PML is the main language in which catastrophe risk is discussed. Underwriters use it to control accumulations, risk managers to set limits under the risk appetite, buyers to size catastrophe reinsurance, rating agencies and supervisors to assess capital, and investors in catastrophe bonds to understand the risk they are taking. The term is older than modern models and once meant an underwriter's judgemental estimate of the largest likely loss to a single building; in catastrophe work it now refers to a modelled point on an exceedance probability curve.

01

Return periods and exceedance probability

A return period is the inverse of an annual exceedance probability:

T=1pT = \frac{1}{p}
  • TT is the return period in years.
  • pp is the annual probability that losses exceed the PML amount.

A 1-in-100 PML of 500 million therefore means the model estimates a 1÷100=11 \div 100 = 1 per cent chance each year that losses will exceed 500 million. A 1-in-250 PML corresponds to a 0.4 per cent annual probability. In statistical terms, the 1-in-100 PML is the 99th percentile, or quantile, of the modelled annual loss distribution. A return period does not mean the loss happens once a century; it can occur in two consecutive years or not at all over several centuries.

02

OEP and AEP bases

PMLs are quoted on one of two bases:

A related metric is tail value at risk (TVaR), the average loss in all simulated years beyond a given return period, which captures the severity of the tail rather than a single point.

03

Use in capital and reinsurance buying

Under Solvency II in the European Union and its UK successor, Solvency UK, the solvency capital requirement is calibrated to a 99.5 per cent one-year value at risk, which corresponds to a 1-in-200 return period, so the 1-in-200 PML is a key input for internal models. Rating agencies and some regulators refer to 1-in-100 and 1-in-250 PMLs when assessing capital adequacy, and Lloyd's monitors syndicates' modelled losses alongside scenario tests.

Insurers commonly buy catastrophe reinsurance up to a chosen return period, for example the 1-in-200 or 1-in-250 OEP PML, with the attachment point set around a more frequent return period that the company can absorb in earnings. The PML curve is also used to price layers: the expected loss to a layer, compared with its premium, gives a view of whether the rate on line is adequate.

04

Limitations

A PML is only as good as the model and the exposure data behind it. Several sources of error are widely recognised:

  • Model uncertainty. Different vendor models can produce materially different PMLs for the same portfolio, and both the frequency and the damage assumptions are uncertain, as discussed under catastrophe model uncertainty.
  • Data quality. Missing or inaccurate locations, construction types and insured values distort results.
  • Non-modelled perils and loss amplification. Wildfire, flood, cyber or contingent business interruption may be poorly modelled, and post-event loss amplification can push actual losses above modelled ones.
  • Climate trends. Models calibrated on historical records may not reflect a changing hazard, a question debated in connection with climate change.

For these reasons, PMLs are usually supplemented by deterministic scenarios and stress tests, and treated as estimates with a range rather than precise figures. Many companies also adjust vendor model output with their own loadings, for example for non-modelled perils, and document these adjustments under model governance so that the PML used for capital and reinsurance decisions can be explained and reproduced.

Last updated 17 Nov 2026