How reinsurers can make up for poorly modelled perils

PartnerRe is championing an integrated risk management and modelling method designed to take into account the cumulative effect of multiple perils, some of which cannot be modelled reliably, including floods and hail.
“Not every type of risk and peril can be reliably modelled, PartnerRe says in its report, Catastrophe Portfolio Modeling: A Complete View of Risk. “Catastrophe models frequently do not model secondary perils, such as hail, snow pressure or even flood, which may add to the loss burden in bad years.”
Also, modelling in and of itself “may not be suitable for all types of reinsurance contract, such as those with a complex structure,” PartnerRe says.
The reinsurer goes on to note: “modelled probable maximum losses (PMLs) have consistently failed to adequately reflect actual catastrophe losses.”
PartnerRe’s approach is to combine traditional modelling with a risk management program that includes the assessment of the loss of capital from natural catastrophes from both a single large event and multiple, but perhaps smaller, events in any one year.
At the single-event level, this is done by imposing an absolute limit or “exposure cap” on regional zones and perils, or so-called “peril zones” (e.g. North East U.S. hurricane).
“The zones are defined so that no single event would substantially erode the aggregate limit from more than one zone,” PartnerRe says.
“This method of accumulation or exposure control is a conservative one that relies on limits and not on the likelihood of those limits being exhausted [i.e. as in the use of PMLs].
“Since it does not rely on catastrophe modelling assumptions, it provides a certainty that our financial strength cannot be impaired beyond the accepted limits.”
The method also takes into account the correlation of losses between peril zones.
In addition, detailed pricing estimates gathered from alternative actuarial analyses can be used to supplement modelling in situations in which poor data or complicated program structures make contracts unsuitable for catastrophe modelling. This can also be done when the perils are either not covered by or are poorly estimated by the catastrophe models.
PartnerRe’s full report is available at: http://www.partnerre.com/