Short-term models overestimate losses
Models that use short-run data to predict insured losses significantly overestimate loss assumptions, according to a report by independent catastrophe risk management firm Karen Clark & Company.
The report found that short-term models used by catastrophe modellers AIR Worldwide, EQECAT and Risk Management Solutions (RMS), to estimate losses for the north Atlantic hurricanes, originally predicted loss levels at least 35 percent above the long-term average for 2006-2010.
“This latest study further supports our previous findings that a short time horizon is not sufficient for credibly estimating insured losses from hurricanes,” said Karen Clark, president and chief executive officer.
The models—introduced by the three major catastrophe modellers in 2006, following the devastating 2004 and 2005 hurricane seasons—are based on long-term average insured losses of $10 billion per year.
“Hurricane activity changes markedly from year to year, and the 2004 and 2005 seasons have proven not to be harbingers of a continuing trend,” said Clark. “Given all the uncertainties, near-term projections do not have sufficient credibility to be used for important insurance applications such as product pricing and establishing solvency standards,” she added.
The cumulative figure, which should have been $40 billion, evolved into $48.8 billion, $54.5 billion and $54.6 billion, for the AIR, EQECAT and RMS models respectively over the period.
The actual cumulative losses were $13.3 billion, “far lower than the model predictions, and only one-third of the long-term cumulative average of $40 billion”, the report states.
The report later reveals that AIR lowered its figure to approximately 16 percent in 2007, while EQECAT made “minor adjustments” to its original estimate of loss increases of between 35 and 37 percent.
RMS also introduced modifications to its model in 2009, but still predicted losses at 25 percent above the long-term average.
Near-term models are based on short-term assessments of the frequency of hurricanes, and represent a radical departure from the way in which catastrophe average annual losses and probable maximum losses are typically derived from historical data.
“Catastrophe models are powerful, broadbased tools that are very good for particular applications. However, model users must recognise that there can be a very wide range of estimates associated with a given model metric, such as average annual loss,” said Clark.
She added: “Model users should be ‘thinking outside the black box’, focusing more on understanding the range of estimates rather than automatically relying on one or two highly uncertain point estimates. This will facilitate more transparent and robust decision-making around catastrophe risk management.”
