Downward counterfactual analysis – or estimating how our observed history could have been worse – is increasingly being used by the re/insurance industry to quantify the potential impacts of previously unseen catastrophic events. In the case of North Atlantic Hurricane, an example downward counterfactual could be "what would losses have been if Hurricane Matthew (2016) had followed one of its early forecast paths and made a double landfall in Florida?". While these downward counterfactuals are useful for site-specific adaptation strategies, the focus on downward (i.e., "worse-only") scenarios precludes us from assigning probabilities to the occurrence of these scenarios.

 

Here it is hypothesised that combined upward and downward counterfactual analysis (i.e., how history could have been either better or worse) may allow us to obtain novel information about historical events' probability of occurrence. To test the hypothesis, we create 10,000 counterfactual NAHU histories from unrealised reforecast data for the period 1985-2016, and compare the statistics to our observed history. While in this formative experiment the counterfactual histories show systematic under-prediction of US landfall risk, the results still point to improved understanding of relative risk of landfalls along the US coastline. It is hoped this work will allow others to attach to the concept and evolve methods appropriately to remove biases; this will ultimately help to better predict and manage the risk of high impact landfalls on the US coastline.

For more information read the open-access book chapter: