My short article, when urban flood models disagree: hydrological equifinality as actuarial model risk, was recently published in the European Actuarial Journal. In a nutshell, I tried to address the question “how much does a flood that has not happened yet cost?”
For insurers, answering that question involves a surprisingly long chain: rainfall has to become runoff, runoff has to become water in streets and buildings, damage has to be estimated, and finally translated into insured losses. Here, I focus on something hydrologists have known for a long time: several perfectly plausible models can fit the same observations and still tell quite different stories about what might happen next, this is known as equifinality, and I argue that actuaries should think of it as a form of model risk.
The issue is especially important in cities, where drainage systems, sewers, micro-topography, flood defences and highly concentrated property values can make small modelling choices financially significant. Two reasonable flood models may therefore lead to different premiums, underwriting decisions, reinsurance needs, capital requirements, or even different conclusions about whether a prevention measure is worth investing in. So uncertainty is not simply about whether a “100-year flood” might occur; it is also about the modelling chain through which water eventually becomes an insurance claim. My main argument is that insurers should make this disagreement visible, by comparing plausible models and asking whether their decisions remain robust across them. We need to price flood risk, of course, but we should also price, or at least acknowledge, our uncertainty about the models we use to describe it.