The last few years have witnessed an increase in the frequency and intensity of extreme weather events. From floods, to hail, to excessive wind—adverse atmospheric events are a poignant reminder of how vulnerable our society is across a broad range of threats posed by environmental extremes. Indeed, as climate change effects become more pronounced, we face a new era of risk with increased weather-related damages and losses. However, there still exists a limited number of actuarial, statistical, machine learning and climate studies that quantify and predict the impact of climate variability on the insurance industry.
Local extreme weather events cause more insurance losses overall than larger natural disasters. The evidence is provided by long-term observations of weather and insurance records that are also a foundation for the majority of insurance products covering weather related damages. Insurers around the world are concerned, however, that the current records used in the actuarial practice are likely to underestimate the future climate-induced risks, which may lead to a highly negative impact, not only on the insurance sector, but on the well being of our society as a whole.
An interdisciplinary research team led by Vyacheslav Lyubchich of the University of Maryland Center for Environmental Science includes an environmental scientist (Nathaniel Newlands), an actuarial specialist (Tahir Mahdi) both from Agriculture and Agri-Food Canada (Government of Canada), and statisticians from Boston University (Azar Ghahari) and the University of Texas at Dallas (Yulia R. Gel). In a WIREs review, the team provides an expansive overview of the state-of-the-art statistical and data science approaches for assessment of climate-induced risk in insurance.
The particular focus of this review is on atmospheric events with so-called low-individual but high cumulative impact. Such non-catastrophic events have been recently shown to bring more cumulative losses than natural disasters, both in agricultural and home insurance sectors. The literature overview indicates that there exists a noticeable interdisciplinary gap in how climate scientists, insurers, statisticians, and computer scientists address the problem of risk assessment, while collaborative efforts are still relatively scarce. Moreover, the potential of many machine learning methods remains untapped, as well as of advanced statistical techniques.
The analysis highlights rainfall as the most important factor for modeling home insurance risks. Future predictions of the risks, however, are rather uncertain. The risk forecasts vary dramatically based on statistical and climate models, as well as based on climate scenarios forcing those climate models. The point on which all forecasts agree is that there will be a general increase of home insurance risks that needs the attention of both the industry and the public.
Statistical sciences play a vital role in addressing these challenges, and this is one of the first attempts to review the currently available methods for risk assessment due to natural hazards. Such advancements offer many potential benefits for both insurers and policyholders in the design of more user targeted insurance products and claims management.
Kindly contributed by the Authors.