Climate Change, Mortality and Machine Learning

The debate on climate change has rapidly evolved in recent years with the insurance industry’s focus shifting towards developing climate risk measurement techniques. While some progress has been made in the recent past, it is becoming increasingly evident that the actuarial community’s understanding of climate risk is not yet as developed as its expertise on traditional insurance risks such as mortality risk. 

Our research in this space has mainly focussed on bridging the gap between complex climate models and complex actuarial models and we aim to continue our work with the goal of leveraging our actuarial know-how as well as that from climate science and data science to discuss the following questions:

Which climate data challenges are we facing and how can we overcome them?

While the actuarial community works on developing new ways to measure the economic impact of the risk posed by global warming, it is exceedingly important for actuaries to gain an understanding of how to work with climate data, how to bridge the gap between climate models and actuarial projection models and how to produce relevant KPI.

To begin with, we take a deep dive into climate data. Taking our recent work on mortality and health risks as an example, we discuss processing of past climate data identifying relevant climate variables and statistical analysis tools. We consider different approaches available to calibrate mortality risk on historical data and discuss how climate model results can be used to estimate the impact of climate change on mortality. We also explore how machine learning techniques can be used to enhance historical data and to remove bias from the modelled climate results.

Having discussed physical risks, we move on to discuss how the current alignment of an institutional investment portfolio to Net Zero can be estimated. In particular, we consider how an insurance company can “take the temperature” of their asset portfolio. Of course, a portfolio warming assessment is a complex process requiring data improvements and a lot of expert judgement, e.g., in order to translate carbon budgets into benchmarks. However, we believe that the portfolio warming is a useful KPI and its estimation can help insurers better understand and manage climate risks inherent to their asset portfolios.