Darryl Brundle and Ashley Kanter on sidestepping mistakes in temperature and mortality modelling
Of the deaths related to heat exposure from 1991–2018, 37% have been attributed to global warming caused by mankind (bit.ly/HeatDeath_Climate). A warmer world will change society in ways that are hard to assess. It may lead to us reducing meat consumption, or requiring more energy as air conditioning becomes a necessity. Our lives will be impacted by both the immediate effects of climate change and the governmental and societal responses to these changes.
This complex feedback loop of change and response is a challenge to insurers and pension schemes, which face mortality and longevity risk. As we assess the impact of different climate scenarios on mortality, actuaries need to capture the right data and consider the challenges involved in untangling the interaction between weather and mortality. How do we start, and what pitfalls should we avoid?
“There are four mistakes that are easy to make when modelling temperature and mortality”
How are temperature and mortality linked?
Around the world we see a broadly U-shaped relationship between temperature and mortality – extreme heat and extreme cold lead to higher mortality risk, while mortality rates are much lower at moderate temperatures (Figure 1). If climate change simply resulted in an overall temperature increase, you might see a reduction in cold weather deaths and an increase in warm weather deaths. However, this hides a lot of complexity that will need to be factored into any risk modelling.
There are four mistakes that are easy to make when modelling temperature and mortality:
- Climate change does not just mean warmer weather – increases in average global temperature are expected to result in more volatile weather. This means the balancing act between hot and cold includes a third element to consider when assessing links to mortality rates.
- Correlation is not causation – the U-shaped relationship should not be taken to mean that the temperature itself is always responsible for changing mortality. In the UK, for instance, deaths in winter are generally driven by higher flu rates, rather than cold weather itself. While temperature and mortality are clearly linked, simply modelling that relationship does not fully examine the underlying driver behind it.
- The temperature-mortality relationship is not necessarily fixed – while it’s convenient to have a simple U-shaped relationship to start with, the shape of the curve represents the present relationship, not necessarily the future one. Governments and society can drive actions that will change the relationship. For example, countries that regularly experience very cold weather generally have homes that are well adapted for it. Humans know how to survive in extremes, but the right infrastructure needs to be widely available to a population.
- Climate can vary significantly at a local level – weather data is often collected at a very granular and local level, but mortality data will not always be credible for such small geographical areas. This can lead to data having to be averaged over larger areas – and with this can come a reduction in precision.
Climate change will impact our society and should be of interest to anyone involved in managing mortality risks. Assessing the size of the risk, or sometimes even the direction of impact, is complex but not impossible. Interpreting the data and helping people to understand the risks to society are both well within the actuarial skillset.
Darryl Brundle is head of longevity at Legal and General
Ashley Kanter longevity research and innovation specialist at Legal and General
Image Credit | Getty