Andrew Hetherington

Actuary working in London, UK. All views are my own.

Continuous Mortality Investigation: Another Year, Another Model Parameter

A riddle for you.

It’s 2020. I wake up, pour myself a bowl of cornflakes (milk first, cereal second — obviously) and check the news. Then I go out into the garden and lock myself in the shed with a note on the door saying to come and get me when it’s all over. Who am I?

If you answered “literally anyone” then you would probably be correct. However, if you answered “an actuary in charge of modelling and projecting improvements in mortality into the future” then you get extra brownie points. Because this year, longevity specialists have had the rather unenviable task of trying to work out how to quantify the long-term effects of a pandemic on human life expectancy.

After quite a lot of thinking, these specialists working for the Continuous Mortality Investigation (CMI) have published a working paper setting out their intentions for the 2020 version of their annual model for projecting how life expectancies will change over the coming decades.

A quick overview of the basics

For those that may not be familiar:

  • In general, life expectancies (simply put, the highest birthday the average person will reach before they die) have been increasing over the years due to a number of factors including better medical care and a decline in smoking.
  • Over recent years, data show that the size of these yearly improvements has been decreasing, leading to questions concerning how long these improvements can last.
  • In response to these emerging questions, new parameters were added to the 2016 and 2019 versions of the CMI’s model in order to allow users to customise the model output according to what they think might happen in the future.

The new model

Earlier this month, the CMI set out a new feature for their new model version due to be published in early 2021. The model will allow users to place more or less weight on the data collected in individual years for the calculation of future mortality projections. In particular, the question of how much weight to place on data collected in 2020 was raised, and the default “core” model will place a 0% weighting on 2020 data — ignoring any possible long-term effects of the pandemic, in effect.

Observing this, we might raise two major questions. The first is perhaps the most obvious one, and that is: how will future mortality be affected by Covid-19? The second arises after taking a step back and realising that there is a growing number of extra knobs and dials that can have a material impact on the results of this model. Is the fact of having more model parameters, more control and customisability, unquestionably a good thing?

How will future mortality be impacted by Covid-19?

This is a complex question to which there isn’t really an answer yet.

Clearly, there have been more deaths than normal in 2020, so we might expect this to dampen any future improvements in life expectancies. On the other hand, it could also be argued that future mortality improvements might be amplified since the pandemic has differentially killed a larger number of older and possibly more frail members of the population, leaving a higher proportion of younger and healthier people behind.

But let’s not be short-sighted. It’s possible that the longer-term effects such as further emergences of the virus, strained healthcare services, and a weakened global economy could contribute the largest impact on what to expect from future mortality rates.

More parameters = a good thing?

The optimist says that more parameters means greater flexibility and an enhanced ability to tailor the CMI model to specific purposes. The pessimist says that more parameters means more opportunities… to get it wrong.

Trust in the relevant experts certainly becomes more important as the problem of setting assumptions for the future mortality of a certain group of interest grows in complexity. No one profession is equipped to come up with a good answer. There are innumerable factors that feed into a phenomenon as complex as how long people live and each one may seem like an impossible task in its own right. Public health experts will be needed to analyse healthcare trends. Statisticians will be needed to uncover and decode patterns in mortality data. Economists will be needed to predict the future of the economy and how this will feed into our daily lives. Actuaries will be needed to liaise with the different experts and deliver actionable advice to their clients. Interdisciplinary communication will be important.

It all comes back to working together

And so, it seems like we are back at the conclusion which we have arrived at before — we need to effect large-scale collaboration between different professions and different viewpoints to find the best answers. In the same way the this virus has greatly accelerated our society’s capacity to develop, test, and roll out vaccines, perhaps we will also see our experts’ ability to work together improve, driven by necessity rather than an abstract idealism.

Plenty of food for thought going into 2021. No-one has been able to completely avoid the difficulties posed by an emergent disease, but opportunities abound and progress has been made. The future looks hopeful, but we’re not out of the garden shed yet.

More info and credits

Andrew Hetherington is an actuary and data enthusiast working in London, UK. All views are Andrew’s own and not those of his employer.

  • Connect with me on LinkedIn.
  • See what I’m tinkering with on GitHub.

Photo by Immo Wegmann on Unsplash.