Covid-19: Keeping the mortality spike in context


IFoA Fellow Phil Caine asks why actuaries aren't 'ripping up the modelling rule book' as a result of the Covid-19 pandemic.

Most of us will never before have seen mortality rates in the news so much as in 2020. Facebook groups are comfortably discussing exponential growth. The Government is addressing the nation with the sorts of graphics that were once just the domain of actuaries. Calculations of R0 are influencing Christmas shopping.  Mortality statistics by age, gender and socio-economic class are on the front pages.  And most importantly, of course, we have the tragic individual stories we’re sadly becoming accustomed to hearing.

A small number of weeks from the main model in use by the actuarial profession (published by Continuous Mortality Investigations) being refreshed to incorporate 2019 data, the events of Spring 2020 had made significant deviations from that model a certainty.  So why aren’t we ripping up the modelling rule-book?    

To most, Covid-19 is a “Black Swan” - an event that now appears completely plausible, but whose possibility hadn’t been seriously considered.  However, this was less of a surprise to those with reason enough to consider it – notably epidemiologists, pharmaceutical companies, and those living through recent pandemics across the world.  Did actuaries have reason enough to have been less surprised?  It depends on our precise field, but I’d suggest mostly not.  

In my own field of defined benefit (DB) pensions, schemes were hit pretty hard in the first part of the century by developments in mortality rates going the “wrong” way (as far as funding costs go at least).  Scheme actuaries are well accustomed to mortality moving the dial every time their assumptions are refreshed.  There are good analogies to investment experience - it’s difficult to predict the short-term direction with any confidence, and the best available analysis can only go so far in explaining events over any past period.  

Admittedly pandemics haven’t been a “hot topic”.  Whilst winter flu rates tend to impact on our models, both for today’s mortality and its projection, it has actually been quite common to question whether unusual years (by pre-2020 standards at least) should be fully reflected.  The impacts of Covid-19 are certainly now on the agenda when we look at longevity measurement and volatility, but they’re not dominating it.   The reason lies in the distinction between the short term and the long term.

Deaths from the pandemic in the UK, as measured by the Office for National Statistics, surpassed 70,000 in November – that’s well over 10% of a typical year’s deaths.   If this level of excess was a new norm, we’d be looking at life expectancies around a year lower, and UK pension schemes could save a few percent of liabilities, effectively assuming death rates had returned to levels seen around a decade ago.   The short term in isolation means very little to us, though.  10% higher deaths over 2020, in your typical pension scheme, is likely to reduce liabilities by a small fraction of 1% (by the time you take into account the lesser effects on individuals not yet retired and the liability that might remain for the dependants of the deceased).  In fact, for all but the very largest pension schemes, a 10% annual fluctuation in pension-weighted deaths is very conceivable even in a more normal year due mainly to “concentration risk” (i.e. where members might individually represent a material proportion of a scheme’s total liability).

We can likely expect the ripple to extend into a good part of 2021, though positive news on vaccines suggests another 70,000 excess deaths may be unlikely, and we’d certainly hope to be getting back to normal levels in 2022.  By that time the typical DB pension scheme will probably have seen the sort of funding movements from Covid-19 that wouldn’t be unusual from an eventful day in investment markets.   This is likely not the whole story though.  

For recent retirees and for generations at working age, the death rates that have most impact on life expectancy are decades away.  Typical causes of death will then differ to present day, and the sources of longevity improvements up to that point (and any offsetting negative influences) will also differ to what we’ve seen in the past.  The past success stories from widespread aspirin usage and reductions in smoking rates seem likely to have largely run their course, but the world won’t stand still.  Healthcare changes, medical interventions and the behaviours of society have been key drivers in mortality rates in the last few decades, and benefits from brand new treatments and technologies (some perhaps not even imagined yet) shouldn’t be discounted.   

Large-scale uncertainty was there before Covid-19 and will remain once the world stabilises, but does our knowledge from Covid-19 have a part to play?  Has it changed how we expect things to look in 2040, or in 2060?  Perhaps we’ll spend time exercising rather than commuting, habits will have changed around hand-hygiene and personal interactions, and the clamour to deal with one pandemic might have led to improvements in infrastructure and medical knowledge.  Or perhaps long-Covid has a part to play, reduced socialising has lasting effects on wellbeing, and the economy doesn’t fully recover.

The distant future is critical to pension schemes, and we must keep a regular watch on likely influences.  For now, the direction of longer-term impacts is far from clear.  Should it become obvious that the net effect will be positive or negative, the longevity market will make its move, and responsible funding will need to move with it.