What is the difference between longevity and mortality?

Matthew Edwards, Editor of the IFoA Longevity Bulletin, explains the difference between mortality and longevity, discusses some of the contexts in which the terms are used, and looks at complications that may make them less simple than might appear.

Longevity and mortality both relate to the same underlying concept – the fundamental point of life or death, survival or demise – but they can be used in ways that are confusing.  This note explains the difference between the two terms, discusses some of the contexts in which the terms are used, and looks at complications that may make them less simple than might appear.

Longevity good, mortality bad

The words are generally used in financial services and demographic contexts in this way:

• Longevity is how long somebody can reasonably expect to live (i.e. their life expectancy);
• Mortality is the probability that somebody will die over the next year.

It might seem that these are opposite sides of the same coin.  However, there is a third element which we would need to calculate a person’s longevity given knowledge of their mortality, and that is the concept of ‘mortality improvements’ which we explain below.

Typical contexts

Whether you think in terms of ‘longevity’ or ‘mortality’ is largely a question of the context.  Discussions relating to life insurance (in particular, term assurance) portfolios or business performance tends to hinge around mortality rates.  Discussions about annuitants, pensioners, pension schemes, or life expectancy in general revolve around longevity.

Drivers of mortality and longevity

To the extent that mortality and longevity involve the same concept of ‘life or death’, they are both influenced by the same factors.  The most important ‘measurable’ factors affecting mortality are age, gender, medical history (and related medical aspects such as body-mass index or blood pressure), smoking habits, and socio-economic status.  We use the word ‘measurable’ because an individual’s genes are also likely to be very influential, and so ideally would feature in such a list, but the scientific community is only now starting to quantify this aspect[1].

Of the factors listed above, while some are clear, most of them involve a degree of uncertainty.  For instance, the mortality effect of age has been well-studied for centuries, but these studies depend on having large numbers of people to work reliably – and hence the mortality of centenarians is relatively uncertain, because there are relatively few.  Similarly, the mortality effect of vaping (e-cigarettes) as opposed to ‘traditional’ tobacco smoking is currently unclear, because there is relatively little data about vapers compared with normal cigarette smokers (it will take some years for vapers’ deaths to accumulate in sufficient numbers to allow reliable studies).

Finally, socio-economic status is a highly subjective area, and even raising it as a factor can be politically contentious.  For a long time, insurers would allow for the sum assured of policyholders in their mortality investigations, because policyholders with large sums assured would generally exhibit lighter mortality than those with smaller sums assured (the larger sums assured on policies being associated with more wealthy policyholders).  More recently, most insurers would also consider postcode, perhaps using (in the UK) the ‘Index of Multiple Deprivation’ as a way of grouping postcodes into relatively homogeneous groups.  There are many reasons why different socio-economic groups, or different postcode groups (whether defined simply by region or some other aspect of commonality), might have different mortality outcomes (after allowing for age, gender variations etc).  

The missing link – mortality improvements

It might seem that, if you know the likely mortality of an individual, you could calculate their longevity (life expectancy) on the grounds that these are pretty much two sides of the same coin.  However, to do so, you would need to know their mortality for many years into the future.

For instance, to calculate the life expectancy[2] of a 60-year old, you need to know the mortality probability of the 60-year old this year, and then of a 61-year old in one year’s time, then a 62-year old in two years’ time …

To do this, we need to assume how these mortality rates are going to vary in the future – because the mortality of a 65 year old person now is likely to be different from that of an equivalent 60 year old in five years’ time.  These are ‘mortality improvement’ assumptions, so called because, historically, mortality has generally improved over the years (owing to such factors as improved public health, medical discoveries, and reductions in smoking), and very rarely decreased.  However, coming up with a mortality improvement assumption is much harder than calculating a mortality rate. 

Calculations of mortality rates are, on the whole, relatively simple divisions of ‘number of deaths’ by ‘number of people who might have died’, ensuring that the two numbers are consistent (for instance, by age).  Thus, barring any problems with small groups of people (where the estimates would be less reliable because of random ‘noise’ in the numbers), it is a very reliable number – and about as objective a number as we could hope for.

But mortality improvements cannot be derived so simply, nor objectively.  How will the mortality of a 60-year old vary in five years' time?  We could look at past trends in mortality, and expect those to continue, but they will continue only to the extent that the historical reasons for those improvements (for instance, reductions in smoking, or improvements in healthcare) also continue.  Perhaps these will continue in the same way, but we are clearly now out of the realm of objective historical fact. Or new factors may start to emerge (for instance, relating to climate change), but again by definition these would not have been present in the historical data.

Finally, just as mortality itself varies by such factors as age, gender, and socio-economic level etc, so too do mortality improvements.  For instance, improvements have generally been much higher for people in the 60-80 age range than people in their 90s. There has been much debate as to socio-economic differentials in mortality improvements, and even more debate as to whether such differentials might continue (and if so, for how long).

Risk: longevity bad, mortality good?

Having clarified what we mean by mortality and longevity, and what influences them, we can now touch on the corresponding risks – mortality risk and longevity risk. These are generally used from the perspective of the financial institutions bearing the risk, rather than the individuals whose lives are being implicitly discussed here.

‘Mortality risk’ is the risk of adverse mortality movements from the perspective of (generally) an insurer or reinsurer with a portfolio of life assurance policies (for instance, term insurance).  If mortality increases, they pay out more than expected.

‘Longevity risk’, rather than being just the opposite of mortality risk, is more the risk that longevity increases more than assumed – in other words, the risk that the ‘mortality improvements’ discussed above end up being greater than assumed.  This is clearly a good thing for the underlying individuals, but likely to be financially adverse for the insurers or pension schemes involved – hence the sense of linguistic tension sometimes found in the financial press, where longevity improvements might be referred to by the writer in a negative fashion.

The two risks are also different because the underlying assumptions are very different in nature – mortality rates are relatively easy to estimate reliably and objectively, and hence the risk is not so much about getting those calculations wrong as being about some rare event occurring to make mortality suddenly jump over the year – for instance, a pandemic.  But longevity calculations rely on somewhat subjective mortality improvement assumptions, and those improvement assumptions may well – with the benefit of hindsight in few years’ time – be found to have been, overall, too low.  Hence there is a substantial market in managing longevity risk.

Mortality, longevity and coronavirus

Given the above, we can see that the coronavirus pandemic is (unfortunately) an excellent example of mortality risk: it is an exceptional event that we would not have allowed for a year ago in estimating somebody’s likely mortality this year. However, insurers will have allowed for the mortality risk here by setting aside capital in respect of a pandemic such as this.

The longevity of those who survive the pandemic is currently unclear.  Their mortality might be lower because the disease COVID-19 seems to act disproportionately on people with some existing medical conditions, such as diabetes (so the ‘pool’ of survivors will include fewer people who had relatively high mortality owing to such conditions).  On the other hand, their future mortality improvements could also be lower – for instance, a weaker economy may lead to less healthcare expenditure; or, if smokers are particularly affected by COVID-19 this year, there will be correspondingly less scope for smoking reductions in future years.

Summing up

In conclusion, and very simplistically: ‘mortality’ relates to the probability of death over a year, ‘longevity’ relates to life expectancy..  Mortality is well understood; longevity, less so, because we are uncertain of how much mortality rates will vary in even a few years’ time.

Matthew Edwards
28 April 2020