,
17 tweets,
6 min read
Read on Twitter
I’m afraid I have to take issue with this article from @dannydorling in @ConversationUK. Before diving into detail (thread warning!) let’s focus on big picture - central message is just wrong. UK mortality rates are generally going down, not up. Life expectancy is rising. /1
.@dannydorling is correct to point out that number of deaths in UK is now rising year-on-year. We reached a turning point in 2011 where effect of ageing and growing population overpowered influence of falling mortality rates, so the number of deaths each year began to rise. /2 
The number of deaths each year will continue to rise, with @PHE_UK forecasting an increase to 550,000 a year by 2023 (around 8,000 more deaths each year). /3
(Incidentally, the demand pressures on health and social care provision caused by the increasing numbers of people near the end of their lives is something that’s been underexplored by those studying the reasons for the slowdown in life expectancy gains.) /4
Turning to death rates though, it’s wrong to say that “the UK [has] seen such a large fall in life expectancy since 2014, for both men and women”.
2018 death rates for males were lower than in 2014 and death rates for females were less than 2% higher. /5
2018 death rates for males were lower than in 2014 and death rates for females were less than 2% higher. /5
When comparing deaths rates it‘s essential to age-standardise them so that they are not unduly influenced by the ageing population. This even applies within 5yr age bands, so individual ages must be considered – which looks like it may not have been the case in this analysis. /6
Whilst any increase in death rates is bad news, even this small apparent rise for females doesn’t stand up to scrutiny. UK mortality (like other temperate countries) is seasonal. The timing of peak winter and the flu season matters a great deal!
This is why ONS base life expectancy on 3-year averages, not individual years. 2014 had exceptionally low winter mortality, the lowest in recent years other than the 2019 winter just gone (I discuss the CMI report on Q1 2019 👇) /8
In short, winter death numbers are volatile, which is why ONS, actuaries and other experts take steps to ensure they don’t unduly influence our view on death rates. /9
By picking out an unusual year without significant winter deaths, and only using a 12 month period, @dannydorling paints a misleading picture which ignore the high mortality in 2013 and the exceptional mortality in 2015. /10
The same issue arises with the 2016-17 and 2017-18 comparison. Twelve-month periods are too short too meaningfully compare and the latter contains the unusually high death rates from Q1 2018. /11
It is noteworthy also that we’ve seen unusually low mortality in the second half of 2018 and the first half of 2019. You can see this in the picture upthread 👆 This is also ignored in the article. /12
Lastly, I have to take real issue with the suggestion that @dannydorling is the only one looking at this! He does a great disservice to the many actuaries, epidemiologists, demographers, gerontologists, statisticians and others doing great work in this area. /13
Perhaps greater attention to the work of the experts mentioned above and a more collaborative approach might have helped produce a more accurate and less sensationalist article? /End
I’ve been asked to expand on how we reconcile the fact that death rates are still falling with the reductions in forecast life expectancy that life insurers and pension funds are benefiting from.
The chart below from the CMI shows the pattern of death rates in recent years.
The chart below from the CMI shows the pattern of death rates in recent years.
Broadly speaking, actuaries have moved gradually in recent years from projecting forward the pre-2011 trendline ☝️to forecasts which allow for the significant reduction in the pace of change. My thread for March gives more detail about the latest model.
Similarly, the mini-thread below gives a more visual illustration of how the pattern of deaths is expected to change over the next 30 years.
