This just ain’t true:

Life expectancy in England falls to lowest level since 2011
Excess Covid deaths contributed to life expectancy in England falling by 1.3 years for men and 0.9 years for women

We don’t measure life expectancy. We measure age of death, then take that to be a proxy for life expectancy. But the use of a proxy does require understanding the proxy. So, how many think that Covid will be carrying off 80 year olds in 80 years time?

Then the current reduction in lifespan from Covid is not an influence upon life expectancy, is it?

27 thoughts on “Bollocks”

  1. I could understand it if it was quoted as “Panic reaction to Covid has reduced life expectancy…” on the basis that the NHS morphed into the “National Covid Service” at the expense of just about all other medical procedures – thus hastening the death of younger people from normally treatable illnesses that went untreated.

    As to Covid itself – I was given to understand that for the UK the average age of death “with” Covid was 82.4 years whereas the average age of death (all causes) in 2019 was 81.3 years. Thus I can’t see how Covid itself has done anything other than prolong life! 🙂

  2. The typical male who has just died aged 79 did not have a life expectancy of 79 years at birth.
    Do two errors self correct?

  3. And this is the joy of computer modelling… You can never account for all variables so any output is a suggestion at best.

    Never take anything a computer says as the full picture because we have biased humans coding them (I’m a coder and have seen this a million times) and we all introduce bias without knowing it…

  4. That is utter nonsense – ask Chris Miller or BraveFart, our local experts (I have only used my actuarial skills a handful of times in the last fifty years). [Incidentally, the CMI Bureau has declared that it will ignore ‘s distorted data when calculating life expectancy].

    Life Expectancy is, as any *intelligent* human realises, an estimate of how long *in future* you will live, starting from the then current date. So only “long covid” will impact on life expectancy, not past deaths from a non-recurrent pandemic. [There may be future pandemics but there won’t be a future covid-19 pandemic in the UK because the vaccines will reduce R below 1 for any future outbreak of covid-19].

    “Long covid” can only have a tiny effect on life expectancy since it only impacts a small minority and we have no data on how far, if at all, it shortens life.

  5. So hang on a minute. The deadliest global virus pandemic we’ve had in about a hundred years puts us back to the same general risk of death we had 10 years ago. Or, to put it differently, the increase in ‘life expectancy’ we’ve experienced over just the past 11 years is greater than the loss we’ve experienced because of Covid.

    Isn’t capitalism wonderful!

  6. @Bill

    The deadliest global virus pandemic we’ve had in about a hundred years..

    You missed the “/irony” tag… 🙂

  7. ” … ignore 2020’s distorted data…”

    It’s interesting that they are doing that at ONS over all causes death figures.

    They are giving 2021 and comparing it with (i) 2020 and (ii) average of 2015-2019.

    Since the second (or was it third?) Covid wave petered out by the end of February, all cause deaths have been lower in 2021 than the 5 year 2015-2019 average (March to date).

  8. @ Andrew C
    ONS has some decent honest statisticians, and they do that because they know that 2020 is not a sound basis for comparison.
    The figures for 2021 may be expected to be a bit lower than average because covid-19 helped to kill off a large number of elderly people with co-morbidities, a significant minority (maybe 10%, maybe 20%) of whom would otherwise have died in 2021 and that more than offsets the covid-19 deaths in 2021.

  9. John77

    “The figures for 2021 may be expected to be a bit lower than average because covid-19 helped to kill off a large number of elderly people with co-morbidities, a significant minority (maybe 10%, maybe 20%) of whom would otherwise have died in 2021 and that more than offsets the covid-19 deaths in 2021.”

    Similarly, there was about 18 months of below average data from Summer 2018 through to Winter 2019. In exactly the same way, the increase in Spring 2020 could almost certainly be seen as an offset for that earlier reduction.

    John, you’re excellent with numbers. If you took the data from here (grabbing the June 2021 dataset):
    https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/monthlymortalityanalysisenglandandwales

    and summed the months into years based on July to June, so that whole winters all appear in one year and which splits Covid into two distinct winters (same we would do if it was flu each year), what do you see?

    What I noticed was that with an 5 year trend line through to 16/17 (ie including data for 18/19 but not beyond because that’s what we want to check), and then a reasonable trend projection from that point, the dip in 18/19 was pretty much balanced out by the peak in 19/20, and left 20/21 (to June 21) as pretty much a nothing.

    Monthly peaks, yes sure, eg April 20 was very real. But simply as an annual perspective, I don’t see anything too unusual in 19/20 that one might not have expected given the dip in the year prior. (Same as in 13/14 and 14/15 or as we’ve seen lots elsewhere with this sort of analysis.) Or, if not Covid, then something else at some point should have done the same or similar.

    I’d upload my graph, but I’m not sure I can on here. But I suspect you could recreate it in 20 minutes flat if you are interested?

    The problem with calendar is that calendar 15-19 includes the full dip (to Winter 19). Calendar 20 then includes the peak. Part of the excess of 20 over 15-19 should therefore be eliminated (for useful comparability). And which becomes very clear if one uses July-June as the basis.

  10. Actually, if it’s helpful at all, and should have done this first, here are the (simple) annually averaged numbers (the slightly more sophisticated/accurate way of averaging them is not materially different):

    20/21 978
    19/20 1,020
    18/19 907
    17/18 977
    16/17 967
    15/16 958
    14/15 1,000
    13/14 933
    12/13 997
    11/12 976
    10/11 998
    09/10 1,012
    08/09 1,066
    07/08 1,077
    06/07 1,081
    05/06 1,116
    04/05 1,165
    03/04 1,200
    02/03 1,217
    01/02 1,217

  11. Very useful, PF. John77 flatters me:
    That is utter nonsense – ask Chris Miller or BraveFart, our local experts (I have only used my actuarial skills a handful of times in the last fifty years).
    I haven’t practised as an actuary in almost as long, but I still keep my hand in, and I agree with your assessment. There was a period when it seemed as though average age at death was increasing, due to Covid taking mainly those already in heaven’s waiting room.

  12. @ PF
    Thanks: I shall try to look at them – but I shan’t be able to complete anything tonight because I’ve only just seen your post.

  13. “ the vaccines will reduce R below 1 for any future outbreak of covid-19”

    The vaccines seem to have a pretty poor record on controlling transmission (vaccine passports?) though they do work at what they were designed for which is keeping hospital numbers low, hence the ongoing ‘cases’ reporting as it inflates the issue.
    Interestingly I noticed that there seemed very little evidence of reinfection among Covid patients before the vaccines came along now ‘breakthrough’ transmission seems to be all the rage. I’m still bemused though over the ongoing official refusal to accept natural immunity from having had Covid as being valid. I get you can look at asymptomatic and question if the person really had Covid (not that they want to question the validity of testing) so maybe shouldn’t allow that to be registered, but symptomatic and hospitalisation cases are treated the same.

  14. “We measure age of death, then take that to be a proxy for life expectancy.” Well no. Average age at death of those dying in a particular year is not life expectancy. What we do is construct a life table of the proportion of people dying at each single year of age over some period and use that to calculate the expectation of age at death. So if we were to use the year 2020 on its own as a basis for constructing the life table then we would indeed see a reduction in life expectancy compared to doing the same thing using data from 2019. But this would be a perverse thing to do as other commenters have pointed out.

  15. “‘breakthrough’ transmission seems to be all the rage”: based, presumably, on PCR tests? Which aren’t very good. It’s been 18 months and more; why don’t we have a better test?

  16. We already knew con-vid was a pack of lies start to–hopefully soon–finish. A bad winter flu lied to the skies to enable shite vax which leads to vax pass 1st step to the real goal:worldwide social credit tyranny CCP-style.

    They can shove all of it up their globo “elite” arses.

  17. You miss the fact that “life expectancy” is an incomplete specification, so to understand what someone means you need to figure out the missing part. It means the average (arithmentic mean) age at death for a group of people. You need to figure out which people are intended. If it’s people born in 1900, the figure can be calculated quite accurately as they are all dead now. If it’s people alive at the start of 2020, it’s an estimate based on what we know. The intent of the article is to express the idea that the estimate we made at the start of 2020 was wrong due to Covid. In the light of what we now know, our estimates should be lower as we will now allow for Covid specifically, but also any other disease which might arise. We allow for other diseases because previous estimates would have underestimated their effect – we had gone 100 years since the last major pandemic, and seen SARS and MERS defeated, so we thought the risk was lower than it actually was – which is why we handled this pandemic so badly.

    @PF “what do you see?”

    Here you are:
    2000/2001 122507
    2001/2002 236044
    2002/2003 236453
    2003/2004 235169
    2004/2005 231070
    2005/2006 226212
    2006/2007 222882
    2007/2008 224746
    2008/2009 225755
    2009/2010 221043
    2010/2011 222227
    2011/2012 221337
    2012/2013 229712
    2013/2014 223788
    2014/2015 240355
    2015/2016 238669
    2016/2017 243698
    2017/2018 252269
    2018/2019 242138
    2019/2020 279422
    2020/2021 274726
    2021/2022 20616

    The 2000/2001 is wrong because the data starts with January 2001, the 2021/2022 is wrong because we don’t yet have figures past July 2021. That looks pretty clear to me. 2019/2020 saw a jump of 37,284 over the previous year (and that was one of the higher years), and 2020/2021 was not quite as large, but still obviously exceptionally big. The average of the full years before 2019/2020 is 231,865, so 2019/2020 was 20% greater – quite obviously exceptional.

    @BniC – “I’m still bemused though over the ongoing official refusal to accept natural immunity from having had Covid as being valid.”

    That’s because it’s a political decision, and our politicians are useless.

  18. @ Charles
    My life expectancy is the number of years that I expect to live starting from *NOW* (so much less than 70 more years). You are talking about “life expectancy from birth” which cannot actually be measured until every single member of that cohort has died (as you note) and is a convenient fictional title for the reciprocal of the current weighted average mortality rate. The statisticians calculate the mortality rate at each age in a particular year and then the average age at death if the population experienced those mortality rates throughout their lifetime – ignoring the improvements in medicine during each person’s lifetime that have reduced/will reduce mortality rates from those current as his/her date of birth.
    That’s a bit turgid, but I hope you get the point that the figures bandied about are a useful illustration rather than a 3-D CT scan by Doctor Who.

  19. Charles!

    From table 1, You’ve taken males only rather than total. But more importantly, you’ve taken total number of deaths, ie column B. Try taking column O instead. Which is per capita (per 100,000), otherwise you are wasting your time. Column O is also age stratified, which gives like for like information from period to period (and is also comparable across countries). Try averaging O13 to O24 and you get 978 (per 100,000), as I’ve included above. There is a more accurate method, weighting for deaths, but I can assure you that the calculated numbers are consistently materially identical to using a simple average.

    Test one or two if you don’t believe me. Then perhaps look at the data I provided above and comment more usefully on that (if inclined?). I hope that helps.

  20. Sorry – weighting for population, which you can notionally calculate from deaths (and the per capital number given), but you get the gist.

  21. @dearieme

    “‘breakthrough’ transmission seems to be all the rage”: based, presumably, on PCR tests? Which aren’t very good. It’s been 18 months and more; why don’t we have a better test?

    In many ways the biggest things that aren’t very good about PCR tests are that the results are slow to arrive (particularly if the labs are busy during a peak, and making people with a sniffle isolate for days while waiting for a result is costly), they generally can’t tell you what variant a person has (we got lucky with S-gene target failure being a good proxy for Alpha/Kent/B.1.7.7) and they’re quite expensive to administer and process (not just for people wanting to go on hols or whatever, but to health services undertaking large-scale testing).

    But in terms of sensitivity and specificity, they’re pretty good really. Looking at the very low rate of false positives in the ONS Covid survey, if they’re finding vaccinated people testing positive for the virus, it’s usually going to be a correct result in the sense that the person genuinely is – or has recently been – infected. That’s reinforced by the way studies relying on PCR testing find patterns that make sense (eg rising protection after vaccination for a while, waning after that, two jabs have more protection than one etc) whereas if the results were essentially random because they were dominated by false positives, all those trends would just be flat. You might reasonably add a fourth item to the list of problems, that even if a positive result is “correct” in that a virus was detected, it can’t say whether you’re genuinely “ill” with Covid or whether you’re actively infectious. But there’s a debate to be had about what exactly it’s reasonable to expect a test to do compared, say, to getting patients to list their symptoms. Would be cool if tests assessed infectiousness but there must be some tricky technical hurdles to measuring that directly.

    Knowing what variant is currently of more interest to the epidemiologists than to patients/doctors since treatment is the same regardless – the UK has done a lot of sequencing of positive tests, which is useful for variant surveillance or understanding chains of transmission (away from Covid specifically, it’s the kind of tech hospitals are increasingly using to manage outbreaks of various hospital-acquired infections), but clearly this adds cost and time. Sequencing should surely dispel any doubts the most conspiratorial have that the virus is real and (barring contamination) if your sample contains enough genetic information to be sequenced, you know exactly what virus you had… but it’s unreasonable to demand to see a full sequence before you accept that someone really did have a breakthrough infection, and the sensitivity is worse because not all samples will contain enough genetic material to be sequenced, even if it’s enough to be caught by PCR. Some multiplex PCR tests have been developed to discriminate between variants (by deliberately extending the SGTF trick that helped track Alpha) but the use case for that is mostly for countries that can’t afford extensive sequencing for their variant surveillance. Lateral flow tests are good for making very rapid decisions, particularly lower stakes ones, but if you’re worried about sensitivity and specificity you probably prefer a PCR test. (There’s an argument that the lower sensitivity of lateral flow devices is actually a *good* thing, because they tend to flag up people who are at their most infectious and so most important ones to catch, even if people with mild or old infections slip through their net, but that comes back to the debate about what exactly you expect a test to do, which is in turn related to the purpose you’re using the test for – PCR and LFD have different use cases, so it doesn’t make sense to declare one as superior to the other.)

  22. @PF
    Yes, I only gave figures for males and should have used column N for people. But you can’t use column O as that is age standardised – see https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/methodologies/userguidetomortalitystatisticsjuly2017#death-rates-ratios-and-standardisation – which is based on a hypothetical standard population. This means, for example, that a death of someone who is 40 contributes more to the rate than a death of someone who is 4 (weighted 7% vs 4%). Consequently, you cannot average them. If do a sanity check, taking the total deaths and death rate you get the population of England to be 4,741,301 in July 2021, which is rather far from the true figure. And the figure from January 2001 would be 3,492,310.

  23. Charles

    If do a sanity check, taking the total deaths and death rate you get the population of England to be 4,741,301 in July 2021

    Err, no… you conveniently “forgot” that it’s monthly.

    July 2021 – 40,467 monthly deaths at 853.5 (which is annualised per 100K) is 56.9 million – and that’s for England not for the UK, ie the right ball park – rather than your 4.7m…..

    But more importantly, whatever the actual number is, it’s a notional calculation (the base position was taken in 2013, from memory, and the adjustments work forwards and backwards from there), ie what you need to understand that this is a standardised methodology to ensure useful comparison.

    To use absolute numbers (in a changing population over time) is utterly meaningless, as you must surely understand? Otherwise, if a country increases its population every year, the deaths may continue to increase every year, but which will tell you nothing. Only by taking into account deaths per capita and then dealing with the age categories can you come up with anything meaningful (for example if all the age increases are young compared with being older, etc). It really isn’t rocket science.

    FWIW, if you are hung up on the age issue, the most important element of that adjustment (in any one country) is “per capita”. And you could prove that by finding and comparing the “per capita” numbers with the “per capita and age stratified” numbers (here), the movements are actually not so different within the same country (a key use of the these EU devised adjustments was to compare mortality with other countries on a like-for-like basis).

    Per capita, simply at a basic minimum? We’re clear with that one, aren’t we (if we’re commenting in good faith)? Ie, forget column N……..

  24. Charles,

    These are officially released ONS numbers. I’m slightly surprised that you simply dismissed them as gibberish (coming up with an England population of 4.7m) without thinking it through first? Hey ho..??

  25. Yes, I forgot the monthly, but that still leaves the 2001 population wrong, so I went to the official population figures.

    So here’s the death rate per 100,000 using population of England in https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/populationestimates/datasets/populationestimatesforukenglandandwalesscotlandandnorthernireland Since that gives mid-year figures, and you want Jul-Jun I have taken the population of year X/X+1 to be the average of the two:

    Year deaths rate
    2001/2002 495353 1001.7
    2002/2003 499068 1006.9
    2003/2004 497183 998.3
    2004/2005 486832 972.5
    2005/2006 474025 940.5
    2006/2007 465570 916.7
    2007/2008 470210 918.9
    2008/2009 471030 912.9
    2009/2010 456781 878.3
    2010/2011 459819 877.2
    2011/2012 459956 869.9
    2012/2013 477158 895.2
    2013/2014 456548 850.5
    2014/2015 497999 920.7
    2015/2016 486028 891.0
    2016/2017 497550 904.2
    2017/2018 511907 923.3
    2018/2019 484698 868.7
    2019/2020 556363 991.2
    2020/2021 539370 956.0

Leave a Reply

Your email address will not be published. Required fields are marked *