Adjusting for the age profile of a population

[Ted-54]

The economist's model does not appear to account for the age profile of a population - unless I have missed something. For instance, in Germany, from 2019 to 2020, the number of 80-90 year olds increased by nearly 264,000, the biggest increase of any age group (some saw reductions), i.e. 5.84% of the total population in 2020 as against 5,53% in 2019. The over 90 year olds increased by 28,000, 0,99% of the total in 2020 as against 0.96% in 2019. Therefore, an increase in deaths in those groups is not in the least surprising and will affect the overall mortality. If not otherwise accounted for, this will show up as excess deaths against an average calculated from years with smaller proportions of 80 + individuals.

[sondreus]

Hi Ted-54,

That would indeed be an oversight, however our models do account for this by modelling expected mortality increases over time. Whenever we have total mortality data for pandemic years, as we do for Germany, it is this modelled expected mortality we compare it to. You can view the code we use to model mortality increases over time here, at the repo this model dynamically pulls excess deaths data from: https://github.com/TheEconomist/covid-19-excess-deaths-tracker

The only exception to this is for our subnational data (from India, and Indonesia), where models are here: https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/blob/main/scripts/aux_subnational_data_generation.R

And for the three months of data from China, which is detailed in the main data generation script.

Hope this clarifies, and thank you for the comment.

[Ted-54]

Thanks a lot for the clarification.

Ted-54

Von: Sondre Ulvund Solstad @.> Gesendet: Montag, 6. September 2021 10:22 An: TheEconomist/covid-19-the-economist-global-excess-deaths-model @.> Cc: Ted-54 @.>; Author @.> Betreff: Re: [TheEconomist/covid-19-the-economist-global-excess-deaths-model] Adjusting for the age profile of a population (#7)

Hi Ted-54,

That would indeed be an oversight, however our models do account for this by modelling expected mortality increases over time. Whenever we have total mortality data for pandemic years, as we do for Germany, it is this modelled expected mortality we compare it to. You can view the code we use to model mortality increases over time here, at the repo this model dynamically pulls excess deaths data from: https://github.com/TheEconomist/covid-19-excess-deaths-tracker

The only exception to this is for our subnational data (from India, and Indonesia), where models are here: https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/blob/main/scripts/aux_subnational_data_generation.R

And for the three months of data from China, which is detailed in the main data generation script.

Hope this clarifies, and thank you for the comment.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/issues/7#issuecomment-913448784 , or unsubscribe https://github.com/notifications/unsubscribe-auth/AVP2F62GCWQ43FIEBBTZ463UAR2Y5ANCNFSM5DPDP24A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub .

[sondreus]

My pleasure!

[Ted-54]

Hi Sondre,

I hope it’s all right if I come back to you for some help on my item: it was actually prompted by a video by a German called Marcel Barz, who posted a statistical analysis of deaths in Germany from which he cannot infer any effect of the pandemic. Here are two screenshots from his talk, the first shows three tables: 1st deaths by age groups in Germany 2012-2020, 2nd population by age groups, 3rd % of deaths in each age group.

The second screeshot is a repeat of the third table, with rankings for each age group in 2020 in the last column as to how the number of deaths compares with the other years (1 least, 9 highest) and colours to indicate the years with the highest (red) and lowest (green) % mortality in each age group.

As you can see, on his figures which I take to be correct, no single age group had an unusually high mortality in 2020 in Germany. This convinces a lot of sceptics, and I don’t know how to answer them. Can you help?

Sincerely,

Ted Pawloff

Von: Sondre Ulvund Solstad @.> Gesendet: Montag, 6. September 2021 10:22 An: TheEconomist/covid-19-the-economist-global-excess-deaths-model @.> Cc: Ted-54 @.>; Author @.> Betreff: Re: [TheEconomist/covid-19-the-economist-global-excess-deaths-model] Adjusting for the age profile of a population (#7)

Hi Ted-54,

That would indeed be an oversight, however our models do account for this by modelling expected mortality increases over time. Whenever we have total mortality data for pandemic years, as we do for Germany, it is this modelled expected mortality we compare it to. You can view the code we use to model mortality increases over time here, at the repo this model dynamically pulls excess deaths data from: https://github.com/TheEconomist/covid-19-excess-deaths-tracker https://github.com/TheEconomist/covid-19-excess-deaths-tracker

The only exception to this is for our subnational data (from India, and Indonesia), where models are here: https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/blob/main/scripts/aux_subnational_data_generation.R https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/blob/main/scripts/aux_subnational_data_generation.R

And for the three months of data from China, which is detailed in the main data generation script.

Hope this clarifies, and thank you for the comment.

— You are receiving this because you authored the thread. Reply to this email directly, https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/issues/7#issuecomment-913448784 view it on GitHub, or https://github.com/notifications/unsubscribe-auth/AVP2F62GCWQ43FIEBBTZ463UAR2Y5ANCNFSM5DPDP24A unsubscribe. Triage notifications on the go with GitHub Mobile for https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 iOS or https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub Android.

[Ted-54]

PS: sorry, the link for Marcel Barz’s talk is https://www.youtube.com/watch?v=nEPiOEkkWzg should you want to see and hear the original. One of his arguments is that, as birth cohorts greatly vary in size, it is necessary to compare percentages of deaths in relation to annual population for each birth cohort (which he does in decennial batches). I suspect he would argue that this approach is necessary to calculate expected deaths reasonably reliably. I have viewed the code which calculates the Economist’s model expected deaths, but I am simply not conversant enough with code to be able to follow it. However, it does not seem to account for birth cohort sizes.

Von: Sondre Ulvund Solstad @.> Gesendet: Montag, 6. September 2021 10:22 An: TheEconomist/covid-19-the-economist-global-excess-deaths-model @.> Cc: Ted-54 @.>; Author @.> Betreff: Re: [TheEconomist/covid-19-the-economist-global-excess-deaths-model] Adjusting for the age profile of a population (#7)

Hi Ted-54,

That would indeed be an oversight, however our models do account for this by modelling expected mortality increases over time. Whenever we have total mortality data for pandemic years, as we do for Germany, it is this modelled expected mortality we compare it to. You can view the code we use to model mortality increases over time here, at the repo this model dynamically pulls excess deaths data from: https://github.com/TheEconomist/covid-19-excess-deaths-tracker

The only exception to this is for our subnational data (from India, and Indonesia), where models are here: https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/blob/main/scripts/aux_subnational_data_generation.R

And for the three months of data from China, which is detailed in the main data generation script.

Hope this clarifies, and thank you for the comment.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/issues/7#issuecomment-913448784 , or unsubscribe https://github.com/notifications/unsubscribe-auth/AVP2F62GCWQ43FIEBBTZ463UAR2Y5ANCNFSM5DPDP24A . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub .

[sondreus]

Hi Ted-54,

Thanks for the link. It does account for birth cohort sizes implicitly, by looking at how mortality changes by year for our projections for "normal year" deaths for 2020 and 2021. So unless there was a particular reason to expect a discrete jump from 2019 to 2020 very unlike year-to-year changes from 2014-2019, then the approach incorporates this information.

[Ted-54]

Hi Sondre,

I hope my persistence does not importune you excessively. I am a layman trying to make sense of contradictory information and to find pertinent answers to relay to others.

3 Questions:

1. Is it possible for you to explain briefly to a layman how your model calculates mortality changes?
2. I attach a simple excel file in which I use Barz’s figures to calculate expected deaths for 80-90 year olds. On that basis, I come up with excess deaths over 65,5 K for 2020 in Germany assuming constant percentage growth from 2015. Using the actual population figures and average deaths for this age group, I come up with 55,5 K excess deaths. Your model comes up with around 45K for all age groups, so is clearly much more sophisticated. Still, could the difference (in my spreadsheet 10K expected deaths) due to an increase of 220K more than the linear calculation arrives at in the actual size of that age group affect your model too? Would it be easy to plug in the actual population figures to find out?
3. If you have looked at Barz’s presentation, are there any fallacies you could detect? As far as I am concerned, the annualised figures are misleading in that they do not show how closely the weekly excess deaths track infections and official Covid deaths. But is there more?

Thank you for your time,

Ted Pawloff

Von: Sondre Ulvund Solstad @.> Gesendet: Dienstag, 7. September 2021 10:55 An: TheEconomist/covid-19-the-economist-global-excess-deaths-model @.> Cc: Ted-54 @.>; Author @.> Betreff: Re: [TheEconomist/covid-19-the-economist-global-excess-deaths-model] Adjusting for the age profile of a population (#7)

Hi Ted-54,

Thanks for the link. It does account for birth cohort sizes implicitly, by looking at how mortality changes by year for our projections for "normal year" deaths for 2020 and 2021. So unless there was a particular reason to expect a discrete jump from 2019 to 2020 very unlike year-to-year changes from 2014-2019, then the approach incorporates this information.

— You are receiving this because you authored the thread. Reply to this email directly, https://github.com/TheEconomist/covid-19-the-economist-global-excess-deaths-model/issues/7#issuecomment-914120219 view it on GitHub, or https://github.com/notifications/unsubscribe-auth/AVP2F6735XJJKWTI4XLOCILUAXHPPANCNFSM5DPDP24A unsubscribe. Triage notifications on the go with GitHub Mobile for https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 iOS or https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub Android.

[sondreus]

Hi Ted-54,

1. You can view our methodology for a non-academic audience here: https://www.economist.com/graphic-detail/2021/05/13/how-we-estimated-the-true-death-toll-of-the-pandemic -- to read more about underlying excess deaths calculations when we have data on total mortality (most of the time for Germany), you can have a look here: https://www.economist.com/graphic-detail/coronavirus-excess-deaths-tracker
2. Not entirely sure what you mean here. I would recommend the people at https://www.mortality.org/, who would be able to tell you more about age-specific mortality data (we use total mortality for our models, as few countries have it by age).
3. I haven't had time to view it unfortunately.

Thanks!