Question on baseline of expected deaths
Closed this issue · 11 comments
Hi there,
I was trying to find the baseline of global expected mortality for the entire year (Jan 1 - Dec 31) in 2020 and in 2021. I was able to find the global excess mortality estimates and the expected mortality by country, but I could not find the number of global expected deaths used to calculate the "central excess death estimate" figures published by Our World in Data (16.99M from Jan 1, 2020 to Dec 20, 2021).
https://ourworldindata.org/excess-mortality-covid#estimated-excess-mortality-from-the-economist
Specifically, I am looking for the “central” estimate for the following:
global ACM in 2020 (Jan 1 - Dec 31)
global ACM in 2021 (Jan 1 - Dec 31)
global expected mortality in 2020 (Jan 1 - Dec 31)
global expected mortality in 2021 (Jan 1 - Dec 31)
Also, what is meant by “central” excess death estimate? Is it the median or the mean of the models used?
Thanks
Thank you for your message.
We do not have this data, because where ACM is missing, we estimate excess deaths directly. This is deliberate: we estimate what we want to estimate directly, rather than intermediate quantities (such as ACM and expected ACM). I can go into more detail on this if needed.
The central estimate is the mean of 10 models all fit on the full sample of data with different starting seeds. It is not the median or mean of all the models used to compute the confidence intervals (which are built using data from a stratified random sample).
Hope this helps!
It does. Could you please elaborate on how you estimate excess mortality directly without ACM and expected ACM? I thought that was the definition of excess mortality, the difference between the two.
Sure, full details are here: https://www.economist.com/graphic-detail/2021/05/13/how-we-estimated-the-true-death-toll-of-the-pandemic
Essentially, one does not need to know the subcomponent of something to estimate it (much like how one can estimate GDP without breaking it into its constituent parts). We have ACM and expected ACM for some countries, with the details here: https://www.economist.com/graphic-detail/coronavirus-excess-deaths-tracker. We use this to come up with excess deaths figures for the these countries. We then use these excess deaths figures as our training data, and estimate it directly.
Hope this helps!
Also, the quotes below led me to believe that The Economist did estimate expected deaths. From the article:
https://www.economist.com/graphic-detail/coronavirus-excess-deaths-tracker
“We have used statistical models to create our baselines, by predicting the number of deaths each region would normally have recorded in 2020 and 2021”
“Update (October 14th 2020): A previous version of this page used a five-year average of deaths in a given region to calculate a baseline for excess deaths. The page now uses a statistical model for each region, which predicts the number of deaths we might normally have expected in 2020”
My apologies, I did not see your previous reply.
No problem. Did my reply clarify sufficiently?
Yes, thank you. I do have one more question. For the countries with available data, what data was used to extrapolate the baseline figures? To clarify my question, the article we both shared mentions an update where a five-year average was initially used to calculate the baseline, and this average was replaced with a statistical model. Was the statistical model that was used to predict the counterfactual based on data from the same five year period, presumably 2015 to 2019?
Yes, 2015-2019, with a week, month or quarter fixed effects and a year trend. You can find the details and code here: https://github.com/TheEconomist/covid-19-excess-deaths-tracker
Thank you for taking the time to answer my questions.