Destatis disaster
The Deutsches Statistisches Bundesamt (destatis, the German equivalent of the UK’s ONS) has been busy finalising the 2021 deaths statistics. The press release contains the following passage:
Aufgrund des zunehmenden Anteils älterer Menschen an der Bevölkerung wird seit etwa 20 Jahren mit einer jährlich steigenden Zahl der Sterbefälle in Deutschland gerechnet. Gleichzeitig stieg die Lebenserwartung vor Beginn der Corona-Pandemie jedoch tendenziell an. Der Effekt der steigenden Lebenserwartung schwächte damit den Alterungseffekt ab. Bei gleichzeitigem Wirken beider Effekte stiegen die Sterbefallzahlen vor Beginn der Pandemie jährlich um durchschnittlich 1 bis 2 %. Bereits 2020 war der Anstieg im Vergleich zum letzten Vorpandemiejahr 2019 stärker ausgeprägt (+5 % auf 986 000 Sterbefälle). Nach den jetzt vorliegenden endgültigen Daten sind die Sterbefallzahlen 2021 um weitere 4 % oder etwa 38 000 Sterbefälle auf insgesamt 1,02 Millionen gestiegen. Ausgehend von 2019 wäre für 2021 eine Sterbefallzahl von 960 000 bis 980 000 erwartbar gewesen, also ein Anstieg um 2 bis 4 %. Tatsächlich ist die Zahl der Sterbefälle von 2019 auf 2021 um 9 % gestiegen. Bezogen auf die beiden Jahre 2020 und 2021 gab es demnach etwa 70 000 bis 100 000 zusätzliche Sterbefälle. Beim Robert Koch-Institut wurden in diesen beiden Jahren fast 115 000 COVID-19-Todesfälle gemeldet. Maßnahmen und Verhaltensänderungen im Zuge der Pandemie können auch dafür gesorgt haben, dass 2020 und 2021 weniger Sterbefälle durch andere Infektionskrankheiten wie beispielsweise die Grippe verursacht wurden.
Translation into bad English (slightly abbreviated):
For around 20 years, we have observed the numbers of deaths rising in Germany. Before the Covid pandemic, rising life expectancy partly compensated for the ageing of the population, leading to a 1-2% annual rise in deaths. The year 2020 saw a rise of 5% compared to the last pre-pandemic year, 2019. In 2021, deaths rose by an additional 4% (38,000 deaths). Based on 2019 figures, between 960,000 and 980,000 deaths would have been expected, a 2-4% rise from 2019 to 2021. The actual figure was 9%. Hence, the two years of 2020 of 2021 have produced between 70,000 and 100,000 excess deaths. The Robert Koch-Institut has recorded almost 115,000 Covid deaths during that period. Pandemic measures and behavioural changes might have reduced the number of deaths attributable to other infectious diseases, such as influenza.
I am quite sure that the people in the engine room at destatis are pretty good at what they are doing. But I am even surer that they are not allowed to write the press releases. I present the above passage as Exhibit A.
I have discussed the data in a previous post. Here’s the numbers again, now including the finalised 2021 deaths (around 2,000 more than previously known):
If we compute year-to-year relative changes in the numbers of deaths, we get:
Yes, the average over these numbers is 1.27%, but calling this a 1-2% annual rise is misleading. Obviously, there are good years (2014, 2016, 2019) and bad ones. What makes a year good or bad is the absence or presence of waves of respiratory diseases. Of course, 2020 saw a rise of 5% compared to 2019, but this was not much of a surprise. To illustrate, here are a few alternative expected numbers of deaths for 2020, all computed by choosing certain mortality rates by age group and applying these rates to the population figures:
“by mean”: mean values of the rates from 2011-2019
“by median”: median values of the rates from 2011-2019
“by min”: minimum values of the rates from 2011-2019
“by max”: maximum values of the rates from 2011-2019
“by 2019”: rates observed in 2019
“by 2018”: rates observed in 2018
For example, by applying the 2018 rates to the 2020 population figures, we compute an expected number of 989,299 deaths, more than the observed number of 985,572 deaths. And even by applying the mean or median rates, we almost hit the observed number. Only two things were special about 2020: a new alpha displaced the old disease(s), and more deaths occurred in the second half of the year.
Here is the same table for the 2021 figures:
By comparing the two tables, we see that simply due to the ageing population, we would have expected around 19,000 additional deaths compared to 2020. The actual difference was twice that, as is mentioned in the destatis press release, and I declare the second 19,000 the actual excess deaths for 2020-2021.
Good people at destatis, think about what the difference was between 2020 and 2021, and then apply a booster to your press release.
Woe to them!