I started looking at mortality data after discovering the Human Mortality Database (HMD) when it was cited as a data source for the paper Global Assessment of COVID-19 Mortality Displacement From 2020 to 2024. The HMD has high quality data on population, births, and deaths over many years, though only for about 35 mostly Western countries. The great feature of this database is that they provide yearly data, both in terms of calendar years and age levels, which means they’ve done the hard work of modeling yearly data from the often less-detailed official country data. Better to have the demographers do that modeling.
The paper was looking at yearly mortality rates since covid to check for evidence of “displacement” covid deaths. That is, was the spike from covid deaths offset by a subsequent reduced mortality rate? The paper didn’t find much evidence— only in 3 of 34 countries. The paper was light on charts, but I made this graph of mortality rates for a few selected countries, ages 50 and up.
Poland was one of the countries showing evidence of a displacement (along with Greece and Lithuania, not shown). I added the hand-drawn dashed line to illustrate the basis of their finding. They extrapolated the linear trend from a few years before covid and measured the post-covid rates against that. In this case, Poland had an upward trend for that period, so even though the latest rates are comparable with pre-covid rates, the extrapolated trend makes it look like a reduction.
Data digging
Given how little post-covid data is available, I started looking for other insights in the data. Here’s a chart of mortality rates in France by birth year and age. Age is on the x axis, and there’s a separate line for each birth year since 1910.
Besides the series of spikes as early birth cohorts become adults during World War II, I found it remarkable that those born in 1919/1920 (orange line) maintained a noticeably higher mortality rate throughout all later years. (I can’t say the exact birth year because it’s computed by subtracting age from death year. For example, a 50 year old who died in 2000 could have been born in 1949 or 1950.)
The 1919/1920 birth cohort effect is most pronounced in France and Italy but possibly present in other European countries.
Better context
Except for special events like war and pandemic, mortality rates generally went down throughout the 20th century. So it’s not unexpected that the 1919/1920 rate is higher than most others. There could be another birth cohort that is similarly bad compared to its temporal peers. To test that, I collected the mortality rates of neighboring birth cohorts (+/- 10 years) for each birth cohort at each age. Relative to that sample of 20 rates, I computed a z-score (number of standard deviations away from the mean) for the birth/age combination in question. Here’s a heatmap showing all the birth cohort/age combinations for all years that have substantial data (1900 – 1970) for France.
In a rough sense, the vertical orange stripes indicate bad years to be born, and the diagonal orange stripes indicate bad years to be alive. There are also some blue stripes for unusually good years. The other orange vertical stripe is for the 1945/1946 birth cohort.
Italy and other countries are similar, but there are also countries with different patterns.
Finland has an unusually strong birth cohort at 194/1942.
Japan has a bad birth cohort at 1946/1947.
For the United Kingdom, the 1919/1920 birth cohort stands out more than it did in the line chart.
Like the UK, the United States has a strong diagonal orange stripe at the top right corresponding to covid deaths. I’m not sure what to make of the other good and bad diagonal bands. Leaded gas, Vietnam war, vaccinations, …?
Why 1919/1920?
Getting back to the 1919/1920 France birth cohort, why people born then more likely to die earlier than other birth years, even much later in life. World War II seems like an obvious culprit, but then you’d expect the effect to be smeared across nearby years.
I learned about another theory related to the Spanish Flu pandemic of 1918-1919. A few studies suggest a lifelong effect from in utero influenza exposure. For example, The Long-Lasting Influenza: The Impact of Fetal Stress During the 1918 Influenza Pandemic on Socioeconomic Attainment and Health in Sweden, 1968-2012 and Early Life Exposure to the 1918 Influenza Pandemic and Old-Age Mortality by Cause of Death. It still seems like the effect would be seen in the 1918/1919 cohort as well, but I don’t have a better explanation.
