France: in 2017 the UN almost hit the 2023 population — its errors cancelled out
The 2017 projection for 2023 was off by just 0.1%, 33.8k people. Its component errors point in opposite directions.
More
A population projection is built from four parts: how many people live there at the start, how many will be born, how many will die and how many more will arrive than leave. The UN has since estimated each part anew, which shows where the projection went wrong. Here the total came out almost right, while the parts missed in opposite directions. In detail: the UN later put the 2015 population 459k higher; 382k fewer babies were born than expected; 217k more people died.
A bar is how much one part moved the total against the projection: right means more people than projected, left means fewer; the number is the size of the move. More births and fewer deaths add people; fewer births and more deaths take them away. “2015 population higher / lower”: the UN later estimated the population the projection started from differently from what it assumed then. The last bar is the total; the parts add up to it within a small residual.
- UN 2017 projection for 2023
66.4M
- UN 2024 estimate
66.4M
- Population in 2015: in the projection → estimated now
64.5M → 64.9M
- Births 2015–2023: assumed → estimated now
6.07M → 5.68M
- Deaths 2015–2023: assumed → estimated now
4.77M → 4.99M
- Net migration 2015–2023, arrivals minus departures: assumed → estimated now
+640k → +825k
- Residual of the calculation
−10.8k
The UN 2017 projection for 2023 against the UN 2024 estimate. The gap is split along the population balance: the 2015 population (estimated now minus the one the projection started from), then births, deaths and net migration over 2015–2023 (estimated now minus assumed; for deaths the other way round, since more deaths mean fewer people). Residual −10.8k: rounding and the five-year periods of older projections. Chosen: the most recent projection of at least five years that was off by at most 1% and by less than a third of the change it expected, while its parts missed in opposite directions and cancelled out by 96% (1 − |sum of the parts| / sum of their sizes). Countries whose errors another card already tells are skipped.
Both the projection and the current figure are UN estimates: in countries without full registration of births and deaths they come from a model, not a count.
A question for this picture
Did the projection understand how the population changes, or merely land on a similar total?