Accuracy

How successive UN World Population Prospects revisions projected world population versus what materialized -- scored release over release, in the platform’s own versioned numbers.

Milestone walk

World population projected for 2050 and 2100, one step per revision, in publication order.

World population, 2050/2100 by revision

9.31B20109.55B10.85B20129.73B20159.77B20179.74B10.88B20199.71B10.35B20229.66B10.18Bcurrent2050210002B4B6B8B10B

un-wpp (archival revisions) · multiple · release 2026.08.3

CSV

Bias, by revision and horizon

Total fertility rate. Positive mean bias: the revision’s projected TFR came in above what materialized (over-projected). Negative: under-projected.

RevisionHorizon (years)Mean biasMAPEOver-projected
20101-5-0.09810.0%68.8%
20106-10-0.04313.7%71.9%
201011-15+0.03718.9%85.4%
20121-5-0.0147.6%69.1%
20126-10+0.03311.5%77.3%
201211-15+0.11116.7%88.7%
20151-5+0.0389.9%73.2%
20156-10+0.11115.0%86.6%
20171-5+0.0459.1%78.4%
20176-10+0.11814.3%83.5%
20191-5+0.10812.0%86.6%
20221-5+0.0516.2%72.9%

release 2026.08.3

Country probe

Pick a country to compare its actual total fertility rate against what each revision projected.

Select up to 2 a country (0/2 selected).

Total fertility rate: actual vs. revisions

Select a country above to see its projection history.

— · — · release —

CSV

Methodology

How scores are computed
Every projection score compares one revision’s projected value, at a given future year, to the value the scoring release’s current data now reports for that same year and entity.
error
projected − actual
relError
error / actual
horizon
year − estimateEndYear

Mean bias is the average of error across scored rows; MAPE is the average of |relError|. Rows are bucketed by horizon into 1-5, 6-10, and 11-15 years out; horizons beyond 15 years are not scored.

Actuals are the scoring release's current best estimates and are themselves revisable.