Living Worlds

Research · Lifetimes · Oct 5, 2026

The UN's Bayesian model, stacked

Not met135 countries

The test at a glance

What it tests
Accuracy against reality: life expectancy in 2019, alone and stacked.
Data
UN World Population Prospects 2024, the countries of the first backcast; fitted up to 1999.
Held out
2019.
Compared against
The UN's Bayesian hierarchical model (bayesLife 5.3.1), Living Worlds' pooled-gain model, and the two stacked.
Bar
Error at most 1.2 years outside shock countries AND at most 1.8 overall, for both sexes.
Result
Pooled-gain better alone than stacked; neither met.

Living Worlds' own pooled-gain model did better alone than stacked with the UN's Bayesian hierarchical model; neither reached the target. Scored on 135 countries.

Protocol, result and scorer

Protocol Result

Lifetimes

More in this area.

Oct 5, 2026 Lifetimes Life expectancy, predicted twenty years ahead Not met Oct 10, 2026 Lifetimes Life expectancy twenty years on, against 2024 national figures Not met Oct 11, 2026 Lifetimes Women's life expectancy twenty years on, a year nobody had read Met (women only) Oct 11, 2026 Lifetimes Men's life expectancy twenty years on, a year nobody had read Met Oct 5, 2026 Lifetimes Shocks as events Met for women, not for men Oct 5, 2026 Lifetimes Honest intervals twenty years ahead, a fresh year Met for women, not for men Oct 5, 2026 Lifetimes Men as women minus a gap Primary met; bias remains Oct 6, 2026 Lifetimes Life expectancy in European regions, ten years ahead Met for men, not for women Oct 6, 2026 Lifetimes European regions on a window nobody had read Met Oct 6, 2026 Lifetimes Lived, not learned: children ever had Met Oct 6, 2026 Lifetimes Siblings, from partnerships that make fathers Not met Oct 10, 2026 Lifetimes Siblings on the mother's side, in four years never read Met Oct 11, 2026 Lifetimes Siblings on both sides, in three years never read Met
All research