Measuring and projecting life expectancy

Every series the site tracks has changed its own history at least once. Excess-mortality estimates for 2020 and 2021 moved by a quarter or more when the ONS changed method in 2024, and differed by a factor of two or more between agencies depending on the baseline chosen, whether ages were standardised, and how the pre-pandemic trend was extended; census results in 2021 and 2022 rebased population estimates and every rate calculated from them, in some countries by several per cent; late registrations lift recent death counts for months; a shift in a cause's share can be a change in coding as readily as a change in deaths, since certification practice differs between countries and automated coding arrived at different times. None of this is error in the ordinary sense — it is how statistics improve — but its size is rarely reported alongside the figure, and conclusions drawn from a vintage do not always survive the next one. The site's revisions stream records each change as it happens; the question here is the pattern behind them: which series revise most, in which direction, and whether revisions are getting larger as methods become more model-based and less counted.

Projection has the same character with a longer horizon. Official projections of life expectancy have been wrong in the same direction for decades — too low, until the 2010s, when in several countries they became too high as improvement stalled — and the question is whether anyone forecasts mortality better than the offices, and how one would know. The models are well described: extrapolative, cause-specific, expert-judgement, ensembles, and now machine-learning approaches that claim more from the same data. Their track records are less well described, because forecasts are seldom scored against outturns in public and the horizon that matters is decades. What the evidence does show is that the errors come mostly from turning points — the stall, the pandemic — that no extrapolative model anticipates, and that cause-specific and expert-judgement models fared no better at them. The site's projection-versus-outturn pages are the evidence for this driver as they accumulate.

Status

building

Direction

mixed

Would change our view

A published comparison scoring competing forecasts against outturns over a decade or more; a method that demonstrably anticipated a turning point rather than fitting it afterwards; agreement among the main agencies on a standard excess-mortality baseline; or an office publishing its revision history as a matter of routine.

Trackers

Gaps

Public forecast archives and revision histories in most countries; any scoring of the actuarial and insurance models, which are private; forecasts for ages over 90, where the data are thin and the stakes highest; the Japanese and Korean offices' practice.