HUMANITYVECTOR
Menu

Life expectancy at birth · World · 1 year ahead

2025: the model assigns an 80 percent probability to 72.84 to 73.83 years

Point estimate 73.69 years from Ridge regression on lags and drivers, against 73.48 years observed in 2024. ridge beats last_value on out-of-sample MAE by 35.2 percent over 16 cutoffs (minimum 8, margin 5 percent).

ModeledMetric page
Distribution over the history

Fan chart

60 observed points to 2024, then 3 forecasts with 80 percent intervals55.9365.4975.05196519731981198919972005201320212029
Solid: observed (years). Dashed and shaded: model point estimates and 80 percent intervals from empirical residual quantiles. Filled dot: the record on this page. Green or red dot: the realised value, inside or outside its interval.
Quantiles
p5 72.66 · p10 72.84 · p25 73.61 · p50 73.65 · p75 73.68 · p90 73.83 · p95 74.52
Source
empirical quantiles of 12 in-window 1-step residuals; no normality assumed
The record

Immutable fields

Key
life-expectancy:WLD:1
Id
941623c9-70a5-400f-b5a4-381081987311
Created
2026-09-07 22:16 UTC
Data cutoff
2026-09-07 22:16 UTC
Origin
2024 = 73.48 (obs 70501)
Target period
2025 (366 days)
Point
73.69 years
Interval
72.84 to 73.83 at 80 percent
Model
ridge models_v1.targets_v1
Baseline
last_value
Snapshot
9ea2a4457533d18622b9f1a2dab363b2d13d70e7479424f3769e97c9c0ba822c
Data mode
ingested_at
Status
active
Drivers

Statistical association, not causal

Ridge coefficient times the standardised input at the origin, in years. These are associations learned in the training window, labelled as such; nothing here is a causal effect.

FeatureInputCoefContribution
y[t-0]73.48-0.0146-0.0225
y[t-1]73.33-0.0124-0.0191
cereal-yield[t-1]4,231-0.0096-0.0172
electricity-access[t-1]91.6-0.0070-0.0162
world-population[t-1]8,062,923,417-0.0074-0.0129
Outcome

Not yet realised

The observation for 2025 has not been ingested. The weekly forecast.evaluate job attaches it when it arrives; the record above does not change.

Scenario assumptions

What the forecast takes for granted

  • No structural break in life-expectancy between the data cutoff and the target period.
  • The source keeps publishing the series on the same definition; a rebasing or redefinition invalidates the comparison.
  • The lagged statistical associations estimated in the training window hold over the horizon.
  • Driver series used, lagged: world-population, cereal-yield, electricity-access. Their own future values are not forecast; only values known at the cutoff enter.
Falsifiers

What would show this forecast wrong

  • If the first published value of life-expectancy (WLD) for 2025 is below 72.84 years or above 73.83 years, the 80 percent interval is falsified.
  • The model assigns a 10 percent probability to a value below 72.84 years and a 10 percent probability to a value above 73.83 years; interval misses should occur about one time in five over many forecasts, and a run of misses well above that rate falsifies the calibration.
  • The point forecast is above the last observed value (73.48 years); a realised value at or below that level falsifies the direction call.
Competing models

Scorecard for this target and horizon

ModelStatusCutoffsMAERMSEDirection80% coverageSkill vs last value
Ridge regression on lags and drivers
models_v1.targets_v1
champion160.27490.559388%56%35.2%
Drift
models_v1.targets_v1
challenger160.33760.566188%63%20.4%
Holt (exponential smoothing with additive trend)
models_v1.targets_v1
challenger160.31590.716588%50%25.5%
Last value
models_v1.targets_v1
baseline160.42410.58160%69%0.0%
Linear trend
models_v1.targets_v1
challenger160.34750.57775%31%18.1%
Champion: Ridge regression on lags and drivers. ridge beats last_value on out-of-sample MAE by 35.2 percent over 16 cutoffs (minimum 8, margin 5 percent).
Model card

Ridge regression on lags and drivers

Ridge regression on lags and drivers

Purpose. Point and 80 percent interval forecasts of life-expectancy (WLD) 1, 2, 5 years ahead. Owner: Forecast Lab (HV 3.0).

Method

A direct multi-horizon ridge regression: the change from the origin to the target period is regressed on lagged target values (lags [1,2]) and lagged driver series (lags [1]), features standardised, penalty lambda 1. Drivers: world-population, cereal-yield, electricity-access. Every driver enters lagged, so no feature uses a period after the origin.

Uncertainty

Intervals are empirical: the model is refitted at earlier origins inside the training window, the realised errors at the same horizon are collected, and the interval is the 10th to 90th percentile of those errors around the point forecast. No normality is assumed.

Training and evaluation

Training window: 65 periods, 1960-01-01 to 2024-01-01. Evaluation: rolling-origin backtest over 16 cutoffs using pseudo real-time data (the earliest ingested value per period; no vintage archive exists for this series, so revisions between the cutoff and today are not reproduced).

Known limitations

  • Structural breaks after the cutoff are not modelled.
  • Driver contributions are statistical associations, not causal effects.
  • Promotion to champion requires beating the last-value baseline on out-of-sample MAE by at least 5 percent over the minimum number of cutoffs; otherwise the baseline stays champion.
  • Retraining: every scheduled forecast.run refits on the data then available; backtests re-run from /admin/forecasts.
Version chain

Supersedes

This is the first record for its key.