# UN WPP 2024 population projections — NTA master-database countries

*Built 15 August 2026*

## Source

United Nations, Department of Economic and Social Affairs, Population Division,
**World Population Prospects 2024**, distributed as the official `wpp2024` R data
package (v1.1-3, dated 2024-11-26, maintainer Hana Sevcikova, UN Population Division),
obtained from `https://github.com/PPgp/wpp2024`.

Underlying objects used:

| Object | Contents |
|---|---|
| `popAge1dt` | Estimates, single-year age, by sex, 1950–2023 |
| `popprojAge1dt` | Projections, single-year age, by sex, 2024–2100, medium / high / low variants |
| `countries` | UN M49 country codes and names |

## Files

| File | Rows | Contents |
|---|---|---|
| `WPP2024_pop_projections_NTA.csv` | 1,656,501 | Projections 2024–2100, annual, single-year age 0–100+, three variants, by sex and both sexes |
| `WPP2024_pop_estimates_NTA.csv` | 530,654 | Estimates 1950–2023, annual, single-year age 0–100+, by sex and both sexes |
| `WPP2024_pop_NTA_compact_5yr.csv` | 458,944 | Both-sexes only, five-year year steps (1950–2020, 2023, 2025–2100), estimates and all three variants stacked — sized for browser loading in NTA Lab |

## Schema

`Country, variant, year, age, popM, popF, pop`
(the compact file drops `popM`/`popF`).

- **Country** — spelled exactly as in the NTA master database (`US`, `South Korea`,
  `Swaziland`, `Sao Tome`, `Laos`, `Russia`, `Turkey`, `Vietnam`, `Congo`, `Moldova`,
  `Taiwan`), so the file joins directly on the master `Country` key.
- **variant** — `estimates`, `medium`, `high`, `low`.
- **year** — calendar year, 1 July reference date.
- **age** — completed years; **100 is an open interval (100+)**.
- **pop** — persons (**not** thousands). WPP publishes in thousands; values were
  multiplied by 1,000 and rounded to whole persons. `pop = popM + popF`.

## Variant definitions

The high and low variants differ from the medium **only in fertility**: total fertility
is set 0.5 births per woman above (high) and below (low) the medium projection from
2025 onward. Mortality and international migration follow the medium assumption in all
three. Consequently, for any cohort already born before 2024, the three variants are
numerically identical — verified in the build. Divergence enters at age 0 in 2024 and
propagates up the age distribution one year per year, so by 2100 the variants differ
only below roughly age 76.

If you later want scenarios that vary mortality or migration (constant-mortality,
zero-migration, no-change, or the probabilistic 80/95% intervals), those series are in
the same package and can be added on the same schema.

## Country coverage

All 71 master-database countries are present; no country had to be dropped. This is
better coverage than the CPI file, which excludes Argentina and Swaziland.

Two name resolutions worth recording:

- **Congo** → UN code 178, Republic of the Congo (not DR Congo, code 180) — the same
  reading used when the CPI file was built.
- **Taiwan** → UN code 158, published by the UN as "China, Taiwan Province of China".
  This is a genuine WPP series, not an imputation, and is not included in the China
  (code 156) totals.

## Validation performed

- 2024 both-sexes totals reproduce published WPP 2024 figures: China 1,417.7 m,
  India 1,457.4 m, US 346.4 m, Japan 123.4 m, South Korea 51.7 m, Taiwan 23.2 m,
  Nigeria 235.1 m.
- Variant ordering holds: summed across the 71 countries, 2100 population is
  8.94 bn (high), 6.24 bn (medium), 4.22 bn (low).
- Pre-2024 birth cohorts identical across variants: pass.
- `pop = popM + popF` by construction.

## Notes for use with NTA profiles

- Master-database profiles have heterogeneous terminal ages (80, 85, 90, 91, 99, 100).
  Any application of these projections to a profile needs the profile's terminal age
  collapsed against the projection, or the projection top-coded to match. NTA Lab's
  existing population alignment step already does this for the historical population
  file and should be pointed at the same logic here.
- Estimates and projections are on a consistent WPP 2024 basis, so the estimate file
  will *not* equal the population figures already stored in the master workbook, which
  come from country sources and earlier WPP rounds. Keep them as separate series rather
  than splicing.