# F1-conf-02 — Balanced-panel mean life expectancy, 2000–2022

## Question
How much did life expectancy at birth (WDI `SP.DYN.LE00.IN`) change from 2000 to 2022 across a fixed set of countries? The estimand is the unweighted mean taken only over countries with a recorded value in every year from 2000 to 2022 (the "balanced panel"). For comparison, a naive 2022 mean over all available countries is reported as a descriptive baseline.

## Data and exclusions
The input `data/life_expectancy_panel.csv` has 439 data rows (`n_input_rows`) covering 20 distinct iso3 codes (`n_countries_total`). Problems found (`data_issues`):
- **TUR is duplicated.** It appears for all 23 years under two names, "Turkey" and "Turkiye", with identical values. This is an artefact of assembling the file from two extractions. One row per (iso3, year) was kept, and the country is labelled "Turkiye".
- **CHE 2007 is empty.** That cell is treated as missing and is not imputed.
- **Four countries start in 2011.** BGD, ETH, IND and NGA have rows only for 2011–2022.

Exclusion ledger (`out/exclusions.csv`):
- 439 → 416 (23 duplicate TUR rows removed)
- 416 → 415 (1 empty value removed)
- 415 → 345 (70 rows from incomplete countries removed)

This leaves `n_analysed` = 345 country-years: 15 countries × 23 years.

## Method
1. Drop duplicate (iso3, year) rows. The script stops if duplicates disagree in value; none did.
2. Drop empty values.
3. Keep only countries with a recorded value in each of the 23 years.
4. For each year, take the simple mean over the balanced set.

The baseline is the mean of the 2022 values over all distinct countries with a 2022 record. The calculation is deterministic and uses no random seed (`random_seed` = null).

## Results
- **Balanced set.** It contains 15 countries (`n_countries_balanced`): AUS, AUT, BEL, CAN, DEU, ESP, FRA, GBR, ITA, JPN, KOR, NLD, SWE, TUR, USA (`balanced_countries`). Five countries are incomplete (`n_countries_incomplete`): BGD, CHE, ETH, IND, NGA.
- **Balanced means.** The mean was 78.060 years in 2000 (`mean_2000_balanced`) and 81.553 years in 2022 (`mean_2022_balanced`). The change is **+3.493 years** (`change_2000_2022_balanced`).
- **Baseline.** The naive 2022 mean over all 20 countries with a 2022 value is 78.692 years (`mean_2022_all_available`). It is lower than the balanced mean mainly because it includes BGD, ETH, IND and NGA. Those four have lower life expectancy and are not in the balanced set. CHE also counts in the baseline but not in the balanced set.

## Uncertainty
All 20 countries in the file have values, so these figures are descriptive statistics for a fixed set of countries, not estimates from a sample. No sampling standard errors are reported. Two things could change the numbers: the WDI values themselves (modelled estimates that get revised between releases), and which countries end up balanced. For example, CHE is excluded only because of one missing cell in 2007.

## Limitations and open issues
- The balanced set is 14 high-income or upper-middle-income OECD countries plus Türkiye. The result does not describe the full 20-country set or low-income countries.
- Means are unweighted, not population-weighted.
- Rounding the values to 3 decimals in the source file has a negligible effect.
- The task is complete (`feasibility` = "complete"). No requirement was left unmet.