When we get something wrong

A correction is anything the site said or did that was wrong and has been put right. That includes a number that did not match its publisher, a figure that had been rounded on its way here, an edition filed under the wrong year, a rule that produced a misleading result, a country placed in the wrong region, a page that should not have been published, and a page that could not be used with a screen reader. It does not include an improvement, a new section or a re-import that changed nothing; those are on the Updates page.

A publisher restating its own series is not a correction of ours, but the site records the restatement on that statistic’s page, and treats a reader’s confusion about it as something to fix in the wording.

1. The cause is found, not just the symptom. A wrong number usually has a wrong process behind it. When two series turned out to have been rounded on their way through a redistributor, every statistic on the site was checked the same way before the two were fixed.
2. The fix is made at the source. The dataset is re-imported, the rule is changed, or the page is rewritten. Where a wrong result had already been published in a fixed edition, the edition page says so; a published edition is not silently rewritten.
3. The rule that prevents it is written down. Most entries below ended in a rule: store what the publisher published, read from the producer, check the licence first. The rules are on the Methodology page.
4. It is recorded here, on the day the fix goes live, saying what was wrong in plain words. The person who reported it is thanked by name if they wish.

Write to contact@nationsinquiry.com or use the contact page. The most useful message names the page, the country and the figure, and says what you think it should be and where you saw it — a link to the publisher’s own table settles most questions in a minute. A person reads every message, and you will hear back whether or not it turns out to be a mistake.

1. September 18, 2026

   A chart summary listed the wrong countries

   What was wrong

   The summary under "Countries With The Lowest Average Salaries in 2024" — shown on its own page, on every card linking to it, and as the description search engines were given — listed the HIGHEST-paying countries instead: Switzerland on 8,111, Luxembourg on 6,633, the United States on 6,455. It was an automatic excerpt taken from the wrong table when the article was imported from WordPress, and it had been published that way since 12 January 2025. Two other charts carried the same kind of imported text, including one that began with the word "Edit", and one chart had no summary at all.

   What changed

   All twelve summaries were rewritten from the charts themselves and now say what each chart actually found. Every chart also publishes its rows as a table, so a reader no longer has to take a summary's word for what the picture shows.

   The chart

2. September 18, 2026

   A chart blamed a continent's wages on its politicians

   What was wrong

   The line printed across the top of "Countries With The Lowest Average Salaries in 2024" reads: "African nations are struggling badly with their economic conditions due to corrupt and bad policy makers." That is a causal claim about a continent, presented as a finding of a chart that shows one wage average converted at market exchange rates. It is not a view this site holds, and the data below it does not support it.

   What changed

   The sentence is part of the published image and cannot be edited out of it, so the chart is kept as published and carries an editor's note beside it saying plainly that the claim is an opinion and not a finding. A house style guide for future charts (brand/chart-posters.md) now rules out standfirsts that explain why a country is where it is; that belongs in the article, where it can be sourced and argued.

   The chart

3. September 11, 2026

   The first Country Index scored countries out of 100

   What was wrong

   Version 1 of the Country Index averaged every statistic on the site into one number per country and printed it as a score out of 100. That number had nothing anchoring it — it was a mean of positions that moved whenever a dataset was added — but it read like a measurement somebody had taken. It also let the index drift toward whichever subject had the most statistics imported.

   What changed

   Version 2 replaced it. A fixed set of key statistics votes, one topic at a time; every value becomes a place within its own field before it is averaged; and only the position and its field are published. The old scores are withdrawn and are not comparable with the new positions.

   How version 2 works

4. September 10, 2026

   Poverty and inequality figures were the World Bank's rounded copies

   What was wrong

   The poverty headcount, the Gini index and the income share of the lowest fifth were read through the World Bank's general data API, which rounds them to one decimal place. Belgium's poverty headcount, published as 0.02 per cent, arrived as 0 — and five more countries below 0.05 per cent read as zero with it.

   What changed

   All three are now read from the Bank's own Poverty and Inequality Platform, which publishes them to two decimal places. Where the platform offers both an income and a consumption survey for the same country and year, the site takes whichever one the Bank's own published figure uses.

   Datasets

5. September 9, 2026

   Under-five mortality was rounded to one decimal place

   What was wrong

   The series was read through the World Bank's data API, which redistributes UNICEF's estimates rounded to one decimal place. UNICEF itself publishes them to fourteen, with uncertainty bounds.

   What changed

   The series is now read from UNICEF's own data warehouse, at full precision. The comparison that found this was then run for every statistic on the site, and it is a required step before any new statistic is added.

   Where a number is read from

6. September 7, 2026

   Values were rounded on import in two indices

   What was wrong

   The Human Capital Index was stored rounded to two decimal places on a scale of 0 to 1, which tied 172 economies onto 48 distinct values in one edition. The Logistics Performance Index lost two editions the same way: the source separates 159 economies with no ties at all, and rounding manufactured about fifty.

   What changed

   Both were re-imported digit for digit, and a rule was adopted for every import on the site: store exactly what the publisher published, and trim precision only when a number is drawn on screen.

   Numbers are stored as published

7. September 7, 2026

   The Logistics Performance Index carried no ranks, and one edition had the wrong year

   What was wrong

   The index was read from a World Bank endpoint that serves scores only, so its pages showed no rank although the Bank publishes one for every edition. The same endpoint year-stamped the final survey 2022; the report itself, and the Bank's Data360 platform, call it 2023.

   What changed

   The index is now read from Data360, which carries the Bank's own rank for all seven editions plus full-precision scores. The 2022 edition was refiled as 2023, and the old address forwards to the new one.

   Logistics Performance Index

8. August 29, 2026

   Thirteen indices were published without permission to publish them

   What was wrong

   Thirteen indices had been imported before the site had a rule about licences. On checking, their publishers' terms did not permit commercial reuse: some carried a non-commercial licence, some an explicit reservation of all rights, and some published no terms at all.

   What changed

   All thirteen were removed, with their pages and their data. A licence check now comes before any import, and a dataset is published only when its publisher's own terms permit it. The licence and the date it was checked are printed on every dataset page.

   Sources & publishers