How IQ Metrics Research works: methods and standards

01 Key points

1. Every number has a named source. National statistics agencies, international surveys such as NAEP, PISA and PIAAC, or our own aggregated test results. Each source is linked at the foot of the study that uses it.
2. Licences are checked before data is used. Sources under CC BY 4.0, CC0 or US public domain are used. Sources that forbid commercial reuse are refused, because iqmetrics.org sells a test.
3. Our own test results are reported as totals only. Groups with fewer than 100 people are left out, and results are described as average scores on the IQ Metrics test, never as the IQ of a country.
4. Close results are shown as ranges. When a source reports standard errors, rankings are re-run within that uncertainty and shown as likely-rank ranges.
5. No study is sponsored. The research desk is paid for by sales of the IQ Metrics test. No organisation pays to appear in a study, and nobody outside the team sees one before it is published.

02 What we publish

IQ Metrics Research publishes four kinds of work. Each kind carries its own label on the research page, so a reader always knows whether a number is our own measurement, a combination of public data or a summary of someone else’s study.

Data study

An index or ranking built by combining public datasets, such as federal test scores and census data. The method, weights and full table are published.

Own-data report

Totals from people who took the IQ Metrics test: averages and differences between groups such as countries or age bands. Totals only, never individual results.

Research explainer

A plain-English account of published research, with a full citation to the original paper and a note on what it can and cannot show.

Technical report

Documentation of how our own test and our studies work, like this page.

03 Where does the data come from?

We use a source only when its licence allows reuse on a commercial website with credit. Copying one public table earns nobody’s trust, so studies combine several sources and say exactly which number came from where.

Table 1

Sources we use, plan to use and refuse

Licence status as checked in October 2026. "Used" means the figures are in a published study; the study's own source list says which.

Source	Publisher	What it gives us	Licence	Status
NAEP, the Nation’s Report Card	National Center for Education Statistics	State test scores, standard errors and significance tests	US government work, public domain	Used
American Community Survey	US Census Bureau	Education, income and population by state	US government work, public domain	Used
Annual Survey of School System Finances	US Census Bureau	Spending per pupil by state	US government work, public domain	Used
PIAAC Skills Map	National Center for Education Statistics	Adult literacy and numeracy estimates by state	US government work, public domain	Used
NCPT dataset	Lumos Labs, on Zenodo	Scores of 757,427 adults on 15 thinking tests, 2013 to 2020	CC BY 4.0	Used
World Bank Open Data	World Bank	Population, learning and education indicators by country	CC BY 4.0	Used
Organisations’ own published figures	Mensa International, national Mensas, Guinness World Records	Membership counts and records, quoted as each organisation states them	Quotation with credit; every page is saved on the day it is read	Quoted
Wayback Machine	Internet Archive	Dated copies of a page, to show how a published figure changed over time	Public archive; quoted with the capture date	Used
Google News results	Google	How many stories a topic got, month by month; never the stories’ text	Counts only	Counted
SAT and ACT state averages	College Board; ACT	State averages and the share of students tested	Copyright; shown with credit, not redistributed in our CSV	Shown with credit
Published research papers	Journals such as Intelligence and Psychological Science	Earlier findings to compare our results against	Cited, never copied	Cited
PISA and PIAAC international data	OECD	Student and adult skills by country	CC BY 4.0	Planned
OpenAlex	OurResearch	Counts of research papers by country and field	CC0	Planned
IQ Metrics test results	IQ Metrics	Totals by country and age band	Our own data; totals released under CC BY 4.0	Planned
Global Health Observatory	World Health Organization	Not used	CC BY-NC-SA 3.0 IGO (non-commercial)	Refused
Global Burden of Disease	Institute for Health Metrics and Evaluation	Not used	Non-commercial user agreement	Refused

Figures that an organisation publishes about itself, such as a membership count, are quoted exactly as written, with the date we read the page, and we keep a saved copy of that page. We do not adjust them, and a study that uses them says so in its method.

04 How a study is built

Every data study goes through the same eight steps. Steps two to six run in code, so the study can be rebuilt from scratch when a source releases new data.

1. 01

   Ask one question

   Write down what is being measured and what is not, before any data is pulled. A study of school test scores is a study of school test scores, not of intelligence.

2. 02

   Pull from the source

   Every dataset is downloaded by a script from the publisher, and the raw files are kept, so each number can be traced back to the file it came from.

3. 03

   Clean and match

   Places and groups are matched across sources by code, not by name, and every exclusion is written into the method.

4. 04

   Combine

   Measures on different scales are standardised before they are combined. The weights and formulas are published with the study.

5. 05

   Carry the uncertainty

   Standard errors and margins of error from the source are carried through to the final ranking.

6. 06

   Check the words against the data

   The page is generated from the dataset, and the build stops if a number written in the text disagrees with the data.

7. 07

   Review

   The Editor-in-Chief reviews every study before publication, with the method open beside the text.

8. 08

   Publish and refresh

   The study goes live with its CSV, citation and version number, and is refreshed at the same address when the source updates.

05 Our own test data

Reports built from IQ Metrics test results follow stricter rules, because they describe real people.

• Only totals are published: counts, averages and differences between groups.
• A group with fewer than 100 people is left out of every table and chart.
• Only each person’s first attempt is counted, so retakes do not inflate a group’s average.
• Results are reported as average scores on the IQ Metrics test, with the date range and the number of people behind each figure.
• Nothing that identifies a person, such as a name, email address, IP address or certificate number, is ever part of a research dataset.

06 How certain are the rankings?

A ranking turns small differences into whole places. When two states are separated by less than the margin of error, calling one 21st and the other 22nd says more than the data does. So where a source reports its uncertainty, we re-run the ranking many times, each time drawing every value from within its reported error, and publish the range of places each one lands in.

In The smartest states in the US, 2026, for example, the index was re-run 10,000 times within NAEP’s standard errors and the Census Bureau’s margins of error. The first place held in every run, while states in the middle of the table moved across ten places or more.

07 Can I reuse and cite the data?

Studies, charts and datasets are published under the Creative Commons Attribution 4.0 licence ^[1]. You may copy, adapt and republish them, including commercially, as long as you credit IQ Metrics Research and link to the study you used.

Source: IQ Metrics Research, iqmetrics.org/research/

The credit line we ask for, with a link to the study page

Every study ends with a ready citation in APA, MLA, Chicago and BibTeX styles, and every chart has a link of its own, so you can point readers to the exact figure.

08 How can I check a number myself?

A study is only worth citing if a reader can check it without asking us. Every number on a research page can be traced in four steps.

1. Under the chart or table, select Data. It downloads the exact rows that figure was drawn from, as a CSV.
2. Open the study’s CSV. It holds every input the study used, one row per place, group or test, with a column dictionary beside it.
3. Follow the source. The reference list links each publisher’s page, and the method names the table or file the number came from.
4. Compare the text with the data. Each page is generated from its dataset, and the build stops if a number written in the text disagrees with the data behind it.

09 Who pays for the research?

The research desk is paid for by sales of the IQ Metrics test. No study is sponsored, no source or organisation pays to appear in one, and nobody outside the team sees a study before it is published. Where a study describes an organisation, such as Mensa, the page says that IQ Metrics is not affiliated with it.

A study that would help sell the test is held to the same rules as one that would not. Our own-data reports say on every chart that our test takers are a self-selected online sample, and no research page calls a skills index or a test score an IQ.

10 Working with journalists

Every study is built to be quoted. There is no embargo, no sign-up and no permission to ask for: the credit line and a link are the only conditions.

• Ready to paste: a news-style summary, numbered statistics with a copy button that adds the credit, and quotes attributable to the named analyst, all in the study’s press kit.
• Images: every chart as a print-ready PNG, a 1200 x 630 share card with the credit built in, and a card for any single state, age or country where the study has a lookup.
• On request: the breakdown behind any number, the significance test for any pair of places, and a short written comment from the analyst on the method.
• Reply time: usually within one working day through the contact page.

Source: IQ Metrics Research, iqmetrics.org/research/

On air or in print, "IQ Metrics" and the page address are enough

11 How are mistakes corrected?

When we find a mistake, or a reader reports one through the contact page, the study is corrected on the same page and the change is recorded in its version history with the date and what changed. Studies are refreshed when their sources release new data, and reports built on our own test results are refreshed every January, always at the same address.

The research page lists every version note across all studies, newest first, so a reader can see what changed without opening each study.