This page sets out how work on The New Economist is researched, sourced, rated and corrected. It is published so that readers can hold the site to a stated standard rather than an implied one. Where the site falls short of what is written here, that is a correction, and it will be logged.
What we publish
Four kinds of piece, each with a different job.
- Explainers set out how a mechanism works. They are written to remain accurate for years and are revised when the underlying institution or rule changes.
- Evidence reviews ask what the research actually shows on a contested question, and carry an explicit confidence rating.
- Data explainers take an official statistical release and explain how to read it, including what it does not measure.
- Analysis is pegged to something in the news but written to the durable mechanism underneath it, so that it still makes sense a year later.
How sources are chosen
- Primary sources come first: statistical releases from the Office for National Statistics, the Bank of England, HM Revenue & Customs and the Department for Work and Pensions; fiscal documents from the Office for Budget Responsibility and HM Treasury; and peer-reviewed research.
- Every empirical claim links to the source that supports it. Where a paper has a DOI, the DOI is the link, because it survives journal reorganisations.
- Press coverage of a study is not treated as evidence of what the study found. The paper is read.
- Where a source is a working paper or preprint rather than peer-reviewed work, the piece says so at the point of citation.
- Research behind a paywall is read in full through institutional access and summarised in our own words. Paywalled text is never reproduced here.
- Think tank and advocacy research is cited where the method is sound, with the funding and affiliation stated so readers can weigh it themselves.
How evidence is weighted
Economics rarely offers clean experiments, so the question is always how credibly a study can separate cause from correlation.
- Randomised trials and genuine natural experiments carry the most weight, followed by quasi-experimental designs — difference-in-differences, regression discontinuity, instrumental variables — where the identifying assumption is stated and plausible.
- Panel and cross-sectional studies inform a question but rarely settle a causal one, and are described as such.
- A single study establishes very little. We look for replication across different designs, countries and time periods before treating a finding as established.
- Where a literature shows signs of publication bias, or where results weaken sharply in pre-registered work, that is reported alongside the headline finding.
- Effect size matters as much as statistical significance. A precisely estimated but trivial effect is reported as trivial.
- Where a finding holds in one country’s institutional setting but has not been tested in the UK, that limitation is stated rather than glossed.
Confidence ratings
Evidence reviews carry a rating so you know how much weight a conclusion can bear. The rating attaches to the specific claim it sits beside, not to the article as a whole — a single piece may contain a strong finding and a contested one.
| Rating | What it means |
|---|---|
| Strong | Findings replicate across multiple credible designs, research groups and settings. Both the direction and the rough magnitude are well established. A single new study is unlikely to overturn the conclusion. |
| Moderate | The direction of the effect is clear and consistent, but the magnitude is uncertain, or the evidence rests largely on one design or one national context. |
| Mixed | Credible studies using sound methods reach different conclusions, and the disagreement is not explained by obvious differences in method, period or setting. |
| Contested | The evidence base is too thin, too dated or too methodologically disputed to support a confident answer. The positions are set out without adjudicating between them. |
Each review also states how many distinct sources underpin it. A high source count is not itself a strength — ten papers sharing one dataset are weaker than three using independent ones — but it tells you the shape of what was read.
What we do not do
- No financial, investment or tax advice. Personal finance pieces explain how a mechanism works and what the trade-offs are. They are not a recommendation, they take no account of your circumstances, and they are not a substitute for advice from a regulated adviser.
- No forecasts dressed as findings. Where a projection is cited, whose projection it is and how that forecaster has performed are both stated.
- No anonymous sourcing. Nothing here rests on unpublished briefings or unnamed insiders.
- No writing to a conclusion. Where the evidence points somewhere politically inconvenient — to any party — it is published as it stands.
Use of AI tools
AI tools are used here for drafting assistance, structuring and copy-editing. They are not used to generate factual claims, statistics or citations, because current systems produce plausible-looking references that do not exist.
Every figure, quotation and citation in a published piece is checked against the primary source by a person before it goes live. Nothing is published unread. A person is accountable for every article here, including anything a tool helped draft, and that accountability is exercised through the corrections process below.
Independence and funding
- The site is independently owned and self-funded, supported by display advertising.
- Advertising is served programmatically. Advertisers have no sight of editorial content before publication, no influence over what is written, and are not told what their advertising will appear beside.
- We do not publish sponsored posts, paid links or affiliate content. If that ever changes it will be labelled on the piece itself and stated here first.
- No funding is accepted from political parties, campaign groups or lobbying organisations.
Corrections
- Substantive errors — anything affecting a factual claim, figure or conclusion — are corrected and logged on the corrections page, with the date and a description of what was wrong. Errors are described, not silently overwritten.
- Typographical and formatting fixes are made without a note.
- Where a piece is revised because new evidence has been published, the revision is dated and the reason given, and the confidence rating is reassessed.
- If you think something here is wrong, write to editor@neweconomist.co.uk. Corrections are welcome regardless of how they are phrased.
Authorship
Pieces are published under the masthead rather than under individual bylines. That is an editorial decision: the argument on the page should stand on the sources cited and the reasoning shown, not on who is making it. Anonymity is not a licence to cut corners, which is why the sourcing rules above are specific enough to be checked and the corrections log is public.
Research draws on peer-reviewed literature accessed through university library subscriptions, alongside published statistical and fiscal sources. Where a piece reaches into an area outside the site’s usual range, that limitation is stated in the article rather than left for the reader to find.
The test we would ask you to apply is the one that works regardless of a byline: follow the links. Every empirical claim here points at something you can read yourself. If a link does not support what it is attached to, that is an error, and we want to hear about it.
These standards were last reviewed in August 2026. Substantive changes to this page are themselves logged as corrections.