Our statement on the Code of Practice for Statistics

Published 24 October 2026. Next review by October 2027.

refnation publishes the results of eight political referendums a day. Journalists, researchers and members quote those results, so they need to know how far they can rely on them. refnation is not an official statistics producer and does not have to follow the Code of Practice for Statistics. We have chosen to apply it voluntarily, because its three pillars of trustworthiness, quality and value are the right test for any figure that enters public debate.

This statement says how we apply each of the Code's ten principles, where we don't yet meet them, and what we will do about it. It is part of the Office for Statistics Regulation's voluntary application scheme, and we have asked to join its Register of TQV Voluntary Application. Applying the Code is our commitment; it is not an endorsement of our method by the Office for Statistics Regulation.

refnation is run by Refnation Ltd (company 17472835, England and Wales). Anyone in the UK aged 16 or over can join for free. Each member has one account and one vote per referendum. Each referendum is on the ballot for one day, settles at midnight, and its result is then final.

Members join by choice, so refnation does not simply count heads. It works out each member's political leaning from how they actually vote, and every headline result gives all seven leanings, from Hard Left to Hard Right, an equal say. The full method is at How results are calculated.

refnation's record starts on 29 September 2026, when this rule began. Referendums that settled earlier are kept as an archive and marked as early results: they were plain counts, and some include automated votes from our launch period, as our corrections log explains.

refnation is not a member of the British Polling Council, and its referendums are unofficial and carry no legal weight.

Show integrity. refnation is not affiliated with any party, campaign or media organisation, and it is not paid to run or word any referendum. Every result is calculated by the same published rule. Vote counts are kept by the database as votes arrive, and they stop changing when a referendum settles. When a vote or a referendum's wording has had to be corrected, the correction is listed, dated, in our corrections log.

Lead responsibly. The founder, Shane Stocks, is accountable for the method, the published figures and this statement. Questions and complaints go to press@refnation.uk.

Be transparent. The method is published in full, with a worked example. On every referendum page, the “by leaning” bars average to exactly the headline figure, so anyone can check a result. The method page is dated and says when the current rule took effect: the equal-say rule started on 29 September 2026, and referendums that settled before then keep the plain count they were published with. When we compare our results with national polls, we publish every comparison, misses included, along with the rules we used to pick the poll; see refnation vs the polls. Each comparison is a dated snapshot, so a figure someone has quoted never changes under them. The counts behind every settled result can be downloaded as a CSV from its page: the headline, the seven leanings it is the average of, and the votes by party, region and age. Corrections and changes to the method are listed, dated, in our corrections log.

Manage data responsibly. Refnation Ltd is registered with the Information Commissioner's Office (registration ZC254853) and handles personal data under UK GDPR. Votes are public by design, and our privacy notice says so: each referendum lists who voted and how. Age, gender, region, voting intention and how a member voted in the 2024 general election are all optional. A member's 2024 vote is never shown to anyone, and is used only combined with other members' answers. Figures we send to journalists are aggregates only.

Prioritise quality. A referendum shows a plain count of Yes and No until it has 50 votes, because an equal-say result on fewer votes swings too much. A result that rounds to 50% is reported as a dead heat, with no winner. Group figures always show how many voted, and in figures for journalists any group under 50 is marked as small.

Be rigorous. Each referendum is reviewed and, where needed, reworded to be neutral before it runs, and it states what a Yes and a No vote mean. Each referendum is scored for the political direction its answers point in, and a member's leaning is the running average of their scored votes. Since 29 September 2026, each vote keeps the leaning its voter had when they cast it, so a member whose views change later cannot change a published result; figures by party, region and age use members' current details. We check our results against national polls on the same issues about once a month.

Be open about quality. We state our limits plainly, on the method page and here:

• The headline result balances political leaning. It does not balance age, sex or region, as a poll does.
• Many members don't give their age, region or gender, so each group figure covers only those who did, and the number is always shown.
• Some groups are small: Welsh members, for example, number about 40 on a typical referendum.
• In our latest comparison with national polls, refnation came down on the same side on 24 of 27 issues; on the other 3 it did not. The typical gap was 7.5 points, against 15.7 for a plain headcount of the same votes. Every row is published.
• In October 2026 we tested whether weighting by age, sex and region, to ONS population estimates, would bring results nearer the polls on those same 27 issues. It made them further away (a typical gap of 12.9 points, against 7.5) and doubled the margin of error, so we do not use it. The full result is on the method page. Members differ from the UK mainly in their politics, not their age, sex or region, so a weight that corrects politics is what's needed.

Be relevant. One referendum each day is written on that day's news, and the rest are chosen from referendums members have asked. Every result is published, broken down by leaning, party backers, region and age, as soon as its referendum settles at midnight.

Be clear. We explain that a headline result is a result, not a headcount: “36% Yes” does not mean 36% of members voted Yes. We publish how to describe and cite our figures on the press page, and we use plain English throughout.

Be accessible. Every settled result is free to view, without an account, on its own page. While a referendum is live, its result is shown only to people who have voted on it, so early figures can't sway later votes. We test our main public pages with automated accessibility checks against the WCAG 2.2 AA rules, on both a phone-sized and a desktop screen. On 24 October 2026 every page in that check passed with no failures: the home page, the results and verdicts pages, a live and a settled referendum, a member's page, the method, the poll comparison, and the about and privacy pages. We have not yet had a full manual accessibility audit, which would test what automated checks cannot, such as using the site with a screen reader or a keyboard alone.

• The headline is not weighted to the UK population. Our test of weighting by age, sex and region (October 2026) made results less accurate, not more. Since 24 October 2026 we ask members how they voted in the 2024 general election, the weight most UK pollsters use. Once enough members have answered, we will test weighting to the 2024 result against the same poll comparisons, and publish what we find either way. By 30 November 2026.
• No one outside refnation has reviewed the method. Ask an independent statistician to review it, and publish their review. By 31 December 2026.
• refnation is run by one person, so checks are not independent of the person producing the figures. Automated checks on every result, and a second reviewer as the team grows. By ongoing.

We will review this statement every year, and whenever the method changes, and confirm our continued application to the Office for Statistics Regulation in its annual review. Comments are welcome at press@refnation.uk.