Brexit: what actually changed?
Ten years after the referendum. What was promised, what happened, and where Britain is demonstrably better or worse off.
How this section is framed
Not “Brexit was a disaster”. Not “Brexit was a success”. Both collapse dozens of separate outcomes into one political judgement.
This section takes the outcomes one at a time — trade, investment, migration, agriculture, fisheries, regulation, Northern Ireland, sovereignty — and reports what is measured, what is modelled, and what cannot yet be known. On several of them the evidence points clearly in one direction. On several it points clearly in the other. Where it points nowhere, the page says so.
The result
17,410,742
Leave
51.9% of valid votes
16,141,241
Remain
48.1% of valid votes
3.8pp
Margin
1,269,501 votes
72.2%
Turnout
Electorate 46,500,001
The margin was 1,269,501 votes — 3.8 percentage points, or about 1.9% of those who voted switching sides. Turnout, at 72.2%, was higher than any general election since 1992. Both facts matter: the result was close, and it was not a low-turnout accident.
23 June 2016. Percentages are of valid votes; 25,359 ballots were rejected. Electoral Commission, Results and turnout at the EU referendum ↗
The counterfactual warning
Read this before the numbers
The 4% figure is a comparison with a Britain that stayed in the EU. That Britain does not exist, was never observed, and cannot be checked against reality. It is the output of a model, published by the government’s own independent forecaster, and reviewed twice since.
What it does not mean
- It does not mean GDP is 4% lower than it was in 2016. GDP is higher than it was in 2016.
- It does not mean every sector, region or household is 4% worse off. Some are better off and some are much worse off.
- It does not mean 4% has already happened. The OBR judges about two-fifths of it had occurred before the trade deal even took effect, through investment uncertainty.
What it does mean
That the government’s independent forecaster expects the UK economy’s productive capacity to settle around 4% below where it would otherwise have been, and builds that into every fiscal forecast the Chancellor works to.
And it cuts both ways. A counterfactual cuts both ways. It is equally wrong to say “GDP grew, so Brexit cost nothing” — growth in the actual world tells you nothing about growth in the world that did not happen. Neither side gets to point at observed data and claim it settles a counterfactual question.
Six numbers, ten years on
4%
lower long-run productivity
The OBR’s assumption relative to remaining in the EU. Reviewed twice and unchanged.
Counterfactual15%
lower trade intensity
Both exports and imports, in the long run, relative to remaining.
Counterfactual14% lower
goods exports to the EU vs 2019
Real terms. Non-EU goods exports were 8% lower — so not all of it is Brexit.
28% higher
services exports to the EU vs 2019
Real terms. The finding that complicates every simple version of this story.
41%
of UK exports still go to the EU
£384bn in 2025. The EU also supplied 50% of imports.
171k
net migration, year to Dec 2025
EU net migration is now negative. Non-EU net migration was 350,000.
The full section
- How it happened1973 to 2026, and what the research says drove the vote.
- What was promisedClaims from both campaigns, checked. Remain’s failed too.
- The economyThe OBR’s 4%, and what it does not mean.
- TradeGoods down 14%. Services up 28%. Both are true.
- ImmigrationThe composition changed far more than the level.
- Where Britain gainedReal autonomy, mostly unconverted into measurable benefit.
- Northern IrelandGains and costs in the same place, at the same time.
- Who paid for the campaignRemain outspent Leave by 45%, and lost.
- The scoreboardTwelve areas, both columns filled. No total.
The key distinction
Brexit produced real political and constitutional gains in autonomy. The strongest official economic evidence simultaneously indicates a significant aggregate economic cost relative to remaining in the EU. Both can be true, and the argument that has consumed British politics for a decade largely consists of each side insisting the other’s half is imaginary.
Two things follow, and they cut in opposite directions. The OBR’s 4% productivity estimate is not proof that every individual, sector or region is worse off — it is an aggregate against an unobserved counterfactual, and plenty of people and firms are doing better than they were. Equally, individual Brexit successes do not overturn the aggregate economic evidence: pointing at a signed trade agreement does not answer a productivity estimate two orders of magnitude larger.
Why there is no total. This page does not add the columns up. Converting "decisions are now made in Westminster" into pounds so it can be netted against a productivity estimate would require an exchange rate between sovereignty and GDP that nobody has, and inventing one would be the least honest thing on the page.
The useful question is not "was Brexit good or bad" — that compresses a dozen separate outcomes into one word. It is "which of these rows do you weigh most heavily, and why". Two people can read this table, agree on every fact in it, and reach opposite conclusions. That is what a genuine political disagreement looks like, as distinct from a factual one.
One thing almost everyone has backwards
Registered Remain campaigners reported spending £19.3m. Registered Leave campaigners reported £13.3m. Remain outspent Leave by 45% on the Electoral Commission’s own record — and lost. The full funding breakdown.
Related: the media case study on how the campaign was covered, and how votes become seats — the referendum was decided by a simple majority of votes cast, a rule Westminster elections do not use. Also immigration for the current picture, and the economy for growth, inflation and pay.