Brexit's Wasted Window

Take the timeline seriously. The referendum was June 2016. Formal departure was January 2020. TCA goods rules took effect January 2021. The first full year of Brexit operating without a major concurrent shock was roughly the first two months of 2021, before supply chain disruption arrived. Then came Ukraine. Then the energy crisis. Then the Truss mini-budget gilt crisis in September 2022. Any model measuring "the cost of Brexit" over this window is measuring Brexit plus Covid plus an energy shock plus a domestic political crisis and calling the total Brexit. John Springford at the Centre for European Reform ran the most prominent such model from June 2018 until early 2023, when he stopped. "In the end," he said, "the energy-price shock killed the model." The NBER paper extends the same approach over a longer, more contaminated window. The OBR warned in 2018 that these comparisons "will probably become less reliable over time." They were right.

NBER Working Paper 34459, published November 2025 by five economists (four of them Bank of England staff), is the paper most commonly cited for the 8% figure. It applies five different country-weighting schemes to a panel of 33 advanced economies and produces a range of minus 6 to minus 8% for GDP per capita on the macro side; the micro component, drawing on the Decision Maker Panel firm survey, produces minus 6%. The "8%" in circulation is the macro method's upper bound, produced by the synthetic control version that weights the United States most heavily as part of the UK's comparator. The US since 2016 has benefited from the CHIPS Act, the Inflation Reduction Act and cheap domestic energy, none of which have anything to do with EU membership. This makes the upper bound sensitive to US-specific performance that is unrelated to the variable being tested. The paper is a pre-publication working paper. It has not been through journal peer review.

The credible centre runs from minus 4% to minus 6%, with recent academic work extending toward 8%. The OBR's official benchmark is minus 4% long-run productivity loss, unchanged since 2016, and the government's own working figure. The NBER's 8% is the upper end of that range, from a paper not yet through peer review but within the range of serious estimates, and it requires accepting that the synthetic control correctly attributed all post-2016 UK underperformance to Brexit and none to the sequence of non-Brexit shocks that filled the measurement window. HM Treasury's 2016 short-run forecast predicted four consecutive quarters of contraction and 500,000 job losses; employment rose and there was no recession through 2016 to 2019, and the Treasury dropped the gravity model methodology entirely. The long-run forecasts (minus 6.2 to 7.5% GDP) now roughly overlap with the NBER range but were wrong on timing by a decade; the expected two-year adjustment became a nine-year slow burn. At 8% of current GDP the implied cost is roughly £200 billion per year. The absence of a 2017–2019 recession and the UK's second-best G7 productivity growth suggest the observable damage is real but the upper bound is not where the weight of evidence sits.

These methods are useful but highly sensitive to assumptions over this particular period. Doppelgänger models work by attributing any post-2016 divergence between the UK and its synthetic comparator to Brexit. They have no mechanism for separating Brexit from Covid, from the 2022 energy crisis, from the Truss gilt crisis or from the specific domestic policy choices made after departure. Germany, which features heavily in most comparator pools, has been through its own crisis since 2022: energy exposure, automotive EV transition, manufacturing contraction. When Germany underperforms for its own reasons, the UK looks relatively better in ways that have nothing to do with Brexit. Country-specific shocks contaminate the whole panel, and this is why there is such a wide spread of results.

Set aside the comparator debate. Some things require no synthetic control. UK goods exports to the EU were 18% below 2019 levels in real terms in 2024, consistent with the frictions from departure from the customs union: rules-of-origin compliance, regulatory divergence, border checks that did not exist before. That fall is clearly in the ONS data and is not contested across political lines.

Services exports to the EU ran 19% above 2019 levels over the same period; services to non-EU countries were 23% above. The OBR's 2016 assumption (that long-run trade volumes, goods and services combined, would be around 15% lower) applies to both, and on current data looks overstated. It has not been revised. Financial services exports fell around 6% and the EU's share of UK financial exports dropped from 37% to 29%; the passporting loss is visible there. But financial services are one sector. The broader services recovery was not in the models, and it matters for the aggregate cost estimate.

Business investment is the most signal-rich data point. UK investment has grown roughly 4% since 2016; across comparable advanced economies the figure is roughly 25%. At around 18% of GDP, the UK is among the lowest-investing economies in the G7. The Decision Maker Panel firm survey (part of the same NBER paper that produces the 8% GDP figure) found that firms with heavy EU trade exposure cut investment by 16% specifically. That figure comes from UK firms reporting on their own decisions, not from a comparator country.

Investment is also forward-looking. Lower expected returns → lower investment → lower capital per worker → lower productivity → lower GDP per capita. The chain started running from the moment the referendum result was declared in 2016, before legal departure, before Covid, before Ukraine. A company deciding in 2017 whether to build capacity in the UK or on the continent was deciding based on expected regulatory environment and market access over the following decade. The investment data shows the UK losing ground on that forward expectation, and it shows it starting from 2016, not from 2021.

There is one figure that sits awkwardly in the standard Brexit cost narrative: UK productivity growth (output per hour worked) from 2016 to 2024 was near the top of the G7 by ONS measures, behind only the United States and ahead of France, Germany, Italy and Japan. This figure is from ONS International Comparisons of Productivity and appears almost nowhere in the mainstream coverage. It does not mean Brexit was harmless. What it means is that the Brexit damage shows most clearly in GDP per capita, not in fundamental productive efficiency: the population grew faster than capital stock, pushing per-capita output down even as output per hour held up reasonably well. Both problems have different causes and different solutions. Falling GDP per capita from rapid low-wage immigration is not the same problem as falling productivity from trade friction, and conflating them produces both a wrong diagnosis and wrong policy responses. It also helps explain why UK GDP per capita underperformed its own history while productivity did not: more workers, less capital per worker, same or better output per hour. UK productivity stagnation started in 2009, seven years before the referendum. That context also does not appear in the debate.

Start with what Brexit promised in economic terms. Departure from the EU created policy freedoms that were not available inside it: a more competitive tax environment, planning reform, domestic energy expansion, regulatory divergence favouring UK growth sectors. Whether or not you accept the full case, what the UK actually did with those freedoms is a matter of domestic record. No comparator country required.

Corporation tax was 19% when the UK left the EU. The Conservative government that delivered Brexit raised it to 25% in April 2023, the largest single rate increase in decades, above the G7 average and more than double Ireland's 12.5%. A 6-point rise in corporation tax materially reduces post-tax returns on marginal investment, compounding the Brexit uncertainty effect with a specific, additional disincentive. Every economic argument made for post-Brexit competitive advantage assumed at minimum that the tax environment would stay attractive. The government chose the opposite.

Planning reform was promised in the 2019 Conservative manifesto and not delivered. Net additional dwellings in England ran at roughly 220,000–235,000 a year, essentially unchanged from 2016, and the 300,000-per-year target was not met in any subsequent year. The planning system appears in every government growth strategy as the primary constraint on development and is then consistently left unchanged, which is a reasonable description of the approach taken to most of the supply-side agenda: identify the problem, acknowledge it, leave it.

Energy policy ran in the wrong direction. The Energy Profits Levy, a windfall tax on North Sea producers introduced May 2022, raised to 35% in January 2023 and repeatedly extended, reduced the incentive to invest in domestic production at precisely the moment European energy vulnerability had become obvious. Outside the EU, the UK had more flexibility on North Sea licensing and grid investment than it had inside it, none of which translated into competitive energy prices; UK industrial electricity costs remained among the highest in Europe throughout the period.

Immigration policy is the most paradoxical case. Brexit reduced EU-origin workers in the UK by roughly 785,000 by 2024; non-EU workers increased by roughly 992,000 over the same period. Britain ended up with more foreign-born workers post-Brexit than before it, just a different composition, skewed towards non-EU workers in lower-wage sectors. The points-based system opened Health & Care visa routes at scale, bringing in workers where many of the largest inflow categories had median earnings well below the OBR's fiscal break-even threshold of £41,700. Net migration peaked at around 944,000 in the year to March 2023, four times the 2010s average. The OBR models the mechanism directly: additional net migrants add to total GDP but put downward pressure on GDP per capita when capital does not adjust, which it has not. The UK added workers faster than it added capital. That's most of the story. Business investment grew 4% while population grew roughly 5%. The doppelgänger models attribute the resulting per-capita drag to Brexit. Some of it is Brexit-related, via the capital suppression that started in 2016; some of it is the specific immigration policy the post-Brexit government chose to run on top of it. Ending free movement was a Brexit outcome; opening large-scale low-wage non-EU routes to replace it was not required by departure itself. The alternative was available. The UK has world-ranked universities, an English-language advantage, established financial and professional services, and a legal system that foreign firms trust — a genuine basis for competing aggressively for high-value global talent rather than filling volume gaps in social care. A points-based system calibrated to capital formation rather than headcount would have looked quite different: weighted toward skills that attract investment rather than skills that substitute for it, with salary thresholds set to filter toward net fiscal contributors rather than away from them. Singapore, the UAE and Canada at various points have run versions of this. Britain had the architecture, after 2020, to do the same. It did not use it that way.

Two G7 anomalies complicate any clean comparator story. Italy, a founding EU member in the single market throughout, ranked second in the G7 on GDP per capita growth over this period — ahead of Germany, France and Japan — largely on the back of the Superbonus construction scheme, which had nothing to do with EU membership. Germany at +4.4% barely outperformed the UK at +4.5% while in the middle of its own domestic crisis on energy and manufacturing. Trade friction and capital dilution are separate channels; both are visible in the data, but neither translates cleanly into a G7 ranking. That's the gap between what the models measure and what the comparison shows.

The failure to deploy the post-Brexit policy toolkit was not primarily technical. Planning reform is difficult but other countries have done it; competitive corporate taxation is a political choice rather than a constraint; domestic energy investment is a function of incentive design. None of these were beyond reach, which makes the absence of any serious attempt at them more interesting than if they had simply been too hard.

One diagnosis comes from inside the 2019 government. Dominic Cummings, writing in April 2026, described what he observed in Parliament: "MPs very clearly had lost interest in the economy and productivity. Blank stares. No interest." That observation is consistent with the legislative record from 2020 onwards. The substantive productivity-focused programme stalled. What followed was three Conservative prime ministers in twelve months, a budget that caused a gilt crisis, and years of parliamentary energy going into debates with no connection to investment, capital formation or growth. Other economies with political disruption of their own deployed aggressive investment policy in response to shocks: the United States with the CHIPS Act and Inflation Reduction Act, Italy with the Superbonus construction scheme. The UK did not.

Not only a Conservative failure. The Labour-adjacent think tank sector showed the same pattern: the centre-left commentariat that might have pushed a serious post-Brexit growth programme spent the same years publishing on culture and identity rather than planning reform, tax competition or energy supply. The political class, across parties, appears to have concluded around 2020 that economic productivity was less interesting than cultural identity, and the investment gap accumulated while that conclusion was being acted on.

The repercussions continue. Labour took office in 2024 and raised employer national insurance contributions, reducing the returns to hiring. Planning reform has begun but remains slow. Corporation tax sits at 25%. The North Sea windfall tax persists. The immigration rate has fallen from its peak but the capital adjustment it required still has not happened. The decisions being made now are downstream of the same political culture that produced the post-Brexit decade, one in which productivity and economic strategy are second-order questions, not primary ones. That is a problem regardless of what you think about the referendum.

Brexit did not cause this. What it did was create a moment when the political class would have had both the justification and the public mandate for serious supply-side reform, and a well-functioning political culture would have taken it. Closing even part of the investment gap would likely have meant several percentage points of GDP — through the capital-per-worker channel alone — that simply never showed up. Instead the gap widened — the result of disruption sitting on top of a decade of policy running in the wrong direction, and the window narrows with each year that passes.