The economic impact of Brexit — Choose Your Own Doppelganger

About this site

How has Brexit affected UK GDP per capita since 2016? One way to estimate this is by constructing a 'doppelganger' based on how much other countries have grown since 2016, and compare this to the UK. The method works by taking a weighted average of the growth of other countries since 2016 and comparing that to the UK. It's been used most recently in an NBER working paper by Nick Bloom and coauthors that estimates that Brexit reduced UK GDP per capita by 6-8%.

This website lets you see how different choices affect estimates of the impacts of Brexit. The key choices are which countries to include in the comparison and how much weight to put on each country. Below, you can choose which countries to include from the set of 33 countries used in the Bloom et al. paper, and what weight to put on all the selected countries. Weights need to add up to 100% across the included countries. You can choose one of the weighting methods that Bloom et al. use, or enter your own custom weights.

Weighting scheme

Comparator countries

Untick to drop a country from the comparison — the other countries' weights are reshuffled to make up the difference. In Custom mode, you can type your own weights into the right-hand column.

Use Country 2015 GDP
$bn		 Distance
km from London		 UK trade
£bn		 Weight
%

Estimated UK GDP-per-capita gap

UK (2016Q1=100)
—
Counterfactual, same period
—
Gap (UK vs counterfactual)
—
Countries used / weights sum to
—
Notes and technical detail

The basic idea

We can't see what the UK economy would have done in a world without Brexit. The next best thing is to compare the UK with other rich, similar countries that didn't leave the EU. If they grew faster than the UK after 2016, that's evidence Brexit slowed UK growth.

The catch: which other countries should we compare with, and how much weight should each get? Different choices give different answers. The five named schemes — unweighted, GDP-weighted, gravity, UK trade-weighted, synthetic control — are the five different choices the paper makes; the headline finding is robust to all of them.

The synthetic control schemes

The synthetic control method picks a small handful of countries whose economies were on a very similar trajectory to the UK before the Brexit vote, weighted so that the blend tracked the UK very closely up to 2016. After 2016, the same blend stands in for the “UK that didn't Brexit”. Any gap between the UK and the blend after 2016 is the estimated Brexit effect. The matching is done both on the UK's pre-2016 GDP-per-capita path and on five country-level characteristics (GDP per head, trade openness, investment ratio, share of value-added in industry, educational attainment). Implemented as Abadie–Gardeazabal nested optimisation: Frank–Wolfe inner solver on the country simplex, multi-start Nelder–Mead outer over the predictor weights.

Three variants are available:

  • Doppelgänger UK — the canonical version. The predictor weights are chosen to minimise pre-Brexit outcome MSPE over the whole 2006Q1–2016Q1 pre-period.
  • Cross-checked — predictor weights are chosen to do well across several chunks of pre-Brexit time using rolling-origin cross-validation (three folds, each with 4-quarter validation windows at growing training-window endpoints). Follows the train/validation approach proposed by Abadie, Diamond & Hainmueller (2015, AJPS) and used in Stata's synth command. Reduces the risk that predictor weights are cherry-picked to fit pre-period noise.
  • With pre-Brexit test window — the predictor weights are chosen using only the years before your chosen cutoff; the later pre-Brexit years are visualised as a held-out test window on the chart. The cutoff slider lets you trade off how much training data the optimiser sees against how big a test region you can inspect.

Country groups

G7: the seven major advanced economies (Canada, France, Germany, Italy, Japan, USA, UK). The UK is excluded from the comparator set because it's the treated country.

EU‑15: the EU members before the 2004 enlargement (Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden). The UK was the 15th but is excluded as the treated unit.

EU‑27: the current EU after the UK's exit (27 countries).

Eurozone: the 20 EU countries that use the euro (as of 2025; Croatia joined 2023).

Data sources

  • UK GDP-per-capita uses ONS series ABMI (chained-volume measure, seasonally adjusted) divided by population spliced from Eurostat (2004–2019) and World Bank (2020–).
  • Other European countries from Eurostat namq_10_gdp (chain-linked volume index, 2015 = 100, SA); US/Canada/Japan from OECD Quarterly National Accounts (chain-linked volume).
  • Ireland: per the paper, the app uses CSO Ireland's Modified Final Domestic Demand (NAQ05, code 0155, constant prices, SA) in place of headline GDP, to remove the volatility caused by multinational corporations' intangible-asset and aircraft-leasing flows.
  • Cyprus and Malta are excluded by default following the paper.

Comparison with the paper

The paper's headline finding is that UK GDP was 6–10% lower than comparator countries by year-to-2025Q1 (mean across schemes: 8%). The headline window above adjusts automatically; with current data it covers the most recent four quarters (year to 2025Q4).

Data-vintage caveat: the paper's UK index at year-to-2025Q1 (104.3, 2016Q1 = 100) is ~1 percentage point lower than the current ONS series (~105.3), because ONS Blue Book 2025 (released September 2025) revised UK GDP up by ~0.5 pp per year over 2021–2024 — after the paper's data was assembled. This app uses post-revision data, so each scheme's estimated gap appears ~1 pp narrower than the paper's stated figures. The comparator side matches the paper to within ~0.3 pp per scheme once that UK shift is removed.

Underlying paper: Bloom, Bunn, Mizen, Smietanka & Thwaites, The Economic Impact of Brexit, NBER Working Paper 34459 (November 2025).

Robustness check: placebo-in-space test

What if Brexit had nothing to do with it — what if the UK was just unlucky? To check, we can pretend that each other country was the one Brexit-treated, and build a synthetic control for it in exactly the same way. If the UK's post-2016 drop is unusually large compared to the cloud of placebos, that's evidence Brexit is doing distinctive work. If the UK sits in the middle of the cloud, the gap is hard to attribute to Brexit specifically. (Standard permutation-style inference, after Abadie, Diamond & Hainmueller 2010.)