How the ranks are calculated

Short answers

How are the ranks calculated?

Each metric is turned into a z-score across the 28 countries (the UK and the 27 EU member states), flipped where lower values score higher and capped at plus or minus 3. A pillar score is the mean of a country's available z-scores in that pillar, and the overall score is the mean of the two pillar scores. Countries are ranked by score, and ties share a rank.

Which countries are ranked?

The UK and the 27 EU member states, at national level. Norway, Switzerland, the United States, Canada, Australia and Japan appear on trend charts as reference lines and are not ranked.

Which year does each metric use?

The most recent year with values for at least 25 of the 28 countries. A country with no value that year uses its latest value from up to two years earlier, and that value is flagged.

What happens when a country has missing data?

A country needs at least 70% of a pillar's metrics to receive a pillar score and rank. Every pillar score shows a completeness badge, such as 9 of 11 metrics, so a score built from fewer metrics is visible as such.

Can I reproduce a rank myself?

Yes. The scores can be rebuilt from the z-scores file on the data page, and a worked example there follows one rank step by step in a spreadsheet.

What does the site not measure?

Sovereignty, regulatory autonomy, migration policy and other non-economic factors people weigh are outside the data. The site does not say whether the UK would be better or worse placed in the EU.

What is ranked

The 28 countries are the UK and the 27 EU member states. Each is ranked on 21 metrics. There are11 economy metrics and 10 quality-of-life metrics. Rank 1 is the highest score on the method, and every country is placed on the same metrics for the same years.

Norway, Switzerland, the United States, Canada, Australia and Japan appear on trend charts as grey reference lines. They are not ranked. They show what happened in comparable economies outside the EU, so UK-specific change can be read apart from wider trends.

The home page shows a subset of ten, called the Core 10: five economy and five quality-of-life metrics, chosen by principle (one per topic, no composites, low year-to-year volatility). The full set is on therankings page.

The ranked metrics

"Scoring direction" says how a value enters the score. It is a convention for the arithmetic, not a view on which outcome is preferable. Latest year is the most recent year in the snapshot for that metric; the latest common year, used for the score, is set by the rule in step 1 below.

Context metrics: shown, never ranked

4 measures are published on country pages to help explain the numbers. None is part of any score. Each has a reason for being kept out of the ranking.

Trend series (Since 2016)

The Since 2016 page uses five indexed series: real GDP, real GDP per head, real investment, real trade (exports plus imports of goods and services) and GDP per hour worked. Each is indexed to 2015 = 100, using non-projection values from 2000. The EU27 aggregate, the big four (Germany, France, Italy, Spain) and the control-group median are drawn beside the UK.

Average annual growth is the compound annual growth rate of the level between the start and end year, to two decimal places. Growth is shown for 2000 to 2007, 2010 to 2015 and 2016 to the latest year with data for both the UK and the EU27. A further column leaves out 2020 and 2021 and takes the geometric mean of the remaining annual growth rates. UK minus EU27 is in percentage points.

Trends show what happened, not why. Covid, energy prices and policy changes overlap the same years.

Scoring, step by step

1. Latest common year. For each metric, the year used is the most recent year with values for at least 25 of the 28 countries. Projections are never counted.
2. Carry-forward. A country with no value in that year uses its most recent value from up to two years earlier. That value is flagged and keeps its own year on the country page. An older value counts as missing.
3. Z-score. Each value is turned into a z-score across the countries that have one: subtract the mean, then divide by the population standard deviation. For a metric where lower values score higher, the sign is flipped. Each z-score is capped at +3 and −3. If every country has the same value, every z-score is 0.
4. Pillar score. The mean of a country's available z-scores in that pillar, with equal weights by default. A country needs at least 70% of the pillar's metrics to receive a pillar score and rank. Otherwise the pillar score and rank are left blank.
5. Completeness badge. Each pillar score is shown with a count such as "9 of 11 metrics", so a score built from fewer metrics is visibly so. This matters most for Bulgaria, Croatia, Cyprus, Malta and Romania.
6. Overall score. The mean of the two pillar scores. It is blank if either pillar score is blank.
7. Ranks and ties. Rank 1 is the highest score. Ties share a rank, so two countries on the same score both take rank 2 and the next place is 4. Only countries with a score take a place.
8. Rank history. For each year from 2000, each metric is scored at the earlier of that year and its own latest common year, using the same rules. The last point of every history therefore equals the headline rank.
9. Rank moves. A move of one place or less is shown in grey; a larger move in dark ink. Neither colour is used for good or bad.

Weights panel

The default view gives every pillar and every metric the same weight, labelled "Equal weights (default)". Therankings page has a weights panel. Each pillar and each metric can be set from 0 to 3 in steps of 0.5. The new weights are applied in the browser, and the page address holds them, so a view can be shared.

• Z-scores are used exactly as published. They are not recalculated when weights change.
• A pillar score is the weighted mean of its z-scores. Metrics set to 0 are left out, and the 70% rule applies to the weighted metrics.
• The overall score is the weighted mean of the pillar scores. Ranks are recalculated with the same tie rule.
• Custom weights show a "Custom weights" label. The panel is a reader tool and does not change the headline rank.

Sensitivity checks

Two alternative calculations are published beside the headline rank, so readers can see how far the result depends on the method. Neither changes the headline.

• Without multinational-sensitive measures. Real GDP growth, GDP per hour worked and trade openness are dropped for every country, and the same rules are applied to what remains.
• Rank-based scoring. Each z-score is replaced by a score from rank position alone: the top country on a metric scores +1.5, the bottom −1.5, and the rest are spread evenly between. The cap does not apply.

Ireland's headline rank is high because its GDP is distorted by multinational accounting. The alternatives show how much the rank depends on those measures. The full tables for all 28 countries are on therankings page.

Sources, precedence and missing data

• Each metric lists its sources in order of precedence. For each country and year, the first source in that order that has an actual value is used. Each value keeps the name of the source it came from, the series code, the vintage and the date it was fetched.
• Projections (IMF forecasts) come only from the primary source. They are shown dashed on charts, are never used to fill a gap, and are never ranked.
• A value that no source provides is shown as "no data". Nothing is imputed. Where a country has no value, the composite scores use the metrics it does have, as long as it meets the 70% rule.
• Every figure names its dataset, series code, vintage and fetch date, and the full list with licences is on thedata page.

Flags and exclusions

A large year-on-year move needs a reviewed reason before it reaches the site. There are two ways to handle one.

• Flags (48). The value is real and the note explains it: an event, such as Ireland's multinational revisions in 2015, or a method change. The value is kept. Of the 48 flags,12 are real events, 1 is a method break, and35 are plausible moves whose year-specific cause has not been found.
• Exclusions (2). The source value appears to be wrong and is dropped. The next source in precedence applies, or the value is treated as missing. Each exclusion is rechecked when the source updates. The two current exclusions are in the changelog.

The lists are in the project repository (reference/flags.toml and reference/exclusions.toml), Neither file changes a value on its own. A flag that matches no observation, or an exclusion that matches none, stops the monthly run.

Validation and the monthly refresh

Before any figure reaches the site, a run must pass these checks:

• The ranked country count is 28.
• Every observation names a known metric and country, and no year appears twice for the same source.
• Every metric has some year with values for at least 25 of the 28 countries.
• Every value lies within the plausible range set for its metric.
• Any year-on-year move above the metric's threshold carries a reviewed flag.

A monthly job, on the third of each month, runs the tests and fetches every source. A run that fails validation writes nothing. A passing run opens a pull request with the new data for review. Nothing is merged or published without that review. If a source fails, its previous file is kept and an issue is opened.

Known gaps

No reliable feed covers the UK and all 27 EU states on these topics, so they are not included, rather than patched with figures that are not comparable.

• Household energy prices. Eurostat no longer publishes UK data, and the IEA series is paid.
• Price levels and cost of living. Eurostat's price-level indices exclude the UK, and the World Bank comparison updates only every few years.
• EU-SILC poverty and deprivation, and Eurobarometer. The UK series end around 2018 and 2019.
• Healthcare waiting times. National definitions differ too much to rank.
• Polling on rejoining or remaining. It is available but left out, to keep the site neutral.

Caveats shown in context

• Ireland. GDP is distorted by multinational accounting. GNI per head is ranked and is shown beside GDP. The sensitivity ranks above apply.
• Covid years, 2020 and 2021. Values are volatile and often revised. Trend averages are shown with and without them.
• Revisions. Sources revise past years. Each snapshot keeps its vintage, and large revisions are recorded in the changelog.
• Projections. IMF forecasts are shown dashed and never ranked.
• Small states. Luxembourg, Malta and Cyprus can sit at the extremes on some measures. They are flagged, not removed.

What this site can't tell you

Changelog

Material method decisions, exclusions, large revisions and corrections are listed here, newest first. Corrections also appear on the about page.