If you compare turnout for the same election across two reputable sources, you will sometimes find they disagree — not by a rounding error, but by whole percentage points. Neither is necessarily wrong. Turnout is a fraction, and countries, commissions and publishers disagree about what belongs in the numerator and what belongs in the denominator.

This matters more than it sounds. Turnout is the number people reach for when arguing about legitimacy, and in some systems it is legally decisive. It deserves to be understood as a constructed measure rather than a raw observation.

The numerator: who counts as having voted

Three conventions are in common use:

  • Ballots issued — every voter who was handed a ballot paper, whether or not they returned it.
  • Ballot papers in the box — the usual choice, and the one Poland's electoral commission uses: turnout is valid ballot cards drawn from the boxes, divided by eligible voters. Note that a blank or spoiled ballot still counts as participation here, which is correct — the person turned up and expressed something, including possibly a deliberate protest.
  • Valid votes only — blank and invalid ballots excluded. This produces a lower figure and, in my view, misdescribes what turnout is meant to measure.

The gap between the second and third is not trivial. São Tomé and Príncipe's 2026 presidential election recorded 58,906 ballots cast of which 2,332 — 3.96% — were blank or invalid. Choosing the wrong numerator moves the published turnout by nearly four points.

Vote shares have the same problem, one level down. Poland's commission calculates each candidate's percentage against valid votes. Russia's Central Election Commission reports shares against all votes cast, invalid ones included. Copy a Russian percentage into a table whose other rows use valid-vote shares and you have quietly introduced an error of a percentage point or so into every row.

The denominator: who counts as eligible

This is where the differences get large:

  • Registered voters — the electoral roll. The most common basis, and the one this site uses, because it is what commissions publish.
  • Voting-age population — everyone old enough, registered or not. Used by comparative political scientists, and always lower than the registered-voter figure in countries with opt-in registration.
  • Resident citizens — used where a large diaspora is on the roll but rarely votes. Finland reports turnout on this basis, which is why its figures look higher than a naive roll-based calculation would give.

Diaspora rolls are the single biggest source of confusion. São Tomé's roll of 142,182 includes emigrants of whom roughly four in five did not vote; that drags the headline turnout down to 41.43% and makes the country look far more apathetic than the domestic figure alone would suggest.

The cleanest illustration I know: Norway, one day, two elections

On 11 September 2023 Norwegian voters went to the polls and cast two ballots — one for their municipal council, one for their county council. Same day, same polling stations, largely the same people. The published turnout figures:

Municipal election
62.20%
Kommunestyrevalg 2023
County election — same day
57.30%
Fylkestingsvalg 2023

Nearly five points apart, and the 2015 pair (59.90% against 54.80%) shows the same gap. Part of it is genuine — county government is less visible, and some voters take the municipal ballot and skip the other. But part of it is pure arithmetic: Oslo does not hold a county election at all, because it is simultaneously a municipality and a county. Oslo's voters are in the municipal denominator and absent from the county one. Two correct numbers, two different populations.

What this means when you compare countries

Cross-country turnout league tables are the most-shared and least-reliable election statistic on the internet. Before drawing a conclusion from one, three questions are worth asking:

  • Is voting compulsory? Compulsory-voting countries sit 15 to 30 points above voluntary ones, and that says nothing about civic enthusiasm.
  • Is registration automatic or opt-in? Automatic registration puts every adult in the denominator, including people who would never vote. Opt-in registration filters the denominator down to the already-motivated and flatters the percentage.
  • Which election is it? Comparing a presidential turnout with a municipal one is meaningless. Norway's 2025 parliamentary turnout was 80.10%; its municipal elections run twenty points lower. Same country, same electorate, different stakes.

Iceland's 82.52% in the 2026 EU referendum is genuinely high on any denominator — but the honest way to say that is "high for a referendum, in a country with automatic registration and no compulsory voting", not "Icelanders are twice as engaged as Americans".

What we do

Every turnout figure on this site is the one the electoral authority published, on the basis that authority uses, entered to two decimal places exactly as given. We do not recompute it to a common standard, because doing so would mean publishing a number no official source would recognise. Where a country's basis is unusual — Finland's resident-citizen denominator, Norway's Oslo gap — the election page says so.

The trade-off is that our figures are exactly comparable within a country over time, and only roughly comparable between countries. That is the honest shape of the data, and pretending otherwise would be the actual error.