The Four Million Who Never Vote: Who They Are and What Would Change if a Third of Them Did

Setting aside those who cannot vote because of illness, work or travel, about four million voters living in Spain abstain election after election: between 10% and 12% of the electoral roll. They are younger, less educated, more of them are unemployed, and most have no interest in politics. They live in specific neighborhoods, the same ones since 2008. If a third of them voted, turnout would rise by about four points. If they voted like their neighbors, the seat allocation would barely change. If they voted as they say in the CIS (Spain's state polling institute), Vox would win two or three seats, mostly at the PP's expense. In no case does the PP and Vox majority in the forecast change.

26 October 2026 · Miguel · Leer en español

Not everyone who doesn’t vote is the same

In the 2023 general election, 29.7% of voters residing in Spain did not vote. Not all of them are abstainers. Some could not vote: they were ill, working, traveling or did not receive their mail-in ballot in time. Others appear on a roll that lags behind reality, with people who have moved or died. And others abstain sometimes and vote at other times. This article is about those who remain: people who do not vote election after election and not because they cannot. We call them systemic abstainers.

No official statistic follows a person from one election to the next, so we looked for them in two ways. One is the polling-station results of the seven general elections from 2008 to 2023, summed over the 32,027 census sections that kept the same code in all of them, with 32 million voters. The other is the ten CIS post-election surveys of general elections, from 1996 to 2023, with 61,427 interviews. In them, each respondent says whether they voted, whether they voted in the previous ones and, if not, why.

The same streets, election after election

Abstention rises and falls from one election to the next: from 24.4% in April 2019 to 30.3% in November, seven months later. But what rises and falls is the level; the distribution across neighborhoods changes little. The sections with the most abstention in 2008 are, for the most part, those with the most abstention in 2023: the rank correlation between the two elections is 0.77. Of the fifth of sections with the highest abstention in 2008, 63% are still in that fifth fifteen years later. Between two consecutive elections, the figure is between 67% and 83%. If we decompose the variation in abstention across the 32,027 sections, 78% is specific to each section and persists from one election to the next. Only 8% is due to some elections mobilizing more than others.

How many there are

Each section has a floor: the lowest abstention it has recorded in the seven elections, almost always in 2008 or April 2019. Someone who never votes is at that floor, because they abstain even in the election with the highest turnout. The average section has a floor of 22.3%. But not even the 1% of sections with the highest turnout falls below 10.2%: that is, roughly, technical abstention, that of people who cannot vote or no longer live there. Subtracting it leaves 12.1% of the roll as systemic, about 4.3 million of the residents in 2026. This is a ceiling, because the floor includes people who abstained in the best election by chance.

The CIS gives a lower bound. In 2023, 31% of those who did not vote say they could not, or cite health, work, family or problems with the mail-in vote. Of the rest, half (51%) also did not vote in November 2019. Applied to actual abstention, this gives about 10.4% of the roll, some 3.7 million. Surveys capture less abstention than there is: in the 2023 post-election survey, only 8.5% say they did not vote, against 29.7% in reality. And people who do not vote tend to answer surveys less. So that figure tends to fall short.

The CIS also shows a trend. In 2008, 78% of those who did not vote had also not voted in 2004; in 2023, 51%. Abstention is now more intermittent: more people move in and out. That partly explains why turnout jumps so much between consecutive elections.

Who they are

Compared with those who always vote, systemic abstainers:

• are younger: an average age of 47, against 53;
• have less education: 22% have a college degree, against 41%;
• include more unemployed people: 19%, against 9%, and fewer retirees (20% against 30%);
• include more men: 56%, against 48%.

What most sets them apart, however, is their relationship with politics. Seventy-two percent say they have no interest in it at all, against 21% of voters, and 39% do not place themselves on the left-right scale, against 6%. Those who do place themselves are in the center, at an average of 4.8 out of 10, almost the same as voters (4.7). In 2023, the CIS asked those who did not vote why; more than one reason could be given. Among the 765 who were not prevented from voting, half (51%) said they did not trust any party or politician or did not feel represented. Twenty-two percent spoke of discontent, protest or disappointment. Eighteen percent said they are not interested in politics and another 18%, that they never vote. Nine percent believe their vote is useless.

Where they live

The map of systemic abstention is the map of inequality. In the 10% of sections with the highest floor (35% on average, against 12.5% for the lowest 10%), the 2021 census unemployment rate is 29%, against 12%. Income per person is €11,800, against €18,100. Only 20% of residents have a college degree, against 38%, and 19% were born abroad, against 9%. Of all the traits, the one most associated with a section’s floor is unemployment: the correlation is 0.59. Next come income (−0.39) and education (−0.37).

By province, underlying abstention, with technical abstention already subtracted, is highest in Ceuta (23%), Melilla (23%), Las Palmas (21%), the Balearic Islands (21%), Santa Cruz de Tenerife (19%), Cádiz (19%), Huelva (18%) and Málaga (16%). It is lowest in Cuenca, Ávila, Segovia and Valladolid, at around 7%, and in Toledo and Madrid, at around 8%. Part of the difference between provinces comes from the roll and not from the voters: on the islands and in tourist areas there are more residents who move and remain registered where they no longer live.

What if a third of them voted?

We start from the October 26 forecast and add a third of each province’s systemic abstainers: 1.43 million more voters, which raises turnout by about four points. We then repeat the seat allocation in each of the 50,000 simulations. What remains is to know whom they would vote for, and for that we make two assumptions:

• Like their neighbors. They vote like their province in each simulation, but tilted toward how the sections where systemic abstainers are most numerous voted in 2023.
• As they say. They vote as they say in the 2023 CIS post-election survey, the only recent one that asks voluntary abstainers whom they would have voted for. Seventy-seven systemic abstainers answered the question with a specific party: the PSOE and Vox, with a quarter each, ahead of the PP (19%) and Sumar (8%). This is a very small sample. That is why, in each simulation, the split is drawn with its sampling error. We repeated the calculation with the 228 voluntary abstainers of 2023, whether or not they had voted in 2019. In that group the split is more similar to that of voters: PSOE 23%, PP 21% and Vox 17%.

The result depends mostly on whom they would vote for. If they vote like their neighbors, almost nothing changes: on average, half a seat across the whole Congress. In each province, the new votes split almost like the ones already there. If they vote as they say, Vox gains 2.5 seats on average and the PP loses 2.1. The PSOE gains 0.2 and Frente Amplio loses 0.2. With the broader group of volunteers, the effect is cut by more than half: Vox gains 0.8, the PSOE 0.4 and the PP loses 0.9. The provinces where a seat most often changes hands are the large ones: Barcelona, Madrid, A Coruña, Málaga, Pontevedra, Cádiz and Sevilla.

Under none of the assumptions does the balance between blocs change. The probability that the PP and Vox together win an absolute majority, which is 94% in the forecast, stays between 93.5% and 95%, and the probability that the PP alone has one remains zero. What does change is the split within the right: with systemic abstainers voting as they say, Vox gains at least one seat in 84% of simulations and the PP loses at least one in 78%.

The numbers help explain why. A third of systemic abstainers is 1.43 million votes, 5.4% of the forecast’s voters. To move seats between blocs, those votes would have to split very differently from how their provinces already vote, and they do not. What they do have is a younger profile, more distant from politics and with a greater declared inclination toward Vox. That inclination weighs precisely where the right divides the last seats.

What we don’t know

• We don’t know whom they would vote for. What 77 respondents say about what they would have done is not a vote. Abstainers who do decide to vote in a specific election usually do so pushed by something in that campaign, and they need not resemble those who stay home.
• A third is an assumption, not a prediction. The turnout the forecast expects already includes some intermittent abstainers. We have found no sign that systemic abstainers will mobilize this time.
• Sections are not people. A section’s floor says that many people there do not vote, not who they are. The individual profile comes from the CIS, with its limits: it captures less abstention than there is, and recall of whether one voted in previous general elections is imperfect.

How we calculated it

• Sections. Interior Ministry polling-station results (residents, excluding the overseas vote), summed over the 32,027 census sections that keep the same code in the seven general elections from 2008 to 2023 and have at least 100 voters in all of them. Floor: each section’s minimum abstention across the seven. Technical abstention: the 1st percentile of that floor (10.2%). Systemic: the floor minus the technical level, weighted by roll size and scaled to the 2026 resident roll. Profile: Household Income Distribution Atlas and 2021 Census (INE, Spain’s statistics office).
• CIS. Post-election surveys of the general elections of 1996, 2000, 2004, 2008, 2011, 2015, 2016, April and November 2019 and 2023 (studies 2210, 2384, 2559, 2757, 2920, 3126, 3145, 3248, 3269 and 3420), with their weighting. Systemic: did not vote in that election or the previous one and cite no impediment. Profiles compare November 2019 and 2023 together. Interest in politics was only asked in 2019, with 319 systemic abstainers.
• Scenario. The 50,000 simulations of the October 26 forecast. Valid votes in each province: simulated turnout × 2026 resident roll × (1 − 2023 invalid votes), plus the overseas vote. A third of the province’s systemic abstainers (minus invalid votes) is added with the split from each assumption and D’Hondt is repeated with the 3% threshold. In the “as they say” assumption, regional parties receive from the new voters the same share they have in their province. The vote for IU, Podemos or Sumar is split between Frente Amplio and Podemos as in the simulation, and those who name no party are not counted.
• Code. cis_poselectorales.py and analisis_abstencion.py, in the altnalysis repository.