Foto: Providus

The record-low overall turnout masks very different trends in the Riga suburbs, in shrinking regions and in the diaspora.

With turnout of 51.79%, the 15th Saeima election ranks among the Saeima elections with the fewest voters. 806,641 people took part, 109,727 fewer than in 2022. The result is also below the 54.56% recorded in 2018. Yet this overall figure does not point to a single large group of “missing” voters or to a nationwide wave of voter apathy. Behind it lie at least three distinct processes, each requiring a different response: very low diaspora turnout, the fading of the unusually high mobilization of 2022, and a deeper fall in turnout in some aging and shrinking areas. The overall decline also conceals significant territorial and demographic differences. The important question is therefore not only “where did the voters go?” but also in which groups of society and which areas turnout changed the most.

Full version of the article in pdf here,

Summary in the presentation format here.

Diaspora: more voters, fewer ballots

The largest long-term changes are taking place abroad. Since 2018 the number of eligible voters registered abroad has grown from around 135,000 to 199,452 (2026), while the number who voted has fallen from 31,946 to 23,196. Turnout abroad has dropped from 23.7% to 11.6%: this year roughly one in nine voted.

This noticeably lowers the overall figure. Excluding voters abroad, turnout in Latvia was 57.5% in 2018, 64.5% in 2022 and 57.7% in 2026. In eight years the gap between turnout in Latvia and the official overall figure has grown from about 3 to about 6 percentage points. This means that the record-low overall turnout partly reflects a change in the structure of the electorate itself. As the number of eligible voters registered abroad grows, and their turnout is far lower than in Latvia, the overall figure can fall even if the activity of voters living in Latvia does not change significantly.

If the number registered abroad keeps growing and voting arrangements stay the same, this will continue to push overall turnout down even if turnout among voters in Latvia stays the same.

2022 was a mobilization peak, but 2026 is not simply a return to the earlier level everywhere

In Riga, turnout in this election was 63.8%. That is lower than in 2022 (67.5%) but almost the same as in 2018 (63.2%). The overall figure for the Riga electoral district (46.4%) looks considerably lower because the district also includes voters registered abroad. This needs to be borne in mind when comparing districts.

The largest declines are outside Riga and its suburbs. In Latgale around 24,600 fewer people voted than in 2022, and the number of voters also fell significantly in many rural municipalities. In Rēzekne municipality the number of voters fell by 27%, and in Ludza municipality by 26%. It is important to stress that turnout in Latgale in this election (47.2%) was higher than in 2018 (43.4%), so what mainly faded there was the unusually high mobilization of 2022.

This calls for caution in using 2022 as the only point of reference. Comparing only 2022 and 2026 shows a very large fall. But overall turnout among voters living in Latvia in 2026 has in fact returned to its 2018 level. So the question is not only why voters stayed at home in 2026, but also what produced the unusually high mobilization of 2022 that was not repeated this time.

Even so, a “return to normal” does not explain everything. In rural municipalities the number of voters in 2026 is around 9% lower than in 2018 as well. This points to a possibly deeper and longer-lasting decline in participation in parts of Latvia.

The fall in turnout is closely linked to the demographic structure of areas

A comparison of municipalities shows a pronounced link between an area’s demographic structure and the change in the number of voters. The fall in turnout is concentrated in municipalities with older populations. This link holds even when the analysis also takes into account the ethnic composition of the municipality and the electoral district. This does not mean that one can conclude from these data that “older people did not vote in this election.” The data describe areas, not the behavior of individual voters. But they do point to a close link between the fall in turnout and the demographic structure of areas.

By contrast, the number of voters grew in several growing municipalities around Riga: by 8% in Mārupe municipality and by 6% in Ropaži municipality.

If this trend continues over a longer period, the demographic differences between growing and shrinking areas will also shift their electoral weight – an ever larger share of the votes actually cast will be concentrated in the growing areas.

Ethnic composition does not explain where voters stayed at home

One of the explanations put forward for the low turnout is the demobilization of Russian-speaking voters. Municipal-level data do not suggest that the fall in turnout was larger in areas with a higher share of ethnic Russians. The link between the share of ethnic Russians in a municipality and the change in the number of voters in 2022–2026 is very weak. Even when demographic structure and region are taken into account at the same time, the share of ethnic Russians is not statistically significantly linked to the fall in turnout.

This, of course, does not allow us to conclude how any individual Russian-speaking voter behaved in the election – municipal-level data are not individual-level data. But they do allow us to conclude that the overall geography of demobilization in 2026 cannot be explained simply by where Russian speakers are concentrated.

Demography is more closely linked to turnout, ethno-demography to the geography of party support

Interestingly, ethnic structure, which does a poor job of explaining the fall in turnout, is much more closely linked to the geography of party support. The territorial patterns of support for the United List and Latvia First differ particularly sharply: in municipalities where one party’s support is higher, the other’s is usually considerably lower.

This does not mean that municipal data allow us to infer directly how particular ethnic groups voted. But it does show an important difference: the ethno-demographic composition of areas is much more closely linked to the geography of political choice than to the fall in turnout itself.

Behind the overall turnout figure of 51.79% there is therefore also a very uneven geography of party support. Establishing to what extent it was produced by differing mobilization of electorates or by voters switching between parties will require additional data and further analysis.

What we don’t know yet: did ID documents prevent some voters from taking part?

Another possible factor is personal identification documents. Before the 2022 election a transitional provision of the Law on Personal Identification Documents was in force: documents that had expired since March 1, 2020 were deemed valid for in-person identification in Latvia until April 30, 2023. In August 2022 the Office of Citizenship and Migration Affairs (OCMA) reported 26,661 people aged 15 and over who had no valid replacement document. There was no similar exemption this year.

How many eligible citizens were actually unable to vote on election day this year because of their documents is not publicly known. This factor can therefore be neither confirmed nor ruled out, and its effect should not be assumed without OCMA data. Providus calls on the OCMA to publish these data and has prepared a freedom of information request to the OCMA for statistics on document validity on election day in 2018, 2022 and 2026.

Early voting does not explain the overall decline. Its share rose from 4.1% (2022) to 10.9% of eligible voters (2026), yet overall turnout still fell. Voters used a different way of voting, but in total fewer people took part.

What to do

In the EIU Democracy Index, Latvia is a “flawed democracy” – not a democracy whose quality can be taken for granted. In 2024 Latvia’s score rose to 7.66 out of ten, but the quality of democracy depends not only on electoral procedure but also on whether citizens have real opportunities to get involved and to understand how the political system works. It is therefore not enough to announce a single overall turnout figure and conclude that society is, or is not, apathetic. The state and the electoral administration must be able to publish timely, accurate and easy-to-use data that show where turnout is falling, which barriers affect voters and which problems still need to be examined. This is a precondition not only for good analysis but also for informed, critical and civically active voters.

• Obtain OCMA data. We need to know how many eligible citizens had no valid document on election day, broken down by municipality and by country. These data should be compared with Central Election Commission (CEC) statistics on refusals to issue ballot papers.
• Monitor document validity ahead of elections. If the problem is large, targeted information for voters should be considered.
• Review diaspora voting arrangements. This year almost nine in ten voters registered abroad did not vote. Ways to reduce practical barriers should be assessed, including simplifying postal voting and other forms of remote voting.
• Publish two turnout figures already during election day. Turnout in Latvia and abroad should be shown separately.
• Analyze turnout in more detail. Turnout should also be assessed regularly by municipality and by areas with different demographic structures. If it is falling persistently in aging and shrinking areas, that is a different problem from low diaspora participation and needs different solutions. In addition to research on electoral turnout, also using anonymized data held by the CEC and the OCMA, Latvia should perhaps consider a systematic democracy audit, this time led by the State Audit Office. The last democracy audit was carried out in 2014.

Latvia has no single state institution whose direct and sole task is “the quality of democracy.” As Latvia is a parliamentary state governed by the rule of law, responsibility for the democratic process and for maintaining its quality rests with the entire system of public administration and the three constitutional branches of power, together with civil society.

Conclusion

Overall turnout of 51.79% says little in itself about the political mood of Latvian society. Turnout among those living in Latvia has in fact returned to its 2018 level – 57.7% compared with 57.5% – while the overall figure is being pushed down more and more by the growing number of eligible voters registered abroad and their very low turnout.

Within Latvia, meanwhile, change is uneven. In some areas the high mobilization of 2022 has faded, while in rural municipalities the number of voters is lower than in 2018 as well. Low turnout therefore cannot simply be explained by “Russian speakers who stayed at home.” The fall in turnout is closely linked to the demographic structure of areas, while ethnic composition does not explain this geography of demobilization. At the same time, ethno-demographic composition is closely linked to the territorial distribution of party support.

The 2026 election therefore cannot be reduced to a story about voter apathy. It is the sum of several demographic, territorial and participation processes. This means that election outcomes are shaped not only by the distribution of political preferences but also by how turnout differs between areas and social groups.

Methodology. Data: Central Election Commission (CEC) (2026 data provisional; 806,641 / 51.79%), Office of Citizenship and Migration Affairs (OCMA) population register statistics as of July 1, 2026. Turnout excluding voters abroad calculated by the authors. The number of eligible voters registered abroad in 2018 is an estimate derived from turnout published by the CEC. Early voting data (169,605) from the CEC. A supplementary municipal-level analysis tested the relationship between the change in the number of voters in 2022–2026 and areas’ demographic structure, ethnic composition, electoral district and 2026 party results. Municipal-level relationships should not be interpreted as evidence of individual voter behavior.