Studentska lista leads Ujedinjena Srbija by 113 seats to 106 in Izborni monitor’s October 22 projection for Serbia’s October 25 parliamentary election. Neither reaches the 126-seat majority. The figures come from a model combining agency polls with an online visitor survey, rather than a new representative national poll.
What the projection shows
The Election Monitor’s final projection assumes 62% turnout and gives the following estimates:
| List | Estimated vote share | Projected seats |
|---|---|---|
| Studentska lista | 42.88% | 113 |
| Ujedinjena Srbija | 40.34% | 106 |
| Faktor stabilnosti | 4.02% | 10 |
| Evropska Srbija | 3.51% | 9 |
| SVM | 1.54% | 5 |
| Pokret za ujedinjenje | 0.62% | 2 |
| SDA Sandžaka | 0.56% | 2 |
| Rasim Ljajić – Ljudski | 0.44% | 1 |
| Albanska koalicija | 0.43% | 1 |
| Ruska stranka | 0.29% | 1 |
| Autentična desnica | 2.69% | 0 |
These are model estimates, including assumptions for smaller lists. The ten lists allocated seats total 250. Studentska lista and Evropska Srbija together would have 122; the lists the model groups with the outgoing majority would have 125. Both combinations fall short.
How the model is built
The final blend assigns 80% to the agency polling average and 20% to the project’s own online survey, with 601 responses. The project is edited by Vencel Varjaši, also known as Varjasi Vencel.
Its published methodology weights agency polls by recency, using a 21-day half-life, and by the square root of sample size. It also balances polling houses so that frequent publication does not automatically give one agency disproportionate influence. Missing party figures are omitted rather than treated as zero.
For seat allocation, the model uses D’Hondt, a 3% threshold calculated from all ballots cast, and minority-list exemptions with a 1.35 multiplier below the threshold. Minority support is partly estimated from historical vote counts; some small lists receive an equal share of the “other” category.
Its uncertainty simulation incorporates poll resampling, sampling noise, a shared three-point polling error and variation in minority vote totals. Reproducing the 2023 seat allocation from known votes checks the allocation mechanism; it does not establish forecasting accuracy.
Why the online component matters
The survey form explicitly describes a non-representative sample of website visitors. Responses are weighted by sex, age, residence, education, ethnicity and reported 2023 vote.
Weighting can correct some observable imbalances. It cannot, by itself, establish that volunteers adequately represent people who never encountered the survey or declined to answer. A conventional random-sample margin of error should therefore not be attached to the 601 responses.
The distinction changes the interpretation: the site’s displayed agency average puts Ujedinjena Srbija at 45.6% and Studentska lista at 42.2%, while the final blend reverses their order. These are different model outputs, not evidence of a measured late swing. Without a full numerical reconciliation, the separate effects of online responses and subsequent adjustments cannot be isolated.
The form says its visitor survey is excluded from the standard polling average. The final projection nevertheless explicitly includes it. Readers should keep those two products separate.
What happened with the Hungarian projection?
Varjaši’s Hungarian project, Választási Monitor, was presented in cooperation with investigative outlet Átlátszó on February 13. Its archive records an April 11 estimate alongside figures labelled as the final official outcome:
| Party | Projected votes | Final votes* | Projected seats | Final seats* |
|---|---|---|---|---|
| Tisza | 49.2% | 53.18% | 116 | 141 |
| Fidesz–KDNP | 38.9% | 38.61% | 76 | 52 |
| Mi Hazánk | 5.6% | 5.63% | 6 | 6 |
*Final figures as reproduced in the project’s archive.
The projection identified the winner but understated Tisza by 25 seats and overstated Fidesz by 24. It implied an ordinary Tisza majority, whereas the archived result exceeded the 133-seat threshold for a two-thirds majority. That record provides useful context, but one election cannot establish a model’s general reliability, particularly across different electoral systems.
How Votendi treats this estimate
This is coverage of an external experimental projection, not a Votendi forecast or an independently verified representative survey. It should not be added as a fresh poll to an average that already contains its underlying agency surveys: that would count some evidence twice.
Compare the underlying evidence in our Serbia election hub and our separate analysis of the October 22 Ipsos poll. The projection discussed here is dated October 22; this article was published October 23.
