State of Polls

Methodology

Polls

Wikipedia’s editors keep tables of published polls for elections around the world, with the pollster, fieldwork dates and sample size for each. The site rereads those tables every hour, and every page links to the article its numbers come from.

If a poll is missing or wrong here, it’s probably missing or wrong on Wikipedia too, and a fix there will show up here within the hour. The tables are the work of Wikipedia’s editors and are used under CC BY-SA 4.0.

Averaging technique

Each poll’s weight in the average depends on three things.

The trend chart shows the average once a week for the last two years, and leaves out weeks with fewer than three polls behind them.

The average describes polling that has already happened. It isn’t a forecast, and polls are sometimes all wrong in the same direction.

Blackouts

Many countries ban publishing polls in the days before a vote: five days in Spain, the eve and the day itself in France, polling day in New Zealand. During a country’s window, its figures stay as they were when the window opened, and they start updating again once voting has ended.

Policy scale

Parties are scored from 0 to 10 on eight issues. The scales are the same everywhere, so a 7 on immigration in Denmark and a 7 in Mexico mean the same position. Each end of a scale is spelled out, and a few well-known parties serve as fixed reference points. On the economy, for instance, the US Democrats are a 5.

The scores, and the short reasons given for them, are written by an AI model working from each party’s Wikipedia article and public record. A second pass goes through one issue at a time, comparing every party on the site, and adjusts scores that don’t fit. The reasons are listed under “Why these scores” on each page. Some will be wrong; if you spot one, write to contact@stateofpolls.com.

In Israel the religion scale measures the role of Judaism, and in Indonesia and Palestine the role of Islam.