How are polling averages calculated?

Raw results are published unchanged, including the undecided category. These records are excluded from party averages on the home page, monthly charts and share images; shares are never redistributed or normalized. Experimental comparisons may show them within separate methodology groups.

The chart shows individual polls, the monthly average and the observed minimum–maximum range. This range is not a confidence interval. The difference from the previous month is calculated only if that party has an observation in the immediately preceding calendar month; missing months are not skipped to infer a change. Changes are in percentage points and do not establish voter movement or statistical significance.

The analysis lab uses a 90-day fieldwork-end window. Records missing source checks, fieldwork/publication dates, sample size, method or treatment of undecided voters are excluded. Different method/undecided groups are kept separate. At least three polls and two pollsters are required. We compare equal weighting, equal weighting of pollster averages, and square-root sample weighting multiplied by a 30-day half-life weight. Results do not change the main monthly average and are not election forecasts. Historical views are limited to the archive available today.

The reference line divides the sum of raw party votes across 81 provinces by total valid votes. It excludes overseas votes and allocated alliance votes. The chart states the source and coverage. DEM Parti is not automatically matched with the historical Yeşil Sol Parti; Other categories are not compared. No 2023 accuracy scores are produced because a historical polling set is unavailable.

The archive file's capture date and notes are stored separately. Uploading a file does not verify its content. CSV includes original sources, dates, methodological information and the export date. Empty fields mean information is unknown. A share image is a snapshot; an embedded chart updates with publishable records. Candidate, bloc and undecided-voter surveys are kept separate; bloc totals are not automatically derived from parties.

Published records must have an HTTP or HTTPS source link, must not be marked as samples, must have no review note and must not be future-dated. The archive covers all years; the main chart covers the past year. A source link alone does not guarantee a study's accuracy or representativeness.

A party's reported shares in the same calendar month are added together and divided by the number of polls reporting a share for that party. Every poll has equal weight; neither sample size nor a pollster's past performance is used as a weight. For example, a party with 30% and 34% in two polls has an average of 32%.

If a party's share is not reported, it is not treated as zero; the observation count for that party is lower. Party averages may therefore not add up to 100%. Date ranges use their final day; records specifying only a month use that month. Fieldwork start, fieldwork end and publication dates are recorded separately when verified in the source. A legacy record's chart date should not be interpreted as its exact publication day.

Question wording, target electorate, fieldwork method and treatment of undecided voters may differ. Known methods and undecided treatment are shown beside each poll; missing fields remain unspecified. Review the original source and its check status together. Two studies by the same pollster are not automatically comparable. We therefore do not generate automatic increase/decrease interpretations or claims of statistical significance.

No. These are descriptive averages of records in the archive. Sampling error, methodological differences between pollsters and archive coverage may affect the result. A simple average does not yield a confidence interval or a probability of winning an election. The current month is incomplete, so its average changes as records are added.

Voluntary responses submitted to the site do not form a probability sample and are not representative of Türkiye's population. They are not included in pollsters' averages. No illustrative sample series appear in the public chart.

The monthly summary covers records added to the site from the first day of the month through today, using the UTC calendar. For older records without an added date, the chart date is used. It is not a list of every study published that month. A record specifying only a month can fall within this window without its exact publication day being known.

When sharing data, include the pollster, date label, original source link and Trends Türkiye page. Use the contact page to report incorrect records. When a source is updated, results and summaries are recalculated on the next request.