2026 Poll Aggregator

Methodology

Poll aggregators are essential tools for getting as close as possible to the real state of public opinion during a campaign. Relying on a single pollster isn't ideal. Aggregating polls with different methodologies offers the best source of information about the state of a race over time.

Our methodological choices for building this aggregator aim to find the best way to combine the information published in hundreds of polls:

• Recent polls count more: we use every poll we could find data for since January 2026, when TSE registration became mandatory. Polls closer to the election, however, are more likely to correctly anticipate the outcome expressed at the ballot box; so we apply a weight adjustment to each poll based on its fieldwork end date as reported to the TSE.
• Polls from pollsters with a track record count more: pollsters we assess in the Pollster Performance (2002-2022) series carry more weight in the aggregator, but are also corrected for the bias they showed in past cycles, in a simplified version of what's known as house effects.
• State-level presidential polls are included too: many pollsters only poll in a handful of states. When their polls include a question about presidential voting intention, we fold those numbers into a multilevel model. State polls are weighted by that state's electorate size and also compared against the 2022 election result. Since these numbers depend on more methodological choices, you can choose to see the presidential numbers with national polls only, in the view above.
• Uncertainty around the numbers is shown as confidence intervals: our confidence intervals are built via bootstrap resampling, meaning the size of the uncertainty is measured from the available data without assumptions about the distribution of errors. You can see the confidence intervals as the shaded areas on the chart; where they overlap, that represents a chance of a statistical tie between the candidates involved.

We don't include polls that have been legally challenged or restricted by a court decision we're aware of. If you notice a poll is missing, you can reach us here.

Methodology (Governors)

State races have no national-projection step — it's a direct aggregation of polls within each state, with exponential recency smoothing (14-day half-life) and a 95% bootstrap interval (resampled by pollster). There's no historical-bias correction by pollster in this tab (the house effects series used for President is specific to the presidential race).

The featured pair (or trio) of candidates per state was picked manually based on available poll coverage — not every race has a stable matchup, and this choice may change as the picture evolves. Secondary candidates with missing data in some polls have their value imputed by the mean of the polls where they're named individually (the same technique used for secondary candidates in the presidential race).

Partial coverage: only states with enough polls for a minimally stable reading are listed; others will be added as more data comes in.