Our Mission
USPollingData.com was built to make high-quality polling data accessible, transparent, and comparable across borders. We track the generic congressional ballot, presidential approval ratings, swing-state snapshots, and issue polling for the United States — and mirror that view against current trends in Germany and the EU.
The site is designed for readers who want more than a headline number: the methodology behind each average, the house effects of individual pollsters, and the structural differences between American and European political systems. We believe informed citizens make better decisions, and that understanding how polling actually works is the first step toward not being misled by it.
We started with a simple observation: American and European political cycles increasingly mirror each other. Populist surges, incumbent collapses, third-party rises, and trust crises show up in Washington, Berlin, and Brussels within months of each other. Covering them in isolation gives readers only half the picture.
We do not endorse any party or candidate. Our role is to present data accurately and explain what it does — and does not — mean.
How We Aggregate
Every day our data pipeline collects newly published polls from all major US pollsters. Each poll is entered into our database with its field dates, sample size, methodology, and topline results. From this dataset we compute rolling averages for key series — Generic Ballot, Presidential Approval, Senate and House balance — using the following model:
- Recency weighting — polls conducted more recently receive higher weight. We apply exponential decay with a half-life of approximately 21 days. A poll released yesterday counts roughly twice as much as one from three weeks ago. This prevents stale surveys from anchoring the average when public opinion is shifting.
- Sample size weighting — larger samples carry more statistical weight, proportional to the square root of N. A poll of 4,000 respondents counts approximately twice as much as one of 1,000. We use the square root (rather than raw N) to prevent single very large surveys from crowding out the field.
- Rolling 7-day average — the displayed figure is the weighted average of all polls whose field end date falls within the past seven days, with older polls included at reduced weight up to 30 days. This provides a responsive but stable estimate.
- Pollster transparency flag — polls from organizations with known methodological concerns (non-disclosed panels, no transparency about question wording, or a documented history of extreme house effects) are included in the raw data but flagged in the display. Readers can filter them out.
We do not currently adjust for pollster house effects in the main average, though we publish house-effect estimates alongside each firm’s historical record. All raw data — date, pollster, sample size, margin of error, and topline results — is published alongside every average so readers can apply their own weighting.
For a deeper explanation of polling methodology, see our Major US Polling Firms guide.
Data Sources
We aggregate polls from the following organizations. All sources are publicly available; we do not receive data under embargo or pay for exclusive access. For each series we also incorporate Wikipedia aggregator averages as a cross-check signal.
| Pollster / Source | Methodology | 538 Grade | Notes |
|---|---|---|---|
| Gallup | Phone / Online | A | Oldest US pollster; gold standard for approval and issue tracking |
| YouGov | Online panel | A- | US, EU & DE polling; large international footprint |
| Quinnipiac | Live telephone | A- | RV/LV breakdowns; strong swing-state coverage |
| Marist College | Live telephone | A | Long-running approval series; NBC/WSJ partner |
| Monmouth University | Live telephone | A+ | Highly respected; careful screening methodology |
| Pew Research Center | Online / Phone | A+ | Non-partisan; deep issues polling; not horse-race |
| AP-NORC | Mixed mode | A | Associated Press partnership; strong issue coverage |
| Siena College | Live telephone | A | NY Times partner; battleground state specialist |
| Reuters / Ipsos | Online panel | B+ | Continuous tracking; frequent field dates |
| NBC News | Live telephone | A- | National survey; registered voter universe |
| Fox News | Live telephone | A- | Bipartisan team (SSRS); included for breadth |
| CNN / SSRS | Live telephone | A- | Periodic national surveys; LV screen applied |
| Emerson College | IVR + Online | B+ | Frequent state-level polling; active in primaries |
| CBS News / YouGov | Online panel | A- | Large-scale national surveys |
| Wikipedia Poll Avg. | Aggregator | — | Cross-check signal; includes all publicly listed polls |
Additional pollsters are included on a case-by-case basis. Internet-only polls without disclosed sample methodology are excluded from the main averages but stored in the raw data archive.
Methodology
Our polling averages use a weighted model similar in spirit to FiveThirtyEight’s approach. Each poll enters the average with a weight determined by recency and sample size as described above.
Registered-voter and likely-voter screens are treated separately where relevant. For the Generic Ballot we display RV and LV averages side by side; for Presidential Approval we note which universe each poll used. As the 2026 midterm election approaches, the displayed default will switch from RV to LV to reflect the electorate that will actually vote.
Margin of error is displayed for each individual poll in the raw data table. For our rolling average, the effective MOE is lower than any single poll because aggregation reduces sampling variance — but we do not publish a single average MOE, because systematic bias (house effects, non-response bias) is harder to quantify than random error and would give a false sense of precision.
All raw data — date, pollster, sample size, margin of error, and topline results — is published alongside each average so readers can apply their own weighting if they choose.
The Transnational Angle
American political events have immediate consequences in European capitals — and vice versa. The rise of populist movements, shifts in NATO policy, trade disputes, and currency markets all cross the Atlantic within hours. Understanding US polling numbers in isolation misses the broader pattern.
Our EU and Germany sections track Bundestag party preferences, European Parliament seat projections, and chancellor/president approval ratings using the same display logic as the US data. This lets analysts and readers spot structural parallels: when trust in incumbents falls simultaneously in Washington, Berlin, and Brussels, it tells a different story than a purely domestic reading would.
Germany in particular functions as a useful comparison point. It is the largest EU economy, its governing coalition dynamics are regularly compared to the US two-party system, and its media ecosystem is sophisticated enough to produce high-quality poll aggregates. We source German polling from the established Sonntagsfrage aggregates: Forsa, Allensbach, INSA, Civey, Ipsos Germany, and Kantar.
EU-level data comes from Eurobarometer — the European Commission’s official survey program — as well as national partner pollsters in France, Italy, Poland, and Spain. We track European Parliament seat projections updated after each major wave of national polls.
Our view is that readers in Germany, Austria, and other German-speaking countries deserve access to US polling data in a format that makes the comparison intuitive. The Transnational Desk was designed from the ground up to serve both audiences simultaneously.
Independence & Non-Partisanship
USPollingData.com is an independent, non-partisan data project. We have no political affiliation and receive no funding from political parties, campaigns, political action committees, or ideologically motivated donors.
We include polls from across the methodological spectrum — including from pollsters with known partisan house effects (Rasmussen, Emerson) — because excluding them entirely would itself introduce a selection bias. We flag house effects prominently in our data display and explain them in our methodology notes.
When polls show divergent results, we show the full distribution rather than cherry-picking the average that supports a preferred narrative. Our goal is for readers to come away understanding the range of uncertainty, not just the central estimate.
If you believe we have miscoded a poll, misrepresented a methodology, or introduced a systematic error in our averages, please contact us. We treat correction requests seriously and publish corrections prominently.
The Team
Responsible for the polling aggregation model, house-effect calculations, and daily data pipeline. Background in statistical modeling and electoral data science.
Writes the race analyses, explainers, and issue trackers. Background in comparative political science and US electoral mechanics. Covers Senate and House battlegrounds.
Tracks EU Parliament seat projections, Bundestag polling, and transatlantic political trends. Sources German-language polling and produces the comparative analysis.
- All polls are sourced from publicly available field reports — we do not use embargoed or proprietary data
- Corrections are published prominently; updated articles note the change date and what was corrected
- House effects from partisan pollsters (Rasmussen, Trafalgar) are flagged explicitly, not hidden in averages
- No article is published on behalf of a campaign, PAC, or political organization
- Methodology disputes are addressed publicly — readers can email us and we respond to all substantive corrections
The site is built and maintained independently. We have no investors or outside stakeholders whose interests could influence editorial decisions. CXMXO GmbH is the legal operator; full details in the Imprint.
Operator & Contact
USPollingData.com is operated by CXMXO GmbH. Full legal details are available in the Imprint.
For questions about missing polls, methodology, data corrections, or press inquiries, please reach out at:
We respond to all substantive corrections. If you believe a poll has been miscoded or a calculation is wrong, please include the pollster name, field dates, and the result you expected.