A growing share of election polls in the United States – almost 70% of national polls in 2024 – are now weighted on political party affiliation or reported vote choice in a previous election. This kind of political weighting is controversial. But evidence shows that it has made polls modestly more accurate.
This report explains how political weighting works, how Pew Research Center does it, and some of its limitations.
It has become more necessary to weight on party or past vote because of growing evidence that Republicans have become slightly less likely than Democrats to take part in polls. Historically, neither party’s supporters were more likely than the other’s to participate in polling. Whether or not that change is permanent, it is true today and requires addressing.
Note: For a primer on survey weighting and how it’s used, read “What is weighting?” below.
Key takeaways
- Political weighting can make polls more accurate. Weighting 2020 and 2024 Pew Research Center postelection polls on party affiliation improved accuracy by about 1-2 percentage points per candidate – or about 2-3 points on the margin between candidates.
- Weighting on political variables has become more necessary in the past decade for two main reasons: a systematic bias that emerged around 2016 which led polls to understate Republican candidates’ share of voter support, and poll-to-poll volatility in the balance of Republican and Democratic voters driven by who participates rather than real changes in opinion. Our report examines this in detail.
- Three common approaches are weighting on self-reported political party affiliation, self-reported past vote choice, and voter file information about party affiliation such as party registration. Each approach has its own strengths and weaknesses, which the next section of this report goes over in more detail.
- Under certain conditions, weighting on self-reported political party has the potential to align a poll with an outdated snapshot of the country. Panel-based surveys can neutralize this problem by keeping track of respondents’ past and present party affiliation. Read more about this approach and its benefits, including a case study from our research.
- There’s more than one factor to consider when choosing what type of political weighting to use – if any. Pollsters need to decide whether to weight a poll on party affiliation, past vote, both or neither. This in turn is affected by the population (all adults or just voters), whether the survey includes political content, and whether an appropriate benchmark is available.
Because there are long-standing questions about how this weighting works, which political outcomes or characteristics to use, and a variety of downsides to consider, this piece seeks to explain:
- Three ways to correct a poll’s political balance
- Why weighting for party affiliation is easier with a panel
- Why many polls need to adjust for party affiliation or past vote
- FAQs about political weighting and its limitations
What is weighting?
Surveys are still the best way to measure what the public is thinking. But they aren’t perfect. It’s hard to contact people for surveys and hard to get them to cooperate. We often don’t get enough people who have less than a high school education or who are younger than 30, groups that are often underrepresented compared with their share of the overall population.
One of the best tools pollsters have for dealing with this is called weighting. To weight a survey, we need targets for the characteristics that are out of balance in our sample. (A target is a template for what a survey sample would look like if it was selected from the population in a way that was truly random.) Weighting is the process of adjusting the raw percentages of groups in our actual survey sample so they match these target shares as closely as possible.
For a few characteristics of the population, we can compare our survey with a trusted source like the U.S. Census Bureau and use that to make sure that our survey has the right percentages (matching the makeup of the actual population) of older and younger people; White, Black, Hispanic and other racial or ethnic groups; and people with a high school education.
Unfortunately, there is no trusted government estimate of party affiliation that we can use to correct a poll’s political balance. But pollsters, including Pew Research Center, have found ways to do this. We outline some of these approaches in the sections of this report that follow, and we address some common questions about political weighting and its limitations.
