Every polling average makes methodological choices about which polls to include and how much influence each should have. This methodology is used for both the NBC News presidential approval average and the generic congressional ballot average.
While the underlying calculations differ slightly to reflect the characteristics of each type of polling, the overall approach is the same.
Our approach emphasizes five principles:
- Recency: Newer polls matter more than older polls.
- Precision: Polls with a lower margin of error are weighted higher. The margin of error depends on the number of people surveyed, and how much the result needs to be adjusted through weighting in order to ensure that the poll is representative.
- Stability: Individual outliers and day-to-day fluctuations have limited influence until they are supported by additional evidence.
- Variety: Frequent pollsters have their older polls downweighted more sharply so that no one firm can dominate the average.
- Consistency: Polls measuring different populations are placed on the same registered voter scale before being combined.
Averages not predictions. Instead, they aim to provide the clearest possible aggregate measure of the current available polling.
About the presidential approval average
Which polls are included?
The NBC News presidential approval average includes publicly available national polls measuring presidential job approval. Polls of all adults, registered voters and likely voters are all eligible for inclusion, as are polls with a partisan affiliation provided those surveys meet certain methodological transparency requirements. Polls with partisan affiliations are noted as such whenever possible.
NBC News collected all the surveys included in the average. NBC News also sponsors its own independent national polls, the NBC News Poll and the NBC News Decision Desk Poll. Those polls are included in the average, and you can read more about those polls below.
To be eligible for inclusion in the average, a polling organization must, at minimum, meet methodological transparency standards by reporting details on when a poll was conducted and how many respondents were interviewed. If a poll does not publicly release information on field dates, sample size and/or margin of error, it is not included.
Many polling organizations release results for more than one sample population. When a poll reports multiple universes — for example, both adults and registered voters — it is counted only once in the overall average to avoid giving extra weight to a single survey. In such cases, preference is given to registered voters. More on that below.
Each poll’s reported approval and disapproval percentages, field dates, sample size and margin of error are factored into the average.
If a polling organization releases results broken down by party identification, we will display those. For the presidential approval average, we display the percent of self-identified Democrats, Republicans and independents who report approving and disapproving of the president.
Why do some polls count more than others?
Not every poll contributes equally to the average. Older polls gradually receive less weight as new surveys are released. Likewise, polls with smaller sample sizes or larger margins of error contribute less than more precise surveys. Poll margins of error are capped at plus or minus 2.0; polls with a smaller margin of error do not gain additional weight.
When a polling organization releases surveys frequently, its older polls are discounted so that no single pollster dominates the average simply because it publishes more often.
Surprising poll results that are far from the current average are given less weight until other polls confirm the result.
Why does a poll’s sample population matter?
Polls do not always survey the same population. Some polls interview all adults, while others report results only among registered voters (RV) or likely voters (LV). Registered voters are respondents who are verified as registered voters through a voter file, or respondents who say they are registered to vote. Likely voters are registered voters whom the pollster identifies as likely to vote, typically based on questions about their voting intentions, past voting behavior or other measures. Pollsters can identify these groups while conducting the survey or select them from a broader group of respondents when reporting the results.
These groups consistently produce different levels of presidential approval, even when surveyed by the same organization at the same time.
The NBC News polling average combines these polls from different populations together to produce one overall number. To do this, all adult and likely voter polls are adjusted in the average to match registered voter polls. We use registered voter polls as our baseline as they are the most common population surveyed and are the common link between the three populations.
To adjust all adult and likely voter polls, we leverage the fact that some pollsters release polls that include multiple populations – both all adults and RV, or both RV and LV. We use these paired results to dynamically calculate how all adult and LV polls need to be adjusted to be on the same scale as RV polls.
This adjustment allows polls measuring different populations to be combined without introducing artificial shifts into the average.
For polls that do release results from multiple populations, only one result from each poll is entered into the average, with the preference order being registered voters, then all adults, then likely voters.
For example, if a poll releases both a registered voter and a likely voter result, the registered voter result will be entered into the average. For transparency, all of the populations reported by a poll are entered as separate results in the full list of polls.
Why display multiple averages with different sample populations?
In addition to the overall average, approval trends are calculated separately for adults, registered voters and likely voters, using only polls of that type.
These groups often hold measurably different opinions. Displaying the populations separately allows readers to see how approval varies across populations.
Because relatively few likely-voter polls are conducted early in an election cycle, that series may not be displayed until sufficient data are available.
How are outlier polls factored into the average?
Occasionally, a poll reports a result that differs sharply from the results of other publicly available poll sources on that question or topic.
Rather than allowing a single surprising poll to dramatically move the average, the model limits its influence unless additional polling confirms the result. If other polls show the same movement, the full effect is incorporated into the average.
This approach helps distinguish genuine changes in public opinion from normal survey-to-survey variation.
What does the trend line represent?
The published average is intentionally smoother than the individual polls.
Instead of moving sharply every time a new survey is released, the trend changes gradually as evidence accumulates. This makes the average easier to interpret and helps emphasize sustained changes in public opinion rather than short-term fluctuations.
About the generic congressional ballot average
The NBC News generic congressional ballot average uses the same methodology as our presidential approval average, with adjustments to reflect the nature of congressional preference polling.
Because congressional preferences generally change more slowly than presidential approval, the generic ballot average is designed to place slightly more emphasis on longer-term trends than short-term movement.
The generic congressional ballot is asked differently across polls. Some polls ask which party should control Congress and others ask which party’s candidate you would vote for if the election in your Congressional district were being held today. The average includes the Democrat and Republican percentages across all different question wording.
Why does the average change to a likely voter scale?
Like the presidential approval average, we made adjustments to polls to place them all on a registered voter scale, because that is the most common survey sample type for pre-election polling. However, as is typical after Labor Day in an election year, polling organizations have begun releasing results among likely voters instead of, or in addition, registered voters.
Because the congressional ballot preference question is a question directly about how people will vote, the polling average shifts to a likely voter scale after Sept. 7 to more accurately reflect the environment. That shift is noted on the chart.
In both cases, all polls are still being entered into the average. What changes is what population is outwardly represented by the average. Once the average shifts to the likely voter scale, you can see that both lines on the chart — the percent preferring Democrat, and the percent preferring Republicans — move upwards. That’s because there are fewer undecided likely voters than undecided registered voters.
How do we look at results by partisanship?
For the generic ballot tracker, we display the percent of those who identify as Democrats, Republicans and independents who prefer a Congress controlled by the Democratic Party or Republican Party. We display those numbers separately for Democrats, Republicans, and independents.
These results are provided for those polls that have publicly available crosstabs. When available, we prioritize including results where Democratic and Republican “leaners” are broken out into their respective parties.
The same weighting mechanism described above is applied to the trends by party identification. However, these trends make no adjustment for different sample populations: numbers are entered into the average as-is.
More about NBC News polling
NBC News sponsors two polls that are conducted multiple times throughout the year — the national NBC News poll and the national NBC News Decision Desk poll.
The national NBC News poll is conducted among registered voters and led by Republican pollster Bill McInturff of Public Opinion Strategies and Democratic pollster Jeff Horwitt of Hart Research Associates. The surveys are typically conducted using a mix of telephone interviews and an online survey sent via text message. NBC News maintains full editorial control over topics, questions and wording.
The NBC News Decision Desk poll is powered by SurveyMonkey, an online platform for surveys. Respondents are selected from the more than 2 million people who take surveys on the platform each day, and results are reported among all adults unless otherwise noted upon release of the poll. Data checks are used to ensure quality respondents. Partnering with the platform allows the NBC News Decision Desk poll to capture a larger number of respondents than other public opinion polls, making in-depth subgroup analysis possible, including among hard-to-reach groups such as Gen Z.
Are we missing a poll? Spot a possible error? Email polls@nbcuni.com.
MORE POLITICS COVERAGE
