Approval Ratings, Explained: How the Polls Work and Why Averages Beat Single Numbers

What an approval rating actually is

An approval rating is a survey estimate. A pollster contacts a sample of adults, registered voters, or likely voters and asks some version of one question: do you approve or disapprove of the way the president is handling his job. The answer is a percentage with uncertainty attached. It is not a count. It is an estimate built from a sample, weighted so the sample looks like the country on characteristics such as age, sex, education, race, and region (Pew Research Center, Public Opinion Polling Basics). Gallup has asked the question in nearly the same form since the 1930s, which is why approval is the longest continuous performance metric in American politics (Gallup, Presidential Approval Ratings).

The uncertainty is not a footnote. A typical national poll of 1,000 respondents carries a margin of sampling error around plus or minus 3 percentage points, and that figure covers only sampling error. It does not cover errors from who refuses to answer, how questions are worded, or how likely voters are modeled (Pew Research Center, Key Things to Know About Election Polls in the U.S., 2024). A president at 42 percent in one poll and 46 percent in another may be at the same true level of support in both.

Why polling got harder

The machinery behind these numbers changed completely in one generation. Telephone response rates collapsed. Pew Research Center's telephone surveys got responses from 36 percent of sampled households in 1997. By 2018 the figure was 6 percent (Pew Research Center, How Public Polling Has Changed in the 21st Century, 2023). Most major pollsters responded by moving to online panels, mail recruitment, text messaging, or mixed methods. Pew counted dozens of distinct methodological approaches in use among national pollsters by the early 2020s, where a generation earlier nearly everyone used live telephone interviews (Pew Research Center, 2023).

Different methods produce different results. That is not scandal. It is arithmetic. A survey that reaches people by mail recruits a different slice of the public than one that reaches people through an opt-in web panel. The weighting choices that correct for those differences are judgment calls, made differently shop by shop.

House effects: why the same week produces different numbers

A house effect is a pollster's persistent tendency to show results a point or two friendlier to one party or one number than the average of all polls. House effects come from method, not motive. Sample source, weighting targets, question order, and likely voter screens all push results in consistent directions for a given firm (Pew Research Center, 2024). A pollster with a two point Republican-leaning house effect is not lying when it shows the president two points above the average. It is measuring the same reality through different instruments.

The practical rule follows directly. Never compare this week's number from Pollster A to last week's number from Pollster B and call the difference movement. Compare Pollster A to its own previous result. Trend within a pollster is meaningful. A gap between two pollsters is usually just the gap between two methods.

Why averages beat single polls

Averaging works because it cancels noise. Sampling error is random, so combining many polls shrinks it. House effects are systematic, but they point in different directions across firms, so a broad average nets them against each other. This is why every serious quantitative forecaster, from the old FiveThirtyEight to today's averages, built on aggregates rather than single surveys. It is also why the professional association for pollsters, AAPOR, pushes disclosure standards: an average is only as good as the transparency of what goes into it (AAPOR, Standards and Ethics).

Averages are not magic. When the whole industry shares a blind spot, the average inherits it. AAPOR's task force found that 2020 pre-election polls overstated the Democratic presidential margin by 3.9 points in national polls and 4.3 points in state polls, the largest national error in 40 years, and the error could not be fully explained after the fact (AAPOR, Task Force on 2020 Pre-Election Polling, 2021). The lesson is not that polls are worthless. The lesson is that a polling average is a measurement with an error bar, and the error bar is wider than the decimal points imply.

How to read the number this month

Apply the rules to the present. Marist's June 2026 survey put President Trump's approval at 36 percent (Marist Poll, June 2026). Broader averages across pollsters through 2026 have run in the high 30s to about 40 percent (Wikipedia, Opinion polling on the second Trump presidency, accessed July 2026). The single poll and the average disagree by a few points. Per everything above, the average is the better estimate, and the Marist result is one input to it, not a replacement for it.

What makes approval worth tracking is not any one reading. It is the long, comparable record. Approval below about 50 percent has historically preceded large midterm losses for the president's party (Gallup, Midterm Seat Loss Averages 37 for Unpopular Presidents, 2018). That regularity is why we report the average, name the pollsters, and skip the daily drama. One poll is a data point. The average is the scoreboard.