A job-approval poll measures how a defined group rates an officeholder’s performance at a particular time. Read the question and methods alongside the headline percentage. That combination tells you whose views were measured, what they were asked and how confidently you can describe a change.
Start with the question
Approval, favorability and voting intention ask different things. Someone can approve of an official’s handling of an issue, dislike that official personally and still prefer them to an opponent. Keep those answers separate when describing public opinion.
Read the exact wording and answer choices. A question about “the job the governor is doing” has a different scope from one about handling housing costs. A scale such as excellent, good, fair and poor needs its own definition of a positive response. Check how the pollster combines categories and handles people who express no opinion.
Earlier questions can influence later answers. The American Association for Public Opinion Research’s journalist guide explains wording, order and comparisons across surveys. Keep a copy of the questionnaire with the result you cite.
Identify the people and dates
| Population | What the label tells you | What to check |
|---|---|---|
| Adults | Views among the adult population covered by the survey. | Geographic coverage, language and age definition. |
| Registered voters | Views among people classified as registered to vote. | Whether registration is self-reported or checked against a voter file. |
| Likely voters | Views among people meeting a turnout screen or model. | How likely voting was defined for this election. |
| A subgroup | Views within a subset, such as an age group. | The subgroup’s own respondent count and uncertainty. |
Record the interview dates, publication date and geographic area separately. If interviews run from Monday through Thursday and a major announcement happens Friday, the release published Saturday still describes the earlier field period.
For a trend, begin with the same pollster, population, question and method. Keep a method change visible in your notes. A national adult survey and a state likely-voter survey can each be useful, but joining their percentages into one line would give the line a changing meaning.
Open the methods statement
AAPOR’s disclosure standards provide a useful checklist: sponsor and research organization, question wording, population, recruitment, survey mode, field dates, sample sizes, weighting and reported precision. Look for the full release or technical appendix when a news story gives only a percentage.
Recruitment and mode answer separate questions. Recruitment explains how people entered the sample; mode explains how they answered, such as online or by telephone. An online survey may recruit respondents through a probability-based process or an opt-in panel. The word “online” alone leaves that distinction unresolved.
Weighting changes respondents’ contributions to the estimate to align selected characteristics with benchmarks. Record which characteristics and benchmarks were used. AAPOR’s survey best practices place weighting within the broader work of sampling, questionnaire design and data quality. Treat a missing methods statement as missing evidence when assessing the result.
Read uncertainty at the right level
A margin of sampling error describes sampling uncertainty under the stated design. For probability samples, check the confidence level and whether the estimate accounts for weighting or other design features. Smaller subgroups usually have wider uncertainty than the full sample. Apply the subgroup’s information to a subgroup claim.
Coverage gaps, nonresponse and question effects require separate attention. A very large respondent count can still leave systematic problems. For a nonprobability sample, examine the pollster’s model and assumptions; a model-based interval has a different basis from a conventional margin of sampling error. AAPOR explains that distinction in its statement on credibility intervals.
To assess whether two estimates differ beyond sampling uncertainty, use the uncertainty of the comparison itself. The pollster’s analysis should account for the samples’ designs and any overlap. Comparing a change with just one poll’s margin can give a misleading significance claim.
Work through a comparison
Fictional example: two waves and a different question
Survey A reports job approval among adults at 42% in March and 46% in April, using the same question and method. The observed increase is 4 percentage points: 46 − 42. Relative to the earlier estimate, it is about 9.5%: 4 ÷ 42 × 100.
Each wave reports a margin of ±3 percentage points. Those two margins alone do not establish the statistical significance of the four-point change. Check the comparison’s uncertainty before describing an established shift.
Survey B reports 51% favorability among likely voters in April. Enter that as a separate measure. It answers a different question among a different population, so it cannot serve as Survey A’s next observation.
A careful summary would be: “In this fictional adult survey, estimated approval rose from 42% to 46%; the comparison’s statistical uncertainty still needs checking.” That sentence preserves the result and the open question.
Keep a short reading record
Use the poll-reading notes worksheet to record one result and any comparison. Save the original release URL, question, field dates, population, method and calculation. End with one sentence stating what the evidence supports and another naming any missing information.
When a campaign uses a poll in an advertisement, follow the claim back to the release and read the political-ad claim-checking guide. Apply the same questions across parties and sponsors. A usable record makes it easier to revise your interpretation when the next comparable survey arrives.
About this guide
Sources checked October 6, 2026. Follow the linked source collections for current guidance and revised records. Explore Politics and Civic Life.