In simple terms
A survey asks some people, not everyone. Ask a different sample and the answer would come out a little differently.
The margin of error puts a size on that wobble. It does not say the survey is wrong. It says how precise the number is.
How it works
Sampling error is the difference between an estimate based on a sample and the value you would get from the whole population. The margin of error measures how large that difference is likely to be.
Add and subtract it from the estimate to get a range. An estimate of 564,757 households with a margin of error of 10,127 gives a range of 554,630 to 574,884.
The confidence level attached to the range differs by publisher, and it is easy to miss. Opinion polls conventionally use 95 percent. The US Census Bureau publishes American Community Survey margins of error at a 90 percent confidence level instead.
Size depends on how varied the population is and on the sample size. Larger samples give smaller margins.
Why it matters
Two limits catch people out.
First, the figure only covers sampling error. Badly worded questions, poor coverage and people declining to answer are not in it, and the size of those errors can never be known precisely.
Second, a headline margin applies to the whole sample, not to subgroups. In one survey of 1,000 people, a subgroup of 200 respondents carried a margin of plus or minus 6.9 percentage points.
This is why a narrow poll lead is often not a lead at all.
Where you’ll see it
- Election polls, usually in small print under the result.
- Official statistics such as census and labor force estimates.
- Market research reports quoting a confidence level.
Example
A poll puts a candidate at 47 percent with a margin of error of 3 percentage points. The plausible range is 44 to 50 percent.
Often confused with
The gap between two numbers. Comparing two figures involves the error in both, so a difference has to be larger than the headline margin before it means much.
Key facts
- The Census Bureau defines sampling error as the difference between an estimate based on a sample and the corresponding value that would be obtained from the entire population.1
- American Community Survey margins of error are published at a 90 percent confidence level; adding and subtracting the margin gives a range expected to contain the true population value 90 percent of the time.1
- The size of the error depends on the underlying variability in the population and on the sample size, and the larger the sample size the smaller the error.1
- AAPOR gives 95 percent as the conventional confidence level for polls, meaning the true answer is expected to fall inside the margin in 95 cases out of 100.2
- AAPOR notes that surveys are also subject to errors of question design and interviewing, so the precise amount of error in any poll finding can never be known.2
- The overall sampling error applies to the total sample, not to subgroups: in a 1,000-person survey, a subsample of 200 respondents had a margin of plus or minus 6.9 percentage points.2
Related concepts
In the news
Quick checkDoes a margin of error account for people refusing to answer a survey?Show answer
No. It covers sampling error only. Nonresponse, coverage gaps and question wording sit outside it and cannot be measured precisely.
Sources
- U.S. Census Bureau. Understanding and Using American Community Survey Data: What All Data Users Need to Know, chapter 7. 2018 (accessed 14 September 2026)
- American Association for Public Opinion Research. Margin of Sampling Error / Credibility Interval. Undated (accessed 14 September 2026)
Editorially reviewed by Specialty Digest Editorial TeamLast reviewed September 17, 2026Researched and drafted with AI assistanceReport an issue