Margin of Error Calculator

Work out how far a poll or survey result could be from the truth, how many people you need to poll for a target margin, or the margin on an average. Drop in your sample size and result, pick a confidence level, and the page returns the margin drawn to scale, in plain words, with the R code to reproduce it.

Try:
Your numbers
%
Settings
95% confidence

±3.1 points

Margin
-
Lower
-
Upper
-
z
-
How it is computed
1
2
3
The same thing in R

  
No data leaves your browser Verified against R's qnorm(), qt() and t.test() Free

Which margin, and which formula

The three modes answer three different questions. Here is what each one uses and when to reach for it.

ModeAnswersFormulaUse it when
Percentage marginHow far off could my % be?z·√(p(1−p)/n)You have a poll or yes/no survey result and its sample size.
Required sample sizeHow many people do I need?z²p(1−p)/m²You are planning a survey to a chosen margin (Cochran).
Average marginHow far off could my mean be?t·s/√nYour outcome is a number (spend, time, score), not a percentage.

The percentage margin uses the normal approximation, the same symmetric band pollsters quote as “plus or minus.” The average margin uses the Student t critical value with n−1 degrees of freedom because the standard deviation is itself estimated from the sample; switch to z only when the population SD is genuinely known. A finite-population correction, √((N−n)/(N−1)), kicks in when you supply a population size and your sample is a sizeable slice of it.

Questions people ask

What is the margin of error in a survey?

It is how far your result could reasonably sit from the true figure in the whole population, purely from the luck of who landed in your sample. A result of 52% with a 3-point margin means the real value is very likely between 49% and 55%. It gets smaller as the sample grows, following z·√(p(1−p)/n).

What sample size do I need for a 3% margin of error?

At 95% confidence and the cautious 50% assumption, a 3-percentage-point margin needs about 1,068 respondents. Use the Required sample size mode, enter a 3% target, and the page rounds the Cochran formula up to a whole number of people. A tighter 2% margin needs about 2,401.

Why do these calculations often use 50 percent?

The term p(1−p) is largest at 50%, so 50% gives the widest, most cautious margin. Pollsters quote that worst case so the stated margin holds however the answers split. If you already have a result far from 50%, enter it and the margin drops a little.

Does a bigger population need a bigger sample?

Barely. Past a few thousand people the margin depends on the sample size, not the population, which is why a 1,000-person sample serves a city or a whole country equally well. It only matters when your sample is a large fraction of a small group; enter a population size and the finite-population correction shrinks the margin.

Does this calculator match R?

Yes. The percentage margin uses the same qnorm construction as R, the average margin reproduces the half-width of t.test() to the decimal, and the sample-size mode matches the Cochran formula. Every mode was checked against R 4.6.0.

Keep going

Tools that pick up where the margin leaves off.