The standard error tells you how precisely a sample estimates the thing you care about. If you drew many samples from the same population, their means would vary from sample to sample, and the standard error is the standard deviation of that variation. It is the basis for confidence intervals and for significance tests. This calculator gives the standard error of a mean and of a proportion, with the margin of error at 95% confidence.
How the calculation works
- Standard error of the mean = standard deviation ÷ √n
- Standard error of a proportion = √(p × (1 − p) ÷ n)
- 95% margin of error = 1.96 × standard error
- Finite population correction = √((N − n) ÷ (N − 1)), applied when you give a population size
The proportion is shown in percentage points, so 4.9 means 4.9 points on a percentage scale.
Example
A sample of 100 with a standard deviation of 15 gives a standard error of the mean of 15 ÷ 10 = 1.5, and a 95% margin of error of about ±2.94. For a proportion of 40% in the same sample, the standard error is √(0.4 × 0.6 ÷ 100) = 4.9 points, with a margin of about ±9.6 points. To halve these errors, you would need 400 observations.
Why sample size matters
The standard error shrinks with the square root of the sample size. Going from 100 to 400 people halves the error, but going from 400 to 1,600 is needed to halve it again. That is why very large samples bring diminishing returns, and why a survey of about 1,000 people is often a good compromise.
Standard error or standard deviation?
Report the standard deviation to describe how variable individuals are, and the standard error (or a confidence interval) to describe how precise your estimate of the mean is. Do not use one in place of the other. The standard error accounts for random sampling variation only; it does not capture bias from a poorly chosen sample or from measurement errors.