Population vs. Sample Standard Deviation Explained
Why dividing by n-1 instead of n changes the result, when each formula applies, and a verified side-by-side calculation showing exactly how much the difference matters.
Published July 20, 2026
The exact same dataset produces two different standard deviation values depending on one choice: whether you’re treating the data as the entire population you care about, or as a sample drawn from a larger population you’re trying to draw conclusions about. That choice changes a single number in the formula — but it changes the result every time.
The one formula difference
Population: σ = √(∑(x − μ)² ÷ N)
Sample: s = √(∑(x − x̄)² ÷ (n − 1))
Every other part of the calculation — find the mean, subtract it from each value, square the deviations, sum them — is identical between the two formulas. The only difference is the denominator in the final averaging step: N (the full count) for population, or n−1 (one less than the count) for sample.
Why the sample formula divides by one less
This adjustment — called Bessel’s correction — exists because a sample’s own mean is calculated from the same limited data being measured, which makes the sample’s deviations systematically a bit smaller than the true population’s deviations would be. Dividing by n−1 instead of n compensates for that bias, producing a less biased estimate of the true population standard deviation.
A verified side-by-side comparison
Using the dataset 2, 4, 4, 4, 5, 5, 7, 9 (sum of squared deviations = 32): the population formula gives 32/8 = 4, so σ = 2.000. The sample formula gives 32/7 ≈ 4.571, so s ≈ 2.138 — about 7% higher on this particular dataset. The gap between the two formulas shrinks as the sample size grows, since the difference between dividing by n and n−1 becomes proportionally smaller.
The gap between population and sample standard deviation is largest for small datasets and becomes nearly negligible for large ones — dividing by 99 instead of 100 barely changes the result, while dividing by 4 instead of 5 changes it substantially. This is exactly why the choice matters most, and is easiest to get wrong, with small sample sizes.
Which one to use
The dataset genuinely represents every member of the group you care about — every student in a specific class, every transaction in a specific month.
The dataset is a subset drawn from a larger group, and you're trying to draw conclusions about that larger group — a survey sample, an experimental trial group.
| Scenario | Formula |
|---|---|
| Every test score in one specific class | Population |
| A survey of 500 people meant to represent a country | Sample |
| Every transaction a business made last quarter | Population |
| A clinical trial’s participants, generalized to a wider population | Sample |
FAQ
What happens if I use the wrong formula? The population formula on sample data slightly underestimates the true variability; the sample formula on full population data slightly overestimates it — neither is a large error for big datasets, but it’s a meaningful, avoidable inaccuracy for small ones.
Is there a rule of thumb when it’s unclear which applies? When in doubt, and especially with smaller datasets, the sample formula is the more conservative, more commonly appropriate default in research and statistics contexts, since most real-world data collection is sampling from a larger population rather than capturing it entirely.
Does this same distinction apply to variance, not just standard deviation? Yes — variance is standard deviation squared, and the identical population-vs-sample denominator choice (N vs. n−1) applies directly to the variance calculation itself, before the square root is taken.
How does the Standard Deviation Calculator handle this choice? It includes an explicit “Population” vs. “Sample” toggle, applying the correct denominator automatically once selected — the underlying steps otherwise match exactly what’s shown in How to Calculate Standard Deviation Step by Step.
Why is it called “Bessel’s correction”? It’s named for Friedrich Bessel, a 19th-century mathematician and astronomer credited with the statistical adjustment, though the general idea of correcting for sample bias has been refined by many statisticians since.
Does a larger sample size ever make the choice of formula irrelevant? Not entirely irrelevant, but the practical difference shrinks substantially — for a sample of several thousand, dividing by n versus n−1 changes the result by a negligible amount compared to a sample of 5 or 10.