Recipe

Apply the finite population correction to a standard error

Use when: you're computing a standard error for sampling WITHOUT replacement from a finite population of known size $N$

Goal: the corrected standard error, or a justified decision that the correction is small enough to skip

Taught in The finite population correction

Quick recipe

Compute the uncorrected standard error
Decide whether the correction is worth applying
Compute the correction factor
Multiply to get the corrected standard error

1
Start from the ordinary $SE=\sigma/\sqrt n$ formula, using the population standard deviation $\sigma$ and the sample size $n$ , exactly as if sampling with replacement.

Example

$N=500$ , $n=40$ , $\sigma=15$ : $SE_{\text{unc}} = 15/\sqrt{40} \approx 2.372$ .

Why: the correction factor in Step 3 only ever shrinks this baseline number; it isn't a substitute for computing it first.
2
Compare the sample size $n$ to the population size $N$ .

Example

$n/N=40/500=0.08$ , 8%, not small: apply the correction. For $n/N=40/50{,}000=0.0008$ , the correction would be safe to skip.
- $n$ is a small fraction of $N$ (a common rule of thumb: $n/N$ under about 5%) — the correction factor would sit extremely close to $1$ , so it may be safely skipped and Step 1's result used directly.
- $n$ is a substantial fraction of $N$ ( $n/N$ isn't small) — continue to Step 3 and actually apply the correction.
Why: the correction factor $\sqrt{(N-n)/(N-1)}$ approaches $1$ as $N$ grows large relative to $n$ , so computing it exactly only matters once the sample takes a nontrivial slice of the population.
3
Compute $\sqrt{(N-n)/(N-1)}$ , using the same $N$ and $n$ from Step 1.

Example

$N=500$ , $n=40$ : $(500-40)/(500-1)=460/499 \approx 0.9218$ , and $\sqrt{0.9218} \approx 0.9601$ .

Why: this factor is always at most $1$ , since $N-n \le N-1$ whenever $n \ge 1$ , which is exactly why without-replacement sampling never has MORE spread than with-replacement sampling of the same size.
4
Multiply the Step 1 uncorrected $SE$ by the Step 3 correction factor.

Example

$SE = 2.372 \times 0.9601 \approx 2.277$ .

Why: this product is the definition of the corrected $SE$ , the standard deviation of the sampling distribution once sampling without replacement from a finite population is accounted for.

Common mistakes

Applying the correction factor to $\sigma$ directly, instead of to the already-computed $SE$ from Step 1.
Using $N$ in place of $N-1$ in the correction factor's denominator.
Skipping the correction reflexively whenever $N$ "looks big," without actually checking the $n/N$ ratio from Step 2.
Ending up with a "corrected" $SE$ larger than the uncorrected one, a sign the correction factor was computed or applied wrong, since it can never exceed $1$ .