Finite Population Correction Factor
Why a finite population makes a sample mean less variable
The usual standard-error formula assumes independent draws. A simple random sample drawn without replacement from a finite population has negatively related draws when more than one unit is sampled: once a unit is selected, it cannot be selected again.
That dependence reduces the sampling variability of the sample mean or proportion for samples of at least two units; a single draw has an FPC of one.
For a population of size N and sample size n, multiply the with-replacement standard error by sqrt((N - n)/(N - 1)). This finite population correction is below one when 1 < n < N, equals one when n = 1, and is zero for a census.
The correction changes precision, not bias, representativeness, or the sampling design. It belongs only when the target population is finite and the sampling is without replacement; it is not justified merely because a sample happens to have a known size.
Which sampling design most directly justifies applying a finite population correction to a standard error?
A with-replacement standard error is 12 units. A simple random sample without replacement has N = 401 and n = 101. What is the corrected standard error, rounded to two decimals?