Does that matter? My understanding is that the groups are a random sample so it should be unbiased estimator of the true pop