Astronomically, you make assumptions about the homogeneity (or lack of it) of a population, or you try to model its variation, and then you sample it, making statistical extrapolations based on those models. After all, no one has ever actually counted all the stars in the galaxy.

The census people probably do the same thing, but if there's power or money involved, the results are bound to be challenged.

You should see how popular my suggestion (to the County) was to reapportion commission district boundaries by purely statistical methods (a standard technique in regional geography).