On the Interpretation of Censuses as Samples