This working paper proposes two considerations prior to the study of city size and emphasizes their importance in the testing of one of the empirical regularities most widely used by the literature, Zipf’s law. The first refers to the importance of using a consistent database for analysis. The second looks at the unresolved problem of what exactly is meant by city.

These two considerations are taken into account when carrying out a contrast of Zipf’s law, based on the generalization of the Pareto distribution. This contrast permits the rejection of this law for the assumed concept of city in all the years considered.