Geographers are well aware of the Modifiable Areal Unit Problem (MAUP). Basically, MAUP causes the results to change when you change the unit of aggregation. This is especially true of things like rates.
As I've written elsewhere, what unit of geography you use for urban analysis matters. Anytime anyone uses city boundaries, be suspicious. By cherry-picking cities (either within the nation, or within a metropolitan area), you can generate all sorts of spurious correlations, and really bias what your results are.