If we have a look at the
inductive dimension (that already can be defined on any topological space), we see that it is simply useful to assume the dimension of an empty set to be -1.
Roughly because the empty set is the boundary (boundary can be defined in any topological space) of every point, and every point should have dimension 0. And the dimension of the boundary (of subsets) should be (at least) one less than the dimension of the regarded set.