However, the scientific criticisms of his methodology hardly look damning (according to wikipedia). It doesn't look perfect, but it doesn't look terrible either.

The criticisms wikipedia lists: he used cocaine prices rather than cocaine consumption as a variable, if you remove smaller counties +florida the drop in murders becomes statistically insignificant (drops in assault and robbery are still significant), and a couple of critics believe the model has bias and coding errors. As I said, I haven't read it in detail, so I won't comment on how significant these are.