Currently, the cluster tract csvs ID cities by cbsa even irrespective of geography type (especially problematic because we are calculating this on max_cities). We should probably change this to reflect the ID used in the dual graph names. This doesn't have to lead to bugs but its confusing and it makes it easy to treat auto clusters as being in the metro area not the actual geography they are calculated for.
Currently, the cluster tract csvs ID cities by cbsa even irrespective of geography type (especially problematic because we are calculating this on max_cities). We should probably change this to reflect the ID used in the dual graph names. This doesn't have to lead to bugs but its confusing and it makes it easy to treat auto clusters as being in the metro area not the actual geography they are calculated for.