Clusters make a nice story, but you're going about it backwards. What are you trying to predict? The scientific method suggests we create a hypothesis first,

Clusters make a nice story, but you're going about it backwards. What are you trying to predict? The scientific method suggests we create a hypothesis first, then design an experiment (or identify a natural experiment), then collect data, and finally perform a hypothesis test. Deviations from that process increase your risk of spurious results.