I don't think corruption can be measured by arrests, as the mechanisms to arrest the corrupt are themselves corruptible. Nor do I think that per-capita is the

I don't think corruption can be measured by arrests, as the mechanisms to arrest the corrupt are themselves corruptible. Nor do I think that per-capita is the correct way to compare jurisdictions of dissimilar sizes. With corruption, it is not the number of residents that matters, but the organization of the independently corruptible units of government, and the amounts of money that can be quietly sucked out of them. The desired metric--that we want to approximate from the more easily obtainable data--is corruption profits per corruptible individual. Each corruptible individual has an incentive to move to places where that metric is higher, thus performing sleazebag arbitrage between those areas where corruption is endemic. Are the examples you cited outliers, with explainable reasons for poorly fitting the trend? Could it be that they are all territories, where crimes of corruption by officials at lower levels in the administrative hierarchy would be investigated by the federal Dept. of Justice from the start, without any states' attorneys-general getting a crack at them first or asking the feds for assistance or joint prosecution? If you separate out DC and the island territories, does the remainder of the set hold up?