As I've said my personal ignorant opinion is "graph spread in absolute terms, graph burden in per capita terms". So it depends really on the context IMO. https://twitter.com/CT_Bergstrom/status/1249480343968575488

Like 10,000 deaths in Scarsdale and 10,000 deaths in Tokyo, those are very different burdens. 100 cases in Scarsdale and 100 cases in Tokyo, those might both be 300 cases in a week or whatever the rate is.

Because really, both are proxies. When you present deaths per capita what you're trying to express, I think, is something like "what percentage of people are grieving, what percentage of businesses lost an employee, what percentage of family members lost a relative", etc.

(Or even more, what percentage of people are mourning multiple friends and family members, what percentage of businesses, or churches or nonprofits or whatever, are inoperably devastated, etc.)

Which is why "per capita" is a crude adjustment there, depending on how the deaths are themselves clustered and distributed etc. IDK I'm sure this isn't revelatory, I'm thinking out loud.

Because if those 10,000 Tokyo deaths all happened in one neighborhood with the population of Scarsdale, well, then we're back to the same thing again.