The fact that it’s a smaller territory means that effects that would even out in larger territories don’t apply. I.e. it’s a relative effect.

Here’s an example: suppose a large retirement community of Chinese immigrants exists in Hong Kong, raising the average age. In order for this kind of retirement community to shift the age distribution the same amount in a country 10x larger, that retirement community would have to be 10x or so larger as well. A 10x larger retirement community is way less likely.

This isn’t to say that all small things lie on the extremes of the distribution, only that an extreme value is likelier to come from a smaller sample.

> That seems like a robust dataset to me.

Perhaps, but there's a wide distribution of age of death, and the author is fretting about a 5 year difference in life expectancy between 2 populations that are doing fine.

Assuming someone can ever fully explain the cause of this life-expectancy difference, it's going to be multi-faceted and complex and amount to nothing more than a long list of generic advice and policy prescriptions that will go nowhere and probably be culturally incompatible.

Size isn't the salient factor. Consider two "cities," one of which draws its political borders around its inner suburbs and one that doesn't. They may have identical populations, even, but they'll produce very different results on many common statistical measures.

This is why you should almost never compare two places based on political borders or designations. (You have to instead normalize around something less arbitrary. "City" and "state" aren't proper categories for classification, because what gets included from city to city and state to state can vary across multiple dimensions.)