The way I have seen this idea stated in the past (e.g. quadratic cost of all-to-all communication) was that the organization lacing a hierarchical structure would fall apart at quite a small size, maybe somewhere around ~100 people.
If one wants to use it to explain the different outcomes between Switzerland and California, they have to explain why something would work for 8 million people (which is not at all a negligible number) and 40 million. What exactly happens at, say, 20 million boundary that breaks the system?
A government has a different problem than just bureaucracy; it has to aggregate preferences from a much larger group in order to do its job. The aggregation of needs, preferences, and other important information just gets harder as the group to be governed gets bigger. I think it’s faster than quadratic, as well; I’d expect that first, second, and probably third derivatives are all strictly increasing functions of population; for a bureaucracy with no customers/clients outside itself the first derivative is positive and usually the second, but probably not the third.
But that would only push the upper limit on efficient governance downwards, no? So the limit would not be 100 people, but rather 30. Still, the question we are discussing is whether there’s a limit somewhere between 8 million and 40 million, which is like five orders of magnitude difference.
Where is part 3?
Here (i just entered the authors profile and it’s one of the newest posts, directly above part 2)