The general observation that, as population densities exist, government types are more insulated from the people and are less democratic or responsive seems to be a rough generality in the real world
I don't think that's the case. Some of the most densely-populated places on the planet include Tokyo, Delhi, Mexico City, Sao Paulo, Mumbai, Osaka-Kyoto and New York, none of which are particularly "undemocratic."
and in our game. The dice conventions ensure variations exist. If you use extended legal, government, etc codes, you can get variances by what sections of government or legal codes as far as the rating goes.
There is little if anything in the world generation rules which connect population ~density~ with governmental interference (which is better measured by Law Level rather than Government type). Population is generated without any modifiers for spatial area; a Size 1 world and Size A world are equally likely to have Pop A (and no matter an Atmosphere or Hydrography which might further restrict the "livable space").
The primary relevant substance baked into the world generation rules is that higher Pop worlds will tend to have higher Law Levels. Sure, it may be challenging to open-carry a firearm--or make non-permitted renovations to your house--in a high Pop jurisdiction but few people living in Tokyo or Delhi or Mexico City or Sao Paulo or New York would tell you, in a unguarded moment, that they're living under an authoritarian regime.
In other words, this particular game mechanic seems to be largely divorced from real-world empirics. If anything, it reflects 1970s-era U.S. Midwestern suburban / rural (mis)perceptions of "the big city" (with a bit of similarly Midwestern Cold War era view of the rest of the world thrown in).
And in any case, if things get bad enough, we'll have government type "Matrix" where we're all just the batteries in a power system for our overlord AI masters....
Yes, but before we get there we'll have the ~Wall-E~ highly-immersive distraction, grav-chair-and-Slurpee government type. . . .