We know several things:
- We're wrong a lot...indeed, when doing anything new, we should expect single-digit chances of being right.
- Feedback systems win: What system is in place to fix the broken parts of the existing system...so that when it works wrong (as it necessarily will), it stabilizes.
- He who has the scarce resource/skill makes the rules. For much of history, the scarce resource was the effective leverage of violence via unified fighting tactics in a cohesive unit. For a while in the 1500-1700s, this was less true.
- Strong central governments are massively incented to behave badly, as per Bruce Bueno de Mesquita's amazing game theory work. Indeed, one can predict very well the value to the people of the government by asking what proportion of the actual population is required in order to allow the government to stay in power: more is strictly better.
Further support from the thesis comes from the suggestion that the reason that China lost modern history is because of over-centralization. Too uniform of a government, with not enough variation, experimentation, and ability to exit caused a stagnation that Mancur Olson predicts from ANY centrally controlled system.
Indeed, even Jerry Pournelle's Iron Law of Bureaucracy suggests that the only thing that can possibly keep a large system from decay is a threat of failure.
You want a good government? The necessary and sufficient condition is: easy exit for citizens to a neighboring unit, thus ensuring competition for citizens, and governments which serve the citizenry, rather than the opposite. This mostly requires smallness in order to guarantee easy exit. Don't got it? Bend over, this is gonna hurt. Roughly...all other models are guaranteed to fail.
Some further evidence in this direction just surfaced at Dan Mitchell's blog. Go check it out.