"We can try to calculate how much information this intensity of selection can provide about institutions. (The following calculations are very rough, but should get the right orders of magnitude; cf. Rivoire and Leibler, 2010.) Take the high estimate of a 33 percent chance of destruction over 500 years. This works out to an annual risk of polis death of about 0.2 percent. This is not nothing, but it's not very much either. If we applied selection like this to a population starting with a standard Gaussian distribution, removing the lowest 0.2 percent of the population each year, by the end of 500 years we'd have increased the mean by about a standard deviation. Said another way, the entropy rate of selection is about 0.02 bits/year, so, accumulated over the whole of the archaic and classical periods, we get about 11 bits of information on which constellation of institutions works best. This is the maximum amount of information which could be extracted from the "signal" of city destruction (and yes, that does bring to mind the one good passage in Thomas Pynchon), and it would be reduced by correlations over time, etc.
"I think that selection is actually vital to the development of institutions. It's just that the selection which matters is not the rare destruction (or duplication) of whole polities or societies, but rather the continual choice by individuals about whether to conform to institutionalized expectations about their behavior, or to do something else. Renan said that a nation is a daily plebiscite; in a real sense, this is true of every human institution. They survive if, and only if, they are continually reproduced in human practices. This is where selection gets its purchase on institutions."
--Cosma Shalizi, "Review of Josiah Ober, Democracy and Knowledge: Innovation and Learning in Classical Athens"