## The Human Yardstick

Thursday, June 19th, 2014

Sir John Glubb found that many empires lasted roughly 250 years, which raised the question of what caused such an extraordinary similarity in the duration of empires, under such diverse conditions, and such utterly different technological achievements:

One of the very few units of measurement which have not seriously changed since the Assyrians is the human ‘generation’, a period of about twenty-five years. Thus a period of 250 years would represent about ten generations of people. A closer examination of the characteristics of the rise and fall of great nations may emphasise the possible significance of the sequence of generations.

1. Marc Pisco says:

I thought you meant this guy.

2. Marc Pisco says:

…who, I hadn’t known, ended up president of ANSI and ISO, very appropriately.

3. Gwern says:

Eh… His list of empires is short, his table on page 4 excludes several he mentions in his text, like the Babylonian Empire, he’s definitely tilting the numbers the way he wants (what are the odds of 3 out of the 11 empires listed all lasting exactly 250 years when the range is 207–267?*), and later research which looks more comprehensive shows no cliff at 250 years but just an exponential. That dataset suggests a mean somewhat close to 250 (220 years) but the exponential distribution, if true, strongly argues against any kind of “life cycle of empires”.

* Assuming that he had for some reason chosen in advance to list only empires lasting 207 to 267 years and empire-years are distributed uniformly at random, then we can see how many “collisions” we’d expect using a variant on the birthday paradox — in Haskell: ``let n = 11 in let d = (267-207) in n - d + (d*(((d-1)/d)^11))` ~> 0.87`. So we’d expect less than 1 collision (and certainly not 2 collisions on the same year-count), if he weren’t rounding numbers or something like that.

4. Candide III says:

Gwern: +1. It’s like those historical theories that Napoleon and Julius Caesar were actually the same person etc. that seem to surface with some regularity.