Average life expectancy for a population is normally given from birth; but it seems to me more useful to distinguish youth mortality (e.g. from birth-defects or malnutrition) from that of adulthood (violence, accidents) and age (heart disease &c).
I’d like to see life expectancy stated in the form of two numbers: the age at which future l.e. stops increasing (which measures childhood mortality; if c.m. is very low I guess the critical age is negative), and the l.e. at that age (which measures what we usually think of as longevity).
This train of thought was prompted by the mention on some blog or other (sorry I’ve now forgotten whose; the background was a pale buff, I think, if that helps) about two brothers who are in business together at ages 100 and 91. The bloguist mused about how it feels to celebrate a centenary birthday and know that one is unlikely to see another. (I commented that a centenarian is more likely to see 101 than he had ever been before!) I’m musing about what it’s like to know somebody for that long. How old can a ‘kid brother’ be, i.e. do people outgrow such hierarchies? (We know that High-Elves don’t, at least not in 2739 years!)
Later: Steven Gallaher shares some pointers and observations. The second derivative starts to look more interesting than the first.