Howard Winter wrote: >>>>> An excerpt from the 2006 CIA factbook: >>>>> >>>>> Rank Country Death rate (deaths/1,000 population) >>>>> 107 United States 8.26 >>>>> .......... >>>>> 166 Israel 6.18 >>>>> .......... >>>>> 219 Gaza Strip 3.80 >>>>> ........... >>>>> 226 Northern Mariana Islands 2.29 >> >>>> That is total death rate. [...] Where are the figures on VIOLENT death? >>> >>> I think those must be them, because if they are the total death rates >>> then life expectancy is a bit overestimated: 1000/2.29 means that >>> (assuming they are averages per year), a Northern Mariana Islander has a >>> life expectancy of over 436, while a USAmerican can look forward to >>> "only" living to 121. >> >> I'm not sure how you come to these figures, but they seem to assume some >> things around birth rate, total population number stability and age >> distribution that are not quite in alignment with reality. I'm not sure >> whether it's that simple -- but I'm not really a statistics expert. > > OK, my use of the phrase "life expectancy" is probably wrong, what I was > calculating is how long it takes for a batch of 1000 people to die, and > given just the death rates (which is all we had in the quote above) this > can be calculated given that the total population doesn't change very > much. > >> For example, the CIA gives for the USA: birth rate 14/1000, death rate >> 8/1000, net migration 3/1000, pop. growth 0.9%/y, life expectancy 78y. The >> Northern Mariana Islands numbers in the same sequence: 19/1000, 2/1000, >> 8/1000, 2.5%, 76y. I'm not sure where your 121y and 436y fit in here. > > Can you point to the URL for that data? CIA World Factbook: https://www.cia.gov/cia/publications/factbook/geos/us.html https://www.cia.gov/cia/publications/factbook/geos/cq.html That's a handy link to keep around. A nice summary of country data for a quick fact check. > What I'm saying is that if net migration and birth rate keep the total > population about the same, if the all-causes death rate is 8/1000 and > stays there it will take 125 years for a batch of 1000 people to die, > from simple arithmetic - how can it be less? I've learned one thing about statistics: it's (almost) never simple arithmetic :) First of all, you need to nail down the assumptions (which may be more than you think), and then you need to check whether the difference between assumptions and reality doesn't make a significant difference. If so, you need to improve your model, ... (and here we go again :) Probably the main assumptions to make your calculation work are population stability, flat age distribution and no migration (or population changes equal to migration rate) -- and all that stable over a period longer than the calculated life expectancy. Not sure whether there are others, but just these assumptions are rarely given. > This is barely believeable, but for the Mariana Islanders it would 500 > years at 2 per year, which is clearly nonsense unless people are > emigrating to somewhere else to die, or the death rate is going to make > a dramatic upswing at some point. If there is significant population > growth, my logic may be defeated since older people are a smaller part > of the population than they were when they were younger, so their > contribution to the rate/1000 falls as they get older. A quick > calculation shows that 0.9% annual growth takes 77 years to double the > population, so that may be the flaw in my argument! See... that's one assumption you made (stable population). Just use a different unit for population growth: instead of 100 as reference (as in %), use 1000 as reference, like the other rates. 0.9% (US) becomes 9/1000, and 2.5% (the islands) becomes 25/1000 -- numbers that look pretty significant compared to the birth and death rates, too significant to be just assumed away. >> I also don't know whether you've seen the various messages where I posted >> FBI (and government) numbers of violent death rates. They are much smaller. >> >> You can of course doubt both the CIA and the FBI -- but in this case, the >> numbers look rather realistic :) > > I followed a link posted earlier to an FBI site that gave figures for > cities (not aggregated to States or the whole of the USA) and I seem to > remember that deaths due to murder and non-accidental manslaughter were > in the region of 200 to 700 per 100,000 population (so 2 to 7 per > thousand). I posted this link earlier: http://www.fbi.gov/ucr/cius_04/offenses_reported/offense_tabulations/table_06.html You must have gotten some numbers crossed; it's easy in that table as it's not very well organized (IMO). In any case, overall in the US this rate is around 5/100k (from other sources), and in some bad metro areas it reaches 50/100k. I don't think you'll find numbers as high as the ones you're citing in that page. (Cross the rows that say "Rate per 100,000 inhabitants" with the column "Murder and nonnegligent manslaughter".) You'll find numbers mostly below 10. In order to get to the really high rates, you need to pick individual bad areas (like Washington, D.C. or New Orleans) and calculate the rate by dividing the numbers of deaths by the population; the rates for these areas are not shown in the table, and the numbers you see in that column are numbers of deaths. A bit confusing... Gerhard -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist