Monday, February 04, 2008

13,417 Reasons to Actually Read What You Wrote

All over the web, and in my inbox, is a message which says that there were 13,417 military deaths in the Clinton administration, and “only” 9,016 deaths during the George W Bush administration. The email then goes on to claim that this is a media coverup of the greatest proportion. (I'm not going to repeat, or directly link to, the email here, look it up yourself and follow the comments, they are instructive.)

Helpfully, the unknown author then provides a link to the document providing the original data: (a PDF file)

Careful reading of which shows that the numbers said author used are wrong, and his/her conclusion bogus.

I could go on to say that this shows that some people see only what they want to see, and don't even bother to look at what is in front of their noses — but that would be wrong, and anyway I've already done it in email.

Instead I present a graph of a couple of columns from the above reference: the total number of deaths among active duty military from 1980 to 2006, and the total number of deaths due to hostile action. The military is about 50% smaller now than it was when Reagan was President, so we'll look at the rates. The numbers are normalized to 10,000 “FTE” military personnel, where the Full Time Equivalent includes Reserve and Guard troops on active duty. That is, “Total” is the number of deaths per 10,000 military in a given year, and “Hostile” is the number of deaths per 10,000 military due to hostile action.

Obviously, if you were in the military in the Clinton years you had a much better chance of surviving than during the rest of the survey.

The military is a dangerous profession. Let's not trivialize the danger by trying to make cheap political points with bad data.

And read the report before you comment.

Note: since the paper gives data by year, the graph above assumes that a President takes office on January 1, 1980 + 4*n (n = 0, 1, 2, 3, 4, 5, 6), not January 20 of the same year. That's an unavoidable error at the level of this study, but I doubt it makes much difference in the final results. If you have the number of casualties those days, let me know and I'll redo the graph.