Wednesday, December 08, 2004
Blue States, Red Blood
For some reason I started thinking about where in the U.S. casualties were from, to get a sense of the impact of the war.
Where would you think the casualties are from? It depends on how you would code the person's "residence," right? Your map would look different if you recorded the address for where the person was posted within the U.S. versus where the person was recruited from, right?
Let's assume the address they use is where the person enlisted from. Where would you expect more pins on the map?
This site plots the "home of record" for US casualties thru 11/15/2004 (I'm checking on how this is defined). It turns out the casualties seem pretty heavily clustered in blue states, with heavy clustering in the east from MD, NJ, PA, and NY and out west in LA and SD.
If the "home of record" is where the person's regiment is headquartered, well this is a map that shows how America's military bases are distributed across the country. My guess is that it's probably the person's "hometown" before they joined the military since so many dots appear where there are no bases (though these could be reservists).
The other thing to keep in mind is that these areas have higher population densities, so it's natural to expect them to have greater numbers of "dots" than elsewhere. And the states are smaller, so the dots look proportionately bigger.
It's tempting to conclude that it's the blue state folks whose kids are dying in a war supported by the red states. And that seems to be the case in terms of aggregate numbers of casualties.
But, their main site has more details that slice-and-dice the data other ways. I combined the deaths by state with Census figures for population as of 7/1/2003. These are the casualty rates per 1,000,000 population:
A somewhat different picture here, as NY and NJ move down to the bottom 20% (lowest 10 states). CA and PA hang in the middle of the pack. It's probably not surprising to see the smallest population states at the top of the list because with such small aggregate populations, it doesn't take too many casualties to get a high score: Vermont, for example, has 10 casualties, yet its rate is 16 per 1m; Wyoming has 5, with about 10 per 1m. And, it's probably not too surprising to have the reverse happen: Alaska has 1 casualty, so its rate is very low even with a small aggregate population.
The lesson isn't that statistics lie and we should chuck the whole thing. The lesson is to know what question you're asking, what your data represent, and whether the questions match the data you have.
Forgive me, but I'm practicing being professorial as I plan to teach an undergraduate research methods class next semester. This struck me as a useful pedagogical tool.
Anybody who can help me figure out how to get rid of the huge space between the text and the table in Blogger would be my hero. (I don't see any HTML code that is forcing the extra space to be added.)
Updates: Jed is my hero. He figured out that Blogger was inserting a BR command with each return in the table. That is kind of obnoxious, but at least it's fixable by running all the rows and cells together in one long string.
Also, I heard back from the casualty mappers: the "home of record" is the place where the person enlisted.
Comments: Post a Comment