Sex Crime Wave Involving East Asians and American Indians?

I blogged about this back in 2002, but I thought I'd mention it again; it's one of my favorite simple statistics puzzles. The federal Sentencing Commission provides a table breaking down all sentenced federal offenders, broken down by offense type and race. The race categories are White, Black, Hispanic, and Other. (I realize that demographers generally view Hispanic as an ethnicity, not a race, but the Commission follows the more common lay usage here.)

For sexual assault, 51.3% of all offenders are "other." Among the U.S. public as a whole, probably only about 6% would fall into the "other" category (subject to some uncertainty that's not relevant here, since any estimate would be far below 51.3%). What's up? I'll post the answer to the comments in a few days, if no-one else gets it before then (though I suspect they will).

John Jenkins (mail):
Sexual assault on Indian Reservations would be prosecuted in federal court, while sexual assault cases not on Indian Reservations will be prosecuted in state court. I would guess the rest of the federal offenders are largely military personnel and residents of the District of Columbia.

Since American Indians make up the virtually exclusive residents of Indian Reservations (and East Asians would include residents of places like Guam, which I believe is a federal protectorate and therefore has a district court) it's only natural that they be the perpetrators of a majority of sexual assault cases prosecuted in federal court.
1.13.2006 5:30pm
DNL (mail):
"Other" would include:
Native American
1.13.2006 5:42pm
Thales (mail) (www):
There's an article about some of the strange and possibly prejudicial evidence case law this phenomenon creates called something like Lustful Indians.
1.13.2006 5:47pm
MikeC&F (mail):
John Jenkins has it. Proof that while *almost* everything is a federal crime, the feds still leave some criminal prosecutions to the states.
1.13.2006 5:49pm
countertop (mail):
I think DNL has it - other includes unkown.
1.13.2006 6:32pm
byomtov (mail):
I'm with Jenkins.
1.13.2006 6:36pm
Makes sense to me. I didn't think about it too much before looking at the comments, but I don't think I would have figured it out on my own.

1.13.2006 6:42pm
Goober (mail):
Huh. Woulda thought that sexual abuse would be the most likely crime to be reported but not investigated to prosecution (I'd assume because the rapist / abuser was never successfully identified, whereas murder is much more likely to be cleared), and hence race couldn't be conclusively reported. I.e., what DNL says.

Supportive of this are the comparatively high "other" rates for assault and burglary / breaking-and-entering, which also presumably would remain uncleared at high rates. I'm not sure why manslaughter would have such high "other" rates, too, although the very small absolute numbers (5, 4, 5 in the racially identified category, 34 in the "other" category) is I suppose consonant with the unsolved-crimes hypothesis.

But John Jenkins makes an interesting case, too. Eh, I'll wait till Professor V gives the answer. (And if his response is "This might shock you but you're -both- right," I'm leaving....)
1.13.2006 7:22pm
Eugene Volokh (www):
OK, since John Jenkins figured it out so quickly, might as well confirm it -- he's right.
1.13.2006 7:35pm
Other can't include unknown. This is a table of sentenced federal offenders.
1.13.2006 7:38pm
Beerslurpy (mail) (www):
John Jenkins beat me to it. That is a federal court sentencing commission, so it wouldnt deal with the 99.999 percent of rapes and murders that are dealt with by state governments. I would bet you that "other" consists almost entirely of american indians or residents of various federal possessions.

You can see which crimes typically fall under federal jurisdiction from the numbers, such as trafficing of illegals and narcotics. I also note that federal firearms prosecutions are high as well, probably as part of that ashcroft program whose name slips my mind.
1.13.2006 7:38pm
John Jenkins (mail):
Exile is the program you're thinking of, I think.
1.13.2006 8:01pm
David Matthews (mail):
I've used this one before in my Introductory Statistics class. Thanks for the reminder; I'll use it again!
1.14.2006 4:06pm