The recent talk about hate crimes reminded me of some data I looked at a while ago, but never got around to blogging. I thought, then, that I'd bring it up now, especially since my sense is that it hasn't gotten much coverage elsewhere.
Let's begin by briefly discussing The Nation's two crime measures, the UCR and the NCVS. The UCR (Uniform Crime Reports) reports on crimes that are reported to, and then reported by, the police. The NCVS reports on what is said by people surveyed through the National Crime Victimization Survey; it includes crimes that aren't reported to the police, and aspects of crimes that aren't reported to or by the police.
Each has possible flaws: The UCR doesn't cover unreported crimes, and in some ways reflects police classification decisions more than anything else. The NCVS is more prone to error as a result of respondent mistakes or falsifications, since no-one investigates the respondents' claims. To give you a sense of the difference between the two, note that the UCR Hate Crime Statistics 2005 reports 7163 hate crime incidents in 2005; the NCVS Hate Crime Reported by Victims and Police (2000-2003 data) estimates "[a]n annual average of 210,000 hate crime victimization incidents." Figure out for yourself which you think is most plausible; my sense is that the NCVS tends to be more reliable (except as to homicide, which the UCR measures pretty well, and which the NCVS for obvious reasons doesn't measure at all).
In any case, what does the UCR tell us about 2005 race-based hate crime incidents? Of the 3919 such incidents, 21% were anti-white and 67% were anti-black. Keep in mind that 75% of the population is white and 12% is black (numbers subject to change slightly depending on how one allots the "other" category and multiracial categories; Hispanics are not classified as a separate race). The UCR data suggests that the victimization rate for race-based hate crimes is 20 times higher for blacks than for whites.
What does the NCVS tell us about 2000-2003 hate crime data? It reports that 56% of all hate crime victims, according to data reported by respondents, were victims of race-based attacks. It unfortunately doesn't break down the race just of those offenders who report race-based incidents, but it reports that the total annual victimization rate for all hate incidents per 1000 people is 0.9 for whites and 0.7 for blacks; 85% of all victims were white, 9% were black.
Thus, assume all the non-race-based attacks were against whites, and all the attacks against blacks were race-based — just to take the scenario that would produce the highest ratio of black/white victimization by race-based hate-crimes. Then, out of 1000 incidents in which there were 850 white victims and 90 black victims, 410 (850 minus the 440 non-race-based) would be race-based attacks on whites, and 90 would be race-based attacks on blacks. This means the maximum victimization rate ratio for race-based hate crimes would that the rate is a bit under 1.4 times higher for blacks than for whites.
I also asked Joe Doherty at the UCLA law school's Empirical Research Group to analyze the NCVS raw data for race-based incidents only, for 1992-2003. The results — which let us avoid having to make any assumptions about how the non-race-based incidents shake out by victim race — were that whites accounted for 64% of the victimizations, blacks for 11%, and Hispanics for 15.5%. Joe's analysis broke Hispanics out into a separate category, but setting them aside means that the victimization rate for race-based hate crimes would be about the same for blacks as for whites.
Which is the right ratio? For that matter, which is the right estimate of total hate crime incidents — 7000 (UCR), 7000 * 7 (to roughly for agencies that didn't submit any incident reports), or 210,000 (NCVS)? You be the judge. My tendency is to assume that the NCVS is generally sounder here as it is elsewhere, though there's no doubt that there are threats to validity in the NCVS as well as the UCR. But even if this is uncertain, just keep in mind the dangers of relying exclusively on UCR-based accounts.