the new 2004 national crime victimization survey numbers were released this week. according to this large-scale nationally representative survey, violent crime is holding steady at the lowest levels recorded in the three decades for which we have had reliable data. that’s the good news, but there is also some disturbing news regarding race differences.

the ncvs asks respondents to self-identify race, using categories for white, black, other “only,” and, since 2003, “two or more races” (“other race” includes american indians/alaska natives, asians, and native hawaiians/other pacific islanders identifying a single racial background). when i first presented last year’s numbers, i was depressed to find extraordinarily high rates of violent victimization among persons of “two or more races:” 67.7 victims of violent crime for every 1,000 people age 12 and older, relative to rates of 29 for blacks, 21.5 for whites, and 16 for those of other races. i’ve always lectured on racial disparities in victimization, but had never seen gaps like this before. this is not my primary area of study, so i may be missing something obvious. still, any good criminologist who does survey work should be able to address such questions.

i first thought it might be an artifact, given that it was the first year that the “two or more races” option was presented, but the same ugly pattern holds for the 2004 numbers (click the graph for a legible version). those reporting “two or more races” were victimized in 2004 at 51.6 per thousand, double the black rate of 26 per thousand, and far exceeding the white and “other” rates of 21 per thousand and 12.7 per thousand, respectively. only about .9% of respondents self-identify as “two or more races,” but we’re talking about over 1,000 people here, given the NCVS sample size of 149,000. those of hispanic ethnicity are somewhat less likely to be victimized than non-hispanics, but the main difference seems to be between “2 or more” (shown in the red bar) and all of the other bars shown.

several press accounts have reported these differences, but to my knowledge none have printed comments from victimization experts about them. so, my hope-i’m-wrong disturbing hypothesis is that a large portion of the greater violent victimization is due to discrimination. is it the case that multiracial persons are more likely to be victimized everywhere — in black, white, hispanic, asian, and american indian communities alike? i don’t know the census data well enough to comment on response patterns to such questions or even the socioeconomic differences between multiracial families and others. perhaps some of the victimization differences are due to ses or neighborhood poverty or family structure (all of which are no doubt linked to discrimination in some form). but am i missing something else here?

such a large bivariate gap cries out for serious multivariate analysis. does the pattern hold for property victimization as well? what happens when controls for neighborhood and ses are introduced? if we still see disproportionate violence against multiracial persons, it also cries out for targeted policy interventions to address this problem.