I was speaking from memory there, and what I had in mind was the 2008 NCVS figures on overall interracial violence, which doesn't include homicides. NCVS data is compiled from victim reports. Homicide victims don't report too well.
And I should have clarified the precise meaning of "more frequent."
According to the Department of Justice, in 2013 (the last year I can find data for) there were 409 homicides featuring a black offender and a white victim, versus 189 featuring a white offender and a black victim. The relevant DoJ table is here. ("White" includes Hispanics for this data.)
If you divide 409 by 189 you get 2.164, so homicide-for-homicide, a true statement would be:
Black killing of whites was more than twice as frequent as white killing of blacks in 2013.That ignores population disparities, though. According to Barack Obama's DoJ, each million of the nation's 280 million whites murdered an average 0.68 blacks in 2013 (189 divided by 280), while each million of the nation's 40 million blacks murdered an average 10 whites (409 divided by 40).
If you divide 10 by 0.68 you get 14.7 So here's another true statement:
While homicide is a very rare event and we are dealing with tiny probabilities here, any given black was almost fifteen times more likely to have killed a white in 2013 (probability 0.001000 percent) than any given white was to have killed a black (probability 0.000068 percent).