Black-On-White Crime: If You Want To Get To Grips With Hatefacts, You Need To Understand The Hatenumbers
Print Friendly and PDF

My August 29th column on the effort by Tim Wise to explain away the great imbalances in inter-racial crime stats brought in a lot of email. I’m going to give over a column to dealing with some of the issues raised by readers.

Inter-racial crime: The context.

It’s hard to overstate the degree to which the entire topic of racial factors in crime is socially radioactive. You could ask Ron Unz, who was purged from the board of The American Conservative, in spite of his having given major financial backing to TAC over several years, for writing a data-rich, highly numerate article about exactly this topic.

The Unz purging inspired a must-read short (700 words) blog post by the irascible, but usually incontrovertible, Greg Cochran, from which:

Someone I know at TAC opined that everyone knows this stuff, and talking about it is just mean. I think he is mistaken: you have to state important facts every so often, or nobody knows them anymore.

If the race’n’crime topic drives self-described conservatives like the TAC board to such measures, you can imagine what it does to liberals.

Or if you can’t, check out this blog post by Tim Wise fan Benjamin David Steele. Mr. Steele was so disturbed by the statistics on black crime posted at Those Who Can See the other day, he wrote a fevered 5,400-word response.

And Mr. Steele was just clearing his throat there: He contributed a further 14,700 words to his own comment thread, including 50-odd hyperlinks. Talk about protesting too much!

(If you don’t want to read 20,000 words of crimestop rationalization, here’s an abstract: “Racist! Racist! Racistracistracistracistracist!”)

Inter-racial crime: Where are the numbers?

If the race’n’crime topic is radioactive, the subtopic of inter-racial crime—black-on-white, white-on-black—looks about to go critical. It doesn’t just glow in the dark, it glows in broad daylight.

The fate of the famous Table 42 caught a lot of readers’ attention.

Here’s the story. First, go to the Bureau of Justice Statistics website. Got it? In the header menus there you want “Publications & Products | View by Topic | Victims.” That gets you to a multi-page list which you have to scroll through looking for the item “Criminal Victimization in the United States—Statistical Tables.” It was on the second page when I did this.

Select this item. You will then be looking at a page which “Presents 110 tables with detailed data on major variables measured by the National Crime Victimization Survey (NCVS).”

Just a word about the NCVS. It has pages of its own on the BoJS website. The main page tells us that:

NCVS is the Nation’s primary source of information on criminal victimization. Each year, data are obtained from a nationally representative sample of about 40,000 households comprising nearly 75,000 persons on the frequency, characteristics and consequences of criminal victimization in the United States. Each household is interviewed twice during the year. The survey enables BJS to estimate the likelihood of victimization by rape, sexual assault, robbery, assault, theft, household burglary, and motor vehicle theft for the population as a whole as well as for segments of the population such as women, the elderly, members of various racial groups, city dwellers, or other groups. The NCVS provides the largest national forum for victims to describe the impact of crime and characteristics of violent offenders.

There are two key points to notice about the NCVS data.

  • It’s victim reports, by household. The offense may or may not have been reported to the police, and may or may not have ended up in court.
  • It’s gathered by sampling. There are (linear extrapolation from here) about 118 million households in the U.S.A., so the 40,000 or so being sampled by the NCVS workers represent about one household in 3,000.  

Statistical theory includes a large and important sub-theory about sampling, and how best to get from the statistics of samples to the statistics of the parent population, and to estimate the degree of accuracy in having done so.

If you can master that sub-theory, any big insurance company, quality-control agency, or polling firm will give you a really good job. Suffice it to say here that the NCVS does not just multiply their sample numbers by 3,000. They use an elaborate weighting methodology, described here.

The late Larry Auster misunderstood this point in the 2007 articleThe Truth on Interracial Rape in the United Statesthat caused David Horowitz to drop him from FrontPageMag. Larry subsequently, grudgingly, acknowledged his error.

Also please note that if you are not experienced at navigating your way through government databases, and do not have a kindly, patient mentor like me to help you, this stuff is by no means easy to locate. The expression “good enough for government work” explains some of the difficulty; but not, in my dark suspicions, all of it.

OK, so we are looking at the “Criminal Victimization in the United States—Statistical Tables” page. What do we see? Many lists of tables, chiefly dated from 1996 to 2008.

For inter-racial crime you want the list headed “Victims and Offenders.” Each file in the list is a collection of tables. Each one includes that famous Table 42: “Percent distribution of single-offender victimizations, based on race of victims, by type of crime and perceived race of offender.”

This is the only published source on inter-racial crime nationwide, broken out by victim and offender races. The FBI’s Uniform Crime Reports, although very fascinating and useful in their own way, do not have such data. And of course the UCR database is built from “offenses known to law enforcement,” not victim reports.

Table 42 tells us, for example, that in the year 2000 there were 134,180 cases of aggravated assault featuring a white victim, of which 17.2 percent reported the offender’s race as black. (“White” in these tables includes self-identifying Hispanics.)

Fascinating stuff, but . . . where are the numbers for 2009, 2010, 2011, and—surely by now—2012?

Well, the data is somewhere in the BoJS files, but they haven’t been publishing it. They have published some crime data with racial breakdowns—here, for example—but nothing equivalent to Table 42.

There might be innocent explanations for this; but bearing in mind that the Bureau of Justice Statistics reports to the Department of Justice, headed since 2009 by Eric “My People” Holder, if you are thinking the worst, you are probably right.

Can’t we make them give us the numbers?

A reader suggests filing a Freedom of Information Act Request. I shall do so. I urge others to do so, too, as the agency’s response may vary in time and quality depending on the form of words in the request. (Requests are free-form.) The BoJS comes under the Office of Justice Programs for request-filing purposes.

Inter-racial rape statistics.

All right, what is the truth about interracial rape? Numbers like those Larry Auster fielded in his 2007 article show up a lot on race-realist websites. What should we make of them?

It’s hard to know. For one thing, as Larry pointed out, NCVS doesn’t have numbers for rape. What we have in Table 42 is numbers for “rape/sexual assault,” which we are told includes “verbal threats of rape and threats of sexual assault.” Down in the definitions at the end of the tables file we get this for “sexual assault”:

A wide range of victimizations, separate from rape or attempted rape. These crimes include attacks or attempted attacks generally involving unwanted sexual contact between victim and offender. Sexual assaults may or may not involve force and include such things as grabbing or fondling. Sexual assault also includes verbal threats.

So the NCVS is casting its net pretty wide there, and leaving its data open to year-on-year changes in fashion—driven by evolving mores, celebrity antics, and prominent news stories—as to what, exactly, constitutes “fondling” or “threats.”

For another thing, the fact that NCVS is working from samples makes some of these numbers iffy.

The much-noted 2008 figure of 0.0 percent for white-on-black in this category, for example, comes with an asterisk indicating “Estimate is based on 10 or fewer sample cases.”

Bear in mind again that the NCVS is sampling one household in 3,000, then processing the numbers they get through some complex logic to get full national estimates, which they round to the nearest ten.

If you are reporting figures to one place of decimals, as here with the percentages, the usual procedure is to report anything from 0.000000 to 0.049999 as 0.0, anything from 0.050000 to 0.149999 as 0.1, and so on.

So NCVS gathered their sample data, cranked the handle on their sampling algorithm, and came up with less than 0.05 percent of the estimated 46,580 cases of rape/sexual assault nationwide involving a black victim and a white offender. That would be 23 or less.

How much that’s worth, based on those “10 or fewer sample cases” asterisks and the broad definitions as noted, we really can’t have much idea.

If you multiply the percentages by the numbers for rape/sexual assault in Table 42, processing the 0.0 percentages as I just did, you get this for the full reporting range. Numbers derived from asterisked percentages in Table 42 are asterisked here. Only two numbers aren’t; I’ve boldfaced them.


B on W

W on B



< 24*



< 7*



< 10*



< 19*



< 20*



< 13*












< 35*






< 23*





So for 2002 the ratio black-on-white versus white-on-black is two; for 2007 it’s over two thousand!

Much as I hate to supply grist to Tim Wise’s mill, I have to say we are looking at seriously unreliable data here. The unreliability arises not from error or malice, but from the issues raised by tiny sample sizes. Some nonzero number of persons in the U.S.A. are bitten by sharks in any given year, but we should not expect to estimate the number accurately by asking one person in 3,000 if he’d suffered a shark bite.

The entire information content of the table I just showed is: For rape/sexual assault as defined, there are 30,000 or so black-on-white and some smaller, probably much smaller, number for white-on-black in an average year.

More than that, the scrupulous number-cruncher should not say.

Turning the homicide numbers upside down. With homicide we are on much better ground. The definition of “fondle” may change with the phases of the moon, but there is something irrefutable about a corpse.

In my August 29th column I had good sport with Tim Wise’s misleading assertion that:

Any given black person in the United States is about 2.8 times more likely to be killed by a white person than any given white person is to be murdered by a black person.

As I explained:

He’s working from this database, using 2010 data. He gets his 2.8 from dividing 277,000 by 100,000. The first of those numbers is the white population (196.8m) divided by black-on-white homicides from the database (704). The second is the black population (42.0m) divided by white-on-black homicides from the database (413), though he didn’t do a precise division, which would give final result 2.7, not 2.8.

Several readers pointed out that you can just turn those numbers around, cross-dividing to get this equally valid statement:

Any given black person in the United States is about 8.0 times more likely to kill a white person than any given white person is to kill a black person.

And that’s on a crude first approximation, like the one Tim used. If you apply my filters—exclude felons killed by police, etc.—that 8.0 swells to 12.4.


If you want to get to grips with hatefacts, you need to properly understand the hatenumbers.

If only the federal government can report the hatenumbers, don’t be surprised if, under a Left-liberal, anti-white administration, they suddenly decide not to do so.

And if that happens, given that these numbers glow in the dark—some of them in broad daylight—don’t be surprised if nobody complains about the decision, or even notices it.

Except for me, and!


John Derbyshire [email him] writes an incredible amount on all sorts of subjects for all kinds of outlets. (This no longer includes National Review, whose editors had some kind of tantrum and fired him. He is the author of We Are Doomed: Reclaiming Conservative Pessimism and several other books. His most recent book, published by com is  FROM THE DISSIDENT RIGHT (also available in Kindle).His writings are archived at

Readers who wish to donate (tax deductible) funds specifically earmarked for John Derbyshire's writings at can do so here.

Print Friendly and PDF