New perspectives on the DOJ data

(Note: David Horowitz, saying that the “statistics in [Auster’s] article were incorrect,” links to this blog entry which he describes as my correction of my article. While the statement that there were under ten “white”-on-black rapes and sexual assaults in the U.S. in 2005 cannot be accurate, as discussed below, Horowitz’s statement that the statistics in my article were incorrect falsely suggests that I incorrectly reported the Department of Justice figures. In fact, I reported them accurately, as I discuss in my response to Horowitz’s “correction.” I also will be posting a fuller article on interracial rape statistics in the near future.)

Finally, three and a half days after my brief article on interracial rape was published at FrontPage, and ten days after it was posted at VFR, and after so many people have denounced the article as racist without one of the denouncers making any argument that the information from the DOJ study that I quoted was not true, two commenters have come forward explaining and qualifying the DOJ data. Their point is that the study was a survey, not a collection of information for the whole country, and that this explains the amazing datum that there were zero rapes by whites (meaning by whites and Hispanics) of blacks in 2005. Of course, this does not change what has been known for many years from Jared Taylor’s important studies of federal crime statistics: that white-on-black rape is a tiny fraction of black-on-white rape, which of course was the main point of my article. I will reply to these comments later.

Ken H. writes:

Lawrence Auster says: “I must say it seems amazing to me that there were zero white-on-black rapes in the U.S. in 2005.”

The DOJ figures are not reported rapes. The page you link to is a random survey, not a record of police reports. That’s why it is called the “National Crime Victimization Survey.”

The footnote “Estimate is based on about 10 or fewer sample cases” does NOT mean that there were 10 or fewer victims nationwide. It means there were 10 or fewer in the survey sample.

Notice how, also in Table 42, there are an estimated 33,400 white victims of Robbery:Completed:With injury. That also has an asterisk by it, meaning that the estimate is based on ten or fewer instances in the sample.

It’s not at all surprising that there are zero white-on-black rapes in the 2005 sample, as I’ll show.

If you combine the last ten years of data, you can get better estimates for the rates. The numbers I get are (comparable to table 42) For single-offender rape/sexual assault When the victim is black, the (perceived) race of the offender is white about 6 percent of the time. When the victim is white, the (perceived) race of the offender is black about 12 percent of the time.

If this is combined with the victimization rate data (from ten years of table (6), we can estimate that each year, for every 100,000 white females, about 25 are raped by black men; for every 100,000 black females, about 16 are raped by white men.

The sample size was 67,000 persons. About 4500 of those would be black females. Of those 4500, we would expect about 15 to have been victimized by rape or sexual assault over one year. The expected number of white-on-black rapes in the sample is about 4500 .00016 = 0.72. Using the equation for a Poisson distribution, there is about a 50 percent chance that no white-on-black rapes would be found in a given year’s sample, a 35 percent chance of one rape, and a 15 percent chance of more than one.

The apparent actual numbers (inferred from table 42 in the different years):

1996 2
1997 0
1998 1
1999 0
2000 1
2001 2
2002 2
2003 0
2004 0
2005 0

Zero rapes: 5 years
One rape: 2 years
Two rapes: 3 years

A. Zarkov writes:

It’s not quite accurate to say that “there were zero white-on-black rapes in the U.S. in 2005,” although the number is surely very small. The survey is just that, a survey and not a census. The numbers in Table 42 are therefore estimates, not actual counts. If you read the survey methodology starting on page 130 you will see that 38,600 households were surveyed covering a total of 67,000 people. If you look at the footnotes for Table 42 you will see that there were 10 or fewer white-on-black rapes in the sample, not the entire country. Their methodology does not give a reliable estimate for white-on-black rapes, so they put 0 percent with a footnote. The reason they can’t get a reliable estimate is the number in the sample is too small, so a change say one might lead to a large change in the percentage.

There are at least two other facts worth noting. First Table 42 does not report rapes; it reports rapes/sexual assault. According to the glossary on page 143 “Sexual assaults may or may not involve force and include such things as grabbing or fondling. Sexual assault also includes verbal threats. Therefore the number of actual rapes is less then the numbers in Table 42. Of course this means there could have been zero white-on-black rapes in the sample. It also means the actual number of rapes is less that what you see in Table 42. The second item has to do with the response rates for the survey. The household response rate for 2005 was 91 percent, which is pretty good, but the response rate for persons was less, 84 percent. Low response rates are a real problem for surveys. If the people who didn’t respond are significantly different from the people who did then the survey can be biased. For example suppose the black women raped by white men tended not to respond to the survey. There are various ways of trying to cope with non-responses, and it’s not clear that DOJ did anything to adjust the data. Frequently surveyors simply ignore the problem even when the response rates are extremely small, say 5 percent.

It would be useful to compare the survey with incarceration statistics and other crime data. Since inter-racial rape is such a hot button issue, I think one should be careful to make only absolutely correct statements with the appropriate qualifications even at the risk of sounding pedantic.

Dimitri K. writes:

I am kind of familiar with the subject of statistics myself, but it is hard to understand anything from the explanations those guys provided. There is a lot of talk about distributions, but the main point (what the survey is and how it was made) is not explained at all. However, don’t get fooled. Whatever method is used, 37000 versus 10 is 37000 vs ten - it is a significant number. It says it all.

Sage McLaughlin writes:

Regarding Dimitri K.’s comment that, “Whatever method is used, 37000 versus 10 is 37000 vs ten—it is a significant number. It says it all,” there’s something important to be said. The point that Ken and Mr. Zarkov have been trying to convey, I believe, is that just looking at those ratios does NOT, in fact, tell you “it all.” It may tell us something important, to be sure. But it doesn’t tell us everything we need to know in order to understand the phenomenon. Dmitri seems determined in his reply to cling to the view that he knows everything worth knowing about this phenomenon by comparing two ratios, and by God, nobody’s going to tell him that there’s more to the story.

Look, I’m the last person to sugar-coat data like this, and I’d be the first to say that people’s inability to face the facts of interracial crime is dangerous and cowardly. I’m disgusted by it. And yet, we have to be careful about claiming that the facts ever really do speak for themselves: they never do. We have to be careful, furthermore, that our facts are actually, you know, factual. We’re not feminists after all, so if we’re going to throw around rape statistics it’s incumbent on us to make sure we at least make some effort to be sure they’re correct. The difference between no rapes of a certain sort and very few rapes of a certain sort may seem trivial, but it’s not. I don’t let leftists get away with throwing around sensationalistic numbers on the argument that they reflect a “deeper reality,” whether or not they’re actually true. And I don’t want people on the right to do that either. I spent a lot of spare time investigating rape numbers some years ago, and have invested too much time in debunking feminist statistical chicanery to give myself a pass on this sort of thing.

While the ratios are significant and tell us something important, that point is going to get lost and our credibility will be significantly damaged if we make claims to the effect that not a single rape by a white man against a black woman took place in this country last year, if in fact that’s not the case. The best thing about being a traditionalist, I’ve found, is that I don’t have to lie or adhere to politically convenient fictions. I don’t have to be afraid of facts or truth. If there is some explanation of the FBI’s CVS numbers that clarifies what those numbers really tell us, we should welcome it. (Granted, people will try to throw excessive mud in the water by getting into overly complicated examinations of the data, but we don’t have to go for that.)

LA writes:

I’ve gone back to the document in the National Crime Victimization Survey. It just presents the data in a group of tables, without providing any explanation of how the data were derived. There is no way for the reader to know, without going to outside sources of information, either that these figures are projections or what the original survey numbers are on which the projections are based. A related document explains the NCVS in general terms:

BJS criminal victimization data collections

National Crime Victimization Survey (NCVS) is the Nation’s primary source of information on criminal victimization. Each year, data are obtained from a nationally representative sample of 77,200 households comprising nearly 134,000 persons on the frequency, characteristics and consequences of criminal victimization in the United States. 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.

So it’s a survey, involving a truly vast number of respondents, 134,000, and it’s the most authoritative source of information on crime victimization in the U.S. At the same time, it remains a survey, a projection of figures from a sampling.

We know there must be more than ten black victims of rape/sexual assault by whites each year, and clearly the document should have been more clear in explaining that the “less then ten,” on which the 0.0 percent figure is based, results from the fact that this is a survey, not hard numbers. According to The Color of Crime, the rate of black-on-white rape/sexual assault is seven times higher than the rate of “white”-on-black rape/sexual assault (“white” in quotation marks indicates white and Hispanic). And that’s after correcting the differential down from a much higher number, as I will explain when I write on this at more length. The Color of Crime further estimates that if Hispanic perpetrators were removed from the equation, the ratio of black-on-white rape/sexual assault to white-on-black rape/sexual assault would be about ten. Again, I will have more on this in the next couple of days.

For the moment, we can say that it’s impossible that the actual number of “white”-on-black rapes/sexual assaults each year is under ten. At the same time, because the national figure of black-on-white rapes/sexual assaults, 37,460, is so large and the sampling on which it is based is so large, it is much more reliable. It may also be possible indirectly to separate out the number of rapes from the number of sexual assaults. Enough on this for now.

* * *

This e-mail by Ortelio came in early Sunday morning and I only got to it very late Sunday night, so it’s being posted lower in this thread than its rightful place, which would be at the beginning. Ortelio does an excellent job of going into several of the pdf documents for different years and comparing the rape/sexual assault figures for all of them.

He writes:

I don’t want to anticipate your evening’s reading, and am in no way an expert in statistics in general or the DOJ’s statistics in particular. But their imperfection, which you have pointed out forcefully in one or two ways, needs to be stressed.

1, Though white on black rape/sex assault is given as 0% in 2005, 2004, 2003 and 1999 (to go back no further), it is given as 14.2% (of 59,490 = 8447) in 2002 and 13.4% (of 29,980 = 4017) in 2001.

2. That sort of fluctuation, like others in the same Table 42, doubtless results from the survey method, which is based on interviewing around 40,000 of the nation’s households (around 70,000 persons), and then extrapolating.

3. The 2005 figures (meaning always Table 42) show 33.6% of the 111,490 rapes/sex assaults on whites being committed by blacks and 44.5% by whites. But the comparable figures for 2004 have only 8.3% of the 139,900 rapes/sex assaults on whites being by blacks, and 65% by whites. Wild fluctuation in figures, given with absurd over-preciseness. No doubt little fluctuation in reality.

4. The definition of “rape/sexual assaults” is even more maddening than you indicated, since it explicitly includes even “verbal assaults”! http://www.ojp.usdoj.gov/bjs/pub/pdf/cvus/cvus05mt.pdf

All this leaves your broad point intact.

Posted by Lawrence Auster at May 06, 2007 03:10 PM | Send