17

I recently saw this picture:

Interracial Rape: 19,283 cases of "black on white", 0 cases of "white on black"

The image links to this DOJ document.

A similar image gives the following sources from the USA Bureau of Justice for the years 2003, 2004, 2005, and 2006 (table 42 in all cases).

Somehow, the perfect 0 score for 5 years feels unrealistic to me. Does this statistic really check out?

Is the data incorrect? Does the image misinterpret the data in any way?

JasonR
  • 9,247
  • 5
  • 51
  • 65
Worse_Username
  • 379
  • 2
  • 8
  • I have checked the Bureau of Justice statistics site shown and this is indeed what the linked Department of Justice statistics show. Why do you doubt these statistics? –  Sep 19 '17 at 21:40
  • This does seem like a case where there's an implicit claim too, but exactly what that implicit claim might be seems a bit fuzzy. – Nat Sep 19 '17 at 21:58
  • @Worse_Username Thanks. But without any actual claim, those are just a bunch of statistics which don't have all that much meaning. It's the interpretation the image is making that is relevant, not the statistics (and as Phi mentioned, if we doubt the official stats, what will we accept as proof?). – tim Sep 19 '17 at 22:01
  • @Worse_Username [Here](https://thesocietypages.org/socimages/2016/10/17/how-the-bureau-of-justice-statistics-launched-a-white-supremacist-meme/) would be a source interpreting those numbers differently (basically, the numbers are real, but the interpretation by the image is not; which is exactly why a notable source for the **actual** claim - not the statistics - is important. The image in that post might actually serve as notable claim though, as it seems far more wide-spread). – tim Sep 19 '17 at 22:01
  • 2
    @Nat Well, it's called the "National Crime Victimization Survey", and the post seems to raise some (or a lot of) valid issues. But either way, this is really my point. The *numbers* aren't a claim, the *interpretation* is. And here we need a notable interpretation. Though as the article I linked to says, this is a claim David Duke made, so it probably is notable. – tim Sep 19 '17 at 22:11
  • @tim Yeah, trying to check into the source of the numbers though. I mean, I'd assumed that these were crime statistics summed up from various police departments and such, where the less-than-10-cases thing meant that there were less-than-10-white-on-black rapes. Seems like the explicit claim's a lot weaker if it's really a questionable extrapolation. – Nat Sep 19 '17 at 22:13
  • Also, if these are the results of surveys, I wonder if we might be able to find aggregated crime statistics that are summed up from crimes reported to police? Seems like raw numbers would be more interesting than extrapolations. – Nat Sep 19 '17 at 22:29
  • This suffers from the same obvious flaw that most criminal statistics do. Statistics can only be applied to data that is recorded. So if 1) many black rape victims choose not to report the rape to the police, or 2) police reports don't record the race of the victim (both of which seem quite plausible), then any statistics will be inaccurate. – jamesqf Sep 20 '17 at 04:32
  • @tim the new image is for 2008. Should the question be changed? – Worse_Username Sep 20 '17 at 09:03
  • @Worse_Username I think it's basically still the same overall claim as it's always referencing the same survey in the same flawed way. But you're right, the question title doesn't fit that well anymore, so please feel free to change it (the easiest way would probably be to replace "2006" with "2008" (which is also the last year the survey included these stats); it's not reflecting the claim 100% as it doesn't mention 2007, but the report for that year exists, so I think it should be fine; alternatively, you could just reference 2008 in the title, or revert/re-edit the question itself). – tim Sep 20 '17 at 09:12
  • Just to be clear, is the claim being forwarded, with the statistical reference, that white people ***never*** rape black victims? – PoloHoleSet Sep 21 '17 at 14:29
  • @jamesqf The data is survey based and is unrelated to whether or not the victim reported the crime. Random households are contacted and there is an attempt to interview all the people 12 years old and older in the household. – DavePhD Sep 21 '17 at 15:31

2 Answers2

23

These are not actually rape statistics, but survey results, which the image then incorrectly extrapolates from.

According to Business Insider the BJS rejects the given interpretation of these numbers, and they do not use them anymore:

Lynn Langton, the researcher who oversees victimization statistics for BJS, acknowledged that the methods of the 2008 report no longer meet the Bureau's standards.

"These posts and memes rely on data that is eight years old [...] Many of the estimates in the 100+ tables (not limited to Table 42) do not meet our current standards for reporting so we no longer report out this information on an annual basis."

The Business Insider article as well as this post by Philip Cohen reject the interpretation of the image for various reasons, among them the tiny sample size (especially regarding black women, because of the relative small black population in the US). The 2008 survey for example works with data containing less than 10 cases of black on white rape, and less than 20 cases of rape of black people (total). The image then extrapolates the given numbers from these cases.

Community
  • 1
tim
  • 51,356
  • 19
  • 207
  • 177
  • 2
    The data isn't just female victims. It's about 20% male, 80% female. There is also 2010 data available in spreadsheet form online, but not pdf. – DavePhD Sep 20 '17 at 15:24
  • 1
    @DavePhD good point; I guess that would probably make the extrapolation even less correct. I don't think the 2010 data matters much, but out of interest: Do you have a link? And does it contain data equivalent to the sub-standard table 42 data? – tim Sep 20 '17 at 15:46
  • 1
    This is it:http://www.discoverthenetworks.org/Articles/Table%2042%20-%20Interracial%20Crime%20-%202010.xls The sample size is even smaller. Seems like 3 people said offender was black and 2 people said they didn't know. It only matters because BJS is saying "These posts and memes rely on data that is eight years old " (which you quote in the answer) while concealing the newer data. – DavePhD Sep 20 '17 at 17:41
  • @DavePhD Thanks. Though what the BJS says still seems to be true. All images I could find reference data from 2008 and earlier (except the dtn post of course). It also does seem like the BJS doesn't report these incorrect estimates anymore, as they do not seem to be part of the official report; They may have still kept them internally, and they may have been leaked; though I'm not sure how much I would really trust the given source regarding the legitimacy of the file; their [post](http://www.discoverthenetworks.org/viewSubCategory.asp?id=1628) using it seems biased and full of odd conclusions – tim Sep 20 '17 at 18:01
11

The 2008 report surveyed 67,090 people (see pdf page 121) out of a population base of 252 million people (see table 5). 67,090 people is 0.000266 of the population.

In table 42, which shows single-offender crimes by victim and offender race, an extrapolated 46,580 black rape or sexual assault victims are reported. This corresponds to approximately 12 people in the actual survey.

The data in table 42 is consistent with 8 of the 12 saying they were raped or sexually assaulted by a black offender and 4 people saying they were raped or sexually assaulted by a person of unknown race (and zero saying "white" or "other").

For more data see:

Fraction of offenders reported to be white by black victims of rape or sexual assault:

2008 0%

2007 0%

2006 0%

2005 0%

2004 0%

2003 0%

2002 14%

2001 13%

2000 7%

1999 0%

1998 7%

1997 0%

1996 14%

So, based upon the available data, about 5% of black victims of rape or sexual assault report that their offender was white.

DavePhD
  • 103,432
  • 24
  • 436
  • 464
  • 4
    Is 12 people enough to statistically justify the extrapolation or is that bad practice? I only took stats in high school and don't remember it. –  Sep 20 '17 at 14:45
  • 10
    @fredsbend That would probably be a good question at stats.SE (at skeptics.SE, sources are generally more welcome than original research). But I think it's pretty reasonable to assume that extrapolating from 0 out of 12 (though actually not out of 12, but out of less than 10, and with some unknown) to 0 out of 46,580 isn't reasonable, especially for a relatively low-frequency even such as inter-racial rape. – tim Sep 20 '17 at 14:59
  • 1
    @fredsbend 12 (of which only 8 even know the race of the offender) is not enough considering that blacks are only 1/8 of the population. The best thing to do would be to combine multiple years of data. Based upon general crime statistics about 16% of crimes against blacks are by whites. – DavePhD Sep 20 '17 at 15:26
  • 5
    Since rape victims [typically know their attacker](https://en.wikipedia.org/wiki/Rape_in_the_United_States#Rate_of_victimization) (as a lover, family member, friend, or acquaintance), and even when they don't the attacker's still likely to live near the victim, it seems like rape victims are disproportionately likely to be raped by someone of their own race. Coupled with the extremely small sample size here, the claim in the question certainly doesn't follow from the cited statistics. – Nat Sep 20 '17 at 17:18
  • 1
    +1 for finding the information about sample sizes. That really helps to explain the reported figures. – Nat Sep 20 '17 at 17:20