One way unserious people tire you out is by taking 15 minutes to write something unserious. It takes you four hours to undo their ignorance. —Ta-Nehisi Coates
In an amazingly bad column about campus rape, George Will discusses some statistics that the White House included in a recent report:1
The statistics are: One in five women is sexually assaulted while in college, and only 12 percent of assaults are reported. Simple arithmetic demonstrates that if the 12 percent reporting rate is correct, the 20 percent assault rate is preposterous. Mark Perry of the American Enterprise Institute notes, for example, that in the four years 2009 to 2012 there were 98 reported sexual assaults at Ohio State. That would be 12 percent of 817 total out of a female student population of approximately 28,000, for a sexual assault rate of approximately 2.9 percent — too high but nowhere near 20 percent.
In a later defense of his column, Will doubled down on Perry’s math:
I cited one of the calculations based on it that Mark Perry of the American Enterprise Institute has performed {link}. So, I think your complaint is with the conclusion that arithmetic dictates, based on the administration’s statistic. The inescapable conclusion is that another administration statistic that one in five women is sexually assaulted while in college is insupportable and might call for tempering your rhetoric about “the scourge of sexual assault.”
George Will defends his column based on the math. So it’s notable that Will’s math is wrong.
Will writes “One in five women is sexually assaulted while in college, and only 12 percent of assaults are reported.” Both statistics can be found on page 142 of the White House’s report on sexual assault on campus (pdf link), although they appear three paragraphs apart.
What Will neglects to tell his readers is that the “20%” and “12%” statistics come from different studies, and the two studies don’t define “sexual assault” the same way. The “20%” statistic,3 from a study published in the Journal of American College Health,4 includes both rape and sexual battery, and both attempted and completed assaults. In contrast, the “12%” statistic, from a study published by the National Crime Victim Research and Treatment Center,5 refers only to completed rapes.
That means the mathematical comparison Will makes is apples and oranges. The 20% statistic counts “A B C and D,” while the 12% statistic is looking only at “D.” What Will’s math does, when you boil it down, is point out that “A + B + C + D” is higher than “D,” and then claims that makes the statistics “preposterous.” Well, no, the statistics aren’t preposterous – George Will’s math is.
Furthermore, the “98 reported sexual assaults at Ohio State” figure Will uses as a basis of comparison (which comes from Mark Perry’s AIE blogpost) enormously undercounts sexual assaults reported by Ohio State students, because it only counts reported rapes that took place either on-campus or at a few select off-campus locations (such as an off-campus building run by the university or a recognized student organization). Notably, private residences are not included.6
But according to the Journal of American College Health study, 61-63% of sexual assaults took place off campus, most commonly in a private residence. Other studies have found similar numbers (for instance, this government study found that 66% of rapes of college women took place off campus). So it’s likely that most sexual assaults in the “20%” statistic wouldn’t be included in George Will’s “98 reported rapes” figure, even if they were reported to police.7
See also: Christie Wilcox at Discover Magazine makes some similar points, and tries to recreate George Will’s math using less dubious numbers.
P.S. I focus on the numbers in this post because I think this aspect has been undercovered. But there are many, many, many other blogs out there that have criticized Will’s awful column along other lines. To scratch the surface:
The conversation we need about our sexual culture – The Washington Post
George Will and Sexual Assault | Ordinary Times
Willful Offense | Ordinary Times
George Will defies description, and a holder of a “coveted status” responds
I was raped, and I stayed silent about my "coveted status."
A College Rape Survivor Responds To George Will’s "Coveted Status" Remarks | Blog | Media Matters for America
George Will’s distasteful conclusions about sexual assault – The Washington Post
George Will: Being a victim of sexual assault is a “coveted status that confers privileges” – Salon.com
ECHIDNE OF THE SNAKES: Being A Victim: A Coveted Status. Or So Writes George Will.
How Twitter voted #survivorprivilege off the island
And – to be fair – a good post by Conor Friedersdorf showing that many of the critiques of Will’s column have misread what Will was saying. But even read properly, Will’s column was pretty awful.
- The report was “Rape and Sexual Assault: A Renewed Call to Action.” [↩]
- Or page 18, if you go by the pdf page numbers rather than the page numbers printed at the bottom. [↩]
- It’s actually 19%, not 20%. [↩]
- “College Women’s Experiences with Physically Forced, Alcohol- or Other Drug-Enabled, and Drug-Facilitated Sexual Assault Before and Since Entering College,” Journal of American College Health, Volume 57, Issue 6, May-June 2009. A detailed report from this study can be read here (pdf). [↩]
- “Drug-facilitated, Incapacitated, and Forcible Rape: A National Study,” 2007. pdf link. [↩]
- My source for this is a source Perry used: Ohio State University’s “2013 Annual Campus Security Report & Annual Fire Safety Report.” (Pdf link.) The relevant info is found on page 46 as the pdf program counts pages. [↩]
- We also know from news reports that the Ohio campus cops sometimes declare reported rapes “unfounded.” I can’t tell if these “unfounded” reports are included in Will’s “98 reported rapes” number or not. [↩]
To be clear, from the report:
(1) “Sexual assault is a particular problem on college campuses: 1 in 5 women has been sexually assaulted while in college.” This is wrong and includes attempts.
(2) “Reporting rates for campus sexual assault are also very low: on average only 12% of student victims report the assault to law enforcement.” This is wrong and refers not to assaults, but only rapes.
So even though both the statistics in the report are wrong, Wills is the liar who’s fucked up his complaint about the White House’s incorrect and contradictory statistics. Rather than focusing on what they actually wrote, he should have instead ignored the contents of the report and used the statistics as they would have be quoted from the original sources if anyone had cared about accuracy. That way his math would have worked.
I think your politics is coloring your judgement.
Welcome to Alas, Franz. :-)
Before I respond, can you clarify something for me? Are you saying 1) Both George Will and the White House fucked up and bear responsibility, or 2) The White House alone fucked up and Will bears no responsibility?
It doesn’t matter whose side I’m on politically.
Wills says the White House misuses statistics and does a calculation showing the 1/5 assaults and 12% assaults reported quoted in the report can’t both be right. You say, actually, the 1/5 the WH said were assaults actually include *attempted assaults*, and the 12% the WH said were reported assaults aren’t really, they’re in fact reports of *completed rapes*. Therefore “math is wrong”, “LIAR”, “the statistics aren’t preposterous – George Will’s math is”.
That just doesn’t follow. It proves the exact opposite. Wills calculation showed the WH stats were inconsistent, and you’ve just provided detailed references to the exact mis-quotations. The fact that there are other statistics out there which make sense isn’t an excuse for the WH using ones that don’t, and doesn’t means Wills erred in pointing that out.
Will is… nothing special. But I’m not sure you can blame Will for the incredibly bad statistics going around.
The most relevant stuff is probably here, on the page which is numbered “10” (I don’t think it’s actually page 10 of the PDF): https://www.ncjrs.gov/pdffiles1/nij/182369.pdf
Here are some basic statistics:
1) 1.7% of the sample was raped* during the sample period (6.91 months, which is also, confusingly, referred to as “one academic year.”) That calculates to about 1 in 58 women who were raped during their time spent at college.
2) 22 percent of the victims had been raped more than once during the study. TWENTY TWO PERCENT, and I remind you that this is only in a 7 month period! (It may be that they mean “victimized” more than once even though they say “raped” more than once. But still.) I sort of have to assume that this is rape by a partner, housemate, or some other situation where the rapist and victim have continuing contact.
3) Everything else is statistics which are simultaneously basic (multiply ___/year by ___ years to get a number) and complex (based on a variety of assumptions which may or may not be correct.) The most obviously questionable assumption is that it assumes the risk of rape at college is the same as the risk of rape outside college. We have pretty good statistics for the rate of rape in the general public, so I don’t think it makes sense to extrapolate like that.
4) The commonly used** statistic that “1 in 5 women will be raped in college” is wrong on multiple fronts. First, it is not limited to rape.** Second, it also includes time when people are NOT in college, which is to say that it takes the “6.91 months of an academic year and extrapolates to the remainder of the year. That doesn’t mean people aren’t raped, but they aren’t raped “in college.” Third, it assumes a five year period.
5) The accurate statistics would be more complex. But perfectly valid. For example, “In a college with 1000 women, if we look only at the time when classes are going on and exclude vacations, roughly 36 women will be raped every single year.” That is an extraordinarily compelling statistic on its own.
*Everyone is talking about “rape” most of the time, and of course even the thread title uses that word. I think that when you talk about “rape” you need to limit discussions to statistics of rape, and not something else like, say, attempted rape. Mostly that is because retrospective self-reports of attempted rape seem much wonkier than retrospective self-reports of rape. Also it is because they are incredibly different in seriousness; one is a capital crime. They should not be combined IMO.
**Not by you. You are always careful to use “sexual assault” when possible.
Sorry, I do intend to respond to Franz (and maybe to G&W), but I’m swamped today. Later!
I agree; notice I didn’t ask you about your politics.
However, it’s hypocritical of you to say that while making comments like “I think your politics is coloring your judgement.” If you want me to not act that way, then you shouldn’t act that way, either.
If I’m reading you correctly, you’re implying that it’s a mis-quotation or otherwise incorrect to include both attempted and completed assaults within the total. Actually, this way of doing things is commonplace within the field. For instance, the FBI’s definition of “forcible rape” includes both attempted and completed rapes. The BJS defines “rape” and “sexual assault” as two different things, which ime is a bit unusual, but defines both of those terms to include both attempted and completed crimes.
Similarly, it’s not unusual for a particular study to only look at completed rapes in some or all measurements and to refer to this as “sexual assault.” This is why terms are defined in papers – because no one expects every paper to use every term identically.
Your argument ignores the difference between a literature review and a meta-study.
The WH document was essentially a selective overview of the literature, and included 88 footnotes referring to 62 different studies and sources. Literature reviews don’t include only studies that define every term the same way and have mathematically comparable numbers; their point is to provide an overview, and overviews by definition include a variety of sources.
All good studies define their terms, because everyone knows different studies can define key terms differently. It’s ridiculous to imply that an overview that draws from over fifty sources is dishonest or in error because not every study cited is mathematically compatible, when the overview never claimed that every study cited is mathematically compatible.
In contrast to the literature review is the meta-study. In a meta-study, data from different studies is mathematically manipulated to produce a new result (or to replicate an old result but with a larger number of observations). This is essentially what Will and Perry did (albeit on a small scale). In a meta-study, it is the author’s responsibility to confirm that numbers taken from different studies are compatible. If the author mathematically treats incompatible numbers as if they are compatible, that’s the fault of the meta-study author, not of the sources she drew the numbers from.
And that is what George Will and Mark Perry did. They took mathematically incompatible numbers from different studies and pretended they were compatible. And when the numbers of course failed to match, Will’s and Perry’s claim is that this shows the statistics are “preposterous.” That’s not a fair conclusion to draw from their evidence.
No it doesn’t. About 1 in 58 women were raped during the 6.91 (let’s call it 7) months covered by the study. “Their time spent at college” would include more time than 7 months. From the study (the 2.8% figure this paragraph uses includes both attempted and completed rapes):
G&W writes:
To clarify, the sample period wasn’t 7 months. Rather, the survey asked about events that occurred since the current academic year began. The amount of time since the current school year began varied from subject to subject, but the average amount of time was 6.91 months.
Anyway, that’s a very good study you linked to. They did a lot of interesting things, like a sub-study which varied the screening question methodology (to test the effect of different approaches to screening questions).
Okay. Sorry for the politics slam.
It is factually a misquotation, there’s no argument.
The NJCH says “Nineteen percent of the women reported experiencing completed or attempted sexual assault since entering college.” The WH says “1 in 5 women has been sexually assaulted while in college”.
The NCVRSC says “Among college women, approximately 12 percent of rapes were reported to law enforcement. ” The WH says “Reporting rates for campus sexual assault are also very low: on average only 12% of student victims report the assault”.
I don’t know why you are trying to debate this. The second statements are just not accurate representations of the first. The point of a literature review is to accurately summarise the literature. The WH review fails.
P.S. Neither of the studies cited mix definitions. They are clear, and it is the WH which introduces the blurring.
[Iseter’s comment, which had no on-topic content at all, deleted. –Amp]
And Iseter – after ten posts in a row in which he disagreed with me in no uncertain terms without being banned or even given a warning that I recall – switches to an all-out personal attack (I’m a feminist because I want to get laid, blah blah blah) and is banned. I guess that this is what Iseter wanted all along, so this banning makes everyone happier, and is therefore a good idea from a Utilitarian perspective.
Why are there so few anti-feminists who are capable of having a disagreement without making personal attacks?
Probably, Amp, he was pissed that you and Franz are having a civil and thoughtful discussion even though you profoundly disagree, hence the childish forced-teaming attempt with Franz.
I suspect you are seeing selection bias. Most anti-feminists who strongly disagree with you aren’t going to hang out here. Those who do are either antifeminists who believe they can constructively engage with you (like ballgame); or people who want to behave like assholes, secure in the knowledge that since they will eventually get shown the door, they can congratulate themselves on being martyrs at the hands of thr Feminist Conspiracy to Destroy Truth and Free Speech, and thus not need actually concern themselves with minutae like getting their facts straight or presenting a well-thought-out argument.
While the JACH paper reports that 19% of female survey respondents had experienced attempted or completed sexual assault, the JACH paper also reports that 19% of seniors had been sexually assaulted since entering college.
I think the White House is referencing that when it says: “1 in 5 women has been sexually assaulted while in college”. As g&w points out, it could also be referencing a completely different study that derives the same 1 in 5 statistic using a different methodology.
For the 12% reporting rate, the White House translates the term “rape” (used by the NCVRSC) to “sexual assaults” (the NCVRSC did not look at the larger category of sexual assault). This is what Amp meant about dealing with inconsistent terminology in constructing a literature review (that the referenced research internally uses consistent terminology is expected but irrelevant). This may be less than ideal, but the use that the White House made of the statistic is not misleading. The White House claims that reporting rates are low and then gives the reporting rate from a study of the most reported category of sexual assaults as supporting evidence. The actual average rate of reporting for all sexual assaults is (presumably) lower than 12%, but that does not contradict the White House claim that reporting rates are lower.
The White House report should not be used as George Will used it. If he wanted to do further statistical analysis from a literature review, he needed to go back to the primary literature referenced, not make guesses about the context of the statistics mentioned in the literature review (and deciding that the 1 in 5 statistic refers to sexual assaults that occur on campus property seems like a particularly shoddy assumption to make, one driven more by having numbers for that rather than actually believing that that was what the White House report was referring to).