Categories
all about the menz antifeminism evil women misogyny MRA playing the victim rape rape culture reddit sexual assault TyphonBlue

CDC: MRA claims that “40% of rapists are women” are based on bad math and misuse of our data

Standard_adding_machine

Feminists often complain, with considerable justification, that Men’s Rights Activists try to turn every conversation about women’s issues into a game of “what about the men?” You’re talking about female rape victims — well, what about the male rape victims?

The trouble with this strategy, from the point of view of the Men’s Rights Activists anyway, is that this little “gotcha” is much less of a “gotcha” then they’d like it to be.

In the case of rape, for example, feminists are well aware that men are raped as well: the “Don’t Be That Guy” ad campaign, which sent so many MRAs into hysterics, focused on male victims as well as female ones. The emergency room rape advocate organization that a friend of mine volunteers for  provides advocacy for victims regardless of gender.

So many MRAs have started playing another game: trying to twist the conversation around in order to cast women as the villains. Rape is a bit tough for them here, since the overwhelming majority of rapists are male. So MRAs talk about the alleged epidemic of female false accusers instead. Or they change the topic entirely and make dead baby jokes (see my post yesterday).

Recently, MRAs have tried a new strategy, seizing on data from The National Intimate Partner and Sexual Violence Survey, a massive study conducted in 2010 under the aegis of the Centers for Disease Control, to claim that “40% of rapists are women.”

This is a claim repeated by numerous MRAs on numerous websites; see, for example, this post by A Voice for Men’s Typhonblue on the blog GendErratic. Here’s the same claim made into an “infographic” for the Men’s Rights subreddit.

Trouble is, this claim is flat-out false, based on an incorrect understanding of the NISVS data. But you don’t have to take my word for it: the NISVS researchers themselves say the MRA “interpretation” of their data is based on bad math. It’s not just a question of different definitions of rape: the MRA claims are untenable even if you include men who were “made to penetrate” women as victims of rape (as the MRAs do)  rather than as victims of “sexual violence other than rape” (as the NISVS does).

I wrote to the NISVS for clarification of this matter recently, and got back a detailed analysis, straight from the horse’s mouth, of where the MRA arguments went wrong. This is long, and a bit technical, but it’s also pretty definitive, so it’s worth quoting in detail. (I’ve bolded some of the text below for emphasis, and broken some of the larger walls of text into shorter paragraphs.)

It appears that the math used to derive an estimated percentage of female rapists … is flawed.  First, we will summarize the assertion and what we perceive to be the basis for the assertion.

According to the web links, the “40% of rapists were women” was derived from these two steps:

1)      Combining the estimated number of female rape victims with the estimated number of being-made-to-penetrate male victims in the 12 months prior to the survey to conclude that about 50% of the rape or being-made-to-penetrate victims were males;

2)      Multiplying the estimated percentage (79%) of male being-made-to-penetrate victims who reported having had female perpetrators in these victims’ lifetime with the 50% obtained in step 1 to claim that 40% of perpetrators of rape or being-made-to-penetrate were women.

None of these calculations should be used nor can these conclusions be correctly drawn from these calculations.

First the researchers clarify the issue of definition:

To explain, in NISVS we define rape as “any completed or attempted unwanted vaginal (for women), oral, or anal penetration through the use of physical force (such as being pinned or held down, or by the use of violence) or threats to physically harm and includes times when the victim was drunk, high, drugged, or passed out and unable to consent.”

We defined sexual violence other than rape to include being made to penetrate someone else, sexual coercion, unwanted sexual contact, and non-contact unwanted sexual experiences. Made to penetrate is defined as including “times when the victim was made to, or there was an attempt to make them, sexually penetrate someone without the victim’s consent because the victim was physically forced (such as being pinned or held down, or by the use of violence) or threatened with physical harm, or when the victim was drunk, high, drugged, or passed out and unable to consent.”

The difference between “rape” and “being made to penetrate” is that in the definition of rape the victim is penetrated; “made to penetrate” by definition refers to cases where the victim penetrated someone else.

While there are multiple definitions of rape and sexual violence used in the field, CDC, with the help of experts in the field, has developed these specific definitions of rape and other forms of sexual violence (such as made to penetrate, sexual coercion, unwanted sexual contact, and non-contact unwanted sexual experiences). We use these definitions to help guide our analytical decisions.

Now the researchers get into the details of the math:

Regarding the specific assertion in question, several aspects of mistreatments of the data and the published estimates occurred in the above derivation:

A.      While the percentage of female rape victims and the percentage of male being-made-to-penetrate victims were inferred from the past 12-month estimates by combining two forms of violence, the percentage of perpetrator by sex was taken from reported estimates for males for lifetime (a misuse of the percentage of male victims who reported only female perpetrators in their lifetime being made to penetrate victimization).  This mismatch of timeframes is incorrect because the past 12-month victimization cannot be stretched to equate with lifetime victimization.  In fact, Table 2.1 and 2.2 of the NISVS 2010 Summary Report clearly report that lifetime rape victimization of females (estimated at 21,840,000) is about 4 times the number of lifetime being made-to-penetrate of males (estimated at 5,451,000).

B.      An arithmetic confusion appears when multiplying the two percentages together to conclude that the product is a percentage of all the “rapists”, an undefined perpetrator population.  Multiplying the percentage of male victims (as derived in step 1) above) to the percentage of male victims who had female perpetrators cannot give a percentage of perpetrators mathematically because to get a percentage of female rape perpetrators, one must have the total rape perpetrators (the denominator), and the number of female perpetrators of this specific violence (the numerator).  Here, neither the numerator nor the denominator was available.

C.      Data collected and analyzed for the NISVS 2010 have a “one-to-multiple” structure (where the “one” refers to one victim and the “multiple” refers to multiple perpetrators).  While not collected, it is conceivable that any perpetrator could have multiple victims.  These multiplicities hinder any attempt to get a percentage of perpetrators such as the one described in steps 1) and 2), and nullify the reverse calculation for obtaining a percent of perpetrators.

For example, consider an example in which a girl has eight red apples while a boy has two green apples.  Here, 50% of the children are boys and another 50% are girls.  It is not valid to multiply 50% (boy) with 100% (boy’s green apples) to conclude that “50% of all the apples combined are green”.  It is clear that only 20% of all the apples are green (two out of 10 apples) when one combines the red and green apples together.  Part of the mistake in the deriving of the “50%” stems from a negligence to take into account the inherent multiplicity: a child can have multiple apples (just as a victim can have multiple perpetrators).

D.      As the study population is U.S. adults in non-institutional settings, the sample was designed to be representative of the study population, not the perpetrator population (therefore no sampling or weighting is done for the undefined universe of perpetrators).  Hence, while the data can be analyzed to make statistical inferences about the victimization of U.S. adults residing in non-institutional settings, the NISVS data are incapable of lending support to any national estimates of the perpetrator population, let alone estimates of perpetrators of a specific form of violence (say, rape or being-made-to-penetrate).

E.      Combining the estimated past 12-month female rape victims with the estimated past 12-month being-made-to-penetrate male victims cannot give an accurate number of all victims who were either raped or being-made-to-penetrate, even if this combination is consistent with CDC’s definition.

Besides a disagreement with the definitions of the various forms of violence given in the NISVS 2010 Summary Report, this approach of combining the 12-month estimated number of female rape victims with the 12-month estimated number of male victims misses victims in the cells where reliable estimates were not reported due to small cell counts failing to meet statistical reliability criteria.  For any combined form of violence, the correct analytical approach for obtaining a national estimate is to start at the raw data level of analysis, if such a creation of a combined construct is established.

So you’re going to need to go back to the drawing board, MRAs.

What is especially distressing here is that the NISVS data could have been the starting point for a serious discussion of male victims of sexual assault by women, which is a real and often overlooked issue. Unfortunately, MRAs have once again poisoned the well by misusing data in an attempt to exaggerate the purported villainy of women and score cheap rhetorical points.

NOTE: A regular in the AgainstMensRights subreddit approached the NISVS researchers with this same question some months back. Unfortunately, the statement they got back from the NISVS contained an incorrect number. The statement I’m quoting here corrects this number and adds more context.

I can provide contact info for the NISVS representative who got back to me on this to any serious (non-troll) person who requests it.

958 Comments
Inline Feedbacks
View all comments
Athywren
Athywren
11 years ago

I’m not Napoleon, but I do have a thing for controlling all of Europe under my iron fist.

Howard Bannister
11 years ago

GODWIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIN

delurking data geek
delurking data geek
11 years ago

fyi, I was using “naive” in a nonjudgemental way, to refer to a calculation without obvious necessary adjustments, not to people.
this is as close as I can find to a definition in this kind of context! http://www.businessdictionary.com/definition/na-ve-forecasting.html

As for the age thing.. the fact that under 18s are victimised but not surveyed is the exact reason I said that lifetime numbers could skew higher but not lower than naively (in the nonjudgemental sense again) expected.

katz
11 years ago

Delurking data geek: The actual values couldn’t skew lower, but I-forget-which-troll was discounting the lifetime numbers because of the ratio:

Argent I could care less about the lifetime data because the ratio is heavily skewed in favor for women for reasons that are completely unknown.

The male:female ratio could skew lower if the age factor was more pronounced for women than for men.

delurking data geek
delurking data geek
11 years ago

katz: That is a legit possibility! I didn’t pick up that part of the context of your posts. Apologies for misunderstanding.

So to summarise, the possible explanations for the discrepancy in ratios are:

1) women disproportionately more likely to be victimised as children (which we already know to be true)
2) men are disproportionately repeat victims.
3) the “victims just forget about MTP assaults” hypothesis, which seems even less likely now it’s competing with two alternative hypotheses.

OK, that all seems much clearer now. Thanks for helping me understand it!

cassandrakitty
cassandrakitty
11 years ago

My favorite part of the troll’s comments about lifetime figures versus figures for one year is the idea that if the lifetime figures show women being raped more often, well, who cares? He doesn’t like that data, so obviously it’s wrong.

katz
11 years ago

I failed to cite the context so it’s my own fault; a lot of people were bouncing around a lot of different numbers.

And then of course there are the “this year is non-representational” factors (titianblue mentioned these): Either random or systematic.

hannasoumaki
hannasoumaki
11 years ago

@argenti or anyone who can speak better about this hi. im sorry to bother you but when i asked questions to typhonblue about her analysis, i received this comment genderratic.com/p/836/manufacturing-female-victimhood-and-marginalizing-vulnerable-men/#comment-119299 and i have some questions?
1. does ‘raw data’ mean the numbers of people raped w/o other factors like rate, age, etc. or does it mean something else?
2. past the first paragraph im confused what theyre trying to say. i think overall they have a problem w/ either that blair came up with ‘37% of female rapists’ and rounded up to 40% or that argenti came up w/ ‘20%’ and doesn’t know why? im following argenti’s math from a few pages back and i dont understand? did adiabat mean they extrapolated from the 12 month numbers wrong?
again sorry for asking such questions. hope everyone has a happy and safe halloween.

CassandraSays
CassandraSays
11 years ago

That Guardian thread is still making flames come out of my ears.

Like the first poster, I was taught to respect women. I’ve never taken advantage of a drunk woman, even when I’ve been under the influence. I would suggest that I am part of the majority. There are those however who are caught up in a lifestyle or a background where that is not the case and they fall victim to the rape scenario.

Yep, in a situation where a man rapes a woman it’s definitely the man who’s the real victim. It’s the scenario, you see – he can’t help but play his part.

Howard Bannister
11 years ago

There are those however who are caught up in a lifestyle or a background where that is not the case

Actually, the whole ‘benevolent sexism’ of ‘respecting women’ is a good predictor of having nasty misogyny and stuff… so, no. No, it’s those just like you who are, in fact, the problem.

nilvoid
nilvoid
11 years ago

Wow, uh, awesome formatting there. That first sentence should read “Here’s another post on tumblr about…”

delurking data geek
delurking data geek
11 years ago

The raw data would be every recorded answer to every individual question on every survey, no analysis whatsoever.

I haven’t looked at this survey in detail, but in general, there is potential for serious privacy issues with publishing raw data in a survey like this because people might be able to figure out information pertaining to the identity of the respondents by examining all their answers to the questions. People take steps to make respondents anonymous where raw survey data is published but sometimes that could be inadequate. Frankly the idea of raw data on crime victimisation being released to the MRM gives me chills. They’d decide that anonymous respondent #32477 was lying on the survey because of women’s genetic compulsion to make false rape accusations, then they’d use half-assed mob logic to decide some random woman on facebook was respondent #32477, then they’d hound said random woman to the point where she feared for her life.

The third paragraph is basically bullshit. They’re complaining that they haven’t got better information, and that’s the CDC’s fault (there’s an amusing implication that this is because of the feminazi conspiracy but I am PART of the feminazi conspiracy and I know for a fact we didn’t do this and we have no direct control over the CDC) so they’ll keep pretending the number they pulled out of their ass is the truth because no one will tell them the truth and that simply isn’t fair. The CDC is forcing them to be full of crap! They are MRAs, nothing they do is ever their own fault.

Keep in mind that even with the raw data, a survey of crime victims simply can’t give them the information they want about perpetrators. And even without the raw data, there was no reason for them to latch onto the past-year number instead of the lifetime number, other than the obvious reason that the past-year data gave them an answer they liked once plugged into their bullshit calculation.

CassandraSays
CassandraSays
11 years ago

OK, so none of the other data supports the conclusions that I’ve drawn from the 2010 data, but as an MRA I really like those conclusions so obviously all the other data is wrong. My confirmation bias, let me show you it.

Ally S
11 years ago

@nilvoid

The Widom and Morris study that Typhonblue cites is useless for these reasons:

1. The survey was conducted within the context of a 2-hour in-person interview. It is likely, then, that what caused the relative disinclination to disclose abuse was the presence of the interviewer. It’s possible that the male respondents felt less comfortable talking to someone in person about child sexual abuse. An anonymous survey could have yielded different results.

2. It only screened for child sexual abuse. That is a problem because the NISVS only had respondents of ages 18 and above; the 12-month figures, therefore, miss a significant number of child sexual abuse incidents, to which the Widom and Morris study could be applied. But it can’t be applied to the 12-month figures because not all child sexual abuse victims are 17 (the youngest possible age of victimization that could be found in the 12-month figures).

3. A quick Google search led me to a criticism of the Widom and Morris study that basically states that the study never actually checked whether the documented cases of sexual abuse were the exact same incidents that the respondents disclosed to the interviewer(s). That means that the data was unreliable from the start.

4. Even if the Widom and Morris study did not suffer from any major methodological flaws and served as valid evidence that the lifetime figures are unreliable, its use would be limited.
In particular, you can’t simply throw away the lifetime rates and deem them unreliable and then gather data from the lifetime rates (in this case, the lifetime figures on the sex of the perpetrators reported by the respondents) to make a conclusion about the 12-month rates.
In other words, if you think that one data set is non-representative, then using information about that data set and applying it to an actually representative data set is complete nonsense.

Ally S
11 years ago

Ew, my previous comment’s last paragraph is ugly. x_x

Argenti Aertheri
11 years ago

That study TyphonBlue cites, does anyone know if they made any attempt to account for the people who didn’t say but did remember it? Cuz didn’t mention what I wanted them to =/= doesn’t remember the incident.

Ally S
11 years ago

Argenti, so far as I can tell, all they did was get a group of people with official documented cases of child sexual abuse and then ask them through 3 or 4 surveys asking them questions like “Has anyone ever done [act X] to you when you were young?” (paraphrasing there). Nothing beyond that.

katz
11 years ago

It only screened for child sexual abuse. That is a problem because the NISVS only had respondents of ages 18 and above; the 12-month figures, therefore, miss a significant number of child sexual abuse incidents, to which the Widom and Morris study could be applied. But it can’t be applied to the 12-month figures because not all child sexual abuse victims are 17 (the youngest possible age of victimization that could be found in the 12-month figures).

Wait, what? So they got respondents who might be 20, 30, or 40, and asked them “were you abused as a minor in the past 12 months?”

Ally S
11 years ago

Katz, the Widom and Morris study asked them about whether they have been sexually abused in their childhood – no questions about what happened within the last 12 months.

Argenti Aertheri
11 years ago

Katz — no, that’s kinda the point. The other study only screened for child sexual abuse, so who knows if it applies the the CDC study where only a tiny percent of respondents would’ve been minors in that 12 month period.

katz
11 years ago

Oh, OK, they collected lifetime data, people are just using 12-month figures gleaned from their study?

Argenti Aertheri
11 years ago

Ninja’ed by the person being asked, whoops!

(My mother just went “OO A PIECE OF CANDY” — we’re sitting on the porch with the bowl between us, this is about as exciting as “OO MY FINGERS”)

katz
11 years ago

Or rather, applying conclusions drawn from their study to the 12-month data.

Ally S
11 years ago

Sorry if I was unclear, katz. Basically, Typonblue is using child sexual abuse disclosure data from the Widom and Morris study (which measured only lifetime incidence) to invalidate the lifetime figure of female-on-male rape found in the NISVS report. The idea here is that men were less likely to disclose sexual assault experiences in the NISVS and that’s why the lifetime rates show a higher female victimization figure. Therefore, the lifetime rates are inaccurate and the 12-month ones must be used instead, according to her.

1 16 17 18 19 20 39