The Marshall Project Takes A Swan Dive Into The Bidet (Update x2)

It’s not easy being the Messiah du jour, especially when there are so many emotions at play, but one would have thought  that Oberlin educated Neil Barsky could control his urges long enough to gain some cred, if not protect his non-profit status, before losing control. Nope. It was less than a month ago when it debuted by attacking criminal defense lawyers as the bane of the system, but it has now gone for naked advocacy with Dana Goldstein’s post on dueling data on campus rape.

On Monday Emily Yoffe published a long piece in Slate arguing that the Obama administration and college officials have become overwrought in their concern about campus sexual assaults. Yoffe heaped particular scorn on one of the common talking points in the campus rape debate: the claim that one-fifth of all female students have been victimized.

Of the many things Emily Yoffe did, heaping scorn was not among them.  If anything, Yoffe’s post was as neutral and unladen by colored adjectives as anything ever written on the subject.  Goldstein’s use of words that carry a negative connotation is not accidental.

A federal report released today seems, at first glance, to back Yoffe. It estimates that fewer than one percent of women aged 18 to 24, and even fewer female college students — .6 percent — have been raped or sexually assaulted.

So Goldstein is reporting a new report that confirms that the pervasively reported statistics are, and always have been, utter nonsense?  Not exactly.

What accounts for the phenomenal gap? And is the “one-in-five” claim as wildly off the mark as Yoffe implies? Maybe not. The two studies are based on surveys that asked different questions of different populations under different circumstances. And while Yoffe accurately pointed out the limitations of the “one-in-five” study, the .6 percent estimate has its own set of serious shortcomings. In short: don’t take either number as gospel.

Wait.  The federal report says .6% of female college students have been raped or sexually assaulted, so the story in breaking the news of the report is “maybe not”?

The report released today by the Department of Justice is based on the National Crime Victimization Survey (NCVS), which relies on interviews with 160,000 Americans over the age of 12. The survey asks respondents if they are attending college, and classifies everyone who says “yes” as a college student, even if they are enrolled solely in online courses or spend only a few hours per week on an actual campus.

Well, that’s going to raise some very serious questions about . . . oh, wait. 160,000?  That’s a pretty good sample. Actually, that’s a huge sample.  The larger the sample, the more valid the survey, and this one is pretty big. But what of the fact that the survey included “attending college” as “enrolled solely in online courses or spend only a few hours per week on an actual campus”?  Well, that is “attending college,” but more to the point, with a sample size that large and the relative numbers of students “attending college” strongly skewed in favor of full-time, on-campus students as opposed to online or “a few hours,” because that’s where the people are, what difference would this make in the outcome?  Assuming all “attending college” respondents were full-time, on-campus, and every single one of them claimed to be raped, it would up the percentage by one point?

Edit:  Apparently, the head-to-head comparison shows that students in college are less likely to be raped or sexually assaulted, so the campus numbers excluding those college outsiders would likely be lower, not higher.

Still, there are advantages to the CSA approach. Although its sample is small, it is tightly focused on full-time students at traditional brick-and-mortar universities. Full-time students who live on campus are the students at the center of the national conversation about college sexual assault. Previous research shows that female students who live in dorms or sorority houses are more likely to be sexually assaulted than female students who live with parents or in off-campus residences. The CSA may have come up with a higher estimate of victimization, in part, because its student respondents at traditional colleges were more vulnerable.

So ignore the study with a huge sample size in favor of a study with a miniscule sample size, because maybe it could possibly make a tiny difference?  Except the CSA study wasn’t only monumentally flawed because of its size, but because it neglected to make the necessary definitional distinction to merit credibility.

Yoffe critiques the CSA for asking respondents about low-level sexual assaults, including “rubbing up against you in a sexual way, even if it is over your clothes.” The NCVS also inquires about a broad range of sex crimes, ranging from rape to unwanted touching. Where the two surveys differ is in how specific and graphic their questions are. The NCVS simply asks if respondents have been the victims of “rape” or have been forced to engage in “unwanted sexual activity” such as “grabbing, fondling, etc.”

The fault with the NCVS study is that it involves, you know, actual, if similarly vague, touching?

The CSA is much more detailed. It inquires about sexual penetration with body parts or objects in specific orifices. It also prompts women to reflect on whether they have ever had sex while they were unable to consent due to being “passed out, drugged, drunk, incapacitated, or asleep.” By excluding questions about the victim’s substance use, the NCVS may have missed some crimes. Separate research suggests 44 percent of American college students binge drink, and that binge drinking is associated with unwanted sexual contact.

Aha! Finally a difference we can sink our teeth into, the NCVS didn’t specifically mention drunk (which can happen to college students), so respondents may have assumed it only meant rapes while stone sober, as opposed to the CSA which meant rape after both students had a few beers, which means the female student is not responsible for her actions and the male student in otherwise consensual sex raped her.  Well then. What comes of this ridiculously strained attempt to spin a valid study, in contrast to one that has been so thoroughly debunked as to render anyone attempting to use it to bolster the campus rape hysteria (yes, President Obama, that would mean you), is that Goldstein has forsaken any claim, either for herself or as a representation of the Messiah du jour, of fairness or neutral credibility in an effort to play spinmeister for the forces of campus rape advocacy. The fact of this new National Crime Victimization Survey is huge, which explains why no one is talking about it. That the Marshall Project spent its political capital attempting to negate its findings reflects the sad but clear fact that it’s engaged in banal feminist advocacy, no matter what its press release says.

Update:  Via Pete Metrinko, The Hill notes the NCVS study with the headline “DOJ: 80 percent of campus rapes unreported.”  This is a logical fallacy. While the negative results of a survey, that respondents state they were not raped or assaulted, produces a valid response, the positive response, that they were raped or sexually assaulted, does not prove the claim that they were.  They may feel that they were, but rape and sexual assault are crimes, that exist upon conviction.

Had the headline read, “80 percent of claimed campus rapes (or sexual assaults) unreported,” it would be accurate. That said, the use of that headline in light of the report is absurd and reflects unmitigated bias.

Update 2:  Mark Bennett has tried his hand at re-crunching the numbers, using assumptions (such as 90% of all rapes and sexual assaults go unreported) that are far more generous than reason would suggest is appropriate, multiplied the numbers by 5 to assume that it is cumulative over a 5 year college career, the best he can come up with is a rate of 3.5%.  Interestingly, he notes:

The “one in five” number will probably never go away. It’s been challenged and debunked before, and still it gets trotted out as gospel truth. But the actual numbers are much smaller and would be smaller still if respondents got to define “sexual victimization” themselves.

When I first read this, I thought Mark had it backwards, but surprisingly to me, it turned out to be counterintuitive:

What the studies have done is said, in effect, “sexual victimization includes an unwanted come-on,” defining sexual victimization more broadly than people define it in everyday life.  “[I]n a comparable survey, the federally sponsored 2007 Campus Sexual Assault study, two-thirds of the women classified as victims of drug- or alcohol-induced rape and 37 percent of those counted as forcibly raped did not consider the event to be a crime.”

In other words, the studies are broader and more expansive as to what constitutes “sexual victimization” than normal respondents would be if left to their own definitions.  Most women would not include an “unwarned come-on” as sexual victimization, but rather as something that just happens in the ordinary course of life and no big deal.

34 thoughts on “The Marshall Project Takes A Swan Dive Into The Bidet (Update x2)

  1. Beth Clarkson

    Regarding sample size, I’m afraid that your perception that bigger is better is unfounded. Accurate survey results are far more dependent on the how representative the sample is compared to the population of interest. A larger sample, once beyond the minimum required for sufficiency, only results in higher precision (smaller confidence intervals) rather than a more accuracy (less bias).

    Regarding the percentage of traditional versus non-traditional students, they may form a very substantial portion of the respondents. It’s not appropriate to dismiss them in inconsequential. On the other hand, it’s also not appropriate to dismiss their experiences as college students simply because they are non-traditional students. For example, it would be reasonable to assume they are more likely to attend night classes and more likely to attend outreach classes taught off campus in neighborhoods that may be far less safe than a campus location. Are these women less deserving of having their experiences counted?

    1. SHG Post author

      A larger sample, once beyond the minimum required for sufficiency, only results in higher precision (smaller confidence intervals) rather than a more accuracy (less bias).

      It’s always good to have a statistician around when you need one. Can you explain the difference between these, as it would appear to me that higher precision isn’t a whole lot different than more accuracy.

      1. Beth

        It’s easier with pictures, but I’ll do my best.

        Accuracy versus bias has to with whether the expectation is centered on the true value or falls consistently to one side or the other.

        Precision refers to how close you expect the results to be to the true value.

        Imagine a dart board with concentric circles. High precision and high accuracy with result in a series of hits clustered in the bulls eye. High precision but poor accuracy will result in a series of darts clustered together somewhere between the bulls eye and the edge. Good accuracy with poor precision will result in darts spread evenly around the dart board but averaging to the bulls eye. Poor accuracy and poor precision implies the darts are missing the dart board all together and landing randomly on the wall to one side.

          1. Dave

            A coworker here tells his son that he better be good at math (and science) because if he isn’t he is going to get stuck going to law school like dear old dad. His son’s response, after coming to a take your kid to work day … “Dad, your job is boring. . He hasn’t been to one of those days since.

      2. delurking

        “Precision” in this context is a statistical term of art, pretty well removed from its conventional meaning. “Accuracy” is also a term of art, but it has pretty close to the conventional meaning.
        If I screw up my survey plan and end up interviewing 5000 cross-dressing male immigrant republican state senators, I can run through some math and report on the precision of the survey results (which is a number related to how many people I interviewed and how spread out their answers were). It may or may not be particularly relevant to the real world.

        The accuracy of the survey results is basically a measure of how closely the survey results match a hypothetical survey taken of the entire population I am extrapolating to. Usually, it cannot be known, it can only be estimated.

    2. Gary Hochman

      While I agree with Beth’s comments on sample size, there were two sampling issues with the CSA study I found far more trouble than the size that Scott mentioned.

      The first, is that the study only looks at 2 schools. The lead author is even quoted as saying “In no way does that make our results nationally representative” in the Slate piece.

      The second, and potentially more troubling, is the potential response bias. While potential particants were randomly selected, they were then sent an email asking them to take the survey as well as a description of what the survey was about. Having been in college during part of the survey period, I remember the huge volume of emails recieved and would not be surprised if people who cared about the subject because they experienced a sexual assault were more likely to respond, while many others just ignored the email. Indeed they say in the study that they had a low response rate to the emails. They made an attempt to look for response bias in 3.2.2, but I found it lacking.

        1. JRF

          Exactly, this is what is often derisively referred to as SLOP; self selected listener opinion polls. You would need an extremely high response rate to even consider this seriously.

  2. JohnnyWalkerPurple

    When did you stop talking about law and start engaging in banal rape apologism and misogyny?

  3. JohnnyWalkerPurple

    Yeah, because we all know that all incidents of rape are reported, especially to bored surveyors.

    1. SHG Post author

      There is nothing “rape survivors” want more than to let others know of their suffering when there is no downside to it. If anything, this was an opportunity for every women who felt any need whatsoever to let it out. Even to “bored” surveyors.

      By the way, your weeping at the injustice of it all doesn’t entitle you to multiple threads here, no matter how emotional you are. Save the spin for the choir.

  4. delurking

    A quantitative point: the NCVS reports on a per-year basis, so it is about 2% over a 4-year college career. The relevant comparison between the studies is between 2% and 20%, which is only a factor of 10.

    1. SHG Post author

      I’m not sure it’s cumulative, but I’m a lawyer so I leave the interpretation of statistics to more knowledgeable folks.

      1. David M.

        It isn’t cumulative, it’s 1 – 0.994 (the likelihood of not being raped in a year of college) ^ 4 years of college, or about a 2.37% chance of being raped.

      2. bCohn

        Rigorously it’s not quite cumulative, but it’s probably close enough for a comment section post. What needs to be done is the likelihood of being sexually assaulted if you’ve been sexually assaulted needs to be determined, then you need to do some analysis to estimate the number of double-counted people. But if we assume that the probability of being sexually assaulted doesn’t jump dramatically, the probability of someone being assaulted multiple times is around 10^-3 or 4. A difference of less than 0.1% when talking about a couple percent is good enough for the internet.

        Also, most every survey asks slightly different questions, which can have very large effects on the results, and as this study mentions the best way to ask these questions isn’t known yet. The questions in the NCVS survey were written in order to distill responses down to criminal cases of sexual assault and rape. It also only asked student between 18-24, which are 32% of students in colleges. The NISVS asks about ‘unwanted sexual contact’. It is a matter of judgment to decide which of these interpretations or methodologies to use, but from a public policy perspective it is not clear that the first is the more correct one to use.

        1. SHG Post author

          The questions in the NCVS survey were written in order to distill responses down to criminal cases of sexual assault and rape.

          I view this as a feature, not a bug. Words need definition or their use is worthless. Much as it pains feminists, the words rape and sexual assault cannot mean whatever anyone wants them to mean, or it renders any response just as meaningless as the words.

          It also only asked student between 18-24, which are 32% of students in colleges.

          I’ll assume the validity of the 32%, though that strikes me as quite low. It was my understanding that the question of attending college was asked of all respondents, regardless of age, but I’ll put that aside as well. But what strikes me (based on the recent internal MIT survey) is that this is the core group “at risk” for sexual assault and rape, and if any group was going to spike the numbers, this would be it.

          The NISVS asks about ‘unwanted sexual contact’.

          Stare rape. Sexual harassment by the undesired asking someone on a date. Got it.

          1. bCohn

            “Words need definition or their use is worthless. Much as it pains feminists, the words rape and sexual assault cannot mean whatever anyone wants them to mean, or it renders any response just as meaningless as the words.”

            I agree completely with this. The NCVS survey was asking a specific, limited question and I believe it has good data on that question. The NJCRS study with the 1/5th statistic, which is through the links in the post, points out that it is using different terminology (‘victimization’) than NCVS and therefore is asking a different question. The sin of conflating the two appears to belong, as it so often does, to reporters.

            “I’ll assume the validity of the 32%, though that strikes me as quite low.”

            That is what the NCVS survey says. Apparently they just selected the information from 18-24 year olds out of their database. I can agree with you that it might not be particularly relevant here.

            “Stare rape. Sexual harassment by the undesired asking someone on a date. Got it.”

            The NJCRS study defines sexual contact as “touching; grabbing or fondling of breasts, buttocks, or genitals, either under or over your clothes; kissing; licking or sucking; or some other form of unwanted sexual contact” which seems decent at first blush. They asked about “stare rape” in a different part of the survey, but it wasn’t used to generate the topline number.

            1. SHG Post author

              That the catch-all, “or some other form of unwanted sexual contact,” particularly the word “contact,” doesn’t specify actual physical contact, leaves too much to the imagination. That they specifically asked about stare rape at all makes me very sad indeed.

              Thinking back to my youth, I now realize that I was a serial stare rapist. On the bright side, it was, as I recall, reasonably well received.

          2. EH

            The 1/5 study is subjective: it is designed to focus on how many people in college feel like they have been sexually assaulted. The recent study is objective and is designed to see if facts meet some objective criteria.

            Unfortunately the folks talking about it are either unintelligent or–more likely–highly intelligent but dishonest.

            That is why they tend to deliberately conflate subjective and objective. And that is why they keep playing semantic tricks, like using “rape” in one situation, and “sexual assault including unwanted approaches” another.

            It’s even more frustrating because subjective information is still very useful in the hands of intelligent folks. If 20% of your student body have a similar problem which upsets them, then it is probably worth discussing. Maybe it can be fixed and maybe not, but it’s worth a chat. And I’d happily discuss the fact that 20% of college women feel like they have a sexual assault issue, if I could trust my opponents to speak honestly.

            1. SHG Post author

              You start out soundly, then devolve.

              If 20% of your student body have a similar problem which upsets them, then it is probably worth discussing.

              Discussing what? The need for therapy, psychotropic medication, growing up, getting over every unwanted glance, narcissistic tendencies? It may be worth discussing, but hardly in the way you think. Maybe a good smack as in Moonstruck will serve far better than any discussion.

            2. Myles

              Are you suggesting that because 20% (more between 15% and 18%, but let’s not get too technical) of female students “feel” as if they’ve been raped or sexually assaulted even though they have not, in fact, suffered either, there is some sort of obligation to validate their feelings or that it would somehow be sound policy to pander to people who suffer false sense of victimization?

              Like SHG, I can’t begin to fathom what one would do about them other than therapy to figure out why they suffer from victimization delusions.

            3. EH

              Discussing means what it means, no more, no less. “Talking about see see what comes out of the discussion.” Not “agreeing to take action;” not “Doing Something About It;” and certainly not Taking It Very Seriously. It just means talking about it a bit without preconceived outcomes.

              After all, there aren’t that many groups in college who can consistently identify something that is a problem for a large hunk of the population. When there is such a large group problem–be it parking, food, long wait times for Psych 101, or what have you–then it’s usually worth having a discussion to see if there’s a non-obvious solution to it.

              The result might be “we should start the freshman year with a lecture on how to reduce the likelihood of becoming a victim;” or “we think we should generally crack down on college drinking;” just as it might be “after some discussion, we don’t think this is a problem which should be institutionally addressed at all.”

              I can easily see how folks might be inclined to dismiss even talking about it, given the copious bullshit spewed by the rapists-are-everywhere crowd. But IMO that is a bit extreme.

            4. SHG Post author

              Your idea of discussing it seems to be large part of the cause of it. Start them with a speech about rape and watch the paranoia and victimization blossom. Feeding falsehoods and a victim culture is hardly a solution to ending it so these kids can move on with a happy college experience untainted by their feelings of victimhood.

  5. Wes

    The ‘believe all accusers’ people are missing a couple things I think, whether willfully or not.

    One, all the surveys about unfounded or falsified rape claims are about *reported* claims, whether to police or a college administration. So whatever number is trotted out for the percentage of falsified rape claims would not apply to general accusations of rape that have not been formally reported, like the UVA case. The percentage of false accusations of sexual assault that are not reported hasn’t been studied, and is probably unknowable.

    Two, even if you accept the idea that 98% of all sexual assault accusations are true, this would not mean that every accusation has a 98% *chance* of being true. A basketball team with a 70% winning percentage doesn’t automatically have a 70% chance of winning its next game. Their chances depend on the strength of the opposing team. Every accusation still needs to be evaluated on its own merits, based on the unique facts of the case.

    1. SHG Post author

      That’s always a problem when seguing between stats and an actual case. The stats have one purpose, but mean nothing as to any individual instance, which stands or falls on its own merits.

  6. DHMCarver

    “Yoffe heaped particular scorn on one of the common talking points in the campus rape debate: the claim that one-fifth of all female students have been victimized.” This sort of statement by advocates like Goldstein drives me nuts, the easy segue from “victimized” to “rape”. This is the core problem with this discussion of “campus rape” — what counts as rape? Advocates for the 1-in-5 number slide easily from broad definitions of sexual assault claiming 1-in-5 women have been assaulted to 1-in-5 women in college have been raped. For people like Goldstein, it often seems that any act from unwanted touching to penetration is rape. (I heard Yoffe on the Diane Rehm show the other day on a panel with a Diane Rosenfeld, who teaches at Harvard Law — and, chillingly, is former Senior Counsel to the Violence Against Women Office, U.S. Department of Justice –, and the lawprof’s loose definition of terms and constant switching between supposed assault and rape was an embarrassment to lawyers and academics.) If you want to craft laws to solve a problem (or a perceived problem), you better be damn sure you are defining your terms with precision. I realize there are many who are happy to cast a wide net, and the bycatch be damned, but I am not one of them.

    1. SHG Post author

      This is why I keep harping on the definitional problems of using rape and sexual assault untethered from definition. It appears that every person is now allowed to define the words in whatever fashion suits them. When I used to read the word “rape” in a story, I knew what it meant. I now have no clue what they mean.

  7. Nigel Declan

    I can’t comment much on the stats or the methodology, but what is clear is that Dana Goldstein might as well just come out and announce that she disagrees with the newer study because it doesn’t confirm her pre-existing beliefs and biases. If one-in-five students were sexually assaulted while on campus then, by God, one-in-five students were assaulted, your newfangled datums be damned.

    1. SHG Post author

      There are no shortage of reporters/writers advocating the same position as Goldstein. What makes her special, and worthy of singling out, is that she writes for the Marshall Project, which has positioned itself as the gold standard of intelligent and neutral reporting on criminal justice. And many people have not merely accepted their claim, but embraced it, including my pal Radley.

      So this is the new gold standard? Self-proclaimed saviors always scare me. Those who tell me we’re kinda on the same team always scare me. The Marshall Project scares me. I can argue with guys like Bill Otis all day, and he’s so far off the rails that I have no fear any knowledgeable person will take him seriously.

      But I am deeply afraid that people will take the Marshall Project seriously, and that it gives the appearance of being more fair and reasonable while being every bit as ignorant and misguided as anyone else. It is the gravest threat to reform we face, because it is being accepted as credible when it has done nothing to merit credibility.

  8. ImARapist

    As a middle age married man (living in Ca) who will be attending college in the fall, I take issue with the fact that under the current rules I am a rapist. If my wife and I go out wine tasting and come home to indulge in carnal relations then I am a rapist. If my wife states that she would like it if I (insert sexual action) while she is asleep then I am a rapist. If I walk up to her and grab her (insert beast,ass,v-gay) when she is not in the mood then I am a rapist.
    The current standards means that a wide swath of acceptible behaior between a married man and his wife would lable the man as a rapist.

  9. Pingback: Papercuts of Oppression | Simple Justice

Comments are closed.