Over at Slate Emily Yoffe argues that universities have overcorrected
their campus rape policies and procedures to the point that they’re harming
innocent males. And though her article is a long and engaging read, its main contribution
lies in its negative critique of the statistics cited by the Obama
administration. The evidence she musters
to support her “overcorrection” claim is largely unpersuasive.
College Rape Statistics
It’s
been repeatedly pointed out that findings from college rape surveys vary widely.
A recent Bureau of Justice study finds that female college students have a .61%
chance of being raped, slightly lower than non-college females from the same
age group. This probability is dramatically lower than oft-cited one in five statistic (see also the CDC findings).
Vox recently
published a good breakdown explaining why we see such disparate findings and I
encourage you to read it. Worth noting is that the BoJ study does not ask about
incapacitated rape, which, as other studies have found, is the most prevalent
form found in colleges. Rapes are likely going underreported, which inevitably
lowers the BoJ’s estimates. The BoJ report also frames its survey questions
differently, only coding a response as a sexual assault if the victim
identifies herself as having been assaulted. This differs from other surveys
which makes coding decisions based upon legal, not victim, definitions. Whether
or not this framing leads to further under-reporting is less clear cut because
it speaks to epistemological concerns
(e.g. who has more epistemic authority, the victim or the researcher? Can it
still be rape even if the woman does not identify herself as a victim?).
Another
difference is that the one in five statistic captures a whole host of other
forms of sexual assault beyond rape as it is legally defined. The study does
not actually say that college women have a 20% chance of getting raped, but
rather that they’ll be victim to unwanted sexual contact. The CDC study, which uses a larger sample size and samples over 200 colleges finds the rate to be about 1 in 8.
So,
yes, a college woman does not have the same likelihood of being raped as a
woman in war-torn Congo. And yes, as Yoffe argues, the study making the one
in five statistic has sampling issues. It also assumes that rapes are
independent events, which other research has found to be problematic. The fact
is these studies have what are called reliability problems: we cannot establish
consistent findings.
Part of
this is due to how the survey questions are composed. But another, I think,
stems from a persistent source of uncertainty in the data that a universally
adopted questionnaire alone will not fix: the fact that rape victims, as
victims of trauma, will not be able to (1) clearly remember the facts of the
event, (2) that they will want to avoid revisiting the issue, and (3) they’ll
oftentimes blame themselves or rationalize away the behavior of the perpetrator.
Because of the nature of the crime, rape studies may very well be inevitably
plagued by a large error term.
Given the gravity of rape, the government should commission a national study, one that standardizes survey strategies and investigates ways to overcoming the reliability problems I just described.
Neverthefuckingless
To
those of you who use this fact to underplay rape statistics: don’t. The
question isn’t whether or not college rape is a problem, but how bad of a problem it is. The error I
describe above is a question of how depreciated the findings are. We’re still
faced with the fact that in a class of third graders anywhere between one to
two of its girls will be raped ten
years from now. And three to five of these girls will be sexually assaulted
when they’re in college. The only question is if it’s one, two, three, four, or
five.
Ok, so
while the data do not suggest that we have a civil war style rape epidemic, we certainly have a rape problem. Yoffe doesn’t discount this
fact. Where I’m less persuaded is in her accounts of supposedly innocent males
who are victims themselves of the government’s and the university’s overcorrection
to this issue.
I do
not give much credence to anecdotal evidence when making inferences about
overall populations. Yes, the men in her story may be getting screwed over…but
then again they might not be. But even if they are, they amount to three data
points. What Yoffe is essentially arguing is that the government and
universities are committing type II errors: they’re finding false positives.
Innocent people are suffering. It’s ironic, though, that a person who finds
statistical findings questionable because of non-random sampling procedures is
making inferences from a data set of three that was not randomly sampled. Her
case studies may very well suffer from, uh oh, confirmation bias.
Type I and Type II errors:
But let’s
talk about our errors a bit more. Let’s imagine two scenarios: in the one, the
goal is to make sure no rapist goes unpunished. On the other is to never let an
innocent male be punished. The two are in tension: because some uncertainty is
inevitable we’re going to have to make a tradeoff between one and the other
because the uncertainty requires us to hedge in one direction over the other.
To ensure that no rapist goes unpunished is inevitably opening us up to
punishing too many people. But to make sure that the innocent go free requires
us to hedge in favor of innocence: actual rapists will be missed. By hedging
against type 1 errors, we open ourselves to making type II errors and vice
versa.
The
question society needs to answer is where they wish to place the weight: toward
catching rapists or ensuring that the innocent roam free. This is not a
question that can be answered by the statistics because it is a non-statistical
question. It’s moral. Given the history of gender relations in the country,
whose interests should we be protective of even at the expense of other’s? If
the goal is to end rape then the
necessary means may simply require a higher propensity to damn the innocent. I, for one, am on the side of ending rape.
This
doesn’t seem fair, but it’s our fate. We can never be certain about the
empirical world. And because we lack certainty, some people are going to get
unlucky. The questions are: who and how?


