Suppose I made the claim that: “Only 6% of reports of rape lead to a criminal conviction, therefore 94% of rape reports are made up”. I would, rightly, be howled down for having committed a gross fallacy. It would be explained to me that accusations of rape often revolve around one person’s word against another’s, which makes them very hard to prove to the criminal standard of proof. Thus many accusations that do not lead to criminal convictions are still most likely true.
Now suppose I instead made the claim that “only 2% of accusations of rape are proven to be false, therefore 98% of accusations of rape are true”. This claim is widely made (example), yet it is just as fallacious and for the same reason. Just as it is usually hard to prove rape claims true, it is also hard to prove them false; many claims cannot be proven either way.
Yet, under Title IX codes in American colleges, the claim that nearly all accusations are true is used to justify a process that more or less presumes an accused to be guilty from the outset, with little need for due process. Afterall, if there is only a 1-in-50 chance that the accused male is innocent, then the accusation is pretty much sufficient in itself. Given that, a mere “preponderance of evidence”, defined as a 51% versus 49% likelihood, is all that is needed to expel a student from college for sexual misconduct, something that would always be on their record and likely blight their career prospects.
Yet this “nearly all accusations are true” claim has no basis in proven fact, it is purely ideological.
First, the studies it is based on don’t say 2%; for example the widely cited Lisak et al (2010) says “2% to 10%” (though that sometimes gets reported as “only 2%”).
And, again, that is only the fraction that were proven false. For the majority of cases studied no conclusion could be drawn either way. The Lisak et al researchers put cases into four categories. These were “false allegations” (6%), “case did not proceed to any prosecution or disciplinary action” (45%), “referred to disciplinary action or prosecution” (35%) and “insufficient information to be coded” (14%).
If you conclude that only the 6% that were proven to be false were actually false, then you are asserting that every one of others was a true allegation. How do you know that?
How many of those in the other three categories were actually false accusations? No one knows, I certainly don’t. If the others were all for-sure true, then why did 45% not even proceed to prosecution? How many led to actual convictions? Elementary logic tells us that if there were any way of reliably concluding that fewer than 10% of allegations are false, then we could just as readily have a conviction rate of over 90% of all allegations reported to the police.
It really is amazing how often that Lisak et al conclusion that: “the prevalence of false allegations is between 2% and 10%” is quoted as though they had established that all the others were true. The honest way of phrasing that sentence would be: “the prevalence of allegations proven to be false is between 2% and 10%”, though even that would be misinterpreted by those committing the fallacy at the head of this post.
All we really know is that the false-accusation rate is higher than the “2% to 10%”, and that it is lower than 94% (the fraction for which there is no criminal conviction). When it comes to date rape, and accusations after a consensual relationship has broken down, the truth or falsity can be very hard to establish and no-one has much idea what fraction are true. Half of them could be false for all we know. Short of putting hidden cameras into all college bedrooms, can you suggest a study protocol that would tell us the false-accusation rate for the sorts of cases that US colleges adjudicate?
To some steeped in identity politics even asking this question would be criticised. The ideological position is that we should always believe the woman, since women have been historically oppressed, and in identity-politic ideology the one further up the oppression hierarchy is the one in the right. Hence, the leap from “2% to 10% of accusations are proven false” to “only 2% to 10% are false”, or worse, “only 2% are false”.
And from there we have a process that assumes guilt from the outset. Accusers are automatically regarded as “victims” and we’re asked to always “believe victims”, and anyone who makes the suggestion that due process might be appropriate is accused of lacking sympathy for rape victims and of wanting to making things easier for rapists and sexual harassers. The three-part series by Emily Yoffe is a good introduction to what is wrong with Title IX currently.
If you were accused of a wrong-doing that could get you expelled from college, would you be happy having the outcome determined by tossing a coin? That would give a 50:50 likelihood of the outcome being guilty versus acquittal. Fair? No, of course not, everyone would be aghast at such a process.
Yet the Title IX regulations mandated by the Obama administration required colleges to use a 51:49 threshold for a guilty verdict. What’s the difference between 50:50 and 51:49? Nothing really, since no-one can assess such probabilities to 1% accuracy. Either gets things wrong half the time.
The 51:49 threshold can be justified only if “only 2% of accusations are false” were actually true. That very low prior probability of the accused being innocent would then make the overall process fair enough. And yet that claim is based on little more than ideology!
Betsy DeVos, Trump’s appointee as Education Secretary, has withdrawn the Obama-era instruction and issued new guidelines. These allow (though don’t require) colleges to use a higher threshold, one of “clear and convincing evidence”. Several other changes remove the implicit presumption of guilt and bolster the fairness of the process. Much of the US left and left-leaning blogosphere is outraged. But reading the new guidelines, this seems to me the first and only action by a Trump nominee that is actually pretty sensible.
Update: Well I knew when writing it that this post wouldn’t be popular with some. It was inspired by a Twitter interaction with @ExoWanderer (who is Jonathan Fraine, a Staff Scientist working on exoplanets at the Space Telescope Science Center). Specifically, it was inspired by his claims:
I asked for a cite for the “less than 2%” claim. I didn’t get one (anyone know of one?), but he did twist my words to assert again:
He also pointed me at studies such as Lisak et al (2010), saying:
This inspired my above post explaining what Lisak et al had actually said, and what conclusions could be drawn from it. On pointing my interlocutor at the resulting blog post, so far I have had no substantive rebuttal, but tweets do include (among similar others):
And finally a call for help:
If my above post is substantially wrong then please can someone point me at substantive evidence of that? This does seem an important issue, and I want to get it right. So far the claims from such as Jonathan Fraine seem to be purely ideological.