Soon after the allegations against Hollywood producer Harvey Weinstein broke, a campaign started on social media under the hashtag #metoo. Thousands of woman posted the hashtag, sometimes along with claims that they, too, had been sexually harassed or assaulted, and occasionally even descriptions of the alleged attacks.
Many people have been quick to jump to an apparently simple conclusion: there is an epidemic of sexual harassment and assault in our society, and the #metoo campaign is lifting the lid on it. Women have been putting up with much more sexual assault than most of us had ever suspected. But the inference is too quick, and faulty.
Let’s begin with what should be a pretty elementary point about polling. In the first class I ever took in the social sciences, the professor began by stressing the importance of good experimental design, and by trotting out a few well-known examples of shoddy methodology. His go-to example of a badly designed survey was a self-selecting online poll, a type of survey that’s so famously unreliable that it even has its own acronym, SLOPpy.
Why are SLOPpy surveys so misleading? Well, it’s mainly because of the self-selecting nature of them. Imagine a radio DJ asking listeners to call in if they’ve ever had a problem with dogs. The people who’ve been bitten by dogs would doubtless jump at the chance, and they’d all call in; most of the population, on the other hand, would be bemused and wouldn’t bother. You’d then get the result that the vast majority of callers reported problems with dogs, even though they might be a tiny majority of the population.
At least with the radio show example, people would actually have to go to the trouble of calling in. With an online poll, the costs in terms of time and effort drop dramatically. Anyone who’s even had a problem with a dog – or just happens not to like them – could simply tick a box and boost those numbers of ‘people who’ve ever had a problem with dogs.’ To an untrained observer, it might well look like we have a problem with dogs that’s never been recognized before.
The #metoo movement is a perfect example of a SLOPpy poll. Women who have claims of harassment make them, and the rest of them aren’t asked about their experiences. The result is that we get a very skewed picture of the extent and frequency of sexual harassment. This is made worse by the argument that some feminists have made about the women who haven’t posted. It’s not that they haven’t been harassed, they say – it’s just that they haven’t spoken up (or, alternatively, that they haven’t been harassed yet, but inevitably will be in the future).
To see how skewed this is a way of getting to the truth about sexual harassment, let’s flip the experiment. Imagine that men who’d been the object of spurious claims of harassment started posting about that. We’d get a lot of posts from angry men saying they’d been falsely accused, and very few by men bothering to point out that they’d never been falsely accused. As for the men who didn’t post at all, we could (to follow the feminists) argue that it’s not that they hadn’t been falsely accused – it’s just that they haven’t spoken up yet (or of course, that they haven’t been falsely accused yet, but inevitably will be in the future).
The use of SLOPpy polling to make large claims about sexual harassment isn’t just confined to the informal realm of opinion. This year in New Zealand a self-selecting poll about sexual harrassment in universities made it onto the first item of the news and into a column by a university vice-chancellor. I won’t dignify the poll by repeating its conclusions, but it found, unsurpringly, that a good percentage of the people who felt motivated to volunteer for a survey on sexual harassment had complaints about sexual harassment.
These problems are just consequences of the type of polling that #metoo effectively represents. But there are other problems too. If my Facebook feed is anything to go by, the movement conflates serious crimes (coerced intercourse) with polite come-ons (a taxi driver giving a woman his card). To look at all the #metoo posts and come to any conclusions about the frequency of sexual assault would be like counting up all the posts that have anything to do with fighting (from pub-fights to a friendly punch on the shoulder) and coming to conclusions about the level of violence in our society.
This bring us to a final problem, which is the huge expansion of the meaning of terms like sexual assault and sexual harassment. I’ve heard the term used recently to refer to a man asking a women up to her hotel room; a man trying to kiss a woman outside a nightclub; and a man telling a woman that she looked beautiful. I’ve even heard a man asking a woman if she had a boyfriend described as ‘inappropriate.’
The fact that such things are part of the #metoo movement means that it isn’t really a good indicator of serious sexual misconduct. It’s not good evidence that we are experiencing an epidemic of rape or groping. But it may be evidence that some women now believe that a man telling her she looks beautiful constitutes harassment.
Few people, I think, really believe that a man who puts his hand on a woman’s waist, only to remove it at the first sign of discomfort, is committing serious sexual assault or harassment. But the small minority who do are much more vocal than the rest, and this effect is facilitated and magnified by social media.[pullquote]
There’s no real evidence that there’s an epidemic of sexual assault or harassment. A small minority of activists want us to believe that SLOPpy polls are scientific studies, that a touch on the knee is part of the same continuum as rape, and that sexual harassment is anything that might make a woman feel uncomfortable, now or in the future. We shouldn’t listen to them