Democracy Has Prevailed.

November 7, 2011

Before This Gets Out Of Hand

Before saying anything else, let me first say that this is NOT a defense of Herman Cain.  This blog post won't take any position on his alleged sexual harassment, though I agree with Bill Green from this week's 4802 (abouth 4:30 in) when he said that his campaign's response to the charges show a level inexperience that's damaged his campaign.

No, this is about something I came across at Thinkprogress:
Herman Cain’s many apologists are deeply and tragically wrong. A recent University of Michigan study found that nine in 10 women workers will experience “offensive sexist remarks or being told that they could not do their job properly due to their sex,” and the women who only endure these conditions will be the lucky ones. A massive one in 10 women in the workplace will at some point be “promised promotion or better treatment if they [are] ‘sexually cooperative‘” with a co-worker or supervisor.
Clicking the link will lead you back to this article from the Daily Mail in the UK.  It starts with this:
Nine in ten women have suffered some form of sexual discrimination in the workplace, a study has found.

A vast majority of women workers have experienced ‘gender harassment’, which includes offensive sexist remarks or being told that they could not do their job properly due to their sex.
That's horrible!  Just horrible.  Too bad the study cited doesn't say what the Daily Mail says it says.  We can snag a few clues from the Mail's next two paragraphs:
The researchers at the University of Michigan found that 10 per cent of the women surveyed had experienced the most severe form of harassment, in which they were promised promotion or better treatment if they were ‘sexually cooperative’.

The study questioned women in two male-dominated environments – the US military and the legal profession. It found that although few were subjected to actual advances, such as being groped, 90 per cent had been subjected to gender harassment. [emphasis added.]
And that's the first objection. The women surveyed worked in two "male dominated" workplaces.  Can we logically extrapolate conclusions of this study to the work place as a whole?  I don't think so.

But let's assume we can - there's still the problem of  the study itself and how it doesn't say what the Daily Mail (and Thinkprogress) says it says.

Let's take the military survey first.  Here's how the study describes it:
Study 1 involved secondary analysis of survey data collected by the U.S. Military. This survey began with a non-proportional stratified, single stage random sample of active-duty members from all branches of the U.S. Military (Army, Navy, Air Force, Marines, and Coast Guard); women and people of color were oversampled. The initial sample contained 60,415 individuals, of whom 53,170 were deemed ‘‘eligible’’ (reasons for ineligibility were various, such as inability to locate the sample member). These individuals were invited to complete surveys either on paper or online, and 19,960 usable surveys were returned (38% response rate). The current study focused on the 9,725 women who responded to the survey. (p 5)
9,725 surveys completed.  Remember that number.  But we're getting ahead of ourselves.  The study uses a four tier assessment of behavior; (1) gender harassment: sexist, (2) gender harassment: crude, (3) unwanted sexual attention, and (4) sexual coercion.

And here's how they're defined:
The gender harassment: sexist subscale measured treatment that conveys explicit antipathy toward members of one gender. The subscale consisted of four items, such as ‘‘made offensive sexist remarks (for example, suggesting that people of your gender are not suited for the kind of work you do)’’ and ‘‘referred to people of your gender in insulting or offensive terms.’’ Four items also assessed experiences of gender harassment: crude behavior; although sexual on the surface, this behavior expresses animosity rather than attraction. Examples included: ‘‘made offensive remarks about your appearance, body, or sexual activities’’ and ‘‘made gestures or used body language of a sexual nature that embarrassed or offended you.’’ The unwanted sexual attention subscale consisted of six items, including ‘‘made unwanted attempts to establish a romantic relationship with you despite your efforts to discourage it’’ and ‘‘touched you in a way that made you feel uncomfortable.’’ The sexual coercion subscale contained four items, e.g., ‘‘implied faster promotions or better treatment if you were sexually cooperative’’ (p 5)
And here are the results:
The largest group consisted of women who reported the lowest levels of harassment (Group 1; n = 3,933). As seen in Fig. 1, the experiences they described almost exclusively consisted of sexist behavior. The second-largest group (n = 1,161) contained women who had encountered both subtypes of gender harassment—sexist and crude— but very little unwanted sexual attention or sexual coercion (henceforth, this group will be referred to as Group 2, the ‘‘Gender Harassment’’ group). Group 3 (n = 429) disclosed episodes of unwanted sexual attention, in addition to moderate levels of sexist and crude behavior. Group 4 (n = 138) had encountered moderate levels of all types of harassment measured by the SEQ-DoD. Group 5 (n = 37), the smallest profile group, reported the most frequent harassment on all four subscales. In sum, 89.4% of harassment victims fell into Group 1 or 2, which described experiences of gender harassment but virtually no unwanted sexual attention or coercion. (p 7)
This is where the 9 out of 10 stat comes from.  And the 1 out of 10 stat as well.  But did you see the (n = ...) numbers? Do you have a calculator? Add them up. You'll get 5,698.

But wait, there were 9,725 surveys used, right?  What happened to the remaining 4 thousand?

The answer is found on Table 2 (p 8).  There were 4,014 non-victims.  Out of the 9,700 or so women surveyed, 4 thousand of them did not suffer any sort of harassment.

The reporting, however, says it's 9 out of 10.

It's just plain wrong.

9 out of 10 of those who said they'd been harassed in some way were harassed with "virtually no unwanted sexual attention or coercion."

And so that 1 out of 10 who'd suffered through some sort of unwanted sexual attention or coercion is really only 1 out of 10 of the 57% who'd been harassed.

So the Daily Mail, by simply misreading the numbers, simply doubled the size of the problem.

Let's take a look at the same numbers in the other survey.  In this second, smaller survey, 1,425 women's surveys were used (p 9).  Of those 851 are listed as "non-victims."  That's about 60%.  So it's the remaining 491 that reported harassment.  If we follow the same pattern as the military survey, that means that 92% of those surveyed, suffered harassment with no unwanted sexual attention, 8% were.  If my math is correct, that's about 2.5% of the total number of surveys.

That's hardly 1 out of 10.

The Daily Mail and Think Progress got this wrong.

None of this should be construed by anyone to think that I think sexual harassment isn't a big deal.  It is a very big deal.

But if you're going to write about it honestly, you have to get the numbers right.  You simply have to.

1 comment:

EdHeath said...

Really, I don't have a lot of confidence in those studies in terms of identifying the amount of sexual harassment that exists in the workplace. A sample of size of 60,000 and they throw out 53,000, yet still end up with about 20,000, with about half the group women? My confidence that the sample is random is not high. The second group, the attorneys, they mailed surveys out, and only about half of the group responded. That could mean that the other half didn't experience any harassment, or that they were too embarrassed to talk about it. Again, not much confidence that we are getting a random sample.

Like you said, it doesn't mean sexual harassment in the workplace or anywhere is somehow not a bad thing, or that it is either not common or that it is common. From those studies all that can be said is that it does exist, but we really don't know in what percentage.