Women are 2.87 times more likely than men to be called pushy

My post looking at gendered descriptions of Jill Abramson has generated a little bit of attention. Notably, The Atlantic posted an article by Olga Khazan titled “Pushy is used to describe women twice as often as men”, citing my work.

I’ve been asked if The Atlantic headline is technically inaccurate. This question is based upon the astute observation that the description of “twice as often” fails to account for the fact that women are under-represented in the corpus. Indeed, in the original post I made the rather difficult to interpret observation that “Women are labelled pushy about twice as frequently as men in COCA even though men are mentioned nearly twice as frequently as women.”

Here’s a lesson in why it’s important to clearly and succinctly report key research results. I believe that the first half of that sentence has been what people focused on, and the second half has been largely ignored. Nonetheless, I think The Atlantic headline is technically an accurate representation of my findings (and in particular it uses language that I used to describe them). However, I also believe that the headline reports a way of looking at the data that virtually no one is interested in. Allow me to explain.

We can ask a question like “what are the genders of the people who are labelled pushy in these texts?” This is the question that I asked when I went about categorizing instances of pushy in my data. For this question, my answer is that pushy was used to describe about twice as many women as men, as The Atlantic headline accurately states. That is a technically correct answer to a question that interests no one.

Alternatively, we can ask a question like “how much more likely is a woman to be called pushy in our public discourse than a man?” For this question, we have to account for the under-representation of women in the corpus. When I account for this, I estimate that a woman is 2.87 times more likely than a man to be called pushy in public discourse. (If it interests you, I’ve placed an explanation of the math I did to arrive at this figure in a section below.)

The second question I believe best captures what people really want to know. Thus, it’s possible to say that a headline like this would be optimal: “Women are 2.87 times more likely than men to be called pushy“. However, I would hypothesize that the inclusion of decimal points in headlines drastically reduces the number of clicks that the article gets.

Anyway, my point is not to nitpick the way people have discussed these findings. In fact, if anyone is responsible for misleading reports of these findings, it’s me. What I wanted to do was simply put out an estimate that better addressed the questions people are asking and, more importantly, provides even more support for the arguments that writers are making about the way gender stereotypes complicate the task of being a leader for women.

The math explained

I have 116 occurrences of pushy in my data. If we assume that pushy is not gendered, then, based on the fact that women are underrepresented in the corpus I used (The Corpus of Contemporary American English), before we read through them and actually categorized them, we would expect that 43 (37%) of the occurrences would apply to women and 73 (63%) would apply to men. What I found was strikingly different (significantly different; the likelihood that I would have gotten this data given the assumption that pushy is not gendered is extremely low, so low that we can reasonably infer that the “pushy is not gendered” hypothesis is wrong).

The word is actually used for a woman 72 times and 44 times for a man in my sample. Thus, women’s rate of being called pushy is 1.67 times as much as it would be if the word were used in a non-gendered fashion (72/43), and men’s is only 0.6 times as much as we would expect it to be (44/73). Using these ratios, we can answer the question “how much more likely is a woman to be called pushy in public discourse than a man?” by dividing them (1.67/0.6). This gives us a rate of 2.87 times more likely. Thus, I estimate that a woman is 2.87 times more likely to be labelled pushy than a man in public discourse.