A friend of mine pointed me toward a 2011 article Women’s Equality Day: What the Heck Do I Tell My Daughter? this evening. I found it an interesting read but it contains a sentence that really caught my attention, “Further, studies have found that with EQUAL resumes, women with children are up to 100 percent less likely to be hired than women without children.” This is a shocking statement. It is saying that studies have found that mothers will not be hired if there is an equivalently qualified nonmother. (Okay, it says “up to,” but let’s ignore that for now.)

A sentence like that needs a citation and the author, Kristin Rowe-Finkbeiner, gives us a link to a 2007 article Motherhood and the math factor: Sociologist Shelley Correll exposes biases that affect women in business and academia by Franklin Crawford. This earlier article contains a very similar sentence, “In [Correll’s] studies, when two equally qualified candidates—one childless, one a mother—applied for a job, the mother was 100 percent less likely to be hired.”

Crawford’s article does not provide a citation for the claim and indeed, two sentences later, Correll is quoted as saying, “[Mothers] were also offered $11,000 a year less pay, on average, than an equally qualified childless candidate.” These two sentences seem to contradict each other: If mothers are not being offered any jobs for which there are nonmother applicants, then certainly they could not be offered less money for the job.

From Correll’s Stanford website, I surmised that the study being discussed in Crawford’s article is Getting a job: Is there a motherhood penalty? in the American Journal of Sociology 112(5):1297–1338. I believe this (update: removed broken link) is a preprint of that article (published version is behind a paywall here).

The first thing to note is that the study described here is not using real hiring data. Instead, 188 participants evaluated fictitious job application material for a “mid-level marketing position” in a company. The experiment is described starting on page 13 of the preprint. (Particularly interesting is the clever way in which they noted if an applicant was a parent described at the top of page 18.) Table 1 on page 43 gives their findings. The relevant entries are in the “Proportion recommended for hire” row at the bottom of the table. In particular, 84.0% of the nonmothers were recommended for hire compared to only 46.8% of the mothers.

Contrary to the assertion that mothers are 100% less likely to be hired, the data shows that mothers are 44% less likely to be hired. This is quite a different statement and so I’m left wondering how this mistake came about and why it was repeated several years later.

My best guess is that the sentence originally read something like, “In her studies, when two equally qualified candidates—one childless, one a mother—applied for a job, the childless candidate was 100\% more likely to be hired.” This statement is slightly imprecise, but accurate. Continuing my guess, after the sentence was written, it was later rewritten to avoid repeating “childless candidate” or to make it sound more negative by phrasing it as mothers being less likely to be hired as opposed to nonmothers being more likely to be hired.

At the end of the day, it seems clear to me that writing “x is y% more/less likely than z” is likely to lead to confusion and should probably be avoided. Giving the actual numbers would have prevented this (admittedly minor in the grand scheme of things) mistake.