In late 2019, I discovered that I was part of Facebook’s gender bias problem.
This was not a possibility that had previously occurred to me. I am an editor with a helicopter industry magazine, Vertical, and have been managing our Facebook page since I created it in 2009. I’m also a female helicopter pilot and flight instructor myself, with no appetite for the blatantly misogynistic posts I’ve seen on some other aviation industry pages.
For many years after the launch of our Facebook page, a consistent 10 percent of our followers were women. This might not sound like a lot, but it’s at least twice the percentage of female pilots and maintenance engineers in the industry. I naively believed our Facebook page — which is overwhelmingly just cool photos and videos of helicopters — was helping to expose more women to the possibility of a career in aviation.
A few years ago, however, Facebook began dramatically limiting the reach of business pages, including ours. Now, if I want to reach more than 10 or 20 percent of our roughly 300,000 followers, I have to pay to “boost” our posts as ads. We’ve always posted high-quality content that is generally rewarded by Facebook’s algorithms, so I don’t have to pay much: a few dollars per post usually does the trick.
I’ve always been proud of the fact that we’ve grown our page organically, without paying for followers; those 300,000 people actually want to hear from us. So when I started boosting posts, the target audience I selected was simply “people who like Vertical Magazine.” I didn’t bother analyzing the results of the boosts beyond their overall reach, which was presented to me alongside each post in my admin view.
Then, in 2019, I read about research showing that Facebook’s ad delivery system can have discriminatory outcomes even when advertisers aren’t trying to discriminate. That prompted me to take a closer look at our boosted posts. Because I knew around 10 percent of our followers were women, it seemed reasonable to assume that around 10 percent of the people seeing our boosted posts would be women, too.
Nope. I discovered that even when my only criteria for an audience was “people who like Vertical Magazine,” Facebook would routinely show my boosted posts to an audience that was over 99 percent male. Double-checking our demographic stats, I also found that our percentage of female followers had edged down to nine percent, which I had to assume was related to Facebook preferentially showing our posts to men under its new distribution paradigm for business pages.
Early last year, shortly before Covid began shutting down the U.S., I reached out to two of the researchers who had been studying this phenomenon, Muhammad Ali and Piotr Sapiezynski of Northeastern University. I explained my dilemma and asked them how I might be able to reach more of our female fans. They suggested that I boost my posts to two distinct audiences: “people who like Vertical Magazine” and “women who like Vertical Magazine.” (The first, more general category was intended to capture non-binary as well as male followers.)
As it happened, Facebook wouldn’t let me boost posts to “women who like Vertical Magazine,” as that audience was too small. So instead I began boosting posts to worldwide audiences of both “people who match” and “women who match” interests of “Vertical Magazine or helicopter.” I increased my advertising budget to accommodate the new category, but still rarely put more than a few dollars behind each boost.
Covid struck and my world turned upside down, but I continued to consistently boost our Facebook posts to these dual audiences. Later in the year, once I had the bandwidth to return to the project, I reached out to Ali and Piotr again and gave them access to my advertising account to analyze the data. We experimented with a few variations on these audiences; for example, limiting them to the U.S. only to ensure that the results weren’t being skewed by countries where the population online was disproportionately male.
They weren’t. In all cases, Facebook delivered my boosted posts about helicopters to barely any women unless I targeted women specifically. And Ali made another disturbing discovery as well: that Facebook was charging me more to show my posts to women relative to their levels of engagement.
See, Facebook advertising doesn’t operate according to a fixed schedule of prices. Many factors influence how many people will see a given ad, including its quality and relevance. For the same budget, ads that generate more positive interactions (such as likes, clicks and shares) will be shown to more people than ads that generate fewer of those interactions, or that generate negative feedback (e.g. people clicking on “hide ad” to never see it again).
Ali found that the predominantly male audiences Facebook was delivering my posts to tended to engage with them about five to seven percent more than the all-female audiences I was targeting. Yet the cost to reach those all-female audiences was dramatically higher, as you can see in this graph for four example boosted posts (where CPM is cost per 1,000 impressions). In one case, the all-female audience actually engaged with the post more than the predominantly male audience, yet Facebook charged me about twice as much to reach them.
My year-long experiment in combating Facebook’s gender bias was successful insofar as our posts reached many more women than would have seen them otherwise. But I’m not thrilled about the fact that my solution to Facebook’s gender bias problem is giving Facebook more money. For now, this devil’s bargain is worth it to me to avoid shutting women out of our social community entirely, but it’s clear to me that fundamental change is needed at the platform. Although I can only point to my own data, I have to assume that this type of bias is ubiquitous across Facebook, hardening stereotypes in ways that could undo decades of real-life progress in heavily gendered industries like aviation.
Before I hear from the “men just like helicopters more than women do” crowd, let me emphasize again that I’m not seeking anything close to full gender parity. I’m prepared to accept that more men than women are interested in aviation (although I would attribute that to the complex interplay of multiple social factors, rather than biological hardwiring).
Regardless, in choosing to follow our Facebook page, a significant number of women have clearly signaled that they want to receive updates from Vertical, an important news source for the helicopter industry. Facebook’s algorithms have decided that they can’t, unless I pay the platform a ransom to enable it.
How is that fair?
