Facebook is currently being sued for housing discrimination. That may seem like a leap, seeing that Facebook isn’t actually doing any real estate deals, but Facebook does run ads for housing. Canadian lawyers are complaining that Facebook’s employment ads discriminate against older workers. Researchers ran fake Facebook ads to test whether they could manipulate ads to show up for men and women in different proportions from what they specified when they set up their ads. They could.
Should this affect your decision to use Facebook ads or not?
Is it Facebook or advertisers?
A lot of the things people get cross with Facebook for are actually things people are choosing to do with Facebook. You beguile your time with Facebook quizzes? Facebook is going to show you quizzes. You upload your photos and agree that Facebook has a license to use them (yes, you did agree to that)? Facebook is going to do what it wants with your pictures.
When you set up a community to show your Facebook ads to, the ethnic heritage of the people isn’t one of the choices. But you can pick location, down to the zip code. You can also choose people who are similar to the people who like your page, and their friends. With a little research, this could allow an advertiser to skew their audience toward a particular ethnic group.
I’m pretty confident that our cause marketing campaign at Facebook sometimes inadvertently target specific ethnic groups. I hadn’t thought about it before, but on reflection, there are plenty of ways to focus on a given demographic. Of course, that’s also true with legacy ad placements in newspapers, magazines, TV, radio, and billboards.
And Facebook doesn’t give those targeting options when it comes to ads for housing, employment, or credit.
But Facebook’s algorithm also skews the way it delivers ads, regardless of advertiser choices. A new study shows that Facebook’s efforts to show viewers relevant ads based on their behavior ends up showing certain ads to certain groups. The researchers created ads with content, including images, that would stereotypically be associated with men or women, white people or black people, etc. Then they ran these ads with identical neutral targeting.
High budget ads, they determined, are slightly more likely to be shown to men. Want ads for lumberjacks tend to be shown to men, too. Ads about hip hop are shown to black viewers and ads about country music are shown to white viewers. Apart from the ad budget issue, this seems reasonable. Grizzlyrose reports that “70% of nonwhites surveyed reported listening to country music at least weekly,” but Brandon Gaille says 90% of country music fans are white. I’m not a country music fan myself, but I guess an advertiser who wanted to target white consumers could choose “country music fans” as a euphemism.
Is it discrimination?
Presumably, African-American country music fans choose content related to country music when they are on Facebook. They listen to country songs, click through on ads for cowboy boots, slow down for country music celebrity gossip posts, and otherwise show their fondness for the genre. Facebook no doubt shows them ads related to their country music preferences.
These Facebook users are a smaller proportion of Facebook’s country music fan community than the white people behaving in the same way. They are not treated differently from others in their interest grouping, but those country music ads will not be shown to other African-American Facebook users who don’t behave like country music fans.
I’m a white person. I never see anything about country music at Facebook, because my behavior at Facebook doesn’t give Facebook’s algorithm any reason to think I’m interested in country music. Facebook doesn’t just show country music content to all white people.
But, according to the study mentioned above, they did show housing ads to black viewers more often than to white viewers if the ad included a picture of a black family and the rental price was low. They showed the ads to white families more often if the price was higher, or if the house was for sale rather than rent. When the same ad used a picture of a white family, the price did not affect the proportion of white viewers targeted.
Your Facebook ads
If for some reason you want to target specific ethnic groups at Facebook, you can probably do so — just as you can in other forms of advertising. You can overtly target males and females, specific age groups, and people with particular interests — like country music fans.
As an advertiser, you know that there are perfectly good reasons to target specific groups of people. We’ve worked with companies that sell hair care products for ethnically mixed children, women’s athletic clothing, and workshops on conscious eldering. However open minded these companies are, they don’t want to target 17 year old white males.
The target personas for each of these companies will be different. For our company, and for many more nowadays when we have access to consumer behavior data, demographics don’t matter. We target people who do X, not people who are in X category.
But so does Facebook. No one is suggesting that Facebook’s algorithm identifies white consumers or black consumers and serves ads based on the identification. Rather, Facebook’s algorithm is serving ads based on user behavior and stereotypes that it has learned.
In the experience of Facebook’s software, most lumberjacks are male. Or at least, most viewers who act like Facebook’s idea of a potential lumberjack job candidate are male. There is no evidence that people at Facebook are intentionally preventing women from seeing ads for lumberjacking jobs.
If you want to be sure to reach a broad audience, don’t choose images associated with a narrow audience. If you want to reach a diverse audience, don’t target zip codes with a homogeneous population.
On the other hand, if you want to reach a specific population and you figure out how to manipulate Facebook ads to achieve that goal, don’t blame Facebook.