Microtargeting is a powerful tool that leaves too much room for misbehavior and disinformation.
September 21, 2020
Facebook recently announced that it would not allow new paid political ads in the week before Election Day. This is too little too late given the deluge of microtargeted ads that have already inundated American voters in the past year.
While broadcast ads are still the norm, more candidates and political players are opting for the narrowcast so that they can reach you directly in your social media feed. During the second half of the 20th century, the main way that major party candidates for the presidency got their message out to nationally dispersed electorates was through broadcast ads on radio and TV. But nearly as soon as the internet allowed, political ads have been moving online. And as platforms have become more sophisticated treasure troves of personal information, political ads are increasingly microtargeted to subsets of the population. An expected $6 billion is likely to be spent on in political campaign this cycle, and 20 percent, or $1.2 billion, will to go to digital ads.
So what’s so wrong with microtargeting online political ads? The trouble is that when they contain misinformation or disinformation, they are the mental equivalent of a drive-by shooting: They are quick. They inflict damage, and then suddenly, they are gone.
As I explained in my book Political Brands, “marketing a [commercial] brand can be microtargeted to a particular demographic, or income group, or subculture. For example, Dove soap, which targets women with images of empowerment that are ostensibly against lookism, and Axe soap, which targets men with sexist appeals, are both sold by the same company (Unilever), just to different target demos.”
The same thing can be done with political propaganda such that an untruth is repeated at a targeted subset of the electorate, until the idea is accepted as the truth. At the end of the day, Unilever wants to you to buy soap, and politicians want you to buy ideology.
In one sense, targeting key voters with tailored messages is not new. In the pre-internet age, political campaigns microtargeted particular voters using snail mail with messages that did not appear in broadcast campaigns. So for example, if the Democratic Party had a sense that a particular Republican household was environmental because they subscribed to Backpacker magazine, that household could be targeted by direct mail with messages that touted a candidate’s stance on the environment. Or if the Republican Party had a sense that particular Democratic household was strongly pro-law enforcement because they had given to the Police Benevolent Association, it could target tailored law-and-order messages to that home.
One factor that mitigated how much of this targeting by mail could occur is that postage costs add up fast. For instance, in 1984 both the Democrats and the Republicans had racked up remarkable postage fees during the election. But now, each digital ad can be sent to a voter at near zero marginal cost. And the effectiveness of targeting by street address can be undercut by how mobile Americans are. By contrast, even when a woman moves from Topeka to Tallahassee, her Facebook page (and other social media) is portable. While mail to her old house may not reach them, ads on her Facebook pages will follow her around like barnacles on a ship.
How are voters targeted? There is a cottage industry of big-data wizzes who will sell political campaigns the ability to microtarget voters. Each company has its own secret sauce, but the basic components are similar: start with the voter file that is publicly available from the state. This gives the name, address and party affiliation (or lack thereof) of the voter. Add to that other publicly available data, which in many states that will include marriages and divorces, gun ownership, or criminal records. In Florida you find boat ownership if you have enough information.
Then added to those voter files is commercially available information, which can include your web browsing history if you have not taken any privacy precautions online. And then there is the really creepy aspect that some political advertisers will try to guess your personality type. This is what Cambridge Analytica allegedly did for the Trump campaign in 2016. This was after Cambridge Analytica targeted black youth voters in in 2010 in Trinidad with a campaign that discouraged them from voting with a slogan of “Do so. Don’t vote.” The point was to try to increase voter apathy. That company is now defunct, but its owners have just moved on to start other companies.
What the voter file allows the microtargeter to do is to send targeted messages to divorced, gun-owning Florida Republicans with a habit on clicking on Fox News stories online. This is likely to very different from the message sent to single Democratic voters who click on Pro Publica headlines daily in Ohio. And platforms like Facebook enable this type of granular targeting of voters. As MapLight explained, “Not only does Facebook allow advertisers to spread political misinformation, but it also allows them to target platform users on the basis of personal information such as age, gender, education, income, multicultural affinity, ZIP code, or interests. While microtargeting can help NGOs, non-profits, and political challengers to reach their audiences in a cost-effective way, it can also enable foreign interference and voter suppression.”
Politicians are stereotyped as talking out of two sides of their mouths. But with microtargeting messages on complex issues, politicians can send out a hundred different messages on the same topic, just microtargeted to different constituencies. Think of health care — a politician could put out dozens of variations on messages on health care, just to different audiences. Microtargeting may contribute to polarization as the algorithms in most social media platforms are designed to tell the viewer what she already wants to hear — confirming bias after bias.
The less scrupulous the politician, the more dangerous microtargeting is. If the politician is willing to lie to voters, then microtargeting is gift that keeps on giving because now the politician can lie personally to each voter in a tailored and manipulative way. If the politician has no scruples, then why not tell identifiable pro-life voters that he is anti-abortion and tell identifiable pro-choice voters, that he is for reproductive rights? Tell the gun owners that he is pro-guns and tell the gun sense voters that he’s for restrictions? This isn’t a hypothetical, as Facebook has continued with its puzzling stance of letting politicians lie in paid advertisements on its platform.
In May, Media Matters excoriated Facebook for allowing paid political ads that touted misinformation about voting fraud. And a civil rights audit of Facebook criticized the platform for being “far too reluctant to adopt strong rules to limit misinformation and voter suppression.” And in July, ProPublica found political ads on Facebook making “unsubstantiated claims that House Speaker Nancy Pelosi is trying to fix the election using mail-in voting.”
Finally, another difference between direct mail targeting and digital targeting is that at least with mail, there is a physical artifact that a voter can show the media if there was a particularly misleading or obnoxious message from a political campaign. But with most digital ads, until very recently they were ephemeral.
In 2018, Google launched its Ad Library of political ads, and in 2019, Facebook finally launched a library of political ads. These are steps in the right direction, although realistically, most users won’t research the political ads that just shocked or titillated them online.
In 2020, more voters should flag troubling or peculiar digital political ads they see in their social media feeds to the media — especially if the ads seem to discourage voting.
The views expressed are the author’s own and not necessarily those of the Brennan Center.
