How Do We Get More Democrats to Vote?
Part 2: Digital
Can we be Facebook friends? Will you “like” my candidate?
Going into 2016, we weren’t sure what to make of digital advertising. Could social media really Get Out the Vote?
The only way to find out was to launch a test!
Campaigns in the Digital Age
Of all forms of political communication, digital is growing most rapidly.
Less than 2% of political advertising in the 2012 cycle was spent on digital media. By the 2016 election, digital media’s share had ballooned to more than 14% of the political advertising budget, corresponding to $1.2 billion of spending on digital ads.[i]
With other industries spending 30–50% of their advertising budget on digital ads, we expect campaign spending on digital ads to only further accelerate.[ii]
Our organization, Progressive Turnout Project, formed in late 2015 in the midst of this political digital advertising boom. Our group’s goal from the start has been to design and test new solutions to a decades-old problem for the Democratic party – debilitating low voter turnout from its base.
So in 2016 (our first cycle running programs), we ran our first experiment to measure digital ads’ impact on increasing Democratic voter turnout.
Why We Test
Despite the booming digital spending in 2016, our staff was still skeptical about digital ads’ potential to convince a person to vote.
Recalling her thoughts at the time, our former Research Director Claire Rogerson says, “We come from the school of thought that the most effective method is having someone from your community knock on your door and connect with you personally about common values. Digital ads are nothing like that.”
But we were building an organization committed to making decisions supported by data rather than hunches. Thus, we went about designing a field experiment targeting digital ads at Democrats who don’t always vote.
After combing through research from political scientists and behavioral scientists, we designed our programs for 2016, emphasizing commitment, accountability, and identity in our messaging.
We wanted our message across mediums to be consistent, but we had much less room to deliver a long, complex message on a digital ad than we would at a voter’s door, in a piece of mail, or on the phones. With just a few words and an image, we had to create a virtual billboard that captured these sentiments.
“Digital ads are very simple and need to convey a message using little text and image. The thing that’s challenging about digital is you can’t have a conversation. You can’t ask a person why they vote or what issues they care about,” said Claire.
Our Digital Program
Our digital ad program had three phases, mirroring the three phases of messaging we used in conversation with voters at the door.
To build our experiment universe, we first selected our ideal targets from the voter file data. Our targets were likely Democratic inconsistent voters in thirteen highly competitive districts. Then we sent these targets to a data vendor to see which could be matched to an identifying digital footprint called a “cookie.”
We received about 140,000 cookie matches back from our vendor. Of those 140,000 people, some were put in a control group, some received media that didn’t include digital ads, and about 80,000 received digital ads.
Targets in our control group received no digital ads or other media from us. This way, we could compare the voting rates of people who received our digital ads to similar people who did not see our ads.
The first phase of digital ads began in mid-October 2016 and ran for about two weeks. The overarching goal of this period was to get people to make a commitment to vote. We used visuals that associated being a voter with being a role model or a good member of the community. The ad text directly asked people to commit to vote.
For the following twelve days, we ran Phase 2: a video ad that ran in 15- and 30-second versions. The video expanded on the role model theme by talking to real people about who views them as a role model and how those people would feel if they found out they didn’t vote. The video ended by asking people to plan out when and how they will vote. You can see the ad here.
We ran this video to our targets about fifteen times and made an estimated 7 million impressions. This was the costliest phase of the experiment – we spent about $60,000 running these ads on Facebook, YouTube, and before other targeted videos.
Phase 3 ran the last four days of the campaign, including Election Day. These ads reminded our target voters that whether or not they vote is a matter of public record and helped voters find their polling location.
Voters who clicked through the ads from any phase were taken to a sign-up page that asked people to commit to vote. If voters completed that form, they were redirected to a website that had a sample ballot, polling place locator, and information about candidates.
To craft the sample ballots, we worked with BallotReady, a non-partisan organization supported by the Knight Foundation, National Science Foundation, and the University of Chicago. BallotReady’s mission is to help voters make informed decisions at every level of elected office by aggregating primary source information on candidates. For our efforts, we pioneered the platform’s endorsed candidate pages to ensure our voters received information about their Democratic candidates prior to voting.
Interpreting Our Results
Once state voting records were available post-election, we pulled individuals’ records to compare the voting behavior of people we targeted with digital ads to that of our control group.
Our results showed that people who saw certain media overlays that included digital ads voted at a 2% higher rate than those in our control group. It’s clear that our digital ads had a positive impact on voter turnout.
The simple act of viewing the ads — even without clicking on them — had an effect on our targeted voters. As our Executive Director Alex Morgan said, “We confirmed digital ads are virtual billboards, but it turns out they’re much more effective than we expected! A very small percentage of people click through the ad, but people will see it.”
Why do we think our digital ads had an impact? Ultimately, we theorize the effectiveness of digital ads comes down to their delivery method. Digital ads are like a virtual billboard, but they are a billboard we can deliver to exactly whom we want to see it.
The most challenging part of talking to turnout voters is contacting them, as the virtue of their lifestyle makes them harder to reach. Inconsistent voting Democrats are younger and more mobile than other voting groups, so their contact information is often out-of-date. Digital is an effective and efficient way to find people and speak to them.
“In the end, we believed we could change people’s voting behavior if they were in front of us. Digital is one way to get in front of people you can’t reach through other methods,” said Claire Rogerson.
Going forward, we plan to design digital ads that deliver information to voters without requiring interaction. Since we can’t count on people clicking the ad to get more information, it’s necessary to provide everything we want the voters to know in the ad itself.
In our 2016 work, we tested digital as one part of a larger media package. In future tests, we want to test how digital ads influence voter behavior when they aren’t in combination with other media overlays.
These future tests will help tease out the groups of voters for whom our ads are most effective. This is a key aim of many of the tests we run – as we hone in on what outreach method or package of methods moves different types of people to vote, we can build customized programs that are increasingly cost-efficient. Becoming more efficient with our spending will allow us to reach more Democrats in more districts.
[i] Borrell Associates, Inc. The Final Analysis: What Happened to Political Advertising in 2016 (And Forever). 2017.
[ii] Lapowski, Issie. Political Ad Spending Online is About to Explode. Wired. August 8, 2016.