🍑 Swipe Out the Vote 🍑
Movement Labs 2021 Georgia Runoff Dating App Get Out The Vote Project
By Noah Foley-Beining, Khrizia Velacruz, Daniel Jubelirer, Julia Foote, and Rebekah Dixon
Program Outcomes At a Glance
- From Dec 23th -Jan 5th, Movement Labs’ Swipe Out the Vote team recruited volunteers to do voter outreach on dating apps. Our project netted 270 votes at a cost of $22.2 per vote.
- Our average response rate was 61% from contacts that were matched, much higher than traditional GOTV contact methods.
- We persuaded 16% of people to vote who said they weren’t planning on voting.
- We completed an estimated 2,231 conversations on the apps over 14 days.
- Volunteers presented transparently — as out of state and interested in the GA election, hoping to talk to voters about turning out.
- Volunteers who were burnt out from months of phone banking and text banking found this to be a fun, engaging way to volunteer and we saw many volunteers keep returning to our trainings and swipe parties.
Background
In the final months of the 2020 Presidential General Election, several groups experimented with using dating apps such as Tinder, Bumble and Hinge as a means of engaging potential voters and persuading them to vote. Many have anecdotally claimed this to be wildly more effective than traditional GOTV efforts (phone calling, texting, door knocking). Some volunteers found themselves with 600+ matches and a consistent backlog of 100+ messages. Many of these conversations were with apathetic voters and people who were undecided. After just a short conversation about voting, many matches said they planned to vote. These promising experiences in October/November 2020 led to investments in these efforts by Movement Labs in the 2021 Georgia Senate Runoff Election.
Outcomes
Conversations Completed: By the end of our two weeks working on this program, we ended up with 46 confirmed active volunteers who logged 1,339 self-reported conversations.
Estimated Totals: In total we had 132 volunteers attend training (not all remained active) and messaged an estimated 2,231 voters.
Results of tracked conversations: We recorded 22 instances of persuading a match to vote who wasn’t planning to vote before we contacted them, 94 not voting, 148 undecided, 502 already voted, and 248 stating that they did plan on voting.
Engagement Rate: We saw very high rates of engagement when talking to voters on the apps. Though limited in data collection abilities, the advantages of using dating apps for voter outreach are huge. Of our volunteers who kept count of their total number of matches, we saw an average response rate of 60.8%. There is no comparison when looking at other voter contact methods. Phone calling has a response rate of around 6%, Movement Labs texting response rate for GA was around 5%, and door to door canvassing contact rate of 24%.
Targeting
Unlike a traditional phone or doors program which contacts targeted voters from a voter file, dating apps do not allow for explicitly targeting people of a certain political party.
Dating App voter outreach is analogous to community canvassing, where you go to a public area and talk to anyone walking by. You are reaching the general public on the apps.
To ensure we reached our target audience of likely Democratic, low-propensity voters, we took the following steps to attempt to target the right people:
- We gave parameters to volunteers for age and location settings on the apps, based on 2020 general election turnout.
- We had people set their location within zip codes in GA with 65%+ Biden voters in 2020.
- We targeted younger voters (18–35) who skew progressive and who tend to vote at a lower frequency than older voters
- We had people swipe left (indicating they weren’t interested and therefore wouldn’t match) with people with clear conservative keywords or iconography such as MAGA hats, “blue lives matter” in the bio, etc.
- Therefore we aimed to reach folks with politically-neutral bios, who were young and in progressive areas because this gives us a target audience that we want to reach to motivate to vote.
Findings
Our Data
While messaging people in the final two weeks of the runoff, we found that 51% had already early-voted, 25% said they planned on voting on Election Day, 9% said they did not plan on voting or were ineligible, 15% were undecided, and 16% would not share their plans on voting. In total, our volunteers were able to confirm voting plans 84% of the time when a match engaged in conversation.
Out of the people who originally said that they would not be voting, 16% said they would vote after a persuasion conversation and 6% said maybe they would vote after a persuasion conversation. This finding that we may have persuaded up to 22% of people who said they would not vote is a huge number to consider.
While talking to matches we were able to record party affiliation in 235 cases. Out of those, 76% supported Warnock and Ossoff, 3% had a split ballot, and 23% were non-supporters. We did not ask party affiliation in the majority of conversations, because we determined at the outset that simply turning out young voters would be a net gain for Democrats, since young voters in blue districts in Georgia vote overwhelmingly Democratic.
Which apps work best?
Our program was geared towards using Tinder since it is the most popular app in the US, but we did encourage our volunteers to experiment with other apps. In the end, our break down of app usage looked like this (some volunteers did not record which apps they were using):
App
How many voters can one volunteer talk to?
The nature of dating apps causes this to vary widely. For one, femme-presenting individuals on dating apps often have overwhelming numbers of men matching and reaching out. Due to these apps being designed for dating, individuals who are deemed attractive will end up with many matches. Through this experimentation, we have found that femme-presenting volunteers are likely to be able to talk to many more voters than masc-presenting individuals regardless of perceived attractiveness. There is no way for us to track time spent on app other than self-reporting, but volunteers were usually getting matches on Tinder in just a number of minutes after creating their accounts, and were having back and forth conversations shortly after. Another key component is that conversations are peer-to-peer, meaning that the age of each volunteer typically falls within the selected range of their matches. In our experiment, we found those between 20–33 to be the most effective volunteers. Below is a histogram of the age groups that we engaged with most:
Out of our 46 active volunteers (active meaning they recorded at least one conversation), the average was 27 conversations per volunteer. However, we believe that with committed volunteers, that number can be much higher. When looking at our 18 most active volunteers (many of the other 46 only messaged 1–3 people and did not stick with it), we found the average number of conversations to be 62. This indicated that with a volunteer base that understands the effectiveness of this tactic, many persuasive conversations can be had with voters.
What messaging works best?
When training volunteers we encouraged them to bring up voting with their matches on the third message sent with a conversation often looking like this:
Doing so makes the match less likely to accuse the volunteer of being spam or a bot that is only focused on voting. It is crucial to establish a human connection at the onset so the person at the other end trusts the volunteer enough to engage in conversation. One thing to note on this topic: due to the uneven gender dynamics on dating apps, non-femme-presenting individuals may need more engaging openers than a simple “how are you doing?” We encouraged volunteers to open conversations with comments that related to the match’s profile, such as this match who indicated that he was a baseball pitcher in his profile:
Messages that referenced a current event and one that most people would have an opinion on were also effective at getting a reply. This question about $2000 stimulus checks, in particular, was great for quickly pivoting to voting:
We found that nighttime between the hours of 7pm-12am local time was, without a doubt, the time period when most people were responding on Tinder, which is why we scheduled our trainings and Swipe Parties in the evening (click here to read more about optimal times). All in all, over an hour of swiping and sending messages to matches a volunteer could easily have 20 conversations.
Location
Many things needed to be taken into consideration when creating training materials to guide a cohort of volunteers on effecting dating app strategy. We began by deciding on targeting. Unlike traditional GOTV where an organization would have access to a voter database with party affiliation, dating apps offer a much different set of criteria to target on. Since Tinder does not let you filter for political leaning, we decided we would concentrate our efforts on people between the ages of 18–35, as this age group tends to vote liberal. From there, we prioritized the 10 largest cities in Georgia and instructed volunteers to set their radius in the app to a set number of miles that surrounded the city. This way, we targeted young people in cities, who are more likely to be liberal than in targeting rural areas. We felt confident that moving forward with this criteria would ensure we did not turn out a statistically significant number of young conservative voters.
After our two weeks of talking to matches about voting, this is what our data distribution looked like across the state. Click here to explore the interactive map to see the breakdown of confirmed voters.
As you can see, most volunteers spent their time in Atlanta swiping on voters 18–35 years old, who overwhelmingly vote for Democrats. Out of all the conversations with location data, 94% were in Democratic Counties and 6% were in Republican counties.
Key Learnings
Data Collection
The challenge with organizing a volunteer effort on dating apps, as opposed to traditional GOTV, is that there is no way to gather data on volunteers’ successes without self-reporting. This made setting concrete metrics difficult. We also asked people to only track matches where the conversation yielded a notable result, such as the match stating that they are voting, are not voting, already voted, unsure, or the match was unmatched. This meant that conversations that did not yield information about voting did not get recorded. In the future, we recommend a more robust data collection method that involves tracking more data from each conversation.
We believe that adding incentives for volunteers such as stipends, gift cards or special access to campaign events in exchange for reporting would be a cost-effective way to grow the impact of this program.
Account Bans
During our two weeks of swiping with 46 active volunteers, five got their accounts banned (4 from Tinder and 1 from Hinge). However, the Hinge user was able to appeal and regain access to their account. We used Tinder because it is the most popular dating app in the US, but Tinder also has the strictest rules against advocating for voting on its platform. As their community guidelines state “If the purpose of your profile is to advertise your event or business, non-profit, political campaign, contest, or to conduct research, we may delete your account.” Commonalities between all the banned accounts included: bios that mentioned voting, one or more politically-oriented tags, and selectively swiping on potential matches. Based on this, we do not think that bans were caused by triggering an algorithm that would be looking out for people swiping yes on everyone. We think it is most likely that bans were caused after a match expressed anger at the volunteer being from out of state, advocating for voting instead of dating, or was upset with their progressive beliefs. As one volunteer wrote in her account ban debrief:
“I had three people ask me directly if I actually lived in Georgia/was actually interested or if I was just on the app to talk to people about voting, so I answered honestly and said I wasn’t currently interested in romantic connection and was there to chat with folks about getting engaged with the runoffs and provide support to anyone who might need help figuring out a plan to vote.”
Read through all of the Account Ban Incidents here.
Demographic Differences
Many new things came to light for us to consider when doing this in the future. One big one was demographic differences. In Georgia at least, we noticed hardly any Asian Americans or Pacific Islanders on Tinder. We then learned that Asian Americans and Pacific Islanders have a larger presence on the dating app Coffee Meets Bagel, which limits the number of matches per day. The app BLK is a dating app specifically for Black people. Grindr and Taimi are designed for the LGBTQ+ community and feature a grid-based matching system that differs from Tinder-style swiping. We had some volunteers on these apps, but in the future, it would be worthwhile for more volunteers from those communities to evaluate the suitability of these platforms further.
Volunteer Recruitment
The fact of the matter is that in order to effectively engage with a young voter on a dating app, the volunteers need to be young as well. Part of the challenge we faced was that much of Movement Lab’s existing volunteer base was on the older side. In the future, partnering with a youth-led organization to recruit larger numbers could be helpful. However, there is skepticism regarding this approach even among younger volunteers. Some don’t see the viability while others feel uncomfortable with being on a dating app due to the romantic/sexual nature of the app. For those that are comfortable with it and want to feel as though they are using their precious volunteer time most effectively, we hope that his report demonstrates that this tactic can be much more time-efficient than phone banking, text banking, or canvassing.
Closing Reflections
We think this approach has huge potential, especially in smaller races with more limited spending, where the dating app conversations may be the only place young people hear about a candidate.
Grassroots organizing in Georgia is working. When we experimented with Swipe Out The Vote in Texas during the 2020 general election we found many young people who were apathetic, undecided, and generally uninterested in voting in the election. We went in expecting to find a similar scenario among Georgia Tinder users. This was not the case. 51% of the people we had conversations with said that they had already voted and 25% said that they were planning on voting When asking matches if they were planning on voting, we were often met with “yeah of course.” This suggests that the years of grassroots organizing work in Georgia has had a tremendous impact on how young people view the importance of political participation. Additionally, over $800 million was spent on the GA Senate runoff when factoring in outside PACs and the campaigns themselves, which is over $100 per voter. GA voters were inundated with outreach at an unprecedented level.
If you are young and want to volunteer your time to help get out the vote, dating apps are the most effective use of your time. The quantity is lower than texting, but the quality of conversation is much more human. You will not be bothering people and getting repeatedly hung up on like you would with phone banking. This is deep canvassing, and people often open up even more than they would if you were knocking at their door. Young people engage in serious, in-depth conversations on dating apps and they can be persuaded to vote one way or another via this medium. Matches were often eager to share that they had been convinced to vote this election. However, it is important to remember that targeting the right locations and getting the messaging right is crucial to ensure that this technique is successful. Through dating apps, young people can have authentic conversions with other young people about voting.
Special thanks to our star volunteers Rebekah Dixon, Phoenix Mandel, Ryan Mann, Tahoe Roe, Yvonne Liccione who were instrumental in this experiment.
