Experimental Voter Contact in the 2020 Elections
By Laura Flores
In addition to our main organizing and voter contact programs, we wanted to test out new and interesting tactics for the final stretch of the 2020 cycle. Here’s what we learned, and why it matters:
Tactic 1: Online Communities
What is it: NextGen is always looking to meet young voters where they are at. One thing we know for sure is that 18–35-year-olds are very online. Most people are a part of one or multiple online communities, even if they don’t view it that way. An online community could be a university class Facebook page, your neighborhood good eats Facebook group, a subreddit about a local NBA team, or a group of hashtags frequently used on TikTok. We had previously tested online community outreach for organization-wide days of action like Vote By Mail Day. For GOTV, we wanted to test our methods on a larger scale and learn about how we can tap into existing online networks to provide helpful and actionable voting information to young people.
How we did it: We focused on three platforms: Facebook, Reddit, and Tiktok. For Facebook and Reddit, we identified groups, pages, and subreddits by geographical indicators and made educated guesses about communities we have found to be particularly “youth-dense.” For example, a “University of Arizona Class of 2022” subreddit is both in one of our target locations and filled with college students in our desired demographic. “Milwaukee Gigs” is another example — a Facebook group that’s in our target location and geared towards gig economy workers. While we can’t guarantee everyone in such a group is in our age demographic, we know young people likely represent a high percentage of gig economy workers in any given major city, and many of the posts in the group were from younger Facebook users.
For Tiktok, we identified posts that we wanted to comment on by using location-based hashtags or trending hashtags. The good thing about TikTok is that the majority of its users are in our target age range, so narrowing by age wasn’t a primary concern. In some cases, we used campus-based hashtags, like #ForksUp at Arizona State University.
What we found:
This tactic would be extremely helpful for an organization that is still trying to build their audience or who needs to get information out on a tight deadline. Instead of trying to build your own audience, organizations can tap into existing audiences and find new voters who may not be typically politically active. The most receptive groups were those where we had been active for a longer period of time, groups that had a topic that closely aligned with our organization’s message (ex., environmental appreciation groups), and groups that had less tight knit communities. Our posts were removed most frequently in groups with small tight-knit communities with high-levels of moderation. Our content was taken down the least in groups that anyone could post in and had high membership numbers, but those posts received very low engagement.
Reddit was the most difficult to organize. The organization requires you to have enough “karma” to post in groups that have a genuine sense of community, while content gets swallowed by a deluge of other posts in larger groups. On Reddit, there were other organizations doing the same thing we were: posting voter information into location-based groups and campus groups during GOTV. This was a low-cost tactic that was overall positive but would have been more successful if we had started to embed ourselves into these online communities earlier and recruit members to be the designated messengers rather than having to post as an outsider.
TikTok users were very receptive to our comments and would often pin our messages to the top of their comment section or reply to us. We found that the best way to engage with users on the app was to leave between 25 and 50 comments every day or every couple of days. TikTok’s algorithm negatively weights accounts with a high volume of the same or similar comments, and comments from those accounts will not be visible to all users (this practice is commonly referred to as “shadow-banning”). Although TikTok’s platform doesn’t allow for hyperlinks in comments, we would push traffic towards a link in our bio. If we had three people dropping 25–50 comments per day, we would see between 16% and 33% of those comments drive traffic to our landing page.
Would we do it again: Yes. Tapping into an online community is a really valuable way to find new voters and earn trust in a community. There is a lot of potential to get to know people in the groups and recruit someone within that group to become the messenger for your campaign or organization. The difficulty that we ran into is that we should have started joining and interacting with these communities earlier so that we could have been more embedded in them by GOTV. Similar to in-person organizing, it takes time to build relationships online. For every online community platform, but TikTok specifically, our tactic was most effective when we engaged regularly and in small increments, as well as on marquis days of action. In the future, this would be a great tactic to incorporate into an organizer’s daily checklist. Given the low-cost and minimal time commitment, this is a tactic we would recommend others try.
Tactic 2: Amazon Mechanical Turk
Another tactic we tested was organic content distribution through Amazon Mechanical Turk, a digital tasking service that serves as a kind of job board for small online gigs. We hired 18–35-year-olds in Maine, Iowa, Arizona, and North Carolina to fill out a survey after posting a non-partisan election reminder graphic on their social media accounts.
Why we chose it: The goal of the project was to test our influencer program model with regular young people’s networks and see if relational tactics on social media are a valuable asset to a digital communications or organizing program. The great thing about using Amazon Mechanical Turk is that posting the task is actually voter contact in and of itself (since you can target the task listing by demo). We targeted taskers by state and age range. For example, the only people who were able to see our tasks in North Carolina were between the ages of 18 and 35. Even if they did not choose to complete our task, they still received our voter information because they saw the posted task. We wanted to try this form of voter contact because it has the potential to scale if found to be effective, and because it offered a way to access organic feeds and trusted messengers anywhere.
How we did it: Amazon mechanical Turk allows companies/organizations to post a variety of tasks including survey completion requests. For this trial, we decided that the best way to set up our task was to pay Taskers to complete the survey after they had posted a non-partisan, voter information graphic on social media. Taskers were required to post the graphic on social media with a pre-written caption, take a screenshot, submit the screenshot for records, and answer questions regarding their experience. Taskers answered questions about how they felt their post was received by their network, the number of likes, comments, or DM’s they received after posting the graphic, and whether or not they would do this again for free if the information was easily available and reliable. To start, we ran a trial in North Carolina and we learned that the most fruitful age ranges were between 25 to 30 and 30 to 35. We took that information and scaled the trial to include Maine, Iowa, and Arizona. In the final week leading up to the election, we had Taskers post Get Out The Vote information.
What we found: Because Amazon Mechanical Turk is a place where people go to do business it was important for us to make sure our information was factual, non-partisan and informational. We learned early on that our survey was too long and we were requiring too many steps in the process. Once we simplified our survey, more people chose to complete our task. We had the most survey completions at $15 per task. We also found that the majority of people said they would post this graphic again for no payments if voter information was trustworthy and accessible. On average for every five survey completions, we saw between 40 and 50 link clicks. This varied slightly based on state and by age range.
Would we do it again: Yes, but we need to do further testing with Mechanical Turk and other digital tasking services. There were a number of nuances that we learned along the way and would have been helpful in the first trial in North Carolina. Some of those small nuances include using only amazon task ID numbers versus names, lowering the number of days between the tasker completing the task and getting paid, shortening the task completion time window, and creating many small batches of tasks instead of a few large batches. Overall, this tactic has a lot of potential to reach hard-to-find audiences in statewide races. It is more costly than the online communities bucket, and has a higher cost-per-click than many ads, but it provides a way to distribute content organically anywhere. It’s also flexible in terms of scale — if you can find a few thousand dollars within an organization/campaign budget, it is an interesting experiment in how your messaging resonates with your target audience and their network.
