As robocalls increase and pickup rates decrease, better voter contact solutions are a must
For decades, campaigns have shuffled eager volunteers towards rows of phones to make calls to potential voters — to raise awareness, make their case, and turn people out to vote. Now, though, amid a mountain of robocalls, telephone pickup rates are in the gutter and campaigns are scrambling to replace this channel. According to a report by Ozy, in 2018 alone, 26.3 billion robocalls were placed to cellphones and landlines — approximately seven calls per month, per person.
Peer-to-peer texting offers political campaigns a personalized, yet less invasive, way to connect with voters at a much lower cost than other traditional outreach methods such as canvassing, a higher read and response rate than emails, and significantly higher penetration rate than phone calls. In the 2018 cycle alone, Democratic campaigns and politically-adjacent organizations sent 350 million text messages through Hustle and GetThru (formerly Relay), the two largest peer-to-peer texting platforms in the Democratic market, 6x more than during 2016 and 2017.
However, because texting is such a new outreach tool for political campaigns, there’s very little data on if and what makes for an effective text messaging strategy - specifically one that will turn out voters. In 2018, Tech for Campaigns (TFC) was active in 20 states and ran 29 texting programs for Democratic campaigns and committees in 9 of those states, sending more than one million texts. TFC data scientists were able to compare voting records to our work and begin to assemble best practices from 2018.
The findings below are select learnings from our get-out-the-vote (GOTV) texting analyses. While we think of randomized controlled trials (RCTs) as the gold standard for analysis, RCTs are notoriously difficult to run in active, competitive political environments, so these analyses are correlational, not causal - though we controlled for as many variables as possible.
We have already started to incorporate these findings into our campaign work and hope it helps to bring a more data-driven mindset to this growing channel.
2018 Campaign Learnings
People who were texted were 1% more likely to turn out to vote
While 1% may not seem like a lot, competitive races, especially state legislative races, are decided by just tens or hundreds of votes. In one state alone, the difference between the texted and untexted voters in the districts in which TFC ran texting programs would have been enough to win 5 of the 7 races Democrats lost in 2018.
Texting younger voters can have an outsized impact
Registered voters between the ages of 27 to 50 turned out at a rate almost 8% higher than those in that same age group that were not texted.
Younger, registered voters (50 and younger), turned out at much higher rates when texted when compared to older (51+) groups. There is statistically significant evidence to support this finding when controlling for factors including age, sex, race/ethnicity of the registered voter as well as whether they voted in the general elections of 2016, 2014, and/or 2012.
Content matters - especially on the issues
Text messages sent as part of TFC’s 2018 program used two different content approaches: issue-based texts (ex: healthcare, education, environment and voting information such as polling hours and location) or non-issue based texts (ex: only polling station hours and location, event information, etc.). Issue-only texts outperformed non-issue texts in predicting turnout. This holds true when we examine significance across specific variables including an individual’s historical voting patterns, party support, and age. While voters still need basic voting information, our findings suggest personalized, localized and candidate-specific messaging is more effective than stock messaging.
Conversation is critical
As campaigns make their decisions about voter contact method and investments, they are doing so in the dark because voting records are typically not available until after the election. However, leading indicators such as response rate to texts are critical to understand whether and how a texting program is contributing to outcomes. Our analysis found that voting rates are significantly higher among those who responded to texts. While purely correlative, it may be a useful way to predict who will vote and enable campaigns to redirect GOTV resources elsewhere instead of continuing to contact these voters.
A Final Note
At Tech for Campaigns we believe in a data-driven approach to voter outreach. As texting continues to grow in importance, we encourage taking a thoughtful, test-and-learn approach to all programs so that the channel continues to be useful for voters and impactful for elections. If you'd like to get in touch regarding running a data-driven texting campaign for your state legislative, statewide, or Congressional race, get in touch.
Methodology and Sources
Tech for Campaign’s texting programs included a total of 591 text variants and more than 1 million sent through Hustle, GetThru (formerly Relay), and Groundbase texting platforms. Texting projects included several different objectives: fundraising, event turnout, and get-out-the-vote texts. We conducted our analysis based on campaigns in 4 states which focused on GOTV texts only.
Given the difficulties involved in running randomized controlled trials in the field during political campaigns, we used a purely correlational approach, controlling for as many variables as possible for a given analysis - including age, sex, race/ethnicity of the registered voter, party support score, and whether people voted in the general elections of 2016, 2014, and/or 2012.