To subscribe to the monthly C&E email newsletter and event announcements click here.

At the beginning of the 2016 cycle, we felt that text messages held promise based on some existing studies done by our peers in the civic engagement world. They were promising, but not done with large sample sizes because on the whole most campaigns organizations hadn’t invested much in mobile programs yet. We were eager to build on their work and run some larger studies as well as testing a new mode of contact.

In 2014 the Analyst Institute conducted a study of texting with around 150,000 participants. They found this increased turnout in that midterm year by 0.9 and 1.4 percent for “plan-making” texts (those that get the voter to go through the mental process of planning how they will get to the polls). The program operated at an incredibly low cost when looking at cost per vote, especially compared to other modes of contact.

At vote.org, we ran three experiments in 2016 using SMS for voter registration and two varieties of GOTV. But before we get into the findings, some explanation of terminology is in order.

There are two types of text messaging exchanges. Many are likely familiar with “warm” texting. That means you’ve got a relationship with the subscriber and they’ve opted to hear from you. Organizations have been building these kinds of lists for a while and much like email you can blast messages out to all your subscribers at once.

“Cold” text messaging involves using people to send messages one at a time to voters based on cell phone numbers purchased from a commercial data vendor. But doesn’t that violate the Telephone Consumer Protection Act (TCPA)? Well, no, it doesn’t, because the texts are not being sent using “automated telephone equipment.” Rather, they’re being sent one at a time by humans hitting send from a real phone number. It’s not much different than phone banking (where you also make unsolicited calls), it’s just much much faster and less intrusive.

We’ve previously shared our research on using “cold” texts and one-to-one messaging to both turn out voters and to register voters. Our final 2016 experiment dealt with two topics, the first being the traditional “warm” texts. Vote.org was able to build a giant opted-in text list in 2016 because of the heavy use of the tools we offer on our site. This “warm” texting experiment was the largest one conducted to date with a total of 324,935 voters.

Our 2016 study showed important findings with effects on par with other non-partisan presidential year GOTV programs:

First and most important, texting works, both the “cold” and “warm” varieties. Overall, our study showed a statistically significant 0.5 percent increase in turnout (0.65 percent without factoring in social pressure texts). Sending people their polling place information and engaging in plan-making works to increase turnout.

Second, for one of the test conditions we used some light “social pressure” messaging. Voters were reminded that whether or not they voted was public record. This has been found to be effective time and time again in other modes of contact. That was not the case with this texting experiment. Social pressure messaging underperformed our other GOTV messages by a margin of 0.4 percentage points with to 0.65 points without.

Finally, doing GOTV using “warm” texting (0.65 percent) outperformed “cold” texting (0.2 percent), but the opportunity for scaling is much higher and the acquisition costs much lower for “cold” texting.

The second part of our experiment focused on using email for GOTV. Email is the tried and true digital workhorse. Many organizations rely on it for fundraising, communication, and to drive action from their members. It’s very effective for these tasks and has proven itself time and time again. We set out to figure out if it was useful for GOTV.

Our experimental universe was 510,079 voters, half of whom received a series of emails introducing Vote.org, providing polling place locations, engaging in plan-making, and a couple of reminders to vote.

Ultimately, email showed no increase in turnout effects in our experiment. So while it remains a solid bet for a number of other uses, GOTV dollars would best be spent elsewhere if the intent of those messages is to increase voter turnout. If civic engagement organizations aren’t building SMS programs now they should shift some resources into doing so, as it’s a much more promising channel for GOTV efforts.

Raven Brooks is the chief operating officer of Vote.org