Organizing in 15 different languages
Lessons from the 2018 campaign to flip Texas’ 22nd congressional district
This memo was written after the 2018 midterms as I was finishing my role as Field Director for Sri Preston Kulkarni, who was the Democratic nominee for Congress in Texas’ 22nd district. As we head into the 2020 general election, I hope it can be helpful to campaigns looking to organize diverse communities and build long-term power.
Fresh off the 2018 midterms, the Kulkarni for Congress team is memorializing our historic campaign in Texas’ 22nd congressional district. Sri Preston Kulkarni’s campaign received national attention for our innovative approaches to voter contact and engagement. We are proud of the inclusive campaign we ran, and would like to share some of the details of our organizing tactics. Focusing on two distinct methods – multilingual outreach and relational organizing, we hope to begin a conversation with organizers and activists across the country about how campaigns can engage communities of color and leverage peer-to-peer relationships and social pressure for large-scale voter mobilization.
Organizing micro-communities & Multilingual outreach
Texas’ 22nd district is one of the most diverse districts in the country. Located just outside of Houston in Fort Bend county, over 100 languages are spoken in the district, which is home to a rapidly growing constituency of Indian, Chinese, Vietnamese, Pakistani, Filipino, and other Asian immigrant communities. Our volunteer-led campaign reached out to these communities in 15 different languages. Having grown up in Fort Bend county, our candidate instinctively appreciated the richness of the district’s diversity. We prioritized talking to voters who are typically ignored by campaigns, and as a result, we earned buy-in from the different subcommunities and saw significant increases in voter turnout. We received widespread attention for this approach to community organizing, which has been standard practice for candidates of color and which we hope will become a model for more campaigns in the future.
There are many reasons campaigns typically do not do outreach to voters in over a dozen languages: beyond the fact that many consultants will advise campaigns to stick to more “traditional” organizing methods, there is no existing infrastructure or support for Democratic campaigns to segment the voter file in such a micro-targeted way. At best, Democratic campaigns are provided with (often inaccurate) race and ethnicity models. In our case, the district’s voter file included roughly 80,000 voters tagged as race/ethnicity = “Asian.” This broad categorization of course includes voters ranging from Pakistani-Americans to Japanese-Americans. We exported this 80,000 voter list to a giant Google Sheet and distributed it to our volunteer network who segmented the file by hand into subcommunities through surname analysis. It was a large undertaking, but our volunteer team was able to successfully identify 90% of Asian voters in just under two weeks.
This list segmentation, produced through tireless volunteer effort, proved to be invaluable. Not only were we able to begin multilingual phone banking and canvassing, we used these hand-curated voter lists for micro-targeted digital ads, direct mail, handwritten postcards, and peer-to-peer texting. Multilingual outreach transforms what is otherwise a “cold call” into a “warm call.” When volunteers open up calls with common greetings (‘Assalamualaikum’ or ‘Namaste, this is…’) and can pronounce voters’ names without error, we see a significant decrease in hangups and increases of 4–10 times the typical positive ID rate. For many voters, we had inaccurate or missing phone numbers. This turns out to not be an issue when a community leader can simply look up the voter’s phone number in the community directory.
Most importantly, these lists in the hands of community leaders are tremendously empowering. We were able to have one-on-one meetings with community leaders and say “here are the members of your community who did not vote in 2016. What are the ways we can engage them this time?” We produced physical, paper binders of voters complete with phone numbers and voting history, which we gave to volunteer leaders who took ownership for mobilizing their community. We included summary information on the historical voting participation of their community (and how they stack up against other communities). Leaders held house parties where attendees would pour over the binders and Google Sheets and take responsibility for reaching out to entire families of voters who they knew personally. All of this information was uploaded to our database, while the physical binders remain in the hands of community leaders well after the election has ended.
The real momentum began upon the start of early voting. Through dozens of active WhatsApp groups, we provided daily updates to the volunteers and community groups detailing which members of their community had and had not voted. We created a “leaderboard” indicating voter turnout across all the community groups, which inevitably fostered friendly competition between our volunteers.
There is nothing particularly glamorous about lugging dozens of binders across a sprawling Texas congressional district or doing VLOOKUP formulas in dozens of Google spreadsheets until 3 in the morning. But community organizing, especially in an electoral context, is about meeting people where they are, and good organizing in general requires doing the best with what you have. Right now, we don’t have the best digital tools or an adequately enriched voter database to do this type of organizing at a national or even statewide scale. Netting votes is hard enough with the proper tools for engagement. Oftentimes, it is about piecing together handfuls of votes where you can. In our case, it was about building a coalition of immigrant communities, some of whom had fewer than 1,000 voters, and finding creative ways to empower those communities with voter information that is typically only available to paid campaign staff.
With these successes, challenges, and limitations in mind, our team is now considering the following:
- How can we equip every campaign that is eager to do community organizing with the digital tools and resources necessary to micro-target specific communities of voters?
- What would it take to segment the voter file into ethnolinguistic groups at scale, and perhaps at a national level, ahead of the 2020 elections? As one data manager at the Texas Democratic Party put it, how can we “Amazon Mechanical Turk” the voter file? How can this handwork contribute to machine-learning techniques that can automate list segmentation?
- This model is particularly applicable to districts with large immigrant communities. Can we extend this model of organizing to other communities, such as faith-based or professional communities and college campuses?
- Finally, we are very aware of the ethical implications of campaigns microtargeting communities of color. How can we open up the voter file to volunteer activists digitally while safeguarding against potential privacy and safety issues for voters?
We want to note that we implemented these organizing tactics alongside a standard “bread-and-butter” field program that included door knocking, cold calling, and texting to generic GOTV and persuasion universes. We do not believe that this type of organizing could or should replace a campaign’s standard, “big” field operation. Rather, we recognize that the standard field program, however large it may be, will not reach many communities of voters meaningfully, if at all. The same is true of generic digital and direct mail programs. A “one size fits all” approach to campaigning is ineffective with large swaths of the electorate and alienates key communities of activists and volunteers. Moreover, the aggregate-level data and modeling appended to the VAN (which informs targeting and universe-building) inevitably contributes to generic campaigning and cookie-cutter messaging.
Good data, list segmentation, and a willingness to invest time and resources into outreach to communities of color offers an antidote to tone deaf, top-down organizing and provides an opportunity for campaigns to be more inclusive and culturally competent (not to mention electorally successful). The VAN is an extremely powerful albeit imperfect tool, and each cycle it remains exclusive to a handful of paid campaign organizers. Recognizing this, it is the responsibility of VAN administrators and campaigns to use the database to implement intentional, robust community organizing programs.
A final note on microtargeting: microtargeting is the practice of sorting people into categories in order to then tailor communications to the individuals in each group. Our volunteers sorted people into categories that were used to tailor not the message itself, but the messenger. The missing piece to most of today’s microtargeting efforts is finding and elevating credible messengers to deliver the campaign’s message. We did not use our list segmentation to customize messaging to different ethnolinguistic groups. Instead, our candidate worked tirelessly to earn buy-in from community members who then were equipped with tools to persuade their networks to do the same — whether it was a phone call, text message, hand signed postcard, or 30 second endorsement video served on Facebook, voters were receiving messages developed and delivered by a member of their own community. The entire operation, from start to finish, was reliant on the support and consent of activists on the ground. Quite literally, none of it would have been possible without the campaign’s large and diverse base of dedicated volunteers. This leads us to our final recommendation: focus less on message-testing communications and invest more in earning the support of critical grassroots constituencies. Trust community members to be influential and authentic spokespeople for the campaign and empower them with the resources to engage their own networks.
Relational organizing & social pressure (at scale!)
Our campaign’s intense focus on community organizing evolved into an understanding that friend-to-friend outreach coupled with social pressure is the single most effective tactic for driving out the vote. This aligns with a mounting body of scientific research that shows the statistically significant impacts of social pressure and relational organizing, independently of one another. What is unique about our campaign is that we did relational organizing at scale, and we coupled it with social pressure. The following is an overview of our relational organizing program:
- First, we collected over 12,500 “relational IDs” (friends, family, and acquaintances of our supporters). We did this by cutting custom lists of voters for each volunteer, who spent typically less than one hour scanning the list and “claiming” voters who they personally know. (For example, Amy is a young mom in the Riverstone subdivision and active in her daughters’ PTA. Amy receives a list of the 1,400 female voters ages 30 to 40 who are zoned to her daughters’ elementary school, and she can personally identify 167 voters in the list.)
- Next, we hired a full-time relational organizer who built a team of volunteer relational organizers to follow up regularly with each of our 250 volunteers who participated in the program. The goal was to develop vote plans with each of their voters.
- Finally, we leveraged Texas’ lengthy twelve-day early voting period for GOTV and accountability. On each night of early voting, we produced a report for each volunteer detailing which of their friends/family have and have not yet voted. This report included the voter’s contact information, polling location, and voting history.
We collected the 12,500 IDs over the course of only four weeks, and were astounded by the pace at which we were able to scale the program. Importantly, the voters in our relational ID network were younger and had less reliable voting participation relative to the district’s registered voters.
Our campaign was fortunate to have the support of a talented coder who built an email bot that produced and sent the emails to the volunteers on each night of early voting. The reports included a “voting status” column that was updated daily using the early vote rosters provided by the counties. Below is an example:
Integrating near-real-time early voting data into the reports, along with the voters’ voting histories, introduced a social pressure element to our relational organizing program. Our volunteers were able to tell their friends, “Hey, I noticed you haven’t voted yet. Early voting ends in three days! Can you make it to the Randall’s tomorrow?” Research continues to show the effectiveness of social pressure tactics when done right. It has also revealed potential blowback when not done right. The risk of blowback to social pressure messaging is potentially minimized when friends and family are the messengers rather than campaigns or volunteer “strangers”.
While an imperfect measure of the impact of this program, a comparison of the turnout of voters in our relational network against districtwide turnout shows the voters in our relational network outperforming the average voter significantly. This holds true across basically all groups (age, race/ethnicity, gender, and past voting history).
Reflecting on the program, our team is eager to explore ways to improve upon these methods and hopefully trial this relational organizing model with other campaigns. Based on countless anecdotes from our volunteers, the daily early vote reports allowed our supporters to turn out their friends and family in a methodical and effective way. Mapping 12,500 relationships in just four weeks was a thrilling experience for us as organizers, and we believe it merits a discussion about the potential for campaigns to do relational organizing at scale in 2020. Moreover, we are eager to explore relational organizing not just for increasing turnout, but also for implementing persuasion programs. The research increasingly shows that in attempts to persuade voters, more important than the message is the messenger. How can we leverage relational organizing to maximize crossover voting when possible? We feel that this program deserves to be trialed in other campaigns and could become a critical and standard feature of Democratic campaigns in the near future.
