Black and Latino voters are being erased from databases: Here’s how to fix it 

Editor’s note: This piece was updated to correct an estimate. We regret the error.

According to my estimates of U.S. Census data and information revealed in a recent Stanford study, 24.76 million Black and Latino people are missing or listed with incorrect information in the voter databases sold by vendors, making them unreachable. While 40 percent of Black and Latino people are essentially invisible, only 18 percent of white people are missing or mislisted, according to the study. 

Finding these millions of missing Black and Latino voters could determine the outcome of the 2024 elections. In 2020, Joe Biden won battleground states like Arizona, Georgia and Wisconsin by only 10,457, 11,779 and 20,682 votes.

Vendors supplement government voter registration lists with demographic data and consumer information, such as from credit bureaus. For voters, being missed perpetuates a cycle of exclusion from civic engagement; if they aren’t in the voter database, they won’t get contacted about an upcoming election. Incorrect data also distorts the algorithms that assign vote propensity scores. Traditional campaigns will not generally contact voters whom technology has categorized — correctly or incorrectly — as unlikely to vote.

These applications are, in part, manmade — which means human biases seep in. For example, in 2022, one machine learning company announced that it could predict the race or ethnicity of an individual using only their name and address. Can something as complex as race be predicted by two factors alone? In general, companies building and selling voter databases do not provide enough transparency to allow users to assess the racial or other biases that may have distorted their products.

Working intimately with voting organization data leaders, I have learned that state-based organizations are using workarounds to find people who are missing or incorrectly listed in the databases they have purchased. First, they are supplementing the industry’s voter databases with their own list-building through relational organizing, that is, asking volunteers to reach out to friends and family and adding their contact information to their databases. 

For example, during my work with Voces de la Frontera in Wisconsin, I learned that through internal research after the 2018 elections, it found that thousands of voters had been purged from their voter contact database. Over the next several years, Voces was forced to manually add names and contact information back into their list so that they could reach voters in the next election cycles.

Second, organizers are moving away from arbitrary vote propensity score cutoffs. One organization, the New Georgia Project, rejects the industry term “low propensity voter” and embraces “high opportunity voters” instead. In their experience, these voters will become civically active if people invest in them. 

Living United for Change in Arizona (LUCHA) told us that, in 2022, it focused on reaching first-time and infrequent voters, attempting contact with 447,000 voters with low propensity scores and speaking face-to-face with more than 12,000 first-time voters. This approach — reaching voters who would otherwise be ignored — was and will be outcome-altering.

But it should not be left up to resource-strapped nonprofits to fix the voter databases (that they have paid for) and do the arduous work of bringing overlooked voters into the democratic process while traditional campaigns ignore them. All decision-makers need to become more active in creating an expansive democracy. Specifically:

1. Political data vendors need to put a pause on selling inaccurate data and unfair models and grapple with the huge gaps in industry and government files. At a minimum, vendors should be more transparent about their methodologies and answer questions about the quality of their data.
2. People working on campaigns should prioritize reaching out to missing and mislisted voters in 2023 and 2024. As the experiences of New Georgia Project, Voces de la Frontera, LUCHA and others show, organizers cannot limit their outreach to those listed in voter files — not everyone is included. At a minimum, organizers should have a healthy skepticism about the products the political industry is selling.
3. Political donors and campaigns should reconsider the metrics that they use for measuring voter contact efforts. Finding missing and mislisted voters requires more time-intensive outreach, such as a neighbor coming to the house for a thoughtful conversation rather than a quick door knock from a paid canvasser.
4.  The original sources of voter information — state and county governments — should end arbitrary voter roll purges. From November 2020 through July 2021, more than 8.6 million voters were purged from official registration lists nationwide.

Nearly a half-century ago, my grandmother, Loretta J. Williams, reflected on the exclusion of Black Americans from mainstream American life. “With the decline of legalized caste systems … one notes the maintenance or continuance in development of such parallel institutions,” she wrote in “Black Freemasonry and Middle-Class Realities.” 

Forty-three years later, millions of Black and brown people are still being erased from a crucial aspect of political engagement: voter databases and reliant algorithms. Those who care about our democracy should hear alarms blaring.

Miriam McKinney Gray is a lab director and senior data analyst for the Democracy and Power Innovation (DPI) Fund, a collaboration between state-based organizing groups, political and social science researchers, and aligned funders in support of research and innovation to improve strategy and practice at the intersection of organizing, civic engagement, and power in communities of color. The DPI Fund is hosted and managed by the Rockefeller Family Fund.