Bureaucratic Bias or Voter-Side Factors? Testing Competing Explanations for Racial Gaps in Vote-By-Mail Ballot Signature Rejections

While voting eligibility has evolved throughout American history, a consistent pattern remains: White voters have often received preferential treatment from lawmakers and election administrators. During the Jim Crow era, for example, White voters were shielded from voting restrictions through mechanisms like White primaries and grandfather clauses, which exempted them from literacy tests, poll taxes, and understanding requirements. These barriers—commonly referred to as “administrative burdens” (Herd et al. 2023; Moynihan et al. 2015)—were systematically imposed on marginalized groups to suppress their political participation. Central to this process were local election administrators exercising tremendous discretion in their enforcement of voting laws and applying them along racial lines, essentially determining voter eligibility on a case-by-case basis with race being the determining factor. This racial profiling was crucial to the maintenance of White political dominance in the South (Keele et al. 2021; Key 1984; Kousser 1974). After the passage of the Voting Rights Act, blatantly discriminatory policies were phased out, but a certain degree of racialized bureaucratic discretion in elections has persisted and is making a resurgence in some states, especially in areas with localized racial resentment (Morris and Shapiro 2025). The controversial SAVE Act, if signed into law, would unleash local election administrations to make even more discretionary decisions on citizenship (Bedekovics and Bryant 2025). Many of these decisions will come down to determinations of the authenticity of signatures in registration forms, USPS-delivered mail ballots, and citizenship through the verification of passports/birth certificates.                                                                                                                                                                                                                                                                In the current vote-by-mail process, a voter’s identity must be verified as a safeguard against fraud, and in most states, this is accomplished through signature verification (National Conference of State Legislatures 2024). The justification for using signatures is that they function as a behavioral biometric—believed to be unique to each individual—thereby making large-scale fraud more difficult to execute (Hameed et al. 2021). However, a core concern with this method is that the individuals tasked with verifying signatures are granted substantial discretion, often making subjective, case-by-case judgments with limited oversight. These decisions are frequently made under pressure and, in many cases, without formal training (Page and Pitts 2009). The process has been described as “witchcraft” (Graham 2020), and “ripe for error” (Lau and Nelson 2020), though artificial intelligence is rapidly changing the landscape with many counties embracing more efficient, though perhaps more problematic computer vision AI tools, like those often used for facial recognition (Grother et al. 2019). Each year, tens of thousands of eligible voters have their mail ballots rejected due to perceived signature mismatches. In some states, these voters are not notified and are given no opportunity to correct—or “cure”—their ballots, resulting in their effective disenfranchisement without notice (Meredith and Kronenberg 2023). Moreover, research suggests that more than 80 percent of rejected ballot signatures are mistakenly deemed invalid (Street 2024). In Washington State alone, which conducts its elections entirely via mail, this error rate translates to over 100,000 wrongly rejected ballots between 2019 and 2024.                                                                                                                                                                                                                      The implications of these errors are especially significant in close elections, where the number of rejected ballots can easily exceed the margin of victory. For example, the 2004 Washington gubernatorial election was decided by just 133 votes. When the ballot rejection process is uneven or biased, it means that thousands of voters face systemic barriers to participation. Such disparities not only undermine core democratic values of political equality (Dahl 2007) but may also violate the Equal Protection Clause of the Constitution and provisions of the Voting Rights Act.

Read more below…