Do We Need Another Voter ID Study? Evidence from the 2020 National Voter File

• In Event: Unequal Turnout in Comparative Perspective

Abstract

Are Black, Hispanic and young people less likely to vote in 2020 and prior elections when residing in a state with voter ID requirements? Few topics have resulted in such conflicted debates as the effects of state voter identification laws on turnout, especially in their potential to restrict some demographic groups such as racial and ethnic minorities, from the ballot. Drawing on national voter files (2020 and 2012) that cover nearly all Americans this study reports the potential negative effects of ID laws on Hispanic and youth voters.

The 2020 presidential election broke a century long record with 67 voter turnout of eligible citizens. Yet, by 2020 more than 35 states had passed some form of voter ID requirements. While identification laws are popular, and national surveys indicate 3 in 4 Americans favor government issued photo identification to vote (Pew 2021) any law that restricts access to the voting ballot requires careful review. Though some previous studies find ID requirements reduce turnout, particularly of Blacks and Hispanic voters, others have uncovered null results. Most previous work has not included young people but focused on race.

Previous studies have found that voter identification requirements reduce turnout for Black and Hispanic Americans to a greater extent than white non-Hispanics (Kuk et al 2020; Hajnal, et al 2017; 2018, but see Grimmer et al 2018), decrease turnout but without racial and ethnic disparities (Alvarez et al 2008; Hood and Bullock 2012), or have little impact (Rocha and Matsubuyashi 2014). These discrepant findings are often attributed to several empirical limitations such as use of self-reported turnout, non-representative state samples in national surveys, underrepresentation of specific voters in Census data, and issues with survey vote validation procedures (Ansolabehere, Fraga and Schaffner 2021; Grimmer et al 2018).

Additionally, differences in how ID laws are coded may account for some of the differences in findings. NCSL notes four types of state laws: 1) non-strict, non-photo; 2) strict, non-photo; 3) non-strict, photo; and 4) strict, photo identification, with a fifth category of requiring no identification. While most previous research focuses on the most strict category, this study measures all four categories in comparison to states requiring no documentation.

A new study (Cantoni and Pons 2021) using individual level panel data from the Catalist national voter files from 2008-2018 of 1.6 billion observations came to a surprising conclusion. Comparing strict to non-strict ID laws and photo vs non-photo, the results show the laws have no negative effects on voter registration or turnout for the overall population or for any group defined by race or ethnicity. In fact, their models find that Hispanics are somewhat more likely to vote in state requiring voter identification. Others find Latinos are more likely to be asked to show identification to vote (Akteson et al 2010) and a new study from Texas finds these laws are more likely to prevent Blacks and Hispanics from voting (Fraga and Miller 2021).

Limitations of Cantoni and Pons (2021) include that the study did not test all forms of state identification laws, omitting photo non-strict. Pooled statistical interactions used to test the effects on racial and ethnic groups omitted the reference category, making interpretation of the results unclear. Given the population data, subsampling by racial and ethnic groups is preferable, as we do here. Using two-way fixed effects (state and year), the models included a limited number of covariates, omitting factors known to predict turnout, including partisanship and education. It is unclear if identifying assumptions were met, including parallel trends between treated vs. untreated units.

Almost no previous studies have incorporated information on election administration quality in their analyses despite its importance (Ritter and Tolbert 2020). Previous research finds Black, Hispanic, less educated, and those who English is not the primary language are more likely to be asked for voter IDs (Ansolabehere 2008; Atkeson et al. 2010, 2014; Cobb et al 2012).

This study uses the national voter files merged with state election administration data to examines the casual impact of ID laws on individual voter turnout, and for racial and age subgroups. We use the 1% random draws of 2012 and 2020 voter files (Catalist) that cover nearly all Americans (7 million cases) with voter histories (panel data). We also use the full Catalist population data for all 330 million Americans with turnout by racial and ethnic groups for states and counties. The study seeks to provide a definitive answer to the question of whether ID requirements create unique barriers to voting. The results uncover the negative effects of some forms of ID laws on Hispanic and youth voters. Despite the focus on race, youth voters are the most disadvantaged by identification laws.

Authors

• Caroline J. Tolbert, University of Iowa

• Joseph Coll, University of Iowa