Study Significance
The authors evaluated strict voter identification laws because these laws are promoted to deter fraud but raise concerns about selective disenfranchisement; the study tests whether such laws actually reduce registration or turnout or produce unequal effects across demographic groups.
Data and Measurement
They used a nationwide panel of voter records covering 2008–2018 with about 1.6 billion observations and measured voter registration and turnout, along with subgroup indicators (race, gender, age, party) and measures of campaign contact, contributions, political engagement, and fraud (actual or perceived).
Comparison Design and Controls
They implemented a difference-in-differences design exploiting variation in state adoption of strict ID laws, reporting many specifications and a most-demanding model that included state, year, and voter fixed effects plus state- and voter-level time-varying controls.
Analysis and Estimation
📊 Nationwide Panel and Natural Experiment: Uses a difference-in-differences design on a nationwide panel of voter records covering 2008–2018 with 1.6 billion observations. A large range of specifications is reported; the most demanding model includes state, year, and voter fixed effects plus state- and voter-level time-varying controls.
Main Findings and Interpretation
Across specifications, strict ID laws had no negative effect on registration or turnout overall or for any group defined by race, gender, age, or party; the most demanding specification produced point estimates of −0.1 percentage points for both registration and turnout (95% CIs: registration [−2.3; 2.1], turnout [−3.0; 2.8]).
The estimated effect on nonwhite turnout relative to whites was +1.4 percentage points (95% CI: [−0.5; 3.2]), and the likelihood that nonwhite voters were contacted by a campaign increased by 4.7 percentage points, suggesting party mobilization may have offset modest barriers for minorities; the laws also had no detectable effect on fraud, actual or perceived.
The authors conclude that strict ID requirements did not measurably reduce registration or turnout in their data and suggest efforts to improve elections may be better directed at other reforms.
