This study asks whether strict voter identification (ID) laws affect voter turnout in the United States and whether effects differ by election type and by when states adopted these laws. Using state-level turnout from 1984–2020 and a synthetic difference-in-differences approach to address time-varying confounders and improve power, the paper compares aggregate and disaggregated effects across presidential versus midterm elections and between early- and late-adopting states. The central contribution is evidence that aggregate effects are close to zero but that heterogeneous, modest impacts appear in some contexts and may help reconcile mixed prior findings.
State-level turnout panel (1984–2020) Synthetic difference-in-differences Heterogeneous effects by election type and adoption timing
Data and methods
The author compiles state-level turnout data covering 1984–2020 and applies a synthetic difference-in-differences design intended to control for time-varying confounders and to increase statistical power. The analysis disaggregates treatment effects by election type (presidential vs. midterm) and by whether states were early adopters (pre-2008) or late adopters (post-2008) of strict voter ID laws.
Main findings
In aggregate, strict voter ID laws show no statistically significant effect on turnout. Disaggregated results reveal heterogeneous patterns: in presidential elections, late-adopting states experience a decrease in turnout (reported as about 2.7 percentage points), while early-adopting states show no significant change. For midterm elections, the primary analysis finds an increase in turnout (reported as about 2.9 percentage points), with larger estimated effects among early-adopting states, though that midterm finding is not robust across all specifications.
- Aggregate results are close to zero, reflecting offsetting small effects across contexts.
- Heterogeneity by election type and adoption timing may explain mixed findings in prior literature.
