Abstract
How much can automatic voter registration (AVR) increase turnout? Unlike many democracies, most American voters face the additional cost of registration, resulting in potential disenfranchisement. AVR is naturally expected to promote turnout, but its causal effects have rarely been quantified due to violations of crucial assumptions. I show that a variation of AVR that targets existing registrants as opposed to eligible non-registrants---termed automatic re-registration (ARR)---increases turnout by 5.8 percentage points. I exploit a natural experiment in a novel administrative dataset; election officials in Orange County, California, notified existing registrants who moved within-county that their residential addresses were automatically updated. The treatment alleviated registrants of re-registration burdens, but only for those who moved before the legal cutoff date, enabling a quasi-random treatment assignment. Contrary to the popular narrative, ARR had no significant effect on the turnout of registered Democrats, but Republicans' and nonpartisans' turnout increased by 8.1 and 7.4 percentage points.
Version notes
Title change + 2022 APSR conditional accept minor changes
Content

Supplementary material

Online Appendix
Supporting Information