Movers and Shakers, Also Voters: Do Automated Address Updates Increase Turnout?

• In Event: Impact of (Non-political) Life Events on Voter Registration

Abstract

Automatic Voter Registration (AVR) , passed by16 states and the District of Columbia to date, simplifies voter registration by making it an automated process for eligible citizens who transact with state government agencies such as the Department of Motor Vehicles (DMV) or, less commonly, agencies that provide social services. While it is clear that this policy can be successful in terms of enlarging and diversifying the voter rolls, it is more challenging to assess its impact on voter turnout. Causal inference is impaired by the absence of a true counterfactual; It is impossible to know what a state’s voter turnout in subsequent elections would have been in the absence of AVR. Both temporal analyses of voter turnout before and after implementation and comparative analyses of voter turnout relative to other states are subject to this important limitation.

The proposed paper uses a regression discontinuity approach to estimate the causal effect of AVR in two states, Oregon and California. The discontinuity leveraged is the date of AVR implementation in each state. Individuals who interact with each state’s DMV prior and subsequent to AVR implementation are alike in every way except for the effect of AVR. We leverage this discontinuity using two research designs premised on the reasons why individuals interact with the DMV.

In the first analysis, we consider driver’s license renewal. We take advantage of the fact drivers licenses in both Oregon and California expire on one’s birthday. A driver’s license expires every eight years in Oregon and every five years in California. In both states, driver’s license renewal is an eligible transaction under AVR. If AVR has an effect on voter turnout, individuals whose birthdates fall in the weeks after implementation of AVR will be overrepresented in the voter rolls relative to individuals whose birthdates fall in the weeks prior to implementation. Take for example the state of Oregon, where AVR was adopted in 2016 but retroactively implemented through 2014. At the time of the 2018 general election, 62.5% of drivers born in the last week of January would been “touched” by AVR as a result of drivers license renewal requirements (specifically, all those whose licenses expired in 2014 through 2018) but only 50% of drivers born in the last week of November (specifically, all those whose licenses expired in 2014 through 2017, but NOT those whose license expired in 2018). As such, an overrepresentation of voters with January birthdates can be attributed to the adoption of AVR. Birthdates are not evenly distributed across the calendar year, but data on the distribution of birthdays from prior voting rolls can be used to determine overrepresentation. In addition to a single, overall estimate of the effect of AVR on voter turnout for the general population, this method allows for subgroup analysis based on gender, age, partisanship, and geography. Individuals may interact with the DMV for reasons beyond drivers licenses renewal, a bias that could cause Type 2 errors in this analysis.

In the second analysis, we take advantage of the fact that many individuals interact with the DMV following a change of address. In California, any change of address reported to the DMV is an eligible interaction under AVR. In Oregon, only a change of address that requires obtaining a new driver’s license, namely a move from out of state, is subject to AVR. In this analysis, we compare voter turnout rates of those who move prior and subsequent to AVR implementation. This sub-sample is useful for analysis because there is a greater certainty of interaction with the AVR within a limited timeline, reducing the likelihood of Type 2 errors, though notably at a cost of generalizability. We will draw our sample for this analysis using data originating from the National Change of Address Registry (NCOA), available for purchase by Experian PLC. The sample will be limited to adults who moved into the state of Oregon in 2013 and 2014 (one year before and one year after the retroactive implementation of AVR) and Californians who changed addresses in the months before and after AVR implementation in April 2018. Current name and address listings for these individuals will be linked with the 2018 Oregon and California Voter files to identify both voter registration status and vote histories. We will compare registration and voter turnout rates between movers prior and subsequent to AVR implementation, leveraging the difference as the causal treatment effect of AVR. Once again, this method allows for subgroup analysis based on gender, age, partisanship, and geography.

Authors

• Ellen Seljan, Lewis & Clark College

• Paul Gronke, Reed College