Election integrity is a cornerstone of democracy. In the US, voters who move will often have active registrations in more than one state, opening up the possibility that they can choose where to vote. This raises fairness concerns. This column uses the universe of US voters in 2020 to measure the number of double-registered individuals and their characteristics. Double registrants respond to cross-state strategic incentives and costs. They disproportionately choose to vote in swing states and in states which automatically send out mail-in ballots.
Faith in political institutions is critical to a well-functioning democracy, but this trust is falling across high-income countries (Zoega et al. 2017, Melnikov et al. 2019). In the US, a significant share of people express doubts about how elections are managed (Marist Poll 2021, MIT Election Lab 2021). Attention has focused on within-state issues like voter ID laws, polling stations, and the certification process. There has been less focus on cross-state election concerns, in part because – unlike other countries with federal systems – US states run their elections independently without national coordination on voter registrations across states.
This decentralised approach can cause problems when voters move from one state to another. Without central coordination, voters who move will often have active registrations in more than one state, opening up the possibility that they can choose where to vote. This raises fairness concerns, since some individuals can choose where to vote while others cannot, and is particularly relevant in the US, where election outcomes are determined by a handful of swing states. Despite these concerns, little is known about the prevalence and strategic behaviour of double registrants.
In recent work (Dahl et al. 2024), we study how many individuals were registered to vote in two states for the 2020 election, along with their characteristics. We also examine whether these double registrants respond to incentives and costs, that is, whether they disproportionately choose to vote in ‘swing states’ (where voters have a higher incentive to vote) or in states which automatically send out mail-in ballots (which makes voting less costly).
Definition of double registrant
A double registrant is an individual listed on two different state voter rolls at the same point in time. The lack of a national voter registry and limited cross-state coordination means double registration mainly arises when individuals move across state lines. Under the National Voter Registration Act of 1993, most states must wait at least four years before removing inactive voters. Therefore, most individuals will be double-registered for at least a few years if they were registered in their old state and register in their new state after a move.
Size of the double-registrant population
We use voter registration data from L2, a prominent non-partisan data provider frequently used by political campaigns and in academic research (e.g. Allcott et al. 2020, Brown et al. 2021). To estimate the number of double registrants, we identify individuals with the same name and birthdate in two different states, adjusting for the likelihood of these matching by chance alone. The intuition is as follows. Because there are 365 days in a year, we would expect two different individuals to have the same birthday 1 out of 365 times. However, when we pair registrations that have the same name in two different states, we find they have significantly greater odds of having the same birthday. This means that some of these pairs are in fact the same individuals, and we can estimate the number of double registrants by seeing how often these paired registrations occur beyond chance. Using this approach, we estimate that there were a total of 6.1 million double registrants, or 3.1% of voters on state voter rolls, as of October 2020.
We repeat this analysis for subgroups based on wealth and income and find much higher rates of double registration among the rich (see Figure 1). Specifically, using property wealth data from Zillow at the zip-code level, the double-registration rate for individuals in the bottom quartile of housing prices is 2.2%, while it rises to 8.0% for those in the top 1%. Similarly, using IRS tax data at the zip-code level, the double-registration rate for individuals in the bottom quartile of income is 2.8% and increases to 6.8% for the highest-income earners.
Figure 1 Double-registration rates by housing wealth and income
A sample of known double registrants
So far, we have discussed the overall number of individuals registered to vote in two states but have not identified specific individuals who are double registrants. To create a sample of known double registrants for analysis, we use characteristics from the L2 database. Specifically, we identify all pairs of voters from different states who have the same first name, last name, gender, birth year, birth month, and, importantly, the same cell phone number. We estimate that registered voters in different states with these matched characteristics have a 99.999% chance of being the same person.
Cross-state strategic voting
This approach gives us a subsample of 583,835 known double registrants, which can be used to analyse which state they choose to vote in. Because elections in swing states have a greater influence on the outcome of a presidential election, this may incentivise voters to prioritise these states over other states whose election outcomes are more certain. In addition, some states lower the convenience costs of voting by automatically mailing ballots to registered voters. We refer to the behaviour of choosing where to vote based on these factors – incentives and costs – as cross-state strategic voting.
Cross-state strategic voting has straightforward predictions. For example, consider two individuals who are registered in California, one who moved from Arizona (a swing state) and the other who moved from Oregon (not a swing state), both of whom still have registrations in their old states. Cross-state strategic voting predicts the first individual is more likely to vote in their old state because Arizona is swing but Oregon is not. Likewise, cross-state strategic voting predicts that double registrants are more likely to vote in states which send out auto-mail ballots.
As background, most states have laws that require individuals to vote in their state of legal domicile, which refers to the state where someone lives and intends to remain. While a person can have multiple homes, they are only allowed one legal domicile at any given time. Governments typically determine domicile by considering factors such as where individuals hold a driver’s licence, pay state taxes, or spend most of their time. This means that the choice of which state to vote in for double registrants is potentially a legal grey area.
In our sample of double-registered voters, we classify the state with the earlier registration as the ‘first’ state and the one with the more recent registration as the ‘second’ state. It is more likely that the second state represents the voter’s current legal domicile, as this is where they last registered. Consistent with this idea, only 10% of double registrants in our sample voted in their first state during the 2020 presidential election.
Evidence of strategic behaviour
Figure 2 provides evidence for both types of strategic behaviour by examining the relative likelihood a double registrant votes in their first state in the 2020 election. We focus on individuals who have a near-certain probability of voting, so we can isolate where they vote rather than whether they vote at all.
The blue bars show how the likelihood of voting in the first state varies with whether the first, the second, or both states are swing. When only the first state is swing, a double registrant is 5.9 percentage points more likely to vote in their first state than when neither state is swing. In contrast, when only the second state is swing, they are 4 percentage points less likely to vote in the first state. Likewise, the purple bars reveal that when only the first state has auto-mailed ballots, individuals are more likely to vote there, while the opposite is true when only the second state has auto-mailed ballots. When both states are swing (or auto-mailed), we find little difference in the likelihood of voting in the first state. These patterns match the predictions of cross-state strategic voting.
A similar analysis for the 2016 election finds similar results.
Figure 2 Evidence of strategic behaviour: Likelihood of voting in 1st state
Discussion
A back-of-the-envelope calculation indicates that cross-state strategic voting did not affect the outcome of the 2020 election. However, it could be important in closer elections such as the 2000 election – where the outcome was decided by 537 votes in Florida. There is also the possibility that campaigns could target double registrants, encouraging more of them to cast a ballot in swing states to increase the impact of cross-state strategic voting.
Perhaps the fundamental concern with cross-state strategic voting is that it undermines the democratic principle of ‘one person, one vote’ by granting certain voters more power than others. However, reforms targeting double registrants face inevitable tradeoffs. While reforms may reduce the number of double registrants, these come at a potential cost to political enfranchisement and state autonomy.
Source : VOXeu
