Do Double-Registered Voters Strategically Choose Where to Cast Their Ballot?

This paper investigates whether U.S. voters registered in two states strategically choose where to vote—favoring swing states or states with lower voting costs like mail-in ballots. Using 2020 data from 203 million voter registrations and linked cellphone records, the authors estimate 6.1 million double-registrants (3.1% of voters). These individuals disproportionately voted in swing states (increasing likelihood by 4.7 percentage points or 48%) and states that auto-mailed ballots (increasing likelihood by 20%). Wealthier and Democratic voters were overrepresented. Evidence of similar behavior was found in 2016. Although double voting was rare (0.031%), strategic state selection raises fairness and integrity concerns.

Full Citation and Link to Article

Dahl, Gordon B., Joseph Engelberg, Runjing Lu, and William Mullins. Cross‑State Strategic Voting. NBER Working Paper No. 30972 (February 2023; revised September 2024). https://ssrn.com/abstract=4354261 or http://dx.doi.org/10.2139/ssrn.4354261 

Extended Summary

Central Research Question

This study asks whether U.S. citizens who are registered to vote in two different states strategically choose where to cast their vote based on electoral incentives (such as whether a state is a swing state) and logistical costs (such as whether a state mails ballots automatically). The paper introduces the concept of Cross-State Strategic Voting (CSSV), which captures this behavior and explores its prevalence, magnitude, and potential impact on electoral integrity and fairness.

Previous Literature

While the literature on strategic voting has traditionally focused on who people vote for (e.g., choosing a less-preferred candidate to avoid wasting a vote), this paper shifts the focus to where people vote. Past research has shown that individuals are more likely to vote when elections are competitive (Blais 2006; Alvarez et al. 2006) and when voting is made easier through reforms such as early voting or mail-in ballots (Gerber et al. 2013; Braconnier et al. 2017). However, prior studies have not considered the strategic implications of being registered in multiple states. Additionally, legal scholarship and empirical work have examined voter fraud and administrative shortcomings (e.g., Minnite 2010; Goel et al. 2020), but little has been done to quantify the extent of double-registration or the behavioral implications thereof.

Data

The primary data source is the L2 voter file, a comprehensive commercial database of voter registration information in all 50 states and the District of Columbia as of October 2020. It includes individual-level data on names, addresses, birthdates, registration dates, and voting history, as well as cell phone numbers and inferred partisanship.

To identify double-registrants, the authors match voters across states using a combination of identifiers: same first and last name, same birthdate (month, day, and year), gender, and cell phone number. The analysis also uses supplementary data such as housing wealth (via Zillow’s home value index), IRS tax records for zip code–level income, PredictIt betting market odds for identifying swing states, and CoreLogic’s property database to identify dual-state homeowners.

Using these sources, the authors construct multiple robust samples of double-registrants: one using cellphone matching, another using uniquely named individuals with identical birthdates, and a third focused on voters who own homes in two different states.

Methods

The study begins by estimating the number of voters registered in two states at the same time. The authors apply the Law of Total Probability to pairs of voters with the same name and birthdate across states, comparing the observed frequency of matching birthdates to the probability expected by chance. This probabilistic matching approach yields an estimate that 6.1 million individuals (3.1% of all voters) were double-registered as of October 2020.

To investigate voting behavior, the authors analyze whether double-registrants voted in the “first” state (where they were registered earlier) or the “second” state (more recent registration). They regress the probability of voting in the first state on indicators for whether each state is a swing state or mails ballots automatically, controlling for potential confounders and testing robustness across subgroups. The authors also analyze subsets of “near-certain” voters—those who voted in the 2020 primary or have a ≥99% predicted probability of voting—using an “identification at infinity” approach to isolate the choice of where to vote from the decision whether to vote.

They further test robustness by examining results from the 2016 election, by comparing across states with different election laws, and by analyzing variations in electoral competitiveness using continuous measures (such as betting market–implied “win gaps”).

Findings/Size Effects

The authors find strong evidence that double-registrants behave strategically in choosing where to vote:

• Swing State Incentive: Double-registrants are 4.7 percentage points more likely to vote in a first state if that state is a swing state and the second is not—an increase of 48% relative to the baseline 9.8% voting rate in first states. Conversely, they are 1.9 percentage points less likely to vote in the first state if only the second is a swing state.

• Voting Cost: If the first state automatically mails ballots and the second does not, double-registrants are 20% more likely to vote in the first state. If only the second state mails ballots, the likelihood of voting in the first state drops by 31%.

• Double Voting: Though extremely rare, the study finds a 0.031% rate of double voting across the entire voter population. Among double-registrants, double voting is more likely when both states are swing (increasing the rate by 41%) or both states mail ballots automatically (raising the rate by 111%).

• Wealth and Income Disparities: Double-registration is disproportionately concentrated among the wealthy. The double-registration rate is 2.2% in the bottom quartile of home values but 8.0% in the top 1%. Similar patterns hold using income. Democratic and Independent voters are also more likely to be double-registered than Republicans.

• Behavior Among Dual-State Homeowners: These individuals display even more pronounced CSSV. Among near-certain voters, the likelihood of voting in a swing first state (versus a non-swing second state) increases by 13.1 percentage points—significantly larger than the baseline sample.

• Cross-Year and Sample Robustness: Similar CSSV behavior is observed in the 2016 election using a unique-name sample. Results remain consistent across swing-state definitions, voter roll quality, migration flows, and tax differences.

• Relative Closeness Effects: Using continuous measures of electoral competitiveness, the authors show that voters are more likely to vote in states where the predicted election margin is narrower—further validating the incentive-based component of CSSV.

Conclusion

The study documents a previously unexamined form of strategic voting: selecting which state to vote in when registered in multiple states. With an estimated 6.1 million double-registrants in 2020, the phenomenon is widespread. These individuals tend to be younger, wealthier, and more mobile—and they often vote in the state where their vote is more impactful or easier to cast. This raises equity concerns, as voters with more resources and mobility can amplify their electoral influence by choosing their state of participation.

While CSSV did not affect the outcome of the 2020 presidential election—due to offsetting behaviors among Democrats and Republicans—it could matter in close elections like those in 2000. The authors argue that campaigns or political operatives might seek to mobilize double-registrants in future close contests, increasing the stakes.

The paper also highlights the tension between election accessibility and integrity. There is no central mechanism to track voter registration across states, and individuals are not automatically deregistered when they move. Policies aimed at reducing double-registration may prevent CSSV but risk disenfranchising legitimate voters unless implemented carefully.

Finally, CSSV undermines the principle of “one person, one vote,” especially given its concentration among the wealthy. As courts and legislators continue to address election fairness and integrity, the authors call for greater attention to cross-state strategic behavior and its implications for democratic representation.