Does Living Close to a Polling Place Make You More Likely to Vote?

Full Citation and Link to Article

Bagwe, G., Margitic, J., & Stashko, A. (2025). Polling place location and the costs of voting. Journal of Public Economics. https://www.aeaweb.org/articles?id=10.1257/pol.20220435&&from=f

Short Summary

This article examines how distance to polling places influences voting behavior, using administrative data from more than 15 million voters in Pennsylvania and Georgia. Employing a geographic regression discontinuity design, the authors find that each additional mile reduces in-person voting by 1–3 percentage points. The effect is stronger among those living closest to polling sites: in Pennsylvania, a one-mile increase leads to a 7.7-point decline for voters within 0.4 miles of their polling place. In Georgia, mail-in voting offsets these losses, but in Pennsylvania—where absentee voting was more restricted—overall turnout decreases significantly .

Extended Summary

Central Research Question

The study investigates how the distance from a registered voter’s residence to their assigned polling place affects voter participation. The central question is whether administrative decisions about polling place locations influence turnout levels by increasing or decreasing the “cost” of voting.

Previous Literature

Research on polling place accessibility has produced widely varied findings. Some studies, such as Clinton et al. (2020) and Yoder (2018), reported negligible effects, while Brady and McNulty (2011) observed that a one-mile increase in distance reduced turnout by about one point. Cantoni (2020), in contrast, found much larger effects, estimating turnout declines of 4 to 12 points per mile in Massachusetts and Minnesota. These discrepancies may stem from differences in geography, methodology, or context. Scholars have emphasized that small costs, such as distance or waiting times, can significantly influence participation, especially in highly contested elections. This study addresses the gaps by applying a rigorous regression discontinuity design in two large swing states with different voting rules.

Data

The analysis uses administrative voter registration and history files from Pennsylvania and Georgia for the 2018 midterm elections, covering over 15 million individuals. The data include home addresses, polling place assignments, and whether voters abstained, voted in person, or voted by mail. Geographic information was used to calculate both road-route distances and travel times. The dataset was supplemented with U.S. Census and American Community Survey data to capture socioeconomic characteristics such as income, education, and transportation modes. Observations with potential geocoding errors or implausibly long distances (over 10 miles) were excluded, improving precision.

Methods

The authors implement a geographic regression discontinuity (RD) design, exploiting the fact that voters living near precinct borders are similar in demographics but may be assigned to different polling places. By dividing borders into segments to satisfy monotonicity, they ensure that moving across a boundary consistently changes polling place distance. This fuzzy RD approach provides local average treatment effects. Additional models include fixed-effects regressions and matching methods to compare results with prior studies. To assess policy implications, the authors also conduct counterfactual simulations to identify turnout-maximizing polling locations using existing infrastructure.

Findings/Size Effects

The results show consistent evidence that distance reduces in-person voting. In Pennsylvania, each additional mile reduces in-person turnout by 2.8 points and overall turnout by 2.6 points, with no significant increase in absentee voting. In Georgia, in-person turnout falls by 1.4 points per mile, but this is offset by a 1.1-point increase in mail-in ballots, resulting in a net effect close to zero. Effects are nonlinear: voters closest to polling stations are most sensitive. For instance, in Pennsylvania, those 0.4 miles away lose 7.7 points per mile, compared with a negligible effect for those more than 1.5 miles away. In Georgia, voters 0.9 miles from polling places show a 2.8-point drop, while those over 3 miles away lose only 1.2 points. Travel time analyses corroborate these results, showing declines of about 1 percentage point for each additional minute by car. Heterogeneity analysis indicates that transportation access, rather than demographics or partisanship, drives sensitivity. Urban voters without reliable car access are disproportionately affected. Importantly, counterfactual exercises in Pennsylvania reveal that moving polling places to more optimal locations could increase turnout by 5 percentage points (10%) using existing buildings, and by up to 13 points (26%) if the number of polling places were doubled.

Conclusion

The study concludes that polling place distance is a significant barrier to participation, but its impact depends on the availability of alternatives. In Georgia, no-excuse absentee voting absorbs the effect, while in Pennsylvania, where mail voting was more limited at the time, distance decreases overall turnout. These findings suggest that election administrators wield considerable power over participation through decisions about precinct design and polling site placement. Optimizing these locations could deliver turnout boosts equivalent to costly mobilization campaigns or expansions in early voting. More broadly, the study emphasizes that small costs of voting matter, and that equitable access requires careful attention to the geographic allocation of polling places.