AbstractStudies of voting behavior in some settings may be hampered by poor data availability or unsuitably large units of aggregation for reported turnout. We propose and demonstrate a practical big-data solution to these kinds of challenges, using fine-grained cell-phone mobility data on millions of GPS locations for more than 300,000 eligible voters in Tokyo. Our approach uses the geolocations of polling stations, combined with GPS data points recorded on election day and a reference day, to measure patterns in individual-level (but anonymized) voting behavior. We first test the validity of the measure by comparing it to official aggregated data on turnout, and then illustrate its substantive utility with an application exploring the well-known relationship between turnout decisions and the cost of voting, proxied by the distance between a voter’s residence and the polling station. Finally, we discuss the potential limitations of the approach and provide step-by-step instructions for other researchers.