Research on political representation has traditionally focused on the design of electoral systems. Yet there is evidence that voting costs result in lower turnout and undermine voters' confidence in the electoral system. Election administrators can selectively manipulate participation costs for different individuals and groups, leading to biased electoral outcomes. Quantifying the costs of voting and designing fair, transparent and efficient rules for voter assignment to polling stations are important for theoretical and practical reasons. Using analytical models, we quantify the differential costs of participation faced by voters, which we measure in terms of distance to polling stations and wait times to cast a vote. To estimate the model parameters, we use real-world data on the 2013 midterm elections in Argentina. The assignment produced by our model cut average voting time by more than 27%, underscoring the inefficiencies of the current method of alphabetical assignment. Our strategy generates better estimates of the role of geographical and temporal conditions on electoral outcomes.