We combine two natural experiments, multiple empirical strategies and administrative data to study voters' response to marginal changes to the fine for electoral abstention in Peru. A smaller fine leads to a robust decrease in voter turnout. However, the drop in turnout caused by a full fine reduction is less than 20% the size of that caused by an exemption from compulsory voting, indicating the predominance of the non-monetary incentives provided by the mandate to vote. Additionally, almost 90% of the votes generated by a marginally larger fine are blank or invalid, lending support to the hypothesis of rational abstention. Higher demand for information and larger long-run effects following an adjustment to the value of the fine point to the existence of informational frictions that limit adaptation to institutional changes.