Using Brazilian data, we estimate the impact of commuting duration on students' performance using distance to closest schools as an instrumental variable. In addition, we also use identification through heteroskedasticity, which does not rely on exclusion restrictions, and estimate bounds for the treatment effect in case there is remaining bias from unobservables. We find strong and consistent evidence that duration of commuting has a negative causal effect on academic achievement. Moreover, the relatively small heterogeneity across quantiles of test score makes us believe that public transportation policies targeted at students can promote education not only for those less well‐off.