The value of time (VOT) determines the allocation of non-labor time to tasks and it is especially salient in travel demand and infrastructure, with US commuters estimated to lose 97 hours due to traffic congestion in 2018. The economics literature so far has focused on travel time savings and estimating the mean VOT, but there is considerable heterogeneity in VOT, particularly with high-VOT individuals, that is still understudied. This paper fills the gap by estimating the full VOT distribution for a population of drivers, using the setting of highway Express Lanes (ELs), which offer time savings in exchange for a posted toll. I construct a novel panel dataset of over 2 million observations that matches repeated EL drivers’ choices with extremely fine traffic data to measure time saved. The EL toll changes every 3 minutes according to a continuous function of traffic in the EL and is then rounded to the closest $0.25, thus creating 32 discontinuities up to the $8 toll cap, which provides the identification strategy. First, the reduced-form evidence from an RDD shows that EL drivers have a mean VOT of $66.56 per hour saved. Second, the RDD results together with aggregate traffic data can be used to back out the full VOT distribution for the entire population of drivers. The distribution has a median of $21.67 per hour and a long right tail. Third, a structural model provides a full characterization of the commuting choice, where drivers choose departure time and then, conditional on that, choose whether to use the EL. The distribution of VOT and the average departure time preference parameters allow to predict the effect of counterfactual congestion policies. I find that reconverting the EL into a standard lane increases per-capita welfare by $25.68 per year. Finally, if the composition of the VOT distribution of drivers is changed to include more low-VOT drivers, keeping the EL in place modestly increases per-capita welfare but disproportionately hurts low-VOT drivers.