We examine the urban learning and schooling premium in Peru using two approaches: 1) estimating the effect of local population density on pupils’ value-added in learning between grades two to eight, 2) quasi-experimental Census estimations on the effects of the duration of urban exposure in childhood on school attainment. Unconditional estimates show that a ten-fold increase in population density is associated with around 0.13 standard deviations higher value-added in learning. Conditional estimates suggest that this association is driven largely by sorting, captured mainly by household socio-economic status, and local area factors, reflecting agglomeration benefits. School resources, in turn, are unimportant. Finally, a higher population density also correlates with unobserved factors that can harm learning. The quasi-experimental estimations confirm a positive relationship between urban exposure and educational outcomes: a longer period of urban education leads to higher educational attainment and a higher likelihood of enrolment in the correct grade for age.