This article estimates the causal effect of primary school entry age on children’s cognitive development in rural northwestern China, using data on nearly 1,800 primary school aged children from the Gansu Survey of Children and Families. Instrumental variable estimates, exploiting the discontinuity structure in children’s school entry age around the enrolment cut-off date, indicate that a 1-year delay in school entry reduces children’s scores on a cognitive ability test administered when they were aged 9–12 by 0.11–0.16 standard deviations (of the distribution of test scores). The negative late-school-entry effect is significantly larger in villages with no preprimary schools. It also persists as children advance to higher grades. These findings suggest that delayed school entry, even if it may be rural parents’ rational response to resource constraints, can be harmful for children’s cognitive development in developing areas with underdeveloped preprimary school systems.