There are large spatial disparities in unemployment durations across the 1,300 municipalities in the Ile-de-France region (Paris Greater Area). In order to characterize these imbalances, we estimate a proportional hazard model stratified by municipality on an exhaustive dataset of all unemployment spells starting in the first semester of 1996. This model allows us to recover a survival function for each municipality that is purged of individual observed heterogeneity. We show that only 30% of the disparities in the observed determinants of the survival rates relate to individual variables. Nearly 70% of the remaining disparities are captured by local indicators, mainly segregation indices.