This article examines the impact of family-level clustering on under-five mortality risk by a Bayesian approach. A proportional hazard model with multiplicative random effect is applied to a sample of 6,804 children's survival times. This data set results from the 1998/99 Demographic and Health Survey conducted in Ivory Coast. When the frailty is Gamma distributed, a variance of 0.32 is obtained, which indicates that family membership significantly affects child mortality risk.