Discussion Paper no. 04-09


Recent research on child mortality suggests that despite overall declines in mortality levels, mortality could remain relatively high due to a clustering of deaths in certain pockets of the population. Using data from the 1998 demographic and health survey (DHS) of Kenya this study examines child death clustering within families. The analysis first examines how known distant determinants of child mortality including socioeconomic and socio-cultural factors affect child survival, and how their effect is modified by proximate determinants of mortality including biodemographic and household environmental conditions. This is then followed by a control for family random effects to establish whether mortality risks in families are correlated net of the measured factors. The results show that biodemographic factors are more important in infancy while socioeconomic factors and household environmental conditions have a greater effect in childhood. Additionally, there is significant familial variation in child mortality risks even after controlling for measured determinants of mortality. This correlation of mortality risks suggests that there exits unobserved or unobservable familial factors related to mortality which lead to a concentration of deaths in certain families.