Child mortality (the mortality of children less than five years old) declined considerably in the developing world in the 1990s, but infant mortality declined less. The reductions in neonatal mortality were not impressive and, as a consequence, there is an increasing percentage of infant deaths in the neonatal period. Any further reduction in child mortality, therefore, requires an understanding of the determinants of neonatal mortality. 209,628 birth and 2581 neonatal death records for the 1998 birth cohort from the city of São Paulo, Brazil, were probabilistically matched. Data were from SINASC and SIM, Information Systems on Live Births and Deaths of Brazil. Logistic regression was used to find the association between neonatal mortality and the following risk factors: birth weight, gestational age, Apgar scores at 1 and 5 minutes, delivery mode, plurality, sex, maternal education, maternal age, number of prior losses, prenatal care, race, parity and community development. Infants of older mothers were less likely to die in the neonatal period. Caesarean delivery was not found to be associated with neonatal mortality. Low birth weight, pre-term birth and low Apgar scores were associated with neonatal death. Having a mother who lives in the highest developed community decreased the odds of neonatal death, suggesting that factors not measured in this study are behind such association. This result may also indicate that other factors over and above biological and more proximate factors could affect neonatal death.