A demographic survey was used to estimate the level and determinants of perinatal mortality in eight lower socio-economic squatter settlements of Karachi, Pakistan. The perinatal mortality rate was 54.1 per 1000 births, with a stillbirth to early neonatal mortality ratio of 1:1. About 65% of neonatal deaths occurred in the early neonatal period, and early neonatal mortality contributed 32% of all infant deaths. Risk factor assessment was conducted on 375 perinatal deaths and 6070 current survivors. Poorer socio-economic status variables such as maternal and paternal illiteracy, maternal work outside the home and fewer household assets were significantly associated with perinatal mortality as were biological factors of higher parental age, short birth intervals and poor obstetric history. Multivariable logistic analysis indicated that some socioeconomic factors retained their significance after adjusting for the more proximate biological factors. Population attributable risk estimates suggest that public health measures for screening of high-risk women and use of family planning to space births will not improve perinatal mortality substantially without improvement of socio-economic conditions, particularly maternal education. The results of this study indicate that an evaluation of perinatal mortality can be conducted using pregnancy histories derived from demographic surveys.