Background: Guinea Bissau is one of the poorest countries in the world, with one of the highest under-5 mortality rate. Despite its importance for policy planning, data on child mortality are often not available or of poor quality in low-income countries like Guinea Bissau. Our aim in this study was to use the baseline survey to estimate child mortality in rural villages in southern Guinea Bissau for a 30 years period prior to a planned cluster randomised intervention. We aimed to investigate temporal trends with emphasis on historical events and the effect of ethnicity, polygyny and distance to the health centre on child mortality. Methods: A baseline survey was conducted prior to a planned cluster randomised intervention to estimate child mortality in 241 rural villages in southern Guinea Bissau between 1977 and 2007. Crude child mortality rates were estimated by Kaplan-Meier method from birth history of 7854 women. Cox regression models were used to investigate the effects of birth periods with emphasis on historical events, ethnicity, polygyny and distance to the health centre on child mortality. Results: High levels of child mortality were found at all ages under five with a significant reduction in child mortality over the time periods of birth except for 1997-2001. That period comprises the 1998/99 civil war interval, when child mortality was 1.5% higher than in the previous period. Children of Balanta ethnic group had higher hazard of dying under five years of age than children from other groups until 2001. Between 2002 and 2007, Fula children showed the highest mortality. Increasing walking distance to the nearest health centre increased the hazard, though not substantially, and polygyny had a negligible and statistically not significant effect on the hazard. Conclusion: Child mortality is strongly associated with ethnicity and it should be considered in health policy planning. Child mortality, though considerably decreased during the past 30 years, remains high in rural Guinea Bissau. Temporal trends also suggest that civil wars have detrimental effects on child mortalit