BACKGROUND Advancing an understanding of childbearing desires is an important precursor to achievement of the policy goal of reducing unintended pregnancies. It has been long debated that concepts of fertility desires and planning may be particularly problematic in sub-Saharan Africa. However, examination of the utility of fertility preference measures and their link to reproductive behaviour is still rare in the region. The aim of this study is to assess the predictive validity of future childbearing desires on subsequent reproduction among women living in the highly unpredictable circumstances of Nairobi slums. METHODS We used data from a longitudinal study (2007-2010) nested in the Nairobi Urban Health Demographic Surveillance System that is located in two slums in Nairobi, Kenya. We analysed baseline fertility desires among 4577 postpartum women. Cox proportional hazard model was employed to examine the effect of fertility desires on subsequent reproduction. RESULTS One-third of the women wanted no more children and 37% wanted to wait for at least five years at baseline. While two-thirds of the women who wanted to have a child soon became pregnant within three years, less than one-third of those wanting no more children became pregnant. The multivariable analysis shows that the probability of becoming pregnant among women who expressed desires to stop or delay childbearing at least for five years was 50% less than among women who wanted to have a child in two to four years. In addition to prospective fertility desires, level of woman's education, residence and ethnicity exerted important influences on implementation of baseline preferences. CONCLUSIONS Our study finds a strong link between baseline fertility desires and subsequent reproduction. A large difference in pregnancy risk was observed between those who wanted no more children and those who wanted another child. The link between a woman's stated desire to stop childbearing and subsequent childbearing is just as strong in the Nairobi slums as elsewhere. In addition, the findings revealed a pronounced gradient in pregnancy risk according to preferred spacing length, which support other evidence on the important contribution of long-term spacing or postponement to fertility decline in sub-Saharan Africa.