Evaluate homelessness during pregnancy as a unique, time-dependent risk factor for adverse birth outcomes. 9,995 mothers of children <48 months old surveyed at emergency departments and primary care clinics in five US cities. Mothers were classified as either homeless during pregnancy with the index child, homeless only after the index child's birth, or consistently housed. Outcomes included birth weight as a continuous variable, as well as categorical outcomes of low birth weight (LBW; <2,500 g) and preterm delivery (<37 weeks). Multiple logistic regression and adjusted linear regression analyses were performed, comparing prenatal and postnatal homelessness with the referent group of consistently housed mothers, controlling for maternal demographic characteristics, smoking, and child age at interview. Prenatal homelessness was associated with higher adjusted odds of LBW (AOR 1.43, 95 % CI 1.14, 1.80, p < 0.01) and preterm delivery (AOR 1.24, 95 % CI 0.98, 1.56, p = 0.08), and a 53 g lower adjusted mean birth weight (p = 0.08). Postnatal homelessness was not associated with these outcomes. Prenatal homelessness is an independent risk factor for LBW, rather than merely a marker of adverse maternal and social characteristics associated with homelessness. Targeted interventions to provide housing and health care to homeless women during pregnancy may result in improved birth outcomes.