Child mortality, defined here as mortality under age five, is not evenly distributed but found in clusters. In a contemporary polygamous population in Ghana with extended families, we separate clustering at the parental and household levels, which are often overlapping and inseparable in other historical studies. For eight years, we followed 28,994 individuals, including 9,288 children under the age of five, in 1,703 households. We identified four determinants that had a significant effect on child mortality: sex of the child, age of the child, drinking source, and socioeconomic status. After correcting for these determinants, we still identified significant clustering of child mortality at the level of the village (covariance [cov] = 0.02, p = .04), household (cov = 0.14, p = .003), father (cov = 0.24, p = .001), and mother (cov = 0.18, p = .05). The present data provide clues regarding the levels at which to look for unidentified determinants of child mortality and suggest that the importance of the father could be larger than previously thought.