Rates of childhood obesity have increased dramatically in the last few decades. Non-causal evidence suggests that childhood obesity is highly persistent over the life cycle. However little is known about the origins of this persistence. In this paper we attempt to answer three questions. First, how do anthropometric measures evolve from birth through primary school? Second, what is the causal effect of past anthropometric outcomes on future anthropometric outcomes? In other words, how important is state dependence in the evolution of anthropometric measures during the early part of the life cycle. Third, how important are time-varying and time invariant factors in the dynamics of childhood anthropometric measures? We find that anthropometric measures are highly persistent from infancy through primary school. Moreover, most of this persistence is driven by unobserved, time invariant factors that are determined prior to birth, consistent with the so-called fetal origins hypothesis. As such, interventions designed to improve child anthropometric status will only have meaningful, long-run effects if these time invariant factors are altered. Unfortunately, future research is needed to identify such factors, although evidence suggests that maternal nutrition may play an important role.