The paper explores the dynamics of child and household poverty in rural Ethiopia using three rounds of household survey and qualitative data collected by Young Lives, a longitudinal study of child poverty. It uses a mixed-method taxonomy of poverty (Roelen and Camfield 2011) to classify children and their households into four groups: ultra-poor, poor, near-poor and non-poor. Survey and qualitative data are then used to analyse the movements in and out of poverty and explore the factors that underpin these movements. The use of mixed methods in both the identification of the poor and analysis of their mobility illustrates that the combined use of qualitative and quantitative information can lead to deeper insights and understandings. The paper reports a reduction in the percentage of poor households from 50 to 20 percent between rounds 1 and 3 (2002-9), following the "stages of progress" posited in Roelen and Camfield (2011). However, these changes were not unequivocally beneficial to children (for example, the acquisition of livestock might mean dropping out of school to herd them). Ultra-poverty proved persistent with little change in the circumstances of the one in ten households classified as ultra-poor, who were vulnerable to illness, lending or "sharecropping-out" land on unfavourable terms and exclusion from the government?s food-for-work scheme.