An assessment of the Young Lives sampling approach in Ethiopia

The sampling methodology adopted by Young Lives is known as a sentinel site surveillance system. In Ethiopia, the Young Lives team used multi-stage, purposive and random sampling to select the two cohorts of children. This methodology randomised households within a study site while the sites themselves were chosen on the basis of predetermined criteria, informed by the Young Lives objectives. To ensure the sustainability of the study, and for resurveying purposes, a number of well-defined sites was chosen. The sites were selected with a pro-poor bias and to ensure a balanced representation of the Ethiopian regional diversity as well as rural/urban differences. This paper assesses the sampling methodology by comparing the Young Lives sample with larger, nationally representative samples. In doing this, the Ethiopia team sought to: (i) analyse how the Young Lives children and households compare with other children in Ethiopia in terms of their living standards and other characteristics; (ii) examine whether this may affect inferences between the data; (iii) establish to what extent the Young Lives sample is a relatively poorer or richer subpopulation in Ethiopia; (iv) determine whether different levels of living standards are represented within the dataset. We found that households in the Young Lives sample were slightly wealthier than households in the DHS sample. Further analysis revealed that households in rural areas and in urban areas, except Young Lives households in Addis Ababa, were wealthier. Young Lives households in Addis Ababa were poorer than households in the DHS sample. A similar picture emerged when we use t-tests to compare the means for a range of living standard indicators between the Young Lives and the DHS samples. Young Lives households in rural areas had better access to public services such as drinking water and electricity supply, while households in Addis Ababa had less access to basic services. These findings were supported by the comparison of common variables in Young Lives and the WMS. However, Young Lives households were less likely to own land or a house, and had smaller livestock holdings than WMS households. To assess trends over time we compared the Young Lives sample with the DHS 2005 sample. Some of the differences, which we observed in the comparison of Young Lives with the DHS were reduced which indicates some improvements in living standards between 2000 and 2005. The analyses show that households in the Young Lives sample were slightly better-off and had better access to basic services than the average household in Ethiopia, as measured by the nationally representative DHS and the WMS. However, our detailed analysis reveals that, while Young Lives households are located at sites with better access to services and utilities, they hold less land, less livestock. And are less likely to own their own house than the average Ethiopia household. This evidence is consistent with the sampling methodology applied with the Young Lives samples in Ethiopia. Despite these biases, it is shown that the Ethiopian Young Lives sample covers the diversity of children in the country. Therefore, while not suited for simple monitoring of child outcome indicators, the Young Lives sample will be an appropriate and valuable instrument for analysing causal relations, modelling child welfare, and its longitudinal dynamics in Ethiopia