Child Poverty Estimation and Analysis Using Household Survey and Geospatial Data, Alessandro Carraro (UNICEF)

Two practical applications combining household survey and geospatial data are presented. First, administrative level one data from household surveys are combined with geospatial data to project child poverty headcount rates at administrative level two. This analysis is carried out using the same indicators and thresholds across countries (to estimate child poverty nationally and at administrative level one) and various machine learning models to generate the estimates (at administrative level two).
The first part of the paper includes a presentation of results and discussion of limitations of this methodology. In addition, the paper includes a practical application combining georeferenced data and survey data. In this case, the child poverty subnational data used in the first part are combined with high-resolution geographical data collected in the UNICEF Children’s Climate Risk Index (CCRI). Combining these two sets allows to analyze the relationship between child poverty and environmental risks.