Using spatial data for targeting the poor: Reaching the most vulnerable

The lockdown of mega cities and the perceived trade-off between lives and livelihoods throws up immense policy challenges. We take up the social protection side of this challenge by constructing a targeting tool that pays attention to economic vulnerabilities, as well as requirements of disease containment.

Our proposed tool geo-maps Karachi on the basis of electricity consumption. Using geo-tagged data of 2.4 million households in Karachi, we can (i) predict electricity consumption of any location in Karachi with 70% accuracy, and (ii) provide distributional moments of any neighbourhood or geographical area within Karachi. Both the prediction and the estimation of consumption can be used to target social assistance geo-spatially.

Additionally, collaborators at the University of Bristol are using housing features and physical barriers, such as roads and nullahs, to create ‘natural’ neighbourhoods. It is these neighbourhoods for which the tool will provide distributional moments. Additionally, these neighbourhoods will contain information on population density based on satellite imagery. The juxtaposition of density data with socio-economic data enables the tool to be responsive, not just to economic vulnerability but also to correlate disease spread.

Our tool is designed in a way that it can, with small changes, use geo-tagged data from different sources. In the future, we plan to implement the tool using cell-phone usage data (depending on availability), thus providing a validation for the current exercise. As an immediate next step, we plan to use asset data from the Peoples Poverty Reduction Program to geo-map asset poverty in 16 districts of Sindh.