A BIG DEAL: How can we use big data to measure poverty in Sudan?

from UN Development Programme
Published on 12 Jan 2016 View Original

Our Perspective

Anisha Thapa, in collaboration with Miguel Luengo – Oroz, Anne Kahl, Abdalatif Hassan, Jennifer Colville and Vasko Popovski

As development practitioners in Sudan, we are facing a major challenge in effectively managing our programmes: the generation of reliable data to regularly monitor development impact, in particular related to changes in household poverty. As part of the cross regional initiative on big data for development exploration, UNDP Sudan is now working to explore how new sources of data can measure key development indicators.

As many other countries, Sudan faces multiple challenges in acquiring frequent and adequate data to assess situation changes and measure implementation of development programmes. Household surveys and censuses are expensive, time consuming, and demand elaborate processes and resources for data collection and analysis. In addition, the intensification of armed conflict in many parts of the country hinder accessibility to these areas for data collection through traditional methods. Timely measurement of indicators, such as poverty thereby suffers, resulting in data gaps that negatively affect efficient adjustments to development interventions and programming.

Electricity and lights at night as proxies for poverty: What we found?

With support from UNDP and UN Global Pulse partnership on big data for development spanning across several COs in the Europe and Central Asia and Arab States region, funded through Global Innovation Facility and Government of Denmark, UNDP Sudan is exploring the potential of alternative big data sources such as electricity consumption and night time lights (from satellite imageries) as proxies to estimate poverty levels in Sudan. We analyzed electricity power consumed and illumination values from 2013 and 2014 with poverty percentages (Figure 1) calculated using standard methods - S3M (2013) and MICS (2014), besides looking at correlations between electricity and lights (Figure 2). As expected, electricity consumption data was correlated with light data, validating the hypothesis that electricity availability (a relevant variable for the multidimensional poverty index) can be measured from the space. We created an interactive visualization (Image 1) that allows to see places where light signal increase – and possibly more interesting- where light decreased or disappeared. When comparing lights at night and poverty indicators, similarly to recent studies in Kenya and Rwanda by the World Bank, we found that the data sources have moderate potential to be a reasonable proxy for poverty (Figure 3). However a more detailed analysis with higher temporal and spatial resolution should be carried out.

1 Proxy poverty index estimated through combination of indicators from two standard surveys in Sudan

2 Relations between lights vs electricity consumption in Sudan (Source: UN Global Pulse)

Image 1 Night Time lights in Sudan 2013 & 2014 (Source: UN Global Pulse)

2 Relations between lights vs electricity consumption in Sudan (Source: UN Global Pulse)

Dashboard with night time lights changes between 2012 and 2013 available at http://dev.unglobalpulse.net/sudan/ .

3 Correlation between illumination and proxy poverty estimates for 2013

Going forward: Mobile Data

In a quest to identify a data source with higher correlation, we are now exploring cell phone usage that could compensate the inherent voids in coverage by electricity and night time lights. With 77% of market penetration of population in Sudan[1] and 73.8% of HHs owning at least 1 mobile phone (MICS, 2014), cell phone usage data provides a substantial opportunity to monitor socio economic behavior that could be a proxy for poverty. Proxies with this extent of coverage will provide measure of poverty levels at various granularity in time and space thereby allowing well informed, evidence based and representative adjustments, monitoring and planning of development interventions.

Immediate next steps entail a deeper analysis of current datasets across additional time period and higher coverage level from states down to localities to confirm initial results. We have proposed to institutionalize this work through a National Implementation (NIM) project based on cell phone usage as big data. UNDP Sudan is working with Ministry of Communications and IT and Central Bureau of Statistics to explore big data potential by analyzing anonymized call detail records (CDR) and its scope to reveal crucial socio economic patterns. A pilot test analysis of sample CDRs against HH survey 2015 is required as a proof of concept, the outcomes of which will be the basis for upscaling activities. The first utilization of the pilot analysis results is aimed within UNDP for the preparation of next UNDAF cycle (2017 – 2020).

Ownership and implementation of Big Data Analytics (BDA) project by national entities, ensures sustainability and its overarching impact within the country, where UNDP’s role will that be of a supporting partner for its set up, capacity building and management. Through the project, we target to form a BDA framework and a national BDA team who would be involved in design of a standardized model out of the pilot analysis. A one day event to share pilot analysis results, big data knowledge and other initiatives is planned to raise interest, increase collaboration and mobilize resources for upscaling efforts. We hope that the outputs of the model and national capacities in place, would meet the needs of regular poverty estimations at low cost and would potentially complement official statistic in GoS’s regular monitoring of SDG # 1 for Sudan leading to a step-change in real-time steering of its development programmes.

[1] http://www.budde.com.au/Research/Sudan-Telecoms-Mobile-and-Broadband-Sta...