Africa, despite lagging behind in infrastructural development, is actually data rich. In 2010, of the 5 billion mobile phones in the world, 80% were in developing countries. East Africa is at the forefront of new technologies such as mobile payments, which are ubiquitous even for those in the lower socio-economic groups, with over 38m of Tanzanians (in a population of 49 million) having a mobile wallet. For a large majority of these 38m people, this is the only financial service that they have access to, banks and other financial institutions being a luxury even then 11m Tanzanians have no access to financial services at all.
Going the last mile on achieving universal financial access for all, including those living on less than $2 a day and those in rural areas will be challenging. This is made near impossible, by the relative lack of demographic data presenting a serious impediment to firms to make investment decisions. Broadly, in developing nations absence of the market intelligence and data infrastructures that we take for granted in developed economies: Firms simply lack the essential insights they require in consumer and household behavioural patterns, financial habits, mobility, and indeed even basic demographics. This knowledge gap deters invaluable inward investment that would help alleviate poverty and promote economic growth; it is a key stumbling block to financial service investment.
As such, the N-Lab team [link to team page], through the Gates Foundation Grand Challenges [link to previous blog] is working on the Kupata Project (Swahili for Access) to fuse open and commercial transaction data streams. These streams could be Call Detail Records or aerial imagery and other forms of data. Through researching and applying new machine learning techniques (such as Dynamic Topic Modelling [link to paper]) by linking these streams together, identifying relationships between features, we aim to provide comprehensive spatio-temporal segmentations. These segmentations would describe financial behaviours, mobility and communication patterns – and the market intelligence that would enable the provision of the next generation of financial services, moving towards full coverage.
Generating these layers will be an intensive process. Currently, map layers showing where financial services are being mapped by a volunteer community (ramanihuria.org/map) – data here: http://overpass-turbo.eu/s/iVJ, or generated through satellite images which can demarcate landuse, showing informality and slums. These are very expensive methods, but necessary in a data poor environment. But, with new data streams (such as CDR) can we research new predictive models of financial service flows in the region where traditional data does not exist? If so, this would be a game changer, allowing areas to be mapped through the digital footprint of mobile phones.
This will reveal not only previously unknown insight into the growth of accessibility to financial services, but deeper understanding into how these regions and their geo-demographics change over time, informing how to better service population growth and urban planning. Ultimately, this work supports the attainment of the United Nations 2030 Sustainable Development Goals, specifically Goal 1: Eliminate Poverty and Goal 8 Sustainable Economic Growth and Employment. This is part of a global agenda aimed at eliminating global poverty by 2030. Now, the opportunity arises to use data for the social good to meet these global development goals.