Predicting
Financial Access

Access to digital financial services is fundamental to enabling those living in poverty around the world to become more economically stable, prosperous, and resilient. One of the Bill & Melinda Gates Foundation’s stated goals is “to help people in the world’s poorest regions improve their lives and build sustainable futures by connecting them with digitally-based financial tools and services.”

Connecting the poor to these financial services is a complex undertaking. Where can we look for answers? Data. Data can unlock barriers and expand access to life changing services by uncovering the existing landscape of financial services as well as critical gaps and needs.

However, the right data doesn’t always exist or isn’t reliable. Why? Current data collection techniques and methodologies are often expensive, error prone, and slow, or provide data snapshots that are low resolution or become quickly outdated.

Spatial Development International’s data science team is exploring what could be a breakthrough for efficiently collecting financial access data. Using statistical and spatial analytical means, we can use probabilistic models to predict the location of financial touch points that were previously not known.

Fortunately, some useful data already exists – like this landscape of financial services touch points that were manually geolocated through data collection surveys by enumerators on the ground. Just looking at this map, one can see where financial services exist – and potentially don’t exist.

For this work, the SpatialDev team can make use of a rich set of existing spatial and tabular data to build a probabilistic prediction model. The results of the model can then direct data collection efforts on the ground and reduce the amount of time and number of surveyors required. Think of it as an intelligently targeted collection – versus a blanket – effort.

The model can consume millions of data points that help us visualize components of the financial services access challenge. A dataset showing urban-rural populations helps us determine the where we need to reach people.

We can also look at mobile network coverage to get a clear picture of where people have access to the infrastructure needed to transfer financial services and other data.

Another set of critical data we can visualize is the location of different financial services such ATMs, banks, microfinance institutions (MFIs), and mobile money agents. We can also set a buffer from those locations to understand how easily people can access these services.

Geosocial media posts can also unlock a picture of social media hotspots, where people are using Twitter and mentioning an activity involving a financial transaction – like shopping, for example. This is important, as it can be used in the model to inform both where unmapped financial access points likely exist as well as good candidates for expanding financial access if the financial services aren’t yet available.

All of these data are brought together along with country-specific knowledge to train a financial touch point prediction model. The results will inform more efficient data collection efforts across sub-Saharan Africa. The larger the variety of data the model consumes from additional country surveys, the better trained it becomes, leading to even greater efficiencies. The model can then output a heat map of its own, indicating where enumerators and citizen surveyors can expect to find financial access points and map them.

The FSP mobile application will be leveraged for on-the-ground efforts to validate the model and collect additional data. The application allows users to view the spatial estimation model, add and edit financial touch points, and review other data previously collected in the field. The spatial estimation model will be made available for others to use independently in their collection efforts. It will also be accessible to financial access providers to better understand market demand and potential for service expansion.

With an intelligent prediction of where financial services exist or are needed, providers will ultimately be able to avoid trial and error for rolling out new services and focus in on where they can provide the most benefit and reap the most reward for their business. Through this work, we can more efficiently collect the data needed to understand the financial services landscape, and make an impact on efforts to lift the world’s poor out of poverty.