Financial inclusion is critical in emerging economies – driving economic opportunity, reductions in poverty, and increases in financial literacy across generations. Traditional credit scoring – with a small stream of data coming from set sources to establish identity and prevent fraud, establish current debt load, and detail historical credit performance – excludes large portions of unbanked people. Microfinance has reached many people in emerging economies, but is labour intensive and can be expensive.
New models of credit scoring for the unbanked and those who are not yet included in the formal economy include using nontraditional data to evaluate and score risk. Those who are working in the informal economy and are low-income are a high risk group for credit and lending, especially when they have no history or savings. But collecting and using nontraditional data to score credit risk may provide another way to offer credit, outside of microfinance and traditional financial markets.
Utilities and telecommunications companies collect a great deal of data which can be used to establish identity and ability and willingness to pay. While some nontraditional data is covered by regulatory restrictions, much is available to be collected–but once collected, how can this large disparate data be aggregated and understood? Many companies are seeking to monetise their data, by selling the data they collect for other purposes. Does this monetisation of data affect its validity for credit scoring?
Big data – described as large, diverse data sets – is being analyzed by algorithms written for AIs, or neural networks. The algorithms are specifically written to look for patterns and make predictions in large collections of disparate data. The sensitivity and specificity of these patterns and predictions is growing, but several challenges remain.
Privacy is both legal and cultural, and there are vast differences across the world in what data is acceptable and legal to collect and use. More significant to the widespread acceptance of using nontraditional data to devise credit scores is the concern over bias in the data sets. While the development of algorithms and neural networks is growing rapidly, the issues of cross-checking, of quality assurance for the new technology, has still not been developed. With data being collected and sold by vendors whose primary service is something else, such as utilities, the questions of accuracy, inclusion, and bias remain untested and unproven.
Large populations in emerging economies are currently being credit-scored with these nontraditional methods. The validity of these new data sources, and the algorithms being used to look for patterns and make predictions, will not emerge for some time. The benefits of the microfinance model, while labour-intensive, include education, social support, and widespread community acceptance. These benefits and the effectiveness of this model to move underserved communities toward greater financial inclusion cannot be replicated by data-intensive credit scoring tools. Together, however, these two models may provide effective in moving underserved people in emerging economies toward greater financial inclusion.
Find out more about Fern Software solutions, click here to request for our brochures for free!