Explo(i)r(t)ing Alternative Data

Data is the new oil. Expansion in data volume, velocity, and variability, as well as new technology, has made data more accessible and easy to analyse. This necessitates a larger acceptance of not only traditional data sources but also alternative data. Experts are turning to alternative data to provide insights into the behaviour of individuals and markets in various industries, particularly in developing nations where traditional data is hard to obtain, especially in the financial services sector. However, before delving into the story of why alternative data is so valuable, let us establish a more concrete foundation by outlining a definition.

What is alternative data?

Alternative data is generally understood to refer to non-traditional data that supplements established data sources. The definition of alternative data varies depending on the sector and context. Today's alternative data could be tomorrow's standard.?

The utilisation of internet search terms, such as Google Trends data, is a well-known example. Statistician Nate Silver predicted the 2016 American presidential election with astounding precision, using internet search data to supplement traditional polling data. This result, however, could not be replicated in 2020, emphasising the significance of not relying too heavily on this type of data. Social media posts are another popular kind of alternative data, and they are increasingly being used in financial markets with natural language processing (NLP) tools to try to infer underlying market sentiment.?

The Economics Department at Stellenbosch University is currently conducting a study using machine learning tools to assess the impact of US presidential speeches on financial market activity. These publicly available texts exhibit some predictive power, which could potentially provide an edge in stock market forecasting. Alternative data is especially valuable in developing economies when data integrity is difficult to verify. Night-time satellite imagery (dubbed "night lights" in the literature) has been utilised in economic studies to provide a better picture of economic progress in countries where official data cannot be trusted.

Alternative Data in Fintech

Having a broad variety of data to support traditional data is becoming increasingly critical in many circumstances. The examples provided do not represent a nearly comprehensive list of alternative data sources. In theory, the collective set of alternative data available is only limited by the constraints of imagination, technology, and legality.

Fintech platforms have begun to aggressively investigate alternative data sources, with credit provision being a popular application. China and the United States are at the forefront of these lending markets. However, Southeast Asia, Latin America, and Africa are experiencing some of the fastest growth in big tech credit. For many financial service organisations, offering credit to higher-risk individuals with short credit histories (the so-called "credit invisibles") in order to increase financial inclusion frequently requires looking beyond traditional indicators.?

Financial inclusion

In a recent article on fintech lending, led by Harvard-based researcher Marco Di Maggio, it was discovered that fintech platforms that use alternative data to determine borrowers' creditworthiness result in higher financial inclusion. Borrowers with low credit scores or short credit histories had a 70% higher chance of being rejected and earning higher interest rates in this study than those who followed established creditworthiness models. Interestingly, providing loans to thin-file borrowers using alternative data criteria resulted in higher profitability for lenders and better economic outcomes for borrowers. This accords with the notion that credit record requirements can work as a barrier to financial opportunities for the credit invisibles.

In the context of lending activities, alternative data refers to information that credit reporting agencies do not typically gather. Analysing spending patterns via bank statements could help determine whether consumers are willing and able to repay their debts. From these statements, one can gather whether frivolous purchases are made or make determinations on whether a potential client's liquidity woes are only temporary. One could also consider payment regularity with respect to subscription services and other bills. Credit providers might frown upon missed Netflix payments or failure to pay electricity and cellphone bills.

Behavioural data science

Understanding the mindset underlying spending and consumer activity has driven large banks such as Standard Bank to invest resources into developing experts in behavioural economics, a subdiscipline of the field that incorporates elements of psychology into economics to better understand the strategic behaviour of economic agents. Alternative data plays an important part in this development.?

Another example of a company that is bringing theories of psychology into their modelling approach is ConfirmU. ConfirmU's psychometric game uses alternative methods of credit risk scoring to open up loan and credit opportunities to individuals with no previous credit history, e.g. the impoverished, unemployed, and recent immigrants or graduates. This will allow them access to potentially life-changing financial opportunities. All this is done while minimising risk to credit providers by identifying financially conscientious individuals who are most likely to make their repayments.?

A case study detailing the nature of this collaboration between Matogen Applied Insights and ConfirmU will be posted in the coming weeks.