Conventional banking has typically focused on the 5Cs of “Credit” – character (identity), collateral (security), capacity (to repay), capital (savings, investments, or other assets), and conditions (usage of loan) – while making lending decisionscom.pk/blog/mapping-creditworthiness-unbanked-population/">. This approach, however, does not work for the unbanked population. Unbanked individuals and households, who are either in poverty or earn low incomes, do not have prior credit histories. They also typically lack bankable assets. Banks usually prefer ‘safe collateral’, such as a house in urban area, a large farmland in rural area, or hard cash in a bank account. As a result, unbanked folks cannot satisfy the requirements of risk-averse formal lenders, who see high risk and low return at bottom of the pyramid.
It makes little sense to expect conventional-banking model of collateral-backed loans to meet borrowing needs of low-income households in Pakistan. The last two decades have seen the rise of microfinance in informal markets, where small loans are offered through personal guarantees or community pledges. However, microfinance, which also follows conventional banking models to extend small credit at somewhat high rates, hasn’t been able to penetrate more than a third of the potential market.
Credit scoring models can help in developing a conducive and cost-efficient lending environment. This notion is particularly relevant for markets that have limited credit bureau coverage and customers who are traditionally excluded from formal credit. This can help financial institutions to asses the credit worthiness of a potential costumer, which shall be backed by statistical data and not on subjective judgment. This will also result in innovative products with different pricing slabs for costumers falling under different score bands.
Need for alternative credit-scoring mechanisms
In Pakistan, it appears that conventional banks are meeting the needs of the upper- and upper-middle incomes groups; microfinance is catering to the middle; while a large bottom is exposed to the informal market. In recent years, several new approaches have emerged in other developing countries, through which creditworthiness of the unbanked population has been mapped by leveraging technology. This has given rise to “digital credit”.
a) Big data techniques
Big data techniques have been used on large sets of available transaction data to create risk profiles of unbanked individuals, families, and businesses. For instance, a Fintech firm interested in offering digital credit can identify, secure and convert large payments data – specifically relating to mobile phone recharge, grocery purchases, and utility bills – to simulate credit history of the target market. These big data methods have been used by several dozen mobile-money deployments worldwide.
Big data can help service providers make lending decisions for the hitherto unbanked, by looking at the future instead of the past. Data mining can also help service providers better engage with their clients once a relationship is established. For instance, correlating usage patterns with other publicly-available databases, service providers can send their clients useful and targeted alerts (weather patterns, crop conditions, market prices, etc.), besides offering meaningful saving and insurance products. A Pakistani domiciled fintech named “CreditFix” ventured into such an idea. CreditFix will be using smartphone consumption pattern and publicly available data points to target low-income groups to help them start and expand their businesses. Loan processing times are significantly shorter than conventional banks and the customers will get risk-based pricing. Better risk profiles would result in cheaper interest rates.
b) Social media and behavioral data
Some companies have demonstrated that non-traditional, behavioral data can also be a predictor for creditworthiness. Lenddo – a Fintech company which works with over 50 financial and utility companies in 20 countries – has developed a model through which applicants’ online behavior and social data are mined to score their credit-worthiness.
Lenddo specifically looks at patterns within users’ social media profiles, email accounts, and wireless communications. It doesn’t look at entire histories, out of privacy concerns. Instead, the computer programs search for keywords related to debt obligations, employment conditions, purchasing behavior, and socialization. Another such example is China’s Rapid Finance (CRF) which uses search engine data from Baidu (a leading Chinese language internet search provider) to reveal correlation between search engine phrases and borrowing needs. For example, small business owners that search keywords “photography” or “hiking” are likely to need an unsecured small loan. However, small business owners searching online for “lottery tickets” are unlikely to be a safe bet for lending platforms.
c) Online purchases
This credit-scoring model specifically analyses users’ online purchases through algorithms to predict creditworthiness. This model was pioneered by Alibaba in 2015. The credit-score, called as Ali-loan, is based on users’ online shopping history – basically buy and sell transactions – at the B2B portal of Alibaba.com. The credit score is then sold to financial institutions in China.
Relevance for Pakistan
Of the three approaches listed above, perhaps the most relevant for Pakistan is using ‘big data’ on ‘payment histories’. This historical data specifically concerns cellular recharge and utility bill payments. In the cellular segment, the nearly 75 percent tele density is a great source of payment data. Besides, gas and electricity bills are also a useful source of payment history.
The branchless banking transaction data is another viable source to build digital credit models. In the last decade, the branchless banking segment has come a long way, in terms of footprint (number of agents and mobile accounts) and transaction mix (over-the-counter (OTC) transactions and mobile payments).The customer-oriented OTC and m-wallet transactions stood at 3.6 million transactions per day in Dec 2018 – up from 0.2 million daily transactions in Dec 2011.
BB service providers maintain that a significant portion of transactions, taking place through their channels are conducted by unbanked and under-banked individuals. In that respect, this channel provides a better targeting for digital credit scoring. Yet, no digital credit deployment currently exists in Pakistan. However, the State Bank of Pakistan had allowed Easypaisa (Telenor Microfinance Bank) to conduct a pilot/feasibility for digital credit in Pakistan back in 2016. FINCA, another microfinance bank has also launched a pilot of offering digital loans.
Challenges in digital credit deployment
While one awaits the pilot’s results, it should be kept in mind that deployment of digital credit comes with its own set of challenges.
a) Loss of focus
It can be easy to lose sight of the mission of improving access to basic financial services. In 2016, an analysis by the World Bank’s Consultative Group to Assist the Poor (CGAP) analyzed ten global deployments of digital credit and identified several common mistakes across the deployments. CGAP found out that
- The deployment’s inadvertent focus was more on “credit-scoring” and less on “collections”.
- Credit-scoring didn’t use a broad set of data to expand the scope of applicants, resulting in poor targeting.
- Loan application process was complex, thus reducing uptake.
- And digital credit was unviable due to pre-existing issues with mobile money, such as high money transfer fees and a limited digital payment ecosystem.
b) Limited oversight over loan usage
While the digital medium has its benefits, its “remoteness” also creates issues for effective oversight. For instance, a 2018 CGAP survey conducted in Tanzania showed that digital credit was being used mostly for daily needs, instead of using it for emergencies or for business needs. It also found that the default rate was over 30 percent, while more than half of borrowers had made late repayments. International Monetary Fund (IMF), in its study “Financial Inclusion: Can It Meet Multiple Macroeconomic Goals?” resolved that economic and financial stability risks increase when access to bank credit is expanded without proper supervision. The systemic risks involved with the extension of digital credit can be mitigated through sufficient supervision and the intensity of the supervision has to be “proportionate” to the risks involved.
c) Consumer consent and data privacy
In digital credit, keeping the users’ trust goes beyond protecting their funds/deposits. As this medium uses payments data, it is important to uphold the highest standards for data privacy and to keep customer data secure and confidential. A recent global survey on regulators’ view on digital credit identified the following issues as the most important for regulators: adequate disclosures, lack of deceptive marketing, pricing fairness, and orderly dispute resolution mechanisms.
 The concept of “5 Cs of Credit” has long been taught in training curriculums at Pakistani commercial banks. This risk-assessment concept is also practiced abroad. For more, see “Credit Management at Wells Fargo”(URL: https://www.wellsfargo.com/financial-education/credit-management/five-c/)
 Pakistan Microfinance Network | MicroWatch | Issue 50 (Oct-Dec 2018)
 CGAP | Credit Scoring in Financial Inclusion(URL: https://www.cgap.org/sites/default/files/publications/2019_07_Technical_Guide_CreditScore.pdf)
 CGAP | Policy Brief | The Proliferation of Digital Credit Deployments
 McKinsey & Company | New credit-risk models for the unbanked(URL: https://www.mckinsey.com/business-functions/risk/our-insights/new-credit-risk-models-for-the-unbanked)
 Lenddo | Credit Scoring: The Lenddo Score(URL: https://www.lenddo.com/products.html#creditscore)
 Forbes | How Social Media Could Help The Unbanked Land A Loan(URL: https://www.forbes.com/sites/chynes/2017/04/25/how-data-will-help-drive-universal-financial-access/#17be5a3257e6)
 GPFI |Alternative Data Transforming SME Finance(URL:https://www.gpfi.org/sites/gpfi/files/documents/GPFI%20Report%20Alternative%20Data%20Transforming%20SME%20Finance.pdf)
 Quartz | “Alibaba’s customers can now get a loan based on their online shopping history”(URL: https://qz.com/436889/alibabas-customers-can-now-get-a-loan-based-on-their-online-shopping-history/)
 Pakistan Telecommunications Authority | Telecom Indicators (as of January 2019)(URL: https://www.pta.gov.pk/en/telecom-indicators/2)
 State Bank of Pakistan | Branchless Banking Newsletters | Various Issues(URL: http://www.sbp.org.pk/publications/acd/branchless.htm)
 State Bank of Pakistan | Broadening Access to Financial Services | Annual Report (FY18)(URL: http://www.sbp.org.pk/reports/annual/arFY18/Vol-1/Chapter-4.pdf)
 CGAP | Policy Brief | The Proliferation of Digital Credit Deployments(URL: https://www.cgap.org/sites/default/files/researches/documents/Brief-Proliferation-of-Digital-Credit-Deployments-Mar-2016_1.pdf)
 CGAP | Helping or Hurting? Ten Facts About Digital Credit in Tanzania(URL: https://www.cgap.org/blog/helping-or-hurting-10-facts-about-digital-credit-tanzania)
 IMF| Financial Inclusion: Can It Meet Multiple Macroeconomic Goals?(URL: https://www.imf.org/external/pubs/ft/sdn/2015/sdn1517.pdf)
 Alliance for Financial Inclusion | Digitally-delivered credit – Policy Note | Results From Regulators Survey(URL: https://www.afi-global.org/sites/default/files/publications/guidelinenote-17_cemc_digitally_delivered.pdf)