The Economic Survey 2019-20 proposes the creation of a GSTN-like body which will verify individuals and corporate customers via KYC and Aadhaar KYC, run data analytics on their financial data, generate their credit profile using artificial intelligence — all to eventually assess their eligibility for loans from public sector banks.
The NPA crisis and PSBs: Released yesterday (January 31), the Economic Survey dedicates an entire chapter to the status of public sector banks and the NPA crisis. The Survey suggests that a large proportion of India’s NPA crisis — non-performing assets peaked at Rs 10.3 trillion in March 2018, of which public sector banks accounted for Rs 8.9 trillion — could have been prevented had data and analytics been used in corporate lending. A “robust credit analytics system” would have picked up factors that contributed to the NPA crisis — such as poor audit disclosures by large defaulters.
Data as ‘gold mine for economic growth’: The Survey further cites “the digital infrastructure that generates and stores an abundance of high-quality structured data on all aspects of the economic lives of firms and individuals” as a factor in the growth of digital payments in the last 3-4 years.
Perhaps more important is that the inclusion is backed by state-of-art digital infrastructure that generates and stores an abundance of high quality structured data on the all aspects of the economic lives of firms and individuals. Such data are, of course, the gold mine for economic growth in the 21st century. They offer essentially unlimited and uncharted possibilities, especially for firms and individuals who have been traditionally excluded from the financial system.
It then elaborates the use of technology in PSBs:
As of now, PSBs employ technology mostly for MIS and reporting while most information processing on loans happens manually which causes inefficiency frauds and loan defaults. Information processing includes all activities related with the ex-ante screening of potential borrowers and the ex-post monitoring of their behaviour.
The Survey goes on to suggest that public sector banks are (also) sitting on a treasure trove of data — they have “local market insights and relations” from over 40 years and a far wider reach — and need to make investments “to exploit this data-rich environment”. To prevent an NPA-like crisis, public sector banks need to screen their borrowers better in order to decide whether they should be given loans, and then monitor them better after they’ve given out the loans. PSBs can “fulfill their role of delegated monitors if all the PSBs can pool their data into one entity”, the Survey says. And since the government owns all of them, it can mandate them to share such data. This data will be pooled in the proposed body — the Public Sector Bank Network or PSBN, which will run analytics to assess the quality of people and businesses asking for loans.
How the Public Sector Bank Network will work
The Survey envisions data collection, KYC, and operations involving AI for the PSBN:
Step 1: Customer Verification: Once a customer approaches a bank for a loan, the PSB will transfer the loan information to the PSBN, which will verify the customer via Aadhaar eKYC. The key players involved in this will be “identity verifying agencies” such as UIDAI, Udyog Aadhaar, and Income Tax Department, says the Survey.
Step 2: Data collection from multiple sources: The PSBN will collate data from the various data sources after confirming their identity via KYC, such as:
- Account Aggregator: for seeking consent and accessing customers’ banking data through
- Government sources: business and income data from government sources such as GSTN and IT Returns using APIs or direct file uploads. Data can also be sourced from customers’ bank statements from the public sector banks.
- Credit Bureau Data: Credit score data from bureaus like CIBIL and Equifax
- Alternate data from telcos or mobile handset data
Step 3: Generating a credit profile using algorithms: Analysing all the above data, PSBN will generate a credit profile of the customer after running the AI/ML algorithm built into it. Different underwriting AI/ML models will be built for different customers — individuals, SMEs, and corporations. Although this is not explicitly clear, the Survey mentions that FinTechs will build the algorithms for screening (before loans are given) and monitoring (after loans are given out).
Step 4: Loan eligibility: Based on KYC and underwriting, the PSBN will assess customer eligibility of loans and transfer all information to the concerned PSB, which will then take a decision on the amount and interest rate of the loan.
PSBN will work because it can make underwriting decisions using data that PSBs have been collecting for the past 50 years; public sector banks will be able to process loan applications faster and reduce turn-around-time (TAT), which will help them compete with ‘New Private Banks’ — a term used for commercial banks established in the 1990s and whose operations were automated, such as Axis Bank, ICICI Bank, Yes Bank, etc.
Although it did not suggest it for use in the PSBN, the Survey pointed out that both structured and unstructured data can be used for credit analytics:
- Structured data includes credit information and credit scores based on loan grants and repayments held in the credit registries or credit bureaus
- Unstructured data, which is “richer” is found in text, images, geo-tagged data, social network data, mobile apps, “as well as other shallow or deep digital footprints of current and potential customers”
Leveraging this data requires new data, analytics, and modelling skills. Banks also need to invest in credit recovery infrastructure, and hire people who can work on analytics.
Govt bodies have ‘leveraged data’ to protect collateral: Survey
The Economic Survey separately noted the use of data and technology for tracking monitoring creditors’ collateral, a measure suggested to improve loan recovery. Wilful defaulters have a natural incentive to misrepresent the value of collateral on loan, and may even pledge fictitious collateral, or may pledge one collateral for multiple loans, says the Survey. “Data can come to the rescue of lenders in such cases,” it added, enlisting possible ways in which PSBs can use fintech, with a passing note on privacy:
- Geotagging can help lenders keep track of location of assets given as collateral, and would make it difficult for borrowers to move such assets by stealth. The Rural Development ministry geotags MGNREGA assets, Department of Land Resources geotags watershed projects. “Lenders are better placed to evaluate the market value of assets, the bulk of whose value derives from their location.”
- GPS systems can give lenders more monitoring power, as such systems can alert lenders when assets are being moved out of a property. GPS systems that come with remote-kill functions — which remotely disable all functions if the someone attempts to tamper with the asset — can also help.
- Integrated data on collateral across all lenders in a geography may be particularly useful in curbing double-pledging of collateral, which is when a borrower takes multiple loans (from different lenders) using the same collateral. As long as lenders rely on human control processes and paper-based documentation to verify trades, such double-pledging easily escapes notice.
- SWIFT India — the messaging platform that PNB used to transmit messages in the Nirav Modi case — recently announced a pilot blockchain effort that allows lenders to log invoices and e-bills submitted to them online, allowing other lenders to verify whether a trade they are looking to finance has already been funded or the underlying collateral already pledged.
The Survey notes that these technologies carry the risk of “infringing upon the borrower’s privacy and dignity” and “enforcement of debt obligations should not encroach into the borrower’s private sphere”. It states that “strong and clear policy guidelines are needed on what data may be collected, how, by whom and for how long”.