Emerging Gold Standard of Underwriting — Transactional data

Rating agencies have been too charitable in rewarding ratings

Emerging Gold Standard of Underwriting — Transactional data

Credit Rating Agencies have put asset managers in a fix

After IL&FS, even DHFL has defaulted. Both were one of the biggest lenders at a point of time. None of the credit rating agencies brought down their rating to D (D is for default) until there was no other choice. In fact, neither IL&FS nor DHFL was never rated negative ever before. Lending activities had stopped months ago. The rating agencies were clearly not even signalling the market, although the fund managers knew that these two lenders were bleeding. They also failed to report impending default of Reliance Capital and Zee Group.

Even though DHFL has cleared its overdue payments within a week of default, now its rating cannot change for 90 days, that’s the cooling off period set by SEBI. Why did DHFL wait to default until credit rating agencies raised the red flag? Certainly, there is a huge information asymmetry, hence even credit rating agencies are defaulting. This is a good sign for us. Who doesn’t want to know the exact financial health of merchants one is dealing with. Credit rating agencies will never be accurate in their ratings because they get outdated data. A credit rating agency relies on an expensive time-taking and variety of information sources, including published annual reports. An audit process is designed to detect fraud or misrepresentation of information, whereas the credit rating process is not.

The equity shares of rating agencies such as ICRA, Care Ratings, and CRISIL are in the doldrums in recent times. Analysts are blaming this on the companies’ failure to report adequately and on time, on the NBFC crisis. Since 21st September, 2018, the shares of CRISIL, ICRA, and Care Ratings have eroded by 15.38%, 18.70%, and 22.51%, respectively. — ET

Until mid of May 2019, CARE Ratings had not downgraded DHFL from ‘A’ to ‘BBB-‘. Please note that CARE A signifies “low” credit risk, while CARE BBB- signifies “moderate” credit risk. There is a growing perception in India that credit rating agencies have been too charitable in rewarding ratings.

  • The number of companies that enjoy AAA status in India is very high compared to other nations. India has 70-odd companies that are rated highest quality
  • Only two companies in the US enjoy this distinction. No company in Germany and UK enjoys AAA rating. Among emerging countries, China has only 14 AAA-rated entities
Photo by Kristopher Roller on Unsplash

Improved underwriting models from transaction level data

Market is screaming for operational efficiency without compromising on speed of meeting demand for rapid access to capital. In latter half of 2018, we began to develop an algorithm to read transactional level bank data to create better underwriting models and improve loan affordability with a vision to empower the lenders (MFs, insurance, securities, etc.) to get accurate and up-to-date insights on credit risk in the most affordable and easy-to-use manner.

At Inkredo, we are testing models that provide insights into customer affordability based on transactional level data in completely automated fashion. This will eventually help lenders make decision with higher confidence than ever before because of greater validation of income and expenditure data.

How do we do it?

Categorisation of transactions: Perhaps the key challenge for using transactional banking data to make underwriting decisions is the ability to categorise spend as discretionary or essential. This is at the heart of deciding whether a working capital is affordable. For example, a spend at Big Bazaar could be for their value range food, premium lager or a second TV; it is not an easy challenge.

Integrating categorised transactions into models: More data means more questions. e.g. It is difficult to consider level of gambling without transactional level data, a Rs.10 lottery ticket may be pardonable, but a Rs.2000 gamble at a casino? This is how we enable data-driven decisions around responsible underwriting.

At an organisation level, we do forensics around rotation of money. e.g. An organisation may rotate Rs.10,000 for ten times in a month between two accounts and thus creating a false impression of transactions worth Rs.1,00,000!

What the likes of ILFS, DHFL and Zee did is not exceptional. It is a rampant practice in India. A business creates a large network of entities. It then borrows using one balance sheet and diverts the funds to other entities linked to it. It then recklessly lends to projects, businesses, and ideas for personal favours.

What is the cost incurred in securing transactional level data?

Lenders have an option to pull bank statements by applying for NBFC-AA (NBFC Account Aggregator) license. This licensing fee is close to couple of crores, and let’s not talk about months of time taken to receive a nod from RBI. This is in turn increases the burden of recovering cost from the borrower.

Whilst customers can be apprehensive about providing hard copies of bank statements, sharing access to bank data online can be even more alarming. For years, Indians were told not to divulge their banking details by banks and government. Given the time and cost of obtaining a license, overcoming the perception barrier of borrowers is even more challenging. The cost of education might be in multiples of licensing fee.

On the other hand, it is much easier and highly cost-effective to obtain a non-machine-readable electronic file authenticated by banks (pdf statements). There are solutions in the market that help you get insights from it without compromising on data leakage.

Concluding Remarks

We are aware that being an early mover to introduce transactional-level data analysis might be a disadvantage given the cost of development and lack of awareness in the target market. It takes a lot of time and resources to build such an engine. However, we believe lenders who embrace transactional data and understand how to operationalise up-to-date information in a digital form will be the ones that prosper.

Additionally, short-term loans do not cover the cost of manual underwriting profitably and customers always have a number of borrowing options if their application is not processed in time. It is a no-brainer that digital transactions increased the speed of money exchanging hands, similarly digital methods of lending practices are significantly improving turnaround times. The emerging gold standard of lending is affording transactional level data analysis.

The sector is definitely going through a crisis but such events change the structure of the industry. Transactional-level data is a baby step in that direction because there is nothing more fundamental and real-time record of financial activities than bank statements.

Want to get a hint of what does transactional level data analysis looks like? Try Inkredo.

Thanks to Samkit Jain for suggestions.

This post was originally published on Medium by Kumar Tanmay.