Author: Prateek Samantaray
No matter which news article you tune in to, the world is focused on a single rallying point: Recession. Often followed by talks of ‘Stagflation,’ the fears around early recessionary trends have been picking up pace in recent times, owing to bearish investor sentiments, rising costs, the aftermath of a global pandemic, and the threat of multi-nation conflict, among other factors.
Nobel-winning economist Paul Samuelson had once quipped that Wall Street had predicted nine out of the last five recessions. While the markets may be touted as cautiously pessimistic in their approach, it is evident that we are on the cusp of a deep economic downturn, marked by steadily increasing inflation rates and a consequent uptick in interest rates.
Whether this downturn is a long-term phenomenon or a short-term contraction is a debate best left to the economists, but the implications of such downturns on consumers, on credit patterns and lending behavior of financial institutions, and how credit unions can combat this downturn by enhancing their lending strategies with AI, forms the crux of our discussion.
Recession and Lending: A Love-Hate relationship:
In order to better understand the relation between recession and debt, and the consequent effects of recessionary trends on credit access, we will dial our clocks back 15 years to 2007. Most schools of thought point out to two dominos through which the financial crisis built up an economic collapse:
An uptick in household debt in the early 2000s due to subprime lending which, in combination with the collapse of housing prices, severely impacted household spending
Inherent issues in the then financial ecosystem, which, among other factors, included excessive risk-taking and an over-reliance on short-term wholesale funding. This culminated in a credit crunch
The textbook correlation between interest rates and recessionary trends is simple: the Fed will try to spur borrowing and economic growth through a decrease in interest rates. In reality, this doesn’t always translate into more loans available for borrowers. Financial Institutions have especially been cautious in lending during a recession, tightening lending standards and thus reducing loan supply. Credit across all channels, be it business loans or individual loans, showed a sharp decline during the recession, and the recovery in loan growth was even slower in the next 4 years.
What’s more, the on-ground reality is completely different: contrary to expectations the Fed rates have registered a record hike despite increasing fears of recession. And the reason for this is simple: the Fed is trying its best to contain inflation by inducing a slowdown in prices, in effect, making borrowing expensive. By forcing slower growth, the Fed hopes that supply will catch up with demand.
That said, consumer and business sentiment towards lending has been unfaltering and robust. On an annualized basis, borrowing jumped 10.1% in April 2022. Total credit increased by $38.1 bn from March 2022, of which revolving credit outstanding, which includes credit cards, increased $17.8 billion. Non-revolving credit, including auto and educational loans, saw an uptick by $20.3 billion. In May, consumer credit increased at a seasonally adjusted annual rate of 5.9 percent. Revolving credit increased at an annual rate of 8.1 percent, while non revolving credit increased at an annual rate of 5.2 percent. Despite the shockwaves from the Ukraine-Russia crisis, and impending recession fears, the shift of the economy post pandemic disruptions has seen a rebuilding of supply chains and inventories, which in effect have driven loan demand over the past 2 quarters. While the aggressive interest rate hikes may cool off loan growth, it is upto the lenders, and especially credit unions, which have been somewhat countercyclical and increased lending during downturns, to capitalize on consumer sentiment and responsibly lend more, albeit to the right borrower.
The numbers don’t lie: Credit unions have seen lesser delinquency and charge-off rates during downturns, and also been observed to lend more. During the pandemic and recession, credit union asset quality actually showed an improvement, with net delinquency rates and charge off rates at 0.54% and 0.47% in September 2020, as compared to 0.70% and 0.56% in the year prior. While the reasons for the same may be varied, a key factor is that credit unions were built as member first, non-profit institutions, that lend with the mission of serving their members. As such, they don’t have an end goal of profit maximization. Credit Unions go the extra mile in identifying vulnerable sections of their member community and support them during downturns.
That said, in the backdrop of the current economic landscape, it becomes all the more important for credit unions to better identify these vulnerable sections of their member community that, while being creditworthy, are being overlooked because of unavailability of enough data about them, or because of over reliance on a scoring metric to gauge their quality.
Figure: Small Business Loans across 3 periods, Pre Recession (2004-07), Recession(2008-09), and Post Recession Recovery (2010-17)
For credit unions, the lending trends have been somewhat contradictory and countercyclical. During the great Recession, credit unions faced much lesser of a brunt than banks, and ironically, while banks were tightening their lending standards and effectively cutting off easy access to credit, credit unions were increasing their member base, and lending more to support local businesses and families. To that effect, the time is ripe for credit unions to plan responsible disbursement of credit, by leveraging intelligent underwriting practices that employ data to predict borrower behavior.
Data, AI, and Oil for the new combustion engine:
‘Data is the new oil’: A phrase originally used first in 2006 by British mathematician Clive Humbley, becomes more and more relevant as we delve into a time where billion dollar companies stand on a foundation of mined and processed data. In 2011, Gartner SVP Peter Sondergaard took the adage one step further:
“Data is the oil of the 21st century, and Analytics and AI are the combustion engines”.
With businesses dealing with reams of data, credit unions find themselves in a unique situation. They have potentially a goldmine of data to unveil and use to their advantage. The problem? Most credit unions don’t believe that there are tools or capabilities to synthesize the data into information that matters. Regulators spend days, months, and years trying to make sense of a myriad of information on financial risks, often without much success. More often than not, the myriads of data is too much to handle, and relying on simple analytics would only serve to overheat the proverbial combustion engine.
Enter AI.
AI has been in the mix for over 6 decades now. But the utility of AI based models have been limited, and financial institutions, which consider themselves fortresses of closely guarded data, in general abhor the concept of an intelligent system that builds its own models which are often difficult to explain, and in general, carry with them concerns of data privacy.
But the times have changed.
Modern artificial intelligence empowered tools go beyond recommending credit decisions: they deploy flexible credit decisioning engines that can draw data from alternate sources, test and learn in real time, and come up with constantly evolving models that take both current and historical data into account. AI-empowered lending technologies, albeit still in their evolving stages, have become advanced enough to simulate what-if situations by themselves, warn FIs with regards to potential distress and financial deterioration, and predict customer credit behavior.
The use of AI in credit decisioning can be a game changer for credit unions, especially in the light of recent recessionary trends. The long standing problem of adverse selection due to the lack of credit information of individuals has hurt both lenders and borrowers. While lenders are hurt by an increasing load of NPAs, creditworthy borrowers are burgeoned with the load of additional interest rates. By accessing alternate data that goes beyond bureau information, AI empowered lending technologies tend to give a clearer picture of borrowers without any bias or discrimination, helping identify credit worthy, low risk borrowers whose applications would othrwise have been declined, and helping these borrowers get loans at the right interest rate.
In a time when conventional data’s reliability has been put into question, alternative data will go a long way in establishing a low risk lending ecosystem for lenders, and in conjunction with machine learning models, can help control and mitigate risk during a slowdown. With alternative data, a bank can look at hitherto untouched sources of rich data, ranging from loan history, employment and stability details, cash flow, etc. The end result: A bigger, complete picture of the borrower’s credit health.
The long standing apprehension regarding explainability of AI models has also been resolved to a large extent, with modern AI technologies simplifying their models and having a separate explainability module for credit unions to make sense of AI enabled decisions.
So what does it translate to?
Some of the biggest credit unions in the USA which have taken the leap of faith and implemented AI in their credit decisioning process have reported astounding numbers: Upto a 40% increase in approvals, and upto a 15% reduction in defaults. There are notable examples of Credit Unions like GESA Credit Union, which adopted AI during the height of the recent recession, and saw an increase in approvals by 20%, an astounding number considering the lending landscape at the time. In recessionary situations, these differences might translate to an FI staying afloat or sinking.
As the oil and combustion engine metaphor becomes more and more relevant, it is worthwhile to reminisce that the first companies which started using the combustion engine were the ones who prospered through difficult times, while the ones who didn’t, got left behind. Like oil, the key here is to use data and AI in a non discriminatory way, leveraging AI’s ability to go beyond conventional data and process and analyze data that is beyond the capabilities of multimember analyst teams. As such, AI becomes indispensable for modern FIs in trying times, especially to make sense of the gray, at a time when credit decisions are limited to the black or the white.
Comments