As the region battles to get the Covid-19 Delta variant under control, the forecasting models that lenders rely on to predict borrower behaviour are being upended again. The data that these models run on has become unreliable due to widespread payment moratoriums and other financial relief programmes across markets.
Governments around the world suspended loan repayments at the start of the coronavirus pandemic to pre-empt potential defaults from affected borrowers. In Southeast Asia, the moratoriums have been extended time and again, leading to an increase in informational asymmetries.
While banks typically depend on internal data and bureau data, which are often supplemented by alternative data, bureau data traditionally provides the biggest indicators of risk, especially in this region. With the moratoriums, bureau data no longer reflects the true credit risk of the customers. And without updated payment data, banks are not able to tell which customer profile is riskier than the other.
Having to make sense of “stale” data was the single biggest challenge that came up repeatedly during our discussion at the Experian ASEAN Strategic Advisory Board on risk calibration in the wake of the pandemic. As one banking leader put it, “The impact of the debt moratoriums on monitoring and underwriting is like flying a plane in the dark through a storm. How do you measure it?”
From judgement calls to simulations
To cope with the lack of useable data and facing an unprecedented environment, banks have been relying on “judgement calls” based on past experience and domain knowledge of their risk managers to strengthen monitoring and increase their ability to predict customer risk exposure. Some added policy overlays to existing score cards in a bid to improve overall quality of data.
Stress testing has been ongoing to measure financial health, as well as running simulations of various scenarios or outcomes based on hypothetical end dates of the various moratoriums. Some banks managed to automate internal systems to identify customers who opted for the debt relief, so that these customers could be managed as a subset instead of being included in the overall estimation.
In countries where loan relief was not applied uniformly across the board, banks faced the added complexity of differentiating between the genuine cases and those that were trying to game the system. For example, a consumer could have informed one bank that he was facing financial distress and hence eligible for a moratorium, and then go on to another financial institution to ask for a new loan.
This is why better harmonisation of regulation and credit pricing has a key role to play in this scenario. It is a delicate balance with governments expected to provide capital flexibility to banks, but on the auditing side of things, banks will have to manage additional provision and policy overlays, and even start recalibrating and potentially downgrading relevant customer segments based on key portfolio indicators.
An exit strategy
When data from an entire year becomes an outlier with no realisable near-term value, the way ahead lies in finding solutions to recalibrate risk behaviour and forecast models to deal with probabilistic and deterministic risks in lending and collections.
Automation in collections has emerged as a central theme for ongoing concerns around customer management with acknowledgement that more can be done to streamline the process. Rueing the manual processes involved in collections for a segment that saw a slight jump in credit losses after some of the loan relief ended, a common sentiment was echoed that “Automation is in place but it is not yet end to end. This was an expensive lesson, and it reminds us that we need to be better prepared to deal with the next pandemic.”
As countries embark on their journey towards Covid-19 recovery, albeit in a start-stop fashion, banks need to rethink their approach to exit strategies. Given the swiftly evolving impact of the crisis, scenario models that worked in the past are not likely to be relevant now. And organisations can no longer afford to waste time on manual and repetitive tasks in this state of constant change.
It is critical that banks map out an analytics and technology strategy that would enable them to act quickly and decisively – based on changing scenarios. For example, by harnessing artificial intelligence and machine learning, they can build, deploy and retrain high-quality analytical models that run on large and varied data sets – including additional new sources of data – faster than before. With the ability to make incremental changes across the entire model lifecycle on the fly, these banks are poised to emerge stronger and more nimble from the crisis.