Expecting the unexpected: Modelling relationships between economic factors and expected credit losses during COVID-19

24 June 2020

When we estimate expected credit losses (ECLs), we form expectations about possible forward-looking macroeconomic scenarios, assign probabilities to each, and then use historical relationships between the macroeconomic variables and our loss parameters to estimate the losses that would arise under each scenario. Sometimes those past relationships are better predictors than others. Take unemployment, for instance. Normally, when people become unemployed it means that they’re out of work without a job to return to. It takes time to find a new one and until they do, cash isn’t flowing in. 

From an ECL perspective, that means that the ‘probability of default’ (PD, being the risk that the loan cannot be repaid) increases. For argument's sake, let’s say that on average it takes a year to find a new job and that as a result, PD typically doubles. Why might that be different during the coronavirus (COVID-19) pandemic? 

For starters, there are new government programs and reliefs that may increase cash flows available to support loan repayment. Secondly, unlike usual circumstances, the employees and businesses themselves may not have failed. In other words, there may be jobs available to return to once lockdown eases. If so, that means less time in limbo and less time without normal cash inflows. With any luck, that should mean less risk (compared to normal unemployment) that the loan will not be repaid. 

Of course, it’s also possible that it could turn out to be worse and that can’t be overlooked either. After all, we don’t know what effects government relief programs will ultimately have, how long lockdown and social distancing will continue, whether there’ll be another wave, or when a vaccine might be developed. Each could have a very significant impact on how and when loan losses will transpire. 

Since transition to IFRS 9, our application of the ‘multiple scenarios’ requirement in estimating ECLs has focused predominantly on forecasting and modelling various economic variables and what we think those variables might be in the future. That worked well while conditions were benign - but now the underlying relationships between those economic variables and the variables we need to estimate ECLs might change too. In other words, our consideration of uncertainties may need to be extended to include reviewing both the economic variables and their relationships with loan losses. 

All of this has meant that over the past few months, banks’ models are often producing unusual results when using recent macroeconomic data. This could be the manifestation of unusual economic conditions, limitations in the models themselves, or both. Typically, models go through calibration during development and validation, including testing them under a range of conditions. That’s not necessarily to say that they’re certain not to work in conditions outside of that range, but they weren’t developed with them in mind and so there’s a chance that they may not. How do you discern between the effects of unusual economic conditions and outliers in evaluating model outputs? We’d start by asking: 

  • What’s causing us to believe there are outliers?
  • Have known limitations been identified during model development and validation?
  • How can we measure it and what alternatives exist (e.g. scalar, overlay or other)? 
  • What additional validation and internal control procedures are necessary?

To assess this, some banks start by comparing results to those generated by stress-testing results, Basel capital models, internal budgets, external sources of projected default rates, historical loss experience (e.g. during the Global Financial Crisis) and other available reference points. Where limitations are confirmed, models might be applied within the calibration range and overlays used to address identified shortcomings beyond them. Depending on the portfolio, individual loan reviews or other approaches may be applied as well. 

Whichever approach is taken, keep in mind that robustness and consideration of all possibilities is key and when it comes to data points and triangulating adjustments, more is better. 

Our guest blogger is Chris Wood, Banking Partner and IFRS 9 specialist, connect with him on LinkedIn here.  


Verify your Comment

Previewing your Comment

This is only a preview. Your comment has not yet been posted.

Your comment could not be posted. Error type:
Your comment has been saved. Comments are moderated and will not appear until approved by the author. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.


Post a comment

Comments are moderated, and will not appear until the author has approved them.