Schrödinger's COVID revisited: Once in a blue moon?

Since the onset of the COVID-19 pandemic, we have been estimating Lifetime Expected Credit Losses (LECL) for a pair of benchmark portfolios, using reasonably standard time-series and credit modelling assumptions.

As discussed in our recent article Stage Fright, under regulatory guidance firms have typically adopted an approach of "innocent until proven guilty". In-effect credit losses for retail borrowers on moratoria exist but are unobservable until moratoria end and banks find out who has resumed paying. Likewise, widespread state support is likely to have provided sufficient liquidity to both retail and commercial borrowers, to the extent that payment defaults have largely been avoided.

There is little doubt that without moratoria and wage support, the banking sector would have experienced a tsunami of payment arrears and default events. Unlike Erwin Schrödinger's famous experiment, the result and impact would have been observable almost instantly in banks' funding costs, as well as in terms of LECL estimates in banks' IFRS accounts. One could speculate that in terms of customer outcomes the result may have been similar (perhaps with widespread forbearance being offered) but a correlated systemic shock may also have rippled through the real economy, in-effect "changing the result by looking at it".

By using market-implied data, our modelled assumptions have been able to peer "inside the box" and estimate just how much banks would have had to set aside each month. Since 31/03/2020, Lifetime Expected Credit Losses (LECL) have approximately halved in both secured and unsecured lending, across the base case and upside/downside sensitivities. However, in more-recent months, despite House Prices Index (HPI) growth, the Credit Cycle Index (CCI) has stubbornly refused to return to historical equilibrium conditions anywhere near as quickly as after previous peaks. The chart below shows CCI and annual change in natural log of HPI ("dlnhpi") actuals and forecasts:

Over the Halloween blue moon weekend we learned of a second English lockdown (at the time of writing, subject to a vote in parliament), extension of the furlough scheme, and a further six month payment holiday for mortgage borrowers. As a result, arrears or contract breaches may remain unobservable until into 2021.

Continuing the parallel, it is worth remembering that Schrödinger's cat was placed in a sealed box, and the experiment should be concluded before the cat dies of hunger or suffocates. Despite the once-in-a-blue-moon opportunity to replenish the food and air in banks' boxes of loans without looking, it's worth bearing in mind that the IFRS 9 standard is agnostic to whether or not payment patterns are observable. Instead, a broad test of "reasonable and supportable information" is applied. Come year end 2020, the question will, as always, be: What percentage of each loan's contractual cash flows can banks reasonably expect to receive?

Appendix: LECL results at 31/10/2020

The tables below express LECL on a benchmark secured and unsecured loan at various maturities, as an uplift factor with respect to year-end 2019.

The methodology and assumptions are broadly applicable to all exposure classes, including consumer and corporate lending; but potentially excluding financial instructions, where the default correlation is typically higher. We have assumed that the credit risk on a basket of UK consumer firms is a reasonable proxy for the systematic (shared, non-diversifiable) risk factor for any UK lending. One could alternatively use one of the public arrears or write-off data series. However, these are currently temporarily blinded by payment moratoria; and also tend to be non-stationary, hampering robust statistical modelling. For further details of our methodology and assumptions please refer to this article.

Secured      

Unsecured