Even if the expected ultimately happens, the ride there under IFRS 9 could still be bumpy…

19 October, 2020

Mark Randall headshot photo Mark Randall
Director, PwC UK

There are huge uncertainties around the impact of coronavirus (COVID-19) on banks’ loan losses. But the good news is that banks are now using IFRS 9’s ‘expected credit loss’ or ECL provisioning model, designed to avoid recognising loan losses ‘too little, too late’. So – putting aside the many other factors that could cause volatility – at least if reality unfolds as a bank expects, its expected losses should leave little room for surprises later on, right?

Wrong. It’s a sensible expectation, but unfortunately it’s not that simple and there could be real surprises, particularly in portfolios with lower volume and high value. A key reason is differences between the timing of loan defaults with their associated provision increases, versus the timing of provision decreases on good loans that repay in full. If they’re in different reporting periods, you could be in for a bumpy ride, with ‘spikes’ in the reported ECL expense that aren’t offset by provision releases. But to explain how this could happen, let’s use a simplified example.

Take a bank that’s lent 100 each to 5 different borrowers, loans A to E. Assume that if any loan defaults then the bank will lose the entire 100, and the bank expects on average that 1 of the 5 loans will default. The bank therefore estimates each loan’s lifetime probability of default to be 20%. That means the bank has a lifetime ECL provision of 20 on each loan (100 × 20%), giving a total provision of 100 across the 5 loans.

Now assume that loan A defaults (and, if I lose you, take a look at the table at the end which summarises this). Loan A’s provision will increase from 20 to 100, since the bank expects to lose the entire 100, and this increase generates a loss of 80. But loan A defaulting doesn’t mean that loans B to E are suddenly certain to repay. Assume there’s still uncertainty about the economy and the prospects for loans B to E, so their default probabilities remain 20%. So their provisions of 20 don’t change, generating no gain or loss. That means that, in the period loan A defaults, our bank will report a loss of 80. So, is the model broken already – as the expected happened, but we still booked more losses?

No, it’s not, but to see why you need to look longer term. If the original expectation of only 1 loan defaulting ultimately happens, then over the remaining life of the ‘good’ loans B to E the provision of 80 held against them will be released until it reaches zero, when they repay in full. Those releases will give gains in the P&L of 80. That shows that the bank’s starting provision of 100 made sense: the subsequent loss on loan A of 80 is offset by the subsequent gains from loans B to E of 80, and the 100 provision is also the loss suffered on loan A.

But this highlights the importance of the timing of defaults versus the timing of provision releases on ‘good’ loans. If loans B to E were long-term loans to companies that remained vulnerable for a number of years, then the gains from releasing their ECL provisions could happen years after the ‘spike’ in ECL losses from the default of loan A. But, if loans B to E were short-term loans whose credit quality quickly improved, most of their ECL releases could occur in the same period as when loan A defaults, so there’s little or no ‘spike’.

So what does this mean in practice? To minimise surprises, management should explain the possibility of these timing mismatches when discussing ECL forecasts with stakeholders and, if they can, provide context on when increases and releases might happen. And stakeholders shouldn’t jump to the conclusion that the IFRS 9 ECL model isn’t working, if they see these spikes when defaults happen. On reflection, the real surprise would have been if something as complex as the IFRS 9 ECL model had worked entirely as we expected…

Summary table

 

Period 1: at the start

Period 2: loan A defaults

Period 3: loans B–E fully repay

 

PD

ECL provision (A)

PD

ECL

provision (B)

ECL income / (expense)

(A – B)

PD

ECL

provision (C)

ECL income / (expense) (B – C)

Loan A

20%

20

100%

100

(80)

100%

100

0

Loan B

20%

20

20%

20

-

0%

0

20

Loan C

20%

20

20%

20

-

0%

0

20

Loan D

20%

20

20%

20

-

0%

0

20

Loan E

20%

20

20%

20

-

0%

0

20

Total

 

100

 

180

(80)

 

100

80

Contact us

Mark Randall

Mark Randall

Director, PwC United Kingdom

Tel: +44 (0)7764 988946

Follow us