One of the most common warnings about the impact of impairment accounting under financial reporting standard IFRS 9 is that impairment levels will be much more volatile. But is this always true?
First, let’s distinguish between “volatile” and “responsive”. IFRS 9 impairment models are intended to be more responsive to expected changes in both macro conditions and micro behaviour, as compared to its predecessor IAS 39.
One of the perceived failures of IAS 39 impairment models was that the recognition of impairment was “too little, too late”. This delayed recognition is actually codified into IAS 39:
“Losses expected as a result of future events, no matter how likely, are not reported.” (IAS 39, paragraph 59)
So even if you could reliably estimate that a certain percentage of your “good” accounts would become “bad” in the next 12 months, you would not recognise any impairment against them today. Even “incurred but not reported” (IBNR) provisions adhere to the principle that the loss events have already happened, even if the individual affected accounts cannot be identified.
In contrast, IFRS 9 accounting recognises at least 12-month expected credit losses against even the best quality exposures, including newly originated accounts. For revolving credit products this includes off-balance sheet exposures. So not only does IFRS 9 anticipate future loss events and future macroeconomic conditions, it also anticipates future drawdown of unused credit line.
So does the new approach create inappropriate volatility? Has the pendulum swung from “too little, too late” to “too much, too soon”?
Consider the following stylised example. Bank A holds only a trivial amount of IBNR provision against non-impaired accounts and that IAS 39 impaired accounts map perfectly to IFRS 9 Stage 3 accounts and that the chart below tracks one deteriorating account over time:
The account attracts more impairment under IFRS 9 until it finally becomes impaired, at which point the two methodologies theoretically align. So which impairment calculation is more volatile? Arguably the IAS 39 approach creates more volatility because the major increase only occurs when the account becomes impaired, whereas the IFRS 9 approach reaches the same level in several material “steps”.
Compare and contrast
In fact there are some ways in which IFRS 9 is actually intrinsically more stable than IAS 39 e.g.:
• Impairing against exposure rather than balance reduces large provision increases, as both balance and risk increase.
• Impairing against lifetime losses reduces artificial seasonality impacts.
• A sound stage allocation methodology will mean more gradually built impairment against higher risk in order accounts.
It is true that individual account provisions under IFRS 9 may be volatile as accounts move between Stage 1 (12-month expected credit loss provision) and Stage 2 (lifetime expected credit loss provision). But couldn’t the same be said under IAS 39 as accounts move between “impaired” and “non-impaired” states, with greater consequences?
Moreover there are several ways to manage volatility under IFRS 9, including:
• Adopting asymmetric stage definitions – define stage allocation rules to prevent large movements between Stage 1 and Stage 2. For example if an account enters Stage 2 due to delinquency then introducing a minimum cure period before it can return to Stage 1 will improve stability.
• Managing unused credit lines carefully – large differences between current balance and exposures at default will amplify the impact of moving from 12-month to lifetime expected loss calculations.
• Balancing financial volatility across the portfolio – robust IFRS 9 impairment calculations will reflect actual portfolio dynamics so managing the asset (and exposure) mix in anticipation of the change will naturally reduce volatility and encourage active asset allocation strategies.
• Managing volatility of macroeconomic forecasts – enforce strong governance over updates to macroeconomic forecasts, including the frequency of updates and the probability-weighting of scenarios, in order to manage input volatility.
• Ensuring model stability – provisions will be based on expected loss models, often themselves derived from individual probability of default (pd), exposure at default (ead) and loss given default (lgd) models. Following a rigorous and robust model design and development process will ensure these models are appropriately responsive without being inappropriately volatile.
• Optimising product design – the gap between lifetime losses and 12-month losses will naturally lessen where remaining contractual life is shorter, amortisation is more rapid and/or interest (discount) rates are higher.
We shouldn’t fear highly responsive expected credit loss models if they are accurate. These models provide valuable early warning systems to enable more time to take mitigating actions. Surely we can be more prepared when we can see storm clouds in the distance, rather than waiting for the rain. Earlier signs of credit deterioration enable more effective line management strategies for revolving products. Pricing and underwriting strategies can be adapted to anticipated changes in the economy. Collections and pre-collections activities can be prioritised based on a longer term view of account health.
The distinction between responsiveness and volatility is not simply semantic. IFRS 9 models are supposed to react quickly to a range of expected changes but not be overly punitive. Taking a holistic approach to the end-to-end IFRS 9 implementation process will ensure that provisions respond appropriately to actual and realistically anticipated changes in portfolio mix, customer behaviour and macroeconomic conditions.