ICAEW.com works better with JavaScript enabled.

Predicting the future with IFRS 9

Brendan van der Hoek looks at some of the challenges for banks using forward-looking scenarios in determining expected credit losses under IFRS 9.

Predict the Future Technical Expected Credit LossesCredit impairment under IFRS 9 is perhaps the single biggest accounting change for banks since the adoption of IFRS more than 14 years ago.

Providing for loans that show all of the signs of being impaired is the easy part. Now that it’s all about expected losses rather than incurred, as it was under previous rules, banks need to have more sophisticated systems, models and processes to ensure that they capture and apply the information needed to adequately determine credit provisions under the expected credit loss (ECL) impairment approach.

Many banks have been applying IFRS 9 for more than a year. The largest UK banks all recently reported their first full year of results under the standard, which included the impact of ECLs and incorporated the latest view of forward-looking information.

Challenge #1

‘Reasonable and supportable’ information
The use of forward-looking information is a key component of the ECL impairment approach. This is not straight-forward and involves judgement. No one can predict the future with certainty so the incorporation of forward-looking information introduces considerable volatility into banks’ results.

ECLs are measured in a way that is not just determined by evaluating a range of possible outcomes, adjusted for the time value of money, but also taking into account ‘reasonable and supportable’ information about past, present and future events and economic conditions.

Banks need to consider information at the reporting date about the current conditions, as well as forecasts of future events and economic conditions.

Information used should be available without undue cost or effort and typically includes a combination of internal and external sources of data. It’s important to note that, as the forecast horizon increases, the specificity of information used to measure ECLs decreases, meaning the degree of judgement required to estimate ECLs increases. Historical information, an important anchor in measuring ECLs, must be adjusted to reflect the effects of current conditions and forecasts of future conditions. Banks should also consider observable market information about the credit risk of the particular financial instrument.

Financial instruments are grouped on the basis of shared credit risk characteristics. This means forward-looking information might need to be tailored for each portfolio. How much weight to give that information depends on the specific credit risk drivers.

To illustrate, most banks lend across a broad customer base resulting in concentrations of risk exposure because of the sectors and geographic areas in which customers are based or work. A bank’s expectations over future unemployment in a particular sector may only be relevant to borrowers who work in that sector.

Challenge #2

Non-linear relationships between different forward-looking scenarios and their associated credit losses could have a material impact for ECL provisions.

Consider a bank with a portfolio of mortgages. If property prices were to fall by 10%, only a 2% increase in ECL is experienced, due to significant remaining over-collateralisation in the book.

However, if property prices were to fall by 20%, a 10% increase in ECL is experienced, as significantly more loans become under-collateralised and experience losses. This example illustrates that ECLs do not increase linearly as property values fall; rather they increase at a greater rate the further property prices fall. The use of a single, forward-looking economic scenario, such as one based on the most likely outcome, would not meet the objectives of IFRS 9 when there is a non-linear relationship.

Challenge #3

How many scenarios are enough?
There is no right or wrong answer. Banks need to exercise judgement in determining the appropriate number of forward-looking, macroeconomic scenarios that need to be considered in measuring ECL to adequately capture material non-linearity.

Many banks use a central scenario (sometimes referred to as a base case), an upside scenario and a downside scenario with some banks opting for additional scenarios for a more severe downside and/or optimistic upside. Alternatively, some banks use a Monte Carlo approach, which involves simulating a large number of alternative scenarios around the central scenario and averaging the outcomes into unbiased expectations of ECLs and probabilities of default. Recent reporting by banks tells us that the macroeconomic environment does influence the number of scenarios that banks might use at any given point. There is little in the way of strong evidence to suggest that a bank that opts for more scenarios is any better placed than a bank that uses less.

Challenge #4

Don't ignore the economics
The economic circumstances may require additional scenarios to capture non-linearities that previously did not need to be considered but are now more likely to materially affect ECL.

Some UK banks recognised additional ECL impairment allowances during 2018 to take account of economic uncertainty that could not be built into their new ECL models. Economic factors in the UK came to the fore in the second half due to the possibility of a no-deal Brexit. After becoming concerned that their scenarios did not adequately capture possible adverse economic outcomes, some banks applied additional downside scenarios in their year-end results, which topped up the amount of their ECL allowances to specifically address economic uncertainty.

Challenge #5

Timing is everything
Reasonable and supportable information of events and current conditions should be reflected in the assessment of significant increases in credit risk and the measurement of ECLs at the reporting date.

Because most banks run their ECL models one month in arrears, the information used prior to the reporting date needs to be updated to reflect conditions at the reporting date. Forward-looking effects will be determined at the reporting date through assumptions based upon expert judgement and models that will produce estimates. Revisions of estimates due to changes in circumstances or as a result of new information do not relate to prior periods. This means that developments after the reporting date would not typically result in adjustments.

Challenge #6

Sensitive to changes
Assumptions made about the future carry a significant risk that actual outcomes could result in material changes to ECLs in subsequent periods. Banks should disclose the sensitivity of ECLs to changes in those assumptions.

Many items in financial statements, including ECLs, cannot be measured with precision and can only be estimated. Actual outcomes are often different. While estimates should be based on reasonable and supportable forward-looking information at the reporting date, it is important for banks to disclose the sensitivity of their assumptions and the potential impact on ECLs if reasonably possible alternative assumptions had been applied.

A major concern with ECL methodology is the susceptibility of ECLs to volatility due to the use of forward-looking information. Some of these effects, largely masked in the earlier part of 2018 due to relatively benign economic conditions, were witnessed towards the end of the year as economic uncertainty crept in. The methodology will only be truly tested in a more widespread, severe economic downturn.

Many banks disclosed sensitivities of the forward-looking information used in the measurement of ECL provisions in their 2018 annual reports. Measurement uncertainty and sensitivity of ECLs is an area of focus by regulators and investors. Further refinement of these disclosures is likely in subsequent periods.

Comparison with US Generally Accepted Accounting Principles
In respect of forward-looking information, the requirements are broadly similar under both IFRS 9 ECL and the US’s current expected credit loss (CECL) approaches, the latter of which becomes operative in 2020. Both require information about past events, current conditions, and reasonable and supportable forecasts of future economic conditions to be considered when measuring ECLs. The CECL model requires entities to recognise lifetime ECLs for all assets, not just those that have had a significant increase in credit risk since initial recognition. This means ECLs will be higher and recognised sooner than under IFRS. The application of forward-looking information means ECLs under either model will not be an exact science.

About the author

Brendan van der Hoek, technical accounting, Santander UK and member of the Banking Committee