IFRS 9 is the new accounting standard from the International Accounting Standards Board for credit losses on portfolios of loans. It will come into effect in most jurisdictions for reporting periods starting January 2018. One of the key principles is that lenders should use relevant data that is reasonably available to assess the appropriateness of credit provisions. While much effort has gone into interpreting and implementing the standard, little consideration has been given to how firms and auditors will interact with each other and the changes this might require.
In practice, audit firms have multiple (private) reference points across their clients and as a result, can take a view on provision levels and component models that may lead them to raise concerns with an individual client’s models. The auditor’s comparators however, are likely to be incomplete and not entirely like-for-like. To counteract these views, lenders will need clear and referenceable evidence to potentially rebut auditors’ initial position. Addressing these audit concerns is likely to involve considerable effort from the bank. Given the significantly increased complexity of IFRS 9 and the need to isolate differences caused by forward-looking assessments vs underlying model assumptions. To support banks on the discussion with their auditors it is going to be very helpful for them to have access to independent, granular and comparable benchmarking to set common expectations and probe specific concerns or counter potential misconceptions the auditor may have, regarding the relative position of the lender to its peers.
The type of questions that auditors will be asking banks might include the following:
How does the bank justify that Stage 2 criteria for identifying significant increase in risk a reasonable and prudent view?
How does the bank demonstrate that it is holding appropriate lifetime provision relative to peers on similar Stage 2 assets?
How does the bank illustrate that it understands the key economic sensitivities in its portfolio on Probability of Default (PD) and Lifetime Expected Loss?
By default, reporting institutions will in due course have the public disclosures detailed in IFRS 7 to support their peer comparison analysis. This however, is unlikely to be ideal and represents significant challenges, particularly in the short term when no bank has published any numbers. Inherently, the complex nature of IFRS 9 means comparing aggregate provision or the individual component outputs may be impossible without an explicit known reference point, stripping out for example, economic outlook effects or different customer treatment strategies. The benefit of having a detailed, account level, peer benchmarking solution is that it allows a true ‘component by component’ comparison isolating interfering factors that might include:
Economic outlooks or the approaches to weighting of multiple scenarios
Stage 1 / Stage 2 / Stage 3 allocation decisions
Risk grade/segment definitions
Treatment of EIR calculations
Treatment of behavioural lifetime calculation
Underlying PD models
Treatment of EAD calculation
LGD models and sensitivities
To support these requirements multiple comparisons are required:
Comparing the bank to its peers in terms of provision level by different segments/stages; this explores any provision rate differences that exist
Comparing the bank’s models against component benchmark models using the bank’s economic outlook and evaluated only on the bank’s assets; this explores pure effects of model differences between the bank and a reference set of models
Comparing the peer assets average model to the benchmark models on the pool using a common set of economic conditions; this identifies how the average of the pool of banks in house models perform against the benchmark
A small sample of the types of output are shown below:
Here the Client PD estimate for risk grade 1 and 2 is lower than the benchmark on the same accounts (risk grade defined by benchmark model). Furthermore, the pool average is lower than the benchmark on a like for like basis (columns 3 and 4). One could therefore conclude that the benchmark model is more conservative than client or the pool average. The Client however is probably more conservative on a like for like basis than the average of its peers for risk grade 1 based on the relative differences. Conversely we can see the client is broadly in-line with peers on risk grade 2.
This graph shows that over time, for Stage 1 accounts at the Client, the lifetime PD has been going down relative to the 12-month PD on the same assets. This has been reflected similarly in the benchmark models on the same accounts. This effect isn’t evident in the pool. This effect therefore looks to be associated with the clients book specifically, but may not indicate a weakness in the model or a need for increased provision.
This detailed level of comparison is something that 4most is looking to deliver in partnership with benchmarking experts, Argus. We are launching IFRS 9 benchmarking in Q4 of 2016 and are looking to provide benchmark members information throughout parallel run in 2017.