Transfers between impairment 'buckets'

During the June 2011 joint Board meetings, the Boards agreed to continue development of an impairment approach for financial assets using three "buckets":

• The first bucket essentially comprises assets that are not in the second or third bucket (generally performing assets with no heightened expectation of future credit events) and would have an allowance amount recognised of less than the lifetime expected losses
• The second bucket would comprise assets that have been affected by the occurrence of observable events or conditions which indicate possible future default and would therefore have an allowance amount recognised of full lifetime expected losses (this category would typically be a portfolio of assets rather than individual assets)
• The third bucket would comprise assets where available information indicates credit losses are expected to, or have, occurred on individual assets and the allowance amount recognised would be the full lifetime expected losses (similar to bucket two).

At the last meeting, the staffs were asked to develop criteria on how and when assets should be transferred from one bucket to the next; the existing criteria in IFRS and US GAAP on current impairment triggers was one method suggested for developing this criteria.

The staffs presented to the Boards an agenda paper which summarised an event based approach (utilising the existing impairment triggers in IFRS and US GAAP) and a credit risk management approach. The credit risk management approach was further broken down into an 'absolute' credit risk model and a 'relative' credit risk model.

Under the event based model approach all assets (whether originated or purchased) would start in bucket one with transfers occurring based on specific events which indicate the existence of credit deterioration or improvement. However, the event based approach would also require differentiation between events that would cause a transfer between buckets and events that would not cause a transfer but would instead change the loss expectations. In considering this approach, the staffs felt it was difficult to identify a principle to differentiate events and changes in circumstances that would require a transfer across categories from those that would only impact the estimate of expected losses.

Under the 'absolute' credit risk model approach, the objective would be to reflect the change in credit quality of assets consistent with an entity's credit risk management practices. As assets are originated or purchased, they would be categorised into one of the three buckets based on its absolute level of credit risk, regardless of whether the asset's pricing appropriately reflects the associated credit risk.

Under the 'relative' credit risk model approach, the objective would be to reflect the credit deterioration or improvement in assets utilising an entity's credit risk management practices. This approach would result in transfers across buckets based on changes in credit loss expectations. As assets are originated or purchased, they would initially start in bucket one and then transfers to buckets two and three would occur based on changes in credit loss expectations. It was noted by the staffs that under this approach, some loans with higher credit risk could be held in bucket one while other loans with lower overall credit risk but for which changes in credit loss expectations have occurred would be held in buckets two or three.

During Board discussions, many Board members were generally supportive of the use of a credit risk model rather than solely an event driven model, although some Board members requested retaining or layering aspects of the event driven model into the credit risk model approach. Many Board members leaned toward the 'relative' credit risk model approach with some having fundamental concerns with the 'absolute' credit risk model approach.

One of the FASB members suggested using a 'more likely than not' threshold of whether the asset continues to perform consistently with the original pricing of the asset to determine whether the asset would be recognised in bucket one or bucket two. When it becomes 'probable' the asset will not perform consistently with the original pricing then the asset would move to bucket three. He also suggested potential disclosures of a roll-forward of asset balances from period to period which would include a migration of the allowance by bucket.

The FASB Chair added that if the event driven approach were incorporated into the credit risk model it was important that the terminology be more clear as use of the term 'events' suggests the model is no longer focused on 'expected' credit losses but rather incurred credit losses. She suggested use of the term 'cues' instead of 'events'.

The Board tentatively decided to proceed with development of a credit risk model using the 'relative' approach but including event indicators as cues for assets transferring between buckets.

'Bucket One' measurement

During the June 2011 joint Board meetings, the Boards tentatively agreed that for assets in 'bucket one' an allowance should be recognised equal to 12 months of expected credit losses based on the initial expectations plus the full amount of any changes in expectations. However, the Boards recognised there may be operational difficulties with such an approach and requested the staffs to further develop this approach ensuring it was operationally feasible and to conduct outreach to identify how to make the approach operational.

The outreach activities performed by the staffs (including EAP members) identified that the recognition approach supported by the Boards was the most operationally challenging of the alternatives considered by the Boards and would require significant systems modifications. Many also commented on the similarities to the original IASB exposure draft which was not considered operational for open portfolios of assets because of the requirement to track initial expectations of credit losses. These constituents also felt that if the migration to bucket two and bucket three occurred earlier than envisioned for transfers to the bad book would have been under the Supplementary Document (SD) proposals then the pressure on the allowance for bucket one would be diminished. As a result, they suggested the calculation for the bucket one allowance be as operationally simplistic as possible as the incremental benefit from a complex calculation methodology would not be justified.

The staffs presented the Boards with four alternative methods for calculating the bucket one allowance balance:

• Method A — 12 months' worth of losses expected to occur on the financial assets, or for the remaining expected life if that is less than 12 months
• Method B — 24 months' worth of losses expected to occur on the financial assets, or for the remaining expected life if that is less than 24 months
• Method C — A simplification of the original approach supported by the Boards by using a 'rolling' loss rate rather than maintaining an original loss rate
• Method D — A simplification of the original approach supported by the Board using remaining lifetime expected losses of assets expected to be transferred to buckets two and buckets three in the next 12 months plus credit deterioration on remaining assets in bucket one.

Based on the feedback provided during the outreach activities, the staffs believe that Method A and Method B are worthy of additional consideration but do not support further pursuing Method C and Method D because of the operational difficulty associated with these approaches. The staffs also wanted to clarify terminology as they felt the terms 'annual loss rate' and 'annualised loss rate' may be being used interchangeably when in fact they are different concepts. An 'annual loss rate' would be those losses expected to occur over the next 12 months while an 'annualied loss rate' would be the lifetime expected losses divided by the lifetime of the assets multiplied by a 12 month period. These two approaches could result in different allowance amounts if assets have loss emergence patterns that are front or back loaded.

One of the IASB members said that he was one of the more vocal supporters of the Boards' previous decision but he has changed his view to supporting either Method A or Method B as the allowance amount for bucket one would not be significantly different and because the Method C and Method D approaches would result in allowance balances that are not comparable across entities. One FASB member questioned how Method D could be described as a simplification of the Board's previous decision as he felt that such a method could not be practically applied. One IASB member said that he could support either Method A or Method B but that if a 12 month loss period was used then he would like to introduce disclosure requirements for assets on a 'watchlist' of being transferred from one bucket to another. One IASB member said he did not support either Method A or Method B because of the resulting day 1 loss that is associated with such an approach. One FASB member expressed concern with Method A as he felt that based on initial research he had performed of US banks over more than the last half century that banks typically have more allowance recorded than 12 months of write-offs, therefore he felt that Method A could result in less allowance being recognised that current US GAAP.

The Boards tentatively agreed to keep the calculation for the bucket one allowance operationally simple and to further consider an approach of using either 12 or 24 months' worth of losses expected to occur on the financial assets, or for the remaining expected life if that is less than 12 or 24 months. The Boards also tentatively agreed to require the use of an annual loss rate (or a 24 month loss rate).