3. A TDR exists when a creditor grants a concession to the debtor for economic or legal reasons related to the debtor's financial difficulties that it otherwise would not consider. A concession is deemed to have occurred when an entity does not expect to collect all amounts due according to the contractual terms of the original loan as a result of restructuring. Therefore, in all TDRs, the entity does not expect to collect all amounts due according to the original agreement.

4. If a creditor determines, based on current information and events, that it is probable it will be unable to collect all amounts due according to the contractual terms, an individual loan is deemed to involve impaired loans. Accordingly, based on the conclusion in paragraph 3 above, all TDRs are deemed impaired. Once a loan is deemed impaired, the creditor is required to measure the individually impaired loan using the present value of expected future cash flows discounted at the loan's original effective interest rate, assuming the creditor does not apply one of two available practical expedients (that is, the loan's observable market price or the fair value of the loan's collateral if collateral dependent). Consequently, TDRs that effect both principal and future interest are measured under current GAAP because a DCF method is applied. ²

5. The following represent the key changes made by the amendments in Update 2016–13 that affect the accounting for TDRs:

6. The FASB staff presented to the TRG members the issue regarding whether entities should forecast all types of reasonably expected future TDRs on a portfolio basis and include the effect of those reasonably expected TDRs in the calculation of expected credit losses. The staff identified two views:

7. The staff notes that under both of these views, all effects of a TDR are considered in the estimate of credit losses. The only difference between the two views is that in View A, the effects of a contractual term extension (that is, additional credit risk exposure over an extended measurement period and the effect of other expected concessions, if applicable) would be forecasted at a portfolio level before assets are specifically identified as being troubled. Whereas in View B, the effects of a contractual term (that is, additional credit risk exposure over an extended measurement period and the effect of other expected concessions, if applicable) extension should be considered only upon specific identification of a reasonably expected TDR.

8. Various forms of TDRs were discussed throughout the discussion at the June 12, 2017 TRG meeting. Some members noted that such loss mitigation activities may have a direct or indirect effect on loss history. For example, the loss history for an entity would reflect both the credit risk for troubled loans and the offsetting loss mitigating effect of successful TDRs. Therefore, some effects of TDRs may already be captured in a portfolio's historical loss information. Some TRG members suggested that because of these factors, future TDRs may already be appropriately reflected in the measurement of credit losses for portfolios of performing financial assets (that is, the estimate may incorporate a lower amount of historical credit losses because of successful TDRs). Under this view, any incremental adjustment to portfolios of performing financial assets could misrepresent the effects of TDRs on performing portfolios and could result in additional cost and complexity.

9. Certain TRG observers noted that if the loss mitigating aspects of TDRs are recognized in loss rates for performing pools, all concessions related to the loss mitigation also should be reflected. Those TRG observers argued that to obtain this symmetry, all of the effects of reasonably expected TDRs should be forecasted on a portfolio basis for all loans, including performing pools, before troubled loans are specifically identified.

10. TRG members noted that while some TDR activity is captured in overall loss history, the subsequent accounting for certain TDRs (for example, interest rate concessions and term extensions, which are often provided as part of the same modification) may not be reflected in that loss history because of nonaccrual policies. ³ In addition, considering various methods are allowed under the new standard, some questioned if certain accounting concessions (for example,interest rate concessions) would be captured if a DCF analysis was not explicitly required.

11. The discussion at the June 12, 2017 TRG meeting with regards to measuring concessions given on TDRs revealed a potential area of confusion between the various methods permitted in the guidance to measure credit losses in paragraph 326-20-30-3 and paragraph 310-40-35-10, which states:

12. The staff believes the combination of the following guidance means that all effects of reasonably expected TDRs ⁴ need to be incorporated into the allowance for credit losses:

13. Stakeholder feedback has indicated that there are no concerns with considering the benefits of loss mitigation strategies or, once a TDR is specifically identified on an individual basis, considering additional exposure to credit risk over an extended term. Rather, questions have been raised on the requirement to reflect economic losses related to concessions in the allowance and when to reflect those losses under the new standard. The staff agrees with the TRG discussion (summarized in paragraph 10) that entities may be able to measure some concessions only through a DCF methodology (or reconcilable method). For further context, the staff prepared the example below, which illustrates several fact patterns for a problem loan and describes why different measurement methodologies may not reflect certain concessions, such as interest rate concessions.

14. Assume that at period 0 an entity holds a par loan with a single borrower that produces contractual cash flows at 5 percent on a principal balance of $100 over 3 periods. Assume that at the end of period 1 the borrower has experienced financial difficulty and that the entity has decided to offer an economic concession by lowering the interest rate to 3 percent in a TDR. Loan A in the following table is a hypothetical loan that would result if the borrower had not experienced financial difficulty and the TDR had not been performed. Loan B represents the reality of the loan described above. The difference in column (a), total interest cash flows, results from differences in cumulative interest income, and if there was any difference in column (b), total principal cash flows, it would result from differences in cumulative principal cash flows.

15. The following table illustrates various credit loss measurement methodologies and the basis for that measurement. Non-DCF methodologies typically are based on net charge-offs (NCOs), which may not capture all losses of interest cash flows, while DCF calculations consider the differences in all interest and principal cash flows (both columns (a) and (b) above). In this example that assumes the entity continues to receive current payments on Loan B, historical NCO data does not provide any information on the economic loss from the interest rate concession. As noted in the table below, neither a loss rate or probability of default and loss given default (PD/LGD) measurement methodology (if based solely on NCO information) ⁵ reflect differences in interest cash flows; therefore, both result in a loss statistic of zero in this example.

16. Additionally, U.S. banking regulators generally require a nonaccrual policy to be in place for problem loans, which exacerbates the issue of capturing interest related losses in historical loss information. These nonaccrual policies remove the effect of "lost interest" from the historical loss data in accounting systems and make the comparison of interest cash flows between the original and modified loan even more difficult. For example, the table below shows the interest income accruals recorded each period for the Loan B cash flows if the loan was placed on nonaccrual status at the end of period 1. Loan B represents a nonaccrual policy with no reversal of previously recorded interest income, while Alternate Loan B represents a nonaccrual policy for which period 1 interest is reversed.

17. As compared with loans that are not accounted for using a nonaccrual policy, the above table shows that an entity’s accounting system no longer is accruing any interest on a loan that has been placed on nonaccrual status. Therefore, the entity would not reflect losses of interest cash flows in its historical loss data when a nonaccrual loan is ultimately charged off. In order to determine the amount of lost interest, the entity would have to manipulate existing data outside of the current system to simulate the interest income that would have been accrued under both the original and modified terms of the loan. This would add another layer of operational complexity and cause the measurement of lost interest in a non-DCF methodology to be difficult to determine with sufficient precision.

18. Based on the differences between DCF and non-DCF measurement methods and application of nonaccrual policies noted above, the staff believes it would be difficult, if not impossible, for an entity to determine, with the appropriate level of precision, an allowance that considers lost future interest with a non-DCF method.

19. The example above highlights that certain measurement methods may not capture losses related to interest rate concessions. Loan term concessions would similarly not be captured by certain measurement methods. However, because various measurement methods are permitted under Update 2016–13, some stakeholders have questioned whether the Board intended to extend the flexibility in measurement method to capturing concessions given on TDRs, even if that means some of those concessions might not be captured.

20. The staff believes that while Topic 326 provides flexibility regarding different measurement methodologies, the requirement in paragraph 310-40-35-10 for concessions to be reflected in the allowance cannot be ignored. Accordingly, it is the staff's view that if a concession is given on a TDR that can be captured only through a DCF methodology, that concession must be measured using a DCF methodology (or a reconcilable method) either on a pool or individual basis. The staff notes that because a DCF methodology is required for impaired loans, including TDRs under current GAAP, this view should not represent a significant operational issue for most institutions.

21. There have been two different stakeholder interpretations of the following paragraph regarding reasonably expected TDRs. Paragraph 326-20-30-6 states in part:

22. The differences in interpretation hinge upon what was meant by the Board's use of the phrase reasonably expected. Some stakeholders maintain that reasonably expected means that the effects of TDRs should be recognized before they are specifically identified. If those effects are to be recognized before specific identification, the assessment of reasonably expected TDRs would be done on a portfolio basis for performing assets. These stakeholders assert that an entity may be able to reasonably expect a specific level, or percentage, of TDRs that will occur in the future. Under this logic, earlier recognition of the effects of TDRs would result as compared with if the assessment were made on an individual financial asset basis because entities generally would be unable to reasonably expect individual TDRs until the borrower is experiencing financial difficulty. These stakeholders note that this aligns with certain objectives of Update 2016-13. Those objectives include earlier recognition of credit losses and elimination of thresholds before losses are recorded.

23. However, other stakeholders maintain that TDRs are reasonably expected only when they can be specifically identified on an individual financial asset basis, which is consistent with the statement in paragraph 326-20-30-6 on executing a TDR with "the borrower." These stakeholders add that the objective of paragraph 326-20-30-6 is to limit the estimation of credit losses to the contractual term except in limited situations (that is, when an expected TDR can be specifically identified) to avoid requiring a "life of relationship" estimate. If the measurement period is extended beyond the contractual term based on portfolio-level forecasts of TDRs, these stakeholders note that the estimate of credit losses would be one step closer to a "life of relationship" approach, which was not the Board's intent. Additionally, these stakeholders observe that the Board's decisions on TDRs in Update 2016-13 only addressed the measurement of TDRs, not the identification of TDRs, which historically has been performed on an individual financial asset basis.

24. As noted in the TRG discussion summarized in paragraphs 8 through 10 of this memo, some of the effects of TDRs already are reflected in historical loss information (for example, the benefit of lower losses resulting from loss mitigation activities). However, stakeholders have noted that for portfolios of performing loans, it is not practical for entities to develop an expectation for specific loans that will become TDRs and the related effects that are not already reflected in historical loss information. Many of these stakeholders do not currently capture data on other effects of TDRs that are not reflected in historical loss information, such as interest rate concessions, to the extent necessary for an appropriate incremental adjustment to be made at the portfolio level before expected TDRs can be specifically identified. The staff also notes that because of the large extent to which TDRs were executed during the last financial crisis, in many cases through government mandated programs, many entities' historical data will reflect an economic period and modification programs that may be inconsistent with the current and future economic periods and programs. Therefore, this data may not be indicative of future TDRs, which means entities may not find it to be relevant information to consider when estimating credit losses.

25. Beyond the practical issues of identifying TDRs within portfolios of performing loans noted in the previous paragraph, stakeholders also observed that such an approach may create conceptual issues when trying to support that loans in a portfolio share similar risk characteristics. If an entity believes certain loans within the portfolio will become TDRs, then those loans inherently have different risk characteristics than the other loans in the portfolio and would need to be evaluated separately (which would be impractical because they have not yet been identified). Some stakeholders believe that to address this potential issue, they would need to create a hypothetical pool of loans for purposes of measuring credit losses, which they believe was not the Board's intent. ⁶

26. Overall, the staff believes there are three alternatives for the timing of when the effects of a TDR not already included in historical loss information are recorded:

27. The staff believes requiring forecasts of a TDR's activity to be made on pools of performing loans is inappropriate because the lack of appropriate historical data in this area most likely would preclude the “reasonably expected” notion from being met for most entities. Because many entities do not have the data or systems to achieve an appropriate level of precision, the staff believes that it is unreasonable to expect an entity to forecast the effects of future TDRs on loans in performing pools.

28. Furthermore, the staff notes that most assets are evaluated for restructuring based on each asset’s unique facts and circumstances surrounding the individual borrower (whether that is based on an individual analysis or a programmatic analysis, such as assets exceeding a certain delinquency threshold). Additionally, the staff acknowledges the potential circularity issues that may arise when determining credit losses for a portfolio in which TDRs are forecasted before they are specifically identified (as discussed in paragraph 25). Accordingly, the staff believes that the incremental effects of TDRs outside of those already reflected in historical loss information should be recognized when an asset is specifically identified as an expected TDR, which would be before execution. The staff believes it was not the Board’s intent to require separate tracking and calculation of the effects of TDRs that may potentially occur within performing asset portfolios from day 1. The staff believes that if such an approach were required, it would result in additional subjectivity and complexity and it would not provide a significant benefit in terms of the information on credit losses being provided.

29. The staff notes that entities typically will have processes in place to help them identify certain loans that they may decide to attempt to recover on through a TDR. These processes may be different depending on the risk characteristics of the loans being considered, but there always will be indicators of potential TDRs before execution of the TDR. For example, entities may have policies in place that require a modification to be considered for all loans past a certain delinquency threshold. Entities may even have additional policies that specify the kinds of concessions that should be offered depending on the borrowers’ specific circumstances. Therefore, the staff believes that entities will need to consider their own policies and practices for modifying assets when determining the point in time a TDR is considered reasonably expected. The staff believes a TDR would always be reasonably expected before execution, and there often would be indications of a reasonably expected TDR before individual negotiations with a borrower begin. Determining the exact timing of when a reasonably expected TDR exists is an area where entities will need to apply judgment based on their individual facts and circumstances. The staff also notes that entities may incur additional costs to develop new processes to capture data on when TDRs become reasonably expected in their systems because this is a new concept that has not previously been applied for financial reporting purposes.

30. The staff believes that consistent with the analysis in paragraphs 26 through 29, identification of a reasonably expected TDR should be on an individual asset basis. The staff notes that the guidance for identifying TDRs was not amended by the new standard, and the requirement in current GAAP to individually identify TDRs also should apply to identifying reasonably expected TDRs.

31. The staff understands that in limited circumstances in which third-party loan servicers are used, there may be cases in which the servicer does not report TDRs to the lender until after they have been executed. In these cases, because of the operational challenges of developing a reasonable expectation of TDRs on an individual asset basis, the staff would not object to an entity estimating reasonably expected TDRs on a portfolio basis based on metrics that represent the individual loans within the pool that are TDRs as of the reporting date, but for which information on the modification has not yet been received from the servicer. While the staff would not object to a portfolio-level approach that is a proxy for View B as described above, the staff would object to View A and View C because those views would not meet the objective of the guidance on identifying TDRs (as described in paragraph 30).

32. The staff believes that because the concept of an individually impaired loan has been removed as an accounting concept, it was the Board's intent for the unit of account with which an entity should measure TDRs to be consistent with that which is applicable to all of Topic 326. That is, once a TDR is individually identified, it may be measured either individually or on a pool basis depending on the risk characteristics of the loan(s). Paragraph 326-20-30-2 states in part:

33. The staff understands that when the effects of TDRs are measured with a DCF methodology under current GAAP, the measurement typically is performed on an individual asset basis, which should continue to be an operable and acceptable approach for most entities under the new guidance.

However, the staff notes that the new guidance allows an entity to pool TDR loans that have similar risk characteristics and measure the effects of TDRs for that pool. If the entity has compiled historical TDR data on past loans with similar risk characteristics, the staff believes a measurement of the effect of expected TDRs at a pool level would be an acceptable approach.

34. The staff believes the result of the new guidance surrounding TDRs and credit losses is to accelerate the recognition of the economic concessions offered in TDRs from the point of execution to the point at which they are reasonably expected. However, the staff notes that the Board did not change guidance on identifying TDRs from current GAAP, which requires TDR guidance to be applied at the point a loan can be specifically identified as a TDR on an individual basis. Therefore, the staff believes that when a loan is individually identified as a reasonably expected TDR (View B), all effects of the TDR should be reflected in the allowance for credit losses.

35. The staff believes it was not the Board's intent to require the effects of TDRs to be forecasted on a portfolio basis before a loan is specifically identified as a reasonably expected TDR (View C). This would require an entity to adjust non-DCF measurement methodologies on performing pools of loans for the effects of future TDRs not already reflected in loss statistics. Because certain concessions in TDRs can be measured only using a DCF method (or reconcilable method), the only way for all effects of TDRs to be reflected in the allowance for credit losses on a portfolio basis would be to require all entities to use a DCF method (or reconcilable method) to measure credit losses on all portfolios of loans for which TDRs are expected to occur. Additionally, there may be separate issues with segmentation of portfolios for which TDRs are expected to occur because loans that are expected to become TDRs inherently have different risk characteristics than other loans, which may introduce additional complexity into an entity's evaluation of similar risk characteristics. The staff believes that these points conflict with the Board's intent to allow flexibility when measuring credit losses. Additionally, the intent to reduce cost and complexity of applying the new guidance because identifying TDRs on a portfolio basis is operationally unfeasible for most entities. Therefore, the staff believes that reasonably expected TDRs are those that can be specifically identified, and at the point of individual identification all effects of the reasonably expected TDR should be reflected in the allowance. The staff notes that this would result in earlier recognition of the effects of concessions compared with current practice, which is generally when a TDR is executed.

36. Because there is flexibility regarding the methods used to measure expected credit losses under the new standard, there may be situations in which some effects of TDRs, such as the effect of loss mitigation, already are reflected in the allowance from day 1, while other effects, such as the economic impact of concessions, are considered later (that is, at the point that a reasonable expectation exists) in a refinement of the initial estimate. Because of the requirement in paragraph 310-40-35-10 that requires concessions to be recorded in the allowance for credit losses, the staff believes that at the point a loan is specifically identified as a reasonably expected TDR, an entity must use a DCF method if the TDR involves a concession that can be captured only using a DCF method (or reconcilable method), such as an interest rate or loan term concession. The staff notes that current GAAP requires the use of a DCF method for impaired loans, including TDRs; therefore, this should not represent a significant operational issue for most entities.

37. The staff notes that the standard requires credit losses to be evaluated on a pool basis unless risk characteristics change to an extent that they must be assessed separately. Because there is no further specification about to the unit of account under which the effects of TDRs must be measured, the staff believes it was the Board's intent to allow entities to measure the effects of TDRs either on a portfolio basis or individual basis depending on which is most appropriate for the facts and circumstances of the assets. Accordingly, the staff believes that the effects of TDRs may be measured either individually or on a portfolio basis, as long as all concessions are being captured.

38. Regarding the identification of TDRs, the Board agreed (7-0) with the staff's recommendation that when a loan is individually identified as a reasonably expected TDR (View B), all effects of the TDR should be reflected in the allowance for credit losses. The Board also agreed that the application of View C for identification of TDRs would not be consistent with current or future guidance under Topic 326.

39. Regarding the measurement of TDRs, the Board agreed (7-0) with the staff's recommendation that at the point an individual loan is specifically identified as a reasonably expected TDR, an entity must use a DCF method if the TDR involves a concession that can be captured using only a DCF method (or reconcilable method).

40. The Board also noted that if an entity uses a DCF method for measurement of credit losses on a performing loan portfolio, any effects of TDRs that are incremental to what is embedded in the historical loss information should not be incorporated as an input to the DCF method until a TDR is individually identified in accordance with the approach in paragraph 38. However, if certain effects of TDRs are embedded in the historical loss information, adjustments may be made to account for differences between historical TDR activity and reasonable and supportable forecasts of future TDR activity.

41. The staff notes that the Board decisions have no effect on the timing of when TDRs should be reflected in the applicable disclosures, which will continue to be at the time a TDR is executed.