Deutsche Bank AG Taunusanlage 12 60325 Frankfurt am Main Germany Tel +49 69 9 10-00 Susan Cosper Technical Director Financial Accounting Standards Board ( FASB ) 401 Merrit 7 PO Box 5116 Norwalk, CT 06856-5116 director@fasb.org May 31 st, 2013 Re: File Reference No 2012-260 - Response to FASB Exposure Draft: Financial instruments Credit Losses (Subtopic 825-15) Dear Ms. Cosper, Deutsche Bank ( DB or the Bank ) appreciates the opportunity to provide feedback on the Impairment Exposure Draft (the ED ). We share the concerns that the FASB has about the delayed recognition of credit losses under the current incurred loss methodology. We welcome the Board s objective of finding a new model for impairment to replace the existing model which will result in earlier recognition of credit losses. It has been DB s belief that the new model should have no incurred loss impairment triggers but should rather recognize credit losses for events that are expected to occur in future and should require a consideration of a wider set of information than that used under the existing rules today. However, while we welcome the FASB s efforts in this area we are not supportive of the proposed Current Expected Credit Loss ( CECL ) model for the reasons we outline in detail below. Firstly, we believe that one of the key principles in the FASB ED that all expected credit losses should be recognized at the point of origination of a loan is flawed as it does not reflect the economics of a lending transaction. For originated loans the likelihood of non-payment by the borrower is factored into the interest rate of the loan via the risk premium agreed upon at inception. The risk premium is not received up-front and similarly lifetime losses should also not be recorded upfront. Therefore, we believe that recording a provision for all expected credit losses for originated loans is not consistent with the objectives of financial reporting as it does not faithfully represent the underlying economics. We strongly believe that a loan provision should reflect changes in credit quality. Secondly, the suggested approach by the FASB for assessing credit losses is not consistent with internal risk management practices. We would support a loan provisioning model that differentiates the level of provision held between those loans that are performing ( good book ) and those which are not ( bad book ), consistent with our internal risk management procedures. DB manages performing and non-performing instruments differently. The assessments of risk and calculations of exposures and expected loss are done on this basis by our Credit Risk Management division. We agree that the provisions for financial instruments held in the bad book should reflect losses immediately based on the full lifetime loss expectation for positions where there is significant credit deterioration or where there is an observable loss event (e.g. insolvency, covenant breach, 90 days or more past due). The Bank believes that the International Accounting Standards Board s ( IASB ) model, which differentiates between instruments which have suffered from credit deterioration since inception and those which have not, is more consistent with internal risk management practice than the FASB model. The IASB model reflects the economics of lending more closely and therefore is more consistent with the objectives of financial reporting. While the IASB model also requires a day one loss for the good book, the measurement of the loss is for a 12 month horizon, as opposed to a lifetime concept, which is consistent with our internal risk management practice and regulatory treatment under the Basel framework today. The provisions for expected losses on the performing book should be recognized on a basis consistent with interest income recognition and with a one year expected loss provision. DB does not agree with taking full
lifetime losses at inception of a loan and recommend that the FASB and IASB work together to develop a converged approach to the calculation of expected loss as this would greatly enhance the comparability of financial statements. Such an approach should take into account the need for reliability and comparability which we believe can be achieved only through concrete provisions like those in the IASB proposal, for example, introducing a one year time frame for computing expected losses on loans which have not experienced significant credit deterioration. The following points are focused on specific aspects of the FASB proposal. A) Interest income recognition DB recommends a consistent model for interest income recognition applied to all assets. Hence, we recommend that the interest rate be decoupled from the expected loss calculations as envisioned in the FAS 114 Effective Interest Rate ( EIR ) model, and that the original rate should continue to apply even after an asset is impaired. We believe this would more closely align the standard with the proposed single measurement objective of holding an amortized cost asset at the present value of its future expected cash flows discounted at the original effective interest rate. We do not support introducing the concept of nonaccrual into the accounting literature because we do not believe that there is a conceptual basis for nonaccrual under the definition of amortized cost and application of the EIR methodology. Furthermore, we support the consistent application of this concept to purchased credit impaired loans and removal of the existing asymmetrical income recognition pattern for these portfolios. Whilst we acknowledge that it is required for regulatory reporting in the US and is an approach used by many US institutions in current practice, we recommend that the nonaccrual concept continue to remain outside of the accounting literature. B) Practical Expedient We welcome the FASB s consideration of a practical expedient approach when assessing credit impairment under the CECL model for debt securities that are classified as Fair Value through Other Comprehensive Income ( FVOCI ) when substantially all of the changes in value are attributable to non-credit related variables (i.e. liquidity). We are, however, concerned with the limitations of the approach as currently defined in the ED and are not clear under which scenarios we would be able to avail ourselves of this option. In particular, as further discussed in our response to question 7, we are of the view that the first condition (825-15-25-2a) would rarely be met in a market environment where the interest rate is likely to increase. This situation would result in the fair value of the debt instrument being lower than its amortized cost even though the expected loss on the individual financial assets are insignificant (the 2nd condition under 825-15-25-2b). Using this practical approach as proposed will mean an accounting result which is inherently inconsistent with amortized cost treatment. Furthermore, this option is not being contemplated by the IASB. We would like to see the Boards work together and propose a converged solution for instruments held at FVOCI. C) Troubled Debt Restructurings ( TDR s ) 2012-260 We do not support the FASB s retention of a special classification for TDR s. We believe that the concept of TDR is no longer relevant under the proposed credit loss model that requires a direct write off. Accordingly, we are of the view that under the proposed credit loss model, a TDR should not be accounted for differently from any other debt restructuring. To distinguish a TDR from other non-troubled restructured debt creates an exception to the general standard of debt modification and extinguishment. This would not be consistent with the spirit of principle-based accounting and is also operationally burdensome for institutions without adding any perceivable benefits. We agree that information on significant debt restructurings and modifications is of benefit to the users of financial statements and suggest that this information be included in a required disclosure for problem loans. Furthermore, removal of the TDR classification and the creation of a problem loan disclosure note covering these types of events will eliminate another divergence between US GAAP and IFRS.
D) Disclosure Requirements We do acknowledge that using an Expected Loss methodology introduces a certain level of complexity and judgment into the impairment process itself and that direct comparability across firms may prove difficult as a result, owing to differences in the factors used to measure risk and in the calculations of expected loss themselves. Hence, we welcome the proposal to include enhanced qualitative disclosures, particularly those that require disclosure of the explanations of the process by which a firm arrives at its expected loss estimates. We believe that enhanced qualitative explanations could obviate the need for significantly increased data requirements. We are also concerned with the volume of the additional disclosure requirements. We also believe that a conscious effort should be made to avoid duplicative disclosures, in particular, if other standards require similar disclosures that may satisfy the requirement. Entities should be permitted to cross reference relevant disclosures when similar information is provided in elsewhere in the financial statements and footnotes. E) Transition Considerations We generally agree that a cumulative effect adjustment to the statement of financial position is the appropriate transition approach for the proposed change in credit impairment. We believe that the transition date for implementation of the new standard and the accompanying disclosures should be earliest of January 1st, 2016 for public companies. This will ensure that the expected loss information we disclose is reliable and is well supported by both qualitative disclosures and accurate risk factor data. Concluding Remarks We hope to see the specific recommended changes to the ED as described in the paragraphs above because we believe that they would better align financial reporting with the underlying economics of the products and Industry standard risk management practices as well as with the models used for our regulatory capital calculations. If these changes are implemented this will not only better reflect how we risk manage our business, but will also yield a financial reporting result that is more reliable to the users of our financial statements. The results of the proposed model can leverage off of existing risk management frameworks and they will be robust, replicable and transparent. We strongly recommend that the FASB resume their joint deliberations with the IASB to work in the development of a converged approach to accounting for impairments and are hopeful that a converged model will be the result which recognizes that the credit risk is managed differently between the performing good book and nonperforming bad book. Our responses to the questions in the ED are attached in the appendix to this letter. We hope that our comments and answers are useful and relevant and look forward to continuing to work with you in future. Should you want to discuss the contents of this letter in more detail, please do not hesitate to contact us. Yours sincerely, Karin Dohm Managing Director Chief Accounting Officer Deutsche Bank AG Michael Fehrman Managing Director Head of Accounting Policy and Advisory Group Americas Deutsche Bank AG CC: Ms. Sue Lloyd, Senior Director, Technical Activities, IASB
APPENDIX Question 1: Do you agree with the scope of financial assets that are included in this proposed Update? If not, which other financial assets do you believe should be included or excluded? Why? Generally, we agree with the proposed scope of the standard, but would like to see financial guarantees included within the scope. Financial guarantees are in substance a form of lending similar to a loan commitment. Furthermore, inclusion of financial guarantees would align the scope of the FASB provisioning model with that of the IASB. We would also like to understand why insurance contracts have been specifically excluded and under which new standard they will be addressed as the economics of these is often also directly comparable. Question 2 8 are for Users only, hence we have not provided any answers. Question 9: The proposed amendments would require that an estimate of expected credit losses be based on relevant information about past events, including historical loss experience with similar assets, current conditions, and reasonable and supportable forecasts that affect the expected collectability of the financial assets remaining contractual cash flows. Do you foresee any significant operability or auditing concerns or constraints in basing the estimate of expected credit losses on such information? DB currently uses historical specific loss information adjusted for current and future predicted events to arrive at its estimates of expected loss. The regulatory requirements under the Basel regime requires financial institutions to either use loss parameters provided by the regulators or to introduce a comprehensive and complex framework to determine and validate parameters used for amongst others their Risk Weighted Assets ( RWA ) and EL calculation. All parameters used for this purposes and which will be leveraged for the envisioned EL calculation to determine the general risk reserves for all asset in scope of this ED are calibrated based on the firm s historic loss experience. The rating process itself considers current conditions and anticipates reasonable and supportable information about possible future developments which might affect the client s ability to fulfill its contractual obligations. Question 10: The Board expects that many entities initially will base their estimates on historical loss data for particular types of assets and then will update that historical data to reflect current conditions and reasonable and supportable forecasts of the future. Do entities currently have access to historical loss data and to data to update that historical information to reflect current conditions and reasonable and supportable forecasts of the future? If so, how would this data be utilized in implementing the proposed amendments? If not, is another form of data currently available that may allow the entity to achieve the objective of the proposed amendments until it has access to historical loss data or to specific data that reflects current conditions and reasonable and supportable forecasts? As outlined in the answer to question 9 above, DB is required by the Basel regime to have access to historical data. We also consider current market conditions during the rating process as well as reliable estimates of future conditions which are used in the calibration of risk parameters. Question 11: The proposed amendments would require that an estimate of expected credit losses always reflect both the possibility that a credit loss results and the possibility that no credit loss results. This proposal would prohibit an entity from estimating expected credit losses based solely on the most likely outcome (that is, the statistical mode). As described in the Implementation Guidance and Illustrations Section of Subtopic 825-15, the Board believes that many commonly used methods already implicitly satisfy this requirement. Do you foresee any significant operability or auditing concerns or constraints in having the estimate of expected credit losses always reflect both the possibility that a credit loss results and the possibility that no credit loss results? DB will not have any operability problems when determining the likelihoods of loss or no loss on a particular position or portfolio. We use probabilities of default which we track based on historical borrower data. There are also various sources of default data observable in the market. The likelihood of loss or no loss for individual instruments or portfolios can be ascertained from these probabilities of default. Question 12: The proposed amendments would require that an estimate of expected credit losses reflect the time value of money either explicitly or implicitly. Methods implicitly reflect the time value of money by developing loss statistics on the basis of the ratio of the amortized cost amount written off because of credit loss and the amortized cost basis of the asset and by applying the loss statistic to the amortized cost balance as of the reporting date to estimate the portion of the recorded amortized cost basis that is
not expected to be recovered because of credit loss. Such methods may include loss-rate methods, rollrate methods, probability-of-default methods, and a provision matrix method using loss factors. Do you foresee any significant operability or auditing concerns or constraints with the proposal that an estimate of expected credit losses reflect the time value of money either explicitly or implicitly? If time value of money should not be contemplated, how would such an approach reconcile with the objective of the amortized cost framework? Time value of money is considered both implicitly and explicitly in the loss estimates used by DB for our internal risk management purposes. Expected losses are a function of not only the probability of default but also the time to default, the prevailing interest rates in the market and the likely losses given default. There is interplay between each of these factors which is implicitly affected by the time value of money. Furthermore, for longer dated exposure the time value of money can be explicitly factored into the model by adjusting the estimates of loss given default on an instrument by instrument or portfolio basis. Question 13: For purchased credit-impaired financial assets, the proposed amendments would require that the discount embedded in the purchase price that is attributable to expected credit losses at the date of acquisition not be recognized as interest income. Apart from this proposal, purchased credit-impaired assets would follow the same approach as non-purchased-credit-impaired assets. That is, the allowance for expected credit losses would always be based on management s current estimate of the contractual cash flows that the entity does not expect to collect. Changes in the allowance for expected credit losses (favorable or unfavorable) would be recognized immediately for both purchased credit-impaired assets and non-purchased-credit-impaired assets as bad-debt expense rather than yield. Do you foresee any significant operability or auditing concerns or constraints in determining the discount embedded in the purchase price that is attributable to credit at the date of acquisition? Yields in the market that are credit risk free can be used to discount the future expected cash flows at inception. The credit spreads in the market have allowed for the computation of the present value of the cash flows not expected to be received. The difference between these two present value amounts can be used as the bad-debt adjustment to the initial carrying value. The price paid determines how much of the adjustment goes to the credit line and how much gets recorded as a discount or premium. Alternatively, the credit portion of the upfront adjustment can be determined by using the dollar equivalent of the one year expected loss of the position. Expected loss is a function of the probability of default which, in turn, can be observed in the ratings of the debt assigned by external rating agencies or in the historical data captured for that borrower by the lender. Additionally, we recommend that preparers have the option of applying this accounting treatment to not only purchased instruments but to all assets, including those that are originated by an entity. Thereby, any credit elements embedded in an origination price can be also adjusted for at the origination date. This will ensure consistent accounting treatment between purchased and originated assets where the terms and economics of two positions are exactly the same, with the only difference being that one is originated and the other purchased. Question 14: As a practical expedient, the proposed amendments would allow an entity to not recognize expected credit losses for financial assets measured at fair value with qualifying changes in fair value recognized in other comprehensive income when both (a) the fair value of the individual financial asset is greater than (or equal to) the amortized cost basis of the financial asset and (b) the expected credit losses on the individual financial asset are insignificant. Do you foresee any significant operability or auditing concerns or constraints in determining whether an entity has met the criteria to apply the practical expedient or in applying it? Yes, but we believe that this provision will be of very limited application as, in the majority of economic scenarios, entities will not always be able to avail themselves of this option. An entity may elect credit losses when the fair value is greater than the amortized cost. If the fair value is greater than the amortized cost this is primarily owing to one of two things an improvement in the market s perception of the credit risk and/or liquidity risk. In ordinary market conditions, these two risk factors would impact value in the same direction. It would be unusual to see a fair value higher than an amortized cost owing to a liquidity factor that is higher than the credit factor; and is positive whilst the credit effect is negative. The fact that market interest rates may increase in future means fair values will be lowered even further by this risk factor. It seems unlikely that the liquidity effect will ever materially outweigh the combined effect of credit and interest rate movements. Furthermore, the fair value expedient is not consistent with the single measurement objective of holding an asset at the present value of its cash flows discounted at the EIR.
Question 15: The proposed amendments would require that an entity place a financial asset on nonaccrual status when it is not probable that the entity will receive substantially all of the principal or substantially all of the interest. In such circumstances, the entity would be required to apply either the cost-recovery method or the cash-basis method, as described in paragraph 825-15-25-10. Do you believe that this proposal will change current practice? Do you foresee any significant operability or auditing concerns with this proposed amendment? Nonaccrual of interest is a regulatory requirement that has become industry practice in the financial reporting of US banks. We believe that this requirement should not be introduced into the new standard as it is not consistent with the single measurement objective of holding an asset at the present value of its cash flows discounted at the EIR. The nonaccrual concept mixes the recognition of credit losses with interest income recognition in a way that is not beneficial to the users of financial statements. Also, it is not clear how nonaccrual accounting would apply to Purchased Credit Impaired ( PCI ) and TDR assets. For PCI assets the interest accrual could end up being stopped immediately after the assets are acquired. For TDR assets the accrual could be switched off immediately upon classification of these assets as TDR. Furthermore, we believe that information regarding loans that are on nonaccrual can be adequately disclosed in sufficient detail in the notes to the financial statements without the need to create a different accounting treatment from the primary proposal of the standard which is amortized cost. Question 16: Under existing U.S. GAAP, the accounting by a creditor for a modification to an existing debt instrument depends on whether the modification qualifies as a troubled debt restructuring. As described in paragraphs BC45 BC47 of the basis for conclusions, the Board continues to believe that the economic concession granted by a creditor in a troubled debt restructuring reflects the creditor s effort to maximize its recovery of the original contractual cash flows in a debt instrument. As a result, unlike certain other modifications that do not qualify as troubled debt restructurings, the Board views the modified debt instrument that follows a troubled debt restructuring as a continuation of the original debt instrument. Do you believe that the distinction between troubled debt restructurings and nontroubled debt restructurings continues to be relevant? Why or why not? DB does not believe that the distinction between troubled debt restructurings and nontroubled debt restructurings continues to be relevant under the CECL model. We understand that it is the current practice of many banks to consider a forgiveness or concession of contractual terms to be a loss that results in a charge off. This approach is expected to be reflected in the basis adjustment to the carrying value of the loan in the expected losses measurement under the credit loss model. We believe that when there are significant modifications and a related write-off on the loan, the lender should be viewed as granting a new loan under a new set of contractual cash flows. Accordingly, we are of the view that a TDR s should not be accounted for differently from any other debt restructurings. To distinguish a TDR from other non-troubled debt restructurings creates an exception to the general standard of debt modification and extinguishment. This would not be consistent with the spirit of principle-based accounting and is also operationally burdensome for institutions without adding benefits. Therefore, we recommend that the TDR classification be removed and that all debt restructurings, troubled or non-troubled, be accounted for under the same accounting principle for debt extinguishment and modification. Furthermore, we believe that modification and restructuring information can be adequately disclosed in sufficient detail in the notes to the financial statements, thereby obviating the need to create a different accounting treatment. Question 17 is for Users only, hence we have not provided an answer. 2012-260 Question 18: Do you foresee any significant operability or auditing concerns or constraints in complying with the disclosure proposals in the proposed Update? While we agree that an institution should provide relevant and useful information for users to understand its credit risk management practices and the related methodologies and risk factors used to arrive at expected losses. We believe that a conscious effort should be made to avoid duplicative disclosures, in particular, if other standards require similar disclosures that may satisfy the requirement. Entities should be permitted to cross reference relevant disclosures when similar information is provided in elsewhere in the financial statements and footnotes.
We are also concerned with the volume of the additional disclosure requirements. For example, we recommend not including the roll forward table for debt instrument held at FVOCI as this information is not readily available and will require significant systems upgrades to obtain. Furthermore, we believe that the result will be of limited value to users, as similar information can be obtained from the disclosures surrounding credit quality and in the notes to debt instruments classified under FVOCI where a reconciliation to fair value is already required. Question 19: Do you believe that the implementation guidance and illustrative examples included in this proposed Update are sufficient? If not, what additional guidance or examples are needed? The examples are sufficient. Question 20: Do you agree with the transition provision in this proposed Update? If not, why? We generally agree that a cumulative effect adjustment to the statement of financial position as of the effective date is the appropriate transition approach for the proposed change in credit impairment. Question 21: Do you agree that early adoption should not be permitted? If not, why? We agree that early adoption should not be permitted. Question 22: Do you believe that the effective date should be the same for a public entity as it is for a nonpublic entity? If not, why? Given the significant resources and effort required to implement the proposed CECL model, we believe that a different effective date should be applied to a nonpublic entity from the effective date for a public entity. As it is permitted in the amendments of many other standards, a one year delay in the effective date would provide the necessary relief for a nonpublic entity to deploy resources and upgrade its systems prior to implementation of the new requirements. Question 23: Do you believe that the transition provision in this proposed Update is operable? If not, why? Subject to response to Question 22, it will be operable. 2012-260 Question 24: How much time would be needed to implement the proposed guidance? What type of system and process changes would be necessary to implement the proposed guidance? DB has a credit risk management system that captures the risk factors that we would require in the calculation of provision using expected loss. We are also able to adjust these input factors to consider average lifetimes and time value of money assumptions. We believe that the detailed disclosures that accompany this new standard should not be brought into effect before January 1 st, 2016. This will ensure that the expected loss information we disclose is reliable and is well supported by both qualitative disclosures and accurate risk factor data.