ESBG response: EBA consultation on PD, LGD, defaulted exposures

​Provides feedback on guidelines

​>> See the .pdf document


ESBG response to the EBA consultation on guidelines on PD estimation, LGD estimation and the treatment of defaulted exposures




ESBG (European Savings and Retail Banking Group)

Rue Marie-Thérèse, 11 - B-1000 Brussels

ESBG Transparency Register ID: 8765978796-80

​February 2017​

​>> See the .pdf document


Doc 0057/2017                                                                                   


Version 3


Dear Sir/Madam,

​Thank you for the opportunity to comment on the EBA consultation on guidelines on PD estimation, LGD estimation and the treatment of defaulted exposures. We would like to share with you the following reflections that we hope will be taken into account by the EBA.


Question 4.1: Do you agree with the proposed requirement with regard to the application of appropriate adjustments and margin of conservatism? Do you have any operational concern with respect to the proposed categorization?

Yes, we agree.

Furthermore, we believe that without guidance on principles for the quantification of the margin of conservatism (MoC), it is unlikely that there will be consistency across banks. However, we would also like to point out that quantifying and documenting the MoC requires great effort. In some cases, there remain doubts whether it will be feasible to carry out the modifications regarding the estimation of the LGD. In other cases, some information may not be available to institutions, be it that they do not receive it or be it that it does not exist, and burdensome deficiencies assessment, analysis and documentation are to be expected. Moreover, important IT developments are required. In short, there are some operational concerns.

Apart from that, the draft Guidelines require a specific margin for each deficiency as defined in the Article 30: "Any occurrence of any of the triggers referred to in paragraph 25 should result in the application of a margin of conservatism (MoC). Where more than one trigger occurs, a higher aggregate MoC should be applied […]". We think that it should be possible to apply one MoC if the identified deficiencies are related. In this case a separate evaluation of the MoC would lead to double cover of the deficiencies which are interconnected. It also might be difficult to find an appropriate methodology to estimate the impact of a particular deficiency separately from other deficiencies and other factors impacting the estimation. Hence, ESBG thinks that a common margin for related deficiencies should be allowed.

Considering all this, ESBG is not entirely convinced that the costs and benefits of quantifying and documenting the MoC are perfectly balanced.


Question 5.1: Do you see any operational limitations with respect to the monitoring requirement proposed in paragraph 53?

With regard to retail and SMEs, we don't see any operational limits. Regarding large corporates/low default portfolios (LDP), we don't see operational limits either, but detect obviously unstable measurements.


Question 5.2: Do you agree with the proposed policy for calculating observed average default rates? How do you treat short term contracts in this regard?

We agree.

One ESBG member indicates that, as of now, they use non-overlapping windows. Short term contracts are not included.


Question 5.3: Are the requirements on determining the relevant historical observation periods sufficiently clear? Which adjustments (downward or upward), and due to which reasons, are currently applied to the average of observed default rates in order to estimate the long-run average default rate? If possible, please order those adjustments by materiality in terms of RWA.

To us, they are not entirely clear. For instance, what is the share of downturn in the relevant observation period? From our point of view, a fixed share of downturn/upturn is necessary to obtain a stable long-run average default rate. Otherwise, institutions would be estimating a moving average with a high volatility.

Apart from that, ESBG thinks that a global definition of downturn period would be very helpful. As a consequence, the long-run average default rate would be comparable between institutions.

Furthermore, an ESBG member reports that even if observed defaults go back to mid-1990s, it is considered not to have a full economic cycle. They are therefore taught to include estimated default rates for the severe down turn in the beginning of the 1990s.

Moreover, clarification would be appreciated on the question of what is an appropriate mix of favourable and unfavourable economic conditions.

Finally, in case an institution needs to infer data from past periods, can macroeconomic indicators be taken into account to infer internal default rates?


Question 5.4: How do you take economic conditions into account in the design of your rating systems, in particular in terms of:

d) definition of risk drivers,

e) definition of the number of grades

f) definition of the long-run average of default rates?

d) One ESBG member states that financial ratios are used directly without adjusting them for current economic conditions, hence giving a point-in-time (PIT) effect. With regard to large corporates, they use non-financial information that tends to be through-the-cycle (TTC).

e) This is not applicable to one ESBG member. They use a common master scale (PD-bands into grades) across all their portfolios.

f) They are weighted between normal periods and downturn periods and almost always higher than the observed average default rates, as one ESBG member reports. They typically use 10-20% weight on severe downturn periods.


Question 5.5: Do you have processes in place to monitor the rating philosophy over time? If yes, please describe them.

One ESBG member points out that they partially carry it out in the validation processes.


​Question 5.6: Do you have different rating philosophy approaches to different types of exposures? If yes, please describe them.

Yes, mortgage and qualifying revolving retail exposure (QRRE) portfolios rely heavily on behaviour information, including current accounts and days past due.


​Question 5.7: Would you expect that benchmarks for number of pools and grades and maximum PD levels (e.g. for exposures that are not sensitive to the economic cycle) could reduce unjustified variability?



​Question 6.1: Do you agree with the proposed principles for the assessment of the representativeness of data?

We would like to point out the following three aspects, which could require further clarification:

  • Exclusion rules: on one hand, the draft Guidelines define that all data should be included in the Reference Data Set (RDS) with possible representativeness adjustment, e.g. in the Representativeness of data explanatory box: "As in accordance with Article 181(1)(a) of Regulation (EU) 575/2013 institutions are required to use all observed defaults it is not possible to remove the observations that are not fully representative from the estimation sample". In Article 21(c) a possibility to introduce some exclusions is introduced: "The rationale and scale of data exclusions broken down by reason for exclusion, using statistics of the share of total data covered by each exclusion, where certain data were excluded from the model development sample". The intention is not clear to us.
  • Outlier rules: in reference to the representativeness of data, the draft Guidelines might be understood as if no outlier rules are allowed: e.g. in the section on the representativeness of data: "As the purpose of the own funds requirements is to address the unexpected loss even if extreme events were observed in the past these should not be excluded from the estimation sample.". On the other hand Article 21(d) supposes the possibility of outlier rules: "The procedures for dealing with erroneous and missing data and treatment of outliers and categorical data, and the procedures for ensuring that, where there has been a change in the type of categorisation, this did not lead to decreased data quality or structural breaks in the data". It is not clear if outlier treatment is allowed and if some limits are imposed. This should, in ESBG's opinion, be clarified. Following the reasoning in Article 21(d), we suggest accepting the exclusion of some marginal amount of outliers.
  • Open defaults: it is required that all observed defaults have to be used for the purpose of LGD estimation, which indicates that also all defaults, where the recovery process is not completed ('open defaults') have to be used, independent of the time in default. As an accurate estimation of future recoveries is hardly possible at the beginning of default, because the workout strategy is mostly fixed after a certain observation period, we propose allowing the option to exclude open defaults with a duration shorter than a reasonable period.


Question 6.2: Do you agree with the proposed treatment of additional drawings after default and interest and fees capitalised after the moment of default in the calculation of realised LGDs?

ESBG believes that benchmarks may not be representative if a different approach is optional when it comes to additional drawings. LGD and EAD should be viewed together for benchmarking purposes if this optionality remains.

​Question 6.3: Do you agree with the proposed specification of discounting rate? Do you agree with the proposed level of the add-on over risk-free rate? Do you think that the value of the add-on could be differentiated by predefined categories? If so, which categories would you suggest?

While we agree to a certain extent with regard to the discounting rate, we don't agree in respect of the risk-free rate.

ESBG would appreciate if the EBA could reconsider the use of a discounting rate of interbank funding rates and a 5% add-on, which seems excessive to us. While it might be simple and contribute to increased comparability, the use of this discounting rate is risk insensitive and would punish an institute with a low risk profile (and would favour institutes with a high risk profile). Moreover, this approach stands in contrast to the IFRS 9. ESBG is of the opinion that an alignment towards IFRS 9 would be preferable.

As a practical alternative, one could also consider using the original effective interest rate as prescribed by IFRS 9 instead, which reflects to some extent different risk characteristics. We understand that it also does not properly fit into the economic loss concept, but it would effectively reduce unintended variability of RWAs while being a pragmatic solution.

In ESBG's view, it would also be useful to specify which situations/costs are supposed to be covered by the risk-free rate and the add-on in order to avoid the inclusion of the same costs in the model twice (e.g. refinancing costs).

Furthermore, it could, in our opinion, be problematic to have the same discount level for all countries with a single currency (e.g. Eurozone countries). Potential differences might be introduced via add-ons.

Yes, the value of the add-on could be differentiated by predefined categories. It would be appreciated if institutions could continue using the ones that they currently use.  Distinguishing between, for instance, corporates, SMEs, mortgages, QRRE and "other" makes sense too.


Question 6.4: Do you agree with the proposed approach with regard to the specification of historical observation period for LGD estimation?



Question 6.5: Do you agree with the proposed treatment of incomplete recovery processes in obtaining the long-run average LGD?

Yes. One ESBG member states that they use actual provisions as a best estimate of future recoveries and believe this is a better estimate for future recoveries than recoveries from observed closed cases.

On a general note, we would like to point out the following: if we understand correctly, for the purpose of the future recoveries and costs estimation, banks should use both closed exposures and incomplete recoveries with factually observed costs and recoveries.

Let us consider a portfolio with 2 closed defaults and 2 incomplete recovery processes:




Months in default












Banks shall construct a recovery pattern for each month in default (M). Thus they would calculate average monthly recovery for all M≤3 using the average monthly recovery based on all defaults. The recovery pattern for M=4 will be based on 3 defaults – C1, C2 and I1 and it would be used to estimate the additional recovery of default I2 in M=4. The recovery pattern for all M>4 would be based only on closed defaults C1 and C2 and this pattern would be applied to defaults I1 and I2 to estimate future monthly recoveries for M>4.

If our understanding is correct, we would agree with the proposed treatment as long as the time from default is shorter than the maximum period. Otherwise, ESBG would like to ask the EBA for a more detailed explanation on the usage of closed and incomplete recoveries.

In cases where the time from the moment of default is longer than the maximum period, we do not understand why future collateral recoveries should not be estimated. In cases where a periodical re-evaluation of the collateral value is required we propose, as banks should have sound collateral management procedures in place, to allow estimating future recoveries stemming from the realisation of the existing collateral as long as the actual evaluation results in a collateral value > 0.


Question 6.6: Do you agree with the proposed principles on the treatment of collaterals in the LGD estimation?

Since collaterals have an important impact in the recovery process, and therefore also on the realised LGD, we would like to state that we support the idea of the inclusion of collaterals in the LGD estimation. Nevertheless it should be taken into account that a part of the collaterals treatment is manual and therefore it can be impacted by imperfections in the data treatment. Historical data is even more inaccurate as current requirements were not fully applicable for the historical data. Use of a common recovery rate for both collateral and non-collateral payments should be allowed in case this is more appropriate due to the abovementioned reasons, even if these curves should be penalised by applying a MoC. Otherwise, a more sophisticated model does not necessarily lead to a more precise LGD estimation. If separate recovery rates are required, the institutions should be given sufficient additional time to assure a sufficient proportion of good quality data in the RDS before using separate collateral and non-collateral recovery rates.

Apart from that, ESBG would like to point out that the strict requirement of (regulatory) eligibility of collateral to allow market values to be used as input variables has to be seen as critical from a methodological perspective. Firstly, although this condition might only impact a minor portfolio in typical developed portfolios, it might be significant in other areas. And secondly, a forced consideration of these cash flows in unsecured recovery rates will have to be offset by additional methodological approaches (e.g. monitoring of the stability of this portfolio, introduction of specific sub-segments to properly reflect these cash flows etc.) to properly reflect the economic value of these contracts. We believe this regulatory restriction increases model complexity and might even add to RWA variability by distorting the economic value of non-eligible collateral. Consequently this restriction should be lifted for IRB LGD and ELBE models, supported by proper estimation and validation of the (economic) value of any (relevant) collateral.


Question 6.7: Do you agree with the proposed treatment of repossessions of collaterals? Do you think that the value of recovery should be updated in the RDS after the final sale of the repossessed collateral?

Yes, to a large extent we agree with the proposed treatment of repossessions of collaterals.

However, we are a bit concerned about Article 152. In case that the value of the repossessed collateral is adjusted by haircuts to reflect the potential price that could be achieved from such a sale, the uncertainty would be double-booked: firstly, in the LGD model of the institution and, secondly, on the books of the institution which repossesses the collateral in a form of investment risk.

Moreover, we would actually prefer that the value is not updated in the RDS after the final sale of the collateral.


Question 6.8: Do you think that additional guidance is necessary with regard to specification of the downturn adjustment? If yes, what would be your proposed approach?

Yes, maybe at component level, i.e. cure rates, recovery levels and workout period.


Question 7.1: Do you agree with the proposed approach to the ELBE and LGD in-default specification? Do you have any operational concerns with respect to these requirements? Do you think there are any further specificities of ELBE and LGD in-default that are not covered in this chapter?

Yes, we agree. Our main concern refers to ELBE, which is very challenging. In our view, LGD in-default should be an add-on to ELBE and depend on the time horizon on the recovery.


Question 7.2: Do you agree with the proposed reference date definition? Do you currently use the reference date approach in your ELBE and LGD in-default estimation?

Yes, we agree. One ESBG member points out that they do not use reference dates currently. Another ESBG member explains that they currently use a single reference date in their models: the time since default began.


Question 7.3: Do you agree with the proposed approach with regard to the treatment of incomplete recovery processes for the purpose of estimating LGD in-default and ELBE?

As indicated above, if we understand correctly, for the purpose of the future recoveries and costs estimation, banks should use both closed exposures and incomplete recoveries with factually observed costs and recoveries.

So the same example that has been used in answer 6.5 could be considered: a portfolio with 2 closed defaults and 2 incomplete recovery processes. Only one reference date – months since default – is used.


DefaultStatusMonths in default


Banks shall construct a recovery pattern for each reference date: month in default (M). All reference dates M<4 would use all defaults to set up recovery patterns as described in 6.5. Recovery pattern for the reference date M=4 can be based only on 3 defaults – C1, C2 and I1. Recovery patterns for all reference dates M>4 would be based only on closed defaults C1 and C2. Future recoveries and costs of incomplete recovery processes would be estimated based on these recovery patterns.

If our understanding is correct, we would agree with the proposed treatment. Otherwise, we would like to ask the EBA again for a more detailed explanation on the usage of closed and incomplete recoveries.


Question 7.4: Which approach do you use to reflect current economic circumstances for ELBE estimation purposes?

One ESBG member explains that ELBE is set as being equal to the individual credit risk adjustments (provisions).

Another ESBG member states that, in the estimation of ELBE, they use the observed recoveries of closed and currently incomplete exposures with the estimated future recoveries. As a result, the average of the ELBE tends to fluctuate with the economic cycle as the incomplete recovery processes reflect the current economic conditions (an analysis had been produced that confirmed this).


Question 7.5: Do you currently use specific credit risk adjustments as ELBE estimate or as a possible reason for overriding the ELBE estimates? If so how?

Within our membership, there are banks that directly use provisions as ELBE.


Question 8.1: Do you see operational issues with respect to the proposed requirements for additional conservatism in the application of risk parameter estimates?



​Question 9.1: Do you agree with the proposed principles for the annual review of risk parameters?

Yes, we agree. In addition, we would like to refer to our comments on question 4.1., where we hinted at operational concerns and possibly burdensome workloads that one could expect.


​Question 10.1: Do you agree with the clarifications proposed in the guidelines with regard to the calculation of IRB shortfall or excess?

Yes, we agree.


​Question 11.1: How material would be in your view the impact of the proposed guidelines on your rating systems? How many of your models do you expect to require material changes that will have to be approved by the competent authority?

Very material, as a matter of fact. One ESBG member explains that a large portion of models would require material changes, particularly in respect of the LGD modelling.


Additional comments

ESBG would like to highlight another topic which is not directly addressed in the questions: multiple defaults treatment (Article 90). The intention why multiple defaults should be treated in a different way compared to PD or CCF parameters is not clear to us.

  • Merging of subsequent defaults into one default is related to the probation period of the default definition, which should already include sufficient reserve to cover the uncertainty about the 'final cure'. The probation period within the default definition should be adapted to the bank´s situation and it should avoid high level of multiple defaults without any additional period introduced during risk parameters estimation (as required by the Article 90). The additional period seems to be appropriate if the default definition shows some deficiencies (e.g. no probation period in the past). We propose to use an additional period only in case of a deficiency in the default definition.

  • In case the abovementioned opinion was not accepted, we would propose to define a unique period – an addition of the probation period within default definition and an additional period in the moment of risk parameters estimation. We propose a shorter period than the current proposal (3+12 months), e.g. 6 months. This approach of a unique period would allow each institution to decide on which level it would apply the period: default definition or risk parameters estimation.

  • Whatever defaults merging period is selected, all risk parameters should be based on the same set of defaults as defined in Article 52 of the RTS on the assessment methodology which requires treating multiple defaults consistently for estimation of CCF, PD and LGD parameters. The principle of treatment of multiple defaults should be specified in the general part of the Guidelines (chapter 4), in our view, and not in the LGD part (chapter 6) to assure the consistency of treatment.


​>> See the .pdf version of this document

About ESBG (European Savings and Retail Banking Group)

ESBG brings together nearly 1000 savings and retail banks in 20 European countries that believe in a common identity for European policies. ESBG members represent one of the largest European retail banking networks, comprising one-third of the retail banking market in Europe, with 190 million customers, more than 60,000 outlets, total assets of €7.1 trillion, non-bank deposits of €3.5 trillion, and non-bank loans of €3.7 trillion. ESBG members come together to agree on and promote common positions on relevant regulatory or supervisory matters.

​European Savings and Retail Banking Group – aisbl

Rue Marie-Thérèse, 11 ■ B-1000 Brussels ■ Tel: +32 2 211 11 11 ■ Fax: +32 2 211 11 99 ■



Published by ESBG. February 2017.


​>> See the .pdf version of this document


European Institutions; European Supervisory Authorities (EBA-ESMA); Risk assessment training and procedures; Risk management; Capital Markets; Capital Requirements Directive and Regulation; SME finance