Response to addendum to Joint Consultation on draft Implementing Technical Standards on the mapping of ECAIs’ credit assessments

Go back

Q2. Do you agree with the proposed definition of sufficient number of credit ratings and rest of requirements imposed to the calculation of the short run default rate when a sufficient number of credit ratings is available?

We understand the problem to define the notion of “sufficiency” and the difficulty to be more precise in the proposed regulation. However, we would like to comment the following points:

- Article 3 (5)
“Items withdrawn prior to the end of the time horizon and not defaulted shall only contribute to the denominator of the short run default rates mentioned in point (a) of paragraph 3 with a weight equal to 50%. Any item for which there is evidence that it has been withdrawn prior to the occurrence of default shall be considered as defaulted item. “

When the item is withdrawn but remains under review so that the ECAI is able to know whether it has defaulted or not within the time horizon concerned, there is no reason to apply such a 50% rule. Therefore it is proposed to introduce an exception to article 3 (5), at the end of the first sentence, that would state “except if it has been verified by the ECAI that the item is not defaulted in the period considered”.

- Article 3 (6):

Regarding the definition of default, we are surprised that the definition set forth in EU regulation No 575/2013 is not referred to, at least as a basis. Taking into account the importance of the default definition for the mapping procedure, it would be wise to perform some tests to ensure the consistency of the ITS default definition towards the default set in CRR and used by IRBs for the identification of default items.

Moreover the explanatory text of article 3.6 states: Types of default events not included in this list should not be used for the default rate calculation and, where not all types of default events have been considered by an ECAI, the qualitative factor specified in Article 9 should be applied to the calculated default rate in order to increase the consistency of the mapping process. This formulation does not seem to leave any leeway for ECAIs that do not consider all the listed types of event, even if they use other events, in particular because of the characteristics of the companies they rate (e.g. SMEs): their default rates cannot be considered as a robust basis for the quantitative analysis. The qualitative factor then apply, which systematically results in considering its default rate as underestimated. Therefore, to fully take into account the situation of ECAI that have adopted other types of definition, in particular if it may be considered as being in line with the CRR definition, it would be rational to introduce more flexibility as from article 3.6.

Regarding the events listed in article 3 (6), we also consider that the notions used should be more precisely defined, in particular:
(b) a missed or delayed disbursement of a contractually required interest or principal payment, unless payments are made within a contractually allowed grace period;
Does it apply both to financial obligations and securities? For financial obligations, as the information is clearly linked to the appreciation of the credit institution concerned regarding the existence and the materiality of a missed or delayed disbursement, it would be sound andconsistent to refer to past due obligor in the same terms as in the CCR definition and to respect the same 3 months period (this 3 month period resulting of a study that had been led by the Basel Committee to take into account bank practices and reduce the cases of technical defaults).

(c) a distressed exchange if the offer implies the investor will receive less value than the promise of the original securities;
We consider that this event should at least not be given the same importance as the others, as it only relates to securities andas it may not concern certain categories of companies. Moreover, such an event is rather linked to LGD than to PD. Therefore, it does not seem relevant to include it in the events that shall be considered by any ECAI. Incidentally, a precise definition of a distressed exchange should be added.

(d) the rated entity is under a significant form of regulatory supervision owing to its financial condition.
More explanations or examples could be given to characterize this sort of event. Otherwise there might be large differences in the interpretation of the type of regulatory supervision concerned.


- Additional comments on recital 5:
Recital 5: This Regulation takes into account that the default definitions used by ECAIs differ, as reflected in Regulation (EC) No 1060/2009 and Commission Delegated Regulation (EU) No 448/2012. In order to ensure that the overall level of the capital required to externally rated exposures is not changed, the types of default events used for the calibration of the benchmark referred to in point (c) of Article 136(2) CRR constitute the default definition proposed for this Regulation.

For the purpose of harmonization, it would be more relevant to stress that the aim should be to approach the CRR definition of default but that a proxy might be used in order to ensure that the overall level of the capital required to externally rated exposures is not changed.

Therefore we propose the following amendments to recital 5:

This Regulation takes into account that the default definitions used by ECAIs may differ from the one set forth in the regulation EU 575/2013, as reflected in Regulation (EC) No 1060/2009 and Commission Delegated Regulation (EU) No 448/2012, even if they shall remain consistent to it . In order to ensure that the overall level of the capital required for externally rated exposures is not changed, the default definition used by ECAI will be considered in the range of the qualitative factors and, if it differs from the CRR definition, the types of default events used for the calibration of the benchmark referred to in point (c) of Article 136(2) CRR will constitute criteria for determining the mapping of credit assessments.

Q3. Do you agree with the proposed requirements imposed to the calculation of the long run default rate when a sufficient number of credit ratings is available?

We globally agree with the process. However we do not fully understand the meaning of some proposals

Article 4 (3):
“…and the remaining short run default rates have been estimated”
How should the remaining short run defaults be estimated? Is the ECAI allowed to use all available data or are there specific rules to follow?

article 4 (4):
“For the purpose of producing the weighted average referred to in paragraphs 2 and 3, the short run default rates calculated according to Article 3 shall refer to the most recent recessionary period. “
Does this paragraph mean that only default rates related to a recessionary period should be taken into account for the calculation of the weighted average? Or does it mean that they should be given more weight, and in this case what is the rule that should be applied?

We consider that such a rule is mainly relevant for the calculation of the long run default rate when less than 20 short run default rates are available. However if an ECAI provides a calculation on more than 20 short run default rates, focusing on the recessionary periods would rather be contradictory with the fact that the weights should ensure an adequate representation of recessionary and non recessionary years in a full cycle. Moreover default rates may be particularly volatile in recession periods.

Therefore we propose the following amendment to article 4 (4):
For the purpose of producing the weighted average referred to in paragraphs2 and 3, the short run default rates calculated according to Article 3 shall refer to the most recent recessionary period.

Q4. Do you agree with the proposed options to calculate the quantitative factors when a sufficient number of credit ratings is not available?

No specific comment

Q5. Do you agree with the proposed use of the default definition use by the ECAI as a relevant factor for the mapping? Do you agree with the proposed assessment of the comparability of the default definition of an ECAI? If not, what alternatives would you propose? In particular, what are your views on the introduction of a specific benchmark for ECAIs relying on the regulatory definition of default and/or for ECAIs providing ratings for companies that do not rely on market financing? Could it be a robust alternative to the current benchmark? Do you think that the adjustment factor depends on certain characteristics of the rated firms such as size and credit quality and if so, how can this be reflected?

The Banque de France strongly supports the intention of the Joint Committee to conduct further analyses in order to determine whether additional specific benchmarks based on the CRR default definition and a large and diverse set of exposures could be developed.

The Banque de France agrees with the fact that the default definition used by the ECAI is a relevant factor for the mapping. However, the methodology explained in article 9 does not sound relevant for all ECAI and for the rating of SMEs:

First, the draft ITS is based on a default definition that largely focuses on the debt securities of the assessed company whereas SMEs generally do not issue that type of assets. The defaults that credit institutions may observe regarding SMEs mainly concern their loans and are governed by the definition set forth in the EU regulation No 575/2013, which is not referred to in the draft ITS. Taking into account the purpose of ECAIs’ credit assessment for credit institutions, it would have been consistent to start from the same definition as the one assigned to IRBs. For ECAI using other type of definition, the same sort of process as the one used for IRBs might have been applied apply (cf. article 175 CRR: “The institutions shall document the specific definitions of default and loss used internally and ensure consistency with the definitions set out in this Regulation”):

In this regard the following amendments have already been proposed to article 9:
ECAI shall document the specific definitions of default used internally and ensure consistency with the definitions set out in Regulation No 575/2013. When using a definition diverging from this latter one, the following types of events should be included when applicable in the definition of default:

Second, the explanatory text underlines the problem of using the benchmark default events considering that “it is not always possible to identify defaults as a consequence of missed payments of principal/interest of bonds traded in the market”. The same applies for “distressed exchange”. Moreover, the CRR default definition, there is a rule which is strictly respected by credit institutions regarding the delay of 3 month before registering a default event. In the benchmark used, a default occurs within 5 days of non reimbursement. We could add that the ITS default definition does not take into account the “unlikeness to pay” which is yet a decisive criterion for banks when registering defaults on bank loans. Therefore the Banque de France considers it is not possible to identify the default events of SMEs on the basis of the ITS definition. In this regard, assessing whether the benchmark default definition is stricter than the one used by ECAI focusing on SMEs does not make sense as long as the benchmark definition simply does not seem appropriate in the case of SMEs.

Third, the ITS proposes that the strictness of a default definition is measured as the additional number defaults recorded by the ECAI due to a reason different from bankruptcy. This might have been an interesting proposal (even if it largely bases on domestic notions of “bankruptcy whereas significant differences may arise across countries in its definition). However the proposed 100% adjustment factor is based on information used to calibrate the benchmark (see the previous point). This is to say it compares the proportion of bankruptcy events of bigger companies to the proportion of their other default events, simply assuming that:
- this proportionwould be the same for SMEs,
- the benchmark default data might be used as a comparison point whereas the explanatory text states this type of comparison might not be possible for some ECAI due to the type of firms they rate.

Fourth, having collected CRR defaults from credit institutions since 2012, the Banque de France considers that the adjustment factor differs according to the characteristics of the rated firms. Therefore, its opinion is that an ECAI that mainly assesses the credit quality of smaller firms might be applied a different scaling factor from the one resulting from the benchmark which focuses on bigger firms. In order to correctly reflect the characteristics of the companies rated in the mapping, it seems important to take into account the data that might have been collected by the ECAI concerned as a reference to assess the relevance of its default track, even if its definition of default events differs from the one proposed, as long as it remains in line with the CRR definition.

In particular, what are your views on the introduction of a specific benchmark for ECAIs relying on the regulatory definition of default and/or for ECAIs providing ratings for companies that do not rely on market financing?

Of course, we agree on the fact that it is important to implement a consistent based approach, in line with international standards, as soon as those standards do exist for all types of items assessed. We also recognize that the benchmark mentioned in the draft ITS has so far been used as common references to set CQS values. However it is not intuitive and obvious to keep using the default rates of international ECAIs observed more than 10 years ago on a specific population of rather large and international listed companies as a benchmark for determining the mapping of any ECAI in particular if it operates on a domestic market regarding companies that are mainly unlisted SMEs and whose defaults are registered by banks on the basis of the CRR definition.

In particular, using default definition and rates largely focused on the debt securities of the assessed company in order to produce mappings that in certain cases mainly concern companies that do not issue this type of market instruments does not appear adequate. Moreover, the importance and intangibility given to the lower and upper bound set in the draft ITS lead to reduce the range of values that can be compatible with each credit quality step and finally to consider them as the sole possible standard.

Our opinion is that a correct estimate of corporate credit risks, and more particularly SMEs credit risks, requires remaining close to the banks’ lending practices when possible. Therefore for ECAI that could either collect default data according to the CRR definition or that rate a large and significant population of SMEs, a way should be managed to elaborate more relevant references. At least, a study should be conducted to examine the question of the relevance of the methodology and the reference used in those both cases.

An alternative benchmark based on the CRR default definition could be a solution to better address the SME dimension in the draft ITS and the mapping process. These data would allow completing the ones that have been considered as sole comparison points for all mappings so far. Consequently it could lead to adapt the CQS values and/or bounds taken into account so that they may be adapted to a larger number of firms, mainly SMEs, exposed to a credit risk of default.

Default data that NCBs which are ICAS (In-house credit assessment system) have been collected since 2012 appear as a new source of information that should allow building this alternative benchmark in a near future in order to have complementary data on the credit risks of SMEs in particular. The NCBs indeed use a harmonized definition of default, in line with the CRR definition, focus on corporate credit claims, collect credit and default data from all credit institutions established in their country and assess a larger population of companies.

ECAIs focusing on companies, in particular SMEs, using bank loans rather than market based financing do not collect the Basel defaults from banks as NCBs do. However they neither collect the same default events as the ones referred to in the draft ITS. Therefore it could appear appropriate to compare the default rates of CRAs focusing on SMEs with the ones observed on those companies rather than with the ones currently used as references.

However this process would need a deeper analysis and a longer set of defaults than the one currently available (since January 2012). Therefore, as long as the relevance of the references currently used is questioned for the mapping of some ECAI, it would be wise to postpone their re-mapping exercise. Until then, the current mapping could persist for the ECAI concerned by this specific issue.

Q6. Do you agree with the proposed use of the time horizon of the rating category as a relevant factor for the mapping? Do you agree with the proposed use of transition probabilities to identify the expected level of risk during the 3-year time horizon?

We agree with the proposed article 10

Q7. Do you agree with the proposed use of the range and meaning of rating categories as a relevant factor for the mapping? Do you agree with the proposed restriction of this factor to adjacent rating categories?

We agree with these proposals.

Q8. Do you agree with the proposed use of the risk profile of a rating category as a relevant factor for the mapping?

We strongly disagree with the proposed article 12
The creditworthiness of items assigned the same rating category shall be determined by considering, at least, their size and the degree of sector and/or geographical diversification of their business activity.
…and the example provided in the explanatory text:
For example, a rating category composed mainly of small and non-diversified firms will very likely not qualify for CQS1, which underlying level of risk is representative of multinational firms present in various continents/sectors and therefore with a high resilience to adverse shocks. This does not mean that none of these firms could ever qualify for such low-risk CQS. It means, however, that unless contradicted by other relevant information, these general risk factors suggest that the mapping of such a low-risk CQS should not be proposed.

The size and diversification are criteria that should normally be taken into account by the ECAI when assessing the credit quality of a company. Assessing the resilience to adverse shocks is also the role of the ECAI. Therefore it is not relevant to introduce an article that leads to exclude some companies from the best rating categories out of hand, because e.g. they would be considered as too small. It would also de facto tend to penalize the few ECAIs which assess SMEs credit quality, even if they have developed a long experience in this field, their rating categories being by nature mainly composed of small firms.

Regarding diversification, we consider that an insufficient diversification of items in a rating category, especially if the number of items is low, might be an issue as there is a risk of contagion that should be consider when determining the adequate credit quality step. However, in this specific case, contagion risk may exist for smaller as well as for bigger firms.

Finally, we strongly recommend that the regulation do not lead to penalize small firms, especially in a context where at the European Commission level as at the national one a lot of initiatives have been taken to facilitate their access to credit and financial market.

Amendments proposed to article 12:
When there is not a sufficient number of items in a rating category, the degree of sector and/or geographical diversification of the business activity of those items shall be considered to determine the degree of risk of the rating category.

Q9. Do you agree with the proposed use of the estimate provided by the ECAI of the long run default rate associated with all items assigned the same rating category as a relevant factor for the mapping? Do you agree with the proposed role played by this factor depending on the availability of default data for the rating category?

We agree with this

Q10. Do you agree with the proposed use of the internal mapping of a rating category established by the ECAI?

No specific comments on this point.

Q11. Do you agree with the proposed specification of the long run and short run benchmarks? In particular, do you think that different benchmarks should be introduced to fully take into account the specific situation of ECAIs relying on the regulatory definition of default and/or of ECAIs providing ratings for companies that do not rely on market financing, including in particular SMEs? Do you agree with the proposed mechanism to identify a weakening of assessment standards?

An automatic mechanism to identify a weakening of assessment standards appears as an interesting tool. However the methodology used for the automatic identification being based on the calculation of a confidence interval per se penalizes ECAI that assess a lot of companies (N). The bigger is N, the higher is the precision of the PD (as the interval is smaller), which is fine. By definition, a smaller interval, especially its lower limit, will more likely differ from the benchmark than a larger one. The risk is aggravated if there are discrepancies between period and economic situation compared when assessing the ECAI concerned to the benchmark.

However, we consider that it would be useful to provide more specification for the benchmark, in particular, regarding the rules followed to calibrate the benchmark. It also seems relevant to specify criteria regarding the sufficiency of ECAI providing data and data collected (items and default rates) as well as regarding rules to be used when those data appear insufficient according to the aforesaid criteria. It would also be interested to explain how divergent data are treated.

Moreover the methodology used for the automatic identification being based on the calculation of a confidence interval penalizes ECAI that assess a lot of companies (N). Indeed, the bigger is N, the higher is the precision of the PD (as the interval is smaller), which is fine. However, by definition, a smaller interval, especially its lower limit, will more likely differ from the benchmark than a larger one. The risk is aggravated if there are discrepancies between the short-term PD estimation period and economic situation compared when assessing the ECAI in comparison with the benchmark. Therefore, we rather suggest that the assessment is conducted on the basis of a comparison between the long term PD for the benchmark and the short term PD for the ECAI concerned. Such an alert procedure would thus allow showing the difficulties of the ECAI’s assessments to follow the trend imposed by the benchmark (the trend is seen on long-term PD not on short-term PD).

Finally to correctly apply the re-mapping methodology, it would be relevant to guarantee the representativeness of the benchmark regarding the ECAI otherwise the mapping procedure would be groundless. To achieve this goal, some correlation tests between benchmark and ECAI (e.g. Khi-2 tests on default definitions, sectors, turnover…) could be done.

Q12. Do you agree with our analysis of the impact of the proposals in this CP? If not, can you provide any evidence or data that would explain why you disagree or which might further inform our analysis of the likely impacts of the proposals?

In the indirect costs (§39) of the initial consultation paper, it is mentioned that:
“The costs, defined in terms of capital requirements under the Standardised Approach, derived from the specification of new individual mapping tables will be entirely met by the credit institutions.”
However, should the credit institutions have to bear additional costs for their lending activities on account of mapping changes (if those changes lead to higher capital requirements), there might be indirect consequences on access to credit (such as an increase in borrowing costs or a drop in lending). Therefore, in this case, the costs would not be entirely met by credit institutions and a possible impact for companies could also be mentioned in the analysis of the impact of the proposals.

Our own estimate of the impact of the proposals of the consultation paper differs from the one presented in figure 1 of section 4 of the addendum to CP (effect estimated at about 8%). Based on the September 2014 credit risk data of our credit register and the distribution of ratings 3++ to 9 among corporates at the same date, the effect of the new Banque de France mapping would be an increase of 15% of the average amount of corporate weighted assets.

Furthermore, the proposed new mapping of Banque de France would trigger the application, under the standard approach for RWA, of a 100% risk weight, instead of the current 50% risk weight, on more than 100 GEUR of bank commitments, mainly towards SMEs (75% of the entities that would be remapped from CQS2 to CQS3 having a yearly turnover below 7,5 MEUR). More generally, we also observe that, in the cases where a mapping was already available, the increase in capital requirements only affects ECAI that rate a significant amount of SMEs.

Upload files

Name of organisation

Banque de France