We globally agree with the three main categories of conditions for the rise by NCAs of risk weight and LGD minima, but wish to raise the following comments.
We would require the confirmation that NCAs mentioned in the RTS are national NCAs, even for institutions supervised at ECB level. It seems implicit, since national specificities have to be taken into account according to this draft RTS. But from this point of view it must be noticed that institutions having activities in different European countries could end up having to comply with different risk weight and LGD minima assessment processes, and with different risk weight and LGD minima.
We consider the categories of conditions specified in the RTS offer a too wide framework for NCAs risk weight and LGD minima assessment processes. The field of data to be potentially analyzed is very large, and their characteristics not defined precisely enough (historical series? Level of detail?). Then, considering the assessment processes could imply strong administrative burden and cost of adapting available databases or building new databases (including anticipation models), and that the impact it is expected to have at the end on the capital allocation inside the institutions are of major importance, we would call for precise governance frameworks for the processes, established between NCAs and the institutions of their jurisdictions. These governance frameworks for the assessment processes would locally determine precisely the data and ratios to be analyzed, and would define more precisely the rules according which the NCA would decide or not to raise risk weight and LGD minima. It would also determine in advance the conditions that would allow for a comeback to the initial levels. A transparent publication of these processes frameworks would be required.
Finally, and most importantly, residential loans are mostly secured in France by guarantees (“cautions”) and not by mortgages. This type of guarantee, that have proved to be at least as solid as mortgage, and that is recognized in EU regulation, should be permanently named and considered equivalent to mortgage in the European financial regulation in general, and in this RTS on NCAs risk weight and LGD minimal values assessment processes in particular.
The reports mentioned in article 2.1.a have been set up in January 2014. As they are semestrial reports, there has been three collections at that stage. Since it could become of major signification for NCAs to determine a possible rise in risk weights and LGD minimal values, it would be important that the regulator gives a feedback on the quality and significance of the data collected at that stage. What is the quality of the reports that have been collected? Are collected data considered relevant by the regulator? Is the perimeter of the data relevant or should it be modified? What historical series are available on the data collected through these reports? Are they sufficient?
Article 1(a) refers to determining loss experience relating to one or more property segment secured by immovable property. It seems that the segmentation between residential and commercial immovable properties is efficient enough. We consider it would be too complex and hardly efficient to look for loss experience data on too narrow segmentation (local markets, types of businesses…).
It would become difficult to handle for institutions if too fine data were required.
Moreover, some concerned institutions are larger banks subsidiaries. They may use the standardized approach whereas the mother bank uses the IRB approach. It seems important that the segmentation used in the assessment processes for risk weight and minima LGD are the same, or at least remain coherent.
Finally, it seems of major importance that the segmentation rules used in the assessment processes are coherent with the current work undertaken by the Basel committee on the revision of the standardized approach.
The data to be collected and analyzed for the assessment processes are widely defined in the draft RTS. From a quantitative point of view, it seems not sufficient to list the categories of data required for the determination of expected losses in the assessment process. It’s important that institutions are aware of the types of data that will be required, their level of detail. If required data are too detailed, it will raise the question of the ability of the institutions to deliver the data. It might oblige institutions to build models that at the end could be close to IRB models.
These questions raise the issue of more precise and transparent governance frameworks for the assessment processes, as exposed in Q1.
From a qualitative point of view, it seems that data used for the calculation of loss expectation should not rely only on too macro –economic considerations and figures on the one hand (article 2.1), and on single mathematical formulas on the other hand (article 2.3). It seems data should be better linked with the immovable property financing market and the life of institutions portfolios. For instance, account should be taken of historical and dynamic data on Loan To Value (plus prices actualization) and Debt to Income ratios series, in a long enough periods of time.
Finally, the choice of the data collected for the risk weight assessment process should also be coherent with the risk factors that will finally be determined when the current work on the revision of the Standardized (and IRB ?) approach is finalized.
The frequency of the risk weights assessment process is important to determine. CRR requires that it is at least once a year. It would mean that once risk weights have been raised, institutions would have to wait at least a year to see them de reduced back. On the opposite, intra-annual frequencies could become difficult to handle by smaller institutions.
These considerations also call for the establishment and the transparent publication of governance frameworks of the risk weight assessment processes between NCAs and institutions, as exposed in Q1.
It seems that the second argument, derived from the 0,3% level of losses mentioned in paragraph 3(a) in articles 125 and 126 CRR is not relevant since it has not been initially designed to set the level of loss at which risk weights are appropriate. There is no immediate rationale to assume that the 0,3% is a benchmark for increasing risk weight up to 100%. Moreover, the demonstration underlying this second argument seems biased since the note #5 in paragraph 1 page 24 mentions that 100% is the applicable risk weight for exposures fully secured by immovable properties, whereas it seems that the right percentage is rather 80%.
Therefore, for the identification of the levels of losses to which the risk weights of 35% and 50% for exposures secured by RRE and CRE are appropriate, we would consider the first argument more accurate. It leads to consider risk weights of 35% and 50% as generally sufficient for an average loss of 1,4% or 2%.
It would be important to confirm that these percentages have to be applied on the global portfolio, and not on a loan to loan basis.
The specifications do not raise any particular comment.
We globally agree with the “other conditions” for setting of higher risk weights. The provision by NCAs of explanations and assessment of procyclical effects seem essential. Yet, it implies the ability for institutions to “backtest” the data analyses of the NCAs, and so, that the NCAs processes are transparent and precise enough (property segments, detail level of indicators, historical series…). As exposed in Q1, this point calls for precise governance frameworks of the processes to be established between NCAs and the institutions of their jurisdictions.
We agree with the fact that it is not appropriate to set indicative benchmarks for the setting of higher minimum LGD values.