Finding Aggregate Data on Banks’ Credit Risk Modeling Approaches
Summary
The document asks whether data are available in aggregate on European banks’ use of credit risk approaches for regulatory capital. It describes the distinction between the Foundation and Advanced Internal Ratings-Based approaches, the additional role of slotting for specialised lending, and the Standardised Approach. The proposed research question is how corporate exposures are distributed across these methods, rather than simply how individual banks disclose their practices.
The answer points to European Banking Authority credit risk benchmarking exercises covering high-default and low-default portfolios. It cites a report that includes a summary table and says that 114 of 117 institutions permitted to use credit models submitted data. It also notes that institutions may appear under different approaches in different countries and that not every institution has exposures in every asset class. These details help interpret sample counts, but the document does not reproduce aggregate shares or fully establish representativeness; readers must consult the reports for the underlying figures and definitions.
Key ideas
- The EBA’s high-default and low-default portfolio benchmarking exercises are potential sources for aggregate credit risk data.
- The cited survey reports submissions from 114 of 117 institutions with permission to use credit models.
- Institution counts can vary by asset class because banks do not all have exposures in every sample.
- A single institution may use different capital approaches across countries, so reported counts may not sum as expected.
- The document identifies a data source but does not provide the requested aggregate exposure shares.
Tags
Full text
# Dominating credit risk modeling approaches for capital calculation in banks # Dominating credit risk modeling approaches for capital calculation in banks In Basel/CRR (capital requirement regulation) there are various approaches for the estimation of capital requirements. For corporate exposures there is the Foundations IRB approach (F-IRBA, own estimates of PD) and the Advanced IRBA (A-IRBA) with own estimates of loss-given-default and conversion factors. For the exposures to specialised lending (e.g. project finance) a so called slotting approach exists in addition. These approaches have increasing demands for model sophistication and especially data availability. The approaches used for capital calculations are reported in the disclosure reports for each bank. my question: are such numbers available in an aggregate basis somewhere? I would be interested in questions like "What is the share of Corporate exposure in F-IRBA at European banks as opposed to Standardized Approach and A-IRBA?". Is such data collected and made available somewhere? PS: I am aware that such questions are not 100% on-topic for a quant site but I made the experience that quants should know how the peer group looks. ## Answer by Magic is in the chain (score 3, accepted) https://quant.stackexchange.com/a/46453 The EBA performs the HDP (high default portfolio) and LDP (low default portfolio) benchmarking exercises, which would be relevant. You can find it on their website. Here are a couple of examples: https://eba.europa.eu/documents/10180/2087449/EBA+Report+results+from+the+2018+Credit+Risk+Benchmarking+Report.pdf https://eba.europa.eu/documents/10180/15947/EBA+Report+results+from+the+2016+high+default+portfolio+exercise+-+March+2017.pdf Re-comment, I think it is clearer in the summary table in the latest survey which includes both HDP and LDP (https://eba.europa.eu/documents/10180/2087449/EBA+Report+results+from+the+2018+Credit+Risk+Benchmarking+Report.pdf): The intro says 117 institutions have permission for credit models, of which 114 submitted the data, so I think the sample is representative and they have even explained why the 3 did not. The reason the number of institutions differ by asset class is because not all of them would have exposure to all classes/samples. And the rows don’t add up to the total institutions could be because of footnote 41, same institution using different approaches in different countries. The report also claims the results are stable compared to previous survey, implying quality is ok I suppose.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.