Internal Risk Rating Development Questions
-
We are building internal Risk Rating models for the C&I and CRE portfolios of a bank that has recently undergone a large expansion. Complications in this effort arise due to the quality of data across the merged banks, inconsistency of variable tracking, and extraneous definitions of default
To that end, we have two pressing questions:
-
To identify “bads” across the combined bank, we have been using the Basel default definition to ensure consistency. However, the banks have intentionally identified many loans that fit the Basel definition as not-in-default because they consider them “administrative defaults”. Is there any precedent in the exclusion of “administrative defaults” in development of Risk Rating models? How do regulators react to this type of treatment?
-
When developing internal model segmentation for the C&I portfolio, the bank has historically segmented along the dimensions of size (defined by Net Sales) and industry (defined by NAICS code). Aside from size and industry, are there any further dimensions that are considered fundamental segmentation criteria in C&I or CRE model development?
-
-
On the point of “administrative” defaults, I haven’t seen the specific regulatory reactions but I will say I’ve seen a client who realized their historical data had a very conceptually reasonable default definition but a certain number of loans were being flagged as meeting one of the default conditions in only 1 or 2 monthly snapshots, and then returning to performing status in the following month, never to return to a default condition again after that
In this case, I think they came up with a reasonable logic to exclude those defaults that had less than 3 months in which they were flagged as default (including nonconsecutive months), and never charged off. Given the way this definition was constructed, it clearly had no impact on the ability to capture actual charge-offs, and spot checking of the individual cases removed generally indicated these were various forms of administrative errors, including purely internal accounting errors unrelated to the client behavior
-
In CRE portfolios, the most common dimensions for segmentation that I have seen are:
- Income-producing vs CLD (Construction & Land Development)
- RE type: multi-family, office, retail, hotel, warehouse, healthcare, other
- Urban vs Suburban vs Rural
- Geographical regions, states
- LTV buckets
- Buckets of Capitalization Ratio
-
On 1: Do you essentially mean technical defaults? If so then I think they are, in fact, generally excluded for IRB modelling purposes, at least in PRA/ ECB regulated banks where I’m more familiar with. But regulator dependent. e.g., see EBA’s Guidelines on the application of the definition of default, Section 2.2.2:
" § The required payment has been made by the obligor before the relevant days past due criterion, including the materiality threshold, has been breached but default has been identified as a result of a long payment allocation process within the institution.”
On 2: A key dimension that regulators look out for in particular is geography (e.g., country of domicile / risk, or region), and that is also one that we usually also segment by for representativeness purposes
-
we have been helping some EU players in their defaults
exclusions. Some key take aways from the process are:- You need to have a clear definition of default criteria as starting point
- You need to identify the clear motives for the exclusions, and provide evidence for each group of exclusions (e.g. proof of payment, proof of misclassification by a manager, technical errors in the default implementation, etc.)
- 2nd line of defense typically review the default exclusions rationale and proofs to ensure compliance
We have done this across low default portfolios like C&I or Project Finance.
Regarding risk drivers:
- For C&I I suggest you taking a look to the ECB Guide to Internal Models and the EBA Guideline on PD and LGD estimation, these give some guidance on the expected risk drivers for these models (in EU)
- For CRE, [the above] risk drivers sound reasonable, I’ll suggest taking a look into the Basel slotting guidelines for Specialised lending, which include CRE
In most cases, I’ve seen that for both portfolios, these risk drivers are
included in the rating model rather than the PD model, so you just use rating as risk driver for the PD -
From a first principles perspective, one could defend exclusion of “administration defaults” by thinking about the use case for those modeled ratings, which is more often to measure risk, allocate/hold capital against, price loans, and forecast losses under stress.
For CRE segmentation, I think location is very important consideration nowadays given the changing behaviors since COVID. I would also think that loans made toward Retail Strip Malls (for example) are very different from Office Space are very different from Hotels/ Motels – given fundamental difference in use case for end users. Not sure if those distinctions are captured by NAICS but worth looking into
-
The other C&I related criteria I have seen for segmentation (may or may not be captured by your size / sector view):
- Rated / quoted vs. not (on the basis these firms have access to additional sources of funding plus an extra source of predictive info, although that can be reflected by other mechanisms)
- Specialised lending vs not (Basel has some rules re: when should be viewed as specialised – to some extent it comes down to legal form)
- Legal form e.g. limited liability vs. partnership vs sole trader (at bottom end)
- Leverage finance / recent transaction e.g. divestiture, M&A … (on the basis that they are more sensitive to changes and historical performance data may be less relevant