RiskTech Forum

S&P: Recent Developments in Bank Risk Rating Systems

Posted: 1 March 2017  |  Source: S&P Global Market Intelligence

While the appropriate level of complexity of a Bank’s Risk Measurement System is specific to each institution and portfolio type – and we know that- one size does not fit all -- we are seeing more and more Banks adopting a “dual risk ratings” process. I should note that a risk rating should not be confused with a credit rating issued by a credit rating agency.

In this dual system, the probability of default (PD) is estimated separately from the loss given default (LGD). The expected loss for a given loan is then calculated as their product.

This method is also among several valid options for estimating expected credit loss explicitly contemplated in F.A.S.B.’s (Financial Accounting Standards Board) proposed standards update, called the current expected credit loss (CECL) model.

This indicates that Banks are already thinking about ways to replace the existing incurred-loss estimation approach to an expected loss type of model. And as we know, the allowance for credit losses is one of the most significant estimates on a Bank’s financial statement and regulatory report because it has a direct impact on earnings.

Dual risk rating systems that separate PD and LGD assessments have initially emerged because a single risk rating may not support all of the functions that require credit risk evaluations. Borrower risk ratings typically support deal structuring and administration, while facility risk ratings support Allowance for Loan and Lease Losses (ALLL) and capital estimates.

So how do Banks build these systems in practice? For banks with sufficient internal data, PD, LGD, and EAD models are typically based in large part on the bank’s own historical default and loss experience. However, most banks lack sufficient data for creating such models. Banks that find themselves in this situation have several options:

• First, it is often possible, for example for Community Banks with relatively straightforward portfolios, to estimate expected loss directly (without going into individual components), based on their historical loss experience coupled with a judgmental assessment of the current economic environment.

• Second, banks can build robust custom PD or LGD models based on external data, such as that which is provided by S&P Global Market Intelligence, which can be sampled in such a way as to represent the bank’s own portfolio.

• And the third option for banks facing data constraints is the use of vendor models and scorecards to estimate PD and LGD—and, in turn, expected loss. These models should be reviewed to make sure they are appropriate relative to the composition of the bank’s loan portfolio. Out-of-sample validation, calibration, and benchmarking are all common exercises we perform to ensure that the model is applicable.

In practice, we find that often times banks prefer to rely on vendor models and scorecards regardless of their internal data situation, since such models have undergone model validation, are maintained by the vendor, and represent leading industry practices.

In summary, the dual risk rating system requires a risk rating on the credit worthiness of the borrower and a risk rating based on the facility of the loan. The two risk ratings are then combined using a matrix such as the one shown on this slide to develop an overall composite loan quality risk rating.

That’s all for today, but if you are interested in learning more about this topic, or any of the solutions we covered, please complete the short form that will appear on your screen.