Skip to content

New to Omni?·Get 20% off your first purchase with code WELCOME20 at checkout.See plans

FRM® credit ratings: what must an internal system do?

An internal rating links borrower risk, default estimates, monitoring and validation. Work through an invented case and avoid common exam traps in FRM Part II.

ConceptOmni Curriculum Team6 min read

Blank cream index cards rise in ordered stacks beside an unmarked ledger and pencil on a navy desk mat.
On this page

An internal credit rating system assigns risk grades through a defined process and connects them to estimates, monitoring and decisions. For an FRM® candidate, the central task is to distinguish the rank ordering of borrowers from the numerical probability of default attached to a grade, then ask whether the system remains useful as conditions change. A tidy scale is not proof of accurate risk measurement.

The rating is an output of a system

A grade is a compressed statement about credit risk. To produce it, a bank needs data, criteria, human review, controls and a way to update the assessment. The Basel Committee describes a rating system broadly enough to include processes and technology as well as the assigned grade and the estimates of default and loss. This matters because two banks can both call a borrower “grade B” while using different definitions, horizons and calibration samples.

Begin by asking what is being rated. Borrower risk concerns the chance that an obligor fails to meet its obligations. Transaction risk also depends on the particular exposure: collateral, seniority, guarantees and other contractual details can affect the loss if default occurs. The Basel Committee says banks should decide whether to rate the borrower, the transaction or both. A candidate who treats one letter as a complete expected-loss measure has skipped this distinction.

Internal grades are useful because they make different exposures comparable enough for oversight. They can guide credit approval, pricing review, watchlists, portfolio concentration analysis and internal capital decisions. They need sufficient gradations to separate credits that are all broadly performing but have different risk. Yet more labels do not automatically produce better decisions: if assessors apply them inconsistently, the apparent precision is false.

Ranking and calibration answer different questions

A good ranking places borrowers more likely to default below those less likely to default under a clearly stated convention. Calibration asks whether the default probability assigned to each grade agrees with realised experience over an appropriate observation period. A model might rank borrowers well but attach probabilities that are too low. Another might give a plausible average probability while failing to distinguish the safest from the weakest borrowers. Both defects affect use of the ratings.

Probability of default is not a realised default count for a single borrower. It is an estimate for a population or grade over a specified horizon. Likewise, a year with no defaults does not prove a very low-risk grade was correctly calibrated, especially if only a few obligors were observed. The Basel Committee notes that low-default portfolios require additional evidence and validation methods because simple comparisons can be weak. Ask about sample size, definitions and economic conditions before declaring the model sound.

Migration adds another layer. A borrower’s grade may improve or deteriorate as information changes. A transition table records how a cohort moves among grades, including default, over a stated period. It is useful for spotting deterioration in the book, but a migration table built from a short benign period should not be treated as a timeless law. The rating process must capture weakening credits early enough for monitoring to respond.

A transition matrix must also have a clear starting cohort and observation horizon. If borrowers leave the dataset when a loan is sold or repaid, the analyst should understand how those exits are handled before interpreting a default proportion. Mixing rating definitions after a methodology change can make apparent migration a measurement artefact. These details explain why a bank needs governance over data and process, rather than merely a spreadsheet of grade movements.

A worked case with invented figures

Imagine a fictional bank with an internal grade called Amber. It maps Amber to an illustrative annual default probability of two per cent. The bank holds an illustrative £10 million exposure to one Amber borrower and assumes an illustrative loss given default of forty per cent. Multiplying exposure, default probability and loss given default gives an illustrative expected loss of £80,000 for the specified horizon. These invented numbers explain the relationship; they are not GARP data, regulatory inputs or a quote for a real loan.

Now suppose the bank reviews an illustrative cohort of one hundred Amber borrowers and observes four defaults during a year. The realised rate is four per cent for that cohort, above the assigned two per cent. It would be careless to announce that the entire rating system is invalid from this one observation. The bank should examine the cohort definition, underwriting changes, economic conditions and performance across more periods, then decide whether Amber’s calibration needs adjustment. It should also check whether weaker grades still show higher default rates than stronger grades.

Consider another fictional cohort whose realised defaults align with the expected average, yet most defaults occurred among borrowers previously classified as safer than Amber. The aggregate arithmetic can look acceptable while ranking is poor. A useful review therefore tests both the level of estimated risk and the ordering. It also asks whether officers delayed downgrades, whether overrides were documented and whether a changing portfolio made historical comparisons misleading.

In the fictional bank, an analyst could also ask whether the Amber exposure is secured or unsecured. The assigned borrower probability might remain two per cent in either case, while the amount recoverable after default changes. The bank therefore needs distinct inputs for default likelihood and loss severity when estimating expected loss. Confusing those inputs can make a numerical answer look precise while answering the wrong credit question.

Through the cycle and point in time

A through-the-cycle approach aims to reflect creditworthiness over a broad range of economic conditions, so it tends to move less with temporary changes in the cycle. A point-in-time approach is more responsive to current conditions and can move sharply during stress. These are modelling choices, not moral grades. Do not infer that the steadier estimate is always the more accurate description of current risk, or that a current stress signal automatically changes a longer-horizon grade.

Suppose the fictional Amber borrower’s sales weaken during an economy-wide downturn. A current-condition estimate may rise before the bank concludes that its longer-run grade should change. The bank still needs to monitor covenants, cash flow and the cause of the weakness. If it becomes persistent or structural, a grade review is warranted. The key exam move is to state the horizon and purpose of the estimate before comparing the two numbers.

Validation and governance complete the answer

The Basel Committee calls for periodic review of individual credit ratings and for a function independent of loan origination to confirm consistency and accuracy. Its internal-ratings requirements also call for comparing realised default rates with estimated probabilities and documenting the analysis. These are controls against a rating scale becoming a sales label. The bank should investigate material deviations, revise estimates where evidence warrants and retain a record of methodology changes.

For the exam, answer in a sequence: define the exposure being rated, explain the grade and its probability mapping, identify decisions that use it, then test ranking, calibration, migration and governance. Do not confuse an external agency opinion with a bank’s own internal scale, and do not turn a grade into a guarantee that a borrower will or will not default. GARP lists credit risk measurement and management within Part II on its FRM exam page; the Omni FRM page provides programme context and Omni’s methodology explains the teaching approach.

One further limitation is that ratings depend on the evidence available at the time. Financial statements can lag, forecasts can be optimistic and a model can miss a change in business conditions. Qualitative review helps only when its reasons are recorded and challenged; otherwise an override can disguise inconsistency. A complete answer therefore includes a feedback loop from new borrower information and realised outcomes to grade review and model validation.

A final self-check is to imagine changing only the collateral on a loan. The borrower’s likelihood of default may be unchanged, while the loss on that specific exposure may differ. If your answer changes the borrower grade solely because collateral improved, you have probably mixed borrower risk with transaction risk. That distinction makes the rest of the system easier to explain.

Sources

  1. General credit risk management | Bank for International Settlements · bis.org · Retrieved
  2. Credit risk – internal ratings-based approach | Bank for International Settlements · bis.org · Retrieved
  3. FRM Exam FAQs | GARP · garp.org · Retrieved

Explore FRM study resources

Educational content, not investment advice. This article is written to help candidates prepare for professional exams. It is not a recommendation to buy, sell or hold any security, and it does not take your personal circumstances into account.

Omni Finance Academy is an independent exam-preparation provider and is not affiliated with or endorsed by CFA Institute, GARP, CAIA Association or CFP Board. CFA®, FRM®, CAIA® and CFP® are trademarks of their respective owners. Full disclaimer.