A lender approves two card applications with identical 720 FICO scores. One borrower defaults within a year; the other never misses a payment. FICO is a probability, not a forecast: the score sorts the population by default likelihood without claiming to predict any one outcome. The exam tests whether you can read a credit scoring framework as a probabilistic ranking tool rather than a deterministic verdict.
Bank supervisors evaluate institutions on five dimensions: Capital adequacy, Asset quality, Management, Earnings, Liquidity. Ratings range 1 (best) to 5 (worst); a 4 or 5 triggers heightened supervisory action. Asset quality looks at concentration and classified-loan ratios; management at risk culture and controls; earnings at quality and stability of income; liquidity at funding diversification and stressed coverage. A bank with strong capital but weak management is not a strong bank: the framework is multiplicative, not additive.
Five inputs: PD (over the horizon), LGD (fraction of EAD lost on default), EAD (dollar exposure), EL = PD × LGD × EAD, and time horizon (one year for capital, longer for stress...
Common mistakes
- Treating FICO as a forecast. FICO and similar scores are probabilities, not predictions. A 720 borrower has a low default probability but is not "won't default." Choices that say "FICO predicts default" oversimplify.
- Confusing TTC with PIT ratings. TTC ratings are smoother and less reactive; PIT ratings are more responsive. Choices that imply agencies use PIT methodologies are wrong: agencies aim for TTC. Trap: a question asks why agency ratings often lag market spreads. The TTC philosophy is the right answer.
- Using the raw sovereign spread as equity ERP. Damodaran's framework requires multiplying the default spread by the equity/bond volatility ratio. A choice that adds the raw sovereign spread to the US ERP without the multiplier understates equity risk. Trap: choice C uses 200 bps directly, missing the 1.5x volatility multiplier.
Bottom line
- Credit scoring ranks borrowers by default probability using empirical models (logistic regression, decision trees). FICO buckets: super-prime 800+, prime 740 to 799, near-prime 670 to 739, subprime below 670.
- Application vs behavioral scoring: application uses static data at origination; behavioral uses ongoing payment patterns to adjust limits, repricing, and collection treatment.
- Through-the-cycle (TTC) ratings smooth across business cycles; point-in-time (PIT) ratings reflect current conditions. Agencies aim for TTC; bank internal IRB models sit closer to PIT.
- Retail credit relies on granularity for diversification; corporate credit relies on industry and geographic spread, requiring different modeling and default-management approaches.
Exam shortcut
When a question gives you a CDS spread and asks for hazard rate, divide by (1 − R) and remember to convert annualized basis points to a decimal. When it gives you a sovereign spread and asks for country ERP, multiply by the equity-to-bond volatility ratio. The raw spread is the trap answer.
The full lesson (about 3,683 words, 25 min read) adds 3 worked examples, all 6 common mistakes, a self-check, free in the app.
Learning objectives
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
Browse all free FRM Part II lessons or jump into free FRM Part II practice questions.