Set credit scoring against credit rating as tools for judging credit quality, and outline the varieties of each.
The idea
Credit scoring and credit rating both rank borrowers by credit quality but differ in form and use. A score is a number produced by a statistical model, such as discriminant analysis or logistic regression, from borrower characteristics. It suits large homogeneous populations such as consumers and small businesses, and comes in application, behavioural, collection and fraud varieties depending on the decision supported. A rating is an ordinal grade on a defined scale, assigned to corporates, banks, sovereigns and debt issues, usually with analyst judgement. Ratings are external, from agencies such as S&P, Moody's and Fitch, or internal, and internal systems are statistical, expert-based or hybrid. A score is only a ranking until it is mapped and calibrated to a PD. In the Altman model, WC is working capital, TA total assets, RE retained earnings, MVE market value of equity and TL total liabilities. The Altman score is Z = 1.2 × WC/TA + 1.4 × RE/TA + 3.3 × EBIT/TA + 0.6 × MVE/TL + 1.0 × Sales/TA. Below 1.81 is distress and above 2.99 is safe.
Formulas
Altman Z-score
Z = 1.2 × X₁ + 1.4 × X₂ + 3.3 × X₃ + 0.6 × X₄ + 1.0 × X₅
- X₁
- working capital ÷ total assets
- X₂
- retained earnings ÷ total assets
- X₃
- EBIT ÷ total assets
- X₄
- market value of equity ÷ total liabilities
- X₅
- sales ÷ total assets
Public manufacturing firms. Below 1.81 is distress, above 2.99 safe. X₄ divides by total liabilities, not total assets.
A practice question
Which statement best distinguishes credit scoring from credit rating?
- A score is a numeric output of a statistical model; a rating is an ordinal grade, often assigned with judgement. - correct
- A score is always issued by an external agency, whereas a rating is always produced by the lender's own model.Why not: Reverses who produces them: agencies issue ratings and lenders build scoring models, and ratings can also be internal.
- Scoring measures only loss given default, whereas rating measures only the probability of default and nothing about recovery.Why not: Confuses what is measured: both mainly rank default risk, and neither is limited to one component in this way.
- Scoring is applied only to sovereigns and large corporates, whereas rating is applied only to consumers.Why not: Swaps the usual domains: scoring is typical for retail and small business, ratings for corporates, banks and sovereigns.
A credit score is a number from a statistical model, typically for large homogeneous populations. A rating is an ordinal grade on a scale, typically for corporates, banks, sovereigns and issues, and often involves judgement.
A harder one
A bank's internal rating process for mid-sized companies starts with a statistical score, which analysts may adjust by a limited number of notches using qualitative information such as management quality. How is this system best classified?
- Pure statistical scoring, because a numeric score drives the whole process.Why not: Ignores the analyst override, which means the output is not purely model-based.
- An external agency rating, because letter grades are produced.Why not: Uses a letter scale as the test; an agency rating requires an external agency, while this is a bank's internal grade.
- A market-based model, because qualitative information reflects market views.Why not: Market-based models infer risk from traded prices; analyst assessment of management quality is not market data.
- A hybrid internal rating system: statistical model plus expert judgement. - correct
Internal rating systems can be statistical, expert-based or hybrid. A model score that analysts may adjust using qualitative information is a hybrid system.
The hardest tier
The original Altman Z-score is Z = 1.2 x X1 + 1.4 x X2 + 3.3 x X3 + 0.6 x X4 + 1.0 x X5, where X1 = working capital / total assets, X2 = retained earnings / total assets, X3 = EBIT / total assets, X4 = market value of equity / total liabilities and X5 = sales / total assets. Zones: below 1.81 distress, 1.81 to 2.99 grey, above 2.99 safe. A manufacturer has total assets 1,000, working capital 100, retained earnings 150, EBIT 80, sales 1,100, market value of equity 450 and total liabilities 500 (all in USD millions). Which result is closest to the correct Z-score and zone?
- Z = 1.96, grey zoneWhy not: Divides market value of equity by total assets (0.45) instead of total liabilities (0.90) in X4.
- Z = 2.33, grey zoneWhy not: Adds the five ratios without applying the model's coefficients.
- Z = 2.23, grey zone - correct
- Z = 1.13, distress zoneWhy not: Omits the sales / total assets term (X5), which contributes 1.10 to the score.
X1 = 0.10, X2 = 0.15, X3 = 0.08, X4 = 450/500 = 0.90, X5 = 1.10. Z = 0.12 + 0.21 + 0.264 + 0.54 + 1.10 = 2.234, inside the 1.81 to 2.99 grey zone.
In the app: 6 practice questions on this outcome, spaced review that brings it back before you forget it.
Start free →