Commercial Banking - Goethe
Transcrição
Commercial Banking - Goethe
Schwerpunkt Finanzen Johann Wolfgang Goethe Universität, Frankfurt (Main) Prof. Dr. Mark Wahrenburg und Dr. Peter Raupach Commercial Banking 4-stündige Vorlesung mit Übung im SS 2002 © Commercial Banking The last decade has seen an increase in the demand and possibilities for credit risk management Driving forces: Impact: New methods for credit risk measurement Rating systems Evolving market for credit risk Credit derivatives and securitization Bank regulation (Basle II) (mainly) Rating systems Shareholder-value management in banks requires risk assessment Credit portfolio management © Commercial Banking Credit portfolio management RAROC / EVA Tools Purpose and objectives Rating systems Purpose Learn how the credit risk of individual obligations can be measured through ratings Objectives Understand dimensions of credit losses Understand differences between rating systems Learn how to implement particular rating systems Learn how ratings are used in the Basle II proposal © Commercial Banking Expected credit loss (from default) is default probability × loss given default × exposure at default Dimensions of credit risk Which losses will you suffer on average? What is the likelihood that a borrower becomes insolvent? Expected loss = EL = Probability of default = PD X What is your exposure to the borrower in case of insolvency? Exposure at default = EAD X How much of your nominal exposure will be lost in case of insolvency? © Commercial Banking Loss given default = LGD Verwendungsbereiche von Rating • Aufsichtsbehörden / Eigenkapitalregulierung – Basel – National bank regulators • Securitization – Bewertung von Collateralized Loan Obligations • Strategisches Management – ROE/EVA – Entry/Exit Entscheidung • Kreditmanagement – Vorprüfung von Kreditanträgen – Kreditvergabeentscheidung (ja/nein) – Marking to Market des Kreditportfolios – Überwachung von Engagements (work-out) – Bildung von EWBs – Pricing / Konditionstafeln – Incentive compensation © Commercial Banking The dimensions of credit risk need to be exactly defined When is loss given default measured? – initiation / end of bankruptcy procedures – emergence of: liquidation, settlement, completion of restructuring) What constitutes a default? – missed payments – distressed exchanges – bankruptcy What probability of default shall be measured? – time horizon (usually 1 year) – point in time perspective or rating through the cycle © Commercial Banking Basle II proposes a definition of default based on four criteria Basle 2 definition of default event: (a) It is determined that the obligor is unlikely to pay its debt obligations (principal, interest, or fees) in full. (b) A credit loss event associated with any obligation of the obligor, such as a charge-off, specific provision, or distressed restructuring involving the forgiveness or postponement of principal, interest, or fees. (c) The obligor is past due more than 90 days on any credit obligation (d) The obligor has filed for bankruptcy or similar protection from creditors. © Commercial Banking Measuring default probability through-the-cycle abstracts from the current position in the credit cycle Permanent deterioration of credit quality Credit quality under neutral conditions Credit quality under neutral conditions Current position in cycle no rating change © Commercial Banking rating change no rating change Because of its medium term perspective, a through-the-cycle approach may contain valuable information Consider investing in a 10-year corporate bond You plan to sell the bond in one year Question: One year default probability sufficient for assessing the risk of your position? Answer: No. If the bond has not defaulted, the price in one year depends on the default probabilities prevailing in one year. Ideally, you have a complete term-structure of default probabilities. A long term average (rating through-the-cycle) is useful © Commercial Banking Rating systems can distinguish between long-term and shortterm risk, and between issue and issuer risk Issuer rating: an assessment of an obligor’s overall financial capacity to pay its financial obligations Issue rating: creditworthiness of an obligor with respect to a specific financial obligation Short-term rating (pertaining to obligations with an original maturity of no more than 365 days) Long-term rating: covers all obligations S&P rating grades High credit quality Low credit quality © Commercial Banking Long-term rating Short-term rating AAA AA A BBB BB B CCC A-1 A-2 A-3 B C Rating modifyers allow better differentiaton of risk High credit quality Investment grade Non-investment grade Low credit quality Rating category Rating modifiers (+, flat, -) AAA AA A BBB BB B CCC AAA AA+, AA, AAA+, A, ABBB+, BBB, BBBBB+, BB, BBB+, B, BCCC Rating may also be continuous, e.g. equal to the estimated default probability © Commercial Banking Bank often map their internal Rating system to external ratings Internal Grade S&P 1 year EDF Equivalent AAA - A Moody's Equivalent Group Aaa - A2 Investment grades - 0.01-0.07% 1 0.08% A- A3 2 0.12% BBB+ Baa1 3 0.18% BBB Baa2 4 0.27% BBB- 5 0.41% BB+ 6 0.61% BB+ 7 0.91% BB 8 1.37% 9 2.05% 10 3.08% 11 4.61% B B1 12 6.92% B- B2 13 10.38% CCC B3 BBB+ 14 CC, C 15 D © Commercial Banking Baa3 Active business grades Ba1 Ba2 Ba3 High yield grades Watchlist Defaulted Default probabilities for a rating grade can be derived from historical default frequencies for issuers of that grade Rating AAA AA A BBB BB B CCC Percentage defaulted after year 2 3 4 5 0.00 0.05 0.11 0.17 0.02 0.07 0.15 0.27 0.12 0.21 0.36 0.56 0.54 0.85 1.52 2.19 3.4 6.32 9.38 12.38 12.36 19.03 24.28 31.66 33.52 41.13 47.43 54.25 1 0.00 0.00 0.04 0.24 1.01 5.45 23.69 10 1.00 0.96 2.06 5.03 23.69 42.24 60.91 less observations available * Quelle: S&P, 1981-1998. © Commercial Banking As with default probabilities, probabilities for rating migrations can be estimated with historical data S&P transition matrix (1981-1998) Rating at year end (%) Initial Rating AAA AA A BBB BB B CCC © Commercial Banking AAA AA A BBB BB B CCC D 91.94 0.64 0.07 0.04 0.04 0.00 0.19 7.46 91.79 2.27 0.27 0.10 0.10 0.00 0.48 6.76 91.68 5.56 0.61 0.28 0.37 0.08 0.60 5.12 87.87 7.75 0.46 0.75 0.04 0.06 0.56 4.83 81.48 6.95 2.43 0.00 0.12 0.25 1.02 7.90 82.80 12.13 0.00 0.03 0.01 0.17 1.11 3.96 60.44 0.00 0.00 0.04 0.24 1.01 5.45 23.69 Rating systems differ in the concept of credit loss they employ Through-the-cycle expected loss – Moody’s ratings Through-the-cycle probability of default – Standard & Poor’s ratings – Fitch IBCA ratings Point in time probability of default – Internal ratings of banks (typically) – KMV (KealhoferMcQuownVasicek) – Moody’s Riskcalc (default probabilities for private firms) © Commercial Banking Rating systems differ in the way they aggregate information There are various ways of deriving ratings from information about credit quality Methods for weighting individual indicators Indicators of credit quality: - leverage - profitability - management - stock price - ... Judgmental / expert system Theoretical reasoning Statistical analysis (Discriminant Analysis, Logistic Regression, Neural Networks, ...) Hybrid © Commercial Banking Rating Grundlage jeden Ratings: Vergleich von „gut“ und „schlecht“ Gesunde Unternehmen Ausfallsample Kunde Merkmale EWB- Merkmale Fall A B C... X A 1 1 2 2 3 3 . . . . . . N K B C... X Vergleich der Ausprägungen ZIEL IST DIE ERMITTLUNG DER INDIVIDUELLEN TRENNSCHÄRFE VON KENNGRÖSSEN © Commercial Banking They are other reasons whey ratings can differ Incentives faced by raters – are raters paid by borrowers? (solicited / unsolicited ratings) – are raters involved in credit approval or other business with the borrower? – it the rater “betting his money” on his rating? – could a rating revision signal bad quality of the initial rating? Frequency of rating revisions Type of information that enters the rating process – public information – personal interviews / private information – analysts information © Commercial Banking Indicators of default probabilities: retail clients Information obtained through loan request – occupation – salary – marital status, children – assets and liabilities – age, sex, education – domicile (ZIP-code) Information from checking account – length and amount of overdrafts – pays bills in time – spending behavior General information – regional unemployment rate – observed trends in default frequency of customer segment © Commercial Banking Indicators of default probabilities: corporate clients Firm-specific accounting information – profit margins – interest rate coverage – leverage – sales Other firm-specific information – forecasts of earnings, sales etc. – management quality – competitive position Financial market variables – stock and bond prices (of borrower or comparable firms) – stock price volatility Macroeconomic variables – profitability in industry – interest rate level, term structure of interest rates © Commercial Banking The most important financial indicators are profitability and leverage Median financial ratios for S&P rating categories EBIT interest coverage Free oper. Cash flow / total debt Return on capital Operating income / Sales Total debt / total capital Total sales (bil $) AAA 18 55 28 29 27 21 AA 11 25 23 21 36 13 A 7 16 20 18 40 5 Source: S&P Special Report, September 20, 2000 © Commercial Banking Qualitative Credit Risk Assesment: Stepwi 1. Bilanzanalyse 2. Qualität der Rechnungslegung 3. Management Qualität & Soft Factors 4. Industrie-Rating 5. Tier Position innerhalb Industrie 6. Länderisiko 7. Garantien dritter Parteien 8. Vertragsstruktur 9. Sicherheiten © Commercial Banking BBB 4 7 14 15 47 2 BB 2 2 12 15 61 0.8 B 1 -5 7 11 75 0.4 1. Bilanzanalyse • Qualität, Liquidationswert der Aktiva – Korrektur um Eigentumsvorbehalte und andere vorrangige Rechte Dritter • Profitabilität – ordentliche / außerordentliche Erträge – Kapitalrentabilität • Cash Flow/ EBITDA – Operating Cash Flow / Fremdkapital – Schuldendienstkapazität (z.B. EBITDA/Zinsen p.a.) • Liquidität – kurzfr. Vermögen/ kurzfr. FK • Leverage (Verschuldungsgrad) © Commercial Banking Unternehmensanalyse ff. • Zyklizität/Schwankungen von Earnings / Cash Flows • Zugang zu Kapitalmärkten • Private Reserven der Gesellschafter/ Zugang zu Hausbankfinanzierung • Marktkapitalisierung, book-to-market-Verhältnis © Commercial Banking Benchmarking von Bilanzkennziffern gegen S&P-Daten S&P Medianwerte Industrial Comp. AAA EBIT / Zinszahlung 16 EBITDA / Zinszahlung 20 Vorsteuer Gesamtkapitalrentabilität 77 Operating Income / Sales 24 Langfristiges FK / Gesamtes FK 13 FK / Gesamtkapital 23 AA 11 15 30 19 22 29 A 6 8 19 16 32 39 BBB 4 6 8 15 43 47 BB 2 4 2 15 54 56 Quelle: S&P Corporate Ratings Criteria, 1998 © Commercial Banking 2. Qualität der Rechnungslegung / Berichterstattung • Pünktlichkeit • Vollständigkeit • Plausibilität von Angaben • Identität von Wirtschaftsprüfer / Aufsichtsrat etc. © Commercial Banking B 1 2 1 13 66 69 3. Soft Factors • Beobachtbare Kontobewegungen • Risiken: Umwelt / Schadenersatzklagen... • Ausbildung / Erfahrung des Management Teams • Risikomanagementprozess des Unternehmens • Plausibilität / Stringenz der Unternehmensplanung © Commercial Banking 4. Industrie-Rating • Wachstumstrends • Kapazitätsauslastung • Profitabilitätstrends • Innovationsintensität • Regulierung • Konjunkturabhängigkeit • Wettbewerbsintensität © Commercial Banking 5. Tier Position innerhalb der Industrie • Marktanteil • Strategie (Kostenführer / Technologieführer, Nischenplayer...) • Ausnutzung von Skaleneffekten • Homogenität der Produkte und Wettbewerbsintensität © Commercial Banking 6. Länderrisiko • Anleihespread über US-Treasuries • Hermes-Exportbürgschaftspreise • Länder-Risikoanalyse mit Zahlungsbilanz-, Makrodaten etc. © Commercial Banking 7. Garantien dritter Parteien • Implizite / explizite Garantie durch Konzernmutter • Bürgschaften von Gesellschaftern, „Friends & Relatives“ • offenstehende Kreditlinien anderer Banken © Commercial Banking 8. Kreditvertragsstruktur • Vorzeitige Kündigungsrechte durch Bank • Vorzeitige Fälligkeit bei „material adverse change“ • Schutz vor Reichtumsverschiebungen durch Gewinnausschüttungen / vorrangiges Fremdkapital © Commercial Banking 9. Sicherheiten • Bewertung und Wertschwankungsrisiko • Korrelation von Wertschwankung und Kreditwürdigkeit • Durchsetzbarkeit der Ansprüche / Verwertbarkeit – kollidierende Ansprüche – Vollstreckungssperre im Insolvenzverfahren © Commercial Banking Statistical methods use historical data to derive optimal weights for credit quality indicators Observations on past credit losses are coded into a variable y (0= no default, 1= no default) Set up functional relationship between y and variables x1, x2,...xN y= f(x1, x2,...xN) Estimation of functional relationship f (.) based on some optimality criterion – minimize squared prediction error [ y - f(x1, x2,...xN)]² – maximum likelihood – minimize probability of misclassification © Commercial Banking Statistical Models require good data feeds for calibration Data basis of several systems Fitzpatrick Beaver Altman Lev Wilcox Deakin Edmister Blum Taffler Libby Diamond Altman, Haldeman and Narayanan Marais Dambolena and Khoury Ohlson Taffler El Hennawy and Morris Moyer Taffler Zmijewski Zavgren Casey and Bartczak Peel and Peel Barniv and Raveh Boothe and Hutchninson Gupta, Rao, and Bagchi Kease and McGuiness Keasey, McGuiness and Short Shumway Moody’s RiskCalc Public Firm Moody’s RiskCalc Private Firm Median © Commercial Banking Year Defaults Non-Defaults (32) 19 19 (67) 79 79 (68) 33 33 (71) 37 37 (71) 52 52 (72) 32 32 (72) 42 42 (74) 115 115 (74) 23 45 (75) 30 30 (76) 75 75 (77) 53 58 (79) 38 53 (80) 23 23 (80) 105 2,000 (82, 83) 46 46 (83a) 22 22 (84) 35 35 (84) 22 49 (84) 40 800 (85) 45 45 (85) 60 230 (88) 35 44 (89) 58 142 (89) 33 33 (90) 60 60 (90) 43 43 (90) 40 40 (96) 300 1,822 (00) 1,406 13,041 (00) 1,621 23,089 40 45 Logitanalyse zur Schätzung der Ausfallwahrscheinlichkeit • Modell: Z = ß1 x1 + ß2 x2 + ... + ε ("Score" ) 1 (logistische Funktion) 1 + e −Z ß : zu schätzendeRegressionskoeffizienten p(def ) = P(def) 1 x : Faktorausprägungen • Dateninput: Sample von Krediten mit Outcome (p = 1 oder 0) • Modellschätzung: Suche ß‘s, die Prognosegüte maximieren • Output: p(def) für gegebene Faktorausprägungen • Probleme: – Wahl einer repräsentativen Stichprobe für Modellkalibrierung – Konstanz der Parameter im Zeitablauf unterstellt © Commercial Banking Auswahl der Faktoren Z Diskriminanzanalyse Z = ß1 x1 + ß2 x2 + ... + ε ("Score" ) Wahrscheinlichkeit Sample: Kein Konkurs Sample: Konkurs Z Z* Schlechte Kredite akzeptiert (Typ 1 Fehler) Gute Kredite abgelehnt (Typ 2 Fehler) Ziel: Suche ß‘s, die (mit Stabw normierten) Gruppenmittelwert-Abstand maximieren. Output: 1. Z* (⇒ Klassenzuordnung) 2. Wahrscheinlichkeit einer Fehlklassifikation (aus Z-Z*) © Commercial Banking Relationship between credit risk indicators and default probability is often non-linear Leverage (%) Median leverage (total debt / total capital) for S&P rating grades compared with historical default probabilities for the respective grades 80 70 60 50 40 30 20 10 0 0% 1% 2% 3% 4% Default probability © Commercial Banking 5% 6% Rating systems should allow for non-linearities in the relation between credit risk and credit risk indicators The scoring method (widely used in banks) uses a linear combination of credit risk indicators xi with weights bi Credit risk score = Σi bi xi (The scoring method does not produce default probabilities since score is not constrained to lie between 0 and 1) The probit regression is one method for nonlinear estimation: 1,0 y=Φ[bi xi] 0,8 Φ= cumulative normal 0,6 distribution function 0,4 0,2 0,0 x © Commercial Banking An aside: standardized rating procedures are not necessarily based on statistical procedures for weighting credit risk indicators Many banks use scoring methods for their ratings Often, the weights have been set judgmentally ⇒ The rating is standardized but not based on a statistical optimization of the weights (e.g. using discriminant analysis) © Commercial Banking Example: Moodys Rating-System RiscCalc • Anspruch: Transparentes Benchmarksystem schaffen Æ nur „hard data“ verwendet Æ nur Bilanzdaten gehen ein Ziele: • Verständlichkeit • Statistische Power • Trennschärfe (Unterscheidung) • Kalibrierung an PODs • Empirisch validierbar © Commercial Banking Schritt 1: Transformation der Inputvariablen © Commercial Banking Schritt 2: Schätzung eines multivariaten Probitmodels y = prob(default) = Φ{ f ( x, B)} Φ(x) 1 x f ( x, B ) = B0 + T1 ( x1 ) B1 + T2 ( x2 ) B2 + ... + T10 ( x10 ) B10 © Commercial Banking Schritt 3: Anpassung des Modells an empirische Ausfallwahrscheinlichkeiten © Commercial Banking Results: Input variables of Moody’s RiskCalc Accounting Variable Weight Profitability Capital Structure Liquidity Size Sales Growth Trading Accounts 23% 21% 19% 14% 12% 12% Source: Moody‘s (2000): RiskCalc private model: Moody‘s default model for private firms © Commercial Banking Option pricint theory can be used to derive default probabilities ( Merton model) Firm is solvent, if value of assets > value of liabilities Firm is insolvent otherwise Default probability is the probability that the value of total assets falls below the value of liabilities This probability can be determined if we know – the current asset value – the distribution of future asset value changes – the value of liabilities © Commercial Banking To estimate the distance to default we need assumptions about the evolution of the asset value Assume: firm’s debt consists only of a 1 year zero bond asset value is distributed lognormally Possible paths Market value in t0 1+µ Equity Assets Liabilities probability of default © Commercial Banking Pay Offs of Debt and Equity in the Merton Model Value of Debt = Value of Assets ./. Value of Put Wert FK in T or: Value of Zerobond ./. Value of Call D Value of Equity = Value of Call D © Commercial Banking Wert der Aktiva in T Valuation of risky debt with the Merton-Model BT F Ausfallbedrohtes FK + Put-Option = risikoloses FK B0 PT + P0 = Fe-rT VT F P0 = −N (−d1 )V0 + Fe−rT N (−d2 ) (Black/Scholes Formel) ln(V0 / F ) + (r +1/ 2σ 2 )T d1 = ; d2 = d1 − σ T σ T © Commercial Banking Berechnung des fairen Spreads von ausfallbedrohtem FK im Merton Modell • Yield to maturity des ausfallbedrohten FK (yT): B0e yT T = F ⇒ yT = − ln( B0 / F ) / T ⇒ yT = − ln ( Fe− rT − P0 ) / F / T einsetzen von P0 : 1 V π T = yT − r = − ln N (d2 ) + −0rT N (−d1 ) T Fe © Commercial Banking Faire Spreads und Laufzeit nach Merton-Modell (illustr.) yT Volatilität hoch Volatilität gering T © Commercial Banking An implementation of the Merton model is the KMV CreditMonitor For listed firms: stock market prices & volatility used Use data of comparable firms Rely on empirical observation that firms typically default if Asset Value < Short-term Debt + α × Long-term Debt = Default Point (α: industry-specific) Use distance to default (DD) as sole default risk indicator DD = Asset Value - Default Point σ Asset Value Transform DD into default probability using an empirical fit (instead of lognormal distribution) © Commercial Banking PD DD KMV ratings often react faster to crises Default probability of US company Acme metals according to KMV and S&P Bankkruptcy on 29.9.98 Default probability in % 20 15 10 5 0 Okt 95 Jan 96 Apr 96 Jul 96 Okt 96 Jan 97 Apr 97 According to S&P Rating Jul 97 Okt 97 Jan 98 Apr 98 Jul 98 KMV EDF © Commercial Banking Difficulties in applying the Merton model Often, only book values available, not market values of equity and debt lagged and incomplete accounting information lognormal distribution assumption may be wrong debt structure is usually more complex than assumed in Merton model debt structure not constant over time © Commercial Banking Numerical valuation of risky credits using binomial trees • Standard in complex option pricing • In every time step, asset value goes up by factor u or down by factor d S uS with prob = q dS with prob = 1-q ∆t • The no arbitrage value is the present value of the expected terminal value under the risk neutral probability © Commercial Banking „Cook book receipt“ for binomial valuation of risky credit 1. Step: Set up a tree for the evolution of customer asset value a) b) c) d) Chose ∆t (for example omne month: ∆t = 1/12) Estimate asset volatility σ and safe interest rate r (r = 10%, σ = 40%) Evaluate current asset value, e.g. S = 100 Derive up- and down factors and pseudoprobability (u, d, q): u = eσ ∆t d = e −σ q = © Commercial Banking e r ∆t ∆t = 1,1224 = 0,8909 −d = 0,5076 u −d q = „risikoneutrale Wahrscheinlichkeit“ eines „up“ Resulting binomial tree for evolution of asset value Assume: credit has 5 month maturity Æ 5 step tree required Asset Value Evolution 178,13 158,71 141,40 125,98 112,24 100,00 141,40 125,98 112,24 100,00 89,09 112,24 100,00 89,09 79,38 89,09 79,38 70,72 70,72 63,01 56,14 © Commercial Banking 2. Step: Determine Pay Off at T and set default condition • Example: The customer has one credit with notional amount = 80 • You believe that: – default bevor T occurs, when asset value drops below 70 – default in T occurs, when asset value is below 80 – recovery rate is 50% © Commercial Banking Derivation of terminal pay offs Asset Value Evolution 178,13 158,71 141,40 141,40 125,98 125,98 112,24 112,24 100,00 112,24 100,00 100,00 89,09 89,09 89,09 79,38 79,38 70,72 70,72 63,01 56,14 Terminal Payoff 80,00 x x 80,00 x x x x x 80,00 x x x x 80,00 x x x 40,00 40,00 40,00 © Commercial Banking Step 3: Derive present value of credit by backward induction [ ] = q * S u + (1 − q) * S d e−r∆t Value of Credit 69,91 80,00 79,34 78,68 78,02 75,10 69,91 80,00 79,34 78,68 73,36 65,75 80,00 79,34 69,13 59,03 80,00 59,79 49,63 40,00 40,00 40,00 © Commercial Banking Step 4: Derive the fair credit premium • 69,91 is the present value of a risky zerobond with face value of 80 • We find: 69,91* (1 + 175%) * 5 = 80 12 Æ 175% is the fair anual interest rate! Æ 175% - r = 165% is the fair credit risk premium! © Commercial Banking Vergleich KMV vs. Logit-, Diskriminanzanalyse • Lineares vs. nichtlineares Modell (Erhöhung des Verschuldungsgrades um x% ⇒ Erhöhung des Z-Wertes um ß*x, unabhängig von Startwert ⇒ Erhöhung der Distance to Default abhängig von Startwert wegen Konvexität der Black-Scholes Formel • Andere Inputdaten • Aktualität, Reagibilität: KMV‘s EDF ändert sich für börsennotierte Unternehmen täglich in Abhängigkeit von Aktienmarktentwicklung ⇒ ständiges „Marking to Market“ von Krediten © Commercial Banking Rating agencies and banks typically combine both quantitative and qualitative criteria into their ratings Financial risk (financial ratio analysis...) Business risk (product range, competitive position, industry and country risk...) Management quality .... ⇒ Judgmentally combined into a rating © Commercial Banking Banks typically use different rating approaches for different clients Retail clients Small and mediumsized businesses Large Caps © Commercial Banking Quantitative information Qualitative assessment External ratings scoring model - - scoring model complementary if available as complement to external ratings primary source Validation techniques for internal ratings become important Selected overview of validation techniques comparison of default rates across rating classes - is it lower for better rated companies? comparison of expected and realized default rates Brier Score: sum of squared errors: If default occured: error = (1 – pd) If no default occured: error = (pd – 0), Power Curve analysis (next slide) © Commercial Banking Analyse der classification errors: Typ 1 und Typ 2 Fehler • Evaluiere für verschiedene z* die Güte der Trennschäfe Wahrscheinlichkeit Sample: Kein Konkurs Sample: Konkurs Z Z* Schlechte Kredite akzeptiert (Typ 1 Fehler) Gute Kredite abgelehnt (Typ 2 Fehler) © Commercial Banking The power curve analysis investigates the discrimatory power of rating systems power curve of perfect model 100% Defaults (in%) realized power curve cumulative power curve of worst model Size of area is a measure of quality 100% companies, ordered from bad to good, in % © Commercial Banking Leistungsfähigkeit der Ratingmodelle steigt (4000 observations, 50 defaults) „Power Curve“ verschiedener Modelle 1.00 % of Defaults Excluded 0.75 C-score Rated C-score Pub Shumway Public 0.50 C-score Private Z-score (4) 0.25 Compustat Daten 1990 - 97 4000 Beobachtungen 50 Ausfälle 0.00 0.00 0.25 0.50 Quelle: Falkenstein Commercial Credit Risk Modelling (1999)% of Sample Excluded © Commercial Banking 0.75 1.00