Introduction: The Anatomy of Credit Loss

While market risk focuses on price fluctuations in trading books (stocks, bonds, derivatives), Credit Risk represents the risk that a borrower, counterparty, or issuer will fail to meet their contractual financial obligations. Whether issuing a corporate loan, underwriting a consumer mortgage, or trading over-the-counter interest rate swaps, financial institutions face constant credit exposure.

To measure, price, and provision for credit risk under modern regulatory frameworks (such as the Basel Accords and IFRS 9 / CECL accounting standards), quantitative risk teams decompose credit loss into three core mathematical components: Probability of Default (PD), Loss Given Default (LGD), and Exposure at Default (EAD). This lesson deconstructs credit risk metrics, structural credit models (Merton Model), credit scoring scorecards, and expected credit loss calculations.

Part 1: The Three Pillars of Credit Risk Quantification

The expected credit loss (ECL) for any credit exposure is calculated as the mathematical product of three core variables:

ECL = PD × LGD × EAD

1. Probability of Default (PD)

Definition: The statistical probability that a borrower will default on their debt obligations over a specific time horizon (typically a 1-year horizon for regulatory capital or lifetime horizon for accounting provisions).

Modeling: Estimated using historical default databases, logistic regression scorecards, and machine learning classification models trained on borrower financial ratios, credit bureau histories, and macroeconomic indicators.

2. Loss Given Default (LGD)

Definition: The percentage of total exposure that the bank expects to lose if the borrower actually defaults, accounting for collateral recovery and legal liquidation costs.

Formula: LGD = 1 – Recovery Rate

Example: If a corporate borrower defaults on a $10,000,000 loan, but the bank seizes and liquidates collateral worth $6,000,000 (net of legal fees), the Recovery Rate is 60%, and the LGD is 1 – 0.60 = 0.40 (40%).

3. Exposure at Default (EAD)

Definition: The total gross financial amount that the bank is exposed to at the exact moment the borrower defaults.

Components: For a term loan, EAD is simply the outstanding principal balance. However, for revolving lines of credit or credit cards, borrowers often draw down remaining credit limits right before defaulting. Risk models use a Credit Conversion Factor (CCF) to estimate the expected drawdown of unused credit lines:

EAD = Drawn_Balance + CCF × (Credit_Limit – Drawn_Balance)

Part 2: Structural Credit Models (The Merton Model)

Developed by Nobel laureate Robert Merton, the Merton Model applies options pricing theory to evaluate corporate default risk.

1. The Balance Sheet Approach

The Merton model conceptualizes a firm’s equity as a European Call Option on the firm’s total underlying assets (V), with a strike price equal to the face value of the firm’s debt (D) maturing at time T.

If the total value of assets V_T falls below the debt obligation D at maturity, the firm defaults, and equity holders receive nothing.

Using the Black-Scholes option pricing framework, the model calculates the theoretical Distance to Default (DD)—how many standard deviations the firm’s asset value is away from its debt default threshold.

Part 3: Regulatory Frameworks: Basel III/IV and IFRS 9 / CECL

Global banking regulators enforce strict capital reserve requirements based on these credit risk metrics.

1. The Basel Capital Accords (IRB Approach)

Under the Internal Ratings-Based (IRB) approach of the Basel framework, sophisticated banks are permitted to use their own internal quantitative models to estimate PD, LGD, and EAD for their loan portfolios. These internal estimates are then plugged into regulatory capital formulas to determine the minimum Tier 1 capital the bank must hold to absorb unexpected credit losses.

2. IFRS 9 and Current Expected Credit Losses (CECL)

Historically, banks used an “incurred loss” model, waiting until a loan actually defaulted before setting aside loan-loss reserves. Following the 2008 crisis, accounting standards shifted to Expected Credit Loss (ECL) models:

Banks must instantly estimate and provision for lifetime expected credit losses the moment a loan is originated, incorporating forward-looking macroeconomic forecasts (e.g., adjusting PD estimates upward if economic unemployment is projected to rise).

ADDITIONAL DEEP TECHNICAL NOTES:

1. Credit Risk Components Deep-Dive

Probability of Default (PD) Modeling:

text
PD Estimation Approaches:

1. Historical Default Rates:
   PD = (Number_of_Defaults) / (Total_Exposures)

2. Statistical Models (Logistic Regression):
   Logit(PD) = ln(PD/(1-PD)) = β₀ + β₁X₁ + β₂X₂ + ... + β_nX_n

   Where X_i are borrower characteristics:
   - Financial ratios (debt/equity, interest coverage)
   - Credit bureau scores
   - Macroeconomic variables
   - Industry factors

3. Machine Learning Models:
   - Random Forest
   - Gradient Boosting (XGBoost)
   - Neural Networks

Loss Given Default (LGD) Modeling:

text
LGD Components:

1. Collateral Recovery:
   Recovery_Value = Collateral_Value × (1 - Liquidation_Cost)

2. Seniority:
   - Senior Secured: 0-20% LGD
   - Senior Unsecured: 30-50% LGD
   - Subordinated: 50-70% LGD
   - Junior: 70-100% LGD

3. Macroeconomic Factors:
   LGD = LGD_Base × (1 + α × (Unemployment - 5%))

4. Industry Factors:
   - Real Estate: Lower LGD (better collateral)
   - Technology: Higher LGD (less collateral)

Exposure at Default (EAD) Modeling:

text
EAD Calculation:

1. Term Loans:
   EAD = Outstanding_Principal

2. Revolving Facilities:
   EAD = Drawn_Balance + CCF × (Undrawn_Amount)

   CCF Factors:
   - Investment Grade: 10-30%
   - Non-Investment Grade: 40-60%
   - Distressed: 70-90%
   - Default: 100%

3. Derivatives:
   EAD = max(MTM, 0) + PFE

   Where PFE = Potential Future Exposure

2. Merton Model Deep-Dive

Merton Model Mathematics:

text
Black-Scholes-Merton Framework:

Equity Value:
E = V_A × N(d₁) - D × e^{-rT} × N(d₂)

Where:
d₁ = (ln(V_A/D) + (r + σ_A²/2)T) / (σ_A√T)
d₂ = d₁ - σ_A√T

Volatility Relationship:
σ_E × E = σ_A × V_A × N(d₁)

Distance to Default:
DD = (ln(V_A/D) + (r - σ_A²/2)T) / (σ_A√T)

Probability of Default:
PD = N(-DD)

Asset Value and Volatility:
Solve simultaneously:
E = V_A × N(d₁) - D × e^{-rT} × N(d₂)
σ_E × E = σ_A × V_A × N(d₁)

Merton Model Implementation Steps:

text
1. Collect:
   - Equity price and volatility (σ_E)
   - Debt amount (D) and maturity (T)
   - Risk-free rate (r)

2. Solve for V_A and σ_A:
   - Use numerical optimization (Newton-Raphson)
   - Initial guess: V_A = E + D, σ_A = σ_E

3. Calculate DD and PD:
   - DD = (ln(V_A/D) + (r - σ_A²/2)T) / (σ_A√T)
   - PD = N(-DD)

4. Calculate Credit Spread:
   - Risk-neutral PD = N(-d₂)
   - Credit Spread = -ln(1 - Risk_Neutral_PD)/T

3. Expected Credit Loss (ECL) Calculation

ECL Components:

text
12-Month ECL (Stage 1):
ECL = PD_12M × LGD × EAD

Lifetime ECL (Stage 2):
ECL = Σ_{t=1}^{T} PD_t × LGD_t × EAD_t × (1 + Discount_Rate)^(-t)

Stage Classification:
- Stage 1: Significant increase in credit risk (SICR)
- Stage 2: Credit-impaired
- Stage 3: Defaulted

IFRS 9 SICR Criteria:

text
Significant Increase in Credit Risk (SICR):

Quantitative Indicators:
1. PD increase > 100% for investment grade
2. PD increase > 50% for non-investment grade

Qualitative Indicators:
1. Negative changes in credit rating
2. Adverse economic outlook
3. Covenant breaches
4. Payment delays > 30 days

Backstop:
- 30+ days past due = Stage 2 (SICR)
- 90+ days past due = Stage 3 (Default)

4. Basel IRB Approach

IRB Capital Formula:

text
Capital = (EL_UL - EL) × EAD

Where:
EL_UL = LGD × N((N^{-1}(PD) + √ρ × N^{-1}(0.999)) / √(1-ρ))

ρ = 0.12 × (1 - e^{-50×PD}) / (1 - e^{-50}) + 0.24 × (1 - (1 - e^{-50×PD}) / (1 - e^{-50}))

Maturity Adjustment:
b = (0.11852 - 0.05478 × ln(PD))²

Maturity Factor:
MA = (1 + (M - 2.5) × b) / (1 - 1.5 × b)

Capital = EL_UL × MA × EAD

Asset Correlation by Asset Class:

 
 
Asset Class Correlation (ρ)
Corporate 0.12 × (1 – e^{-50×PD})/(1 – e^{-50}) + 0.24 × (1 – (1 – e^{-50×PD})/(1 – e^{-50}))
Sovereign 0.12 × (1 – e^{-50×PD})/(1 – e^{-50}) + 0.24 × (1 – (1 – e^{-50×PD})/(1 – e^{-50}))
Retail – Mortgage 0.03 × (1 – e^{-35×PD})/(1 – e^{-35}) + 0.16 × (1 – (1 – e^{-35×PD})/(1 – e^{-35}))
Retail – Revolving 0.03 × (1 – e^{-35×PD})/(1 – e^{-35}) + 0.16 × (1 – (1 – e^{-35×PD})/(1 – e^{-35}))
Retail – Other 0.03 × (1 – e^{-35×PD})/(1 – e^{-35}) + 0.16 × (1 – (1 – e^{-35×PD})/(1 – e^{-35}))

5. Credit Risk Portfolio Models

CreditMetrics Framework:

text
CreditMetrics Process:

1. Rating Migration:
   - Transition matrix (AAA to Default)
   - Probability of migration

2. Valuation:
   - Value portfolio under each rating state
   - Include correlations

3. Distribution:
   - Simulate thousands of scenarios
   - Generate loss distribution

4. Risk Metrics:
   - Credit VaR = Percentile - Expected Loss
   - Expected Shortfall = Average of worst cases

CreditRisk+ Framework:

text
CreditRisk+ Assumptions:

1. Each obligor has small exposure
2. Defaults are rare events
3. Portfolio independent

Model:
1. Calculate default intensities
2. Model as Poisson process
3. Aggregate to portfolio

Loss Distribution:
P(L = l) = Σ_{i=1}^{n} P(n_i defaults) × P(L = l | defaults)

 

PreviousNext