Introduction to Credit
Risk Modeling and
Assessment
Source - Analytical Techniques in the Assessment of Credit Risk: An Overview of Methodologies
and Applications, by Micha/is Doumpos, Christos Lemonakis, Dimitrios Nik/is, and Constantin
Zopounidis.
1
by Aishwarya Nair
Dissecting the CAMEL Framework: How Banks Are Rated on
Health & Resilience
🧱 What is CAMEL? Q1: Why do you think it's
important for a bank to
Q3: What role does a bank’s management play in
its financial health?
CAMEL is an acronym for the maintain a certain level of
five key areas used to evaluate A3: Management makes decisions that directly
capital reserves?
the health and performance of affect the bank’s risk level and performance—like
a bank: A1: Capital reserves serve as a setting policies, forming business strategies, and
financial cushion to absorb managing internal controls.
C – Capital Adequacy losses and protect depositors.
Q4: How do stable earnings help a bank survive
A – Asset Quality Regulators require banks to
tough economic times?
M – Management maintain a minimum capital to
E – Earnings ensure they can withstand A4: Stable and sufficient earnings help a bank
L – Liquidity financial shocks and continue cover its expenses, build capital, and handle losses
operating safely. during economic downturns.
Each component is scored
from 1 to 5, with 1 = strong and Q2: What might happen if a Q5: Why is liquidity important for a bank on a daily
5 = critically deficient. These large portion of a bank’s loans basis?
scores help determine go unpaid?
A5: Banks need to have enough liquid assets to
supervisory actions and risk
A2: If many loans default, the meet short-term obligations—like customer
ratings.
bank suffers losses, which can withdrawals or loan demands. Poor liquidity can
deplete its capital and threaten lead to a crisis.
its solvency. This indicates poor
asset quality.
1
by Aishwarya Nair
Cracking the Credit Risk Equation: PD, LGD, EAD & EL
Demystified
Q1: What do you think a bank wants to estimate 1. Expected Loss (EL) – The predicted amount a bank could lose from
when it lends money to someone? a loan, combining PD, EAD, and LGD.
2. Probability of Default (PD) – Likelihood that a borrower will default
A1: It wants to estimate the expected loss—i.e.,
(miss payments for 90+ days).
how much money it might lose if the borrower
defaults. 3. Exposure at Default (EAD) – The outstanding loan amount at risk at
👉 Formula: the time of default.
4. Loss Given Default (LGD) – The percentage of the loan expected to
Expected Loss = Probability of Default × be lost if default occurs.
Exposure at Default × Loss Given Default 5. Recovery Rate (RR) – The portion of the loan expected to be
recovered after default (RR = 100% – LGD).
1
by Aishwarya Nair
From Risk to Ratio: How RWAs (risk weighted assets) Shape
Capital Adequacy
Q1: Why do you think banks are required to hold capital in proportion to Q2: What might be the pros and cons of
their risk? using fixed risk weights prescribed by
regulators versus customized modeling?
A1: Banks face various risks from lending. Holding sufficient capital acts as a
buffer to absorb potential losses and protects depositors and the financial A2:
system.
Standardized Approach: Simple and
👉 The Capital Adequacy Ratio (CAR) is a measure of a bank’s capital consistent, but lacks bank-specific
relative to its risk exposure: accuracy. Uses external data and
regulatory weights.
IRB (Internal Ratings-Based) Approach:
More accurate and bank-specific but
complex and resource-intensive.
Where α is the regulatory minimum:
8% under Basel II
10.5% under Basel III
1
by Aishwarya Nair
💡 Internal Ratings-Based (IRB) Approach in
Depth
💡Now, how is K calculated?
Here’s the formula to calculate RWA for a loan
using the ASRF (Asymptotic Single Risk Factor)
model:
N: Standard normal cumulative distribution function
N −1: Its inverse
P D, LGD, R, M : As previously defined
β : Maturity adjustment
Where:
Q4: Why is the capital requirement adjusted by a maturity factor β ?
K = capital requirement per unit of EAD, based
on PD, LGD, M, R A4: Longer loan maturities increase the probability of
EAD = Exposure at Default downgrade/default. Thus, capital buffers should be larger for longer-
term exposures.
Q3: What is the purpose of multiplying by 12.5 in
the RWA formula?
A3: Since capital requirements are usually
expressed as a % of RWA, multiplying by 12.5 is
Lower PD → higher β → higher capital requirement
equivalent to dividing by 8% (1 ÷ 0.08), converting
capital requirement to RWA units.
1
by Aishwarya Nair
How Do We Predict Default? From Expert Judgment to Quant
Models
❓Q1: Why might we need expert judgment to ❓Q3: What if we do have a lot of data? Could we let that data guide
evaluate credit risk? our default predictions?
A1: Sometimes, especially in cases where historical A3: Yes! That’s the role of Empirical Models. They use large datasets
or quantitative data is scarce (e.g., a new borrower or of loan performance to find patterns using statistics or machine
project finance), it’s necessary to rely on experienced learning.
credit analysts who can make qualitative
assessments based on non-numeric cues.
❓Q4: Can empirical models find new or emerging risk factors?
👉 This is known as the Judgmental Approach, also A4: Yes. For example, using machine learning, a bank might discover
that corporate governance practices or social media sentiment
called the qualitative approach or expert system.
correlates with default risk—factors a human expert may overlook.
❓Q2: When might this approach be most useful? ❓Q5: Can default risk be modeled using economic theory?
A2: When there is no reliable historical data—like in
A5: Absolutely. That’s the idea behind Financial Models, also known
new project financing or lending to startups—
as market models. They’re rooted in financial theory and used mostly
judgment from experienced analysts can be
for corporate borrowers.
invaluable.
👉 Example: Lending to a startup with no credit ❓Q6: When would financial models be more appropriate than
empirical ones?
history. An expert might rely on the founder's past
track record, industry potential, and business plan A6: For market-traded corporate borrowers, especially where real-
quality. time bond or derivative data is available. Financial models offer
deeper insights into how firm value and market events drive default
risk.
1
by Aishwarya Nair
Inside the Merton Model: Default Distance, Probability & Pitfalls
❓Q1: How do you think we can use option pricing to evaluate a The value of equity (E) is given by:
company’s default risk?
A1: The Merton model treats a firm’s equity as a call option on its
assets. If the value of the firm’s assets (A) at the debt’s maturity
Where:
(T) is more than the debt value (L), shareholders "exercise" the
option and repay the debt. If not, the firm defaults.
👉 The logic is:
If AT > L → firm repays
If AT ≤ L→ firm defaults
So, equity is like a call option with:
Strike price = L (face value of debt)
Underlying asset = A (firm's assets)
1
by Aishwarya Nair
❓Q2: What variables are needed to apply the model? These unknowns are found by solving the system of
equations numerically (simultaneously using the Black-
A2:
Scholes formula and equity volatility formula):
Known Inputs:
E : Market value of equity (e.g., firm’s market cap)
σE : Equity volatility (from market data)
T : Time to maturity of the debt
L: Face value of debt (short-term or total liabilities)
r: Risk-free rate (e.g., U.S. Treasury rate)
Unknowns to estimate:
A: Firm’s asset value
σA: Asset volatility
1
by Aishwarya Nair
❓Q3: How do we measure the likelihood that the firm’s ❓Q4: How far are a company’s assets from the “default
asset value falls below its debt? point” in terms of standard deviations?
A3: That’s the probability of default (PD), and it’s calculated A4: That’s the Distance to Default (DD), calculated as:
as:
Risk-neutral PD:
P D = N(−d2)
This assumes the firm’s assets grow at the risk-free rate.
It tells you how many standard deviations the firm's asset
Real-world PD: value is above its debt level.
Replace r with μ (expected return on assets):
Higher DD → Lower PD
Lower DD → Higher PD
1
by Aishwarya Nair
Q5: What do you think are some limitations of the Merton ✅ Summary: Steps to Use Merton Model
Model?
1. Get market data for E and σE
A5: Great question. The model has several limitations: 2. Estimate A and σAby solving the two equations
1. Assumes assets follow a lognormal distribution (may not 3. Compute d1and d2
be realistic). 4. Calculate:
2. Assumes one class of debt maturing at the same time. PD = N(−d2)
3. Assumes perfect markets (no transaction costs or taxes). DD = value of the PD formula argument
4. Asset value and volatility aren’t directly observable, 5. Interpret results in risk terms
requiring estimation.
5. Static balance sheet – ignores the firm's evolving debt
structure and operational flexibility.
6. Doesn’t handle early default – assumes default only at
maturity.
1
by Aishwarya Nair
Model Showdown: Merton vs CreditRisk+ vs CreditMetrics vs
Moody’s-KMV
🧾 Summary Table – Model Comparison
Model Method Type Data Needed What It Pros Cons
Measures
Merton Structural Market value of Default as option Theory-based, Needs
assets & equity exercise clear link to unobservable
capital structure inputs, assumes
1 debt maturity
Moody’s-KMV Empirical + Historical default Expected Default Realistic default Still requires
Structural data Frequency threshold, uses firm-level data
real data
CreditMetrics Transition Matrix Ratings, yield Value impact of Full portfolio risk Rating
spreads, migrations tracking, market- assumptions,
matrices based complex for
long-term loans
CreditRisk+ Statistical Default Default event Works with Ignores credit
frequency data count minimal data, rating, only
Poisson model is models binary
robust outcomes
1
by Aishwarya Nair
RAROC in Action: Measuring Credit Returns Beyond the Surface
❓Q1: Why do banks need to evaluate a loan based not just ❓Q2: What components make up the revenue a bank earns
on returns, but on risk? on a loan?
A1: Because loans carry credit risk, and profitability should A2: The formula lists:
consider not just the revenue a loan generates, but the
s: Spread (difference between loan rate and bank’s cost
amount of capital at risk due to potential losses. RAROC
of capital)
measures return relative to the risk taken.
✅ Definition of RAROC:
f : Fees earned from the loan
l: Expected loan losses
c: Operating costs
x: Tax rate
👉 Formula for loan revenues:
If RAROC > cost of capital, then the loan is profitable.
1
by Aishwarya Nair
❓Q3: How much capital does a bank need to set aside for a loan? ❓Q4: Why is tax x included in the revenue
formula?
A3: It depends on the method. The two approaches are:
A4: Because taxes reduce the net return to
A. Market-based (Value change method):
the bank. We multiply gross profits by (1 − x)
to get after-tax income, which is what
ultimately matters for RAROC.
Where:
❓Q5: When should you use the unexpected
loan loss approach over the interest rate
L: loan value change method?
D: duration A5: Use it when:
i: interest rate
You have historical data for defaults and
Δi: change in interest rates losses.
B. Unexpected loss approach: You want to measure credit risk more
directly, especially in a portfolio context.
You’re less concerned with interest rate
sensitivity and more with credit
α: confidence factor (e.g., 2.33 or 2.66) performance.
LGD: Loss Given Default
EAD: Exposure at Default
This is useful when historical data is available to model tail-risk more
robustly.
1
by Aishwarya Nair
📌 Key Concept 💡 Takeaway
CAMEL Framework Evaluates a bank’s health across Capital, Asset quality, Management, Earnings, and
Liquidity. Think of it as a report card for bank resilience.
Credit Risk Drivers Quantified through PD, LGD, EAD — the “Credit Risk Trinity” feeding into Expected Loss.
Risk-Weighted Assets (RWA) RWAs adjust for riskiness in assets and form the backbone of regulatory capital under
Basel norms.
Modeling Approaches From expert judgment to empirical ML models and financial-theory-based models like
Merton. Each has its moment.
Merton Model Views equity as a call option on firm assets. Powerful for traded firms, but beware of
assumptions and estimation limitations.
Model Comparisons We saw how CreditMetrics, CreditRisk+, and Moody’s-KMV differ in philosophy and data
needs — no one-size-fits-all.
RAROC A risk-adjusted lens on profitability. Because returns without risk context can be
misleading.
1
by Aishwarya Nair