CFA® Level II – Fixed Income

Credit Analysis Models

www.irfanullah.co

Graphs, charts, tables, examples, and figures are copyright 2012, CFA Institute. Reproduced
and republished with permission from CFA Institute. All rights reserved.
 Contents and Introduction
1. Introduction

2. Measures of Credit Risk

3. Traditional Credit Models

4. Structural Models

5. Reduced Form Models

6. The Term Structure of Credit Spreads

7. Asset Backed Securities

www.irfanullah.co 2
 2. Measures of Credit Risk
• Probability of default (PD) The four credit measures might give different
ranks

• Loss given default (LGD) Present value of the expected loss is the most
precise measure because it quantifies the
difference between a risky bond and an otherwise
• Expected loss = PD x LGD identical and riskless government bond
 PD and LGD depend on state of the economy
Quantification could be in terms of a dollar
and company characteristics difference or a credit spread

Credit spread captures:

• Present value of expected loss 1. PD
 Accounts for risk of cash flows lost in default 2. LGD
Considers time value of money 3. Time value of money
4. Risk of cash flows lost in default

www.irfanullah.co 3
Example 1: Credit Measures

www.irfanullah.co 4
 3. Traditional Credit Models
Two traditional approaches: credit scoring and credit ratings

Credit scoring is used for small owner-operated businesses and individuals

Credit ratings are used for companies, sovereigns, sub-sovereigns, and those
entities’ securities, as well as asset-backed securities

Both approaches are widely used

Provide a link between traditional financial statement-based financial analysis

and more advanced credit risk models (structural form and reduced form)

www.irfanullah.co 5
 Credit Scoring
• Credit scores are generally applied to individuals and very small
businesses
• Credit scores provide an ordinal ranking
 Order riskiness from high to low
• Credit scores do not explicitly depend on current economic conditions
• Credit scores are not percentile rankings
 FICO score: high is good
• PD can vary for different kinds of loans
• Factors impacting credit score
 Timely bill payment
 Level of debt compared to credit limit
 Length of credit track record
 Number of recent credit applications
 Number of credit accounts

www.irfanullah.co 6
Example 2: Credit Scoring

www.irfanullah.co 7
 Credit Rating
Credit ratings rank the credit risk of a company, government (sovereign), quasi-government, or asset-backed
security. Credit ratings provide an ordinal ranking; do not give an estimate of the loan’s default probability.
Major credit-rating agencies include Moody’s, S&P and Fitch. Many institutions also produce internal ratings.

Strengths:
Ratings provide a simple statistic that summarizes a complex credit analysis of
a potential borrower.
Ratings tend to be stable over time and across the business cycle, which
reduces debt market price volatility.

Investment Grade
Weaknesses:
Ratings tend to be stable over time, which reduces the correspondence to a
debt offering’s default probability.
They do not explicitly depend on the business cycle, whereas a debt offering’s
default probability does.
With third-party ratings, the issuer-pays model for compensating credit-rating
agencies has a potential conflict of interest that may distort the accuracy of
credit ratings.

www.irfanullah.co 8
Example 3: Credit Ratings

www.irfanullah.co 9
 4. Structural Models
Structural models are based on the structure of a company’s balance sheet and rely on insights from
option pricing theory. Major vendors are Moody’s KMV and Kamakura Corporation.

Balance sheet of a simple

company at time T

www.irfanullah.co 10
 4.1 The Option Analogy
Owning the company’s debt is economically equivalent to owning a riskless bond that pays K
dollars with certainty at time T, and simultaneously selling a European put option on the assets of
the company with strike price K and maturity T.

Balance sheet of a simple

company at time T

www.irfanullah.co 11
 4.2 Valuation
To use the structural model to determine a company’s credit risk, we need assumptions that
enable us to explicitly value the implied call and put options. These assumptions are:
1. the company’s assets trade in frictionless markets that are arbitrage free
2. the riskless rate of interest, r, is constant over time
3. the time T value of the company’s assets has a lognormal distribution

www.irfanullah.co 12
 4.3 Credit Risk Measures

Credit risk measures can be calculated given the following inputs:

1. Time t asset value: At

2. Expected return on assets: u
3. Risk-free rate: r
4. Asset return volatility: σ
5. Face value of debt: K
6. Time to maturity of debt: T - t

www.irfanullah.co 13
Example 4: Interpreting Structural Model Credit Risk Measures

www.irfanullah.co 14
www.irfanullah.co 15
 4.4 Estimation
• Two ways of estimating the parameters of an option pricing model
 Historical
 Implicit

• For the structural model, historical estimation can not be used because a company’s assets are not
traded in frictionless markets and the value is not observable

• We need to use the implicit model, but…

If the model’s assumptions are not reasonable approximations of the market’s actual structure,
then the implicit estimate will incorporate the model’s error and not represent the true parameter.
This bias will, in turn, introduce error into the resulting probability of default and the expected
loss, thereby making the resulting estimates unreliable.

www.irfanullah.co 16
 Structural Model’s Strengths and Weaknesses
Structural Model Strengths:
1. It provides an option analogy for understanding a company’s default probability and recovery rate.
2. It can be estimated using only current market prices.

Structural Model Weaknesses:

1. The default probability and recovery rate depend crucially on the assumed balance sheet of the
company, and realistic balance sheets cannot be modeled.
2. Its credit risk measures can be estimated only by using implicit estimation procedures because the
company’s asset value is unobservable.
3. Its credit risk measures are biased because implicit estimation procedures inherit errors in the
model’s formulation.
4. The credit risk measures do not explicitly consider the business cycle.

The most fundamental issue with the structural model is that the assumptions are too unrealistic.

www.irfanullah.co 17
 5. Reduced Form Models
Reduced Form Model Assumptions
1. At least one of the company’s liabilities, a zero-coupon bond, trades in frictionless
and arbitrage-free markets
2. Risk-free rate is stochastic
3. The relevant state of the economy can be described by a vector of macroeconomic
state variables
4. Default intensity depends on the state of the economy
5. In a given state of the economy, whether the company defaults depends on
company-specific considerations
6. If default occurs, debt is worth only [1 – t(Xt)] of its face value. Here, [1 – t(Xt)] is
the percentage recovery rate on the debt in the event of default

These assumptions are very general, allowing the default probability and the loss given default to
depend on the business cycle. Given a proper specification of the functional forms for the default
intensity and loss given default and stochastic processes for the spot rate of interest and the
macroeconomic variables, they provide a reasonable approximation of actual debt markets.

www.irfanullah.co 18
 5.1 Valuation
For a zero-coupon bond with face value, K, the value of debt is given by:

“The study of more complex specifications is outside the scope of this reading and left
for independent reading.”!!!

www.irfanullah.co 19
 5.2 Credit Risk Measures
For a zero-coupon bond with face value, K:

www.irfanullah.co 20
Example 5: Interpreting Reduced Form Credit Risk Measures

www.irfanullah.co 21
www.irfanullah.co 22
 5.3 Estimation
As with structural models, implicit estimation is possible

Unlike structural models, historical estimation is a also possible

 Use a technique called hazard rate estimation to estimate the probability of a binary event
 Logistic regression

Qualitative dependent variables are dummy variables used as dependent variables instead of as independent
variables.
Linear regression is not appropriate in these situations. We should use probit, logit, or discriminant analysis. …. The
logit model is based on the logistic distribution rather than the normal distribution.

www.irfanullah.co 23
Using the Logistic Regression Model

www.irfanullah.co 24
Example 6: Using the Logistic Regression Model

www.irfanullah.co 25
 Strengths and Weaknesses of the Reduced Form Model

Reduced Form Model Strengths:

1. The model’s inputs are observable, so historical estimation procedures can be used for the credit
risk measures.
2. The model’s credit risk measures reflect the changing business cycle.
3. The model does not require a specification of the company’s balance sheet structure.

Reduced Form Model Weaknesses:

1. Hazard rate estimation procedures use past observations to predict the future. For this to be valid,
the model must be properly formulated and back tested.

www.irfanullah.co 26
 5.4 Comparison of Credit Risk Models
All three models have been empirically evaluated with respect to their accuracy in measuring a debt
issue’s default probability.

Credit ratings are the least accurate predictors. This is because credit ratings tend to lag changes in a
debt issue’s credit risk because of rating agencies’ desire to keep ratings relatively stable over time, and
consequently, they are relatively insensitive to changes in the business cycle.

Reduced form models perform better than structural models because structural models are computed
using implicit estimation procedures whereas reduced form models are computed using historical
estimation (hazard rate procedures). The improved performance is due to the flexibility of hazard rate
estimation procedures—that is, both their ability to incorporate changes in the business cycle and their
independence of a particular model specifying a company’s balance sheet structure.

www.irfanullah.co 27
 6. The Term Structure of Credit Spreads
The term structure of credit spreads corresponds to the spread between the yields on default-free and
credit risky zero-coupon bonds. In practice, because coupon bonds (rather than zero-coupon bonds)
often trade for any given company, to compute the credit spreads one first needs to estimate the zero-
coupon bond prices implied by the coupon bond prices.

www.irfanullah.co 28
 Credit Spread and Expected % Loss/Year

Estimate of expected
percentage loss per year:

www.irfanullah.co 29
Example 7: Estimate of Expected Percentage Loss per Year

www.irfanullah.co 30
Example 8: Present Value of Expected Loss

www.irfanullah.co 31
www.irfanullah.co 32
www.irfanullah.co 33
www.irfanullah.co 34
www.irfanullah.co 35
 Determinants of the Term Structure of Credit Spreads

In practice the credit spread is determined by the expected percentage loss (from
structural and reduced form models) AND a liquidity risk premium

Credit Spread = yD(t,T) – yP(t,T) = E(Percentage loss) + Liquidity premium

Liquidity premiums are positive because sovereign government bonds trade in more
liquid markets than do most corporate bonds

www.irfanullah.co 36
 7. Asset-Backed Securities
ABS are complex credit derivatives and are
classified by the loans in their collateral pool

A representative ABS structure is shown on the

right

Unlike corporate debt, an asset-backed security does not default when an interest payment is missed. A
default in the collateral pool does not cause a default to either the SPV or a bond tranche. For an ABS, the
bond continues to trade until either its maturity date or all of its face value is eliminated because of the
accumulated losses in the collateral pool or through early loan prepayments.

The credit risk measures used for corporate or sovereign bonds can be applied: probability of loss,
expected loss, and present value of the expected loss.

www.irfanullah.co 37
 Summary
• Probability of default, loss given default, expected loss, present value of expected
loss
• Credit scoring
• Credit rating
• Structural models
 Equity can be viewed as a call option on assets
 Assumptions, strengths, weaknesses
• Reduced form models
 Assumptions, strengths, weaknesses
• Term structure of credit spreads
• Present value of expected loss over a given time horizon
• Credit analysis of asset backed securities

www.irfanullah.co 38
 Conclusion
• Read the summary

• Review learning objectives

• Examples

• Practice problems

• Practice questions from other sources

www.irfanullah.co 39