IOSR Journal of Economics and Finance (IOSR-JEF)

e-ISSN: 2321-5933, p-ISSN: 2321-5925.Volume 13, Issue 3 Ser. VI (May. – June. 2022), PP 28-39
www.iosrjournals.org

Credit Risk Modeling Using Default Models: A Review

George Jumbe 1, Ravi Gor2
1
Research scholar, Department of Mathematics, Gujarat University
2
Department of Applied Mathematical Science and Analytics, Gujarat University

Abstract: Credit risk, also known as default risk, is the likelihood of a corporation losing money if a business
partner defaults. If the liabilities are not met under the terms of the contract, the firm may default, resulting in
the loss of the company. There is no clear way to distinguish between organizations that will default and those
that will not prior to default. We can only make probabilistic estimations of the risk of default at best. There are
two types of credit risk default models in this regard: structural and reduced form models. Structural models are
used to calculate the likelihood of a company defaulting based on its assets and liabilities. If the market worth of
a company's assets is less than the debt it owes, it will default. Reduced form models often assume an external
cause of default, such as a Poisson jump process, which is driven by a stochastic process. They model default as
a random event with no regard for the balance sheet of the company. This paper provides a Review of credit risk
default models.
Key words: Credit Risk, Default Models, Structural Models, Reduced-form Models, Poisson jump process.
---------------------------------------------------------------------------------------------------------------------------------------
Date of Submission: 10-06-2022 Date of Acceptance: 25-06-2022
---------------------------------------------------------------------------------------------------------------------------------------

I. Introduction
The uncertainty about a company's ability to service its debts and commitments is known as credit risk
or default risk. It's the risk of a loss occurring as a result of a borrower's failure to repay a loan or meet
contractual obligations. It refers to the likelihood of a corporation losing money if a business partner defaults. If
the liabilities are not met under the terms of the contract, the firm may default, resulting in the loss of the
company. Credit, commerce, and investment operations, as well as the payment system and trade settlement, all
result in liabilities. Credit risk modelling is difficult due to the fact that company default is not a common
occurrence and usually comes unexpectedly. However, when a creditor defaults, it frequently results in
significant losses that cannot be predicted in advance, therefore effectively measuring and managing credit risk
can reduce the severity of a loss43. Along with market risk and operational risk, credit risk is one of three key
hazards that all financial markets must report and retain capital against. It shows the likelihood of the company
losing money if a business partner fails. We can attribute this failure to a failure to meet contractual duties,
which results in a company loss34. According to Bielecki and Rutkowski, Credit risk has three components: (i)
default risk, which is the risk that the issuer or counterparty will fail to honor the terms of the obligation stated
in a financial contract; (ii) spread risk, which is the risk of loss or underperformance of an issue or issues due to
an increase in the credit spread, and (iii) downgrade risk, which is the risk of credit ratings deterioration 5.

Default models
Default models employ market data to model the occurrence of a default event. They are created by
financial organizations to estimate the likelihood of a corporate or sovereign entity defaulting on its credit
obligations. These models have evolved into two distinct types of models: structural and reduced form models17.

Structural models
This is the first group of default models that look at the structure of the company's capital and are based
on its value. Merton (1974) pioneered structural models, which use the Black-Scholes option pricing framework
to characterize default behavior. They're used to figure just how likely a company is to default based on the
value of its assets and liabilities. They make the assumption that they have complete knowledge of a company's
assets and liabilities, leading in a predicted default time. These models indicate that default risks arise when the
value of a company's assets falls below its outstanding debt at the maturity date. These models are designed to
show a direct link between default risk and capital structure. The fundamental disadvantage of this strategy is
that it ignores the market value of a company's assets and treats debt as an option on those assets, and the default
event is predictable10,39,44.

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 28 | Page

 Credit Risk Modeling Using Default Models: A Review

Reduced form models

The reduced form models are the second type of default models. The relationship between default and
firm value is not explicitly included in these models. Defaulting is viewed as an unanticipated occurrence that
can be influenced by a variety of market circumstances. Reduced form models often assume an external cause of
default, such as a Poisson jump process, which is driven by a stochastic process. They model default as a
random event with no regard for the balance sheet of the company. A Poisson event is the term used to describe
this type of random event. This technique to credit risk modelling is also known as default intensity modelling
because Poisson models look at the arrival rate, or intensity, of a certain event. The probability or intensity of
default, as well as the mean recovery rate, are calculated using reduced-form credit risk models, which employ
the observed market credit spread1. The Jarrow and Turnbull27 model, which uses multi-factor and dynamic
analysis of interest rates to compute the probability of default, is one of the first reduced-form models.

II. Literature review

Structural models
Merton (1974) calculated a company's credit risk by imagining its stock as a call option on its assets.
Merton characterized a firm's asset value as a lognormal process and assumed that if the asset value fell below a
specific default boundary, the firm would default. The default option was available only once, at maturity. The
firm's equity was created via a call option on the underlying assets. The benefit of this model is that it can be
used for any publicly listed company and that data from the stock market may be used instead of financial data.
It can also be used to predict what will happen in the future. The use of this approach in everyday practice, on
the other hand, highlighted some of its flaws. The model's credit spreads, which are premiums to risk-free
interest rates, are typically lower than the real spreads. The assumptions of Merton's model scarcely resemble
reality. Previous experience indicates that the company would struggle to pay its debts for a long time before the
value of its assets falls below the value of its liabilities41. Extensions to the Merton model can help to overcome
some of the model's flaws. The most well-known and commonly utilized is Keaholfer, McQuown, and Vasicek
(KMV), which was discovered in 1974 based on Merton's bond pricing model assumptions48.
The Merton model is based on the idea of considering a company's equity as a call option on its assets, allowing
Black-Scholes option pricing methods to be used. Assume a corporation has an asset A t at time t financed by
stock equity E t and zero-coupon debt D t with a face amount of K maturing at time T  t , and a capital
structure determined by the balance sheet relationship:
At  E t  D t
(1)
The company defaults on its debt at T if A T  K . In the case A T  K the company’s debtholders can be paid
the full amount K . Shareholders’ equity value is given by A T  K . Therefore, equity value at time T can be
written as:

E T  m a x  AT  K , 0  (2)

The asset value A T is assumed to follow a geometric Brownian motion (GBM) process, with risk-neutral
dynamics given by the stochastic differential equation:

d At
 rdt   AdW t (3)
At

where W t is a standard Brownian motion under risk-neutral measure, r denotes the continuously compounded
risk-free interest rate, and  A
is the asset’s return volatility. Applying the Black-Scholes formula for European
call option we obtain:

 r T  t 
E t  At N  d 1   K e N d2  (4)

where N    denotes the N  0 ,1  cumulative distribution function, with the quantities d 1 and d 2 given by:

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 29 | Page

 Credit Risk Modeling Using Default Models: A Review

 A   1 
ln  t    r    T  t 
2
A
 K   2 
d1  , (5)
 A
T t

 A   1 
ln  t    r    T  t 
2
A
 K   2 
d2  . (6)
 A
T t

The continuously compounded credit spread s is given by:

1  At r T  t  
s   ln

N d2   e N   d1 
T t  K 
(7)

Equation (7) allows us to solve for credit spread when asset level and return volatility ( A t and  A ) are
available for given t, T, K, and r . To get A t and  A
we need to assume another geometric Brownian motion
model for equity price E t and applying Ito’s Lemma to show that instantaneous volatilities satisfy:

Et
At A
 E t E
. (8)
 At

Black-Scholes call option delta can then be substituted into (8) to obtain:

At  A N  d 1   E t  E
(9)

where equity price E t and its return volatility  E

are observed from equity market. Finally, (4) and (8) can be
solved simultaneously for A t and  A , which are used in (7) to determine credit spread s.
By modelling the evolution of firm value as a jump-diffusion process, Zhou (1997) created a new structural
approach to estimating credit risk and evaluating default-risky instruments. A firm can default instantly due to a
quick reduction in its value in a jump-diffusion process. They also linked recovery rates to firm valuation upon
default, resulting in endogenous variance in recovery rates in the model. The findings suggested that both the
diffusion and jump processes could be key components of a structural debt valuation model. They proposed that
their paper's valuation framework be expanded to include more institutional elements such as floating rate
coupon payments and bond indenture provisions that may require a firm to return its lenders recovered values at
default time if the bond defaults before maturity. He built a continuous-time valuation framework for hazardous
debt by extending the Merton-Black-Cox-Longstaff-Schwartz approach and modelling the evolution of firm's
value as a jump-diffusion process54.
Crosbie and Bohn 14, Kealhofer33, and Vasicek51 devised the MKMV technique, which uses the Vasicek-
Kealhofer (VK) model to offer a term-structure of physical default risk probability. This model considers equity
to be a perpetual down-and-out option on a company's fundamental assets. Short-term obligations, long-term
liabilities, convertible debt, preferred equity, and common equity are among the five categories of liabilities that
can be accommodated under this model. MKMV derives a firm's market value of assets and associated asset
volatility using the option-pricing equations derived in the VK framework. Empirically, the default point term-
structure is determined. MKMV creates a Distance-to-Default (DD) term-structure by combining market asset
value, asset volatility, and the default point term-structure. Using an empirical mapping between DD and
historical default data, this term structure is transformed to a physical default probability.

 A   1 
  
2
lo g  T
 XT   2 
D DT  (1)
 T

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 30 | Page

 Credit Risk Modeling Using Default Models: A Review

The interpretation of X T in the VK model differs slightly from that of the Merton model. If asset value A falls
below X T at any point in time during the analytical stage of the model, the firm is regarded to be in default.
The VK model estimates a term-structure of this default barrier in the DD-to-EDF empirical mapping step,
resulting in a DD term structure that may be transferred to a default-probability term-structure, hence the
subscript T for default barrier X . According to Huang and Huang22, the default probability created by the
MKMV implementation of the VK model is known as the Expected Default Frequency or EDF credit measure.
After obtaining the EDF term structure, a cumulative EDF term structure up to any term T referenced to as
C E D F T can be calculated. This is then converted to a risk-neutral cumulative default probability C Q D F T
using the following equation:

 C E D FT     sqrt  R   s q r t  T  
1
C Q D FT  N  N
2


(2)

where R is the square of correlation between the underlying asset returns and the market index returns, and 
2

is the market Sharpe ratio. The spread of a zero-coupon bond is obtained as:

1
s   lo g 1  L G D  C Q D F T 
T
(3)
where LGD stands for the loss given default in a risk-neutral framework. The floating leg of a simple CDS (i.e.
a single payment of LGD paid out at the end of the contract with a probability of C Q D F T ) can also be
approximated with this relationship.
Jarrow et al. conducted a robust test of Merton's structural credit risk model that did not rely on either estimated
firm value parameters or projected default probabilities. They also used the Merton model to provide a test for
the consistency of observed changes in debt and equity prices (positive or negative changes). The data
significantly refuted Merton's structural model for all enterprises studied and for all debt difficulties considered.
Only the monitoring of equity prices, debt prices, and the spot rate of interest was required for their testing
approach. Over the period of February 6, 1992, to March 12, 2001, they tested the Merton model on five distinct
companies' debt concerns in various sub-intervals, using both weekly and monthly observation intervals.
However, because they did not look into the consequences of its extensions and generalizations, their testing
methods may be used to look into these as well. They emphasized that Merton's model presupposes complete
markets that are frictionless, competitive, and arbitrage-free, and that the firm's balance sheet can be stated as31:

Vt  Dt T , F  Et
(1)

where V t is the value of the firm’s assets at time t assumed to follow a diffusion process, D t is the value of the
firm’s debt at time t, and E t is the value of the firm’s equity at time t. The debt is assumed to be a zero-coupon
bond with maturity T and face value F. Let rt denote the (default free) spot rate of interest at time t, assumed to
be non-random. Using the call option analogy to the firm’s equity, Merton (1974) develops a pricing formula for
debt D t  D t  V t , rt : T , F  that depends on the firm’s value (and the parameters of its stochastic process, the
volatility), the characteristics of the debt (face value and maturity date), and the spot rate of interest ( rt ). Let
d V t V t   d t   d W t with W t a standard Brownian motion. Then,

T

t ru d u
D t  V t N  h1   F e N  h2  (2)

where

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 31 | Page

 Credit Risk Modeling Using Default Models: A Review

  1
T

 ru d u
Vt    T  t
2
ln  F e t
  2
h1  (3)
 T  t

h 2   h1 T  t (4)

N    is the cumulative standard normal distribution function.

Using Merton's model, Hull et al. investigated credit risk and volatility skews. They suggested a method for
estimating the model's parameters using implied volatilities of stock options. They compared their
implementation of Merton's model against the standard way to implementation using data from the credit default
swap market. Their proposed Merton model implementation outperformed the approach's standard
implementation 24.
Elizalde examined the structural approach to credit risk modelling, taking into account both the situation of a
single firm and the scenario of firms with default dependencies. They looked at the Merton (1974) model and
the first passage models (FPM) (Black and Cox 1976) in the single company situation, assessing its main
properties and extensions. Finally, structural models with state-dependent (SDM) cash flows or debt coupons
were examined. They discovered that, according to Merton's model, a company fails if its assets are less than its
outstanding debt at the time of debt servicing. Defaults in the FPM technique occur when a firm's asset value
falls below a specified threshold. Unlike Merton's approach, default can happen at any time. FPM defines
default as the first time a firm's asset value falls below a certain threshold, allowing default to occur at any time.
SDMs presume that some of the characteristics affecting a firm's ability to generate cash flows or its funding
costs are state dependent, with states representing the economic cycle or the firm's external rating. The empirical
testing of FPM and structural models in general, on the other hand, has not been very fruitful. They agree that
structural bond pricing models do not effectively price corporate bonds, based on estimations from
implementations, because they present the predictability of defaults and recovery rates, which is not true in real
market conditions17.
In the Black-Scholes-Merton paradigm, Kulkarni et al. modeled default probabilities and credit spreads for
selected Indian enterprises. Over the sample period, they found that the objective probability estimates are
greater than the risk-neutral estimates. The model's output performed well when compared to the Altman Z-
score measure. The model, however, did not produce spreads as wide as those seen in the corporate bond
market35.
Tarashev compared the probability of default (PDs) generated by six structural credit risk models to ex post
default rates to assess their performance experimentally. The paper uses firm-level data, in contrast to other
studies seeking similar goals, and demonstrates that theory-based PDs tend to closely match the actual level of
credit risk and explain for its time path. Simultaneously, non-modeled macro variables from the financial and
real sides of the economy aid in significantly improving default rate estimates. Findings show that theory-based
PDs do not adequately reflect credit risk's reliance on business and credit cycles. The majority of the optimistic
conclusions about PD performance are attributed to models with endogenous default. Exogenous default
frameworks, on the other hand, have a tendency to underestimate credit risk 47.
Using term structure of credit default swap (CDS) spreads and stock volatility from high-frequency return data,
Huang and Zhou23 investigated specification analysis of structural credit risk models. Based on the simultaneous
behavior of time-series asset dynamics and cross-sectional pricing errors, they provide consistent econometric
estimate of pricing model parameters and specification tests. The conventional Merton 39 model, the Black and
Cox6 barrier model, and the Longstaff and Schwartz37 model with stochastic interest rates are all significantly
rejected by their empirical testing. The double exponential jump-diffusion barrier model (Huang and Huang22)
outperforms the other two models significantly. The stationary leverage model of Collin-Dufresne and
Goldstein, which we cannot reject in more than half of our sample firms, is the best of the five models studied 13.
However, empirical data show that conventional structural models, particularly for investment grade names, are
unable to represent the dynamic behavior of CDS spreads and equity volatility. Given that equity volatility in
structural models is time-varying, this finding provides direct evidence that using a structural model with
stochastic asset volatility (as in Huang and Huang22, Huang21, Zhang, Zhou, and Zhu53) can significantly
improve model performance, particularly for investment-grade names. This suggests that time-varying asset
volatility, which is not included in typical structural models, could play a role. Although these five models have
different economic assumptions, they can all be integrated in the same underlying structure that comprises asset
process, default boundary, and recovery rate requirements for the underlying company.
Let V be the firm's asset process, K the default boundary, and r the default-free interest rate process. Assume
DOI: 10.9790/5933-1303062839 www.iosrjournals.org 32 | Page
 Credit Risk Modeling Using Default Models: A Review

that, under a risk-neutral measure,

 Nt 
Q

dVt
  rt   dt   vdWt  d    Z i  1    dt
Q Q Q Q
(1)
Vt  i  1 

d ln K t  K l   V    rt     ln  K t V t   d t (2)

d rt      rt  d t   r d Z t
Q
(3)

where  ,  v , K l , V ,  ,  ,  ,  r and     are constants, and W

Q Q
and Z are both one- dimensional
standard Brownian motion under the risk-neutral measure and are assumed to have a constant correlation
coefficient of  . In Eq. (1), the process N is a Poisson process with a constant intensity   0 ¸ the Z i
Q Q Q
's

 ln  Z 1 
Q Q
are i .i .d random variables, and Y has a double-exponential distribution with a density given by:

Q
 d y
Q
 u y
 y  pu  u e 1 y  0   p d  d e
Q Q Q Q
f
Y
Q 1 y  0  (4)

In equation (4), parameters  u ,  d  0 and p u , p d  0 are all constants, with p u  p d  1 .The mean
Q Q Q Q Q Q

percentage jump size 

Q
is given by:

pu  u pd  d
Q Q Q Q

eY  1  Q
Q

  E  Q 1
Q Q
(5)
   1  1
u d

All five models considered in this analysis are special cases of the general specification in Equations (1) to (3).
For instance, if the jump intensity is zero, then the asset process is a geometric Brownian motion. If both  and
 r
are zero, then the interest rate is constant, an assumption made in the three one-factor models. They assume
a constant recovery rate for comparison with other studies and because the CDS database that they used includes
the recovery rate estimates. Under each of the five structural models, it is straightforward to calculate the CDS
spread. Let Q  0 , T  denote the survival probability over  0 , T  under the T -forward measure. Then the CDS
spread of a T-year CDS contract is given by:

 

T
1  RE e x p   r  u  d u I   T 
Q

 0

cds  0,T   4T
(6)
 D  0 , Ti  Q  0 , Ti  4
i 1

where R is the recovery, r is the interest rate process, D  0 ,   the default-free discount function,  the default

   the expectation under the risk-neutral measure. To simplify the

Q
time, and I   the indicator function, and E
computation, they followed the literature to make the standard assumption that the settlement of the contract
occurs on the next payment day. It then follows from Equation (6) that:

4T
1  R D  0 , T i   Q  0 , T i  1   Q  0 , T i  
cds  0,T  i 1
 4T
(7)
 D  0 , Ti  Q  0 , Ti  4
i 1

As a result, the implementation of a structural model amounts to the calculation of the survival
probability Q  0 ,   . In the Merton (1974) and the Black and Cox (1976) models, Q  0 ,   has closed form
solutions. The survival probability in the double exponential jump diffusion model and the two-factor models do
not have a known closed form solution but can be easily calculated using a numerical method6,39.
Schaefer and Strebulaev investigated the impact of structural credit risk models on corporate bond hedging
DOI: 10.9790/5933-1303062839 www.iosrjournals.org 33 | Page
 Credit Risk Modeling Using Default Models: A Review

ratios. They demonstrated that, whereas structural credit risk models are poor predictors of bond prices, they are
highly accurate predictors of corporate bond returns' sensitivity to changes in the value of stock (hedge ratios).
This is significant because it implies that structural models' poor performance may be due to the influence of
non-credit factors rather than a failure to reflect the credit exposure of corporate debt. The key finding of their
study is that even the most basic structural model for corporate debt pricing developed by Merton (1974): the
risk structure of interest rates, produces hedging ratios that are not rejected in time-series tests. They discover,
however, that the Merton model, with or without stochastic interest rates, fails to explain why corporate bonds'
interest rate sensitivity is so low45.
Wang gave an overview of a regularly used structural credit risk modelling approach that the actuarial
community is less aware with. He emphasized that the Merton model is based on the idea of treating a
company's equity as a call option on its assets, allowing Black-Scholes option pricing methods to be used.
Credit spread, according to Merton's model, compensates for credit risk, which is linked to structural
determinants (assets, liabilities, etc.). However, empirical evidence suggests that the Merton model undervalues
credit spreads, especially short-term spreads for high-quality debtors52.
Existing structural models, according to Davydenko, negate liquidity issues as the primary predictors of
default for some enterprises, notably those with significant external funding costs. He emphasized that if
external financing are too expensive, a liquidity crisis may require rearrangement, even though the going-
concern surplus remains significant. According to empirical evidence, structural models need to be theoretically
extended to include the risk of enterprises defaulting due to liquidity shortages and high funding costs15.
The Merton model was used by Valáková and Klietik to analyze credit risk. The possibility of debtor
default or the difference between the value of the company assets and the default barrier expressed as a number
of standard deviations is how the model depicts credit risk. They discovered that default happens in the Merton
model when the market value of the company's assets is less than the book value of its obligations. All relevant
information about the company's risk profile is included in accounting and market pricing of securities issued by
the company, which is a major prerequisite of the technique based on market data analysis 50.
Hoang and Vuong used the Merton model to construct a framework for measuring credit risk with
jumps. They studied a Merton model for default risk, in which a Brownian motion and a compound Poisson
process drive the firm's value. The findings revealed that the firm's worth can fluctuate at random, not only in a
continuous but also cumulatively discrete manner 19.

Reduced form models

Chen and Panjer used a jump-diffusion process to represent firm value evolution, where the
instantaneous percentage of change in firm value is made up of the change from a systematic diffusion process
plus the change from a nonsystematic leap. They presume that the credit spread must be consistent with the
market spread, unlike typical structural models that have distinct credit spreads from market spreads. Because
the diffusion process evolves at a riskless rate, their research determines the implied jump distribution using the
market credit spread. According to their findings, their proposed model establishes an intensity process,
allowing a structural model and a reduced-form model to be merged. Their research developed an exponential-
lognormal jump-diffusion process, which produced a distributed recovery that is consistent with the market
experience11.
Hull and White developed a methodology for valuing credit default swaps when the payoff is
contingent on default by a single reference entity and there is no counterparty default risk. Instead of using a
hazard rate for the default probability, this model incorporates a default density concept, which is the
unconditional cumulative default probability within one period no matter what happens in other periods. By
assuming an expected recovery rate, the model generates default densities recursively based on a set of zero-
coupon corporate bond prices and a set of zero-coupon Treasury bond prices. Then the default density term-
structure is used to calculate the premium of a credit default swap contract. The two sets of zero-coupon bond
prices can be bootstrapped from corporate coupon bond prices and treasury coupon bond prices. They show the
credit default swap (CDS) spread s to be25:

T
1  R  1  A  t    q  t  v  t  d t
  
s 
0
T

 q  t   u  t   e  t   d t   u  t 
0

(1)

where
T : the life of the CDS contract.

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 34 | Page

 Credit Risk Modeling Using Default Models: A Review

q  t  : the risk-neutral default probability density at time t .

A  t  : the ac crued interest on the reference obligation at time t as a percent of face value.
 : the risk-neutral probability of no credit event over the life of the CDS contract.
w : the total payments per year made by the protection buyer.
e  t  : the present value of the accrued payment from previous payment date to current date.

u  t  : the present value of the payments at time t at rate of $1 on the payment dates.

R : the expected recovery rate on the reference obligation in a risk-neutral world.

The risk-neutral default probability density is obtained from the bond data using the relationship:

j 1
Gj  Bj   i 1
q i  ij
q j
 (2)
 jj

where  i j is the present value of the loss on a defaultable bond j relative to an equivalent default-free bond at
time t i .  i j can be described as:

 ij  v  t i   F j  t i   R j  t i  C j  t i   (3)

C j
is the claim made on the jth bond in the event of default at time t i , while R j
is the recovery rate on that
claim. F j is the risk-free value of an equivalent default-free bond at time t i , while v  t i  is the present value
of a sure payment of $1 at time t i . Under this framework, one can infer a risk-neutral default risk density from a
cross-section of bonds with various maturities. As long as the bonds measure the inherent credit risk and have
the same recovery as used in the CDS, one should be able to recover a fair price for the CDS based on the prices
of the obligors traded bonds.
A specification analysis of reduced-form credit risk models was undertaken by Berndt4. They evaluated
numerous one-factor reduced-form credit risk models for actual default intensities using non-parametric
specification tests devised by Hong and Li20. They used Moody's KMV estimations for actual default
probabilities for 106 U.S. enterprises in seven industrial groups from 1994 to 2005 (Bohn et al.9 and Crosbie and
Bohn14). Popular univariate affine model specifications were sharply rejected by the results. They hypothesized
that the logarithm of the real default intensity follows an Ornstein-Uhlenbeck process, known as the Black-
Karasinski (BK) model, for goodness-of-fit and model simplicity. They discovered significant mean-reversion
in actual log-default intensities for the BK model specification, with an average half-time of around 18 months.
Findings also indicated the level of pairwise correlation in log-default intensities differed across industries7.
The reduced form modelling approach for credit risk was provided by Jeanblanc and Lecam in a
unified setting. A credit event was depicted as an inaccessible random time. They proposed two techniques to
modelling default: the intensity-based approach, which is efficient when working with Cox process construction
since the default time is generated from the intensity, and the hazard process approach, which gives the default
time. The results demonstrated that knowing the intensity is not required to price contingent claims using the
first method. The last method is particularly suited to studying models with incomplete observations, which is a
method of studying a model where the default has an economic value (as in structural models) and when the
default is unpredictable, giving nice spreads32.
In a reduced form model of default spreads with Markov-switching macroeconomic factors, Dionne et
al. investigated the ability of observed macroeconomic indicators and the probability of regime shifts to explain
the proportion of yield spreads induced by the risk of default. They applied Bansal and Zhou's Markov-
switching risk-free term structure model to corporate bonds, developing recursive formulas for default
probability, risk-free and risky zero-coupon bond yields, and credit default swap premia2. They used
consumption, inflation, risk-free returns, and default data for Aa, A, and Baa bonds from 1987 to 2008 to
calibrate their model. Macroeconomic factors were associated to two out of three dramatic spikes in default
spreads during this sample period, according to the researchers. Both inflation and consumption growth were
adversely associated to default spreads during these recessions, demonstrating that spread variations can be
linked to macroeconomic undiversifiable risk. They also mentioned that the bond market's illiquidity is most

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 35 | Page

 Credit Risk Modeling Using Default Models: A Review

likely the key reason for the discrepancy in default and credit spreads. They proposed two additions to their
research: (i) taking into account alternative macro factors that are more closely linked to economic recessions
than consumption, and (ii) explicitly including liquidity risk in the model 16.
Liang and Wang proposed a reduced form credit risk model in which common shocks with regime-
switching describe the default dependence structures among default intensity processes. They also came up with
some closed-form formulations for the joint distribution of default timings and the basket default swap pricing
formulas. They used correlated relations to depict the default dependence structure among the default intensity
processes in their model36.
Su and Wang used a reduced form model to analyze the valuation of European options with credit risk.
They assumed that the interest rate follows the Vasicek model, and that the default intensity is determined by a
jump diffusion process. They were able to establish the closed form formula for the option price. They then
calculated the value of the vulnerable European call option by altering the probability measures. Finally, they
calculated the effects of the recovery rate, correlation coefficients, and Poisson intensity on the option price
using numerical analysis46.

Comparison of Structural models and Reduced form models

Using a jump-diffusion technique, Chen and Panjer combined discrete structural models and reduced-
form models in credit risk. They calculated the default probability and mean recovery rate by combining the
market spread and the firm's capital structure. They used the credit spread to find the implied jump distribution.
They demonstrated that credit spreads in the structural and reduced-form models are equivalent. The default
probability and mean recovery rate were calculated using the market spread in the reduced-form model, whereas
the default probability and mean recovery rate were calculated using the capital structure in the structural model.
The degree of freedom is raised by adding a jump process to the diffusion process, but it is lowered by the
market spread, according to the results. The credit spread difference between the structural model and the
reduced-form model was eliminated by using the market spread to calculate the implied jump distribution, and
the structural model and the reduced-form model may now be unified when the default can only occur at
maturity. However, there is no mechanism for determining the jump distribution. The default probability (or
intensity of default) and mean recovery rate are calculated from the market spread using model-specific
assumptions in a reduced-form model, although the capital structure that triggers the default is rarely used12.
From an information-based perspective, Jarrow and Protter evaluated structural vs reduced form credit
risk models. Structural models presume that the modeler has access to the same data set as the business
manager, which includes a thorough understanding of the firm's assets and liabilities. In the vast majority of
cases, this knowledge results in a predictable default time. Reduced form models, on the other hand, presume
that the modeler has the same set of data as the market's inadequate understanding of the firm's status. In the
majority of circumstances, this incomplete knowledge results in an unavailable default time, as a result, they
claim that the main contrast between structural and reduced form models is whether the information set is
observed by the market or not, not whether the default time is predictable or inaccessible. They recommended
reduced form models as the preferred methodology for pricing and hedging because they were built specifically
to be based on the information available to the market30.
Two structural models of credit risk (basic Merton and Vasicek-Kealhofer (VK)) and one reduced-form
model (Hull and White 2003) were empirically compared by Arora et al. Default discrimination and relative
value analysis are two useful reasons for credit models, according to them. They looked at how well the Merton
and VK models could distinguish defaulters from non-defaulters using default probability derived from equities
market data. They examined the HW model's ability to distinguish defaulters from non-defaulters using default
probability derived from bond market data. They discovered that the VK and HW models outperform the simple
Merton model on both the complete sample and relevant sub-samples, with comparable accuracy ratios. They
also assessed each model's ability to anticipate spreads in the credit default swap (CDS) market as a measure of
its relative value analysis capability. Except in circumstances where an issuer has a large number of bonds in the
market, they discovered that the VK model performs better throughout the entire sample and relative sub-
samples. In this scenario, the HW model is the most effective. On the structural side, a simple Merton model
proved insufficient; proper framework adjustments were required to create the difference. The quality and
quantity of data made a difference on the reduced-form side2.
Jarrow examined the structural and reduced form credit risk models utilized in financial economics.
They claimed that reduced form models, rather than structural models, should be used to price and hedge credit-
risky instruments because structural models are static and do not represent the dynamic structure of credit risk.
The sophistication of default contagion models, as well as the estimating processes used, must be increased to
avoid a repeat performance. This enhancement will necessitate extensive study to identify dynamic models that
represent default contagion while having parameters that can be estimated and values that can be computed.
These include stochastic recovery rate models, which are needed to better represent the dynamic nature of

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 36 | Page

 Credit Risk Modeling Using Default Models: A Review

default losses and the reliance of credit and liquidity risk, which is crucial for valuation, hedging, and capital
determination29.
Jarrow explored the theory and evidence of credit market equilibrium. The structural and reduced form
models were reviewed as alternative paradigms for estimating credit risk. Their discussion was based on their
understanding of credit market equilibrium. They demonstrated that in the borrowing and lending relationship,
credit markets include asymmetric information, which influences equilibrium prices. Asymmetric equilibrium
models are consistent with reduced form models, while structural models are not. Structural models should not
be used for pricing, hedging, or risk management, as a result28.
Misanková et al. calculated credit risk using reduced form models based on credit spreads and
calculated the chance of default. Reduced-form models employ credit spreads as an input for calculating the
probability of default, as opposed to structural models, which try to explain credit spreads through structural
characteristics. They calibrated their model using observable market data, which is one of the most appealing
aspects of this technique, and this flexibility is a significant benefit of these models. The results reveal that in
these models, the structural characteristics of the company are insufficient to explain credit default swap (CDS)
spreads, and empirical evidence shows that systematic risk has a significant impact on credit default swap prices
(CDS). CDS pricing, on the other hand, which are derived using structural models, frequently overreact to
increasing market volatility42.

Table 1. Evolution of credit risk models

Main models Related empirical studies Treatment of default event
Structural Merton(1974), Merton(1976), -Aim to provide an explicit relationship between default risk and capital structure
models Zhou (1997), Bohn et al. -Based on the value of the company
(2005), -Examine the structure of the capital of the company
Jarrow et al. (2003), Hull et al. -Endogenously specify default event and recovery rates
(2004), -Calculate the probability of default of a firm based on the value of assets and
Elzade (2005), Black and liabilities
Cox(1976), -Assume complete knowledge of a company’s assets and liabilities resulting in a
Kulkami et al. (2005), predictable default event and recovery rates
Tarashev (2005), -Assume default risk occur at the maturity date if at that stage, the value of a
Huang (2005), Zhang, Zhou company’s assets fall below a debt threshold
and -Don’t observe the market value of a firm’s assets
Zhu (2006), Huang and Zhou -Consider company’s debt as an option on company’s assets
(2008), -Model a firm’s asset value as a lognormal process
Huang and Huang (2003),
Schaefer and
Strebulaev (2008),Wang
(2009),
Davydenko (2012), Valášková
and
Klieštik (2014), Hoang and
Vuong (2015),
Longstaff and Schwartz
(1995),
Mišanková (2015), Saunders
and Allen 2002,
(Chatterjee, 2015)
Reduced-form Acharya and Carpenter -They don’t consider the relation between default and firm’s value in an explicit
models (2002),Jarrow and Turnbull manner
(1995), -Treat default as unexpected event
Chen and Panjer (2002), -They assume an exogenous cause of default driven by a stochastic process (Poison
Berndt (2007), Hong and Li jump process)
(2005), -They model default as a random event without any focus on the firm’s balance sheet
Jeanblanc and Lecam (2008), -Describe a random event of default as a Poison event (default intensity modeling)
Dionne et al. (2011), -Use market observed market credit spread to obtain the probability (intensity) of
Crosbie and Bohn default and recovery rates
(2003),Black and -These models lack economic insights about occurrence of defaults
Karasinski (1991), Bansal and
Zhou (2002),
Liang and Wang (2012), Su
and Wang (2012),
Comparison of Arora et al. (2005), Hull and Because structural models are static in character and do not represent the dynamic
Structural White (2002), structure of credit risk, the literature argues that reduced form models, not structural
models and Jarrow (2009), Jarrow (2011), models, are ideal for the pricing and hedging of credit-risky instruments. Structural
Reduced form Mišanková et al. (2015) models, with minor adjustments, can be effective in the pricing and hedging of credit
models securities.

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 37 | Page

 Credit Risk Modeling Using Default Models: A Review

III. Conclusion and Suggestion for Future Research

We have evaluated studies on default models' approach to credit risk in this paper. We looked at
research on structural models, reduced form models, and comparisons between the two methods. Default
models, in general, focus on the modelling of default events using market data. These models have evolved
through time into two distinct types: structural and reduced form models. Credit risk is linked to underlying
structural factors in structural models, which is a very appealing aspect. They allow for the use of option pricing
methodologies and provide both an understandable economic interpretation and an endogenous explanation of
credit defaults. As a result, structural models can help with not just asset value but also financial structure
selection. The fundamental disadvantage of structural models is their complexity in implementation and the
predictability of defaults and recovery rates, which is contradictory to market reality.
Reduced form models calculate the likelihood (or intensity) of default as well as the mean recovery rate
using the observed market credit spread. They specify recovery rates exogenously. These models suffer from a
lack of economic insights into default occurrence; yet, they provide more functional form selection freedom.
The analytical tractability, as well as the ease of implementation and calibration, are aided by this flexibility
(compared to structural models). Reduced form models, on the other hand, may have high in-sample fitting
features but limited out-of-sample prediction power due to their reliance on past data.
Because structural models are static in nature and do not represent the dynamic structure of credit risk,
empirical data suggests that reduced form models, not structural models, are ideal for the pricing and hedging of
credit-risky instruments. It is necessary to make improvements to the estimating processes employed in
structural models. This enhancement will necessitate extensive study to identify dynamic models that represent
default contagion while having parameters that can be estimated and values that can be computed. These include
stochastic recovery rate models to better represent the dynamic nature of default losses, as well as credit risk
reliance and the incorporation of liquidity risk for valuation, hedging, and capital determination.

References
[1]. Acharya VV, Carpenter JN. Corporate bond valuation and hedging with stochastic interest rates and endogenous bankruptcy. The
Review of Financial Studies. 2002 Oct 1;15(5):1355-83.
[2]. Arora N, Bohn JR, Zhu F. Reduced‐ Form versus Structural Models of Credit Risk: A Case Study of Three Models. The Credit
Market Handbook: Advanced Modeling Issues. 2012 Jan 2:132-64.
[3]. Bansal R, Zhou H. Term structure of interest rates with regime shifts. The Journal of Finance. 2002 Oct;57(5):1997-2043..
[4]. Berndt A. Specification analysis of reduced-form credit risk models. Tepper School of Business, article en développement. 2007
Apr 4.
[5]. Bielecki TR, Rutkowski M. Defaultable term structure: Conditionally Markov approach. IEEE Transactions on Automatic Control,
Special Issue on Stochastic Control Methods in Financial Engineering. 2004.
[6]. Black F, Cox JC. Valuing corporate securities: Some effects of bond indenture provisions. The Journal of Finance. 1976
May;31(2):351-67.
[7]. Black F & Karasinski P. Bond and option pricing when short rates are lognormal. Financial Analysts Journal. 1991; 47(4):52-59.
[8]. Black F, Scholes M. The pricing of options and corporate liabilities. InWorld Scientific Reference on Contingent Claims Analysis
in Corporate Finance: Volume 1: Foundations of CCA and Equity Valuation 2019 (pp. 3-21).
[9]. Bohn J., Arora N & Korbalev I. Power and Level Validation of the EDF CRedit Measure in the U.S. Market. Working Paper,
Moody’s KMV. 2005.
[10]. Chatterjee S. Modelling credit risk. Handbooks. 2015.
[11]. Chen CJ, Panjer H. Unifying continuous structural models and reduced-form models in credit risk: A bridge from ruin theory to
credit risk.
[12]. Chen CJ, Panjer H. Unifying discrete structural models and reduced-form models in credit risk using a jump-diffusion process.
Insurance: Mathematics and Economics. 2003 Oct 20;33(2):357-80.
[13]. Collin‐ Dufresne P, Goldstein RS. Do credit spreads reflect stationary leverage ratios?. The journal of finance. 2001
Oct;56(5):1929-57..
[14]. Crosbie P & Bohn J. Modeling default risk. Moody’s KMV White Paper. 2003.
[15]. Davydenko S. When do firms default? A study of the default boundary. InA Study of the Default Boundary (November 2012). EFA
Moscow Meetings Paper, AFA San Francisco Meetings Paper, WFA Keystone Meetings Paper 2012 Nov 1.
[16]. Dionne G, Gauthier G, Hammami K, Maurice M, Simonato JG. A reduced form model of default spreads with Markov-switching
macroeconomic factors. Journal of Banking & Finance. 2011 Aug 1;35(8):1984-2000.
[17]. Elizalde A. Credit risk models II: Structural models.2005.
[18]. Geske R. The valuation of corporate liabilities as compound options. Journal of Financial and quantitative Analysis. 1977
Nov;12(4):541-52.
[19]. Hoang TP, Vuong QH. A Merton model of credit risk with jumps.2015.
[20]. Hong Y, Li H. Nonparametric specification testing for continuous-time models with applications to term structure of interest rates.
The Review of Financial Studies. 2005 Mar 1;18(1):37-84.
[21]. Huang J. Affine structural models of corporate bond pricing. Smeal College of Business, Penn State University. 21pp. 2005.
[22]. Huang JZ, Huang M. How much of the corporate-treasury yield spread is due to credit risk?. The Review of Asset Pricing Studies.
2012 Dec 1;2(2):153-202.
[23]. Huang JZ, Zhou H. Specification analysis of structural credit risk models. InAFA 2009 San Francisco Meetings Paper 2008 Oct.
[24]. Hull J, Nelken I, White A. Merton’s model, credit risk, and volatility skews. Journal of Credit Risk Volume. 2004;1(1):05.
[25]. Hull JC, White AD. Valuing credit default swaps I: No counterparty default risk. The Journal of Derivatives. 2000 Aug 31;8(1):29-
40.
[26]. Hull JC, White AD. The valuation of credit default swap options. The journal of derivatives. 2003 Feb 28;10(3):40-50.
[27]. Jarrow RA, Turnbull SM. Pricing derivatives on financial securities subject to credit risk. The journal of finance. 1995

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 38 | Page

 Credit Risk Modeling Using Default Models: A Review
Mar;50(1):53-85.
[28]. Jarrow RA. Credit market equilibrium theory and evidence: Revisiting the structural versus reduced form credit risk model debate.
Finance Research Letters. 2011 Mar 1;8(1):2-7.
[29]. Jarrow RA. Credit risk models. Annu. Rev. Financ. Econ.. 2009 Sep;1(1):37-68.
[30]. Jarrow R, Protter P. Structural versus reduced form models: a new information based perspective. Journal of Investment
management. 2004 May;2(2):1-0.
[3]1. Jarrow R, van Deventer DR, Wang X. A robust test of Merton's structural model for credit risk. Journal of Risk. 2003;6:39-58.
[32]. Jeanblanc M, Lecam Y. Reduced form modelling for credit risk. Available at SSRN 1021545. 2008 Aug 24.
[33]. Kealhofer S. Quantifying credit risk I: default prediction. Financial Analysts Journal. 2003 Jan 1;59(1):30-44.
[34]. Kollar B. Credit Value at Risk and options of its Measurement. In2nd international conference on economics and social science
(ICESS 2014), Information Engineering Research Institute, Advances in Education Research 2014 (Vol. 61, pp. 143-147).
[35]. Kulkarni A, Mishra AK, Thakker J. How good is Merton Model at assessing credit risk? Evidence from India. National Institute of
Bank Management. 2005.
[36]. Liang X, Wang G. On a reduced form credit risk model with common shock and regime switching. Insurance: Mathematics and
Economics. 2012 Nov 1;51(3):567-75.
[37]. Longstaff FA, Schwartz ES. A simple approach to valuing risky fixed and floating rate debt. The Journal of Finance. 1995
Jul;50(3):789-819.
[38]. Majerčák P & Majerčáková M. The enterprise valuation and categories of the value, Proceedings of the Financial management of
firms and financial institutions, 9th international scientific conference: 9th-10th September 2013 Ostrava, Czech Republic,pp. 469-
475.
[39]. Merton RC. On the pricing of corporate debt: The risk structure of interest rates. The Journal of finance. 1974 May 1;29(2):449-70.
[40]. Merton RC. Option pricing when underlying stock returns are discontinuous. Journal of financial economics. 1976 Jan 1;3(1-
2):125-44.
[41]. Mišanková M, Kočišová K. COMPARISON OF SELECTED DISCRIMINANT MEASUERES IN CREDIT RISK. SOCIO -
ECONOMIC ASPECTS OF ECONOMICS AND MANAGEMENT. 2015:22.
[42]. Mišanková M., Klieštik T & Adamko P. Reduced–form Models Used for the Calculation of Credit Risk. 2015.
[43]. Mišanková M, Kočišová K, Adamko P. CreditMetrics and its use for the calculation of credit risk. In2nd International Conference
on Economics and Social Science (ICESS 2014), Information Engineering Research Institute, Advances in Education Research
2014 (Vol. 61, pp. 124-129).
[44]. Saunders A, Allen L. Credit risk measurement: new approaches to value at risk and other paradigms. John Wiley & Sons; 2002 Oct
6.
[45]. Schaefer SM, Strebulaev IA. Structural models of credit risk are useful: Evidence from hedge ratios on corporate bonds. Journal of
Financial Economics. 2008 Oct 1;90(1):1-9.
[46]. Su X, Wang W. Pricing options with credit risk in a reduced form model. Journal of the Korean Statistical Society. 2012 Dec
1;41(4):437-44.
[47]. Tarashev NA. An empirical evaluation of structural credit risk models.2005.
[48]. Kliestik T, Misankova M, Kocisova K. Calculation of distance to default. Procedia economics and finance. 2015 Jan 1;23:238-43.
[49]. Valaskova K. Quantification of the company default by merton model. Advances in Education Research. 2014;51:133-8.
[50]. Valášková K & Klieštik T. Assessing Credit Risk by Merton Model. In Proceedings of ICMEBIS 2014 International Conference on
Management, Education, Business, and Information Science, Shanghai, China, EDUGait Press, Canada (pp. 27-30). 2014.
[51]. Vasicek OA. Credit valuation. 1984.
[52]. Wang Y. Structural credit risk modeling: Merton and beyond. Risk Management. 2009 Jun;16(2):30-3.
[53]. Zhang BY, Zhou H, Zhu H. Explaining credit default swap spreads with the equity volatility and jump risks of individual firms. The
Review of Financial Studies. 2009 Dec 1;22(12):5099-131.
[54]. Zhou C. A jump-diffusion approach to modeling credit risk and valuing defaultable securities. Available at SSRN 39800. 1997 Mar.

George Jumbe. “Credit Risk Modeling Using Default Models: A Review.” IOSR Journal of
Economics and Finance (IOSR-JEF), 13(03), 2022, pp. 28-39.

DOI: 10.9790/5933-1303062839 www.iosrjournals.org 39 | Page