Presentacion del Doctor Dr. Viswanathan Arunachalam, profesor asociado en el Departamento de Estadística de la Universidad Nacional
de Colombia.
Resumen:
En el mercado de valores financieros, es importante tener conocimiento sobre las medidas de riesgo que surgen de la distribución del rendimiento de las acciones. Dado que las propiedades de distribución con multimodalidad, simetría, asimetría, colas pesadas, entre otras, nos permiten ser más flexibles en el objetivo de describir los fenómenos presentados en los modelos de valoración, existen varias propuestas sobre la distribución que podría tener la devolución de acciones. Históricamente, se han tenido en cuenta las distribuciones normales conocidas, ya que la distribución de los retornos de las acciones tiene aplicaciones atractivas en finanzas y similitudes con los datos reales. En esta charla vamos a presentar algunos modelos estocásticos para la valoración de riesgo en el sector de seguros, en particular riesgo de Crédito, operacional y derivados financieros.
Q3 2024 Earnings Conference Call and Webcast Slides
Modelos estocásticos para la valoración de riesgo financiero
1. MODELOS ESTOCÁSTICOS PARA LA VALORACIÓN DE
RIESGO FINANCIERO
Viswanathan (ARUN) Arunachalam
Departamento de Estadística
Universidad Nacional de Colombia
Bogotá, Colombia
varunachalam@unal.edu.co
XXIII Seminario financiero: Las finanzas cuantitativas -UNAB
Bucaramanca
20 de Septiembre 2019
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
2. AGENDA
1 FINANCIAL RISK
2 MARKET RISK -OPTION PRICING
3 THE SKEW NORMAL MIXTURES
4 CREDIT RISK
5 OPERATIONAL RISK
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
3. Financial Risk
FINANCIAL RISK
Financial risk is one of the high-priority risk types for every business.
Risk can be referred to like the chances of having an unexpected
or negative outcome. Any action or activity that leads to loss of
any type can be termed as risk. There are different types of risks
that a firm might face and needs to overcome. Widely, risks can
be classified into three types: Business Risk, Non-Business Risk,
and Financial Risk.
Financial Risk: Financial Risk as the term suggests is the risk
that involves financial loss to firms. Financial risk generally arises
due to instability and losses in the financial market caused by
movements in stock prices, currencies, interest rates and more.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
4. Financial Risk
FINANCIAL RISK
Financial risk is one of the high-priority risk types for every business.
Risk can be referred to like the chances of having an unexpected
or negative outcome. Any action or activity that leads to loss of
any type can be termed as risk. There are different types of risks
that a firm might face and needs to overcome. Widely, risks can
be classified into three types: Business Risk, Non-Business Risk,
and Financial Risk.
Financial Risk: Financial Risk as the term suggests is the risk
that involves financial loss to firms. Financial risk generally arises
due to instability and losses in the financial market caused by
movements in stock prices, currencies, interest rates and more.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
5. Financial Risk
TYPES OF FINANCIAL RISKS
Financial risk is caused due to market movements and market
movements can include a host of factors. Based on this, financial risk
can be classified into various types such as
Market Risk : This type of risk arises due to the movement in
prices of financial instrument.
Market risk can be classified as Directional Risk and
Non-Directional Risk.
Directional risk is caused due to movement in stock price, interest
rates and more.
Non-Directional risk, on the other hand, can be volatility risks.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
6. Financial Risk
TYPES OF FINANCIAL RISKS
Financial risk is caused due to market movements and market
movements can include a host of factors. Based on this, financial risk
can be classified into various types such as
Market Risk : This type of risk arises due to the movement in
prices of financial instrument.
Market risk can be classified as Directional Risk and
Non-Directional Risk.
Directional risk is caused due to movement in stock price, interest
rates and more.
Non-Directional risk, on the other hand, can be volatility risks.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
7. Financial Risk
TYPES OF FINANCIAL RISKS
Financial risk is caused due to market movements and market
movements can include a host of factors. Based on this, financial risk
can be classified into various types such as
Market Risk : This type of risk arises due to the movement in
prices of financial instrument.
Market risk can be classified as Directional Risk and
Non-Directional Risk.
Directional risk is caused due to movement in stock price, interest
rates and more.
Non-Directional risk, on the other hand, can be volatility risks.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
8. Financial Risk
TYPES OF FINANCIAL RISKS
Financial risk is caused due to market movements and market
movements can include a host of factors. Based on this, financial risk
can be classified into various types such as
Market Risk : This type of risk arises due to the movement in
prices of financial instrument.
Market risk can be classified as Directional Risk and
Non-Directional Risk.
Directional risk is caused due to movement in stock price, interest
rates and more.
Non-Directional risk, on the other hand, can be volatility risks.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
9. Financial Risk
TYPES OF FINANCIAL RISKS
Financial risk is caused due to market movements and market
movements can include a host of factors. Based on this, financial risk
can be classified into various types such as
Credit Risk: This type of risk arises when one fails to fulfill their
obligations towards their counterparties. Credit risk can be
classified into Sovereign Risk and Settlement Risk. Sovereign risk
usually arises due to difficult foreign exchange policies.
Settlement risk, on the other hand, arises when one party makes
the payment while the other party fails to fulfill the obligations.
Operational Risk : This type of risk arises out of operational
failures such as mismanagement or technical failures. Operational
risk can be classified into Fraud Risk and Model Risk. Fraud risk
arises due to the lack of controls and Model risk arises due to
incorrect model application.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
10. Financial Risk
TYPES OF FINANCIAL RISKS
Financial risk is caused due to market movements and market
movements can include a host of factors. Based on this, financial risk
can be classified into various types such as
Credit Risk: This type of risk arises when one fails to fulfill their
obligations towards their counterparties. Credit risk can be
classified into Sovereign Risk and Settlement Risk. Sovereign risk
usually arises due to difficult foreign exchange policies.
Settlement risk, on the other hand, arises when one party makes
the payment while the other party fails to fulfill the obligations.
Operational Risk : This type of risk arises out of operational
failures such as mismanagement or technical failures. Operational
risk can be classified into Fraud Risk and Model Risk. Fraud risk
arises due to the lack of controls and Model risk arises due to
incorrect model application.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
11. Financial Risk
STOCHASTIC MODELING
Stochastic modeling is a form of a financial model that is used to help
make investment decisions. This type of modeling forecasts the
probability of various outcomes under different conditions, using
random variables.
Deterministic modeling produces constant results
Stochastic modeling produces distribution(various) of results.
A stochastic model incorporates random variables to produce
many different outcomes under diverse conditions.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
12. Financial Risk
STOCHASTIC MODELING
Stochastic modeling is a form of a financial model that is used to help
make investment decisions. This type of modeling forecasts the
probability of various outcomes under different conditions, using
random variables.
Deterministic modeling produces constant results
Stochastic modeling produces distribution(various) of results.
A stochastic model incorporates random variables to produce
many different outcomes under diverse conditions.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
13. Financial Risk
STOCHASTIC MODELING
Stochastic modeling is a form of a financial model that is used to help
make investment decisions. This type of modeling forecasts the
probability of various outcomes under different conditions, using
random variables.
Deterministic modeling produces constant results
Stochastic modeling produces distribution(various) of results.
A stochastic model incorporates random variables to produce
many different outcomes under diverse conditions.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
14. Market Risk -Option Pricing
OPTION PRICING
There is empirical evidence that financial stock returns are not
normally distributed but are characterized by skewness,
leptokurticity, heavy-tailedness and other non-Gaussian
properties.
The skewness and kurtosis of the empirical distribution function
(EDF) of stock returns contribute significantly to the phenomenon
of volatility smile.
In recent years, there have been considerable efforts to report that
the unconditional probability distributions of returns on financial
stocks are not normally distributed.
Specifically, these distributions tend to have heavier tails
(leptokurtic) and asymmetry.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
15. Market Risk -Option Pricing
OPTION PRICING
There is empirical evidence that financial stock returns are not
normally distributed but are characterized by skewness,
leptokurticity, heavy-tailedness and other non-Gaussian
properties.
The skewness and kurtosis of the empirical distribution function
(EDF) of stock returns contribute significantly to the phenomenon
of volatility smile.
In recent years, there have been considerable efforts to report that
the unconditional probability distributions of returns on financial
stocks are not normally distributed.
Specifically, these distributions tend to have heavier tails
(leptokurtic) and asymmetry.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
16. Market Risk -Option Pricing
OPTION PRICING
There is empirical evidence that financial stock returns are not
normally distributed but are characterized by skewness,
leptokurticity, heavy-tailedness and other non-Gaussian
properties.
The skewness and kurtosis of the empirical distribution function
(EDF) of stock returns contribute significantly to the phenomenon
of volatility smile.
In recent years, there have been considerable efforts to report that
the unconditional probability distributions of returns on financial
stocks are not normally distributed.
Specifically, these distributions tend to have heavier tails
(leptokurtic) and asymmetry.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
17. Market Risk -Option Pricing
OPTION PRICING
There is empirical evidence that financial stock returns are not
normally distributed but are characterized by skewness,
leptokurticity, heavy-tailedness and other non-Gaussian
properties.
The skewness and kurtosis of the empirical distribution function
(EDF) of stock returns contribute significantly to the phenomenon
of volatility smile.
In recent years, there have been considerable efforts to report that
the unconditional probability distributions of returns on financial
stocks are not normally distributed.
Specifically, these distributions tend to have heavier tails
(leptokurtic) and asymmetry.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
18. Market Risk -Option Pricing
LOG-SKEW-NORMAL DISTRIBUTION- OPTION PRICING
By assuming that the stock distribution follows a
Log-Skew-Normal mixture (LSNMIX) distribution, we calculate an
explicit formula for option valuation for both European call and put
options and Greek measures.
We also show that some of the well-known models are obtained
as special cases from the proposed model
An example from past S&P500 daily returns to price is presented
to illustrate the model.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
19. Market Risk -Option Pricing
LOG-SKEW-NORMAL DISTRIBUTION- OPTION PRICING
By assuming that the stock distribution follows a
Log-Skew-Normal mixture (LSNMIX) distribution, we calculate an
explicit formula for option valuation for both European call and put
options and Greek measures.
We also show that some of the well-known models are obtained
as special cases from the proposed model
An example from past S&P500 daily returns to price is presented
to illustrate the model.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
20. Market Risk -Option Pricing
LOG-SKEW-NORMAL DISTRIBUTION- OPTION PRICING
By assuming that the stock distribution follows a
Log-Skew-Normal mixture (LSNMIX) distribution, we calculate an
explicit formula for option valuation for both European call and put
options and Greek measures.
We also show that some of the well-known models are obtained
as special cases from the proposed model
An example from past S&P500 daily returns to price is presented
to illustrate the model.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
21. Market Risk -Option Pricing
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
22. Market Risk -Option Pricing
LOG-SKEW-NORMAL DISTRIBUTION
A random variable Y has a LSN distribution with asymmetry
parameter λ ∈ R, denoted as Y ∼ LSN (Λ1) , if its pdf is of the form
fY y; Λ1 =
2
σy
ϕ
ln y − µ
σ
Φ λ
ln y − µ
σ
=
1
y
φSN ln y; Λ1 , y ∈ R+
, (1)
where Λ1 = µ, σ, λ , σ > 0, φSN(·) denotes the pdf of the
skew-Normal (SN) distribution, and ϕ (·) and Φ(·) denote the pdf
and cdf of a standard univariate normal variable, respectively.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
23. Market Risk -Option Pricing
LOG-SKEW-NORMAL DISTRIBUTION
If µ = 0 and σ = 1 then Y is said to have a (standard) LSN distribution,
i.e., Y ∼ LSN (λ) . The parameter λ controls the skewness, which is
positive when λ > 0 and negative when λ < 0. The cdf of (1) is given by
FY (y; Λ1) = Φ
ln y − µ
σ
− 2T
ln y − µ
σ
; 0, λ , (2)
where the function T (z; α, λ) with α ≥ 0 is given as
T (z; α, λ) =sign (λ)
arctan (|λ|)
2π
−
z
α
|λ|x
0
ϕ(x, α, 1)ϕ (y)dydx ,
and sign(·) is the signum function.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
24. Market Risk -Option Pricing
−4 −3 −2 −1 0 1 2 3 4
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Probability Density Functions SN(µ,σ,λ), µ= 0, σ= 1
x
SN(µ,σ,λ)
λ= 0
λ= 2
λ= 5
FIGURE: Comparison of the SN pdf with λ > 0.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
25. Market Risk -Option Pricing
−4 −3 −2 −1 0 1 2 3 4 5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Probability Density Functions SN(µ,σ,λ), µ= 0, σ= 1
x
SN(µ,σ,λ)
λ= 0
λ= −2
λ= −5
FIGURE: Comparison of the SN pdf with λ < 0.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
26. The Skew Normal Mixtures
MIXTURE
we consider the finite mixture model: We assume that if Y is SNMIX
distributed then the transformation X = exp {Y} is distributed as a
LSNMIX. Let us assume that fY(y) is the weighted sum of
m-component SNMIX densities, that is,
fY (y; ˜) =
m
j=1
ωjφSN (y; µj, σj, λj). (3)
We use the notation Y ∼ SNMIX (˜) , where ˜ = (ξ1, . . . , ξm) , and
ξj = (ωj, µj, σj, λj) is the parameter vector that defines the j-th
component and probability weights, ωj, satisfy the conditions
m
j=1
ωj =1, 0 < ωj < 1, for each j. (4)
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
27. The Skew Normal Mixtures
The choice of a finite mixture is attractive from the application view
point because of its flexibility and allows us to consider different kinds
of shaped distributions. For instance, the two component SNMIX
model has the advantage of numerical tractability, because it has only
seven parameters. Assuming ξj = (ωj, µj, σj, λj) , with
µ1 = −1, µ2 = 1, σ1 = σ2 = 1, λ1 = .5 and λ2 = −2. Figure 3 shows the
pdf shape of the SNMIX for three values of ω.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
28. The Skew Normal Mixtures
−5 −4 −3 −2 −1 0 1 2 3 4 5
0
0.1
0.2
0.3
0.4
0.5
Probability Density Functions SNMix(µ1
,σ1
,λ1
,µ2
,σ2
,λ2
,ω)
µ1
= −1, σ1
= 1, λ1
= 0.5, µ2
= 1, σ2
= 1, λ2
= −2
x
SNMix(µ1
,σ1
,λ1
,µ2
,σ2
,λ2
,ω)
ω= 0.06
ω= 0.46
ω= 0.91
Normal
FIGURE: Comparison of the pdf of the SNMIX with varying ω.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
29. The Skew Normal Mixtures
OPTION PRICING FORMULA
Let St be the price of the underlying stock at time t and C(t, τ; K) the
price of the call option with strike price K and maturity date of T = t + τ.
It is assumed that r is the annual risk-free rate. In the absence of
arbitration, the price of European call options can be written as follows:
C(t, τ; K) = E e−rτ
max{ST − K, 0} = e−rτ
E [max{ST − K, 0}]
Ct(K) = e−rτ
∞
K
(ST − K) f (ST) dST.
Here E[.] is the expected value conditional (risk neutral) on any
information that is available at time t, f (ST) is the risk-neutral pdf
(risk-neutral distribution, RND) for the underlying at maturity.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
30. The Skew Normal Mixtures
In an arbitrage-free economy, the stock price discounted by the risk
free rate becomes martingale, that is:
E[e−rτ
ST] =St, τ > 0 (5)
where the standard deviation of ln (St) is σ
√
τ. The net premium of the
derivative, which is the expected value of the option at maturity in a
risk-neutral world, E [max{ST − K, 0}] is calculated using statistical or
numerical methods.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
31. The Skew Normal Mixtures
OPTION PRICE USING THE LSNMIX
We define the distribution of the logarithm of the stock price ST using
its location and scale parameters A and B, respectively, and also ˜, the
parameter of the SNMIX. These parameters satisfy the following
relationship:
ln [ST] =A + BY, with Y ∼SNMIX (˜) . (6)
Then, the pdf of ST is a LSNMIX.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
32. The Skew Normal Mixtures
OPTION PRICE USING THE LSNMIX
PROPOSITION
The price of a European call option is given by
erτ
Ct K; Λ =
m
j=1
2ωjE [ST]
Υj (˜, B)
∞
−δ1j
ϕ (z) Φ [λjzj]dz
−
m
j=1
ωjK [1 − FY (−δ2j; λj)] , (7)
where zj = z + Bσj and FY(·) is given in (2),
δ1j = δ2j + Bσj, δ2j = −
κ − µj
σj
, κ =
ln K − A
B
, (8)
δ2j =
1
Bσj
ln
E [ST]
K
− ln [Υj (˜, B)] −
1
2
Bσj,
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
33. The Skew Normal Mixtures
OPTION PRICE USING THE LSNMIX
PROPOSITION
The LSNMIX European put option price is given by
erτ
Pt K; Λ =
m
j=1
ωjKFY (−δ2j; λj) −
m
j=1
2E [ST] ωj
Υj (˜, B)
−δ1j
−∞
ϕ (z) Φ [λjzj]dz,
(10)
where δ1j and δ2j are given in (8) and Υj (˜, B) is given in (9).
Using the option valuation formula, we can obtain the put-call parity
relationship by subtracting expression (7) from (10) to obtain the
following equality
erτ
Ct K; Λ − Pt K; Λ =E [XT] − K. (11)
In our case, the pdf is a RND and therefore will not violate the put-call
parity relationship.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
34. The Skew Normal Mixtures
BLACK-SCHOLES
Assuming ξj = 1
m , µ, σ, 0 for all j in (3), substituting in
expressions (7) and (10) yields, respectively,
erτ
Ct K; Λ =E [ST] Φ
1
Bσ
ln
E [ST]
K
+
1
2
Bσ − KΦ
1
Bσ
ln
E [ST]
K
−
1
2
=E [ST] Φ (d1) − KΦ (d2) , (12)
where
d2 =
1
Bσ
ln
E [ST]
K
−
1
2
Bσ, (13)
and d1 = d2 + Bσ. Note that when B =
√
τ, these expressions
coincide with the option pricing formula.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
35. The Skew Normal Mixtures
BAHRA(1997)
When ξj = ωj, µ∗
j , σj, 0 for all j, where µ∗
j = B µj − 1
2σ2
j in (3),
substituting in expressions (7) and (10) yields, respectively,
erτ
Ct K; Λ =
m
j=1
ωjE [ST]
Υj (˜, B)
Φ (δ1j) − K
m
j=1
ωjΦ (δ2j) , (14)
where δ1j and δ2j are given in (8) and
Υj (˜, B) =
m
l=1
ωl exp B2
(µl − µj) . (15)
Note that when B =
√
τ, these expressions coincide with the option
pricing formula given in [?].
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
36. The Skew Normal Mixtures
CORNS & SATCHELL (2007) -SKEW BROWNIAN MOTION
When ξj = 1
m , µ, σ, λ for all j in (3), substituting in expressions (7)
and (10) yields, respectively,
erτ
Ct K; Λ =
E [ST]
Φ (ρσB)
∞
−δ1
ϕ (z) Φ [λ (z + Bσ)]dz − K [1 − FY (−δ2; λ)] ,
(16)
where
δ2 =
1
Bσ
ln
E [ST]
2K Φ (ρσB)
−
1
2
Bσ, (17)
and δ1 = δ2 + Bσ. Note that when B =
√
τ, these expressions coincide
with the option pricing formula (Corns & Satchell).
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
37. The Skew Normal Mixtures
NUMERICAL RESULTS
We now present an example from real market data to model the
distribution of the stock prices and compare the numerical values of a
European option under the assumption that the stock movement follow
a LSNMIX distribution. The market data of the S&P500 index for the
daily prices were considered from January 4, 2010 to October 13, 2014.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
38. The Skew Normal Mixtures
Statistics Values
Mean 0.0004
Stan. Dev. 0.0102
Minimum -0.0690
Maximum 0.0463
Skewness -0.4854
Kurtosis 7.6325
JB test 1120.1201
TABLE: Summary of the Descriptive Statistics
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
39. The Skew Normal Mixtures
NUMERICAL RESULTS
Estimate MME MLE
µ1 -0.0039 -0.0025
µ2 0.0036 0.0045
σ1 0.0056 0.0056
σ2 0.0137 0.0132
λ1 1.4683 1.4686
λ2 -0.6444 -0.6457
ω 0.6158 0.6158
TABLE: Estimates for adjusting the SNMIX (˜)
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
40. The Skew Normal Mixtures
NUMERICAL RESULTS
−0.05 −0.04 −0.03 −0.02 −0.01 0 0.01 0.02 0.03 0.04
0
50
100
150
200
250
Density functions adjusted for daily returns on S&P 500 Index
Daily returns
Frequency
Histogram
Normal
Empirical
NMix
SNMix
FIGURE: Returns vs. normal distribution and SNMIX.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
41. The Skew Normal Mixtures
Strike Maturity
(K) τ = 0.25 τ = 0.5 τ = 0.75 τ = 1.0
1300 122,7236 146,2137 168,1804 188,9434
1350 80,8813 106,9346 130,0796 151,5625
1400 47,3226 73,8094 96,9744 118,4221
1450 24,1159 47,8415 69,5369 90,0320
1500 10,5793 29,0350 47,9058 66,5672
TABLE: Comparison prices of call option Black Scholes.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
42. The Skew Normal Mixtures
NUMERICAL RESULTS
Strike Maturity
(K) τ = 0.25 τ = 0.5 τ = 0.75 τ = 1.0
1300 123,0441 143,9754 164,5054 184,3572
1350 78,9742 103,5501 125,8430 146,8139
1400 44,7446 70,3303 92,9945 114,1851
1450 22,2624 45,2095 66,5457 86,9097
1500 10,7043 28,0958 46,5603 65,1000
TABLE: Comparison prices of call option Corrado & Su.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
43. The Skew Normal Mixtures
NUMERICAL RESULTS
Strike Maturities
(K) τ = 0.25 τ = 0.5 τ = 0.75 τ = 1.0
1300 121,8372 143,5546 164,1304 183,8455
1350 76,7124 99,9953 121,4675 141,8595
1400 35,8423 59,5374 81,2789 101,9446
1450 8,3470 25,7672 45,2634 65,0156
1500 1,6477 8,0504 19,5020 34,6045
TABLE: Comparison prices of call option SNMIX.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
44. The Skew Normal Mixtures
88 90 92 94 96 98 100 102 104 106
−10
0
10
20
30
40
50
60
70
Moneyness (%)
Calloption
BS
CS
SNMIX
FIGURE: Call option for different strikes.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
45. The Skew Normal Mixtures
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
105
110
115
120
125
130
135
140
145
150
Maturities
Calloption
BS
CS
SNMIX
FIGURE: Call option for different maturities.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
46. The Skew Normal Mixtures
NUMERICAL RESULTS
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
Strike
Impliedvolatility
BS
SNMIX
FIGURE: Implied volatility.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
47. Credit Risk
CREDIT RISK
The credit risk problem is one of the most important issues in
finance.
It is the risk of counter parties not fulfilling their obligations in full
on the due date.
Losses can result from both default and decline in the market
value due to deterioration of credit quality of an issuer or counter
party. For example, a corporate bond issued by a company has a
credit risk associated with it.
Credit risk analysis basically consists of finding the likelihood of
default of an obligor going into debt.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
48. Credit Risk
CREDIT RISK
The credit risk problem is one of the most important issues in
finance.
It is the risk of counter parties not fulfilling their obligations in full
on the due date.
Losses can result from both default and decline in the market
value due to deterioration of credit quality of an issuer or counter
party. For example, a corporate bond issued by a company has a
credit risk associated with it.
Credit risk analysis basically consists of finding the likelihood of
default of an obligor going into debt.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
49. Credit Risk
CREDIT RISK
The credit risk problem is one of the most important issues in
finance.
It is the risk of counter parties not fulfilling their obligations in full
on the due date.
Losses can result from both default and decline in the market
value due to deterioration of credit quality of an issuer or counter
party. For example, a corporate bond issued by a company has a
credit risk associated with it.
Credit risk analysis basically consists of finding the likelihood of
default of an obligor going into debt.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
50. Credit Risk
CREDIT RISK
The credit risk problem is one of the most important issues in
finance.
It is the risk of counter parties not fulfilling their obligations in full
on the due date.
Losses can result from both default and decline in the market
value due to deterioration of credit quality of an issuer or counter
party. For example, a corporate bond issued by a company has a
credit risk associated with it.
Credit risk analysis basically consists of finding the likelihood of
default of an obligor going into debt.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
51. Credit Risk
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
52. Credit Risk
CREDIT RISK
Recent financial crisis has stressed the importance of the study of
the correlations in the financial market.
In this regard, the study of the risk of default of the counter party,
in any financial contract, has become crucial in the credit risk. It is
the most intensely studied topic in modern financial world.
There are many parameters associated with bond issuer or bond
itself, which quantify the credit risk associated with it. Credit rating
is one of the important parameters. Credit rating of a credit risky
bond, issued by a company, is an indicator of its creditworthiness.
Better the credit rating of a bond, safer it is.
Credit rating plays a very important role in evaluating the
probability of default and thus quantifying the credit risk.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
53. Credit Risk
CREDIT RISK
Recent financial crisis has stressed the importance of the study of
the correlations in the financial market.
In this regard, the study of the risk of default of the counter party,
in any financial contract, has become crucial in the credit risk. It is
the most intensely studied topic in modern financial world.
There are many parameters associated with bond issuer or bond
itself, which quantify the credit risk associated with it. Credit rating
is one of the important parameters. Credit rating of a credit risky
bond, issued by a company, is an indicator of its creditworthiness.
Better the credit rating of a bond, safer it is.
Credit rating plays a very important role in evaluating the
probability of default and thus quantifying the credit risk.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
54. Credit Risk
CREDIT RISK
Recent financial crisis has stressed the importance of the study of
the correlations in the financial market.
In this regard, the study of the risk of default of the counter party,
in any financial contract, has become crucial in the credit risk. It is
the most intensely studied topic in modern financial world.
There are many parameters associated with bond issuer or bond
itself, which quantify the credit risk associated with it. Credit rating
is one of the important parameters. Credit rating of a credit risky
bond, issued by a company, is an indicator of its creditworthiness.
Better the credit rating of a bond, safer it is.
Credit rating plays a very important role in evaluating the
probability of default and thus quantifying the credit risk.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
55. Credit Risk
CREDIT RISK
Recent financial crisis has stressed the importance of the study of
the correlations in the financial market.
In this regard, the study of the risk of default of the counter party,
in any financial contract, has become crucial in the credit risk. It is
the most intensely studied topic in modern financial world.
There are many parameters associated with bond issuer or bond
itself, which quantify the credit risk associated with it. Credit rating
is one of the important parameters. Credit rating of a credit risky
bond, issued by a company, is an indicator of its creditworthiness.
Better the credit rating of a bond, safer it is.
Credit rating plays a very important role in evaluating the
probability of default and thus quantifying the credit risk.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
56. Credit Risk
CREDIT RISK
After the Basel II accord in 2004 (now Basel III), these ratings
became instrumental in evaluating the risk of a bond or loan and
hence became an increasingly important instrument in Credit
Risk, as they allow banks to base their capital requirements on
internal as well as external rating systems.
The level of rating changes from time to time because of random
credit risk and thus can be modeled by an appropriate stochastic
process.
In 1997, Jarrow et al. first time applied Markov processes to
capture the time evolution of credit ratings. These Markov models
are called "migration models".
Other papers deals with the problem of suitability of Markov
process for the description of accurate rating dynamics was
addressed.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
57. Credit Risk
CREDIT RISK
After the Basel II accord in 2004 (now Basel III), these ratings
became instrumental in evaluating the risk of a bond or loan and
hence became an increasingly important instrument in Credit
Risk, as they allow banks to base their capital requirements on
internal as well as external rating systems.
The level of rating changes from time to time because of random
credit risk and thus can be modeled by an appropriate stochastic
process.
In 1997, Jarrow et al. first time applied Markov processes to
capture the time evolution of credit ratings. These Markov models
are called "migration models".
Other papers deals with the problem of suitability of Markov
process for the description of accurate rating dynamics was
addressed.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
58. Credit Risk
CREDIT RISK
After the Basel II accord in 2004 (now Basel III), these ratings
became instrumental in evaluating the risk of a bond or loan and
hence became an increasingly important instrument in Credit
Risk, as they allow banks to base their capital requirements on
internal as well as external rating systems.
The level of rating changes from time to time because of random
credit risk and thus can be modeled by an appropriate stochastic
process.
In 1997, Jarrow et al. first time applied Markov processes to
capture the time evolution of credit ratings. These Markov models
are called "migration models".
Other papers deals with the problem of suitability of Markov
process for the description of accurate rating dynamics was
addressed.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
59. Credit Risk
CREDIT RISK
After the Basel II accord in 2004 (now Basel III), these ratings
became instrumental in evaluating the risk of a bond or loan and
hence became an increasingly important instrument in Credit
Risk, as they allow banks to base their capital requirements on
internal as well as external rating systems.
The level of rating changes from time to time because of random
credit risk and thus can be modeled by an appropriate stochastic
process.
In 1997, Jarrow et al. first time applied Markov processes to
capture the time evolution of credit ratings. These Markov models
are called "migration models".
Other papers deals with the problem of suitability of Markov
process for the description of accurate rating dynamics was
addressed.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
60. Credit Risk
Main issues regarding the suitability of Markov processes are
1 Non-Markov behavior: It was proved that the new rating of a
company depends not only on its last rating but on all the previous
ones, the effect called rating momentum.
2 Ageing or time spent in a rating: In the credit risk problem, a
complete knowledge of the duration inside the states is of
fundamental importance. The credit migration probability depends
on the time spent by a company in a particular rating.
3 Time dependence: It means that in general, transition probabilities
tends to vary with the state of the economy, being high during
recession and low during periods of economic expansion. Rating
evaluation at two different time points is different.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
61. Credit Risk
Main issues regarding the suitability of Markov processes are
1 Non-Markov behavior: It was proved that the new rating of a
company depends not only on its last rating but on all the previous
ones, the effect called rating momentum.
2 Ageing or time spent in a rating: In the credit risk problem, a
complete knowledge of the duration inside the states is of
fundamental importance. The credit migration probability depends
on the time spent by a company in a particular rating.
3 Time dependence: It means that in general, transition probabilities
tends to vary with the state of the economy, being high during
recession and low during periods of economic expansion. Rating
evaluation at two different time points is different.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
62. Credit Risk
Main issues regarding the suitability of Markov processes are
1 Non-Markov behavior: It was proved that the new rating of a
company depends not only on its last rating but on all the previous
ones, the effect called rating momentum.
2 Ageing or time spent in a rating: In the credit risk problem, a
complete knowledge of the duration inside the states is of
fundamental importance. The credit migration probability depends
on the time spent by a company in a particular rating.
3 Time dependence: It means that in general, transition probabilities
tends to vary with the state of the economy, being high during
recession and low during periods of economic expansion. Rating
evaluation at two different time points is different.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
63. Credit Risk
CREDIT RISK
Empirical studies document that prior rating changes may carry
predictive power for the direction of future rating changes, the
property called rating momentum.
If a firm gets a lower rating than its previous rating, the probability
that its next rating will also be lower than the present one is high.
Therefore, the Markov property fails, as the current rating does not
fully determine the transition intensities. In the case of upward
movement, this phenomenon does not hold.
In the credit risk migration model, the rating organisations that
gives the rating, estimates the reliability of the company that
issued bonds. In their paper, they applied time-homogeneous
semi-Markov processes (SMP) to solve the first issue.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
64. Credit Risk
CREDIT RISK
Empirical studies document that prior rating changes may carry
predictive power for the direction of future rating changes, the
property called rating momentum.
If a firm gets a lower rating than its previous rating, the probability
that its next rating will also be lower than the present one is high.
Therefore, the Markov property fails, as the current rating does not
fully determine the transition intensities. In the case of upward
movement, this phenomenon does not hold.
In the credit risk migration model, the rating organisations that
gives the rating, estimates the reliability of the company that
issued bonds. In their paper, they applied time-homogeneous
semi-Markov processes (SMP) to solve the first issue.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
65. Credit Risk
CREDIT RISK
Empirical studies document that prior rating changes may carry
predictive power for the direction of future rating changes, the
property called rating momentum.
If a firm gets a lower rating than its previous rating, the probability
that its next rating will also be lower than the present one is high.
Therefore, the Markov property fails, as the current rating does not
fully determine the transition intensities. In the case of upward
movement, this phenomenon does not hold.
In the credit risk migration model, the rating organisations that
gives the rating, estimates the reliability of the company that
issued bonds. In their paper, they applied time-homogeneous
semi-Markov processes (SMP) to solve the first issue.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
66. Credit Risk
CREDIT RISK
Empirical studies document that prior rating changes may carry
predictive power for the direction of future rating changes, the
property called rating momentum.
If a firm gets a lower rating than its previous rating, the probability
that its next rating will also be lower than the present one is high.
Therefore, the Markov property fails, as the current rating does not
fully determine the transition intensities. In the case of upward
movement, this phenomenon does not hold.
In the credit risk migration model, the rating organisations that
gives the rating, estimates the reliability of the company that
issued bonds. In their paper, they applied time-homogeneous
semi-Markov processes (SMP) to solve the first issue.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
67. Credit Risk
CREDIT RISK - MOTIVATION
We will solve first and second issues, described above, by the use
of Markov regenerative process (MRGP) with embedded
semi-Markov process. Since rating momentum exists only in
downward moving rating, we can say Markov property is satisfied
only when there is a migration from a given rating to a better
rating.
This behavior of the ratings can be modeled by MRGP. Also, the
time spent in a state is a general distribution in MRGP.
Hence, MRGP models both the randomness of time in the
transition between two states and non-Markovian behavior of the
states. The fact that in the proposed MRGP model, the embedded
process is SMP, makes it unique.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
68. Credit Risk
CREDIT RISK - MOTIVATION
We will solve first and second issues, described above, by the use
of Markov regenerative process (MRGP) with embedded
semi-Markov process. Since rating momentum exists only in
downward moving rating, we can say Markov property is satisfied
only when there is a migration from a given rating to a better
rating.
This behavior of the ratings can be modeled by MRGP. Also, the
time spent in a state is a general distribution in MRGP.
Hence, MRGP models both the randomness of time in the
transition between two states and non-Markovian behavior of the
states. The fact that in the proposed MRGP model, the embedded
process is SMP, makes it unique.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
69. Credit Risk
CREDIT RISK - MOTIVATION
We will solve first and second issues, described above, by the use
of Markov regenerative process (MRGP) with embedded
semi-Markov process. Since rating momentum exists only in
downward moving rating, we can say Markov property is satisfied
only when there is a migration from a given rating to a better
rating.
This behavior of the ratings can be modeled by MRGP. Also, the
time spent in a state is a general distribution in MRGP.
Hence, MRGP models both the randomness of time in the
transition between two states and non-Markovian behavior of the
states. The fact that in the proposed MRGP model, the embedded
process is SMP, makes it unique.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
70. Credit Risk
MARKOV REGENERATIVE PROCESS (MRGP)
Firstly, we give the definition of Markov renewal sequence (MRS). A
sequence of random variables {(Xn, Tn), n = 0, 1, . . .} is called a
Markov renewal sequence if
1 T0 = 0, Tn+1 ≥ Tn; Xn ∈ Ω = {0, 1, 2, . . .}
2 ∀ n ≥ 0,
P{Xn+1 = j, Tn+1 − Tn ≤ t | Xn = i, Tn, Xn−1, Tn−1, . . . , X0, T0}
= P{Xn+1 = j, Tn+1 − Tn ≤ t | Xn = i} (Markov property)
= P{X1 = j, T1 − T0 ≤ t | X0 = i} (time homogenity)
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
71. Credit Risk
MARKOV REGENERATIVE PROCESS (MRGP)
Firstly, we give the definition of Markov renewal sequence (MRS). A
sequence of random variables {(Xn, Tn), n = 0, 1, . . .} is called a
Markov renewal sequence if
1 T0 = 0, Tn+1 ≥ Tn; Xn ∈ Ω = {0, 1, 2, . . .}
2 ∀ n ≥ 0,
P{Xn+1 = j, Tn+1 − Tn ≤ t | Xn = i, Tn, Xn−1, Tn−1, . . . , X0, T0}
= P{Xn+1 = j, Tn+1 − Tn ≤ t | Xn = i} (Markov property)
= P{X1 = j, T1 − T0 ≤ t | X0 = i} (time homogenity)
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
72. Credit Risk
MARKOV REGENERATIVE PROCESS (MRGP)
A stochastic process {Z(t), t ≥ 0} on Ω is called an MRGP if there
exists a Markov renewal sequence {(Xn, Tn), n = 0, 1, . . .} such that all
conditional finite dimensional distributions of {Z(Tn + t), t ≥ 0} given
{Z(u), 0 ≤ u ≤ Tn, Xn = i} are same as those of {Z(t), t ≥ 0} given
X0 = i, i ∈ Ω ⊂ Ω. This implies that in this case {Z(T+
n ), n = 0, 1, . . .} or
{Z(T−
n ), n = 0, 1, . . .} is an embedded discrete time Markov chain
(DTMC) and also that Tn’s are regeneration points of {Z(t), t ≥ 0}.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
73. Credit Risk
CREDIT RISK
The main difference between semi-Markov processes and Markov
regenerative process is that in semi-Markov processes, all
transition points are regeneration points but in MRGP not all
transition points are regeneration points.
In MRGP the time spent inside a state can be any distribution
unlike Markov environment, where it has to be negative
exponential.
Therefore, by using MRGP in credit risk problem, we can address
the issue of the duration of the rating inside a class called ageing
problem.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
74. Credit Risk
CREDIT RISK
The main difference between semi-Markov processes and Markov
regenerative process is that in semi-Markov processes, all
transition points are regeneration points but in MRGP not all
transition points are regeneration points.
In MRGP the time spent inside a state can be any distribution
unlike Markov environment, where it has to be negative
exponential.
Therefore, by using MRGP in credit risk problem, we can address
the issue of the duration of the rating inside a class called ageing
problem.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
75. Credit Risk
CREDIT RISK
The main difference between semi-Markov processes and Markov
regenerative process is that in semi-Markov processes, all
transition points are regeneration points but in MRGP not all
transition points are regeneration points.
In MRGP the time spent inside a state can be any distribution
unlike Markov environment, where it has to be negative
exponential.
Therefore, by using MRGP in credit risk problem, we can address
the issue of the duration of the rating inside a class called ageing
problem.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
76. Credit Risk
THE GLOBAL AND LOCAL KERNEL
The global kernel K(t) = [Kij(t)]i,j∈Ω associated with the process is
defined as
Kij(t) = P{Z(T1) = j, T1 ≤ t|Z(0) = i}, i, j ∈ Ω , t ≥ 0
and it follows that
pij = lim
t→∞
Kij(t), i, j ∈ Ω
where P = [pij]i,j∈Ω is the one-step transition probability matrix of the
embedded Markov chain with state space Ω .
Also, the local kernel E(t) = [Eij(t)]i∈Ω ,j∈Ω given by
Eij(t) = P{Z(t) = j, T1 > t|Z0 = i}, i ∈ Ω , j ∈ Ω , t ≥ 0
describes the behaviour of the process between two regeneration
epochs of the embedded Markov chain.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
77. Credit Risk
TRANSTION PROBABILITIES AND RENEWAL EQUATION
Now, the transition probabilities for {Z(t), t ≥ 0} are defined by
Vij(t) = P{Z(t) = j | Z(0) = i}, i ∈ Ω , j ∈ Ω, t ≥ 0.
They can be obtained by solving the generalised Markov renewal
equation
V(t) = E(t) + K(t) ∗ V(t)
or
Vij(t) = Eij(t) +
γ∈Ω
t
0
Vγj(t − y)dKiγ(y), i ∈ Ω , j ∈ Ω.
In the credit risk environment, the first part of above equation can be
interpreted as the probability that firm has a lower rating j at time t
before the next regeneration time point that is before getting better
rating given the firm was at state i at time 0. In second part of above
equation, Kiγ(y) represents the probability that firm will get a better
rating γ in time y and then firm will migrate to rating j in time (t − y)
following one of the possible paths.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
78. Credit Risk
CREDIT RATING MODEL
If we represent Laplace transform of any matrix A by ˜A, then the
solution of above equation in transform domain is given by
˜V(s) = [I − ˜K(s)]−1
× ˜E(s).
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
79. Credit Risk
CREDIT RATING MODEL
The rating issued to a firm by a rating agency gives its degree of
reliability. For example, there are the following ratings given by
Standard and Poor’s,
Ω = {AAA, AA, A, BBB, BB, B, CCC, D}.
The bonds having rating above BB are investment grade bonds
whereas those having BB or below BB are speculative bonds and state
D corresponds to default. Let us assume that these ratings are
Ω = {1, 2, . . . , 8} in the same order as above.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
80. Credit Risk
DETERMINING THE LOCAL KERNAL
Now, we need to determine Eij(t), i ∈ Ω and j ∈ Ω, the probability of
migrating from i to j without any regeneration (upward movement).
Clearly, we have
Eij(t) = 0 if i > j, t ≥ 0, i ∈ Ω , j ∈ Ω.
For i ≤ j, let Ω(i) for each i ∈ Ω be the set of all states reachable from
state i by downward movement that is Ω(i) = {i, i + 1, ..., 8}. Let
{M(i)(t); t ≥ 0}, i ∈ Ω be the subordinated semi-Markov process (since
sojourn time in each state is general distribution in which there are
only downward movements from state i but no upward movement.
Here, by choosing parameters such that probability of downward
movement is high, the problem of rating momentum can be addressed.
The transition probabilities of this semi-Markov process for each i ∈ Ω
will give us Eij(t) ∀j, local kernel of MRGP.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
81. Credit Risk
DETERMINING THE LOCAL KERNAL
Let (Yn, Sn) be the Markov renewal sequence where Yn represents the
state at nth transition and Sn, n ≥ 1 with state space equal to R+
represents the time of nth transition. The kernel Q(t) = [Qjk(t)], j, k ∈ Ω
of subordinated SMP is defined by
Qjk(t) = P(Yn+1 = k; Sn+1 − Sn ≤ t | Yn = j), j, k ∈ Ω
and it follows that
qjk = lim
t→∞
Qjk(t), j, k ∈ Ω
is the transition probabilities of the embedded Markov chain in the
process. Therefore, for j, k ∈ Ω
Gjk(t) = P(Sn+1 − Sn ≤ t | Yn = j, Yn+1 = k) =
Qjk(t)
qjk
if qjk = 0
1 if qjk = 0
.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
82. Credit Risk
DETERMINING THE LOCAL KERNAL
Let Fij(t); i ∈ {1, 2, . . . , 7} and j ∈ Ω represents the distribution of
migrating from i to j. It basically represents the distribution function of
the waiting time in each state i, given that next state is known. Clearly,
Gjk(t) is same as Fjk(t). Furthermore, probability that the process will
be in state j at time t is given by
Hj(t) = P(Sn+1 − Sn ≤ t | Yn = j), j ∈ Ω.
We can observe that
Hj(t) =
8
k=j
Qjk(t).
Now, the transition probabilities of the SMP {M(t), t ≥ 0} are defined by
φjk(t) = P(M(t) = k | M(0) = j), j, k ∈ Ω
and are obtained by solving the following renewal equation
k tArunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
83. Credit Risk
DETERMINING THE LOCAL KERNAL
In the matrix form, it can be written as
φ(t) = H(t) + Q(t) ∗ φ(t).
where H(t) is the diagonal matrix with jth diagonal entry 1 − Hj(t). This
φ(t) will be upper triangular matrix. In the Laplace domain, solution of
above equation is given by
˜φ(s) = [I − ˜Q(s)]−1 ˜H(s).
Therefore, for each i ∈ Ω , this φij(t), j ∈ Ω will give Eij(t), j ∈ Ω of the
MRGP and hence describes the behavior of rating evolution between
two regeneration epochs that is how the rating moves to lower rating
before going to upper rating. Hence, by taking Fij(t) with different
intensities, increasing or decreasing, we can take care of downward
momentum using local kernel.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
84. Credit Risk
DETERMINING THE GLOBAL KERNAL
Let pij be the one-step transition probabilities for the embedded
Markov chain. Then, for i, j ∈ Ω
pij = lim
t→∞
Kij(t).
The global kernel K(t) = [Kij(t)]i,j∈Ω is given by
Kij(t) = P(Z(T1) = j; T1 ≤ t | Z(0) = i) = Fij(t)pij if i > j
since we need only one regeneration epoch before time t. In case if
i ≤ j, we need the probability of going to j by a regeneration which is
possible from states greater than j.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
85. Credit Risk
DETERMINING THE GLOBAL KERNAL
Therefore, when i ≤ j, we have
Kij(t) =
6
l=j+1
t
0
Klj(t − s)dEil(s)
where Eil is element of local kernel E(t) given in earlier. Thus,
Kij(t) =
Fij(t)pij if i > j
6
l=j+1
t
0 Klj(t − s)dEil(s) if i ≤ j
.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
86. Credit Risk
TIME-DEPENDENT SOLUTION
The time-dependent solution of the proposed model is given by
Vij(t) = Eij(t) +
γ∈Ω
t
0
Vγj(t − y)dKiγ(y); i ∈ Ω , j ∈ Ω
where E(t) and K(t) are given earlier.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
87. Credit Risk
TIME-DEPENDENT SOLUTION
This equation can be solved numerically using discretization. Let h be
the step size of discretization, then we have the countable linear
system given by
Vh
ij(kh) = Eij(kh) +
γ∈Ω
k
τ=1
diγ(τh)Vh
γj((k − τ)h), k = 0, 1, . . .
where
dij(kh) =
Kij(kh) − Kij((k − 1)h) if k > 0
0 if k = 0.
.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
88. Credit Risk
TIME-DEPENDENT SOLUTION
Eij(kh) is given by numerical solution of SMP,
Eij(kh) = gij(kh) +
8
l=i
k
τ=1
qil(τh)Elj((k − τ)h)
where
qij(kh) =
Qij(kh) − Qij((k − 1)h) if k > 0
0 if k = 0
and
gij(kh) =
1 − Hi(kh) if i = j
0 if i = j
.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
89. Credit Risk
TIME-DEPENDENT SOLUTION
In the Laplace transform domain the solution can be obtained as
˜V(s) = [I − ˜K(s)]−1 ˜E(s)
where ˜E(s) is given by transform solution of subordinated SMP,
˜E(s) = [I − ˜Q(s)]−1 ˜H(s)
where Q(t) and H(t) are described in subsection 3.1.
On solving the above system of equations, the following measures can
be obtained:
Vij(t), which represents the probability of being in state j after a
time t given that initially it was in state i at time 0. This takes into
account the time spent in a particular state (ageing).
7
j=1 Vij(t), which represents the probability that the system will
never go into the default state in time t.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
90. Credit Risk
TIME-DEPENDENT SOLUTION
In the Laplace transform domain the solution can be obtained as
˜V(s) = [I − ˜K(s)]−1 ˜E(s)
where ˜E(s) is given by transform solution of subordinated SMP,
˜E(s) = [I − ˜Q(s)]−1 ˜H(s)
where Q(t) and H(t) are described in subsection 3.1.
On solving the above system of equations, the following measures can
be obtained:
Vij(t), which represents the probability of being in state j after a
time t given that initially it was in state i at time 0. This takes into
account the time spent in a particular state (ageing).
7
j=1 Vij(t), which represents the probability that the system will
never go into the default state in time t.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
91. Credit Risk
NUMERICAL EXAMPLE
we illustrate the applicability of the proposed model with the help of a
numerical example. Since real data is not available, we start with the
transition matrix given in Jarrow, Lando and Turnbull (1997) to show
how the proposed model works.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
92. Credit Risk
Table 1. 1 year transition probability matrix P
AAA AA A BBB BB B CCC D
AAA 0.8910 0.0963 0.0078 0.0019 0.0030 0.0000 0.0000 0.0000
AA 0.0086 0.9010 0.0747 0.0099 0.0029 0.0029 0.0000 0.000
A 0.0009 0.0291 0.8896 0.0649 0.0101 0.0045 0.0000 0.0009
BBB 0.0006 0.0043 0.0656 0.8428 0.0644 0.0160 0.0018 0.0045
BB 0.0004 0.0022 0.0079 0.0719 0.7765 0.1043 0.0127 0.0241
B 0.0000 0.0019 0.0031 0.0066 0.0517 0.8247 0.0435 0.0685
CCC 0.0000 0.0000 0.0116 0.0116 0.0203 0.0754 0.6492 0.2319
D 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.0000
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
93. Credit Risk
NUMERICAL EXAMPLE
Assuming Weibull distribution for Fi,j(t), i, j ∈ Ω with parameters as
given in Carty and Fons (1994) [1], we have generated the rating
migration matrices at different time instants. The transition matrices
constructed by the proposed model at times 5, 12 and 15 years are
given in Table 2,3 and 4 respectively. For example, the element 0.0659
in row BBB and in column BB represents the probability that a firm will
have a rating BB at time 12 given that it initially started with rating BBB.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
94. Credit Risk
NUMERICAL EXAMPLE
Table 2. 5 year transition probability matrix, Vij(5)
AAA AA A BBB BB B CCC D
AAA 0.9997 0.0002 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000
AA 0.0001 0.9967 0.0026 0.0004 0.0001 0.0001 0.0000 0.0000
A 0.0001 0.0002 0.9943 0.0034 0.0008 0.0005 0.0003 0.0004
BBB 0.0001 0.0002 0.0006 0.9801 0.0102 0.0041 0.0021 0.0026
BB 0.0002 0.0002 0.0002 0.0006 0.9717 0.0161 0.0043 0.0062
B 0.0001 0.0001 0.0002 0.0002 0.0005 0.9710 0.0114 0.0168
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
95. Credit Risk
NUMERICAL EXAMPLE
Table 3. 12 year transition probability matrix, Vij(12)
AAA AA A BBB BB B CCC D
AAA 0.9003 0.0879 0.0065 0.0017 0.0027 0.0005 0.0003 0.0001
AA 0.0006 0.9422 0.0445 0.0064 0.0025 0.0023 0.0007 0.0008
A 0.0003 0.0006 0.9108 0.0554 0.0132 0.0086 0.0047 0.0064
BBB 0.0002 0.0002 0.0007 0.8735 0.0659 0.0268 0.0132 0.0195
BB 0.0002 0.0002 0.0003 0.0007 0.8377 0.0962 0.0239 0.0408
B 0.0001 0.0001 0.0002 0.0002 0.0004 0.8658 0.0473 0.0859
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
96. Credit Risk
NUMERICAL EXAMPLE
Table 4. 15 year transition probability matrix, Vij(15)
AAA AA A BBB BB B CCC D
AAA 0.8907 0.0960 0.0075 0.0019 0.0027 0.0006 0.0004 0.0002
AA 0.0005 0.9198 0.0613 0.0092 0.0035 0.0034 0.0010 0.0013
A 0.0002 0.0003 0.8874 0.0696 0.0168 0.0112 0.0058 0.0087
BBB 0.0001 0.0001 0.0004 0.8432 0.0811 0.0338 0.0156 0.0257
BB 0.0002 0.0002 0.0002 0.0005 0.7974 0.1192 0.0290 0.0533
B 0.0001 0.0001 0.0001 0.0002 0.0003 0.8345 0.0555 0.1092
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
98. Credit Risk
Table 6. Discrete time distribution function of first time of default in a
time horizon of 15 years given rating at time 0.
1 2 3 4 5 6 7 8
AAA 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
AA 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0001 0.0002
A 0.0000 0.0000 0.0001 0.0002 0.0004 0.0007 0.0013 0.0020
BBB 0.0000 0.0002 0.0007 0.0015 0.0026 0.0042 0.0061 0.0084
BB 0.0001 0.0007 0.0018 0.0036 0.0062 0.0094 0.0134 0.0181
B 0.0006 0.0025 0.0059 0.0108 0.0170 0.0245 0.0331 0.0427
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
99. Credit Risk
Table 6. Discrete time distribution function of first time of default in a
time horizon of 15 years given rating at time 0.
9 10 11 12 13 14 15
AAA 0.0000 0.0000 0.0001 0.0001 0.0002 0.0002 0.0002
AA 0.0003 0.0005 0.0007 0.0008 0.0010 0.0013 0.0015
A 0.0029 0.0040 0.0052 0.0064 0.0076 0.0087 0.0098
BBB 0.0109 0.0136 0.0165 0.0195 0.0226 0.0257 0.0289
BB 0.0232 0.0288 0.0347 0.0408 0.0470 0.0533 0.0597
B 0.0529 0.0637 0.0748 0.0861 0.0976 0.1092 0.1207
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
100. Operational Risk
INTRODUCTION
The Operational risk (OpRisk) management has become an
increasingly important component in financial institutions and
plays a key role in the development of integrated risk
management.
OpRisk frequently arises due to other type of risks, and the size of
an OpRisk may be intensely impacted by market or credit risk
factors.
The definition of OpRisk is defined by Basel II as “The risk of
direct or indirect loss resulting from inadequate or failed internal
processes, people and systems or from external events”.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
101. Operational Risk
INTRODUCTION
The Operational risk (OpRisk) management has become an
increasingly important component in financial institutions and
plays a key role in the development of integrated risk
management.
OpRisk frequently arises due to other type of risks, and the size of
an OpRisk may be intensely impacted by market or credit risk
factors.
The definition of OpRisk is defined by Basel II as “The risk of
direct or indirect loss resulting from inadequate or failed internal
processes, people and systems or from external events”.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
102. Operational Risk
INTRODUCTION
The Operational risk (OpRisk) management has become an
increasingly important component in financial institutions and
plays a key role in the development of integrated risk
management.
OpRisk frequently arises due to other type of risks, and the size of
an OpRisk may be intensely impacted by market or credit risk
factors.
The definition of OpRisk is defined by Basel II as “The risk of
direct or indirect loss resulting from inadequate or failed internal
processes, people and systems or from external events”.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
103. Operational Risk
In recent years, considerable importance has been given to model
the Oprisk in the financial world. As a result of its relevance, the
Basel Committee on Banking Supervision (BCBS) has included a
specific section on Oprisk in the document “International
Convergence of measures and capital standards: a Revised
Framework”, known as Basel II, a specific section on OpRisk.
OpRisk measurement focuses on the calculation of capital for
OpRisk and Basel II offers three methods for calculating OpRisk.
One of the three OpRisk methods, the “advanced measurement
approach (AMA)” is the most frequently used method.
The Loss distribution approach (LDA) in the banking industry is
one of the paradigms of the AMA in the banking sector and
actuarial mathematics, in particular insurance sector. An
understanding of loss processes is a crucial element in the
determination of OpRisk.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
104. Operational Risk
In recent years, considerable importance has been given to model
the Oprisk in the financial world. As a result of its relevance, the
Basel Committee on Banking Supervision (BCBS) has included a
specific section on Oprisk in the document “International
Convergence of measures and capital standards: a Revised
Framework”, known as Basel II, a specific section on OpRisk.
OpRisk measurement focuses on the calculation of capital for
OpRisk and Basel II offers three methods for calculating OpRisk.
One of the three OpRisk methods, the “advanced measurement
approach (AMA)” is the most frequently used method.
The Loss distribution approach (LDA) in the banking industry is
one of the paradigms of the AMA in the banking sector and
actuarial mathematics, in particular insurance sector. An
understanding of loss processes is a crucial element in the
determination of OpRisk.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
105. Operational Risk
In recent years, considerable importance has been given to model
the Oprisk in the financial world. As a result of its relevance, the
Basel Committee on Banking Supervision (BCBS) has included a
specific section on Oprisk in the document “International
Convergence of measures and capital standards: a Revised
Framework”, known as Basel II, a specific section on OpRisk.
OpRisk measurement focuses on the calculation of capital for
OpRisk and Basel II offers three methods for calculating OpRisk.
One of the three OpRisk methods, the “advanced measurement
approach (AMA)” is the most frequently used method.
The Loss distribution approach (LDA) in the banking industry is
one of the paradigms of the AMA in the banking sector and
actuarial mathematics, in particular insurance sector. An
understanding of loss processes is a crucial element in the
determination of OpRisk.
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu
107. Operational Risk
THANK YOU VERY MUCH
Arunachalam (UNAL) Modelos estocásticos para la valoración de riesgo financieroXXIII Seminario financiero: Las finanzas cu