Materials for the course

The course is based on book “Mathematical Modeling and Computation

in Finance: With Exercises and Python and MATLAB Computer Codes”,
by C.W. Oosterlee and L.A. Grzelak, World Scientific Publishing Europe
Ltd, 2019. For more details go here.

I Youtube Channel with courses can be found here.

I Slides and the codes can be found here.
Lech A. Grzelak Financial Engineering- Interest Rates and xVA 1 / 30
List of content

3.1. Equilibrium vs. Term-Structure Models

3.2. The HJM Framework
3.3. The Instantaneous Forward Rate
3.4. Arbitrage Free Conditions under HJM
3.5. Ho-Lee Model and Python Simulation
3.6. Hull-White Model
3.7. Hull-White Model and Simulation in Python
3.8. Summary of the Lecture + Homework

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 2 / 30

 Equilibrium vs. Term-Structure Models

Equilibrium vs. Term-Structure Models

I A no-arbitrage model is a model designed to be consistent with

today’s term structure of interest rates.
I The difference between equilibrium (endogenous) and non-arbitrage
models (exogenous) is that today’s term structure of interest rates is
an output in an equilibrium model.
I In a no-arbitrage model, today’s term structure of interest rates is an
input. This means that we take the observed actual rates while
constructing the model and estimate the unobserved rates.
I The HJM framework described a clear path from the equilibrium
towards term-structure models.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 3 / 30

 Equilibrium vs. Term-Structure Models

Equilibrium vs. Term-Structure Models1

I Historically, equilibrium models start with assumptions about

economic variables and derive a process for the short rate, which
means that the current term structure of interest rates hence is an
output rather than input in the model.
I Such models are also called endogenous term-structure models. The
(instantaneous) short rate at time t is the rate that applies to an
infinitesimally short period at time t.
I Some popular equilibrium models. Namely, the Vasicek:

dr (t) = λ (θ − r (t)) dt + ηdW (t),

and the Cox, Ingersoll and Ross (CIR) model:

p
dr (t) = λ (θ − r (t)) dt + γ r (t)dW (t).

I These models are one-factor models, which have several shortcomings,

e.g., the interest rates are perfectly correlated between different maturities.
1 Great overview of the short-rate model can be found in book of Brigo-Mercurio:

Interest Rate Models- Theory and Practice.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 4 / 30
 Equilibrium vs. Term-Structure Models

Equilibrium vs. Term-Structure Models

0.4 1.4

0.3 1.2

0.2 1

0.1 0.8

0 0.6
r(t)

r(t)
-0.1 0.4

-0.2 0.2

-0.3 0

-0.4 -0.2

-0.5 -0.4

-0.6 -0.6
0 10 20 30 40 50 0 10 20 30 40 50
time time

Figure: Within the Vasicek Model model context, the impact of variation of
mean-reversion λ, and of the volatility parameter η on the Monte Carlo paths.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 5 / 30

 Equilibrium vs. Term-Structure Models

Equilibrium vs. Term-Structure Models

0.18 0.4

0.16 0.35

0.14
0.3

0.12
0.25
0.1
r(t)

r(t)
0.2
0.08
0.15
0.06

0.1
0.04

0.02 0.05

0 0
0 10 20 30 40 50 0 10 20 30 40 50
time time

Figure: For the CIR model, the impact of variation of mean-reversion λ, and of
the volatility parameter γ on the Monte Carlo paths.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 6 / 30

 The HJM Framework

The HJM framework

I The Heath-Jarrow-Morton framework represents a class of models

that are derived by directly modeling the dynamics of instantaneous
forward-rates.
I The framework constitutes the fundament for interest rate models as
it provides an explicit relation between the volatility of the
instantaneous forward rates and arbitrate-free drift.
I Both the standard short-rate and Libor Market models can be
derived in the HJM framework however, in general.
I The HJM models are non-Markovian so only a number of models
with a closed-form solution exists.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 7 / 30

 The HJM Framework

The HJM framework

I 1. Equilibrium models.
I 2. Short-rate models in the HJM Framework.
I 3. Market Models models in the HJM Framework.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 8 / 30

 The Instantaneous Forward Rate

Instantaneous Forward Rate

Definition (Forward and instantaneous forward rates)

Suppose that at time t we enter into a forward contract to deliver at
time S a bond that will mature at time T . Let the forward price of the
bond be denoted by P(t, S, T ). At the same time, a zero-coupon bond,
P(t, S), that matures at time S is purchased. Further, a bond, P(t, T ),
that matures at time T is also bought. Assuming no-arbitrate and
market completeness the following equality has to hold:

P(t, T ) = P(t, S)P(t, S, T ).

Now, we define the implied forward rate, R(t, S, T ), at time t for the
period [S, T ] as:

P(t, S, T ) = exp (−(T − S)R(t, S, T )) .

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 9 / 30

 The Instantaneous Forward Rate

Instantaneous Forward Rate

Definition (Forward and instantaneous forward rates cont.)

By equating the two equations we find:

P(t, T )
e−(T −S)R(t,S,T ) = ,
P(t, S)

which reads the forward rate R(t, S, T ) to be given by:

log P(t, T ) − log P(t, S)

R(t, S, T ) = − .
T −S
By setting the limit T − S → 0 we arrive at the definition of the
instantaneous forward rate

def ∂
f (t, T ) = lim R(t, S, T ) = − log P(t, T ).
S→T ∂T

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 10 / 30

 The Instantaneous Forward Rate

Instantaneous Forward Rate

I In the HJM framework the dynamics of the instantaneous forward

rates,f (t, T ), is analyzed.
I We start with an assumption that for a certain, fixed, maturity
T ≥ 0, the instantaneous forward rate f (t, T ) under real-world
measure P is driven by the following dynamics:

df (t, T ) = αP (t, T )dt + σ(t, T )dW P (t), f (0, T ) = f0 (T ),

for any time t < T , with a corresponding drift αP (t, T ).

I Under this model we also define a money-savings account as:
Z t  Z t 
M(t) = exp r (s)ds ≡ exp f (s, s)ds .
0 0

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 11 / 30

 Arbitrage Free Conditions under HJM

Arbitrage-free HJM

I As we see, under the HJM framework the short rate r (t), is defined
as the limit of the instantaneous forward rate r (t) ≡ f (t, t). The
zero-coupon bond, P(t, T ), with maturity T , follows:
 
1
P(t, T ) = M(t)EQ · 1 F(t)
M(T )
" ! #
Z T
= EQ exp − r (s)ds · 1 F(t) .
t

I What are the tradables in this market and what quantities are the
martingales ?
I How to find αQ (t, T ) in

df (t, T ) = αQ (t, T )dt + σ(t, T )dW Q (t), f (0, T ) = f0 (T ), ?

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 12 / 30

 Arbitrage Free Conditions under HJM

Arbitrage-free HJM cont.

I Although the ZCB P(t, T ) can be priced as an expectation its value

can be directly related to today’s yield curve via:
∂
f (t, T ) = − log P(t, T ).
∂T
I From above we can easily determine the following relation:
!
Z T
P(t, T ) = exp − f (t, s)ds ,
t

I Let us start with deriving the dynamics of discounted ZCB:

  " Z T Z t !#
P(t, T ) def
d = d exp − f (t, s)ds − r (s)ds = dZ (t).
M(t) t 0

I After a lot of derivations for dZ (t) and setting the drift to zero we
find.
Lech A. Grzelak Financial Engineering- Interest Rates and xVA 13 / 30
 Arbitrage Free Conditions under HJM

Arbitrage-free HJM cont.

Lemma (HJM no arbitrage drift condition)

For the instantaneous forward rates given by following SDE:

df (t, T ) = αQ (t, T )dt + σ(t, T )dW Q (t),

the no-arbitrage drift condition is given by

Z T
αQ (t, T ) = σ(t, T ) σ(t, s)ds.
t

Proof.
The proof easily follows by deriving the dynamics of Z (t) and equating
all the drift terms to zero.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 14 / 30

 Arbitrage Free Conditions under HJM

Short-Rate dynamics under HJM

I By using the relation that f (t, t) = r (t) and by integrating the
SDEs we obtain the short rate dynamics under the HJM framework
of the form:
Z t Z t
f (t, T ) = f (0, T ) + αQ (s, T )ds + σ(s, T )dW Q (s),
0 0
which for time T = t simply becomes:
Z t Z t
r (t) ≡ f (t, t) = f (0, t) + αQ (s, t)ds + σ(s, t)dW Q (s),
0 0

I Now by applying Leibniz integral rule 2 we obtain following short rate

dynamics:
 Z t
∂ ∂ Q
dr (t) = f (0, t) + αQ (t, t) + α (s, t)ds
∂t 0 ∂t
Z t 
∂
+ σ(s, t)dW Q (s) dt + σ(t, t)dW Q (t).
0 ∂t
2

d
Z b(α) ∂ ∂
Z b(α) ∂
f (x, α)dx = f (b, α) b(α) − f (a, α) a(α) + f (x, α)dx
dα a(α) ∂α ∂α a(α) ∂α
Lech A. Grzelak Financial Engineering- Interest Rates and xVA 15 / 30
 Ho-Lee Model and Python Simulation

Ho-Lee Model
I We specify a certain form of a volatility σ(t, T ) for the
instantaneous forward rate f (t, T ) and determine the resulting
short-rate dynamics. The first, and the simplest possibility is to
consider σ(t, T ) to be constant, i.e.:

σ(t, T ) = σ.

I From previous derivations we find:

Z T
Q
α (t, T ) = σ σds = σ 2 (T − t).
t

I This can be used in Equation for the short-rate dynamics, i.e.:

 Z t
∂ ∂ Q
dr (t) = f (0, t) + αQ (t, t) + α (s, t)ds
∂t 0 ∂t
Z t 
∂
+ σ(s, t)dW Q (s) dt + σ(t, t)dW Q (t).
0 ∂t

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 16 / 30

 Ho-Lee Model and Python Simulation

Ho-Lee Model

I Therefore the short-rate dynamics under the HJM model with

σ(t, T ) = σ is given by:
 
∂
dr (t) = f (0, t) + σ 2 t dt + σdW Q (t).
∂t

∂
I By setting θ(t) = f (0, t) + σ 2 t, we arrive at:
∂t
dr (t) = θ(t)dt + σdW Q (t),

which is well-recognized as the Ho-Lee short-rate model.

I Now we can compute ZCBs P(t, T ) using this model
h RT i
P(t, T ) = E e− t r (s)ds F(t) = eA(t,T )+B(t,T )r (t) .

I Functions A(t, T ) and B(t, T ) will be presented later.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 17 / 30

 Ho-Lee Model and Python Simulation

Ho-Lee Model: Python Exercise

Python Exercise:
I Define Pmrkt (t, T ) = exp(−r (T − t)) (it can be much more involved
or implied from the market), for some r , calculate f (t, T ) and use it
for simulating r (t).
I Consider the Ho-Lee model with a freely chosen parameter σ.
I It is important to properly choose the initial value for the process
r (t):

∂ log Pmrkt (0, )

r (0) = f (0, 0) ≈ − , for  → 0.
∂
I Using Monte Carlo Paths calculate
h RT i
Pmodel (t, T ) = EQ e− t r (s)ds F(t) .

I Are Pmrkt (0, T ) and Pmodel (0, T ) the same for all T ?

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 18 / 30

 Ho-Lee Model and Python Simulation

Ho-Lee Model: Python Exercise

Figure: Comparison of the ZCBs from the Ho-Lee model vs. Market P(0, T )
for different T .

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 19 / 30

 Hull-White Model

Hull-White Model
I Let us now consider a short-rate model generated by the HJM
volatility given by:

σ(t, T ) = σ · e−λ(T −t) with λ > 0.

As before, the short-rate dynamics under HJM arbitrage free

assumptions is given by:
 Z t
∂ Q ∂ Q
dr (t) = f (0, t) + α (t, t) + α (s, t)ds
∂t 0 ∂t
Z t 
∂
+ σ(s, t)dW (s) dt + σ(t, t)dW Q (t).
Q
0 ∂t

I By Lemma we find:
t
σ 2 −λ(t−s)  −λ(t−s)
Z 
αQ (s, t) = σe−λ(t−s) σe−λ(u−t) du = − e e −1 ,
s λ
which implies that αQ (t, t) = 0.
Lech A. Grzelak Financial Engineering- Interest Rates and xVA 20 / 30
 Hull-White Model

Hull-White Model

I The remaining terms are as follows:

Z t
∂ Q σ 2 −2λt λt
α (s, t)ds = e (e − 1),
0 ∂t λ

and
∂
σ(s, t) = −λσe−λ(t−s) = −λσ(s, t),
∂t
with σ(t, t) = σ.
I The dynamics for r (t) is therefore given by:
Z t Z t
∂ ∂ Q
dr (t) = f (0, t)+ α (s, t)ds−λ σ(s, t)dW Q (s) +σdW Q (t).
∂t 0 ∂t 0

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 21 / 30

 Hull-White Model

Hull-White Model
I We see that Brownian motion dW Q (t) is present in two terms. In
Rt
order to find explicitly the solution for the integral 0 σ(s, t)dW Q (s)
we can use definition of the short-rate which yields:
Z t Z t
r (t) = f (0, t) + αQ (s, t)ds + σ(s, t)dW Q (s),
0 0

therefore the first integral can be determined via:

Z t Z t
σ(s, t)dW Q (s) = r (t) − f (0, t) − αQ (s, t)ds.
0 0

As:
t
σ 2 −2λt λt
Z

2
αQ (s, t)ds =
2
e e −1 ,
0 2λ
we obtain the following dynamics for process r (t):

σ2 
 
1 ∂ 
dr (t) = λ f (0, t) + f (0, t) + 2 1 − e−2λt − r (t) dt + σdW Q (t).
λ ∂t 2λ

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 22 / 30

 Hull-White Model

Hull-White Model

I So finally, by taking:

1 ∂ σ2
f (0, t) + f (0, t) + 2 1 − e−2λt ,


θ(t) =
λ ∂t 2λ
the dynamics of the process r (t) yields:

dr (t) = λ(θ(t) − r (t))dt + σdW Q (t),

which can be easily recognized as the Hull-White short rate process.

I It is important to properly choose the initial value for the process
r (t):

∂ log P(0, )
r (0) = f (0, 0) ≈ − , for  → 0.
∂

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 23 / 30

 Hull-White Model

Simulation of the Hull-White Model

0.4 1.4

0.3 1.2

0.2 1

0.1 0.8

0 0.6
r(t)

r(t)
-0.1 0.4

-0.2 0.2

-0.3 0

-0.4 -0.2

-0.5 -0.4

-0.6 -0.6
0 10 20 30 40 50 0 10 20 30 40 50
time time

Figure: Within the HW model context, the impact of variation of

mean-reversion λ, and of the volatility parameter η on the Monte Carlo paths.

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 24 / 30

 Hull-White Model and Simulation in Python

Hull-White Model: Python Exercise

Python Exercise:
I Define Pmrkt (t, T ) = exp(−r (T − t)) (it can be much more involved
or implied from the market), for some r , calculate f (t, T ) and use it
for simulating r (t).
I Consider the Hull-White model with a freely chosen parameters λ
and σ.
I It is important to properly choose the initial value for the process
r (t):

∂ log Pmrkt (0, )

r (0) = f (0, 0) ≈ − , for  → 0.
∂
I Using Monte Carlo Paths calculate
h RT i
Pmodel (t, T ) = EQ e− t r (s)ds F(t) .

I Are Pmrkt (0, T ) and Pmodel (0, T ) the same for all T ?

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 25 / 30

 Hull-White Model and Simulation in Python

Hull-White Model: Python Exercise

Figure: Comparison of the ZCBs from the Hull-White model vs. Market
P(0, T ) for different T .

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 26 / 30

 Summary of the Lecture + Homework

Summary

I Equilibrium vs. Term-Structure Models

I The HJM Framework
I Arbitrage Free Conditions under HJM
I Ho-Lee Model and Python Simulation
I Hull-White Model
I Hull-White Model and Simulation in Python
I Summary of the Lecture + Homework

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 27 / 30

 Summary of the Lecture + Homework

Homework Exercises
The solutions for the homework can be find at
https://github.com/LechGrzelak/QuantFinanceBook
I Exercise
Consider the “exponential-Vasicek” model given by the following
system of equations:

r (t) = ey (t) ,
dy (t) = (θ − ay (t))dt + σdW (t) y (t0 ) = y0 .

I Show that the dynamics for r (t) yields:

σ2
 
dr (t) = r (t) θ + − a log r (t) dt + σr (t)dW (t).
2
I Show that
θ σ2
lim E[r (t)] = e a + 4a .
t→∞

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 28 / 30

 Summary of the Lecture + Homework

Summary of the Lecture + Homework

I Exercise 11.9
Consider the Vašiček short-rate model,
dr (t) = λ (θ − r (t)) dt + ηdW (t),
with parameters λ = 0.05, θ = 0.02 and η = 0.1 and initial rate
r (t0 ) = 0. At time t0 , we wish to hedge a position in a 10y
zero-coupon bond, P(t0 , 10y ), using two other bonds, P(t0 , 1y ) and
P(t0 , 20y ).
a. Determine two weights, ω1 and ω2 , such that ω1 + ω2 = 1 and
ω1 P(t0 , 1y ) + ω2 P(t0 , 20y ) = P(t0 , 10y ).
b. Perform a minimum variance hedge and determine the weights ω1
and ω2 , such that
Z 10y  Z 10y 
Var ω1 P(t, 1y ) + ω2 P(t, 20y )dt = Var P(t, 10y )dt ,
0 0

while ω1 + ω2 = 1. What can be said about this type of hedge

compared to the point addressed?
c. Change the measure, to the T = 10y -forward measure, and, for the
given weights determined earlier, check whether the variance of the
estimator increases.
Lech A. Grzelak Financial Engineering- Interest Rates and xVA 29 / 30
 Summary of the Lecture + Homework

Summary of the Lecture + Homework

I Exercise
I The “exponential-Vasicek” model does not allow for negative
interest rates. In order to “fix” that problem we introduce a
so-called shift parameter ζ,

r˜(t) = r (t) − ζ, ζ ∈ R+ .

I Find the dynamics for r˜(t), simulate Monte Carlo paths and
compare to r (t).
I Discuss the impact of shift on ZCBs, P(t0 , T ).

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 30 / 30