Lecture Notes: Interest Rate Theory

Lecture Notes: Interest Rate Theory

Josef Teichmann

ETH Zürich

Fall 2010

J. Teichmann Lecture Notes: Interest Rate Theory

Lecture Notes: Interest Rate Theory

Foreword
Mathematical Finance
Basics on Interest Rate Modeling
Black formulas
Affine LIBOR Models
Markov Processes
The SABR model
HJM-models
References
Catalogue of possible questions for the oral exam

1 / 107
Lecture Notes: Interest Rate Theory
Foreword

In mathematical Finance we need processes

I which can model all stylized facts of volatility surfaces and
times series (e.g. tails, stochastic volatility, etc)
I which are analytically tractable to perform efficient calibration.
I which are numerically tractable to perform efficient pricing
and hedging.

2 / 107
Lecture Notes: Interest Rate Theory
Foreword

Goals

I Basic concepts of stochastic modeling in interest rate theory.

I ”No arbitrage” as concept and through examples.
I Concepts of interest rate theory like yield, forward rate curve,
short rate.
I Spot measure, forward measures, swap measures and Black’s
formula.
I Short rate models
I Affine LIBOR models
I Fundamentals of the SABR model
I HJM model
I Consistency and Yield curve estimation
3 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Modeling of financial markets

We are describing models for financial products related to interest

rates, so called interest rate models. We are facing several
difficulties, some of the specific for interest rates, some of them
true for all models in mathematical finance:
I stochastic nature: traded prices, e.g.˜prices of interest rate
related products, are not deterministic!
I information is increasing: every day additional information on
markets appears and this stream of information should enter
into our models.
I stylized facts of markets should be reflected in the model:
stylized facts of time series, trading opportunities (portfolio
building), etc.

4 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Mathematical Finance 1

A financial market can be modeled by

I a filtered (discrete) probability space (Ω, F, Q),
I together with price processes, namely K risky assets
(Sn1 , . . . , SnK )0≤n≤N and one risk-less asset S 0 , i.e.˜ Sn0 > 0
almost surely (no default risk for at least one asset),
I all price processes being adapted to the filtration.
This structure reflects stochasticity of prices and the stream of
incoming information.

5 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

A portfolio is a predictable process φ = (φ0n , . . . , φK

n )0≤n≤N , where
i
φn represents the number of risky assets one holds at time n. The
value of the portfolio Vn (φ) is
K
X
Vn (φ) = φin Sni .
i=0

6 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Mathematical Finance 2

Self-financing portfolios φ are characterized through the condition

K
X
Vn+1 (φ) − Vn (φ) = φin+1 (Sn+1
i
− Sni ),
i=0

for 0 ≤ n ≤ N − 1, i.e.˜changes in value come from changes in

prices, no additional input of capital is required and no
consumption is allowed.

7 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Self-financing portfolios can be characterized in discounted terms.

fn (φ) = (S 0 )−1 Vn (φ)
V n
fi = (S 0 )−1 S i (φ)
Sn n n
K
X
V
fn (φ) = φin S
fi
n
i=0
for 0 ≤ n ≤ N, and recover
n X
X K
fn (φ) = V0 (φ) + (φ · S)
V e = V0 (φ) + φij (Seji − Sg
i
j−1 )
j=1 i=1

for self-financing predictable trading strategies φ and 0 ≤ n ≤ N.

In words: discounted wealth of a self-financing portfolio is the
cumulative sum of discounted gains and losses. Notice that we
apply a generalized notion of “discounting” here, prices S i divided
by S 0 – only these relative prices can be compared.
8 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Fundamental Theorem of Asset Pricing

A minimal condition for modeling financial markets is the

No-arbitrage condition: there are no self-financing trading
strategies φ (arbitrage strategies) with

V0 (φ) = 0, VN (φ) ≥ 0

such that Q(VN (φ) 6= 0) > 0 holds (NFLVR).

9 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Fundamental Theorem of Asset Pricing

A minimal condition for financial markets is the no-arbitrage

condition: there are no self-financing trading strategies φ
(arbitrage strategies) with V0 (φ) = 0, VN (φ) ≥ 0 such that
Q(VN (φ) 6= 0) > 0 holds (NFLVR).
In other words the set

K = {V
fN (φ)| V0 (φ) = 0, φ self-finanancing }

intersects L0≥0 (Ω, F, Q) only at 0,

K ∩ L0≥0 (Ω, F, Q) = {0}.

10 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

FTAP
Theorem
Given a financial market, then the following assertions are
equivalent:
1. (NFLVR) holds.
2. There exists an equivalent measure P ∼ Q such that the
discounted price processes are P-martingales, i.e.
1 i 1
EP ( S |Fn ) = 0 Sni
SN0 N Sn

for 0 ≤ n ≤ N.

Main message: Discounted (relative) prices behave like martingales

with respect to one martingale measure.
11 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

What is a martingale?

Formally a martingale is a stochastic process such that today’s best

prediction of a future value of the process is today’s value, i.e.˜

E [Mn |Fm ] = Mm

for m ≤ n, where E [Mn |Fm ] calculates the best prediction with

knowledge up to time m of the future value Mn .

12 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Random walks and Brownian motions are well-known examples of

martingales. Martingales are particularly suited to describe
(discounted) price movements on financial markets, since the
prediction of future returns is 0.

13 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Pricing rules

(NFLVR) also leads to arbitrage-free pricing rules. Let X be the

payoff of a claim X paying at time N, then an adapted stochastic
process π(X ) is called pricing rule for X if
I πN (X ) = X .
I (S 0 , . . . , S N , π(X )) is free of arbitrage.
This is equivalent to the existence of one equivalent martingale
measure P such that
X
πn (X )
EP 0
|Fn =
SN Sn0

holds true for 0 ≤ n ≤ N.

14 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Proof of FTAP

The proof is an application of separation theorems for convex sets:

we consider the euclidean vector space L2 (Ω, R) of real valued
random variables with scalar product

hX , Y i = E (XY ).

Then the convex set K does not intersect the positive orthant
L2≥0 (Ω, R), hence we can find a vector R, which is strictly positive
and which is orthogonal to all elements of K (draw it!). We are
free to choose E (R) = 1. We can therefore define a measure Q on
F via
Q(A) = E (1A R)
and this measure has the same nullsets as P by strict positivity.
15 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Proof of FTAP

By construction we have that every element of K has vanishing

expectation with respect to Q since R is orthogonal to K . Since K
consists of all stochastic integrals with respect to S̃ we obtain by
Doob’s optional sampling that S̃ is a Q-martingale, which
completes the proof.

16 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

One step binomial model

We model one asset in a zero-interest rate environment just before
the next tick. We assume two states of the world: up, down. The
riskless asset is given by S 0 = 1. The risky asset is modeled by

S01 = S0 , S11 = S0 ∗ u > S0 or S11 = S0 ∗ 1/u = S0 ∗ d

where the events at time one appear with probability q and 1 − q

(”physical measure”). The martingale measure is apparently given
1−d
through u ∗ p + (1 − p)d = 1, i.e.˜ p = u−d .
Pricing a European call option at time one in this setting leads to
fair price

E [(S11 − K )+ ] = p ∗ (S0 u − K )+ + (1 − p) ∗ (S0 d − K )+ .

17 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Black-Merton-Scholes model 1
We model one asset with respect to some numeraire by an
exponential Brownian motion. If the numeraire is a bank account
with constant rate we usually speak of the Black-Merton-Scholes
model, if the numeraire some other traded asset, for instance a
zero-coupon bond, we speak of Black’s model. Let us assume that
S0 = 1, then
σ2t
St1 = S0 exp(σBt − )
2
with respect to the martingale measure P. In the physical measure
Q a drift term is added in the exponent, i.e.˜

σ2t
St1 = S0 exp(σBt − + µt).
2

18 / 107
Lecture Notes: Interest Rate Theory
Mathematical Finance

Black-Merton-Scholes model 2

Our theory tells that the price of a European call option on S 1 at

time T is priced via

E [(ST1 − K )+ ] = S0 Φ(d1 ) − K Φ(d2 )

yielding the Black-Scholes formula, where Φ is the cumulative

distribution function of the standard normal distribution and
2
log SK0 ± σ 2T
d1,2 = √ .
σ T
Notice that this price corresponds to the value of a portfolio
mimicking the European option at time T .

19 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Some general facts

I Fixed income markets (i.e. interest rate related products)

form a large scale market in any major economy, for instance
swaping fixed against floating rates.
I Fixed income markets, in contrast to stock markets, consist of
products with a finite life time (i.e. zero coupon bonds) and
strong dependencies (zero coupon bonds with close maturities
are highly dependent).
I mathematically highly challenging structures can appear in
interest rate modeling.

20 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Interest Rate mechanics 1

Prices of zero-coupon bonds (ZCB) with maturity T are denoted
by P(t, T ). Interest rates are given by a market of (default free)
zero-coupon bonds. We shall always assume the nominal value
P(T , T ) = 1.
I T denotes the maturity of the bond, P(t, T ) its price at a
time t before maturity T .
I The yield
1
Y (t, T ) = − log P(t, T )
T −t
describes the compound interest rate p.˜a. for maturity T .
I The forward rate curve f of the bond market is defined via
Z T
P(t, T ) = exp(− f (t, s)ds)
t
for 0 ≤ t ≤ T .
21 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Interest rate models

An interest rate model is a collection of adapted stochastic

processes (P(t, T ))0≤t≤T on a stochastic basis (Ω, F, P) with
filtration (Ft )t≥0 such that
I P(T , T ) = 1 (nominal value is normalized to one),
I P(t, T ) > 0 (default free market)
holds true for 0 ≤ t ≤ T .

22 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Interest Rate mechanics 2

I The short rate process is given through Rt = f (t, t) for t ≥ 0

defining the “bank account process”
Z t
(B(t))t≥0 := (exp( Rs ds))t≥0 .
0

I The existence of forward rates and short rates is an

assumption on regularity with respect to maturity T . A
sufficient conditions for the existence of Yields and forward
rates is that bond prices are continuously differentiable with
respect with T .

23 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

I Notice that the market to model consists only of ZCB,

apparently the bank account has to be formed from ZCB via a
roll-over-portfolio.
I The roll-over-portfolio consists of investing one unit of
currency into a T1 -ZCB, then reinvesting at time T1 into a
T2 -ZCB, etc. Given an increasing sequence
T = 0 < T1 < T2 < . . . yields the wealth at time t
Y 1
B T (t) =
P(t, Ti )
Ti ≤t

I We speak of a (generalized) “bank account process” of B T

allows for limiting – this is in particular the case of we have a
short rate process with some integrability properties.

24 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Simple forward rates – LIBOR rates

Consider a bond market (P(t, T ))t≤T with P(T , T ) = 1 and

P(t, T ) > 0. Let t ≤ T ≤ T ∗ . We define the simple forward rate
through
 
∗ 1 P(t, T )
F (t; T , T ) := ∗ −1 .
T − T P(t, T ∗ )

We abbreviate

F (t, T ) := F (t; t, T ).

25 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Apparently P(t, T ∗ )F (t; T , T ∗ ) is the fair value at time t of a

contract paying F (T , T ∗ ) at time T ∗ , in the sense that there is a
self-financing portfolio with value P(t, T ∗ )F (t; T , T ∗ ) at time t
and value F (T , T ∗ ) at time T ∗ .

26 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Indeed, note that

P(t, T ) − P(t, T ∗ )
P(t, T ∗ )F (t; T , T ∗ ) = ,
T ∗− T 
∗ 1 1
F (T , T ) = ∗ −1 .
T − T P(T , T ∗ )

We can build∗ a self-financing portfolio at time t at price

P(t,T )−P(t,T )
T ∗ −T yielding F (T , T ∗ ) at time T ∗ :
I Buying a ZCB with maturity T at time t costs P(t, T ),
selling a ZCB with maturity T ∗ amounts all together to
P(t, T ) − P(t, T ∗ ).
I at time T we have to rebalance the portfolio by buying with
the maturing ZCB another bond with maturity T ∗ , precisely
an amount 1/P(T , T ∗ ).
I at time T ∗ we receive 1/P(T , T ∗ ) − 1.
27 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Caps
In the sequel, we fix a number of future dates
T0 < T1 < . . . < Tn
with Ti − Ti−1 ≡ δ.
Fix a rate κ > 0. At time Ti the holder of the cap receives
δ(F (Ti−1 , Ti ) − κ)+ .
Let t ≤ T0 . We write
Cpl(t; Ti−1 , Ti ), i = 1, . . . , n
for the time t price of the ith caplet, and
Xn
Cp(t) = Cpl(t; Ti−1 , Ti )
i=1
for the time t price of the cap.
28 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Floors

At time Ti the holder of the floor receives

δ(κ − F (Ti−1 , Ti ))+ .

Let t ≤ T0 . We write

Fll(t; Ti−1 , Ti ), i = 1, . . . , n

for the time t price of the ith floorlet, and

n
X
Fl(t) = Fll(t; Ti−1 , Ti )
i=1

for the time t price of the floor.

29 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Swaps
Fix a rate K and a nominal N. The cash flow of a payer swap at
Ti is
(F (Ti−1 , Ti ) − K )δN.
The total value Πp (t) of the payer swap at time t ≤ T0 is
 Xn 
Πp (t) = N P(t, T0 ) − P(t, Tn ) − K δ P(t, Ti ) .
i=1
The value of a receiver swap at t ≤ T0 is
Πr (t) = −Πp (t).
The swap rate Rswap (t) is the fixed rate K which gives
Πp (t) = Πr (t) = 0. Hence
P(t, T0 ) − P(t, Tn )
Rswap (t) = , t ∈ [0, T0 ].
δ ni=1 P(t, Ti )
P
30 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Swaptions

A payer (receiver) swaption is an option to enter a payer (receiver)

swap at T0 . The payoff of a payer swaption at T0 is
n
X
Nδ(Rswap (T0 ) − K )+ P(T0 , Ti ),
i=1

and of a receiver swaption

n
X
+
Nδ(K − Rswap (T0 )) P(T0 , Ti ).
i=1

31 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Note that it is very cumbersome to write models which are

analytically tractable for both swaptions and caps/floors.
I Black’s model is a lognormal model for one bond price with
respect to a particular numeraire. If we change the numeraire
the lognormal property gets lost.
I The change of numeraire between swap and forward measures
is a rational function, which usually destroys analytic
tractability properties of given models.

32 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Spot measure

The spot measure is defined as a martigale measure for the ZCB

prices discounted by their own bank account process

P(t, T )
B(t)

for T ≥ 0. This leads to the following fundamental formula of

interest rate theory
Z T
P(t, T ) = E (exp(− Rs ds))|Ft )
t

for 0 ≤ t ≤ T with respect to the spot measure.

33 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Short rate models

We can assume several dynamics with respect to the spot measure:

I Vasiček model: dRt = (βRt + b)dt + 2αdWt .

p
I CIR model: dRt = (βRt + b)dt + 2α (Rt )dWt .
In the following two slides typical Vasiček and CIR trajectories are
simulated.

34 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Vasicek−trajectories

0.10
0.08
0.06
X

0.04
0.02
0.00

0 200 400 600 800 1000

35 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

CIR−trajectories

0.10
0.08
0.06
X

0.04
0.02
0.00

0 200 400 600 800 1000

36 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Forward measures

∗
For T ∗ > 0 define the T ∗ -forward measure P T such that for any
T > 0 the discounted bond price process

P(t, T )
, t ∈ [0, T ]
P(t, T ∗ )
∗
is a P T -martingale.

37 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Forward measures

For any T < T ∗ the simple forward rate

 
1 P(t, T )
F (t; T , T ∗ ) = ∗ − 1
T − T P(t, T ∗ )
∗
is a PT -martingale.

38 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

For any time T derivative X ∈ FT we have that the fair value via
“martingale pricing” is given through

P(t, T )ET [X |Ft ].

The fair price of the ith caplet is therefore given by

Cpl(t; Ti−1 , Ti ) = δP(t, Ti )ETi [(F (Ti−1 , Ti ) − κ)+ |Ft ].

By the martingale property we obtain therefore

ETi [F (Ti−1 , Ti )|Ft ] = F (t; Ti−1 , Ti ),

what was proved by trading arguments before.

39 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Swap measures

For T < T1 < . . . < Tn define the swap measure PT ;T1 ,...,Tn by the
property that for any S > 0 the process

P(t, S)
Pn , t ∈ [0, S ∧ T ]
i=1 P(t, Ti )

is a PT ;T1 ,...,Tn -martingale.

40 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

Swap measure

In particular the swap rate

P(t, T0 ) − P(t, Tn )
Rswap (t) = , t ∈ [0, T0 ]
δ ni=1 P(t, Ti )
P

is a PT0 ;T1 ,...,Tn -martingale.

41 / 107
Lecture Notes: Interest Rate Theory
Basics on Interest Rate Modeling

For any X ∈ FT we have that the fair price is given by

n
X 
P(t, Ti ) EtT0 ;T1 ,...,Tn [X ].
i=1

42 / 107
Lecture Notes: Interest Rate Theory
Black formulas

I Black formulas are applications of the lognormal

Black-Scholes theory to model LIBOR rates or swap rates.
I Black formulas are not constructed from one model of
lognormal type for all modeled quantities (LIBOR rates, swap
rates, forward rates, etc).
I Generically only of the following quantities is lognormal with
respect to one particular measure: one LIBOR rate or one
swap rate, each for a certain tenor.

43 / 107
Lecture Notes: Interest Rate Theory
Black formulas

Black formula

Let X ∼ N(µ, σ 2 ) and K ∈ R. Then we have

log K − (µ + σ 2 )
   
X + µ+ σ2
2 log K − µ
E[(e − K ) ] = e Φ − − KΦ − ,
σ σ
log K − (µ + σ 2 )
   
log K − µ σ2
E[(K − e X )+ ] = K Φ − e µ+ 2 Φ .
σ σ

44 / 107
Lecture Notes: Interest Rate Theory
Black formulas

Black formula for caps and floors

Let t ≤ T0 . From our previous results we know that

Cpl(t; Ti−1 , Ti ) = δP(t, Ti )ET +

t [(F (Ti−1 , Ti ) − κ) ],
i

Fll(t; Ti−1 , Ti ) = δP(t, Ti )ET +

t [(κ − F (Ti−1 , Ti )) ],
i

and that F (t; Ti−1 , Ti ) is an P Ti -martingale.

45 / 107
Lecture Notes: Interest Rate Theory
Black formulas

Black formula for caps and floors

We assume that under P Ti the forward rate F (t; Ti−1 , Ti ) is an

exponential Brownian motion

F (t; Ti−1 , Ti ) = F (s; Ti−1 , Ti )

1 t
 Z Z t 
2 Ti
exp − λ(u, Ti−1 ) du + λ(u, Ti−1 )dWu
2 s s

for s ≤ t ≤ Ti−1 , with a function λ(u, Ti−1 ).

46 / 107
Lecture Notes: Interest Rate Theory
Black formulas

We define the volatility σ 2 (t) as

Z Ti−1
2 1
σ (t) := λ(s, Ti−1 )2 ds.
Ti−1 − t t

The P Ti -distribution of log F (Ti−1 , Ti ) conditional on Ft is

N(µ, σ 2 ) with
σ 2 (t)
µ = log F (t; Ti−1 , Ti ) − (Ti−1 − t),
2
σ 2 = σ 2 (t)(Ti−1 − t).
In particular
σ2
µ+ = log F (t; Ti−1 , Ti ),
2
σ 2 (t)
µ + σ 2 = log F (t; Ti−1 , Ti ) + (Ti−1 − t).
2
47 / 107
Lecture Notes: Interest Rate Theory
Black formulas

We have

Cpl(t; Ti−1 , Ti ) = δP(t, Ti )(F (t; Ti−1 , Ti )Φ(d1 (i; t)) − κΦ(d2 (i; t))),
Fll(t; Ti−1 , Ti ) = δP(t, Ti )(κΦ(−d2 (i; t)) − F (t; Ti−1 , Ti )Φ(−d1 (i; t))),

where
F (t;Ti−1 ,Ti )

log κ ± 1 σ(t)2 (Ti−1 − t)
d1,2 (i; t) = p 2 .
σ(t) Ti−1 − t

48 / 107
Lecture Notes: Interest Rate Theory
Black formulas

Proof

We just note that

Cpl(t; Ti−1 , Ti ) = δP(t, Ti )E[(e X − κ)+ ],

Fll(t; Ti−1 , Ti ) = δP(t, Ti )E[(κ − e X )+ ]

with X ∼ N(µ, σ 2 ).

49 / 107
Lecture Notes: Interest Rate Theory
Black formulas

Black’s formula for swaptions

Let t ≤ T0 . From our previous results we know that

n
T0 ;T1 ,...,Tn
X
Swptp (t) = Nδ P(t, Ti )EPt [(Rswap (T0 ) − K )+ ],
i=1
n
T0 ;T1 ,...,Tn
X
Swptr (t) = Nδ P(t, Ti )EPt [(K − Rswap (T0 ))+ ],
i=1

and that Rswap is an PT0 ;T1 ,...,Tn -martingale.

50 / 107
Lecture Notes: Interest Rate Theory
Black formulas

Black’s formula for swaptions

We assume that under PT0 ;T1 ,...,Tn the swap rate Rswap is an
exponential Brownian motion

1 t
 Z Z t 
2
Rswap (t) = Rswap (s) exp − λ(u) ds + λ(u)dWu
2 s s

for s ≤ t ≤ T0 , with a function λ(u).

51 / 107
Lecture Notes: Interest Rate Theory
Black formulas

We define the implied volatility σ 2 (t) as

Z T0
2 1
σ (t) := λ(s)2 ds.
T0 − t t
The PT0 ;T1 ,...,Tn -distribution of log Rswap (T0 ) conditional on Ft is
N(µ, σ 2 ) with
σ 2 (t)
µ = log Rswap (t) − (T0 − t),
2
σ 2 = σ 2 (t)(T0 − t).
In particular
σ2
µ+ = log Rswap (t),
2
σ 2 (t)
µ + σ 2 = log Rswap (t) + (T0 − t).
2
52 / 107
Lecture Notes: Interest Rate Theory
Black formulas

We have
n

X
Swptp (t) = Nδ Rswap (t)Φ(d1 (t)) − K Φ(d2 (t)) P(t, Ti ),
i=1
n

X
Swptr (t) = Nδ K Φ(−d2 (t)) − Rswap (t)Φ(−d1 (t)) P(t, Ti ),
i=1

with
Rswap (t)

log K ± 1 σ(t)2 (T0 − t)
d1,2 (t) = √2 .
σ(t) T0 − t

53 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Market Models

I Let 0 = T0 < . . . < TN = T be a discrete tenor structure of

maturity dates.
I We shall assume that Tk+1 − Tk ≡ δ.
I Our goal is to model the LIBOR market
 
1 P(t, Tk )
L(t, Tk , Tk+1 ) = −1 .
δ P(t, Tk+1 )

54 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Three Axioms

The following axioms are motivated by economic theory, arbitrage

pricing theory and applications:
I Axiom 1: Positivity of the LIBOR rates

L(t, Tk , Tk+1 ) ≥ 0.

I Axiom 2: Martingale property under the corresponding

forward measure

L(t, Tk , Tk+1 ) ∈ M(PTk+1 ).

I Axiom 3: Analytical tractability.

55 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Known Approaches

Here are some known approaches:

I Let L(t, Tk , Tk+1 ) be an exponential Brownian motion. Then
analytical tractability not completely satisfied (“Freezing the
drift”).
P(t,Tk )
I Let P(t,T k+1 )
be an exponential Brownian motion. Then
positivity of the LIBOR rates is not satisfied.
I We will study affine LIBOR models.

56 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Affine Processes

Let X = (Xt )0≤t≤T be a conservative, time-homogeneous,

stochastically continuous process taking values in D = Rd≥0 .
Setting
IT := {u ∈ Rd : E[e hu,XT i ] < ∞},
we assume that:
I 0 ∈ IT◦ ;
I there exist functions φ : [0, T ] × IT → R and
ψ : [0, T ] × IT → Rd such that

E[e hu,XT i ] = exp(φt (u) + hψt (u), X0 i)

for all 0 ≤ t ≤ T and u ∈ IT .

57 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Some Properties of Affine Processes

I For all 0 ≤ s ≤ t ≤ T and u ∈ IT we have

E[e hu,XT i | Fs ] = exp(φt−s (u) + hψt−s (u), X0 i).

I Semiflow property: For all 0 ≤ t + s ≤ T and u ∈ IT we have

φt+s (u) = φt (u) + φs (ψt (u)),

ψt+s (u) = ψs (ψt (u)).

I Order-preserving: For (t, u), (t, v ) ∈ [0, T ] × IT with u ≤ v

we have

φt (u) ≤ φt (v ) and ψt (u) ≤ ψt (v ).

58 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Constructing Martingales ≥ 1

For u ∈ IT we define M u = (Mtu )0≤t≤T as

Mtu := exp(φT −t (u) + hψT −t (u), Xt i).

Then the following properties are valid:

I M u is a martingale.
I For u ∈ Rd≥0 and X0 ∈ Rd≥0 we have Mtu ≥ 1.

59 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Constructing the Affine LIBOR Model

I We fix u1 > . . . > uN from IT ∩ Rd≥0 and set

P(t, Tk )
= Mtuk , k = 1, . . . , N.
P(t, TN )
I Obviously, we set

P(0, TN )
uN = 0 ⇔ = 1.
P(0, TN )

60 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Positivity
Then we have
P(t, Tk )
= exp(Ak + hBk , Xt i),
P(t, Tk+1 )
where we have defined
Ak := AT −t (uk , uk+1 ) := φT −t (uk ) − φT −t (uk+1 ),
Bk := BT −t (uk , uk+1 ) := ψT −t (uk ) − ψT −t (uk+1 ).
Note that Ak , Bk ≥ 0 by the order-preserving property of φt (·) and
ψt (·). Thus, the LIBOR rates are positive:
 
1 P(t, Tk ) 1
L(t, Tk , Tk+1 ) = −1 = (exp(Ak + hBk , Xt i) −1) ≥ 0.
δ P(t, Tk+1 ) δ | {z }
≥1

61 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Martingale Property
I We define the equivalent probability measures

dPTk Mtuk
:= uk , t ∈ [0, Tk ].
dPTN Ft M0

I By Bayes’ rule these are forward measures:

u
u P(t, Tj ) P(t, Tj ) M j
Mt j = ∈ M(PTN ) ⇒ = tuk ∈ M(PTk ).
P(t, TN ) P(t, Tk ) Mt
I We deduce the martingale property
 
1 P(t, Tk )
L(t, Tk , Tk+1 ) = − 1 ∈ M(PTk+1 ).
δ P(t, Tk+1 )

62 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Analytical Tractability

I X is a time-homogeneous affine process under any forward

measure:

ETk [e hv ,Xt i ] = exp(φkt (v ) + hψtk (v ), X0 i).

I The functions φk and ψ k are given by

φkt (v ) := φt (ψT −t (uk ) + v ) − φt (ψT −t (uk )),

ψtk (v ) := ψt (ψT −t (uk ) + v ) − ψt (ψT −t (uk )).

63 / 107
Lecture Notes: Interest Rate Theory
Affine LIBOR Models

Option Pricing

The price of a caplet with reset date Tk , settlement date Tk+1 and
strike rate K is given by
h + i
Cpl(Tk , K ) = P(0, Tk+1 )ETk+1 e Ak +hBk ,XTk i − K ,

where K = 1 + δK . By applying Fourier methods, we obtain

ETk+1 [e (R−iv )(Ak +hBk ,XTk i) ]K1+iv −R

Z
P(0, Tk+1 )
Cpl(Tk , K ) = dv ,
2π R (iv − R)(1 + iv − R)

where R ∈ (1, ∞).

64 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Definition of a Markov Process

I A family of adapted Rd -valued stochastic processes
(Xtx )t≥0,x∈S is called time-homogenous Markov process with
state space S if for all s ≤ t and B ∈ B(Rd ) we have
y
P(Xtx ∈ B | Fs ) = P(Xt−s ∈ B)|y =Xs .
In particular Markov processes with state space S take values
in S almost surely.
I We can define the associated Markov kernels
µs,t : Rd × B(Rd ) → [0, 1] with
y
µs,t (y , B) = P(Xt−s ∈ B).
I They satisfy the Chapman-Kolmogorov equation
Z
µs,u (x, B) = µt,u (y , B)µs,t (x, dy ).
Rd
for s ≤ u and Borel sets B. 65 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Feller Processes

I For t ≥ 0 and f ∈ C0 (Rd ) we define

Z
Tt f (x) := f (y )µt (x, dy ), x ∈ Rd .
Rd

I X is a Feller Process if (Tt )t≥0 is a C0 -semigroup of

contractions on C0 (Rd ).
I We define the infinitesimal generator
Tt f − f
Af := lim , f ∈ D(A),
t→0 t
which conincides with the concept of infinitesimal generator
from functional analysis.
66 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Stochastic Differential Equations as Markov processes

I Let b : Rd → Rd and σ : Rd → Rd×m . Consider the SDE

dXtx = b(Xt )dt + σ(Xt )dWt , X0x = x

I We assume that the solution exists for all times and any initial
value in some state space S as a Feller-Markov process.
I Set a := σσ > . We have C02 (Rd ) ⊂ D(A) and
d
1 X ∂2
Af (x) = aij (x) f (x)+hb(x), ∇f (x)i, f ∈ C02 (Rd ).
2 ∂xi ∂xj
i,j=1

I Example: For a Brownian motion W we have A = 12 ∆.

67 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Kolmogorov Backward Equation

We assume there exist transition densities p(t, x, y ) such that
Z
P(Xt ∈ B | Xs = x) = p(t − s, x, y )dy .
B

I Recall that the generator is given by

d d
1 X ∂2 X ∂
Af (x) = aij (x) f (x) + bi (x) f (x).
2 ∂xi ∂xj ∂xi
i,j=1 i=1

I Kolmogorov backward equation: For fixed y ∈ Rd we have

∂
p(t, x, y ) = A p(t, x, y ),
∂t
i.e. the equation acts on the backward (initial) variables. It
also holds in the sense of distribution for the expectation
functional (t, x) 7→ E (f (Xtx )). 68 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Kolmogorov Forward Equation

I The adjoint operator is given by

d d
1 X ∂2 X ∂
A∗ f (y ) = (aij (y )f (y )) − (bi (y )f (y )).
2 ∂yi ∂yj ∂yi
i,j=1 i=1

I Kolmogorov forward equation: For fixed x ∈ Rd we have

∂
p(t, x, y ) = A∗ p(t, x, y ),
∂t
i.e. the equation acts on the forward variables. It also holds in
the sense of distributions for the Markov kernels p(., x, .) for
any initial value x ∈ S.

69 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Example: Brownian Motion

I The Brownian motion W has the transition densities
1 (y −x)2
p(t, x, y ) = √ e − 2t .
2πt
I The infinitesimal generator is given by the Laplace operator
A = 12 ∆.
I Kolmogorov backward equation: For fixed y ∈ Rd we have
d 1 ∂2
p(t, x, y ) = p(t, x, y ).
dt 2 ∂x 2
I Kolmogorov forward equation: For fixed x ∈ Rd we have
d 1 ∂2
p(t, x, y ) = p(t, x, y ).
dt 2 ∂y 2
70 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Example: The SABR Model

I The SABR model for β = 0, α = 1, ρ = 0 is given through

dX1 (t) = X2 (t)dW1 (t),

dX2 (t) = X2 (t)dW2 (t).

I Its infinitesimal generator equals therefore

x22 ∂ 2 ∂2
 
A= + .
2 ∂x12 ∂x22

I Kolmogorov backward equation: For fixed y ∈ R2 we have

x22 ∂ 2 ∂2
 
d
p(t, x, y ) = + p(t, x, y ).
dt 2 ∂x12 ∂x22

71 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Example: The SABR model

I Kolmogorov forward equation: For fixed x ∈ R2 we have

 2
∂ 2 y22

d ∂
p(t, x, y ) = + p(t, x, y ).
dt ∂y12 ∂y22 2
I Notice that in general the foward and backward equation are
different.

72 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Construction of finite factor models

By the choice of a finite dimensional Markov process (X 1 , . . . , X n )

and the choice of an expression

Rt = H(Xt )

for the short rate, one can construct – due to the Markov property
– consistent finite factor model
Z T Z T
E (exp(− H(Xs )ds) = P(t, T ) = exp(− G (t, r , Xt )dr ),
t t

for all 0 ≤ t ≤ T , G satisfies a certain P(I)DE.

73 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

Vasiček’s model is due to is Gaussian nature relatively simple: We

apply the parametrization x = T − t for 0 ≤ t ≤ T :
1
Λ(x) = (1 − exp(−βx))
β
ρ2 ρ2
A0 (t, x) = r ∗ (x + t) + Λ(x + t)2 − Λ(x)2 −
2 2
Z t
− (Λ0 (x))2 r ∗ (0) − Λ0 (x) e −β(t−s) b(s)ds
0
d ρ2
b(t) = r ∗ (t) + βr ∗ (t) + (1 − exp(−2βt))
dt 2β
A1 (t, x) = Λ0 (x)
dRt = (b(t) − βRt )dt + ρdWt
for real constants β and ρ and an ”arbitrary” initial value r ∗ . This
solves the HJM equation with volatility σ(r , x) = ρ exp(−βx) and
initial value r ∗ .
74 / 107
Lecture Notes: Interest Rate Theory
Markov Processes

The CIR analysis is more involved:

A0 (t, x) = g (t, x) − c(t)Λ0 (x)

g (t, x) = r ∗ (t + x)+
Z t
+ρ2 g (t − s, 0)(ΛΛ0 )(x + t − s)ds
0
d
c(t) = g (t, 0), b(t) = c(t) + βc(t)
dt
A1 (t, x) = Λ0 (x)
1
dRt = (b(t) − βRt )dt + ρRt2 dWt

for real constants β and ρ and an ”arbitrary” initial value r ∗ . This

1
solves the HJM equation with volatility σ(r , x) = ρ(ev0 (r )) 2 Λ(x)
and initial value r ∗ .
75 / 107
Lecture Notes: Interest Rate Theory
The SABR model

I The SABR model combines an explicit expression for implied

volatility with attractive dynamic properties for implied
volatilities.
I In contrast to affine models stochastic volatility is a lognormal
random variable.
I it is a beautiful piece of mathematics.
I all important details can be found in [2].

76 / 107
Lecture Notes: Interest Rate Theory
The SABR model

We consider a model for forward prices F and their stochastic

volatility Σ

dFt = Σt C (Ft )dWt (1)

dΣt = v Σt dZt (2)

with two correlated Brownian motions W , Z with hW , Z it = ρt.

We assume that C is smooth of 0 and that
Z x
du
<∞
0 C (u)

for x > 0. For instance C (x) = x β for 0 ≤ β < 1. Notice that

usually the SABR price F is symmetrically extended to the whole
real line and some (inner) boundary conditions of Dirichlet,
Neuman or mixed type are considered (see the discussion in [2]).
77 / 107
Lecture Notes: Interest Rate Theory
The SABR model

We aim to calculate the transition distribution at time

GF ,Σ (τ, f , σ)dF dΣ, when the process starts from initial value
(f , σ) and evolves for some time τ > 0. This is done by relating
the general SABR model via an invertible map to

dXt = Yt dWt , dYt = Yt dZt ,

with decorrelated Brownian motions W and Z . The latter

stochastic differential equation is related to the Poincare halfplane
and its hyperbolic geometry. We shall refer to it as Brownian
motion on the Poincare halfplane.

78 / 107
Lecture Notes: Interest Rate Theory
The SABR model

Poincare halfplane

Consider the set of points H2 := R × R>0 and the Riemannian

metric with matrix y12 id at (x, y ) ∈ H2 . Then one can calculate
the geodesic distance on H2 , i.e.˜the length of the shortest path
connecting to points, via

(x − X )2 + (y − Y )2
cosh(d(x, y , X , Y )) = 1 + .
2yY
Furthermore one can calculate in terms of the geodesic distance d
the heat kernel on H2 . A derivation is shown in [2].

79 / 107
Lecture Notes: Interest Rate Theory
The SABR model

Applying the invertible map φ

Z f
1 du
(f , σ) 7→ p ( − ρsigma), σ)
1 − ρ2 0 C (u)

to the equation

dFt = Σt C (Ft )dWt , dΣt = Σt dZt

leads to the Poincare halfplane’s Brownian motion perturbed by a

drift term. From the point of view of the SABR model one has to
add a drift such that the φ−1 transformation of it is precisely the
Brownian motion of the Poincare halfplane.

80 / 107
Lecture Notes: Interest Rate Theory
The SABR model

Regular perturbation techniques

Since we are interested in the original SABR model we have to
calculate the influence of the drift term appearing when
transforming from the Poincare halfplane to the SABR model, this
is done by regular perturbation techniques:
I consider two linear operators A, B, where B is considered
small in comparison to A.
I consider the variation of constants formula ansatz for
d
exp(t(A + B)) = A exp(t(A + B)) + B exp(t(A + B))
dt
= A exp(t(A + B)) + f (t).
I this leads to
Z t
exp(t(A+B)) = exp(tA)+ exp((t−s)AB exp(s(A+B))ds,
0
and by iteration to. 81 / 107
Lecture Notes: Interest Rate Theory
The SABR model

∞
X Z
k
exp(A + B) =  exp(s1 adA )B exp(s2 adA )B × · · ·
k=0 0≤s1 ...≤sk ≤1

× exp(sk adA )B ds1 · · · dsk ,

where the adjoint action ad is defined via

exp(s adA )B = exp(sA)B exp(−sA).

82 / 107
Lecture Notes: Interest Rate Theory
The SABR model

Local volatility

A good approximation for implied volatility is given by local

volatility, which can be calculated in many models by the following
formula
∂
C (T , K )
σ(t, K )2 = ∂T∂2
,
∂K 2
C (T , K )
being the answer to the question which time-dependent volatility
function σ to choose such that dSt = σ(t, St )dWt mimicks the
given prices C (T , K ), i.e. for T , K ≥ 0 it holds

C (T , K ) = E ((ST − K )+ ).

83 / 107
Lecture Notes: Interest Rate Theory
The SABR model

The reasoning behind Dupire’s formula for local volatility is that

the transition distribution of a local volatility model satisfies
Kolmogorov’s forward equation in the forward variables

∂2 ∂
σ(t, S)2 p(T , S, s) = p(T , S, s).
∂S 2 ∂T
On the other hand it is well-known by Breeden-Litzenberger that

∂2
p(T , K , s) = C (T , K ),
∂K 2
which leads after twofold integration of Kolmogorov’s forward
equation by parts to Dupire’s formula.

84 / 107
Lecture Notes: Interest Rate Theory
The SABR model

In terms of the transition function of the SABR model Dupire’s

formula reads as
C (K )2 Σ2 G (T , K , Σ, f , σ)dΣ
R
2
σ(T , K ) = R ,
G (T , K , Σ, f , σ)dΣ

which can be evaluated by Laplace’s principle as shown in [2] since

one has at hand a sufficiently well-known expression for G .

85 / 107
Lecture Notes: Interest Rate Theory
HJM-models

Lévy driven HJM models

Let L be a d-dimensional Lévy process with Lévy exponent κ, i.e.

E (exp(hu, Lt i) = exp(κ(u)t)

for u ∈ U an open strip in Cd containing iRd , where κ is always

defined. Then it is well-known that
Z t Z t
exp(− κ(αs )ds + hαs , dLs i)
0 0

is a local martingale for predictable strategies alpha such that both

integrals are well-defined. Notice that the strategy α is Rd -valued
and that κ has to be defined on αs .

86 / 107
Lecture Notes: Interest Rate Theory
HJM-models

3 trajectories of a stable process

6
5
4
3
2
1
0
−1

0 2 4 6 8 10
87 / 107
Lecture Notes: Interest Rate Theory
HJM-models

3 trajectories of VG process

8
6
4
2
0

0 2 4 6 8 10
88 / 107
Lecture Notes: Interest Rate Theory
HJM-models

3 trajectories of NIG process

1.5
1.0
0.5
0.0

0 2 4 6 8 10
89 / 107
Lecture Notes: Interest Rate Theory
HJM-models

We can formulate a small generalization of the previous result by

considering a parameter-dependence in the strategies αS . There is
one Fubini argument necessary to prove the result: we assume
continuous dependence of αS on S, then
Z tZ T Z S Z tZ T
d
Ntu = exp( κ( U
αs dU)ds + hαsS dS, dLs i)
0 u dS u 0 u

is a local martingale. Notice here that Ntt is not a local martingale,

but Z t
(dNtu )u=t
0
is one, since – loosely speaking – it is the sum of local martingale
increments.

90 / 107
Lecture Notes: Interest Rate Theory
HJM-models

The general HJM-drift condition for Lévy-driven term

structures
If
Z t Z T Z t
d
f (t, T ) = f (0, T ) + κ(− αtT )dt + hαt , dLt i
0 dT t 0

defines a stochastic process of forward rates, where continuous

dependence in T of all quantities is assumed, such that
Z t
M(t, T ) = P(t, T ) exp(− f (s, s)ds)
0
Z T Z t
= exp( f (t, S) dS − f (s, s)ds)
t 0

is a local martingale, since

91 / 107
Lecture Notes: Interest Rate Theory
HJM-models

its differential equals

Z t
−f (t, t)M(t, T )dt+f (t, t)M(t, T )dt+exp(− f (s, s)ds)(dNtu )|u=t ,
0

where the first two terms cancel and the third one is the increment
of a local martingale as was shown before.

92 / 107
Lecture Notes: Interest Rate Theory
HJM-models

HJM-drift condition in case of driving Brownian motion

When the HJM equation is driven by Brownian motions, we speak

2
of an Itô process model, in particular κ(u) = ||u||
2 .
If we assume an Itô process model with the HJM equation reads as
d
X Z x
i
df (t, T ) = α (t, T ) αi (t, y )dy dt+
i=1 0

Xd
+ αi (t, T )dBti ,
i=1

where the volatilities αi (t, T )0≤t≤T are predictable stochastic

integrands.

93 / 107
Lecture Notes: Interest Rate Theory
HJM-models

Musiela parameterization

The forward rates (f (t, T ))0≤t≤T are best parametrized through

r (t, x) := f (t, t + x)

for t, x ≥ 0 (Musiela parametrization). This allows to consider

spaces of forward rate curves, otherwise the domain of definition of
the forward rate changes along running time as it equals [t, ∞[.

94 / 107
Lecture Notes: Interest Rate Theory
HJM-models

Forward Rates as states

This equation is best analysed as stochastic evolution on a Hilbert
space H of forward curves making it thereon into a Markov process
ei (t, .) = σ i (rt ), σ i : H → H
σ
for some initial value r0 ∈ H. We require:
I H is a separable Hilbert space of continuous functions.
I point evalutations are continuous with respect to the topology
of a Hilbert space.
I The shift semigroup (St r )(x) = r (t + x) is a strongly
d
continuous semigroup on H with generatorR dx .
x
I The map h 7→ S(h) with S(h)(x) := h(x) h(y )dy satisfies
0

||S(h)|| ≤ K ||h||2
for all h ∈ H with S(h) ∈ H.
95 / 107
Lecture Notes: Interest Rate Theory
HJM-models

An example

Let w : R≥0 → [1, ∞[ be a non-decreasing C 1 -function with

1
1 ∈ L1 (R≥0 ),
w 3

then we define
Z
||h||w := |h(0)| + |h0 (x)|w (x)dx
R≥0

for all h ∈ L1loc with h0 ∈ L1loc (where h0 denotes the weak

derivative). We define Hw to be the space of all functions h ∈ L1loc
with h0 ∈ L1loc such that ||h||w < ∞.

96 / 107
Lecture Notes: Interest Rate Theory
HJM-models

Finite Factor models

Given an initial forward rate T 7→ f (0, T ) or T 7→ P(0, T ),

respectively. A finite factor model at initial value r ∗ is a mapping

G : {0 ≤ t ≤ T } × Rn ⊂ R2≥0 × Rn → R

together with an Markov process (Xt )t≥0 such that

f (t, T ) = G (t, T , Xt1 , ..., Xtn )

for 0 ≤ t ≤ T and T ≥ 0 is an arbitrage-free evolution of forward

rates. The process (Xt )t≥0 is called factor process, its dimension n
is the dimension of the factor model.

97 / 107
Lecture Notes: Interest Rate Theory
HJM-models

In many cases the map G is chosen to have a particularly simple

structure
n
X
G (t, T , z 1 , ..., z n ) = A0 (t, T ) + Ai (t, T )z i .
i=1

In these cases we speak of affine term structure models, the factor

processes are also affine processes. Remark that G must reproduce
the initial value

G (0, T , z01 , ..., z0n ) =: r ∗ (T )

for T ≥ 0. The famous short rate models appear as 1- or

2-dimensional cases (n = 1, 2 – time is counted as additional
factor).

98 / 107
Lecture Notes: Interest Rate Theory
HJM-models

The consistency problem

It is an interesting and far reaching question if – for a given

function G (t, T , z) – there is a Markov process such that they
constitute a finite factor model together.

99 / 107
Lecture Notes: Interest Rate Theory
HJM-models

Svensson family

An interesting example for a map G is given by the Svensson family

G (t, T , z1 , . . . , z6 ) = z1 + z2 exp(−z3 (T − t))+

+ (z4 + z5 (T − t)) exp(−z6 (T − t)),

since it is often applied by national banks. The wishful thought to

find an underlying Itô-Markov process such that G is consistent
with an arbitrage free evolution of interest rates is realized by a
one factor Gaussian process.

100 / 107
Lecture Notes: Interest Rate Theory
References

[1] Damir Filipovic.

Term structure models: A Graduate Course.
see http://sfi.epfl.ch/op/edit/page-12795.html
[2] Patrick Hagan, Andrew Lesniewski, and Diana Woodward.
Probability Distribution in the SABR Model of Stochastic
Volatility.
see http://lesniewski.us/working.html
[3] Martin Keller-Ressel, Antonis Papapantoleon, Josef
Teichmann.
A new approach to LIBOR modeling.
see http://www.math.ethz.ch/~jteichma/index.php?
content=publications

101 / 107
Lecture Notes: Interest Rate Theory
Catalogue of possible questions for the oral exam

[4] Damir Filipovic, Stefan Tappe.

Existence of Lévy term structure models.
see
http://www.math.ethz.ch/~tappes/publications.php

102 / 107
Lecture Notes: Interest Rate Theory
Catalogue of possible questions for the oral exam

Catalogue of possible questions for the oral exam

I What are ZCBs, yield curves, forward curves, short rates,

caplets, floorlets, swaps, swap rates, roll-over-portfolios?
I What is a LIBOR rate (simple forward rate) on nominal one
received at terminal date and what is its fair value before?
Some relations of cap, floors, swaptions like in exercise 2 of
Sheet 2.
I Black’s formula for caps and floors – derivation and
assumptions?
I Black’s formula for swaptions – derivation and assumptions?

103 / 107
Lecture Notes: Interest Rate Theory
Catalogue of possible questions for the oral exam

Catalogue of possible questions for the oral exam

I Change of numeraire theorem: exercise 1,2 of Sheet 7.

I What is the LIBOR market model (Exercise 3 of Sheet 7)?
I What is the foward measure model (Exercise 2 of Sheet 8).
I What is an affine LIBOR model and what are its main
characteristics (Axiom 1 – 3 can be satisfied)?
I What is the Fourier method of derivative pricing (Exercise 1 of
Sheet 9)?

104 / 107
Lecture Notes: Interest Rate Theory
Catalogue of possible questions for the oral exam

Catalogue of possible questions for the oral exam

I Lévy processes and their cumulant generating function

(Exercise 1 of Sheet 11).
I What is the HJM-drift condition for Brownian motion and
which Hilbert spaces of forward rates can be considered?
I Derivation of the HJM-drift condition for Lévy driven interest
rate models.
I Why are models for the whole term structure attractive?
What lie their difficulties?

105 / 107
Lecture Notes: Interest Rate Theory
Catalogue of possible questions for the oral exam

Catalogue of possible questions for the oral exam

I What is the general SABR model and how is it related to the

Poincare halfplane?
I What is the geodesic distance in the Poincare halfplane
(definition) and how can we calculate it (eikonal equation),
Exercise 5 on Sheet 9? What is the natural stochastic process
on the Poincare halfplane – can we calculate its heat kernel?
I What is local volatility and how can we calculate it (Dupire’s
formula)?
I Implied volatility in interest rate markets: which tractability
do we need from models and what do market models, forward
measure models, affine models, the SABR model, HJM
models provide?

106 / 107
Lecture Notes: Interest Rate Theory
Catalogue of possible questions for the oral exam

Catalogue of possible questions for the oral exam

I Short rate models: the Vasicek model (Exercise 1 of Sheet 4).

I Short rate models: the CIR model (Exericse 2 of Sheet 4, not
every detail).
I Short rate models: Hull-White extension of the Vasicek model
(Exercise 2 of Sheet 6).
I Is short rate easy to model from an econometric point of view?

107 / 107