Materials for the Course

Interest rates modeling

Paweł Kliber
Derivative instruments in the
world
Notional amounts outstanding

Foreign exchange contracts

3% 4%
0%
1% 11%
Interest rate contracts

Equity-linked contracts

Commodity contracts

Credit default swaps

Unallocated

81%

Average 2013-2015, Source: Bank for International Settlements

Average 2013-2015, Source: Bank of International Settlements
Interest rate derivative
instuments
 Main instruments
• Standarized instruments (traded on
exchange)
– Bond futures
• Unstandarized instruments (OTC):
– FRA
– Swap
– Cap
– Floor
– Swaption (option on swap instrument)
 Bond futures
• Obligations for the contract holder to buy or
sell a bond on a specified date at a
predetermined price.
• The bond is notional (does not exist in reality)
– e.g. in Poland 10-years T-bond with 5%
coupon rate.
• Seller has a set of real bonds (specified by
some criteria) to choose from.
• Delivery price is corrected by Conversion
Factor (CF, calculated for specific bond)
 FRA
• Instrument used to predetermine interest rate
for a period in the future
• Party A: pays party B interest with
predetermined rate (FRA rate)
• Party B: pays party A interest according to
rate that will be known in the future
• E.g. FRA 3x6 5% – the period starts in 3
months and ends in 6 months. A pays 5%
and B pays according to market rate that will
be in 3 months
 Example
• Firm X plans an investment in 6 month from now.
After the investment starts, it will have to maintain
10,000 euro for the next six month. The firm risks
that in 6 month the interest rates will change
• It can:
– Take a credit of 10,000 in 6 month (with interest rate
that is unknown now)
– Take position in FRA 6X12 contract, with current
market FRA rate equal to 5.5% and in 6 month take a
credit of 10,000 with current market 6-month interest
rate (WIBOR6M) plus spread of 0.5%.
 Example
WIBOR 6M Interest for the FRA Net
(market rate) loan
7% -(7+0.5)% (7-5.5)% -6%
6% -(6+0.5)% (6-5.5)% -6%
4% -(4+0.5)% (4-5.5)% -6%
3% -(3+0.5)% (3-5.5)% -6%
 Interest Rate Swap (IRS)
• Exchange of two cash flows (or series of FRAs)
• One party (payer) pays according to predetermined rate K and
second party (receiver) pays according to actual market rates
• Initial value is zero: the rate K (swap rate is negotiated
 Swaption
• Option that gives right (but not obligation)
to become one part in IRS contract
• Payer swaption – owner can become
payer in the contract
• Receiver swaption – owner can become
receiver
 Cap and floor
• Options on interest rate – a series of
caplets or floorlets
• Caplet payoff: N  ( L(Ti 1 , Ti )  K ) 

• Floorlet payoff: N ( K  L(Ti 1 , Ti ))

where: X   max{X , 0}
N – notional amount
 Contract scheme
L(TN-1,TN)
L(T1,T2)

L(T0,T1)
L(T3,T4) L(T4,T5)
K L(T2,T3)

............
.
cap

............
.
floor
............
.
Types of interest rates
- theory
 Spot and forward rates
• Spot rate – from now to some moment in
the future.
• Forward rate – between two moments in
the future (but fixed now)
Notation
t – present moment (if not specified, t=0)
T, S – some moments in the future (t<S<T)
Spot rate

R(t,T) or
L(t,T)

t T

Forward rate

F(t;S,T) or
L(t;S,T)
the rate is set at
t

t S T
 Setup
• We assume that for every moment T there is
an zero-coupon bond that will pay 1€ at T.
• Symbol for this bond: P(∙,T) or PT
• Its price at the moment t: P(t, T) or PT(t)
• P(t, T) is current (at t) value of 1€ at T, so it
is also the discount factor from T to t:
P(t, T) = d(t, T)
P(0,T) = d(T)
 Bonds’ prices term structure
d(T)=P(t,T) t is given („now”)

T
The curve starts at 1 and is downward-sloping
 Simple spot rate
1  P(t , T )(1  (T  t ) L(t , T ))
or
1
P(t , T ) 
1  L(t , T )  (T  t )
thus
1  P(t , T )
L(t , T ) 
(T  t )  P(t , T )

1€
P(t,T) €

t T
Continuously compound spot
rate
1  P(t , T ) exp (T  t ) R(t , T )
or
P(t , T )  1 exp  (T  t ) R(t , T )

thus
ln P(t , T )
R(t , T )  
T t
1€
P(t,T) €

t T
 Short rate (instantaneous)
When T→ t
ln P(t , T )  ln P(t , t )
R(t , T )   
   r (t )
T t T
r(t) is a spot rate for infinitely short period
[from t to t+dt]
If it is known in advance:
T
R(t , T )    rs ds
t

 T 
P(t , T )  exp   rs ds 
 t 
 Forward rates
At the moment t:
– Buy 1 zero-coupon bond PS, [ -P(t, S) €].
– To finance it sell P(t,S)/P(t,T) bonds PT [ +P(t, S) €].
At the moment t the value of this investment is 0.
The discount factor between T and S [at the moment t] is thus:
P(t , S )
1  d (t; S , T )
P(t , T )
P(t , T )
d (t ; S , T ) 
P(t , S )
P(t,S)/P(t,T)
€
1€

t S T
 Simple forward rate
P(t , T )
 d (t; S , T )  1  (T  S ) L(t; S , T )
P(t , S )

1  P(t , T )  P(t , T )  P(t , S )

L(t; S , T )   P(t , S )  1  (T  S ) P(t , S )
T S  

P(t,S)/P(t,T)
€
1€

t S T
 Continuously compound forward
rate
 d (t; S , T )  exp  (T  S ) F (t; S , T )
P(t , T )
P(t , S )

1  P(t , T )  ln P(t , T )  ln P(t , S )

F (t; S , T )   ln    
T  S  P(t , S )  T S

P(t,S)/P(t,T)
€
1€

t S T
 (Instantaneous) forward rate
When S→ T
ln P(t , T )  ln P(t , S )  ln P(t , T )
R(t; S , T )   
   f (t , T )
T S T

f(t,T) is a forward rate for infinitely short period

[from T to T+dT].
r (t )  f (t , t )

Always T
R(t , T )    f (t , s)ds
t

 T 
P(t , T )  exp   f (t , s)ds 
 t 
 Summary
1  P(t , T ) P(t , T )  P(t , S )
L(t , T )  L(t; S , T ) 
(T  t )  P(t , T ) (T  S ) P(t , S )

ln P(t , T ) ln P(t , T )  ln P(t , S )

R(t , T )   F (t; S , T )  
T t T S

 ln P(t , T )
r (t )  f (t , t ) f (t , T )  
T

 T 
P(t , T )  exp   f (t , s)ds 
 t 
 Summary
• There are several kinds of interest rates
• Spot rates are rates for periods starting
from now
• Forward rates – for periods in the future
(but interest rates are set now)
• Instantaneous rates – are rates for
infinitelly short periods – purely theoretical
constructions but used in financial
statistics
Estimating the term
structure of interest
rates
Types of interest rates (risk free)
• Interbank rates:
– LIBOR/LIBID (London Interbank Interest Rates)
– EURIBOR (Euro Interbank Interest Rates): panel of
banks from EU + some outside banks (US, UK,
Japan)
– STIBOR (Stockholm), WIBOR (Warsaw), PRIBOR
(Prague), ….
– Eonia – rates based on real transactions (only
overnight)
• Central bank interest rates
• Rates derived from the prices of government
bonds
 Shapes of term structure
3.5
3 3.5
2.5 3
2.5
2
2
1.5
1.5
1
1
0.5 0.5
0 0
0 5 10 15 20 0 5 10 15 20

0 5 10 15 20
Central bank and term structure
• CB has more control over short term rates
(policy to control EONIA through open
market operations)
• The influence on long term rates can be
ambiguous. Assume that CB rates grow
– Long term rates should also grow accordingly.
– But it can be anti-inflationary policy. If it is
believed to be successful, the rates will fall.
– If it is expected to fail, there is no reason for
rates to decrease.
Some theories of the term structure
• Expectation hypothesis – in the equilibrium the
long term rate should equal compound of
expected short term rates:
exp T  R(T )  exp t  R(t ) Et exp (T  t )  R(t , T  t )
• Liquidity preference – in the equilibrium long
term rates should be higher then then it would
result from expectation, because people prefer
more liquid assets.
• Partial markets – loans with different maturities
build different markets. Motivations of the agents
on this markets are different. One can consider
equilibrium on each of this market and general
equilibrium.
 Methods of term structure
k 1
estimating
Fc F
P(t , Tk )    
i 1 exp( R (t , Ti )(Ti  t )) exp( R(t , Tk )(Tk  t ))
k 1
  cFd (t , Ti )  F d (t , Tk )
i 1
100
P(t , T )   100exp(R(t , T )(T  t ))  100d (t , T )
exp(R(t , T )(T  t ))

• Bootstrap method
• Splines
• Nelson-Siegel method
• Svensson method
 Bootstrap method
Requires several bonds with ‘’overlapping”
moments of payments.
P(0.25)  F0.25d (0.25)  d (0.25)  P(0.25) / F0.25
ln d (0.25)
R(0.25)  
0.25
P(0.5)  c0.5 F0.5d (0.25)  (1  c0.5 ) F0.5d (0.5)

 d (0.25)  P(0.5)  c0.5 F0.5d (0.25)/1  c0.5 F0.5 

ln d (0.5)
R(0.5)  
0.5
P(0.75)  c0.75F0.75d (0.25)  c0.75F0.75d (0.5)  (1  c0.75 ) F0.75d (0.75)
ln d (0.75)
 d (0.75)  ... R(0.75)  
0.75
 Bootstrap method
One can solve this at once solving system of
equations
 P(T1 )   CF1T1  0  d (T1 ) 
P  CF d
0
    
          or
 P(T )  CF  CFKTK  d (TK ) 
 K   KT1 CFKT2

1
d  CF P

lnd (T ) 
R(T )  
T
 Splines
m
P(0, Tm )   CFm,T j  d (T j )   m
j 1

• CF is the cash flow connected with the

bond
•  is error term
• Discount factor d() is the sum of
„segmented” polynomials
• Let us define:
s
d (T )  1    i g i (T )
i 1

where gi are smooth functions

 Splines – defined on the
intervals
g is a polinomial on some interval [T , T ] and 0 outside of this
i i i-1
interval:

gi

Ti Ti+1
 McMulloch splines [1975]
For i < s:
0 for T  Ti 1
 T  T 3
 i 1
for Ti 1  T  Ti
 6Ti  Ti 1 

g i (T )   Ti  Ti 1 2 Ti  Ti 1 T  Ti  T  Ti 2 T  Ti 
3

    for Ti  T  Ti 1
6Ti 1  Ti 
2
6 2 2

T  T  2Ti 1  Ti  Ti 1  T  Ti 1  for T  Ti 1
 i 1 i 1  6 2 

For i=s:
gi (T )  T
 m
 s

P(Tm )   CFm, j 1    i g (T j )    m
j 1  i 1 
m s m
P(Tm )   CFm, j    i  CFm, j g i (T j )   m
j 1 i 1 j 1

One can use linear regression to obtain α

m

Dependent variable:P(T )   CF m
j 1
m, j

s independent variables: CFm, j g (T j )

j 1
for i=1, 2, …, s
 Nelson-Siegel method (NS)
T T

  3T  
f (T )  1   2 e  e

T
1
R(T )   f ( s)ds
Spot rates can be obtained by integrating:
T0

  
T T
 
R(T )  1   2   3  1 e    e 
T  3
 
Allows for a ”hump” in the term structure.
• There are 4 parameters
 1+2 equals the short rate
 1 is a consol rate (rate of the coupon bond with
infinite maturity): 1 = s()=f()
 Spread between consol rate and short rate is -2.
It can be interpreted it as a slope of yield curve.
 3 determines the shape of the yield curve. For
3>0 it has a hump and for 3 < 0 it is bend
downwards.
  > 0 is the speed with which the rates tend to
console rate
 Discount factors and bods’ prices
   T   T 
d (T )  exp T1    2   3 1  exp      3T exp  
 
       
m    T   T 
P(Tm )   CF j exp T1    2   3 1  exp      3T exp  
j 1        
Parameters are estimated by minimizing the sum of
errors between theoretical and empirical prices.
K
min
 , , , 
  i2
1 2 3
i 1

where  m  P(Tm )  pm

pm - market price of the bond m

 More advanced approach
Weighted least squares
K
min
 , , , 
 wi i2
1 2 3
i 1

Di1
wi 
 j 1 D j 1
K

where Di is duration of bond i (its sensitivity

wrt interest rate)
P(Ti )
Di 
r
 Penality for lack of smoothness
K 
min  wi i  S (1 , 2 , 3 ,  )
2
1 , 2 , 3 , 
 i 1 
where
2
T
  f ( s;1 , 2 , 3 ,  ) 
2
S (1 , 2 , 3 ,  )    
 ds
0
s 2

 Svensson method (SV)
• Two additional parameters.
• Model is more elastic. It allows for two
‘’humps”.
• Parameters do not have easy
interpretation.
• Parameter estimation is multidimensional
optimization problem – not an easy task.
Equations for forward and spot
rates
t t t t t 
f (t )   1   2 exp    3 exp    4 exp 
 1  1  1  2  2 

1  exp(t / 1 )  1  exp(t / 1 )   t 
R(t )   1   2 1  3   exp   
 
t  t  1  
 (1  exp(t /  2 )  2   t 
  4   exp  

 t  2 
 Term structure estimation in various
countries
Country Method Minimized error Maturities
Belgium NS, SV prices (weighted) up to16 years
France NS, SV prices (weighted) up to10 years
Germany SV yields up to 10 years
Italy NS prices (weighted) up to 30 years
Japan splines prices up to 10 years
Spain SV prices (weighted) up to 10 years
Sweden SV, splines yields up to 10 years
Switzerlan SV yields up to 30 years
d
UK splines (variable penalty yields up to 30 years
roughness )
US splines prices up to 10 years
Source: BIS Papers No 25 (2005)
Example: Bundesbank data
Term structure in Germany (2008-2010)
Term structure in Armenia
Discrete models of
rates’ dynamics
 Model
• Probabilistic model – the sample space Ω
and probabilistic measure P.
• Discrete model:
- finite number of moments t=0,1,2,…,T.
- finite sample space Ω (the set of possible
states of the world)
 Bonds
• For each moment t=0,…,T there is a bond
that matures at t.
• Its price at s equals P(s,t)=Pt(s).
• Pt(t)=1.
• At each moment t the term structure of
bonds’ prices is given: (Pt(t)=1, Pt+1(t),
Pt+2(t) …, PT(t)).
 Rates
Forward rates(instantaneous)
P (t )
f (t )  F (t; ,  1)  1
P 1 (t )
Short rate
Pt 1 (t  1)  Pt 1 (t ) 1
r (t )    1  ft (t )
Pt 1 (t ) Pt 1 (t )
 Possible approaches
• Models for bonds’ prices
• Models for forward rates
• Models for short rate

The approaches are equivalent. From one

model one can obtain another.
 From bonds’ prices to forward rate and short rate –
formulas were given
From forward rates to bonds’ prices:
.
P 1 (t )
P (t ) 
1  f 1 (t )
1
P (t ) 
1  ft (t ) 1  ft 1 (t )  1  f 1 (t ) 

From short rate to bonds’ prices:

1
Pt 1 (t ) 
1  rt

what with Pt+2(t), Pt+3(t), …, PT(t)?

 One factor model
• At each moment an in each situation there
are only two possible changes: u and d.
• Changes are random.
• The prices of all bonds are known. At the
moment t the price Pt+1(t) is known. As
Pt+1(t+1)=1, the rt [short rate from t to t+1]
is known at t.
 Model 1
P1(1,u) = 1
P2(2) = 1
P2(1,u) = 0.8333
P0(0) = 1
P1(0) = 0.9091
P2(0) = 0.7920

P1(1,d) = 1
P2(2) = 1
P2(1,d) = 0.9091

t=0 t=1 t=2

 Model 2

P1(1,u) = 1
P2(2) = 1
P2(1,u) = 0.8333

P0(0) = 1
P1(0) = 0.9091
P2(0) = 0.7438

P1(1,d) = 1 P2(2) = 1
P2(1,d) = 0.9091

t=0 t=1 t=2

 Model 3
1
1
1 0.8333
0.8333
0.7095 1
1 1
0.8696
0.8333
0.7418 1
0.6587 1
1 0.8696
0.9091
1
0.8175 1
0.9091 1
0.9091
0.8092
0.7086
1
0.6225 1
1 0.8333
0.8333
1 0.7171 1
0.9091 1
0.8696
0.7920
0.6934 1
1
1 0.8696
0.9091
0.8085 1
1
0.9091

t=0 t=1 t=2 t=3 t=4

 Model 4
1
1
1 0.8333
0.8333
0.7020 1
1 1
0.8696
0.8333
0.7260
1
0.6443 1
1 0.8696
0.9091
1 0.8444 1
0.9091 1
0.9091
0.7920
0.6822
0.5972 1
1
1 0.8333
0.8333
1 0.7095 1
0.9091 1
0.8696
0.7748
0.6696
1
1
1 0.8696
0.9091
0.8175 1
1
0.9091

t=0 t=1 t=2 t=3 t=4

 Model 5 – two factor model
1
1
0.8333
1
1
0.8696 1
0.8696
0.7604
1
1
0.9091
1
1
0.8696
1 1
0.9091 1
0.9091 1
0.8353 0.9091
0.8275
0.7453 1
1
0.9524
1
1
0.8333
1
1
0.9524 1
0.9091
0.8524
1
1
0.9524
t=0 t=1 t=2 t=3
One can describe it specyfying the
rates of growth
P (t  1, u ) d (t ) 
P (t  1, d )
u (t )  
P (t ) P (t )

P (t  1, x)  x (t ) P (t ), x  u, d

Thus:

1
ut 1 (t )  dt 1 (t ) 
Pt 1 (t )

r (t )  ut 1 (t )  1  dt 1 (t )  1
 Dynamics of short rate

Pt 1 (t  1, u )  Pt 1 (t ) Pt 1 (t  1, d )  Pt 1 (t ) 1
r (t )    1
Pt 1 (t ) Pt 1 (t ) Pt 1 (t )

r (t )  ut 1 (t )  1  dt 1 (t )  1
 Exercise
• Calculate r in the models 1 and 2 as well
as in the models 3 and 4.
• Calculate rates u and d in the models.
 Models 1 and 2

t=0 t=1 t=2

 Models 1 and 2

r1 = 1/0.8333−1 =
X
0.20

r0 = 1/0.9091−1 = 0.10

r1 = 1/0.9091−1 =
X
0.10

t=0 t=1 t=2

 Models 3 and 4
0.20 X
0.20
0.15 X
0.20
0.15 X
0.10
0.10 X
0.10
0.20 X
0.20
0.15 X
0.10
0.15 X
0.10
0.10 X

t=0 t=1 t=2 t=3 t=4

 Forward rates
• At each moment and in each situation the
term structure of (instantanous) forward rates
is known.
P (t )
f (t )  F (t; ,  1)  1
P 1 (t )

• Thus the dynamic of short rate is known.

r (t )  ft (t )
Forward rates in model 1

f1(1) = 0.20 X

f0(0) = 0.10
f1(0) = 0.10
f1(1) = 0.10 X

t=0 t=1 t=2

 Model 6 – forward rates
0.2000 X
0.2000
0.1871 0.1500 X
0.2000
0.1478
0.1268 0.1500 X
0.1000
0.0766
0.1000 0.1000 X
0.1478
0.1610 0.2000 X
0.1423 0.2000
0.1745
0.1000 0.1500 X
0.1733
0.1572 0.1000 0.1500 X
0.1121
0.1000 X

t=0 t=1 t=2 t=3 t=4

Calculate short rate and bonds’

 Model and the economic
equilibrium
• In the equilibrium one cannot build a portfolio that
brings profits without risk
• If such porfolio exist, the prices of the appriopriate
bonds would rise (of fall) – it is not an equilibrium
• The bond should be properly prices relative to
– Short rate
– Other bonds
• Markets for bonds with different maturities are
interconnectes – equilibrium should be on all
markets
 Model 2

At t=0 sell 1000 bonds P1 (+909.1€) and buy

1222 (909.1/0.7438) bonds P2. Cost = 0 €.
At t=1 we should pay 1000 € for P1 bonds.
u: value of P2 = 1222∙0.8333=1018.29 € (… -
1000 = 18.29 €)
d: value P2 = 1222∙0.9091=1110.92 € (… -
1000 = 110.92 €)
 Model 1
Buy n1 bonds P1 and n2 bonds P2.
Initial value should =0, thus
0.9091
n2   n1  1.1478n1
0.7920

u: V1 (u)  n1  0.8333n2  0.0435n1

d: V1 (d )  n1  0.90913n2  0.0435n1

It is impossible that for any n1 both number are

positive.
 General two period model
P1(1,u) = 1
P2(2) = 1
P2(1,u) = P2(0)u2(1)
P0(0) = 1
P1(0)
P2(0) P1(1,d) = 1
P2(2) = 1
P2(1,d) = P2(0)d2(1)

t=0 t=1 t=2

Initial value V0 = P1(0)n1 + P2(2)n2 = 0

P1 (0) 1
n2   n1   n1
P2 (0) P2 (0)(1  r (0))
 n1  u (1) 
V1 (u )  n1  n2 P2 (1, u )  n1  P2 (0)u2 (1)  n1 1  2 .
P2 (0)(1  r (0))  1  r (0) 

n1  d (1) 
V1 (d )  n1  n2 P2 (1, d )  n1  P2 (0)d 2 (1)  n1 1  2 
P2 (0)(1  r0 )  1  r0 

Assume that u2(1)<d2(1).

If both expresions in the brackets have the same sign, then there is
no equilibrium.
Thus we should have
u2 (1)
1
1  r (0)
and

d 2 (1)
1
1  r (0)
Putting all together:
u2 (1)  1  r (0)  d2 (1)
There exists a number 0<q<1, that

1  r (0)  qu2 (1)  (1  q)d2 (1)

Multiplying by P2(0) and dividing by 1+r(0) :

P2 (0)u2 (1) P (0)d 2 (1) P (1, u ) P (1, d )

P2 (0)  q  1  q  2 q 2  1  q  2
1  r (0) 1  r (0) 1  r (0) 1  r (0)

Thus

 P (1) 
P2 (0)  E q  2 
1  r (0) 
q is a martingale probability
 General model
Pt(t,u) = 1
Pt+1(t,u) = Pt+1(t−1)ut+1(t)
Pt+2(t,u) = Pt+2(t−1)ut+2(t)
………….
Pt−1(t−1) = 1 PT(t,u) = PT(t−1)uT(t)
Pt(t−1)
Pt+1(t−1)
……. Pt(t,d) = 1
PT(t−1) Pt+1(t,d) = Pt+1(t−1)dt+1(t)
Pt+2(t,d) = Pt+2(t−1)dt+2(t)
………….
PT(t,d) = PT(t−1)dT(t)

t−1 t
 Portfolio at t-1
Vt 1  n0 Pt (t  1)  n1Pt 1 (t  1)  ...  nT t PT (t  1)
Suppose that its initial value is 0:
Pt 1 (t  1) P (t  1)
n0  n1  ...  nT t T  n1 Pt 1 (t  1) 1  r (t  1)   ...  nT t PT (t  1) 1  r (t  1)  .
Pt (t  1) Pt (t  1)

u: Vt (u )  n1 Pt 1 (t  1) 1  r (t  1)    nT t PT (t  1) 1  r (t  1)  
 n1 Pt 1 (t  1)ut 1 (t )   nT t PT (t  1)uT (t ) 
 n1 Pt 1 (t  1)  ut 1 (t )  1  r (t  1)    nT t PT (t  1)  uT (t )  1  r (t  1) 

d: Vt (u)  n1Pt 1 (t  1)  dt 1 (t )  1  r (t  1)    nT t PT (t  1)  dT (t )  1  r (t  1) 
In the matrix form:
 n1 
 Vt (u )   
  A 
 t
V ( d )  n 
 T t 
where:
 P (t  1)  ut 1 (t )  1  r (t  1)  PT (t  1)  uT (t )  1  r (t  1)  
A   t 1 
 Pt 1 (t  1)  dt 1 (t )  1  r (t  1)  PT (t  1)  dT (t )  1  r (t  1)  

If the prices are in equilibrium the matrix A has rang<2

(otherwise for any (V(u), V(d) there would be a solution).
Rows are linearly dependent, thus the value
ut  k (t )  1  r (t  1)
dt  k (t )  1  r (t  1)

is the same for each k.

Consider a portfolio of Pk and Pt
Vt (u)  Pt  k (t  1)  ut  k  1  r (t  1)  nk
Vt (d )  Pt  k (t  1)  dt k  1  r (t  1)  nk
Assume that ut k (t )  dt k (t ) . The expresion in the bracket
should have the same signs: ut k (t )  1  r (t  1)  dt k (t )
There exists a number 0<q<1 that1  r (t  1)  qut k (t )  (1  q)dt k (t )
One can calculate”
dt  k (t )  1  r (t  1)
q
dt  k (t )  ut  k (t )

q is the same for each k.

 Martingale measure
Multiplying by Pt+k(t-1) and dividing by 1+r(t-
1):
Pt  k (t , u ) Pt  k (t , d )
Pt  k (t  1)  q  (1  q)
1  r (t  1) 1  r (t  1)

Thus
 Pt  k (t ) 
Pt  k (t  1)  E q
t 1 
 1  r (t  1) 
 Theorem

Prices of all bonds are in equilibrium if there

exists a measure such that
Pt  k (t , u ) Pt  k (t , d ) q  Pt  k (t ) 
Pt  k (t  1)  q  (1  q)  Et 1  
1  r (t  1) 1  r (t  1)  1  r (t  1) 

q and 1-q are martingale probabilities

(martingale measure)
 Model 1
r(0)=0.1
0.8333 0.9091
0.7920  q  (1  q)
1.1 1.1

thus q=0.5
 Model 2
r(0)=0.1

0.8333 0.9091
0.7438  q  (1  q)
1.1 1.1

thus q=1.2, 1-q=-0.2<0!

The price of the bond is inconsistent with short rate.

 Model 3
Exercise: calculate martingale probabilities.
Do it for each node separately. For example
for t=1 and upper vertex::
0.8333 0.9091
0.7418  q  (1  q)
1.2 1.2

thus q=0.25. And so on.

 Model 3

0.5
0.25 0.5

0.75 0.25
0.25
0.75

0.75 0.25
0.5 0.75
0.5
0.5
0.5

t=0 t=1 t=2 t=3

 Model 4
Exercise: calculate martingale probabilities.
 Model 4 Inconsiste
nt pricing
0.75

0.5 0.25

0.5 -0.5
0.5
1.5

0.5 0.5
0.75
0.5
0.25
0.25
0.75

t=0 t=1 t=2 t=3

 What is wrong?
Shor rate r=0.10.
P4(3) discounted to t=2:
u: 0.8696/1.1=0.7954 < 0.8444=P4(2)
d: 0.9091/1.1=0.8264 < 0.8444=P4(2)

Bond P4 is too expensive at t=3. There is inconsistency between

short rate and price of this bond. We should sell P4 and buy
P3:
Sell 1 P4 : + 0.8444 €. We could buy 0.8444/0.9091=0.9288
bonds P3.

u: -0.8696 + 0.9288∙ 1 = 0.0592 > 0

d: -0.9091 + 0.9288∙ 1 = 0.0197 > 0
 Short rate model
Assume that we know rt and martingale probabilities.
At T-1 prices of PT we can obtain with
1
PT (T  1) 
1  r (T  1)

At T-2 the price of PT-1 we obtain by discounting and

price of PT – using martingale probabilities:
 PT (T  1) 
PT (T  2)  ETq 2 1  r (T  2) 
 
 Short rate model
Consider any moment t. We have prices Ps
for s>t+1 for every moments after t.
The price of Pt+1 we obtain by discounting:
1
Pt 1 (t ) 
1  rt

The price of Pt+k (k>1):

 Pt  k (t  1) 
Pt  k (t )  Etq  1  r (t ) 
 
 Example
0.5 0.20 X
0.5 0.20 0.5
0.10 X
0.20
0.5 0.5 0.5 0.20 X
0.10
0.5
0.10 X
0.10
0.5 0.20 X
0.5 0.20
0.5 0.5
0.10 X
0.10
0.5 0.5 0.20 X
0.10 0.5
0.10 X

t=0 t=1 t=2 t=3 t=4

Compute bonds’ prices

 Summary
1. Models of bonds’ prices
2. Models of forward rates
One should check if there is an equilibrium.

3. Models of short rate + martingale

measure
Describes equilibrium. Prices should be
calculated.
 Ho-Lee model
If interest rates are deterministic:
P (t )
P (t  1) 
Pt 1 (t )
We add distortion (random factor):
P (t )
P (t  1)  hu (  t  1)
Pt 1 (t )
P (t )
P (t  1)   hd (  t  1)
Pt 1 (t )

hu and hd are distortin functions

In each node there are 2 possible next nodes: u and d

hu (  t  1) hd (  t  1)
u (t  1)  d (t  1) 
Pt 1 (t ) Pt 1 (t )
 Assumption
• Equilibrium.
• Martingale probabilities (q, 1-q) are the
same for each node.
• Recombination: paths u-d and d-u should
lead to the same end state (to the same
prices).
 Solution
The only functions that fulfills the assumptions
are:
1 
hu ( )  hd ( ) 
q  (1  q)  q  (1  q) 

Notice that
PT (t , d )
  T t
PT (t , u )
so δ describes the dispersion of possible prices
(δ-1 can be interpreted as volatility of prices)
 Example
δ=0.8, q=0.9
τ 0 1 2 3
hu (τ) 1.00000 1.02041 1.03734 1.05130

hd (τ) 1.00000 0.81633 0.66390 0.53827

Initial condition. At t=0 :

P1(0)=0.9091,
P2(0)=0.8265,
P3(0)=0.7513,
P4(0)=0.6830,
We can calculate prices at next notes. For
example for node u at t=1:
P1 (1)  1
P2 (0) 0,8265
P2 (1)  hu (1)  ·1,02041  0,9276
P1 (0) 0,9091
P3 (0) 0,7513
P3 (1)  hu (2)  ·1,03734  0,8573
P1 (0) 0,9091

P4 (0) 0,6830
P4 (1)  hu (3)  ·1,05130  0,7899
P1 (0) 0,9091
 Model7
1
1
0.9 0.9557
1
0.9430
1 0.9 0.8833 0.1
0.9276 1
1 0.9 0.8573 1
0.1 0.7646
0.9091 0.7899 0.9
0.8265 1
0.7513 0.7544
0.1 0.9 0.5653 0.1
0.6830 1
0.7421 1
1
0.5487 0.6117
0.1 0.9
0.4044 1
0.6035
0.3618 0.1
1
1
0.4893

t=0 t=1 t=2 t=3 t=4

 Prices of any other instruments
connected with interest rate
In the equilibrium the price should be consistent
with bonds’ prices.
1. Start with nodes where price is known
(payoffs)
2. If for some vertex a prices in all child nodes
are known, compute the price in a by
discounting and taking expected value with
respect to martingale probabilities.
3. Repeat 2 until you obtain prices in all nodes.
Model 7, call option on P4 with execution price 0.7000 and
execution time 3
1. We put payoffs

0.2557 X
0.9

0.9 0.1

0.9 0.0646 X
0.1 0.9
0

0.1 0.9 0.1

0 X
0.1 0.9

0.1
0 X

t=0 t=1 t=2 t=3 t=4

2. Prices at t=2

0.2557
0.9
(0.9∙0.2557+0.1∙0.0646)
0.9 ∙0.8833=0.2090 0.1

0.9 0.0646
0.1 0.9
0
0.9∙0.0646
0.1 0.9 ∙0.7544=0.0439 0.1

0
0.1 0.9

0
0.1
0

t=0 t=1 t=2 t=3

2. Prices at t=1

0.2557
0.9

0.9 0.2090
0.1
(0.9∙0.2090+0.1∙0.0439)
0.9 ∙0.9276=0.1785 0.0646
0.1 0.9
0

0.1 0.0439
0.9 0.1
(0.9∙0.0439+0.1∙0)∙ 0
0.7421=0.0293 0.9
0.1

0
0.1
0

t=1 t=2
t=0 t=3
Prices t=0

0.2557
0.9

0.9 0.2090
0.1

0.9 0.1785 0.0646

0.1 0.9
(0.9∙0.1785+0.1∙0.0293)
∙0.9091=0.1495 0.0439
0.1 0.9 0.1

0.0293 0
0.9
0.1

0
0.1
0

t=1 t=2
t=0 t=3
 Literature
S. Kellison, Theory of Interest, McGrew/Hill, 2008.

D. Filipović, Term-Structure Models. A Graduate Course, Springer 2009.

T. Björk, Arbitrage Theory in Continuous Time, Oxford Univ. Press, 2004
(second edition)
M. Musiela, M. Rutkowski, Martingale Methods in Financial Modelling, Springer
2009 (second edition)
D. Brigo, F. Mercurio, Interest Rate Models – Theory and practice, Springer
2006 (second edition)
J. James, N. Webber, Interest Rate Modelling, Wiley & Sons, 2000.
R. Rebonato, Modern Pricing of Interest-Rate Derivatives, Princeton Univ.
Press, 2002.

R. Ferstl, J. Hayden, 2010, Zero-Coupon Yield Curve Estimation with the

Package termstrc, Journal of Statistical Software, vol. 36, pp. 1-34.