交通大學 財務金融研究所 - 財務工程組 研 究 生 : 何俊儒 指導教授 : ...
DESCRIPTION
Using The LIBOR Market Model to Price The Interest Rate Derivatives: A Recombining Binomial Tree Methodology. 交通大學 財務金融研究所 - 財務工程組 研 究 生 : 何俊儒 指導教授 : 鍾惠民 博士 戴天時 博士. Agenda. Introduction Model and Methodology Application Conclusions. Introduction . - Review of Interest Rate Models - PowerPoint PPT PresentationTRANSCRIPT
Using The LIBOR Market Model to Price The Interest Rate Derivatives: A Recombining Binomial Tree Methodology
交通大學 財務金融研究所 - 財務工程組研 究 生 : 何俊儒指導教授 : 鍾惠民 博士 戴天時 博士
• Introduction
• Model and Methodology
• Application
• Conclusions
Agenda
Introduction -Review of Interest Rate Models-Motivation-Contribution
Black’s Model Short Rate Model
• one factor • do not match the volatility structure
Forward Rate Model• HJM (Instantaneous forward rate)• LIBOR Market Model (LMM)
Non-Markov Property No analytical and Numerical solution
Review of Interest Rate Models
Motivation
We can’t observe the instantaneous short rate and instantaneous forward rate in the traditional interest rate models
We adapt LIBOR market model which is based on the forward LIBOR rate that we can observe from the daily market
When implementing LMM in lattice method, we face the explosive tree due to the non-Markov property
5
6
Motivation The non-recombining nodes make our pricing
procedure inefficient and don’t satisfy the market requirement
Contribution We apply the HSS methodology into LMM that
make the nodes combine and make the pricing the derivatives feasible
The method we proposed make our valuation more efficient and more accurate
Model and Methodology- HSS Methodology- Propositions- The Derivation of Expectation
Poon and Stapleton text
(2005)
HSS Multivariate Binomial
methodology (1995)
Conditional Probability
Up/ Down movement
Drift of the discrete time forward rate
Forward Rate Recombining Binomial
Tree
9
•Ho, Stapleton and Subrahmanyam (1995) [HSS] suggest a general methodology for creating a recombining multi-variant binomial tree to approximate a multi-variant lognormal process
•Our assumption about LMM satisfies the conditions in the HSS methodology and apply it into LMM
HSS Methodology10
HSS Methodology
2 0,2ˆ ˆ,
1 2n 2 2 n
2 1 2N n n
TIME 0 1 2
0X
1 0,1ˆ ˆ,
10 1
nX u
1 1 10 1 1
n r rX u d
10 1
nX d
1 20 2
n nX u
1 2 2 20 2 2
n n r rX u d
1 20 2
n nX d
11
1 0 0,( | )rq x x x
2 1 1,( | )rq x x x
1,0X
1,1X
1,2X
1 0,1,2r 2 0,1, , 4r
2,0X
2,1X
2,2X
2,3X
2,4X
Time 0 1y 2y 3yone year
f (0;2,3)
•We apply this methodology into the LMM and make some change to satisfy our conventionsLet Xt,r= f (t ;Tn ,Tn+1)r , X0 = f (0 ;Tn ,Tn+1)
Ex:
•We have the following propositions
12
PROPOSITION 1•For the forward LIBOR rate which follows the
lognormal distribution, we can choose the proper up and down movements to determine the i-th period of the -maturity forward LIBOR rate and have the form
where1 1 1 2( ; , ) (0; , ) , , , ,iN r r
n n r n n i i nf i T T f T T u d i T T T
1/1 1
1, 1
1 1
1
2[ ( ( ; , )) / (0; , )]1 exp(2 ( ) / )
2[ ( ( ; , )) / (0; , )]
: node's number from top to bottom at time
iNn n n n
ii i i i i
i n n n n i
i i i
i
E f i T T f T TdT T n
u E f i T T f T T dN N nr T
nT
13
1 2n 2 2n TIME 0 1 2
f(0;2,3)
f(1;2,3)
f(2;2,3)
The Binomial Trees in Discrete Time Forward Rate
(0;2)f
11(0;2) nf u
1 1 11 1(0;2) n r rf u d
11(0;2) nf d
1 22(0;2) n nf u
1 2 2 22 2(0;2) n n r rf u d
1 22(0;2) n nf d
14
f(0;1,2)
f(1;1,2)
PROPOSITION 2•Suppose that the forward LIBOR rate are joint
lognormally distributed. If the are approximated with binomial distributions with stages and and given by proposition 1, and if the conditional probability of an up movement at node r at time is
1( ; , )n nf i T T
1 1 2( ; , ), , , ,n n nf i T T i T T T
1i i iN N n iu id
iT1 1
1 1,( ) ( ) ln ln( | )
(ln ln )ln
, ln ln
i i i i ii i i r
i i i
i
i i
E x N r u r dq x x xn u d
di r
u d
15
PROPOSITION 2where
•To determining the conditional probability, it has some skills to use for the term
•We first derive term . Since forward rate is log normally distributed, we have
1
1
1 1 1,
2 2 20, 1 1, 1 0, 1
( ; , )ln(0; , )
( ) ( ) ( )
[ ( ) ] /
n ni
n n
i i i i i i i r
i i i i i i i i i
f i T Txf T T
E x E x b E x b x
b t t t t
1( )i iE x
( )iE x
210,
1
( ( ; , )) 1( ) ln[ ](0; , ) 2
n ni i
n n
E f i T TE xf T T
16
The Derivation of Expectation
•The most important result in Poon and Stapleton text is the drift of the discrete time forward rate and we rewrite as
•Then multiple the term on both side to get the
general form of
1
1
( ; , )(0; , )
n n
n n
f t T Tf T T
1 1( ( ; , )) / (0; , )t i n n n nE f T T T f T T
1 2 2 31 1 21, 2,
1 1 1 2 2 2 3
1,
1
[ ( ; , )] ( ; , )( ; , )1( ; , ) 1 ( ; , ) 1 ( ; , )
( ; , ) 1 ( ; , )
t i n nn n
n n
n n nn n
n n n
E f T T T f t T Tf t T Tf t T T f t T T f t T T
f t T Tf t T T
17
The Derivation of Expectation
•The general form of expectation is given as
1
1
2 2 3 11 1 21, 2, ,
1 1
1 1
1
2 3
1
2 2 1
[ ( ; , )]( ; , )
( ; , ) ( ; , )( ; , )11 ( ; , ) 1 ( ; , ) 1 ( ; ,
( ; , ) ( ; , )(0; , ) ,
)
(0; )t i n n
n n
n n nn n n n
n n n n
n n
n n
n
n
n
E f T T Tf t T T
f t T T f t T Tf t T Tf t T T f t T
f t T T f t T Tf T T
T f t T T
f T T
Special Case
•When nl stages approach infinite , the sum of nl stages also approach infinite, that is
•We can reduce the up and down movements to the briefer form which is easier to calculate and show as follows
•The conditional probability as
1, ,l i 1
ii ll
N n
1, 1
21 exp(2 ( ) / )
2
ii i i i i
i i
dT T n
u d
( ) 0.5lq x ln
Application- Bond Option- Cap- Sensitive Analysis
•The embedded option on ZCB is a bond that can be callable before maturity date with a callable price K
•We have a three years maturity zero coupon bond with a callable price K equal to 0.952381 dollar at year two.
21
option maturity
Time 0 1y 2y 3yone year
bond maturityC0
The Valuation of Embedded Option on Zero Coupon Bond in LMM
22
0.000000000000
0.000500000000
0.001000000000
0.001500000000
0.002000000000
0.002500000000
Embedded Option Value
Embedded Option Value
strike price
Sensitive Analysis – Strike Price
23
0.000000000000
0.001000000000
0.002000000000
0.003000000000
0.004000000000
0.005000000000
0.006000000000
0.007000000000
0.008000000000 Embedded Option Value
Em-bedded Option Value
volatil-ity
Sensitive Analysis – Volatility
The Valuation of Caplets in LMM•The payoff function of a caplet at time is
• It is a caplet on the spot rate observed at time with payoff occurring at time
•The cap is a portfolio consisted of n such call options which the underlying is known as caplet
24
1iT
1max( ( ; , ) ,0)i i iA f T T T K iT
1iT
The Theoretical Value of Caplet
•We use the Black’s formula for the caplet to get the theoretical value of caplet and rewrite as
where
25
1 1 1 2( ) ( , )[ ( ; , ) ( ) ( )]i i i i icaplet t A P t T f t T T N d KN d
21
1
21
2
ln( ( ; , ) / ) ( ) / 2 ,
ln( ( ; , ) / ) ( ) / 2,
i i i i
i i
i i i i
i i
f t T T K T tdT t
f t T T K T td
T t
Example
•Assume , , forward curve is flat equal to 5% and
•We have the one period caplet at time 0 caplet1(0)
26
1(0) (0, 2) [max( (1;1,2) ,0)] 1 1 (0,2) 0.0020112666 0.90702948 0.0020112666 0.0018242781
caplet A P E f KP
1A 10%, 5%K 25, 1,...,10in i
Table 1 Volatility is 10% and stage for every period is 25
27
Maturity Black Lattice Difference Relative Difference (%)
1 0.0018085085 0.0018242781 0.0000157696 0.8719669656 2 0.0024348117 0.0024407546 0.0000059429 0.2440786919 3 0.0028388399 0.0028374958 -0.0000013441 0.0473484800 4 0.0031206153 0.0031282191 0.0000076038 0.2436631792 5 0.0033214311 0.0033204098 -0.0000010214 0.0307505629 6 0.0034637453 0.0034664184 0.0000026731 0.0771737036 7 0.0035616356 0.0035658574 0.0000042219 0.1185369137 8 0.0036247299 0.0036200240 -0.0000047059 0.1298286011 9 0.0036600091 0.0036633313 0.0000033221 0.0907678678
10 0.0036727489 0.0036743568 0.0000016079 0.0437804213
RMSE 0.0000063671 1. Caplet assume δ = 1 and stage 252. Assume volatility is 10%, the forward curve is flat 5%
28
25 27 29 31 33 35 37 39 41 43 45 47 490.000000000000
0.000002000000
0.000004000000
0.000006000000
0.000008000000
0.000010000000
0.000012000000
0.000014000000 RMSE vs. Volatility
RMSE with 10% volatility
stage
Conclusions
Conclusions•Constructing the recombining binomial tree, the
payoff of the interest rate derivatives on each node can be obtained
•Comparing to the theoretical value, we find the theoretical value and lattice method is close when we increase the stages.
•The sensitive analysis results of embedded option are consistent with inference of Greek letters
30
~ The End ~
Thank you for your listening
31
Future works•Deriving the joint probability between the states of
different binomial forward rate trees.•Adjusting the stages between period by period to fit
the strike price to reduce the nonlinearity error•Trying to change the constant volatility to stochastic
volatility to fit the volatility term structure.
32
Review of Interest Rate Models
• Equilibrium models• No-arbitrage models
Instantaneous short rate models Instantaneous forward rate model Forward rate model
33
Forward rate modelLIBOR Market Model (LMM)
•The LMM was discovered by Brace, Gatarek, and Musiela (1997) and was initially referred to as the BGM model by practitioners
•There are two commonly used versions of the LMM:▫one is the lognormal forward LIBOR model (LFM) for
pricing caps▫the other is the lognormal swap model (LSM) for
pricing swaptions
34
Forward rate modelLIBOR Market Model (LMM)
•The LFM specifies the forward rate following zero-drift stochastic process under its own forward measure:
▫ is a Brownian motion under the forward measure defined with respect to the numeraire asset
▫ measures the volatility of the forward rate process
1( ; , )i if t T T
1
1
( ; , ) ( ) ( )( ; , )
i ii i
i i
df t T T t dW tf t T T
( )idW tiP
( )i t
35
Market Conventions of the LMM
•The relationship between the discrete LIBOR rate for the term and the zero coupon
bond price , given as follows:
where is the time line and is called the tenor for the period to
11
1( , )1 ( , )i i
i i i
P T TL T T
1( , )i iL T T 1i i iT T
1( , )i iP T T
0 1 2 nt T T T T i
iT 1iT
36
Forward rate agreement (FRA) Definition
A forward rate agreement (FRA) is an agreement made at time t to exchange fixed-rate interest payments at rate k for variable rate payments, on a principal amount A, for the loan period t+T to t+T+1.
-Ak
Ayt+
Tt+T+1
t+T
t
a T-maturity FRA
Forward rate agreement (FRA) The contract is usually settled in cash at t+T on a discounted basis.
One-period case
,
,,
( )1
Assume 1,
[ ] 01
t Tt T
t T
t t T
t T t t TQt t T t
t T
A y kFRA
yA k f
y fFRA E
y
1 , 1 , 11
1 1 1
, 1 1 11 1 1
[ ] 0 [ ] [ ]1 1 1
1 1 1[ ] [ ] [ ] cov [ , ]1 1 1
t t t t tQ Q Qtt t t
t t t
Q Q Q Qt t t t t t t t
t t t
y f fyE E Ey y y
f E E y E yy y y
The one-period ahead forward price of the n-period bond is just the expected value of the subsequent period spot price of the bond.
1, 1, 1, 1, 1 1, 1, 1
, 1, 1, 1, 1, ,
1 , 1
1, 1
[ ] [ ] [ ] cov [ , ]
1 1[ ]1 1
1thus, [ ]1
Q Q Q Qt t t n t t t t n t t t t t t n t t
t t t n t t n t t T t t T t n
Qt
t t t
Qt t
t t
E B E F E B B B
F B B F
Ey f
E yf
, 11
1 , 1
1 , 1 , 1 11
1cov [ , ]1 1
1 [ ] (1 ) cov [ , ]1
t tQt t
t t t
Q Qt t t t t t t t
t
fy
y f
E y f f yy
The Drift of Forward Rate under lognormal
The Drift of Forward Rate under lognormal
Qt 1, , , 1 , 2
1, 2,, , 1 , 2
,,
,
, 1, 1,
[ ]1 1
1
where cov [ln , ln ],
t t T t t T t t t tT T
t t T t t t t
t t TT T
t t T
QT t t t t t T
E f f f ff f f
ff
f f
is the loan period.
general case
HSS Methodology•HSS assume the price of underlying asset follows
a lognormal diffusion process:
where and are the instantaneous drift and volatility of , and is a standard Brownian motion▫They denote the unconditional mean of the logarithmic
asset return at time i as ▫The conditional volatility over the period i -1 to i is
denoted as▫The unconditional volatility is
ln ( ) ( ( ), ) ( ) ( )d X t X t t dt t dW t
ln X ( )W t
1,i i
i
0,i
41
HSS Methodology•The general form of at node r
where
Lemma 1. Suppose that the up and down movements and are chosen so that
iX
, 0iN r r
i r i iX X u d1
ii ll
N n
iu id1
0
1, 1
1
0
2( ( ) / ) , 1,2, , ,1 exp(2 ( ) / )
2( ( ) / ) , 1, 2, , ,
i
i
Ni
ii i i i i
Ni i i
E X Xd i mt t n
u E X X d i m
42
•They denote and the probabilities to reach given a node at as
HSS Methodology0ln( / )i ix X X
1it 1 1,( | ) ( )i i i r iq x x x or q x 1,i rx
43
HSS Methodology
1 0,1ˆ ˆ,
2 0,2ˆ ˆ,
1 2n 2 3n
2 1 2N n n
TIME 0 1 2
0X
10 1
nX u
1 1 10 1 1
n r rX u d
10 1
nX d
1 20 2
n nX u
1 2 2 20 2 2
n n r rX u d
1 20 2
n nX d
44
1 2n 2 3n TIME 0 1 2
f(0;2,3)
f(1;2,3)
f(2;2,3)The Binomial Tree in Discrete Time Forward Rate
(0;2)f
11(0;2) nf u
1 1 11 1(0;2) n r rf u d
11(0;2) nf d
1 22(0;2) n nf u
1 2 2 22 2(0;2) n n r rf u d
1 22(0;2) n nf d
1 0 0,( | )rq x x x 2 1 1,( | )rq x x x
45
Using Black’s Model to Price European Options
•Let P(t ,T ) as the price at time t of a zero-coupon bond that pays off $1 at time T.
•E(VT) = F0 in a risk neutral world
where
46
1 0 1 2( ) ( , )[ ( ) ( )]i ic t P t T F N d KN d
20
1
20
2
ln( / ) ( ) / 2,
ln( / ) ( ) / 2,
i i
i i
i i
i i
F K T td
T t
F K T td
T t
Table 2 Volatility is 10% and stage for every period is 50
47
Maturity Black Lattice Difference Relative Difference (%)1 0.0018085085 0.0018099405 0.0000014320 0.0791823392 2 0.0024348117 0.0024397802 0.0000049685 0.2040608652 3 0.0028388399 0.0028434461 0.0000046061 0.1622542164 4 0.0031206153 0.0031230795 0.0000024643 0.0789673074 5 0.0033214311 0.0033207434 -0.0000006878 0.0207066243 6 0.0034637453 0.0034620815 -0.0000016638 0.0480341136 7 0.0035616356 0.0035626664 0.0000010308 0.0289416315 8 0.0036247299 0.0036267433 0.0000020134 0.0555455781 9 0.0036600091 0.0036617883 0.0000017792 0.0486107386
10 0.0036727489 0.0036734197 0.0000006708 0.0182649876
RMSE 0.0000025690 1. Caplet assume δ = 1 and stage 502. Assume volatility is 10%, the forward curve is flat 5%
Numerical Method of Caplet•We use the payoff function to compute the price in
lattice method•To get the payoff function at time , we have to know
the evolution of the forward rate at time •We construct the binomial tree of and
known the , r = 0, 1, …, i.•Calculating the expectation of the payoff at time and
then multiple the ZCB of to get the caplet value at time t.
48
1iT
1(0; , )i if T T
iT1(0; , )i if T T
1( ( ; , ) )i i i rf T T T K
1iT
1( , )iP t T