Interest Rates and FX Models
Interest Rates and FX Models
Andrew Lesniewski
Courant Institute of Mathematical Sciences
New York University
New York
Contents
1 Term structure modeling 2
(a) Short rate models, in which the stochastic state variable is taken to be the
instantaneous forward rate. Historically, these were the earliest success-
ful term structure models. We shall focus on a tractable Gaussian model,
namely Vasiceks model and its descendents.
(b) LIBOR market model, in which the stochastic state variable is the entire
forward curve represented and as a collection of benchmark LIBOR forward
rates. These, more recently developed, models are descendants of the HJM
model and have been popular among practitioners.
Short Rate Models 3
Short rates models use the instantaneous spot rate r(t) as the basic state vari-
able. In the LIBOR / OIS framework, the short rate is defined as r(t) = f (t, t),
where f (t, s) denotes the instantaneous discount (OIS) rate, as explained in Lec-
ture 1. The stochastic differential equation describing the dynamics of r(t) is
usually stated under the spot measure, and has the form
where A and B are suitably chosen drift and diffusion coefficients, and W is
the Brownian motion driving the process. Models of this type are referred to as
one-factor models, as there is only one stochastic drivers; models with multiple
stochastic drivers are called multi-factor models.
Various choices of the coefficients A and B lead to different dynamics of the
instantaneous rate. You should consult the literature cited at the end of these notes
for a complete catalog of choices available in the repertoire. We shall focus on the
Vasicek model and its descendent, the Hull-White model.
r (0) = r0 . (3)
Originally, this process has been studied in the physics literature, and is known as
the Ornstein - Uhlenbeck process.
A special feature of Vasiceks model is that the stochastic differential equa-
tion (2) has a closed form solution. In order to find it we utilize the method of
variations of constants. The homogeneous equation
dr(t) = r(t)dt
d(t) = et dt + et dW (t).
Consequently, t
t
(t) = e + es dW (s).
0
The solution to our problem is the sum of the solution (4) with C = r0 (in
order to enforce the initial condition) and the particular solution r1 (t):
t
t
( t
)
r(t) = r0 e + 1 e + e(ts) dW (s). (5)
0
To understand better the meaning of this solution, we note that the expected
value of the instantaneous rate r(t) is
( )
EQ0 [r(t)] = X0 et + 1 et , (6)
(c) The model is one-factor, meaning that there is only one stochastic driver of
the process.
(d)) With non-zero probability, rates may become negative (typically, this prob-
ability is fairly low).
Some of these shortcomings can be easily overcome by means of a slight extension
of the model.
and thus
where r0 (0) = r (0) = r0 . This process is called the extended Vasicek (or Hull-
White) model. From (12),
t
(ts)
r(t) = r0 (t) + e (r(s) r0 (s)) + e(tu) (u)dW (u), (14)
s
and so the instantaneous rate is represented as a contribution from the current yield
curve plus a random perturbation. This representation of r(t) implies that
(ts)
EQ
s [r(t)] = r0 (t) + e
0
(r(s) r0 (s)) . (15)
where b(t) is the basis between the instantaneous LIBOR and OIS rates. For
simplicity of exposition we assume that the basis curve is given by a deterministic
function rather than a stochastic process.
In other words, r(t) = r0 (t) + r1 (t) + r2 (t). A natural interpretation of these vari-
ables is that r1 (t) controls the levels of the rates, while r2 (t) controls the steepness
of the forward curve.
We assume the stochastic dynamics:
where 1 (t) and 2 (t) are the instantaneous volatilities of the state variables r1 (t)
and r2 (t), respectively. The two Brownian motions are correlated,
where
1 et
h (t) = .
Integrating by parts we can transform the double integral in the exponent into a
single integral
T u T
(us)
e (s) dW (s) du = h (T s) (s)dW (s).
t t t
we obtain the following expression for the price of a zero coupon bond:
where T T
h (T s)2 (s)2 ds
A (t, T ) = e t r0 (u)du+h (T t)r0 (t)+ 21 t . (23)
Note in particular that the discount factor P0 (0, T ) has the form
P0 (0, T ) = P (0, T )
T T (24)
h (T s)2 (s)2 ds
= e 0 r0 (s)ds+ 12 0 .
The generalization of formula of (22) to the case of the two-factor Hull-White
model reads:
where now
T
A (t, T ) = e t r0 (u)du
T 2 2
(26)
e 2 t (h1 (T s) 1 (s) +2h1 (T s)h2 (T s)1 (s)2 (s)+h2 (T s) 2 (s) )ds .
1 2 2
where
1 P0 (0, Tmat )
d = log , (28)
P0 (0, T ) K 2
with ( )1/2
T
2(T s)
= e 2
(s) ds h (Tmat T ) . (29)
0
Short Rate Models 9
Since floorlets and caplets can be thought of as calls and puts on FRAs, these
formulas can be used as building blocks for valuation of caps and floors in the
Hull-White model.
(a) Matching the volatilities of selected options. This is a bit more difficult,
and we proceed as follows. We choose the instantaneous volatility function
(t) to be locally constant. That means that we divide up the time axis into,
say, 3 month period [Tj , Tj+1 ) and set (t) = j , for t [Tj , Tj+1 ). Now,
we select sufficiently many calibrating instruments, so that their number
exceeds the number of the s. Next, we optimize the choice of s and ,
requiring that the suitable sum of pricing errors is minimal.
Despite the simple structure of the Hull-White model, most instruments cannot
be priced by means of closed form expressions such as those for caps and floors of
the previous section. One has to resort numerical techniques. Among them, two
are particularly important:
Time does not permit us to discuss these numerical techniques, and I defer you to
literature cited at the end of these notes.
10 Interest Rates & FX Models
under the T2 -forward measure QT2 . Here B0 (T1 , T2 ) denotes the credit spread
between LIBOR and OIS. This is, indeed, almost the definition of the T2 -forward
measure! Consequently, L0 (T1 , T2 ) is given by:
( )
1 1
L0 (T1 , T2 ) = 1 + B0 (T1 , T2 )
P0 (T1 , T2 )
( ) (32)
1 P0 (0, T1 )
= 1 + B0 (T1 , T2 ).
P0 (0, T2 )
It is easy to calculate this rate within the Hull-White model. Let us first consider
the one-factor case. Using (24), we find that
L0 (T1 , T2 )
1 ( TT2 r0 (s)ds 12 (0T2 h (T2 s)2 (s)2 ds0T1 h (T1 s)2 (s)2 ds) )
= e 1 1 (33)
+ B0 (T1 , T2 ).
On the other hand, the rate Lfut0 (T1 , T2 ) implied from the Eurodollar futures
contract is given by the expected value of (31) under the spot measure Q0 , namely
1 ( Q0 [ TT2 r(t)dt ] )
Lfut
0 (T1 , T2 ) = E e 1 1 + B0 (T1 , T2 ).
We have explained this fact in Lecture 4, attributing it to the practice of daily1
margin account adjustments by the Exchange. In order to calculate this expected
1
which we model as continuous
Short Rate Models 11
value we proceed as in the calculation leading to the explicit formula for P (t, T ):
[ T2 ]
EQ0 e T1 r(t)dt
[ T2 t (ts) ]
Q0 T1 (r0 (t)+ 0 e (s)dW (s))dt
=E e
T2 [ T2 T1 ]
= e T1 r0 (t)dt EQ0 e 0 h (T2 s)(s)dW (s) 0 h (T1 s)(s)dW (s)
T2 [ T1 T2 ]
T1 r0 (t)dt Q0 0 (h (T2 s)h (T1 s))(s)dW (s)+ T1 h (T2 s)(s)dW (s)
=e E e
T2 T T
r0 (t)dt+ 21 ( 0 1 (h (T2 s)h (T1 s))2 (s)2 ds+ T 2 h (T2 s)2 (s)2 ds)
=e T1 1 ,
and so
Lfut
0 (T1 , T2 )
1 ( TT2 r0 (t)dt+ 21 (0T1 (h (T2 s)h (T1 s))2 (s)2 ds+TT2 h (T2 s)2 (s)2 ds) )
= e 1 1 1
1( )
= L0 (T1 , T2 ) + 1 + F0 (T1 , T2 )
( )
T2 2 T1
e 0 h (T2 s) (s) ds 0 h (T2 s)h (T1 s)(s) ds 1 ,
2 2
where F0 (T1 , T2 ) = L0 (T1 , T2 ) B0 (T1 , T2 ) is the forward rate calculated off the
OIS curve.
Consequently, the Eurodollar / FRA convexity adjustment is given by
1( )
ED / FRA (T1 , T2 ) = 1 + F0 (T1 , T2 )
( ) (34)
T2 T1
e 0 h (T2 s) (s) ds 0 h (T2 s)h (T1 s)(s) ds 1 .
2 2 2
In the case of a constant instantaneous volatility, (t) = , the last integral can be
evaluated in closed form, and the result is:
2 (( )( )2
ED / FRA 3 1 e2T1 1 e(T2 T1 )
2
( )( )2 ) (36)
+ 1 e(T2 T1 ) 1 eT1 .
where 11 = 22 = 1, 12 = 21 = .
Note that for any real value , h (s) is non-negative and monotone increasing.
Therefore, the convexity adjustments implied by the Hull-White model are always
positive (as they should be!).
References
[1] Andersen, L., and Piterbarg, V.: Interest Rate Modeling, Vol. 2, Atlantic
Financial Press (2010).
[2] Brigo, D., and Mercurio, F.: Interest Rate Models - Theory and Practice,
Springer Verlag (2006).
[3] Hull, J.: Options, Futures and Other Derivatives Prentice Hall (2005).