You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254931816

The Volatility of Interest Rates and Forward Rates in the Hull White Model

Article  in  SSRN Electronic Journal · September 2012


DOI: 10.2139/ssrn.2159308

CITATION READS

1 640

1 author:

Joachim Paulusch
R+V Lebensversicherung AG
16 PUBLICATIONS   7 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Joachim Paulusch on 12 July 2021.

The user has requested enhancement of the downloaded file.


The volatility of interest rates
and forward rates
in the Hull White Model
Joachim Paulusch∗
July 2, 2014

The interest rate model by Hull and White allows to calculate an explicit
formula for the prices of zero bonds. From this pricing formula we deduce
explicit formulas for
• the volatility of the instantaneous forward rate,
• the volatility of the interest rate (both the spot rate and interest rates of
any maturity), and
• the volatility of the forward rate.

Acknowledgment. Sections 1 to 3 are based on joint work with Katharina Thomas [7].

Contents
1 Definition of the model and notations 2

2 Basic formulas for the spot rate 3

3 General pricing formula 5

4 Volatility of interest rates and forward rates 7


Joachim Paulusch
R+V Lebensversicherung AG
Raiffeisenplatz 2
65189 Wiesbaden
joachim.paulusch@ruv.de

Electronic copy available at: http://ssrn.com/abstract=2159308


1 Definition of the model and notations
We consider the interest rate model by Hull and White [4], a stochastic differential
equation for the spot rate r(t):

dr(t) = (α(t) − βr(t)) dt + σ dW (t). (1)

The mean reversion β and the volatility σ are positive parameters and α(t) is a time-
dependent parameter. Note that the deterministic part of (1) tends always towards
α(t)/β: r(t) is supposed to rise if r(t) < α(t)/β and supposed to fall if r(t) > α(t)/β.
The greater the mean reversion parameter β, the greater the attraction to α(t)/β.
Using the cash roll-up defined by (1), i.e.
Z t 
B(t) = exp r(u) du (2)
0

as the numeraire, we want to choose the parameters of the model in such a way that the
probability measure defined by (1) is the risk neutral, i.e. martingale measure which can
be used for pricing. In other words, denoting the price of a zero bond at time t with
maturity T > t (and time to maturity T − t) with P (t, T ) we define (cf. [2], formula
(3.2))
  Z T  

P (t, T ) = E exp − r(u) du Ft . (3)
t

The instantaneous forward rate at time t for the maturity T > t is given by


f (t, T ) = − ln P (t, T ) (4)
∂T
and the interest rate at time t with maturity T > t is

ln P (t, T )
R(t, T ) = − . (5)
T −t

Conversely, the zero bond prices can be derived by


 Z T 
P (t, T ) = exp − f (t, u) du = exp (−(T − t)R(t, T )) . (6)
t

The forward rate at time t with forward date T1 and maturity date T2 is

ln P (t, T2 ) − ln P (t, T1 )
f (t, T1 , T2 ) = − . (7)
T2 − T1

For a more detailed discussion of the interest rate market see [1], chapter 5.1.

Electronic copy available at: http://ssrn.com/abstract=2159308


2 Basic formulas for the spot rate
The proofs in sections 2 and 3 follow Rogers [8]. See [8] for an even more general
account than ours with time-varying coefficients β(t) and σ(t).
Theorem 2.1. The solution of (1) is given by
 Z t 
−βt βu
r(t) = e r(0) + e [α(u) du + σ dW (u)] . (8)
0

More general we have for any 0 < t < T


 Z T 
−βT βt βu
r(T ) = e e r(t) + e [α(u) du + σ dW (u)] . (9)
t

The proof of the theorem is done by checking that r(t) defined by (8) indeed fulfills
(1) and (9). It may be noted that an Ansatz for the solution is to consider the function
eβt r(t) and to assume that r(t) is a solution. This leads to

d eβt r(t) = βeβt r(t)dt + eβt dr(t)




= βeβt r(t)dt + eβt [(α(t) − βr(t)) dt + σ dW (t)]


= eβt (α(t)dt + σ dW (t)) .

This implies Z t
βt
e r(t) = r(0) + eβu [α(u) du + σ dW (u)] .
0

Corollary 2.2 (Expected spot rate). The expected spot rate is


 Z t 
−βt βu
E[r(t)] = e r(0) + e α(u) du . (10)
0

Proof. It is a property of Ito integrals (cf. [6], Definition 3.1.4 and Theorem 3.2.1) that
Z t 
βu
E e σ dW (u) = 0. (11)
0

Corollary 2.3 (Volatility of the spot rate). The variance of the spot rate is
σ2
1 − e−2βt .

Var(r(t)) = (12)

The volatility of the spot rate is the square root of the variance (cf. remarks 2.4 and
2.5):
σ p
Volatility (r(t)) = √ 1 − e−2βt (13)

Electronic copy available at: http://ssrn.com/abstract=2159308


Proof. By Ito isometry (cf. [6], Corollary 3.1.7) we have
" Z 2 # Z t
t
E eβu σ dW (u) = e2βu σ 2 du, (14)
0 0

so
" 2 #
Z t Z t
2 −2βt βu −2βt
e2βu σ 2 du

E (r(t) − E[r(t)]) =E e e σ dW (u) =e
0 0

σ 2 −2βt 2βt 
= e e −1 .

Remark 2.4 (General remark on volatilities). Generally speaking, volatility is the stan-
dard deviation of a market invariant, i.e. a time homogeneous, observable market vari-
able (cf. [5], sections 3.1 and 3.2). The market invariant for the cash account is the
change in the spot rate. The market invariants for zero bonds are the changes in the
interest rates of the respective time to maturity. Consequently one should write
σ p
Volatility (r(t0 + t) − r(t0 )) = √ 1 − e−2βt (t, t0 ≥ 0). (15)

Without loss of generality we set t0 = 0 and note that
Standard deviation (r(t) − r(0)) = Standard deviation (r(t)) , (16)
such that we end up with (13).
Note that the volatility (13) is proportional to the volatility parameter σ and essentially
inversely proportional to the square root of the mean reversion parameter β. The mean
reversion parameter again governs the time scaling of the volatility, which we will point
out in the following remark for further reference.
Remark 2.5 (Time scaling of volatility).√Due to the mean reverting property of our
model, the time scaling is not given by t as one is used to from the equity world.
Corollary 2.3 rather shows that
p
Volatility(r(t)) ∼ 1 − e−2βt . (17)
Hence, if the volatility is derived with respect to time steps of length δt, the factor to
rescale the volatility to the yearly basis is
r
1 − e−2β
. (18)
1 − e−2βδt
This particularly implies that the volatility does not go to infinity with ever increasing t
but tends to a finite value:
σ
lim Volatility(r(t)) = √ . (19)
t→∞ 2β

4
3 General pricing formula
By “general” we mean that the pricing formula is valid for arbitrary parameters α(t).
The general pricing formula is an explicit solution for the bond price (3), namely a
solution of   Z T  

P (t, T ) = E exp − r(u) du Ft .
t

This is an auxiliary result in order to derive the market consistent calibration, i.e. a
choice of α(t) such that (3) holds with the actual market prices. – In view of (3) we
consider the random variable
Z T
Z(t, T ) = r(u) du. (20)
t

Lemma 3.1. Z(t, T ) has a normal distribution with


Z T Z u
−βu
E[Z(t, T )] = B(t, T )r(t) + e eβs α(s) ds du, (21)
t t

and Z T
Z(t, T ) = E[Z(t, T )] + σ B(s, T ) dW (s), (22)
t

with
1
1 − e−β(T −t) .

B(t, T ) = B(T − t) = (23)
β
Proof. We derive with Theorem 2.1 and (11)
Z T  Z T  Z u  
−βu βt βs
E r(u) du = E e e r(t) + e (α(s) ds + σ dW (s)) du
t t t
Z T  Z u 
−βu βt βs
= e e r(t) + E e (α(s) ds + σ dW (s)) du
t t
Z T  Z u 
−βu βt βs
= e e r(t) + e α(s) ds du,
t t

and T
Z T
−βu 1 −βu 1 −βt
− e−βT .

e du = − e = e
t β u=t β
This shows (21). In order to prove (22) we use integration by parts for Wiener integrals
(cf. [3] Lemma 2.24), namely
Z T Z T Z T Z u
0
g(s)f (s) dW (s) = g(T ) f (s) dW (s) − g (u) f (s) dW (s) du.
t t t t

5
This means
Z T u
1 −βT T βs
Z Z Z T
−βu βs σ
e e σ dW (s) du = − e e σ dW (s) + dW (s)
t t β t t β
Z T
=σ B(s, T ) dW (s).
t

The Wiener integral on the right hand side has a normal distribution (see [3] Theorem
2.19), so Z(t, T ) has a normal distribution as well.
Lemma 3.2. We have
σ2
 i
1 h −β(T −t) 2

Var(Z(t, T )) = 2 T − t − 2−e −1 (24)
β 2β
σ2 σ2
= 2 [T − t − B(t, T )] − B(t, T )2 . (25)
β 2β
Proof. We use Lemma 3.1 to calculate
" Z
T 2 #
Var(Z(t, T )) = E (Z(t, T ) − E[Z(t, T )])2 = E
 
σ B(s, t) dW (s)
t
Z T 2 Z T
σ 2
= σ2 B 2 (s, T ) ds = 1 − e−β(T −t) ds
t β2 t
T
σ2
Z
= 1 + e−2β(s−t) − 2e−β(s−t) ds
β2 t
σ2
 
1 −2β(T −t) 1 2 2
= T −t− e + + e−β(T −t) −
β2 2β 2β β β
σ2
 
1 h 2 i
= T −t− 2 − e−β(T −t) − 1
β2 2β
σ2
 
1  2 
= T −t− {1 + βB(t, T )} − 1
β2 2β
σ2
 
1  2 2

= T −t− 2βB(t, T ) + β B(t, T )
β2 2β
σ2 σ2
= [T − t − B(t, T )] − B(t, T )2 .
β2 2β

Theorem 3.3 (General pricing formula). In the Hull White model


P (t, T ) = e−A(t,T )−r(t)B(t,T ) (26)
holds with B(t, T ) given by (23) and
Z TZ u
σ2 σ2
A(t, T ) = α(s)e−β(u−s) ds du − 2 [T − t − B(t, T )] + B(t, T )2 (27)
t t 2β 4β

6
Proof. By (3) and (20)
P (t, T ) = E e−Z(t,T ) .
 

Z(t, T ) has a normal distribution by Lemma 3.1. So we have (cf. e.g. [1] page 72)
 
 −Z(t,T )  1
E e = exp − E[Z(t, T )] + Var(Z(t, T )) .
2

A closer look at (21) and (25) finishes the proof.

4 Volatility of interest rates and forward rates


The price of P (t, T ) is a conditional expectation, conditional on the spot rate as of time
t, to be precise (because of (26)). So P (t, T ), as seen from any time before t, is a random
variable. Hence the instantaneous forward rate f (t, T ) at time t for a maturity T > t is,
as seen from time zero, a random variable as is the spot rate r(t).

Theorem 4.1 (Volatility of the instantaneous forward rate). The variance of the instan-
taneous forward rate is

σ 2 −2β(T −t)
1 − e−2βt .

Var(f (t, T )) = e (28)

The volatility of the instantaneous forward rate is the square root of the variance (cf.
remarks 2.4 and 2.5):
σ p
Volatility(f (t, T )) = √ e−β(T −t) 1 − e−2βt . (29)

Note that
Volatility(f (t, T )) = e−β(T −t) Volatility(r(t)) (30)

Proof. By definition of the instantaneous forward rate (4) and the general pricing for-
mula (theorem 3.3), we have

f (t, T ) = ∂T A(t, T ) + r(t)∂T B(t, T ).

A and B being deterministic in terms of the model parameters, this implies

Var(f (t, T )) = (∂T B(t, T ))2 Var(r(t)).

Now, because of (23)


 
1
1 − e−β(T −t) = e−β(T −t)

∂T B(t, T ) = ∂T
β

7
holds, and the variance of the spot rate is given by Corollary 2.3, namely

σ2
1 − e−2βt .

Var(r(t)) =

Altogether we get

σ 2 −2β(T −t)
Var(f (t, T )) = (∂T B(t, T ))2 Var(r(t)) = 1 − e−2βt

e

as was to be shown.

The interest rate R(t, T ) at time t with maturity T > t is, as seen from time zero, a
random variable as are the instantaneous forward rate f (t, T ) and the spot rate r(t).

Theorem 4.2 (Volatility of the interest rate). The variance of the interest rate is
2
1 − e−β(T −t) σ2

1 − e−2βt .

Var(R(t, T )) = (31)
β(T − t) 2β

The volatility of the interest rate is the square root of the variance (cf. remarks 2.4 and
2.5):
1 − e−β(T −t) σ p
Volatility(R(t, T )) = √ 1 − e−2βt . (32)
β(T − t) 2β
Note that

1 − e−β(T −t)
Volatility(R(t, T )) = Volatility(r(t)) (33)
β(T − t)

Proof. By definition of the interest rate (5) and theorem 3.3, we have

ln P (t, T ) A(t, T ) + r(t)B(t, T )


R(t, T ) = − = .
T −t T −t
Using corollary 2.3 and (23), this implies
2 2
1 − e−β(T −t) σ2
 
B(t, T )
1 − e−2βt .

Var(R(t, T )) = Var(r(t)) =
T −t β(T − t) 2β

Note that the time scaling in t, with fixed time to maturity T − t, of both the volatility
of the instantaneous forward rate and the volatility of the interest rate is the same as the
time scaling of the volatility of the spot rate (cf. remark 2.5).

8
Moreover, note that the volatility of the interest rate coincides with the volatility of
the spot rate except for the factor

1 − e−β(T −t) X (−1)k β k (T − t)k β(T − t)
= =1− + ... (34)
β(T − t) k=0
(k + 1)! 2

This implies that the mean reversion parameter β accounts for the magnitude of the
decrease in volatility for greater times to maturity.
Finally, the forward rate f (t, T1 , T2 ) at time t with forward date T1 and maturity date
T2 is, as seen from time zero, a random variable as are the other quantities considered.
Theorem 4.3 (Volatility of the forward rate). The variance of the forward rate is
2 2
1 − e−β(T2 −T1 )

−2β(T1 −t) σ
1 − e−2βt .

Var(f (t, T1 , T2 )) = e (35)
β(T2 − T1 ) 2β
The volatility of the forward rate is the square root of the variance (cf. remarks 2.4 and
2.5):
1 − e−β(T2 −T1 ) σ p
Volatility(f (t, T1 , T2 )) = e−β(T1 −t) √ 1 − e−2βt . (36)
β(T2 − T1 ) 2β
Note that

1 − e−β(T2 −T1 )
Volatility(f (t, T1 , T2 )) = e−β(T1 −t) Volatility(r(t)) (37)
β(T2 − T1 )

Proof. Again, by definition (7) and Theorem 3.3


ln P (t, T2 ) − ln P (t, T1 )
f (t, T1 , T2 ) = −
T2 − T1
1
= {A(t, T2 ) − A(t, T1 ) + r(t) [B(t, T2 ) − B(t, T1 )]} .
T2 − T1
This means
 2
B(t, T2 ) − B(t, T1 )
Var(f (t, T1 , T2 )) = Var(r(t)). (38)
T2 − T1
The difference between B(t, T2 ) and B(t, T1 ) is, again by (23),

1 −β(T2 −t) −β(T1 −t)


 e−β(T1 −t) − e−β(T2 −t)
B(t, T2 ) − B(t, T1 ) = 1−e − 1−e =
β β
−β(T2 −T1 )
1−e
= e−β(T1 −t) .
β
The assertion now follows with (12) and (38).

9
References
[1] Martin Baxter and Andrew Rennie: Financial Calculus. Cambridge University
Press, 2nd edition (1997)

[2] Damiano Brigo and Fabio Mercurio: Interest Rate Models. Theory and Practice.
Springer Finance (2001)

[3] Thomas Deck: Der Itô-Kalkül. Springer (2006)

[4] John Hull and Alan White: Pricing interest-rate derivative securities. Review of Fi-
nancial Studies 3, pages 573-592 (1990)

[5] Attilio Meucci: Risk and Asset Allocation. Springer Finance, corrected 3rd reprint
(2007)

[6] Bernt Øksendal: Stochastic Differential Equations. Springer, corrected reprint of


the 6th edition (2007)

[7] Joachim Paulusch and Katharina Thomas: Das Hull White Modell. R+V Lebensver-
sicherung AG, working paper (2010)

[8] Chris Rogers: Which model for the term-structure of interest rates should one use?
In: Darrell Duffie and Steven E. Shreve (editors): Proceedings of the IMA Workshop
on Mathematical Finance. Springer, pages 93-116 (1995)

10

View publication stats

You might also like