Technical

Notes
Options, Futures, and Other Derivatives

JOHN C. HULL
 Technical Note No. 1*
Options, Futures, and Other Derivatives
John Hull

Convexity Adjustments to Eurodollar Futures

In the Ho-Lee model the risk-neutral process for the short rate in the traditional
risk-neutral world is
dr = θ(t)dt + σ dz
where r is the instantaneous short rate, θ is a function of time, a and σ are constants, and
dz is a Wiener process. Define P (t, T ) as the price of a bond paying $1 at time T as seen
at time t. As explained in the text, the bond price has the form P (t, T ) = A(t, T )e−r(T −t) .
From Itô’s lemma the process for the bond price in a traditional risk-neutral world is
dP (t, T ) = r(t)P (t, T )dt − (T − t)σP (t, T ) dz
Define f (t, T1 , T2 ) as the forward rate (continuously compounded) at time t for the
period between T1 and T2 .

ln[P (t, T1 )] − ln[P (t, T2 )]

f (t, T1 , T2 ) =
T2 − T1
From Ito’s lemma the process followed by f (t, T1 , T2 ) is

σ 2 (T2 − t)2 − σ 2 (T1 − t)2

df (t, T1 , T2 ) = dt + σ dz
2(T2 − T1 )

The expected change in the forward rate between time zero and time T1 is determined by
integrating the coefficient of dt between 0 and T1 . It is σ 2 T1 T2 /2.
The forward rate equals the spot rate at time T1 . The expected value of the forward
rate at time T1 is therefore the expected value of the spot rate at time T1 . It follows that
the forward rate at time zero equals the expected spot rate minus σ 2 T1 T2 /2.
Because we are in the traditional risk-neutral world the expected value of the spot rate
is the same as the futures rate. The futures rate is therefore greater than the forward rate
by σ 2 T1 T2 /2 when both are expressed with continuous compounding. (There is a small
approximation here in that we know that the futures rate with quarterly compounding
equals the expected future spot rate with quarterly compounding. We assume that the
same is true when both both rates are converted continuous compounding.)
When converting the futures rate to the forward rate we should therefore subtract
σ 2 T1 T2 /2 from the futures rate. This is known as a convexity adjustment. As explained
in the text, there are actually two parts to the convexity adjustment:
1. The difference between a futures contract that is settled daily and a similar contract
that is settled entirely at time T1
2. The difference between the contract that is settled at time T1 and a similar contract
that is settled at time T2
To prove a corresponding result for the Hull-White model
dr = [θ(t) − ar] dt + σ dz

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
we proceed similarly. The process for the bond price is

dP (t, T ) = r(t)P (t, T )dt − B(t, T )σP (t, T ) dz

where
1 − e−a(T −t)
B(t, T ) =
a
The result after crunching through similar but more involved math is that the convexity
adjustment is
B(T1 , T2 ) σ2
[B(T1 , T2 )(1 − e−2aT1 ) + 2aB(0, T1 )2 ]
T2 − T1 4a
This reduces to the earlier result when we take the limit as a tends to zero.

2
 Technical Note No. 2*
Options, Futures, and Other Derivatives
John Hull
Properties of Lognormal Distribution
A variable V has a lognormal distribution if X = ln(V ) has a normal distribution.
Suppose that X is φ(m, s2 ); that is, it has a normal distribution with mean m and standard
deviation, s. The probability density function for X is
(X − m)2
 
1
√ exp −
2πs 2s2
The probability density function for V is therefore
[ln(V ) − m]2
 
1
h(V ) = √ exp −
2πsV 2s2
Consider the nth moment of V Z +∞
V n h(V )dV
0
Substituting V = exp X this is
Z +∞
(X − m)2
 
exp(nX)
√ exp − dX
−∞ 2πs 2s2
Z +∞
(X − m − ns2 )2 2mns2 + n2 s4
   
1
= √ exp − exp dX
−∞ 2πs 2s2 2s2
Z +∞
(X − m − ns2 )2
 
2 2 1
= exp(nm + n s /2) √ exp − dX
−∞ 2πs 2s2
The integral in this expression is the integral of a normal density function with mean
m + ns2 and standard deviation s and is therefore 1.0. It follows that
Z +∞
V n h(V )dV = exp(nm + n2 s2 /2) (1)
0
The expected value of V is given when n = 1. It is
exp(m + s2 /2)
The formula for the mean of a stock √ price at time T in the text is given by setting
2
m = ln(S0 ) + (µ − σ /2)T and s = σ T
The variance of V is E(V 2 ) − [E(V )]2 . Setting n = 2 in equation (1) we get
E(V 2 ) = exp(2m + 2s2 )
The variance of V is therefore
exp(2m + 2s2 ) − exp(2m + s2 ) = exp(2m + s2 )[exp(s2 ) − 1]
The formula for the variance of a stock
√ price at time T in the text is given by setting
2
m = ln(S0 ) + (µ − σ /2)T and s = σ T .
* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 Technical Note No. 3*
Options, Futures, and Other Derivatives
John Hull

Warrant Valuation When Value of Equity Plus Warrants Is Lognormal

This note describes how warrants (or other stock options) can be valued if there is
a single warrant issue and the value of the warrants plus the equity is assumed to be
lognormal.1 It should be emphasized that the result presented here is of theoretical rather
than practical interest. In practice a company usually has many warrant (and executive
stock option) issues outstanding. There are in practice many reasons why the equity of a
firm is not lognormal and the existence of warrants is probably not the most important
one.
It is important to realize that in practice warrants and executive stock options are
valued in exactly the same way as other over-the-counter and exchange-traded options on
the company’s stock. The issue of the stock price distribution at warrant maturity should
not be confused with the dilution issue. (As explained in the text dilution takes place
when a warrant or executive stock option issue is first announced.)
Consider a company with N outstanding shares and M outstanding European options.
Suppose that each option entitles the holder to purchase a share from the company at time
T at a price of K per share.
Define V is the value of the company’s equity (including the warrants) and VT as the
value of V at time T . If the warrant holders exercise, the company receives a cash inflow
from the payment of the exercise price of M K and the value of the company’s equity
increases to VT + M K. This value is distributed among N + M shares so that the share
price immediately after exercise becomes

VT + M K
N +M

The payoff to the warrant holder if the warrant is exercised is, therefore,
 
VT + M K
−K
N +M
or  
N VT
−K
N +M N
The warrants should be exercised only if this payoff is positive. The payoff to the warrant
holder is, therefore,  
N VT
max − K, 0
N +M N

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
See D. Galai and M. Schneller, “Pricing Warrants and the Value of the Firm,” Journal
of Finance, 33 (1978), 1339–42; B. Lauterbach and P. Schultz, “Pricing Warrants: An
Empirical Study of the Black–Scholes Model and Its Alternatives,” Journal of Finance, 45
(1990), 1181–1209.

1
This shows that the value of the warrant is the value of
N
N +M

regular call options on V .

The value of V at time zero is given by

V0 = N S 0 + M W

where S0 is the stock price at time zero and W is the warrant price at that time, so that

V0 M
= S0 + W
N N

The Black–Scholes formula in equation (15.20) of the book, therefore, gives the warrant
price W if
1. The stock price S0 is replaced by S0 + (M/N )W .
2. The volatility σ is the volatility of the shares plus the warrants, not just the shares).
3. The formula is multiplied by N/(N + M ).
When these adjustments are made we end up with a formula for W as a function of W .
This can be solved numerically.

2
 Technical Note No. 4*
Options, Futures, and Other Derivatives
John Hull

Exact Procedure for Valuing

American Calls on Dividend-Paying Stocks

The Roll, Geske, and Whaley formula for the value of an American call option on a
stock paying a single dividend D1 at time t1 is
r !
t1
C =(S0 − D1 e−rt1 )N (b1 ) + (S0 − D1 e−rt1 )M a1 , −b1 ; −
T
r !
t1
− Ke−rT M a2 , −b2 ; − − (K − D1 )e−rt1 N (b2 ) (1)
T

where
ln[(S0 − D1 e−rt1 )/K] + (r + σ 2 /2)T
a1 = √
σ T
√
a2 = a1 − σ T
ln[(S0 − D1 e−rt1 )/S ∗ ] + (r + σ 2 /2)t1
b1 = √
σ t1
√
b 2 = b 1 − σ t1
The variable σ is the volatility of the stock price net of the present value of the dividend.
The function, M (a, b; ρ), is the cumulative probability, in a standardized bivariate normal
distribution, that the first variable is less than a and the second variable is less than b,
when the coefficient of correlation between the variables is ρ. A procedure for calculating
the M function is given in Technical Note 5. The variable S ∗ is the solution to

c(S ∗ ) = S ∗ + D1 − K

where c(S ∗ ) is the Black–Scholes–Merton option price when the stock price is S ∗ and the
time to maturity is T − t1 . When early exercise is never optimal, S ∗ = ∞. In this case
b1 = b2 = −∞ and equation (1) reduces to the Black–Scholes–Merton equation with S0
replaced by S0 − D1 e−rt1 . In other situations, S ∗ < ∞ and the option should be exercised
at time t1 when S(t1 ) > S ∗ + D1 .
When several dividends are anticipated, early exercise is normally optimal only on
the final ex-dividend date as explained in the text. It follows that the Roll, Geske, and
Whaley formula can be used with S0 reduced by the present value of all dividends except
the final one. The variable, D1 , should be set equal to the final dividend and t1 should be
set equal to the final ex-dividend date.

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 Technical Note No. 5*
Options, Futures, and Other Derivatives
John Hull

Calculation of Cumulative Probability in Bivariate Normal Distribution

Define M (a, b; ρ) as the cumulative probability in a standardized bivariate normal

distribution that the first variable is less than a and the second variable is less than b,
when the coefficient of correlation between the variables is ρ. Drezner provides a way of
calculating M (a, b; ρ) to four-decimal-place accuracy.1 If a ≤ 0, b ≤ 0, and ρ ≤ 0,
p 4
1 − ρ2 X
M (a, b; ρ) = Ai Aj f (Bi , Bj )
π i,j=1

where
f (x, y) = exp [a0 (2x − a0 ) + b0 (2y − b0 ) + 2ρ(x − a0 )(y − b0 )]
a b
a0 = p b0 = p
2(1 − ρ2 ) 2(1 − ρ2 )
A1 = 0.3253030 A2 = 0.4211071 A3 = 0.1334425 A4 = 0.006374323
B1 = 0.1337764 B2 = 0.6243247 B3 = 1.3425378 B4 = 2.2626645
In other circumstances where the product of a, b, and ρ is negative or zero, one of the
following identities can be used:

M (a, b; ρ) = N (a) − M (a, −b; −ρ)

M (a, b; ρ) = N (b) − M (−a, b; −ρ)
M (a, b; ρ) = N (a) + N (b) − 1 + M (−a, −b; ρ)

In circumstances where the product of a, b, and ρ is positive, the identity

M (a, b; ρ) = M (a, 0; ρ1 ) + M (b, 0; ρ2 ) − δ
can be used in conjunction with the previous results, where
(ρa − b) sgn (a) (ρb − a) sgn (b)
ρ1 = p ρ2 = p
a2 − 2ρab + b2 a2 − 2ρab + b2
(
1 − sgn (a) sgn (b) +1 when x ≥ 0
δ= sgn (x) =
4 −1 when x < 0

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
Z. Drezner, “Computation of the Bivariate Normal Integral,” Mathematics of Com-
putation, 32 (January 1978), 277–79. Note that the presentation here corrects a typo in
Drezner’s paper.

1
 Technical Note No. 6*
Options, Futures, and Other Derivatives
John Hull

Differential Equation for Price of a Derivative

on a Stock Providing a Known Dividend Yield

Define f as the price of a derivative dependent on a stock that provides a dividend

yield at rate q. We suppose that the stock price, S, follows the process
dS = µS dt + σS dz
where dz is a Wiener process. The variables µ and σ are the expected growth rate in
the stock price and the volatility of the stock price, respectively. Because the stock price
provides a dividend yield, µ is only part of the expected return on the stock.1
Because f is a function of S and t, it follows from Ito’s lemma that

1 ∂2f 2 2
 
∂f ∂f ∂f
df = µS + + 2
σ S dt + σS dz
∂S ∂t 2 ∂S ∂S

Similarly to the procedure described in the text for a non-dividend paying stock, we can
set up a portfolio consisting of

−1 : derivative
∂f
+ : stock
∂S
If Π is the value of the portfolio,
∂f
Π = −f + S (1)
∂S
and the change, ∆Π, in the value of the portfolio in a time period ∆t is as given by:

1 ∂2f 2 2
 
∂f
∆Π = − − σ S ∆t
∂t 2 ∂S 2

In time ∆t the holder of the portfolio earns capital gains equal to ∆Π and dividends on
the stock position equal to
∂f
qS ∆t
∂S
Define ∆W as the change in the wealth of the portfolio holder in time ∆t. It follows that

1 ∂2f 2 2
 
∂f ∂f
∆W = − − σ S + qS ∆t (2)
∂t 2 ∂S 2 ∂S

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
In a risk-neutral world µ = r − q as indicated in the text.

1
Because this expression is independent of the Wiener process, the portfolio is instanta-
neously riskless. Hence
∆W = rΠ ∆t (3)
Substituting from equations (1) and (2) into equation (3) gives

1 ∂2f 2 2
   
∂f ∂f ∂f
− − 2
σ S + qS ∆t = r −f + S ∆t
∂t 2 ∂S ∂S ∂S

so that
∂f ∂f 1 ∂2f
+ (r − q)S + σ 2 S 2 2 = rf
∂t ∂S 2 ∂S
This is the equation in the text.

2
 Technical Note No. 7*
Options, Futures, and Other Derivatives
John Hull

Differential Equation for Price of a Derivative

on a Futures Price

Suppose that the futures price F follows the process

dF = µF dt + σF dz (1)

where dz is a Wiener process and σ is constant.1 Because f is a function of F and t, it

follows from Ito’s lemma that

1 ∂2f 2 2
 
∂f ∂f ∂f
df = µF + + 2
σ F dt + σF dz (2)
∂F ∂t 2 ∂F ∂F

Consider a portfolio consisting of

−1 : derivative
∂f
+ : futures contracts
∂F

Define Π as the value of the portfolio and let ∆Π, ∆f , and ∆F be the change in Π, f , and
F in time ∆t, respectively. Because it costs nothing to enter into a futures contract,

Π = −f (3)

In a time period ∆t, the holder of the portfolio earns capital gains equal to −∆f from the
derivative and income of
∂f
∆F
∂F
from the futures contract. Define ∆W as the total change in wealth of the portfolio holder
in time ∆t. It follows that
∂f
∆W = ∆F − ∆f
∂F
The discrete versions of equations (1) and (2) are

∆F = µF ∆t + σF ∆z
and
1 ∂2f 2 2
 
∂f ∂f ∂f
∆f = µF + + 2
σ F ∆t + σF ∆z
∂F ∂t 2 ∂F ∂F

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
As discussed in the text, the drift is zero in the traditional risk-neutral world.

1
 √
where ∆z =  ∆t and  is a random sample from a standardized normal distribution. It
follows that
1 ∂2f 2 2
 
∂f
∆W = − − σ F ∆t (4)
∂t 2 ∂F 2
This is riskless. Hence it must also be true that

∆W = rΠ ∆t (5)

If we substitute for Π from equation (3), equations (4) and (5) give

1 ∂2f 2 2
 
∂f
− − σ F ∆t = −rf ∆t
∂t 2 ∂F 2

Hence
∂f 1 ∂2f 2 2
+ σ F = rf
∂t 2 ∂F 2
This is the equation in the text.

2
 Technical Note No. 8*
Options, Futures, and Other Derivatives
John Hull

Analytic Approximation for Valuing American Options

Consider an option on a stock providing a dividend yield equal to q. We will denote

the difference between the American and European option price by v. Because both the
American and the European option prices satisfy the Black–Scholes differential equation,
v also does so. Hence,

∂v ∂v 1 ∂2v
+ (r − q)S + σ 2 S 2 2 = rv
∂t ∂S 2 ∂S
For convenience, we define
τ =T −t
h(τ ) = 1 − e−rτ
2r
α= 2
σ
2(r − q)
β=
σ2
We also write, without loss of generality,

v = h(τ )g(S, h)

With appropriate substitutions and variable changes, this gives

∂2g ∂g α ∂g
S2 + βS − g − (1 − h)α =0
∂S 2 ∂S h ∂h
The approximation involves assuming that the final term on the left-hand side is zero,
so that
∂2g ∂g α
S 2 2 + βS − g=0 (1)
∂S ∂S h
The ignored term is generally fairly small. When τ is large, 1 − h is close to zero; when τ
is small, ∂g/∂h is close to zero.
The American call and put prices at time t will be denoted by C(S, t) and P (S, t),
where S is the stock price, and the corresponding European call and put prices will be
denoted by c(S, t) and p(S, t). Equation (1) can be solved using standard techniques. After
boundary conditions have been applied, it is found that

γ
 c(S, t) + A2 SS∗ 2 when S < S ∗
C(S, t) =
S−K when S ≥ S ∗


* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John Hull.

1
The variable S ∗ is the critical price of the stock above which the option should be exercised.
It is estimated by solving the equation
n o S∗
∗ ∗ −q(T −t) ∗
S − K = c(S , t) + 1 − e N [d1 (S )]
γ2

iteratively. For a put option, the valuation formula is


γ
 p(S, t) + A1 SS∗∗ 1 when S > S ∗∗
P (S, t) =
K −S when S ≤ S ∗∗


The variable S ∗∗ is the critical price of the stock below which the option should be exercised.
It is estimated by solving the equation
n o S ∗∗
K − S ∗∗ = p(S ∗∗ , t) − 1 − e−q(T −t) N [−d1 (S ∗∗ )]
γ1

iteratively. The other variables that have been used here are
" r #
4α
γ1 = −(β − 1) − (β − 1)2 + 2
h
" r #
4α
γ2 = −(β − 1) + (β − 1)2 + 2
h
S ∗∗ n
  o
−q(T −t) ∗∗
A1 = − 1−e N [−d1 (S )]
γ1
 ∗n
S o
A2 = 1 − e−q(T −t) N [d1 (S ∗ )]
γ2
ln(S/K) + (r − q + σ 2 /2)(T − t)
d1 (S) = √
σ T −t

Options on stock indices, currencies, and futures contracts are analogous to options on
a stock providing a constant dividend yield. Hence the quadratic approximation approach
can easily be applied to all of these types of options.

2
 Technical Note No. 9*
Options, Futures, and Other Derivatives
John Hull

Generalized Tree Building Procedure

This note describes a general procedure for constructing a trinomial tree for a variable,
x, in the situation where
1. There are nodes at times t1 , t2 ,...tN where ∆i = ti − ti−1 and t0 = 0
2. The standard deviation of x between time ti−1 and ti is si
3. The drift between time ti−1 and ti is mi
Suppose that x0 is the initial value of x. At time ti the√nodes are chosen to be at
x0 + jqi where j is a positive or negative integer and qi = si 3∆i . We will refer to the
node where x = x0 + jqi at time ti as the (i, j) node.
If we are at the (i, j) node, the expected value of x at time ti+1 is x0 +jqi +mi+1 ∆i+1 .
Let the node at time ti+1 that is closest to this expected value be the node (i + 1, k) The
tree is constructed so that we branch from node (i, j) to one of the three nodes (i+1, k −1),
(i + 1, k) and (i + 1, k + 1).
We choose the probabilities on the branches to match the first and second moment.
Define pu , pm , and pd as the probabilities on the upper middle and lower branches. It
follows that
pu + pm + pd = 1
kqi+1 + (pu − pd )qi+1 = jqi + M
k 2 qi+1
2 2
+ 2k(pu − pd )qi+1 2
+ (pu + pd )qi+1 = V + (jqi + M )2
where M = mi+1 ∆i+1 and V = s2i+1 ∆i+1
The solution to these equations is

V α2 + α
pu = 2 +
2qi+1 2

V α2 − α
pd = 2 +
2qi+1 2
V
pm = 1 − 2 − α2
qi+1
where
jqi + M − kqi+1
α=
qi+1

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John Hull.

1
 Technical Note No. 10*
Options, Futures, and Other Derivatives
John Hull

The Cornish–Fisher expansion to estimate VaR

As indicated in the text, αi ’s and βij ’s can be defined so that ∆P for a portfolio
containing options is approximated as
n
X n X
X n
∆P = αi ∆xi + βij ∆xi ∆xj
i=1 i=1 j=1

Define σij as the covariance between variable i and j:

σij = ρij σi σj

It can be shown that when the ∆xi are multivariate normal

X
E(∆P ) = βij σij
i,j

X X
E[(∆P )2 ] = αi αj σij + βij βkl (σij σkl + σik σjl + σil σjk )
i,j i,j,k,l
X X
E[(∆P )3 ] = 3 αi αj βkl (σij σkl + σik σjl + σil σjk ) + βi1 i2 βi3 i4 βi5 i6 Q
i,j,k,l i1 ,i2 ,i3 ,i4 ,i5 ,i6

The variable, Q, consists of fifteen terms of the form σk1 k2 σk3 k4 σk5 k6 where k1 , k2 , k3 , k4 ,
k5 , and k6 are permutations of i1 , i2 , i3 , i4 , i5 , and i6 .
Define µP and σP as the mean and standard deviation of ∆P so that

µP = E(∆P )

σP2 = E[(∆P )2 ] − [E(∆P )]2

The skewness of the probability distribution of ∆P , ξP , is defined as

1 3 E[(∆P )3 ] − 3E[(∆P )2 ]µP + 2µ3P

ξP = E[(∆P − µP ) ] =
σP3 σP3

Using the first three moments of ∆P , the Cornish-Fisher expansion estimates the qth
percentile of the distribution of ∆P as

µP + wq σP

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
where
1
wq = zq + (zq2 − 1)ξP
6
and zq is qth percentile of the standard normal distribution φ(0, 1).

Example
Suppose that for a certain portfolio we calculate µP = −0.2, σP = 2.2, and ξP = −0.4.
If we assume that the probability distribution of ∆P is normal, the first percentile of
the probability distribution of ∆P is

−0.2 − 2.33 × 2.2 = −5.326

In other words, we are 99% certain that

∆P > −5.326

When we use the Cornish-Fisher expansion to adjust for skewness and set q = 0.01,
we obtain
1
wq = −2.33 − (2.332 − 1) × 0.4 = −2.625
6
so that the first percentile of the distribution is

−0.2 − 2.625 × 2.2 = −5.976

Taking account of skewness, therefore, changes the VaR from 5.326 to 5.976.

2
 Technical Note No. 11*
Options, Futures, and Other Derivatives
John Hull

The Manipulation of Credit Transition Matrices

Suppose that A is an N × N matrix of credit rating changes in one year such as those
discussed in the text. The matrix of credit rating changes in m years is Am . This can be
readily calculated using the normal rules for matrix multiplication.
The matrix corresponding to a shorter period than one year, say six months or one
month is more difficult to compute. We first use standard routines to calculate eigenvectors
xi , x2 , . . ., xN and the corresponding eigenvalues λ1 , λ2 , . . ., λN . These have the property
that
Axi = λi xi (1)
Define X as a matrix whose ith column is xi and Λ as a diagonal matrix where the ith
diagonal element is λi . From equation (1)

AX = XΛ

so that
A = XΛX−1
From this it is easy to see that the nth root of A is

XΛ∗ X−1
1/n
where Λ∗ is a diagonal matrix where the ith diagonal element is λi .
Some authors such as Jarrow, Lando, and Turnbull prefer to handle this problem in
terms of what is termed a generator matrix.1 This is a matrix Γ such that the transition
matrix for a short period of time ∆t is 1 + Γ∆t and the transition matrix for longer period
of time, t, is
∞
X (tΓ)k
exp(tΓ) =
k!
k=0

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
See R. A. Jarrow, D. Lando, and S.M. Turnbull, “A Markov model for the term
structure of credit spreads” Review of Financial Studies, 10 (1997), 481–523.

1
 Technical Note No. 12*
Options, Futures, and Other Derivatives
John Hull

The Calculation of the Cumulative Non-Central Chi Square Distribution

We present an algorithm proposed by Ding (1992).1 Suppose that the non-centrality

parameter is v and the number of degrees of freedom is k and we require the cumulative
probability that the variable will be less than z. We define

1  z k/2  z
t0 = exp −
Γ(k/2 + 1) 2 2
z
ti = ti−1
k + 2i
We also define
w0 = u0 = exp(−v/2)
ui−1 v
ui =
2i
wi = wi−1 + ui
The required probability that the variable with the non-central chi square distribution
will be less than z is
X∞
wi ti
i=0

By taking a sufficient number of terms in this series the required accuracy can be obtained.

The Gamma Function

In the above formulas Γ(.) is the gamma function. It has the property that Γ(n) =
(n − 1)! when n is an integer. In general Γ(x + 1) = xΓ(x). The computation of the gamma
function is discussed in Numerical Recipes.2 .
"√ 6
!#
2π X pn
Γ(x) = p0 + (x + 5.5)x+0.5 e−(x+5.5)
x n=1
x+n

where
p0 = 1.000000000190015

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
See C.G. Ding, “Algorithm AS275: Computing the non-central χ2 distribution func-
tion,” Applied Statistics, 41 (1992), 478–82.

2
See W.H. Press, B.P. Flannery, S.A. Teukolsky, and W.T. Vetterling, Numerical
Recipes in C: The Art of Scientific Computing. Cambridge University Press, Cambridge,
1988.

1
 p1 = 76.18009172947146
p2 = −86.50532032941677
p3 = 24.01409824083091
p4 = −1.231739572450155
p5 = 1.208650973866179 × 10−3
p6 = −5.395239384953 × 10−6
To avoid overflow problems it is best to compute ln Γ(x) rather than Γ(x).

2
 Technical Note No. 13 1

Options, Futures, and Other Derivatives

John Hull

Efficient Procedure for Valuing American-Style Lookback Options

A number of researchers have suggested an interesting and instructive approach to valuing lookback
options.2 Consider an American-style lookback put. The initial stock price is 50, the stock price volatility
is 40%, the risk-free interest rate is 10% and the life of the option is 3 months. We suppose that three
steps are used to model the stock price movements.

When exercised, the option provides a payoff equal to the excess of the maximum stock price over the
current stock price. We define G(t) as the maximum stock price achieved up to time t and set

G (t )
Y (t ) 
S (t )

We next use the Cox, Ross, and Rubinstein tree for the stock price to produce a tree for Y. Initially, Y=1
because G = S at time zero. If there is an up movement in S during the first time step, both G and S
increase by a proportional amount u and Y = 1. If there is a down movement in S during the first time
step, G stays the same, so that Y = 1/d = u. Continuing with these types of arguments, we produce the tree
shown in Figure 1 for Y. (Note that in this example t = 0.08333, u = 1.1224, d = 0.8909, a = 1.0084, and
p = 0.5073). The rules defining the geometry of the tree are

1. When Y=1 at time t, it is either u or 1 at time t+t.

2. When Y=u m at time t for m ≥ 1, it is either u m +1 or u m −1 at time t+t.

An up movement in Y corresponds to a down movement in the stock price, and vice versa. The probability
of an up movement in Y is, therefore, always 1−p and the probability of a down movement in Y is always
p.

We use the tree to value the American lookback option in units of the stock price rather than in dollars. In
dollars, the payoff from the option is

SY−S

In stock price units, the payoff from the option, therefore, is

Y−1

1 ©Copyright John Hull. All Rights Reserved. This note may be reproduced for use in conjunction with Options,
Futures, and Other Derivatives by John C. Hull
2 The approach was proposed by Eric Reiner in a lecture at Berkeley. It is also suggested in S. Babb s, ``Binomial

Valuation of Lookback Options,'' Working paper, Midland Global Markets, 1992; and T. H. F. Cheuk and T. C. F.
Vorst, ``Lookback Options and the Observation Frequency: A Binomial Approach,'' Working Paper, Erasmus
University, Rotterdam.
We roll back through the tree in the usual way, valuing a derivative that provides this payoff except that
we adjust for the differences in the stock price (i.e., the unit of measurement) at the nodes. If f ij is the
value of the lookback at the jth node at time i t and Yij is the value of Y at this node, the rollback
procedure gives

 
f ij  max Yij  1, e  rt (1  p) f i 1, j 1d  pfi 1, j 1u 
when j ≥ 1.

Note that f i+1,j+1 is multiplied by d and f i+1,j−1 is multiplied by u in this equation. This takes into account
that the stock price at node (i,j) is the unit of measurement. The stock price at node (i+1,j+1), which is
the unit of measurement for f i+1,j+1 is d times the stock price at node (i,j) and the stock price at node (i+1,j-
1), which is the unit of measurement for f i+1,j−1 , is u times the stock price at node (i,j).

Similarly, when j=0 the rollback procedure gives

 
f ij  max Yij  1, e  rt (1  p) f i 1, j 1d  pfi 1, j u 
The calculations for our example are shown in Figure 1. The tree estimates the value of the option at time
zero (in stock price units) as 0.1094. This means that the dollar value of the option is 0.1094× 50=5.47.

1.4140
0.4140

1.2598 1.2598
0.2598 0.2598

1.1224 1.1224 1.1224

0.1432 0.1224 0.1224

1.0000 1.0000 1.0000 1.0000

0.1094 0.0834 0.0533 0.0000

Figure 1 Procedure for valuing an American-style lookback option.

 Technical Note No. 14*
Options, Futures, and Other Derivatives
John Hull

The Hull–White Two Factor Model

As explained in the text, Hull and White have proposed a model where the risk-neutral
process for the short rate, r, is

df (r) = [θ(t) + u − af (r)] dt + σ1 dz1 (1)

where u has an initial value of zero and follows the process

du = −bu dt + σ2 dz2

As in the case of one-factor no-arbitrage models, the parameter θ(t) is chosen to

make the model consistent with the initial term structure. The stochastic variable u is a
component of the reversion level of r and itself reverts to a level of zero at rate b. The
parameters a, b, σ1 , and σ2 are constants and dz1 and dz2 are Wiener processes with
instantaneous correlation ρ.
This model provides a richer pattern of term structure movements and a richer pattern
of volatility structures than one-factor no-arbitrage models. For example, when f (r) = r,
a = 1, b = 0.1, σ1 = 0.01, σ2 = 0.0165, and ρ = 0.6 the model exhibits, at all times,
a “humped” volatility structure similar to that observed in the market. The correlation
structure implied by the model is also plausible with these parameters.
When f (r) = r the model is analytically tractable. The price at time t of a zero-
coupon bond that provides a payoff of $1 at time T is

P (t, T ) = A(t, T ) exp[−B(t, T )r − C(t, T )u]

where
1
B(t, T ) = [1 − e−a(T −t) ]
a
1 1 1
C(t, T ) = e−a(T −t) − e−b(T −t) +
a(a − b) b(a − b) ab
and A(t, T ) is as given in the Appendix to this note.
The prices, c and p, at time zero of European call and put options on a zero-coupon
bond are given by
c = LP (0, s)N (h) − KP (0, T )N (h − σP )
p = KP (0, T )N (−h + σP ) − LP (0, s)N (−h)
where T is the maturity of the option, s is the maturity of the bond, K is the strike price,
L is the bond’s principal
1 LP (0, s) σP
h= ln +
σP P (0, T )K 2
and σP is as given in the Appendix. Because this is a two-factor model, an option on
a coupon-bearing bond cannot be decomposed into a portfolio of options on zero-coupon

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
bonds as described in Technical Note 15. However, we can obtain an approximate analytic
valuation by calculating the first two moments of the price of the coupon-bearing bond
and assuming the price is lognormal.

Constructing a Tree
To construct a tree for the model in equation (1), we simplify the notation by defining
x = f (r) so that
dx = [θ(t) + u − ax] dt + σ1 dz1
with
du = −bu dt + σ2 dz2
Assuming a 6= b we can eliminate the dependence of the first stochastic variable on the
second by defining
u
y =x+
b−a
so that
dy = [θ(t) − ay] dt + σ3 dz3
du = −bu dt + σ2 dz2
where
σ22 2ρσ1 σ2
σ32 = σ12 + 2
+
(b − a) b−a
and dz3 is a Wiener process. The correlation between dz2 and dz3 is

ρσ1 + σ2 /(b − a)
σ3

A three-dimensional tree for y and u can be constructed on the assumption that

θ(t) = 0 and the initial values of y and u are zero. A methodology similar to that for
one-factor models can then be used to construct the final tree by increasing the values of
y at time i∆t by αi . In the f (r) = r case, an alternative approach is to use the analytic
expression for θ(t), given in the Appendix to this note.
Rebonato gives some examples of how the model can be calibrated and used in prac-
tice.2

2
See R. Rebonato Interest Rate Option Models, (2nd Ed., Chichester, England: John
Wiley and Sons, 1998) pp 306-8.

2
 APPENDIX
The Functions in the Two-Factor Hull-White Model
The A(t, T ) function is

P (0, T )
ln A(t, T ) = ln + B(t, T )F (0, t) − η
P (0, t)

where
σ12
η= (1 − e−2at )B(t, T )2 − ρσ1 σ2 [B(0, t)C(0, t)B(t, T ) + γ4 − γ2 ]
4a
1
− σ22 [C(0, t)2 B(t, T ) + γ6 − γ5 ]
2
e−(a+b)T [e(a+b)t − 1] e−2aT (e2at − 1)
γ1 = −
(a + b)(a − b) 2a(a − b)

e−a(T −t) − e−aT

 
1 1 2 1 2 t
γ2 = γ1 + C(t, T ) − C(0, T ) + B(t, T ) − B(0, T ) + −
ab 2 2 a a2

e−(a+b)t − 1 e−2at − 1
γ3 = − +
(a − b)(a + b) 2a(a − b)
e−at − 1
 
1 1 2 t
γ4 = γ3 − C(0, t) − B(0, t) + +
ab 2 a a2
 
1 1 2 1 2
γ5 = C(t, T ) − C(0, T ) + γ2
b 2 2
 
1 1 2
γ6 = γ4 − C(0, t)
b 2
where F (t, T ) is the instantaneous forward rate at time t for maturity T .
The volatility function, σP , is
Z t
σP2 = {σ12 [B(τ, T ) − B(τ, t)]2 + σ22 [C(τ, T ) − C(τ, t)]2
0

+2ρσ1 σ2 [B(τ, T ) − B(τ, t)][C(τ, T ) − C(τ, t)]} dτ

This shows that σP2 has three components. Define

1
U= [e−aT − e−at ]
a(a − b)

and
1
V = [e−bT − e−bt ]
b(a − b)

3
The first component of σP2 is
σ12
B(t, T )2 (1 − e−2at )
2a
The second is

U 2 2at V 2 2bt
 
U V (a+b)t
σ22 (e − 1) + (e − 1) − 2 (e − 1)
2a 2b a+b

The third is
 
2ρσ1 σ2 −at −aT U 2at V (a+b)t
(e −e ) (e − 1) − (e − 1)
a 2a a+b

Finally, the θ(t) function is

θ(t) = Ft (0, t) + aF (0, t) + φt (0, t) + aφ(0, t)

where the subscript denotes a partial derivative and

1 2 1
φ(t, T ) = σ1 B(t, T )2 + σ22 C(t, T )2 + ρσ1 σ2 B(t, T )C(t, T )
2 2

4
 Technical Note No. 15*
Options, Futures, and Other Derivatives
John Hull

Valuing Options on Coupon-Bearing Bonds in a One-Factor Interest Rate Model

Jamshidian shows that the prices of options on coupon-bearing bonds can be obtained
from the prices of options on zero-coupon bonds in a one-factor interest rate model, such
as Vasicek, Ho-Lee, Hull-White, and Cox-Ingersoll-Ross.1 These models have the property
that all rates are moving in the same direction as the short rate at any given time.
Consider a European call option with exercise price K and maturity T on a coupon-
bearing bond. Suppose that the bond provides a total of n cash flows after the option
matures. Let the ith cash flow be ci and occur at time si (1 ≤ i ≤ n; si ≥ T ). Define:
rK : Value of the short rate, r, at time T that causes the coupon-bearing bond price to
equal the strike price.
Ki : Value at time T of a zero-coupon bond paying off $1 at time si when r = rK .
When bond prices are known analytically as a function of r, rK can be obtained very
quickly using an iterative procedure such as the Newton Raphson method.
The variable P (T, si ) is the price at time T of a zero-coupon bond paying $1 at time
si . The payoff from the option is, therefore,
" n
#
X
max 0, ci P (T, si ) − K
i=1

Because all rates are increasing functions of r, all bond prices are decreasing functions of
r. This means that the coupon-bearing bond is worth more than K at time T and should
be exercised if, and only if, r < rK . Furthermore, the zero-coupon bond maturing at time
si underlying the coupon-bearing bond is worth more than ci Ki at time T if, and only if,
r < rK . It follows that the payoff from the option is
n
X
ci max [0, P (T, si ) − Ki ]
i=1

This shows that the option on the coupon-bearing bond is the sum of n options on the
underlying zero-coupon bonds. A similar argument applies to European put options on
coupon-bearing bonds.

Example
Suppose that a = 0.1, b = 0.1, and σ = 0.02 in Vasicek’s model with the initial value
of the short rate being 10% per annum. Consider a three-year European put option
with a strike price of $98 on a bond that will mature in five years. Suppose that the
bond has a principal of $100 and pays a coupon of $5 every six months. At the end

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
See F. Jamshidian, “An Exact Bond Option Pricing Formula,” Journal of Finance,
44 (March 1989), 205–9.

1
of three years, the bond can be regarded as the sum of four zero-coupon bonds. If the
short-term interest rate is r at the end of the three years, the value of the bond is

5A(3, 3.5)e−B(3,3.5)r + 5A(3, 4)e−B(3,4)r + 5A(3, 4.5)e−B(3,4.5)r + 105A(3, 5)e−B(3,5)r

Using the expressions for A(t, T ) and B(t, T ), this becomes

5 × 0.9988e−0.4877r + 5 × 0.9952e−0.9516r + 5 × 0.9895e−1.3929r + 105 × 0.9819e−1.8127r

To apply Jamshidian’s procedure, we must find rK , the value of r for which this bond
price equals the strike price of 98. An iterative procedure shows that rK = 0.10952.
When r has this value, the values of the four zero-coupon bonds underlying the coupon-
bearing bond are 4.734, 4.484, 4.248, and 84.535. The option on the coupon-bearing
bond is, therefore, the sum of four options on zero-coupon bonds:
1. A three-year option with strike price 4.734 on a 3.5-year zero-coupon bond with a
principal of 5.
2. A three-year option with strike price 4.484 on a four-year zero-coupon bond with
a principal of 5.
3. A three-year option with strike price 4.248 on a 4.5-year zero-coupon bond with a
principal of 5.
4. A three-year option with strike price 84.535 on a five-year zero-coupon bond with
a principal of 105.
To illustrate the pricing of these options, consider the fourth. P (0, 3) = 0.7419 and
P (0, 5) = 0.6101. Also, σP = 0.05445, h = 0.4161, L = 105, and K = 84.535. The
value of the option as 0.8085. Similarly, the value of the first, second, and third
options are, respectively, 0.0125, 0.0228, and 0.0314. The value of the option under
consideration is, therefore, 0.0125 + 0.0228 + 0.0314 + 0.8085 = 0.8752.

2
 Technical Note No. 16*
Options, Futures, and Other Derivatives
John Hull

Construction of an Interest Rate Tree with

Non-Constant Time Steps and Non-Constant Parameters

Consider a one-factor model of the form

df (r) = [θ(t) − a(t)f (r)] dt + σ(t) dz

As in the text, we let x = f (r) and first build a tree for the process

dx = −a(t)x dt + σ(t) dz

The procedure for doing this is given in Technical Note 9. We then convert this tree to a
tree for the process
dx = [θ(t) − a(t)x] dt + σ(t) dz
so that the zero curve is fitted using the approach given in the text. For more details see
“The Generalized Hull–White Model and Supercalibration,” Financial Analysts Journal,
57, 6, Nov-Dec, 2001. The article is also available on John Hull’s website.

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 Technical Note No. 17*
Options, Futures, and Other Derivatives
John Hull

The Process for the Short Rate in an HJM Term Structure Model

This note considers the relationship between the HJM model and the models of the
short rate. Consider a one-factor continuous time model for forward rates. Because
Z t
F (t, t) = F (0, t) + dF (τ, t)
0

and r(t) = F (t, t), it follows from the HJM analysis in the text that
Z t Z t
r(t) = F (0, t) + v(τ, t, Ωτ )vt (τ, t, Ωτ ) dτ + vt (τ, t, Ωτ ) dz(τ )
0 0

Differentiating with respect to t and using v(t, t, Ωt ) = 0 we obtain5

Z t 
2
dr(t) =Ft (0, t) dt + [v(τ, t, Ωτ )vtt (τ, t, Ωτ ) + vt (τ, t, Ωτ ) ] dτ dt
0
Z t 
+ vtt (τ, t, Ωτ ) dz(τ ) dt + [vt (τ, t, Ωτ )|τ =t ] dz(t)
0

It is interesting to examine the terms on the right-hand side of this equation. The
first and fourth terms are straightforward. The first term shows that one component of
the drift in r is the slope of the initial forward rate curve. The fourth term shows that
the instantaneous standard deviation of r is vt (τ, t, Ωτ )|τ =t . The second and third terms
are more complicated, particularly when v is stochastic. The second term depends on the
history of v because it involves v(τ, t, Ωτ ) when τ < t. The third term depends on the
history of both v and dz.
The second and third terms terms are liable to cause the process for r to be non-
Markov. The drift of r between time t and t + ∆t is liable to depend not only on the value
of r at time t, but also on the history of r prior to time t. This means that, when we
attempt to construct a tree for r, it is nonrecombining. An up movement followed by a
down movement does not lead to the same node as a down movement followed by an up
movement.

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
5
The stochastic calculus in this equation may be unfamiliar to some readers. To inter-
pret what is going
R t on, we can replace integral signsPwith summation signs and d’s with ∆’s.
n
For example, 0 v(τ, t, Ωτ )vt (τ, t, Ωτ )dτ becomes i=1 v(i∆t, t, Ωi )vt (i∆t, t, Ωi )∆t, where
∆t = t/n.

1
 Technical Note No. 18*
Options, Futures, and Other Derivatives
John Hull

Valuation of a Compounding Swap

Consider a compounding swap where floating rate cash flows in a swap are com-
pounded forward at LIBOR plus a spread rather than being paid. We can use forward
rate agreements to show that the value of the floating side is the same as the value it would
have if forward rates were realized. In other words, the swap can be valued similarly to a
regular swap by assuming that future interest rates equal today’s forward rates.
Suppose that t0 is the time of the payment date immediately preceding the valuation
date and that the payment dates following the valuation date are at times t1 , t2 , . . ., tn .
Define τi = ti+1 − ti (0 ≤ i ≤ n − 1) and other variables as follows1
L: Principal on the floating side of swap
Qi : Value of floating side compounded forward to time ti (Q0 is known)
Q∗i : Value of floating side compounded forward to time ti if forward rate is realized
Ri : LIBOR rate from ti to ti+1 for i ≥ 1 (R0 is known)
Fi : Forward rate applicable to period between time ti and ti+1 (all known)
sp : Spread above LIBOR at which interest is paid on the floating side of the swap (20
basis points in Business Snapshot example in text)
sc : Spread above LIBOR at which floating interest compounds (10 basis points in Business
Snapshot example in text)
We assume that the spread sc is applied first to Qi and then the result is compounded
forward at Ri to produce Qi (1 + Ri τi )(1 + sc τi ). (This assumption is discussed at the end
of the Note.)
The value of the floating side of the swap at time t1 is known. It is:
Q1 = Q0 [(1 + R0 τ0 )(1 + sc τ0 )] + L(R0 + sp )τ0
The first term on the right hand side is the result of compounding the floating payments
from time t0 to t1 . The second term is the floating payment at time t1 .
The value of the floating side at time t2 is not known and depends on R1 . It is
Q2 = Q1 [(1 + R1 τ1 )(1 + sc τ1 )] + L(R1 + sp )τ1 (1)
However, we can costlessly enter into two FRAs today:
1. An FRA to exchange, at time t2 , R1 for F1 on a principal of Q1 (1 + sc τ1 )
2. An FRA to exchange, at time t2 , R1 for F1 on a principal of L
The first FRA shows that the first term on the right hand side of equation (1) has the
same present value as a cash flow of Q1 (1 + F1 τ1 )(1 + sc τ1 ) at time t2 . The second FRA
shows that the second term on the right hand side of equation (1) has the same present
value as a cash flow of L(F1 + sp )τ1 at time t2 . The value of the floating side of the swap
at time t2 is, therefore the same as the value of a cash flow of
Q1 [(1 + F1 τ1 )(1 + sc τ1 )] + L(F1 + sp )τ1

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull
1
All rates are here expressed with a compounding frequency reflecting their matu-
rity. Three-month rates are expressed with quarterly compounding; six-month rates are
expressed with semi-annual compounding; etc.

1
at time t2 . This means that Q2 at time t2 can costlessly be converted to Q∗2 at time t2 .
Consider next time t3 . The compounded forward amount at time t3 is

Q3 = Q2 [(1 + R2 τ2 )(1 + sc τ2 )] + L(R2 + sp )τ2 (2)

To deal with the first term on the right hand side, we note that a cash flow of Q2 [(1 +
R2 τ2 )(1 + sc τ2 )] at time t3 is worth the same as Q2 (1 + sc τ2 ) at time t2 . This from our
earlier result is worth the same as Q∗2 (1 + sc τ2 ) at time t2 . This in turn is worth the same
as Q∗2 [(1 + F2 τ2 )(1 + sc τ2 )] at time t3 . To deal with the second term, we note that we can
today enter into an FRA to exchange, at time t3 , R2 for F2 on a principal of L. These
two observations show that a cash flow of Q3 at time t3 is worth the same as a cash flow
of Q∗3 at time t3 .
Similarly, a cash flow of Q4 at time t4 is worth the same as a cash flow of Q∗4 at time
t4 ; a cash flow of Q5 at time t5 is worth the same as a cash flow of Q∗5 at time t5 ; and so
on. In particular, a cash flow of Qn at time tn is worth the same as a cash flow of Q∗n at
time tn so that the result is proved.
In practice it may be the case that Qi is compounds forward to Qi [1 + (Ri + sc )τi ]
rather than to Qi (1 + Ri τi )(1 + sc τi ). There is then an approximation. The result is only
true when small terms of the form sc Ri τi2 are ignored.
The example in teh text provides an application of the result in this Technical Note.
(It does make the approximation just mentioned).

2
 Technical Note No. 19*
Options, Futures, and Other Derivatives
John Hull

Valuation of an Equity Swap

As explained in the text an equity swap is always worth zero immediately after a
payment date. To value an equity swap between two payment dates, we define
R0 : Floating rate applicable to the next payment date (determined at the last payment
date)
L: Principal
τ0 : Time between last payment date and next payment date
τ : Time between now and next payment date
E0 : Value of the equity index at the last reset date
E: Current value of the equity index
R: LIBOR rate for the period between now and the next payment date.
If we borrow
E
L
E0
at rate R for time τ and invest it in the index, we create an exchange of
E1 E
L for L(1 + Rτ ) (1)
E0 E0
at the next payment date where E1 is the equity index on that payment date. Since this
exchange can be created costlessly it is worth zero. The exchange that will actually take
place at the next payment date is
 
E1
−1 L for R0 Lτ0
E0
Adding the principal L to both sides we see the actual exchange is equivalent to
E1
L for L(1 + R0 τ0 ) (2)
E0
Comparing equation (1) with equation (2) see that value of the swap to the party receiving
floating is the present value of
E
L(1 + R0 τ0 ) − L (1 + Rτ )
E0
This is
1 + R 0 τ0 E
L −L
1 + Rτ E0
Similarly, the value of the swap to the party receiving the equity return is
E 1 + R 0 τ0
L −L
E0 1 + Rτ

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 Technical Note No. 20*
Options, Futures, and Other Derivatives
John Hull

Changing the Market Price of Risk for Variables

That Are not the Prices of Traded Securities
This Note shows that the rules for changing the stochastic process when there is a
change in the market price of risk are the same, regardless of whether the variable is or is
not the price of a traded security.
Suppose that a single variable θ (not the price of a traded security) follows the process

dθ = mθ dt + sθ dz

and that a traded security f dependent θ follows the process

df = µf dt + σf dz

When the growth rate of θ increases by ∆m so that the drift of θ increases by θ∆m
we know from Ito’s lemma that the growth rate of f increases by ∆µ where

∂f
f ∆µ = θ∆m
∂θ
Also using Ito’s lemma
∂f
σf = sθ
∂θ
Hence
∆µ ∆m
= (1)
σ s
Consider the situation where the market price of risk changes from λ to λ∗ . We know
that
∆µ = (λ∗ − λ)σ
Equation(1) shows that
∆m = (λ∗ − λ)s
This shows that when the market price of risk changes, the stochastic processes of variables
that are not traded securities are adjusted in the same way as the stochastic processes of
those that are.
This analysis can be extended to the situation where there are several sources of
uncertainty so that
Xn
dθ = mθ dt + si θ dzi
i=1

and
n
X
df = µf dt + σi f dzi
i=1

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
In this case we still have
∂f
f ∆µ = θ∆m
∂θ
Using Ito’s lemma
∂f
σi f = si θ
θ
Hence
∆µ ∆m
= (2)
σi si
When
n
X
∆µ = (λ∗i − λi )σi
i=1

Equation(2) shows that

n
X
∆m = (λ∗ − λ)si
i=1

which is the required result.

2
 Technical Note No. 21*
Options, Futures, and Other Derivatives
John Hull

Hermite Polynomials and Their Use for Integration

As explained in the chapter on credit derivatives in the text, the Gaussian copula
model requires functions to be integrated over a normal distribution between −∞ and
+∞. Gaussian quadrature approximates the integral as
Z ∞ M
1 2 X
√ e−F /2 g(F )dF ≈ wk g(Fk ) (1)
−∞ 2π k=1

The approximation gets better as M increases. It has the property that it is exact when
g(F ) is a polynomial of order M .
The determination the wk and Fk involves Hermite polynomials. If you want to avoid
getting into the details of this, values of wk and Fk for different values of M can be
downloaded from a spread sheet on the author’s web site.
The first few Hermite polynomials are

H0 (x) = 1
H1 (x) = 2x

H2 (x) = 4x2 − 2

H3 (x) = 8x3 − 12x

H4 (x) = 16x4 − 48x2 + 12

A recurrence relationship for calculating higher order polynomials is

Hn+1 (x) = 2xHn (x) − 2nHn−1 (x)

and an equation for the derivative with respect to x is

Hn0 (x) = 2nHn−1 (x)

Define xk (1 ≤ k ≤ n) as the n roots of Hn (x) (that is, the n values of x for which
Hn (x) = 0) and
√
∗ 2n−1 n! π
wk = 2
n [Hn−1 (xk )]2
A key result is
Z ∞ n
X 2
f (x)dx ≈ wk∗ exk f (xk ) (2)
−∞ k=1

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 √
Setting x = F/ 2 and
1 2 √
f (x) = √ e−x g( 2x)
π
equation (2) gives
Z ∞ n
1 2 X1
√ e−F /2 g(F )dF ≈ wk∗ g(Fk )
−∞ 2π π
k=1

or alternatively
Z ∞ n
1 2 X
√ e−F /2 g(F )dy ≈ wk g(Fk )
−∞ 2π k=1

where
w∗ √
wk = √k Fk = 2xk
π
This is the result in equation (1), with n = M .
This leaves the problem of calculating the n roots of a Hermite polynomial. A program
for doing this is ‘gauher’ in “Numerical Recipes for C: The Art of Scientific Computing”
by Press, Flanery, Teukolsky, and Vetterling, Cambridge University Press.

2
 Technical Note No. 22*
Options, Futures, and Other Derivatives
John Hull

Valuation of a Variance Swap

This note proves the result in the text for the valuation of a variance swap.
Suppose that the stock price follows process
dS
= (r − q) dt + σ dz
S
in a risk-neutral world where σ is itself stochastic. From Ito’s lemma
d ln S = (r − q − σ 2 /2) dt + σ dz
By subtracting these two equations we obtain

σ2 dS
dt = − d ln S
2 S
Integrating between time 0 and time T , the realized average variance rate, V , between
time 0 and time T is given by
Z T
1 dS ST
VT = − ln
2 0 S S0
or Z T
2 dS 2 ST
V = − ln (1)
T 0 S T S0
Taking expectations in a risk-neutral world
 
2 2 ST
Ê(V ) = (r − q)T − Ê ln
T T S0
or  
2 F0 2 ST
Ê(V ) = ln − Ê ln (2)
T S0 T S0
where F0 is the forward price of the asset for a contract maturing at time T .
Consider Z S∗
1
2
max(K − ST , 0)dK
K=0 K
for some value S ∗ of S. When S ∗ < ST this integral is zero. When S ∗ > ST it is
S∗
S∗
Z
1 ST
2
(K − S T )dK = ln + ∗ −1
K=ST K ST S

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
Consider next Z ∞
1
max(ST − K, 0)dK
K=S ∗ K2
When S ∗ > ST this is zero. When S ∗ < ST it is
ST
S∗
Z
1 ST
2
(S T − K)dK = ln + ∗ −1
K=S ∗ K ST S

From these results it follows that

S∗ ∞
S∗
Z Z
1 1 ST
max(K − ST , 0)dK + max(S T − K, 0)dK = ln + −1
K=0 K2 K=S ∗ K2 ST S∗

for all values of S ∗ so that

Z S∗ Z ∞
ST ST 1 1
ln ∗ = ∗ − 1 − max(K − ST , 0)dK − max(ST − K, 0)dK (3)
S S K=0 K2 K=S ∗ K2

This shows that a variable that pays off ln ST can be replicated using options. This result
can be used in conjunction with equation (1) to provide a replicating portfolio for V .
Taking expectations in a risk-neutral world in equation (3)
  Z S∗ Z ∞
ST F0 1 RT 1 RT
Ê ln ∗ = ∗ −1− e p(K)dK − e c(K)dK (4)
S S K=0 K2 K=S ∗ K2

where c(K) and p(K) are the prices of European call and put options with strike price K
and maturity T and R is the risk-free interest rate for a maturity of T .
Combining equations (2) and (4) and noting that

S∗
   
ST ST
Ê ln = ln + Ê ln ∗
S0 S0 S

2 F0 2 S∗
Ê(V ) = ln − ln
T S0 T S0
"Z ∗ Z ∞ #
  S
2 F0 2 1 RT 1 RT
− − 1 + e p(K)dK + e c(K)dK
T S∗ T K=0 K2 K=S ∗ K
2

which reduces to
"Z ∗ Z ∞ #
  S
2 F0 2 F0 2 1 RT 1 RT
Ê(V ) = ln ∗ − −1 + e p(K)dK + e c(K)dK
T S T S∗ T K=0 K
2
K=S ∗ K
2

This is the result in the text.

2
 Technical Note No. 231

Options, Futures, and Other Derivatives

John Hull

The Black, Derman, and Toy Model

As explained in the text, there are two types of model of the short rate: equilibrium and no-arbitrage
models. In an equilibrium model the process followed by the short-term interest rate is specified. This
totally defines the model. Zero-coupon bond prices and the term structure of interest rates are outputs
from the model. Examples of equilibrium models are the Vasicek and Cox, Ingersoll, and Ross models.
These models each have three parameters. The parameters can be chosen so that the models provide an
approximate fit to the term structure of interest rates, but the fit is not usually an exact one.

A no-arbitrage model is constructed so that it is exactly consistent with the term structure of interest rates
that is observed in the market. This means that the term structure of interest rates is an input to the model,
not an output from it. No-arbitrage models can be constructed in many different ways. An early no-
arbitrage model was the Black, Derman, and Toy model published in 1990.2 This model has the
advantage that it can easily be represented in the form of a binomial tree. To correspond as closely as
possible with the Black-Derman-Toy paper, we assume that interest rates are compounded annually.

The Black-Derman-Toy model is a particular case of the more general Black-Karasinski model. The
short rate follows a mean-reverting lognormal process. However, the way the tree is constructed implies a
relationship between the short rate volatility and the reversion rate.

As in the case of the binomial tree used to value stock options, we consider steps of length t. The tree
models the behavior of the t-period interest rate. The zero-coupon interest rates for all maturities at time
zero are known. We denote the zero-coupon interest rate for a maturity of nt by Rn. The volatility of the
t rate between time (n−1)t and nt is denoted by n.

During each time step the t-period interest rate has a 50% probability of moving up and a 50%
probability of moving down. The tree is shown in Figure 1. The initial t period rate, r, is known and
equals R1. The value of a zero-coupon bond that pays off $1 at time 2t is

1
(1  R2 ) 2 t

The value of this bond at node B on the tree is

1
©Copyright John Hull. All Rights Reserved. This note may be reproduced for use in conjunction with Options,
Futures, and Other Derivatives by John C. Hull
2
See F. Black, E. Derman, and W. Toy, ``A one-factor model of interest rates and its application to Treasury bond
options, Financial Analysts Journal, 46 (1), 33-39.

1
 1
(1  rd ) t

The value of the bond at node C is

1
(1  ru ) t

It follows that the value of the bond at the initial node A is

1  1 1 
0.5  0.5 
(1  r ) t  (1  rd )
t
(1  ru ) t 

Hence

1  1 1  1
t 
0.5 t
 0.5 t 
 (1)
(1  r )  (1  rd ) (1  ru )  (1  R2 ) 2 t

Figure 1 The Binomial Tree

ruuu

ruu
F

ru
C ruud
rud
r A E

B rudd
rd

D
rdd

rddd

2
This is one equation that must be satisfied by ru and rd. To match the volatility, the standard deviation of
the logarithm of the interest rate at time t must be 1 t (Recall: i is the volatility of interest rates
during the ith time period.) This means that3

ru
0.5 ln  1 t (2)
rd

Equations (1) and (2) can be solved to determine ru and rd.

We now move on to determine ruu, rud and rdd. To match volatility we must have

ruu
0.5 ln   2 t (3)
rud

rud
0.5 ln   2 t (4)
rdd

We must also match the price of a zero-coupon bond that pays off $1 at the end of time 3t. Rolling back
through the tree the values of this bond at nodes D, E, and F are

1 1 1
, , and
(1  rdd ) t (1  rud ) t (1  ruu ) t

respectively. The values at nodes B and C are

1  1 1 
0.5  0.5 
(1  rd ) t  (1  rdd )
t
(1  rud ) t 

and

1  1 1 
t 
0.5 t
 0.5 t 
(1  ru )  (1  rud ) (1  ruu ) 

The value at the initial node is therefore

3
To see this note that the variance of the logarithm of the interest rate is
0.5(ln ru)2 +0.5(ln rd)2−[0.5(ln ru+ln rd)]2=[0.5(ln ru−ln rd)]2

3
 1  1  1 1 
 0.5 0.5  0.5 
 t  
(1  r ) t  (1  rd ) t  (1  rdd ) t
(1  rud )  
(5)
1  1  1 1  1
 0.5 0.5  0.5 
 t  
(1  r ) t  (1  ru ) t  (1  rud )
t
(1  ruu )   (1  R3 ) 3t

The interest rates r, ru and rd have already been determined. Equations (3), (4), and (5) therefore provide
three equations for determining ruu, rud, and rdd .

Continuing in this way a complete tree can be constructed. The calculations are made considerably easier
if as we move forward we keep track of the value of a security that pays $1 if a particular node is reached
and zero elsewhere. It is then only necessary to roll back one step when valuing zero-coupon bonds using
the tree.

Determining the i

The determination of the i depends on the data available. Sometimes the historical volatilities of zero-
coupon bond yields are used; sometimes the volatilities of caps or swaptions are used. An iterative search
procedure is always necessary.

When bond yields are being matched, we assume that we have data at time zero on the volatilities of a
bond maturing at time it. We will denote this by i. (We approximate i as the volatility of this bond
yield between time zero and time t.) We denote yun as the yield on a bond maturing at time nt at node
C and ydn as the yield on a bond maturing at time nt at node B. Considering a bond that matures at time
2t,

yu 2
 2 t  0.5 ln
yd 2

Because there is only one period left in the bond's life at the nodes at time t, yu2 = ru and yd2 = rd . As a
result

ru
 2 t  0.5 ln
rd

Using equation (2) leads to

1 = 2

The interest rates ru and rd can then be determined from equations (1) and (2).

Determining the subsequent i requires an iterative search. For example, to determine 2

4
 yu 3
 3 t  0.5 ln
yd 3

This must be solved iteratively with equations (3), (4), and (5) for 2, ruu, rud, and rdd.

In general the procedure to determine i (i > 1) is

1. Choose a trial value of i

2. Calculate the interest rates at time it
3. Calculate the yield volatility for a bond lasting until it from the tree. This involves calculating
the bond yields yu and yd at nodes B and C. The bond yield volatility is 0.5ln(yu/yd)
4. Search iteratively for the value of i that matches the bond yield volatility

Once the tree has been constructed it can be used to value a range of interest rate derivatives.

Example

As an example of the application of the model suppose that the term structure of interest rates is as shown
in Table 1, the zero-coupon yield volatilities are as shown in Table 2, and the time step is one year. In this
case r=0.10, t=1, 1=0.19 (the two-year yield volatility) and equations (1) and (2) give

1  1 1  1
0.5  0.5 
1.10  1  rd 1  ru  1.112
r
0.5 ln u  0.19
rd

Solving these two equations gives ru = 0.1432 and rd = 0.0979.

Table 1

Zero-coupon Yield Curve (Annually Compounded)

Maturity (years) Rate

1 10.0
2 11.0
3 12.0
4 12.5
5 13.0

5
 Table 2

Yield Volatilities

i i
2 19.0%
3 18.0%
4 17.5%
5 16.0%

Equations (3), (4), and (5) give

ruu
0.5 ln  2
rud

rud
0.5 ln  2
rdd

1  1  1 1  
0.5  0.5  0.5  
1.1  1.0979  1  rdd 1  rud  
1  1  1 1   1
0.5  0.5  0.5  
1.1  1.1432  1  rud 1  ruu   1.123

We do not know 2 directly. For each trial value of 2 we solve equations (3), (4), and (5) and then
calculate the price of a three-year bond at nodes B and C. The price of a three-year bond at node B is

1  1 1 
Bd  0.5  0.5 
1.0979  1  rdd 1  rud 

and the bond yield at node B is

1
yd  1
Bd

The price of a three-year bond at node C is

1  1 1 
Bu  0.5  0.5 
1.1432  1  rud 1  ruu 

The bond yield is

6
 1
yu  1
Bu

Carrying out an iterative search we find that 2 = 0.172 does the trick. With this value of 2 the solutions
to the three equations are

ruu = 0.1942

rud = 0.1377

rdd = 0.0976

These in turn give Bu = 0.7507, Bd = 0.8152, yu = 0.1542, and yd = 0.1076. Because 0.5ln(0.1542/0.1076)
= 0.18 the three-year yield volatility is matched.

The complete tree of short rates is shown in Figure 2.

Figure 2 The Short-Rate Tree

25.53

21.79

19.42 19.48

14.32 16.06

10 13.77 14.86

9.79 11.83

9.76 11.34

8.72

8.65

7
8
 Technical Note No. 24*
Options, Futures, and Other Derivatives
John Hull

Proof That Forward and Futures Prices

Are Equal When Interest Rates Are Constant
This Note demonstrates that forward and futures prices are equal when interest rates
are constant. Suppose that a futures contract lasts for n days and that Fi is the futures
price at the end of day i (0 < i < n). Define δ as the risk-free rate per day (assumed
constant). Consider the following strategy.1
1. Take a long futures position of eδ at the end of day 0 (i.e., at the beginning of the
contract).
2. Increase long position to e2δ at the end of day 1.
3. Increase long position to e3δ at the end of day 2.
And so on.
Table 1
The Investment Strategy to Show That Futures and Forward Prices Are Equal

Day 0 1 2 ... n−1 n

Futures price F0 F1 F2 ... Fn−1 Fn

Futures
position eδ e2δ e3δ ... enδ 0

Gain/loss 0 (F1 − F0 )eδ (F2 − F1 )e2δ ... ... (Fn − Fn−1 )enδ

Gain/loss
compounded
to day n 0 (F1 − F0 )enδ (F2 − F1 )enδ ... ... (Fn − Fn−1 )enδ

This strategy is summarized in Table 1. By the beginning of day i, the investor has a
long position of eδi . The profit (possibly negative) from the position on day i is

(Fi − Fi−1 )eδi

Assume that the profit is compounded at the risk-free rate until the end of day n. Its value
at the end of day n is

(Fi − Fi−1 )eδi e(n−i)δ = (Fi − Fi−1 )enδ

The value at the end of day n of the entire investment strategy is therefore

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
This strategy was proposed by J. C. Cox, J. E. Ingersoll, and S. A. Ross, “The Relation
between Forward Prices and Futures Prices,” Journal of Financial Economics 9 (December
1981): 321–46.

1
 n
X
(Fi − Fi−1 )enδ
i=1

This is
[(Fn − Fn−1 ) + (Fn−1 − Fn−2 ) + · · · + (F1 − F0 )]enδ
=(Fn − F0 )enδ
Because Fn is the same as the terminal asset spot price, ST , the terminal value of the
investment strategy can be written

(ST − F0 )enδ

An investment of F0 in a risk-free bond combined with the strategy involving futures just
given yields
F0 enδ + (ST − F0 )enδ = ST enδ
at time T . No investment is required for all the long futures positions described. It follows
that an amount F0 can be invested to give an amount ST enδ at time T .
Suppose next that the forward price at the end of day 0 is G0 . Investing G0 in a
riskless bond and taking a long forward position of enδ forward contracts also guarantees
an amount ST enδ at time T . Thus, there are two investment strategies—one requiring an
initial outlay of F0 and the other requiring an initial outlay of G0 —both of which yield
ST enδ at time T . It follows that in the absence of arbitrage opportunities

F0 = G 0

In other words, the futures price and the forward price are identical. Note that in this proof
there is nothing special about the time period of one day. The futures price based on a
contract with weekly settlements is also the same as the forward price when corresponding
assumptions are made.

2
 Technical Note No. 25*
Options, Futures, and Other Derivatives
John Hull
A Cash-Flow Mapping Procedure
This note explains one procedure for mapping cash flows to standard maturity dates.
We will illustrate the procedure by considering a simple example of a portfolio consisting of
a long position in a single Treasury bond with a principal of $1 million maturing in 0.8 years.
We suppose that the bond provides a coupon of 10% per annum payable semiannually. This
means that the bond provides coupon payments of $50,000 in 0.3 years and 0.8 years. It
also provides a principal payment of $1 million in 0.8 years. The Treasury bond can
therefore be regarded as a position in a 0.3-year zero-coupon bond with a principal of
$50,000 and a position in a 0.8-year zero-coupon bond with a principal of $1,050,000.
The position in the 0.3-year zero-coupon bond is mapped into an equivalent position
in 3-month and 6-month zero-coupon bonds. The position in the 0.8-year zero-coupon
bond is mapped into an equivalent position in 6-month and 1-year zero-coupon bonds.
The result is that the position in the 0.8-year coupon-bearing bond is, for VaR purposes,
regarded as a position in zero-coupon bonds having maturities of three months, six months,
and one year.
The Mapping Procedure
Consider the $1,050,000 that will be received in 0.8 years. We suppose that zero rates,
daily bond price volatilities, and correlations between bond returns are as shown in Table
1.
The first stage is to interpolate between the 6-month rate of 6.0% and the 1-year rate
of 7.0% to obtain a 0.8-year rate of 6.6%. (Annual compounding is assumed for all rates.)
The present value of the $1,050,000 cash flow to be received in 0.8 years is
1, 050, 000
= 997, 662
1.0660.8
We also interpolate between the 0.1% volatility for the 6-month bond and the 0.2% volatil-
ity for the 1-year bond to get a 0.16% volatility for the 0.8-year bond.
Table 1
Data to Illustrate Mapping Procedure
Maturity 3-Month 6-Month 1-Year

Zero rate (% with ann. comp.) 5.50 6.00 7.00

Bond price vol (% per day) 0.06 0.10 0.20

Correlation between 3-Month 6-Month 1-Year

daily returns Bond Bond Bond

3-month bond 1.0 0.9 0.6

6-month bond 0.9 1.0 0.7
1-year bond 0.6 0.7 1.0

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 Table 2
The Cash Flow Mapping Result

$50,000 Received $1,050,000 Received

in 0.3 Years in 0.8 Years Total

Position in 3-month bond ($) 37,397 37,397

Position in 6-month bond ($) 11,793 319,589 331,382
Position in 1-year bond ($) 678,074 678,074

Suppose we allocate α of the present value to the 6-month bond and 1 − α of the
present value to the 1-year bond. Matching variances we obtain

0.00162 = 0.0012 α2 + 0.0022 (1 − α)2 + 2 × 0.7 × 0.001 × 0.002α(1 − α)

This is a quadratic equation that can be solved in the usual way to give α = 0.320337.
This means that 32.0337% of the value should be allocated to a 6-month zero-coupon bond
and 67.9663% of the value should be allocated to a 1-year zero coupon bond. The 0.8-year
bond worth $997,662 is, therefore, replaced by a 6-month bond worth

997, 662 × 0.320337 = $319, 589

and a 1-year bond worth

997, 662 × 0.679663 = $678, 074

This cash-flow mapping scheme has the advantage that it preserves both the value and
the variance of the cash flow. Also, it can be shown that the weights assigned to the two
adjacent zero-coupon bonds are always positive.
For the $50,000 cash flow received at time 0.3 years, we can carry out similar calcu-
lations. It turns out that the present value of the cash flow is $49,189. It can be mapped
into a position worth $37,397 in a three-month bond and a position worth $11,793 in a
six-month bond.
The results of the calculations are summarized in Table 2. The 0.8-year coupon-
bearing bond is mapped into a position worth $37,397 in a three-month bond, a position
worth $331,382 in a six-month bond, and a position worth $678,074 in a one-year bond.
Using the volatilities and correlations in Table 1, the standard formula for calculating
the variance of a portfolio can be used with with n = 3, α1 = 37, 397, α2 = 331, 382,
α3 = 678, 074, σ1 = 0.0006, σ2 = 0.001 and σ3 = 0.002, and ρ12 = 0.9, ρ13 = 0.6,
ρ23 = 0.7. This variance is of the 0.8-year bond is√ 2,628,518. The standard deviation of
the change in the price of the bond is, therefore, 2, 628, 518 = 1, 621.3. Because we are
assuming that the bond is the only instrument in the portfolio, the 10-day 99% VaR is
√
1621.3 × 10 × 2.33 = 11, 946

or about $11,950.

2
 Technical Note No. 26*
Options, Futures, and Other Derivatives
John Hull

A Binomial Measure of Credit Correlation

A credit correlation measure sometimes used by rating agencies is the binomial corre-
lation measure. For two companies, A and B, this is the coefficient of correlation between:
1. A variable that equals 1 if company A defaults between times 0 and T and zero
otherwise; and
2. A variable that equals 1 if company B defaults between times 0 and T and zero
otherwise
The measure is

PAB (T ) − QA (T )QB (T )
βAB (T ) = p (1)
[QA (T ) − QA (T )2 ][QB (T ) − QB (T )2 ]

where PAB (T ) is the joint probability of A and B defaulting between time zero and time T ,
QA (T ) is the cumulative probability that company A will default by time T , and QB (T )
is the cumulative probability that company B will default by time T . Typically βAB (T )
depends on T , the length of the time period considered. Usually it increases as T increases.
From the definition of a Gaussian copula model PAB (T ) = M [xA (T ), xB (T ); ρAB ].
where xA (T ) = N −1 (QA (T )) and xB (T ) = N −1 (QB (T )) are the transformed times to
default for companies A and B, and ρAB is the Gaussian copula correlation for the times
to default for A and B. M (a, b; ρ) is the probability that, in a bivariate normal distribution
where the correlation between the variables is ρ, the first variable is less than a and the
second variable is less than b.1 It follows that

M [xA (T ), xB (T ); ρAB ] − QA (T )QB (T )

βAB (T ) = p (2)
[QA (T ) − QA (T )2 ][QB (T ) − QB (T )2 ]

This shows that, if QA (T ) and QB (T ) are known, βAB (T ) can be calculated from ρAB and
vice versa. Usually ρAB is markedly greater than βAB (T ). This illustrates the important
point that the magnitude of a correlation measure depends on the way it is defined.

Example
Suppose that the probability of company A defaulting in one-year period is 1% and
the probability of company B defaulting in a one-year period is also 1%. In this
case xA (1) = xB (1) = N −1 (0.01) = −2.326. If ρAB is 0.20, M (xA (1), xB (1), ρAB ) =
0.000337 and equation (2) shows that βAB (T ) = 0.024 when T = 1.

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
See Technical Note 5 for the calculation of M (a, b; ρ).

1
 Technical Note No. 27*
Options, Futures, and Other Derivatives
John Hull

Calculation of Moments for Valuing Asian Options

We consider the problem of calculating the first two moments of the arithmetic average
price of an asset in a risk-neutral world when the average is calculated from discrete
observations. Suppose that the asset price is observed at times Ti (1 ≤ i ≤ m). We define
variables as follows:
Si : The value of the asset at time Ti
Fi : The forward price of the asset for a contract maturing at time Ti
σi : The implied volatility for an option on the asset with maturity Ti
ρij : Correlation between return on asset up to time Ti and the return on the asset up to
time Tj
P : Value of the arithmetic average
M1 : First moment of P in a risk-neutral world
M2 : Second moment of P in a risk-neutral world
With these definitions
m
1 X
M1 = Fi
m i=1

Also
m m
2 1 XX
P = 2 Si Sj
m i=1 j=1

In this case √
Ê(Si Sj ) = Fi Fj eρij σi σj Ti Tj

It can be shown that when i ≤ j √

σi Ti
ρij = p
σj Tj
so that 2
Ê(Si Sj ) = Fi Fj eσi Ti
and  
m m Xj−1
1 X 2 X 2
M2 = 2  Fi2 eσi Ti + 2 Fi Fj eσi Ti 
m i=1 j=1 i=1

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 Technical Note No. 28*
Options, Futures, and Other Derivatives
John Hull

Calculation of Moments for Valuing Basket Options

Consider the problem of calculating the first two moments of the value of a basket of
assets at a future time, T , in a risk-neutral world. The price of each asset in the basket is
assumed to be lognormal. Define
n : The number of assets
Si : The value of the ith asset at time T 1
Fi : The forward price of the ith asset for a contract maturing at time T .
σi : The volatility of the ith asset between time zero and time T
ρij : Correlation between returns from the ith and jth asset
P : Value of basket at time T
M1 : First moment of P in a risk-neutral world
M2 : Second moment Pn of P in a risk-neutral world
Because P = i=1 Si , Ê(Si ) = Fi , M1 = Ê(P ) and M2 = Ê(P 2 ) where Ê denotes
expected value in a risk-neutral world, it follows that
n
X
M1 = Fi
i=1

Also,
n X
X n
2
P = Si Sj
i=1 j=1

From the properties of lognormal distributions

Ê(Si Sj ) = Fi Fj eρij σi σj T

Hence
n X
X n
M2 = Fi Fj eρij σi σj T
i=1 j=1

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
If the ith asset is a certain stock and there are, say, 200 shares of the stock in the
basket, then the ith “asset” is defined as 200 shares of the stock and Si is the value of 200
shares of the stock.

1
 Technical Note No. 29*
Options, Futures, and Other Derivatives
John Hull

Proof of Extensions to Ito’s Lemma

Options, Futures and Other Derivatives proves Ito’s lemma for a function of a single
stochastic variable. Here we present a generalized version of Ito’s lemma for the situation
where there are several sources of uncertainty.
Suppose that a function, f , depends on the n variables x1 , x2 , . . . , xn and time, t.
Suppose further that xi follows an Ito process with instantaneous drift ai and instantaneous
variance b2i (1 ≤ i ≤ n), that is,

dxi = ai dt + bi dzi (1)

where dzi is a Wiener process (1 ≤ i ≤ n). Each ai and bi may be any function of all the
xi ’s and t. A Taylor series expansion of ∆f gives
n n n n
X ∂f ∂f 1 X X ∂2f 1 X ∂2f
∆f = ∆xi + ∆t + ∆xi ∆xj + ∆xi ∆t + · · · (2)
i=1
∂xi ∂t 2 i=1 j=1
∂xi ∂xj 2 i=1
∂xi ∂t

Equation (1) can be discretized as

√
∆xi = ai ∆t + bi i ∆t

where i is a random sample from a standardized normal distribution. The correlation,

ρij , between dzi and dzj is defined as the correlation between i and j . In the book’s
proof of Ito’s lemma when there is only one stochastic variable it was argued that

lim ∆x2i = b2i dt

∆t→0

Similarly,
lim ∆xi ∆xj = bi bj ρij dt
∆t→0

As ∆t → 0, the first three terms in the expansion of ∆f in equation (2) are of order ∆t.
All other terms are of higher order. Hence
n n n
X ∂f ∂f 1 X X ∂2f
df = dxi + dt + bi bj ρij dt
i=1
∂xi ∂t 2 i=1 j=1
∂xi ∂xj

This is the generalized version of Ito’s lemma. Substituting for dxi from equation (1) gives
 
n n X n 2 n
X ∂f ∂f 1 X ∂ f X ∂f
df =  ai + + bi bj ρij  dt + bi dzi (3)
i=1
∂xi ∂t 2 i=1 j=1
∂xi ∂xj i=1
∂xi

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.

1
 For an alternative generalization of Ito’s lemma suppose that f depends on a single
variable x and that the process for x involves more than one Wiener process:
m
X
dx = a dt + bi dzi
i=1

In this case
∂f ∂f 1 ∂2f 2 1 ∂2f
∆f = ∆x + ∆t + ∆x + ∆x∆t + · · ·
∂x ∂t 2 ∂x2 2 ∂x ∂t
m
X √
∆x = a ∆t + bi i ∆t
i=1

and
m X
X m
lim ∆x2i = bi bj ρij dt
∆t→0
i=1 j=1

where as before ρij is the correlation between dzi and dzj This leads to
 
2 m X
m m
∂f ∂f 1 ∂ f X ∂f X
df =  a+ + bi bj ρij  dt + bi dzi (4)
∂x ∂t 2 ∂x2 i=1 j=1
∂x i=1

Finally consider the more general case where f depends on variables xi (1 ≤ i ≤ n)

and
m
X
dxi = ai dt + bik dzk
k=1

A similar analysis shows that

 
n n X
n 2 m X
m n m
X ∂f ∂f 1 X ∂ f X X ∂f X
df =  ai + + bik bjl ρkl  dt + bik dzk (5)
i=1
∂xi ∂t 2 i=1 j=1
∂xi ∂xj i=1
∂xi
k=1 l=1 k=1

2
 Technical Note No. 30*
Options, Futures, and Other Derivatives
John Hull

The Return for a Security Dependent on Multiple sources of Uncertainty

In this Note we prove a result relating the excess return to market prices of risk when
there are multiple sources of uncertainty.
Suppose there are n stochastic variables following Wiener processes. Consider n + 1
traded securities whose prices depend on some or all of the n stochastic variables. Define
fj as the price of the jth security (1 ≤ j ≤ n + 1). We assume that no dividends or other
income is paid by the n + 1 traded securities.1 It follows from Ito’s lemma in Technical
Note 29 that the securities follow processes of the form
n
X
dfj = µj fj dt + σij fj dzi (1)
i=1

Since there are n + 1 traded securities and n Wiener processes, it is possible to form
an instantaneously riskless portfolio, Π, using the securities. Define kj as the amount of
the jth security in the portfolio, so that
n+1
X
Π= kj fj (2)
j=1

The kj must be chosen so that the stochastic components of the returns from the securities
are eliminated. From equation (1) this means that

n+1
X
kj σij fj = 0 (3)
j=1

for 1 ≤ i ≤ n. The return from the portfolio is then given by

n+1
X
dΠ = kj µj fj dt
j=1

The cost of setting up the portfolio is

n+1
X
kj fj
j=1

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
This is not restrictive. A non-dividend-paying security can always be obtained from
a dividend-paying security by reinvesting the dividends in the security.

1
If there are no arbitrage opportunities, the portfolio must earn the risk-free interest rate,
so that
n+1
X n+1
X
kj µj fj = r kj fj (4)
j=1 j=1
or
n+1
X
kj fj (µj − r) = 0 (5)
j=1

Equations (3) and (5) can be regarded as n + 1 homogeneous linear equations in the kj ’s.
The kj ’s are not all zero. From a well-known theorem in linear algebra, equations (3) and
(5) can be consistent only if for all j
n
X
fj (µj − r) = λi σij fj (6)
i=1
or
n
X
µj − r = λi σij (7)
i=1

for some λi (1 ≤ i ≤ n) that are dependent only on the state variables and time. Dropping
the j subscript, this show that for any security f dependent on the n stochastic variables
n
X
df = µf dt + σi f dzi
i=1

where
n
X
µ−r = λi σi
i=1

This proves the required result.

2
 Technical Note No. 31*
Options, Futures, and Other Derivatives
John Hull

Properties of Ho–Lee and Hull–White Interest Rate Models

This note presents some of the math underlying the Ho–Lee and Hull–White one-factor
models of the term structure. It follows the approach in Hull and White (1993).1
In a one-factor term structure, model the process for a zero-coupon bond price in the
traditional risk-neutral world must have a return equal to the short rate r. Suppose that
v(t, T ) is the volatility. Then:

dP (t, T ) = rP (t, T )dt + v(t, T )P (t, T )dz (1)

In this note, we will assume that v is a function only of t and T . Because the bond’s price
volatility declines to zero at maturity v(t, t) = 0.
From Ito’s lemma, for any times T1 and T2 with T2 > T1

v(t, T1 )2
 
d ln P (t, T1 ) = r − dt + v(t, T1 )dz(t) (2)
2

v(t, T2 )2
 
d ln P (t, T2 ) = r − dt + v(t, T2 )dz(t) (3)
2
Define f (t, T1 , T2 ) as the forward rate for the period between time T1 and T2 as seen at
time t
ln P (t, T2 ) − ln P (t, T1 )
f (t, T1 , T2 ) = −
T2 − T1
From equations (2) and (3)

v(t, T2 )2 − v(t, T1 )2
   
v(t, T2 ) − v(t, T1 )
df (t, T1 , T2 ) = dt − dz(t)
2(T2 − T1 ) T2 − T1

Define R(t, T ) as the zero rate for the period between t and T .
Z T
R(t, T ) = f (0, t, T ) + df (τ, t, T )
0

so that
Z t Z t
v(τ, T )2 − v(τ, t)2
 
v(τ, T ) − v(τ, t)
R(t, T ) = f (0, t, T ) + dτ − dz(τ ) (4)
0 2(T − t) 0 T −t

* c Copyright John Hull. All Rights Reserved. This note may be reproduced for use in
conjunction with Options, Futures, and Other Derivatives by John C. Hull.
1
See J. Hull and A. White , “Bond Option Pricing Based on a Model for the Evolution
of Bond Prices,” Advances in Futures and Options Research, 6 (1993), 1–13.

1
As T approaches t, R(t, T ) becomes r(t) and f (0, t, T ) becomes the instantaneous forward
rate, F (0, t) so that
t t
∂ v(τ, t)2
Z Z
∂
r(t) = F (0, t) + dτ − v(τ, t) dz(τ )
0 ∂t 2 0 ∂t
or Z t Z t
r(t) = F (0, t) + v(τ, t)vt (τ, t)dτ − vt (τ, t)dz(τ ) (5)
0 0

where subscripts denote partial derivatives. To calculate the process for r we differentiate
with respect to t. Because v(t, t) = 0, this gives
 Z t Z t 
2
dr = Ft (0, t) + [v(τ, t)vtt (τ, t) + vt (τ, t) ]dτ − vtt (τ, t)dz(τ ) dt − vt (τ, t)|τ =t dz(t)
0 0
(6)

Case 1: Ho–Lee; v(t, T ) = σ(T − t)

In the case, equation (5) gives
Z t
2 2
r(t) = F (0, t) + σ t /2 − σ dz(τ ) (7)
0

and equation (6) gives

dr(t) = [Ft (0, t) + σ 2 t]dt + σ dz
This is the Ho-Lee model
dr = θ(t) dt + σ dz
We have proved the equation for θ(t)

θ(t) = Ft (0, t) + σ 2 t

Also from equation (4)

Z t
2
R(t, T ) = f (0, t, T ) + σ tT /2 − σ dz(τ ) (8)
0

From equations (7) and (8)

R(t, T ) = f (0, t, T )+σ 2 tT /2+r(t)−F (0, t)−σ 2 t2 /2 = f (0, t, T )−F (0, t)+σ 2 t(T −t)/2+r(t)

Because
ln P (t, T ) = −R(t, T )(T − t)
It follows that

ln P (t, T ) = −f (0, t, T )(T − t) + F (0, t)(T − t) − σ 2 t(T − t)2 /2 − r(t)(T − t)

2
The forward bond price P (0, T )/P (0, t) equals e−f (0,t,T )(T −t) so that this becomes

P (0, T )
ln P (t, T ) = ln + F (0, t)(T − t) − σ 2 t(T − t)2 /2 − r(t)(T − t)
P (0, t)

This proves:
P (t, T ) = A(t, T )e−r(t)(T −t)
where
P (0, T ) 1
ln A(t, T ) = ln + F (0, t)(T − t) − σ 2 t(T − t)2
P (0, t) 2

Case 2: Hull–White; v(t, T ) = σ(1 − e−a(T −t) )/a

In this case, equation (5) gives
t
σ2 σ2
Z
r(t) = F (0, t) + 2 (1 − e−at ) − 2 (1 − e−2at ) − σe−a(t−τ ) dz(τ ) (9)
a 2a 0

Equation (6) gives

t
σ 2 −at
 Z 
dr(t) = Ft (0, t) + (e − e−2at ) + σae−a(t−τ )
dz(τ ) dt − σ dz(t) (10)
a 0

Substituting for Z t
σe−a(t−τ )dz(τ )
0

from equation (9) into equation (10) we obtain

σ 2 −at σ2 σ2
 
−2at −at −2at
dr(t) = Ft (0, t) + (e −e ) − ar(t) + aF (0, t) + (1 − e ) − (1 − e ) dt−σ dz(t)
a a 2a
or
σ2
 
−2at
dr(t) = Ft (0, t) + aF (0, t) − ar(t) + (1 − e ) dt − σ dz(t)
2a
This is the Hull–White model

dr(t) = (θ(t) − ar) dt + σ dz

with
σ2
θ(t) = Ft (0, t) + aF (0, t) + (1 − e−2at )
2a
From equation (4)

σ 2 [e−2a(T −t) − e−2aT − 1 + e−2at − 4e−a(T −t) + 4e−aT + 4 − 4e−at ]

R(t, T ) = f (0, t, T ) +
4a3 (T − t)

3
 t
σ(e−aT − e−at )
Z
+ eaτ dz(τ ) (11)
a(T − t) 0

From equation (9)

t
σ 2 at σ 2 at
Z
σ eaτ dz(τ ) = −r(t)eat + F (0, t)eat + (e − 1) − (e − e−at )
0 a2 2a2

so that

σ 2 [e−2a(T −t) − e−2aT − 1 + e−2at − 4e−a(T −t) + 4e−aT + 4 − 4e−at ]

R(t, T ) = f (0, t, T ) +
4a3 (T − t)

(e−aT − e−at ) σ 2 at σ 2 at
 
at at −at
+ −r(t)e + F (0, t)e + 2 (e − 1) − 2 (e − e )
a(T − t) a 2a
Now
ln P (t, T ) = −R(t, T )(T − t)
and the forward bond price P (0, T )/P (0, t) equals e−f (0,t,T )(T −t) . After some tedious
algebra we get

P (0, T ) 1
ln P (t, T ) = ln + F (0, t)B(t, T ) − 3 σ 2 (e−aT − e−at )2 (e2at − 1) − r(t)B(t, T )
P (0, t 4a

where
1 − e−a(T −t) )
B(t, T ) =
a
showing that
P (t, T ) = A(t, T )e−B(t,T )r
where

P (0, T ) 1
ln A(t, T ) = ln + F (0, t)B(t, T ) − 3 σ 2 (e−aT − e−at )2 (e2at − 1)
P (0, t 4a

4
Bond Options
Consider a European option with strike price K and maturity T on a zero-coupon
bond where the maturity of the bond is s. The forward price of the bond underlying the
option as seen at time t, FB (t, T, s), is

P (t, s)
FB (t, T, s) =
P (t, T )

Using the results in equations (2) and (3) we get

v(t, T )2 − v(t, s)2

d ln FB (t, T, s) = dt + [v(t, s) − v(t, T )] dz
2

This shows that the P (T, s) = fB (T, T, s) is lognormal when v(t, T ) is function only of t
and T . The variance ln P (T, s) is then
Z T
σP2 = [v(t, s) − v(t, T )]2 dt
0

In the case of Ho-Lee v(t, T ) = σ(T − t) and σP2 = σ 2 (s − T )2 T . In Hull-White

v(t, T ) = σB(t, T ) so that

T
σ2
Z
σP2 =σ 2
[B(t, s) − B(t, T )]2 dt = [1 − e−a(s−T ) ]2 (1 − e−2aT )
0 2a3

In both cases bond options can be valued using Black’s model. The average variance rate
of the forward bond price is σP2 /T . This leads to the results for bond options in the text.