MAFS5250 – Computational Methods for Pricing Structured Prod-

ucts

Topic 1 – Lattice tree methods

1.1 Binomial option pricing models

• From replication to risk neutral valuation

• Continuous limit of the binomial model

• Multiperiod extension

• Dynamic programming procedure

• Estimating delta and other Greek letters

• Discrete dividend models

• Pricing of lookback options

- Hull-White scheme

- Cheuk-Vorst algorithms
1
1.2 Trinomial schemes

• Discounted expectation approach

• Multistate extension – Ritchken-Kamrad’s approach

1.3 Forward shooting grid algorithms (strongly path dependent options)

• Cumulative Parisian feature of knockout provision

• Call options with strike reset feature

• Floating strike arithmetic averaging call

• Alpha-quantile options

• Accumulators

2
1.1 Binomial option pricing models

From replication to risk neutral valuation

Discrete model of the dynamics of the underlying price process

Under the binomial random walk model, the asset price after one period
∆t will be either uS or dS with probability q and 1 − q, respectively.

We assume u > 1 > d so that uS and dS represent the up-move and down-
move of the asset price, respectively. The proportional jump parameters
u and d will be related to the asset price dynamics.

Let R denote the growth factor of riskless investment over one period so
that $1 invested in a riskfree money market account will grow to $R after
one period. In order to avoid riskless arbitrage opportunities, we must
have u > R > d.

3
Construction of a replicating portfolio

By buying the asset and borrowing cash (in the form of riskfree money
market account) in appropriate proportions, one can replicate the position
of a call.

Suppose we form a portfolio which consists of α units of asset and cash

amount M in the form of riskless investment (money market account).
After one period of time △t, the value of the portfolio becomes
{
αuS + RM with probability q
αdS + RM with probability 1 − q.

We have the fortunate coincidence: 2 investment instruments: risky asset

and money market account and 2 states of the world: up and down in
the binomial model. Under this special scenario, the replication approach
of deriving the binomial pricing formula works.

4
The portfolio is used to replicate the long position of a call option on a
non-dividend paying asset.

As there are two possible states of the world: asset price goes up or down,
the call price is dependent on the asset price, thus it is a contingent claim.

Suppose the current time is only one period △t prior to expiration. Let c
denote the current call price, and cu and cd denote the call price after one
period (which is the expiration time in the present context) corresponding
to the up-move and down-move of the asset price, respectively.

5
Let X denote the strike price of the call. The payoff of the call at expiry
is given by
{
cu = max(uS − X, 0) with probability q
cd = max(dS − X, 0) with probability 1 − q.

Evolution of the asset price S and the money market account M after one
time period under the binomial model. The risky asset value may either
go up to uS or go down to dS, while the riskless investment amount M
grows to RM with certainty.

6
Replicating procedure

The above portfolio containing the risky asset and money market account
is said to replicate the long position of the call if and only if the values of
the portfolio and the call option match for each possible outcome, that
is,
αuS + RM = cu and αdS + RM = cd.
Solving the pair of equations, we obtain
cu − cd ucd − dcu
α= ≥ 0, M = ≤ 0.
(u − d)S (u − d)R

Apparently, we have 2 states of the world that generate 2 equations

via matching the outcomes. There are two unknowns α and M to be
determined, so we have equal number of states and unknowns.

• The uninteresting case occurs when cu = cd = 0. This leads to

α = M = 0. If the call is surely to be out-of-the-money under the two
possible states, then its present value is zero.
7
 • Since α is always non-negative and M is always non-positive, the
replicating portfolio involves buying the asset and borrowing cash in
the corresponding proportions (excluding the degenerate case of α =
M = 0).
• The number of units of asset held is seen to be the ratio of the
difference of call values cu −cd to the difference of asset values uS −dS.
This is called the hedge ratio.

The call option can be replicated by a portfolio of the two basic securities:
risky asset and riskfree money market account. By invoking the law of one
price, the call value is identical to the value of the replicating portfolio.

Query : Can we adopt the above replicating procedure if the discrete

asset price process follows the trinomial random walk model
(3 states of the world in the next move)? One has to use
the risk neutral valuation principle for deriving the trinomial
pricing formula, which holds under the assumption of absence
of arbitrage. Recall that absence of arbitrage ⇔ existence of
risk neutral measure under discrete models.
8
Binomial option pricing formula

The current value of the call is given by the current value of the replicating
portfolio, that is,
R−d c + u−R c
u−d u u−d d
c = αS + M =
R
pcu + (1 − p)cd R−d
= where p= .
R u−d

Note that the probability q, which is the subjective probability of up-move

or down-move of the asset price, does not appear in the call value.

The parameter p can be shown to be 0 < p < 1 since u > R > d and so p
can be interpreted as a probability.

9
Risk neutral pricing measure

From the relation

R−d u−R
puS + (1 − p)dS = uS + dS = RS,
u−d u−d
one can interpret the result as follows: the expected rate of returns on
the asset with p as the probability of upside move is just equal to the
riskless interest rate. In other words, we observe
1 ∗ ∆t
S= E [S |S],
R
where E ∗ is expectation under this probability measure. We may view p
as the risk neutral probability that the asset price goes up in the next
move.
[  ]
∗ S 
∆t
Since E S equals the current asset value S, we say that the dis-
R 
counted asset value process is a martingale under the risk neutral pricing
measure.

10
Discounted expectation of the terminal payoff

The call price can be interpreted as the expectation of the payoff of the
call option at expiry under the risk neutral probability measure discounted
at the riskless interest rate.

The binomial call value formula can be expressed as

1 ∗ ∆t
c = E [c |S],
R
where c denotes the call value at the current time, and c∆t denotes the
random variable representing the call value one period later.

Besides applying the principle of replication of claims, the binomial option

pricing formula can also be derived using the riskless hedging principle
(similar to the derivation of the continuous Black-Scholes equation) or
finding the state prices of the up-state and down-state.

11
Determination of the jump parameters (u and d) that respects the asset
price dynamics

• The jump parameters are related to the variance of the continuous

asset value process under the risk neutral measure.
• For the continuous asset price dynamics of Geometric Brownian mo-
tion under the risk neutral measure, we have
( )
σ2
d ln St = r − dt + σ dZt
2
( )
St+△t σ2
so that ln becomes normally distributed with mean r − △t
St 2
and variance σ 2△t, where r is the riskless interest rate and σ 2 is the
variance rate.
• Expressed in the form of the exponentiation of a normal random vari-
St+△t 2 △t
able, the mean and variance of 2
are R and R (e σ − 1), re-
St
spectively, where R = er△t.

12
 • For the one-period binomial option model under the risk neutral mea-
St+△t
sure, the mean and variance of the asset price ratio are
St
pu + (1 − p)d and pu2 + (1 − p)d2 − [pu + (1 − p)d]2,
respectively.

• By equating the mean and variance of the asset price ratio in both
the continuous and discrete models, we obtain
S ∆t
E[ ] = pu + (1 − p)d = R
S
S ∆t 2 S ∆t 2
]} = pu2 + (1 − p)d2 − R2 = R2(eσ △t − 1).
2
E[( ) ] − {E[
S S
R−d
The first equation leads to p = , the usual risk neutral probability.
u−d

We have 3 unknowns: u, d and p, but only two equations. How to find the
third condition? It is superfluous to match moments beyond the second
order since the mean and variance fully specify a normal random variable.

13
A convenient choice of the third condition is the tree-symmetry condition
1
u= ,
d
so that the lattice nodes associated with the binomial tree are symmet-
rical.

e 2 = R2eσ △t, the solution is found to be

2
Writing σ
√
1 e2 + 1 +
σ e 2 + 1)2 − 4R2
(σ R−d
u= = , p= .
d 2R u−d

How to obtain a nice approximation function to u instead of using the

above daunting expression?
√
By expanding u in Taylor series in powers of △t, we obtain
√ σ2 4r2 + 4σ 2r + 3σ 4 3
u = 1 + σ △t + △t + △t 2 + O(△t2).
2 8σ

Observe that
√
the first three terms in the above Taylor series agree with
those of eσ △t up to O(△t) term.
14
This suggests the judicious choice of the following set of parameter values
√ √ R−d
u = e △t,
σ d=e−σ △t , p= .
u−d

St+△t
With this new set of parameters, the variance of the price ratio in
St
the continuous and discrete models agree up to O(△t).

We define xt = ln St. If we write ∆x = ln St+∆t − ln St as the discrete

change√in the log asset price over ∆t, then the proportional upward jump
σ ∆t
√
u=e is equivalent to ∆x = ln u = σ ∆t. This is consistent with the
Brownian increment ∆x of the log asset price (without drift rate though)
over the differential time interval ∆t. Note that the increment due to
σ2
drift rate r − over ∆t is O(∆t) while the random Brownian increment
2 ( )
√ σ 2
is O( ∆t). The higher order term r − ∆t arising from the drift can
√ 2
be dropped compared to O( ∆t) term arising from Brownian diffusion.

15
Continuous limit of the binomial model

We consider the asymptotic limit △t → 0 of the binomial formula

c = [pc∆t + (1 − p)c ∆t ] e−r△t .

u d
In the continuous analog, the binomial formula can be written as

c(S, t − △t) = [pc(uS, t) + (1 − p)c(dS, t)] e−r△t.

Instead of choosing c(uS, t+∆t) and c(dS, t+∆t) in the formula, the above
form is more convenient for subsequent analytic derivation since cross
derivative terms do not appear in the later Taylor expansion procedure.

Assuming sufficient continuity of c(S, t), we perform the Taylor expansion

of the binomial scheme at (S, t) as follows:

16
 −c(S, t − △t) + [pc(uS, t) + (1 − p)c(dS, t)]e−r△t
∂c 1 ∂ 2c 2 + · · · − (1 − e−r△t )c(S, t)
= (S, t)△t − (S, t)△t
∂t { 2 ∂t2
−r△t ∂c
+ e [p(u − 1) + (1 − p)(d − 1)]S (S, t)
∂S
1 2
∂ c
+ [p(u − 1)2 + (1 − p)(d − 1)2]S 2 2 (S, t)
2 ∂S }
1 3
∂ c
+ [p(u − 1)3 + (1 − p)(d − 1)3]S 3 3 (S, t) + · · · .
6 ∂S
First, we observe that

1 − e−r△t = r△t + O(△t2),

and also

e−r△t [p(u − 1) + (1 − p)(d − 1)] = r△t + O(△t2),

e−r△t [p(u − 1)2 + (1 − p)(d − 1)2] = σ 2△t + O(△t2),
e−r△t [p(u − 1)3 + (1 − p)(d − 1)3] = O(△t2).

17
Combining the results, we obtain

−c(S, t − △t) + pc(uS, t) + (1 − p)c(dS, t) e−r△t

[ [ ] ]
∂c ∂c 2
σ 2 ∂ c2
= (S, t) + rS (S, t) + S (S, t) − rc(S, t) △t + O(△t 2 ).
∂t ∂S 2 ∂S 2

Since c(S, t) satisfies the binomial formula, so we obtain

∂c ∂c σ 2 2 ∂ 2c
0 = (S, t) + rS (S, t) + S 2
(S, t) − rc(S, t) + O(△t).
∂t ∂S 2 ∂S

In the limit ∆t → 0, the binomial call value c(S, t) satisfies the Black-
Scholes equation.

18
Multiperiod extension

Let cuu denote the call value at two periods beyond the current time with
two consecutive upward moves of the asset price and similar notational
interpretation for cud and cdd. The call values cu and cd are related to cuu,
cud and cdd as follows:
pcuu + (1 − p)cud pcud + (1 − p)cdd
cu = and cd = .
R R

The call value at the current time which is two periods from expiry is
found to be
p2cuu + 2p(1 − p)cud + (1 − p)2cdd
c= 2
,
R
where the corresponding terminal payoff values are given by

cuu = max(u2S − X, 0), cud = max(udS − X, 0), cdd = max(d2S − X, 0).

19
The coefficients p2, 2p(1 − p) and (1 − p)2 represent the respective risk
neutral probability of having two up jumps, one up jump and one down
jump, and two down jumps in the two consecutive moves of the binomial
process.

Dynamics of asset price and call price in a two-period binomial model

20
With n binomial steps, the risk neutral probability of having j up jumps
and n − j down jumps is given by Cjnpj (1 − p)n−j , j = 0, 1, . . . , n, where
n n!
Cj = is the binomial coefficient.
j!(n − j)!

The corresponding terminal payoff when j up jumps and n − j down jumps

occur is seen to be max(uj dn−j S − X, 0).

By the risk neutral valuation principle, the call value obtained from the
n-period binomial model is given by
n
∑
Cjnpj (1 − p)n−j max(uj dn−j S − X, 0)
j=0
c= .
Rn

21
How to get rid of the “max” function in the option type payoff function?

We define k to be the smallest non-negative integer such that uk dn−k S ≥

X
ln Sd n
X, that is, k ≥ . It is seen that
ln ud
{
0 when j<k
max(uj dn−j S − X, 0) = .
uj dn−j S − X when j≥k
The integer k gives the minimum number of upward moves required for
the asset price in the multiplicative binomial process in order that the call
expires in-the-money.

The call price formula can be simplified as

n
∑ n
∑
n j n−j uj dn−j −n npj (1 − p)n−j .
c=S Cj p (1 − p) n
− XR C j
j=k
R j=k

22
Interpretation of the call price formula

The last term in above equation can be interpreted as the expectation

value of the payment made by the holder at expiration discounted by the
n
∑
factor R−n, and Cjnpj (1 − p)n−j is seen to be the probability under the
j=k
risk neutral measure that the call expires in-the-money.

The above probability is related to the complementary binomial distribu-

tion function defined by
n
∑
Φ(n, k, p) = Cjnpj (1 − p)n−j .
j=k

Note that Φ(n, k, p) gives the probability for achieving at least k successes
in n trials of a binomial experiment, where p is the probability of success
in each trial.

23
 up d(1 − p)
Further, if we write p′ = so that 1 − p′ = , then the call price
R R
formula for the n-period binomial model can be expressed as

c = SΦ(n, k, p′) − XR−nΦ(n, k, p).

• The first term gives the discounted expectation of the terminal asset
price given that the call expires in-the-money.

• The second term gives the present value of the expected cost incurred
by exercising the call.

• In the replicating portfolio, we require long holding of Φ(n, k, p′) units

of the underlying asset and short holding of XR−nΦ(n, k, p) dollars of
the money market account.

• The parameter n is related to time to expiry. The another parameter

k is related to expectation of being in-the-money at expiry, which
exhibits implicit dependence on volatility (via u and d) and strike price
X.

24
Mathematical representation

The call price for the n-period binomial model can be expressed as the
discounted expectation of the terminal payoff under the risk neutral mea-
sure
1 1
c = n E ∗ [cT ] = n E ∗ [max(ST − X, 0)] , T = t + n∆t,
R R
where cT is the terminal payoff, max(ST − X, 0), of the call at expiration
1
time T and n is the discount factor over n periods. That is,
R
1 ∗
SΦ(n, k, p′) = E [ST 1{ST >X}]
Rn
Φ(n, k, p) = E ∗[1{ST >X}] = P ∗[ST > X].
The expectation operator E ∗ is taken under the risk neutral measure
rather than the subjective probability measure associated with the actual
(physical) asset price process.

25
Dynamic programming procedure for pricing an American option

How to price the early exercise premium in an American option?

Without the early exercise privilege, risk neutral valuation principle leads
to the usual binomial formula
∆t + (1 − p)V ∆t
pV u
Vcont = d .
R

To reflect the optimal decision of either continuing to hold the Ameri-

can option or exercising the option, the following dynamic programming
procedure is applied at each binomial node

V = max(Vcont, h(S)),
where h(S) is the exercise payoff when the asset price assumes the value
S. The stochastic optimization of the optimal stopping rule (early exer-
cise feature) associated with an American option can be realized by the
dynamic programming procedure applied at each binomial node.

26
American put option

The intrinsic value of a vanilla put option is X − S n

j at the (n, j) node,
where X is the strike price. Here, n is the number of time steps from
the tip of the binomial tree and j is the number of up-moves among the
n steps. The dynamic programming procedure applied at each node is
given by
 
n+1 n+1
n  pP j+1 + (1 − p)P j 
P j = max  , X − Sn
j,
R

where n = N − 1, · · · , 0, and j = 0, 1, · · · , n. Here, N is the total number

of time steps in the binomial tree.

27
Example 1 – Pricing an American put option

Consider a 5-month American put option on a non-dividend-paying stock

when the stock price is $50, the strike price is $50, the risk-free interest
rate is 10% per annum, and the volatility is 40% per annum. That is,
S = 50, X = 50, r = 0.10, σ = 0.40, T = 0.4167.

Suppose that we divide the life of the option into five intervals of length
of 1 month (= 0.0833 year) for the purpose of constructing a binomial
tree.

With ∆t = 0.0833, we have

√ √
u = eσ ∆t = 1.1224, d = e−σ ∆t = 0.8909, R = er∆t = 1.0084,
R−d
p= = 0.5073, 1 − p = 0.4927.
u−d

28
At each node:

Upper value = Underlying Asset Price

Lower value = Option Price

Shading indicates the node at which the option is exercised

Strike price = 50

Discount factor per step = 1/R = e−r∆t = 0.9917

Time step, ∆t = 0.0833 years, 30.42 days

Growth factor per step, R = 1.0084

Risk neutral probability of up-move, p = 0.5073

Proportional up-jump factor, u = 1.1224

Proportional down-jump factor, d = 1/u = 0.8909

29
30
 • The stock price at the j th node (j = 0, 1, · · · , n) at time n∆t (n =
0, 1, · · · , 5) is calculated as S0uj dn−j . For example, the stock price at
node A (n = 4, j = 1) (i.e., the second node up at the end of the
fourth time step) is 50 × 1.1224 × 0.89093 = $39.69.

• The option prices at the final nodes are calculated as max(X − ST , 0).
For example, the option price at node G is 50.00 − 35.36 = 14.64.

Backward induction procedure

• First, we assume no exercise of the option at the nodes. This means

that the option price is calculated as the present value of the expected
option price one time step later. For example, at node E, the option
price is calculated as

(0.5073 × 0 + 0.4927 × 5.45)e−0.10×0.0833 = 2.66

whereas at node A it is calculated as

(0.5073 × 5.45 + 0.4927 × 14.64)e−0.10×0.0833 = 9.90.

31
Check to see if early exercise is preferable to waiting

• At node E, early exercise would give a value for the option of zero
because both the stock price and strike price are $50. Clearly it is
best to wait. The correct value for the option at node E, therefore,
is $2.66.

• At node A, it is a different story. If the option is exercised, it is worth

$50.00 − $39.69, or $10.31. This is more than $9.90. If node A is
reached, then the option should be exercised and the correct value
for the option at node A is $10.31.

• Option prices at earlier nodes are calculated in a similar way. Note

that it is not always best to exercise an option early even when it is
in the money at that node.

32
• Consider node B, the American put is in-the-money since the asset
price $39.69 is below $50. If the option is exercised, it is worth
$50.00 − $39.69, or $10.31. However, if it is held, it is worth

(0.5073 × 6.38 + 0.4927 × 14.64)e−0.10×0.0833 = 10.36.

The option should not be exercised at this node, and the correct
option value at the node is $10.36.

• Working back through the tree, the value of the option at the initial
node is $4.49. This is our numerical estimate for the option’s current
value.

• In practice, a smaller value of ∆t, and many more nodes, would be

used. With 30, 50, 100, and 500 time steps we obtain values for the
option of 4.263, 4.272, 4.278, and 4.283.

33
Convergence of the price of the option with respect to increasing number
of time steps

The convergence trend is oscillatory. Overshooting the theoretical true

value at the current choice of the time step becomes undershooting when
the number of time steps is increased by one.
34
 ∗ (τ )
Early exercise boundary SP

The optimal exercise policy is characterized by the early exercise curve

∗ (τ ), where the American put option should be exercised when the
SP
∗ (τ ) for a given time
stock price falls below the critical threshold value SP
to expiry τ .

continuation
region

exercise
*
SP (W )
region

35
• The numerical approximation of SP ∗ (τ ) can be deduced from the bi-

nomial tree calculations by recording the node points at which the

American put value assumes the exercise payoff.

• The early exercise boundary is approximated by taking the (arithmetic

or geometric) average of the asset prices at neighboring nodes at the
same time level at which continuation value is taken at the upper
node while exercise value is taken at the lower node.

• As a numerical example (refer to the numerical results shown on

p.29), we approximate the early exercise boundary at t = 0.25, t =
44.55+35.36 √ 50.00+39.69
0.3333 and t = 0.4167 as or 44.55 × 35.36,
√ 56.12+44.55
2√ 2
or 50.00 × 39.69, 2 or 56.12 × 44.55, respectively.

36
Callable American call – game option between the issuer and holder

• The callable feature entitles the issuer to buy back the American
option at any time at a predetermined call price.
• Upon call, the holder can choose either to exercise the call or receive
the call price as cash.
• Let the call price be K. The dynamic programming procedure applied
at each node to model the game between the issuer and holder can
be constructed as follows:
  
n+1 n+1
n   pCj+1 + (1 − p)Cj 
Cj = min max  , Sjn − X  ,
R


max(K, Sjn − X) .

37
Justification of the dynamic programming procedure

 
n+1
pCn+1 + (1 − p)Cjn+1
• The first term max  , Sjn − X  represents the
R
optimal strategy of the holder, given no call of the option by the
issuer.
• Upon call by the issuer, the payoff is given by the second term
max(K, Sjn − X) since the holder can either receive cash amount K or
exercise the option.
• From the perspective of the issuer, he chooses to call or restrain
from calling so as to minimize the option value with reference to the
possible actions of the holder. The value of the callable call is given
by taking the minimum value of the above two terms.

38
Recall the well known distributive rule: αx + αy = α(x + y). In the
current context, we may treat “taking max” as multiplication and “taking
min” as addition. An equivalent dynamic programming procedure can be
constructed as follows:
  
n+1 n+1
n  n  pCj+1 + (1 − p)Cj 
Cj = max Sj − X, min  , K  .
R

• From financial intuition, the option will be called when the contin-
uation value is above the call price K. Independent of whether the
option to be either called or not called, the holder can always choose
to exercise to receive Sjn − X if the exercise payoff has a higher value.

39
Estimating delta and other Greek letters

• The delta (∆) of an option is the rate of change of its price with
respect to the underlying stock price. It can be calculated as
∆f
∆S
where ∆S is a small change in the stock price and ∆f is the corre-
sponding small change in the option price.

• At time ∆t, we have an estimate f11 for the option price when the
stock price is S0u and an estimate f10 for the option price when the
stock price is S0d.

• When ∆S = S0u − S0d, ∆f = f11 − f10. Therefore an estimate of

delta at time ∆t is
f − f10
∆ = 11 .
S0 u − S0 d

40
Gamma calculations

To determine gamma (Γ), note that we have two estimates of ∆ at time

2∆t.

When S = (S0u2 + S0)/2 (halfway between the second and third node),
delta is (f22 − f21)/(S0u2 − S0); when S = (S0 + S0d2)/2 (halfway between
the first and second node), delta is (f21 − f20)/(S0 − S0d2).

The difference between the two values of S is h, where

h = 0.5(S0u2 − S0d2).

Gamma is the change in delta divided by h:

[(f22 − f21)/(S0u2 − S0)] − [(f21 − f20)/(S0 − S0d2)]
Γ= .
h

41
Theta calculations

Theta is the rate of change of the option price with time when all else is
kept constant. If the tree starts at time zero, an estimate of theta is
f − f00
Θ = 21 .
2∆t
Note that f21 is the option value at two time steps from time zero and
with the same asset price.

Vega calculations

Vega can be calculated by making a small change, ∆σ, in the volatility

and constructing a new tree to obtain a new value of the option. The
time step ∆t should be kept the same. The estimate of vega is
f∗ − f
ν= ,
∆σ
where f and f ∗ are the estimates of the option price from the original
and the new tree, respectively.

42
Example 2

• Consider again Example 1. We have f1,0 = 6.96 and f1,1 = 2.16. An

estimate for delta is given by
2.16 − 6.96
= −0.41.
56.12 − 44.55
• An estimate of the gamma of the option can be obtained from the
values at nodes B, C, and F as
[(0.64 − 3.77)/(62.99 − 50.00)] − [(3.77 − 10.36)/(50.00 − 39.66)]
= 0.03.
11.65
• An estimate of the theta of the option can be obtained from the
values at nodes D and C as
3.77 − 4.49
= −4.3 per year
0.1667
or −0.012 per calendar day.

• These are only rough estimates. They become progressively better

as the number of time steps on the tree is increased.
43
Discrete dividend models

Let S be the asset price at the current time which is n△t from expiry,
and suppose a discrete dividend of amount D is paid at time between one
time step and two time steps from the current time.

Consider the naive construction of the binomial tree. The nodes in the
binomial tree at two time steps from the current time would correspond
to asset prices
u2S − D, S−D and d2S − D,
since the asset price drops by the same amount as the dividend right after
the dividend payment.

44
• Extending one time step further, there will be six nodes

(u2S − D)u, (u2S − D)d, (S − D)u, (S − D)d, (d2S − D)u, (d2S − D)d
instead of four nodes as in the usual binomial tree without discrete
dividend.
• This is because (u2S − D)d ̸= (S − D)u and (S − D)d ̸= (d2S − D)u,
so the interior nodes do not recombine.
• In general, suppose a discrete dividend is paid in the future between
(k −1)th and kth time step, then at the (k +m)th time step, the number
of nodes would be (m + 1)(k + 1) rather than k + m + 1 nodes as in
the usual reconnecting binomial tree.

45
 Binomial tree with single discrete dividend

In this pictorial representation, we have m = n = 2, so that there are

(2 + 1)(2 + 1) = 9 nodes after 4 time steps.

46
Splitting the asset price into the deterministic dividends component and
risky component

• Splitting the asset price St into two parts: the risky component Set
that is stochastic and the remaining part that will be used to pay the
discrete dividend (assumed to be deterministic) in the future.
• Suppose the dividend date is t∗, then at the current time t, the risky
component Set is given by
{ ∗ −t)
St − De −r(t , t < t∗
Set =
St, t > t∗ .

• Let σ
e denote the volatility of S e and assume σ e to be constant rather
t
than the volatility of St itself to be constant.

47
• Assume that a discrete dividend D is paid at time t∗, which lies be-
tween the kth and (k + 1)th time step.
• At the tip of the binomial tree, the risky component Se is related to
the asset price S by
Se = S − De−kr∆t.

• The total value of asset price at the (n, j)th node, which corresponds
to n time steps from the tip and j upward jumps, is given by
e j dn−j + De−(k−n)r∆t1
Su {n≤k} ,
n = 1, 2, · · · , N and j = 0, 1, · · · , n.

48
A reconnecting binomial tree with single discrete dividend D

Here, N = 4 and k = 2, and let Se denote the risky component of the

asset value at the tip of the binomial tree. The asset value at nodes P, Q
and R are Se + De−2r∆t, Sue + De−r∆t and Sd,e respectively.

49
Example 3

Consider a 5-month American put option on a stock that is expected to

pay a single dividend of $2.06 during the life of the option. The initial
stock price is $52, the strike price is $50, the risk-free interest rate is 10%
per annum, the volatility is 40% per annum, and the ex-dividend date is
1
in 3 months (which is 0.2917 years).
2

We construct a tree to model Se (risky component of the asset price

process), the stock price less the present value of future dividends during
the life of the option. At time zero, the present value of the dividend is

2.06e−0.2917×0.1 = 2.00.
The initial value of Se is therefore 50.00.

50
 • Assuming that the 40% per annum volatility refers to S, e the earlier
figure on P.30 provides a binomial tree for S. e Adding the present
value of the dividend at each node leads to the figure on P.53, which
is a binomial model for S.

• The probabilities at the nodes are 0.5073 for an up movement and

0.4927 for a down movement. Working back through the tree in the
usual way gives the option price as 4.44.

Remark The exercise payoff is calculated using the actual asset price S,
e
not the risky component S.

Note that the lowest node at time 0.25 on P.53 has the put option value
equals 14.22, which is not equal to the exercise payoff (see the binomial
tree on P.30 for comparison). This result corresponds to the higher asset
value, which is 37.41 on P.53 instead of 35.36 on P.30 for the same
binomial node. The exercise payoff of the put option at S = 37.41 and
K = 50 is 12.59, which is less than the continuation value of 14.22.

51
At each node:

Upper value = Underlying Asset Price

Lower value = Option Price

Shading indicates where option is exercised

Strike price = 50

Discount factor per step = 1/R = 0.9917

Time step, ∆t = 0.0833 years, 30.42 days

Growth factor per step, R = 1.0084

Risk neutral probability of up move, p = 0.5073

Proportional up jump factor, u = 1.1224

Proportional down jump factor, d = 1/u = 0.8909

52
53
Tree when stock pays a known dividend yield at one particular time. The
dividend amount is equal to δ times the prevailing asset price. In this
case, the interior nodes do recombine. Here, δ is the dividend yield.

54
Pricing of lookback options

A path-dependent derivative is a derivative where the payoff depends on

the path followed by the price of the underlying asset, not just its final
value. Two important properties:

1. The payoff from the derivative must depend on a single function, F ,

of the path followed by the underlying asset.

2. It must be possible to calculate the updated value of F at time t + ∆t

from the known value of F at time t and the updated value of the
underlying asset at time t + ∆t.

For example, the realized maximum of a discrete asset price process over
successive time steps is given by
max = max(S max , S
Si+1 i i+1), i = 1, 2, . . . , n.

55
Research papers on numerical methods on lookback options

1. Hull, J. and A. White, “Efficient procedures for valuing European

and American path-dependent options,” Journal of Derivatives (Fall,
1993), p.21-31.

2. Cheuk, T.H.F. and T.C.F. Vorst, “Currency lookback options and

observation frequency: a binomial approach,” Journal of International
Money and Finance, vol. 16(2) (1997), p.173-187.

3. Andreasen, J., “The pricing of discretely sampled Asian and lookback

options: a change of numeraire approach,” Journal of Computational
Finance (Fall, 1998), p.5-30.

4. Babbs, S., “Binomial valuation of lookback options,” Journal of Eco-

nomics and Dynamic Control, vol. 24 (2000), p.1499-1525.

56
5. Kwok, Y.K. and K.W. Lau, “Accuracy and reliability considerations
of option pricing algorithms,” Journal of Futures Markets, vol. 21
(2001), p.875-903.

6. Yu, H, Y.K. Kwok and L.X. Wu, “Early exercise policies of American
floating and fixed strike lookback options,” Nonlinear Science, vol. 47
(2001), p.4591-4602.

7. Petrella, G. and S. Kou, “Numerical pricing of discrete barrier and

lookback options via Laplace transforms,” Journal of Computational
Finance (Fall 2004), p.1-37.

57
American floating strike lookback put option on a non-dividend-paying
stock (Hull-White, 1993)

• If the American floating strike lookback put option is exercised at time

τ , the exericse payoff is the amount by which the maximum stock price
between time 0 and time τ exceeds the current stock price. That is,

max St − Sτ .
t∈[0,τ ]
Note that the strike in the put payoff is reset to a new value when a
new maximum asset value is realized.

• We suppose that the initial stock price is $50, the stock price volatility
is 40% per annum, the risk-free interest rate is 10% per annum,
the total life of the option is three months, and that stock price
movements are represented by a three-step binomial tree. That is,
S0 = 50, σ = 0.4, r = 0.10, ∆t = 0.08333, u = 1.1224, d = 0.8909, R =
1.0084, and p = 0.5073.

58
Binomial tree for valuing an American lookback put option

Rolling back through the tree gives the value of the American lookback
put as $5.47. In the first time level, we have

V11 = max((3.36 × 0.5073 + 6.12 × 0.4927)e−0.1×0.0833, 0) = 4.683

V01 = max((2.66 × 0.5073 + 10.31 × 0.4927)e−0.1×0.0833, 50 − 44.55) = 6.38.

59
• The top number at each node is the stock price. The next level
of numbers at each node shows the possible maximum stock prices
achievable on all paths leading to the node. The bottom level of
numbers show the values of the derivative corresponding to each of
the possible maximum stock prices.

• The values of the derivatives at the final nodes of the tree are calcu-
lated as the maximum stock price minus the actual stock price.

• To illustrate the rollback procedures, suppose that we are at node

A, where the stock price is $50. The maximum stock price achieved
thus far is either 56.12 or 50 (depending on the path history of the
asset price movement). Consider first where it is equal to 50. If there
is an up movement, the maximum stock price becomes 56.12 and
the value of the derivative is zero. If there is a down movement, the
maximum stock price stays at 50 and the value of the terminal payoff
is 5.45 = 50 − 44.55.

60
• Assuming no early exercise, the value of the derivative at A when the
maximum achieved so far is 50 is,

(0 × 0.5073 + 5.45 × 0.4927)e−0.1×0.08333 = 2.66.

Clearly, it is not worth exercising at node A because the payoff from
doing so is zero.

• A similar calculation for the situation where the maximum value at

node A is 56.12 gives the value of the derivative at node A, without
early exercise, to be

(0 × 0.5073 + 11.57 × 0.4927)e−0.1×0.08333 = 5.65.

Early exercise gives a value of 6.12 and it is the optimal strategy.

• There may be multiple realized maximum asset values at each node.

The different possible values of the path dependent function at a
given node are linked to the corresponding path dependent function
at the nodes that are one time step earlier.

61
There are 2 possible realized maximum at node A, one is 50.00 while the
other is 56.12.
56.12 56.12
56.12 56.12
0.00 0.00

50.00 50.00
50.00 A 56.12 A
2.66 6.12

44.55 44.55
50.00 56.12
5.45 11.57

When the realized maximum at A When the realized maximum at

is 50.00, the realized maximum A is already 56.12, the realized
becomes 56.12 when the asset maximum remains at 56.12
price moves up while the realized independent of whether the
maximum remains at 50.00 when asset price moves up or down.
the asset price moves down.

62
Alternative binomial algorithm (Cheuk-Vorst, 1997)

When the stock price St is used as the numeraire, the payoff of the floating
strike lookback put takes the form:
V S max
Vet =
t t
= − 1, where Stmax = maxu∈[0,t] Su.
St St
We construct the truncated binomial tree for the process:
Stmax
Yt = , Yt ≥ 1.
St

• At the tip of the binomial tree, Y0 = 1.

• When Yt = 1 where St = Stmax, then

u when St+∆t = dSt
Yt+∆t = .
1 when St+∆t = uSt

• When Yt = uj for some j ≥ 1, St < Stmax, then


uj+1 when St+∆t = dSt
Yt+∆t = .
uj−1 when St+∆t = uSt
63
 • Let Vejn denote the numerical approximation to Vet = Vt/St at the
(n, j)th node of the binomial tree for Yt, where t = n∆t, n ≥ 0 and
Yt = uj , j ≥ 0.

For j ≥ 1, note that when the underlying St jumps up from state j to

state j + 1 with probability p, Yt jumps down from state j to state j − 1.
In terms of Ṽjn, the binomial scheme for pricing the lookback option is
given by

ṼjnS(tn) = e−r∆t[pṼj−1
n+1 n+1
uS(tn) + (1 − p)Ṽj+1 dS(tn)],
so that
[ ]
n
Ṽj = e −r∆t n+1 n+1
pṼj−1 u + (1 − p)Ṽj+1 d .

64
 • The continuation value is then given by
 [ ]
e −r∆t (1 − p)Vej+1 d + pVej−1 u ,
n+1 n+1
j≥1
[ ] .
e −r∆t e n+1
(1 − p)Vj+1 d + pVje n+1
u , j=0
Note that when j = 0, the upward jump of St keeps Yt to stay at the
same value j = 0.

Dynamic programming procedure for an American floating strike lookback

option
 { [ ]}
max Y − 1, e −r∆t (1 − p)Ve n+1
d + p e
V n+1
j≥1
e n { j [ j+1 j−1 u]} ,
Vj = .
max Y − 1, e−r∆t (1 − p)Ve n+1d + pVe n+1u , j=0
j j+1 j

Dimension reduction is achieved by taking the stock price S(tn) as the

numeraire. The exercise payoff of the American(floating ) strike lookback
Stmax
option can be expressed solely in terms of Yj = .
St j

65
 • The upper figures are values of Yt while the lower figures are option
values at the nodes.

Cheuk-Vorst’s procedure for valuing an American-style floating strike

lookback option. Though dimension reduction is achieved, the numerical
scheme has very slow rate of convergence.
66
European floating strike currency lookback call (S = 100, r̃d = 0.04,
r̃f = 0.07 and T = 0.5) with payoff: ST − min Sτ . Here, N is the number
[0,T ]
of time steps in the binomial tree.

• The binomial results seem to converge very slowly to the analytical

one (converge from below). The poor rate of convergence arises
from the ineffective modeling of the recording of the newly realized
maximum of the asset price (based on their intuitive derivation of
the binomial formula at j = 0). See Qn4 of HW2 for an alternative
derivation of the numerical boundary condition at j = 0.
67
• Cheuk-Vorst’s algorithm suffers from an extremely slow rate of con-
vergence when compared to other finite difference schemes for pricing
continuously monitored lookback options. Finite difference schemes
can incorporate boundary conditions effectively.

68
Binomial schemes for European fixed strike lookback call options

Terminal payoff = max( max S(ti) − K, 0),

0≤i≤N

where K is the fixed strike. Write M (tj ) = max0≤i≤j S(ti) as the realized
maximum asset value up to time tj , a known quantity at tj . Note that

max S(ti) = max(M (tj ), M (tN ; tj+1)),

0≤i≤N
where M (tN ; tj+1) = max S(ti) is the random path dependent state
j+1≤i≤N
variable for the future realized maximum asset value between time tj+1
and tN .

69
The fixed strike lookback call value at time tj and with known M (tj ) can
be expressed as

cf ix(S(tj ), M (tj ), tj )
= e−r(tN −tj )EQ[max(max(M (tj ), M (tN ; tj+1)) − K, 0)].

70
The terminal payoff can be decomposed into 2 terms:

max(max(M (tj ), M (tN ; tj+1)) − K, 0)


max(M (t ; t
N j+1 ) − K, 0) if M (tj ) ≤ K
=
M (t ) − K + max(M (t ; t
j N j+1 ) − M (tj ), 0) if M (tj ) > K
= max(M (tj ) − K, 0) + max(M (tN ; tj+1) − max(M (tj ), K), 0).

The decomposition reveals

• M (tj ) ≤ K
M (tj ) has no effect on the final option payoff.

• M (tj ) > K
Guaranteed to receive at least M (tj ) − K at maturity, plus higher
payoff if a higher realized maximum value is achieved at later time
instants.

71
How to achieve dimension reduction?

Define the adjusted exercise price K ′(tj ), where

K ′(tj ) = max(M (tj ), K).

• Since M (tj ) records the maximum among discrete nodal values of

the stock prices in the binomial tree, so M (tj ) is equal to S0un for
some integer n. One may set the original strike price K be equal to
S0um for some integer m (at least as a numerical approximation). In
fact, it would not be quite a restriction if the number of time steps
is sufficiently large. Taking M = max(n, m), then K ′(tj ) = S0uM .

This paves the adoption of dimension reduction in the discrete pricing

model by normalizing the price function with respect to asset price.

72
We relate the adjusted strike K ′(tj ) with S(tj ) in terms of an index k (as
power of u) via
S(tj ) ′ = S(t )u−k ,
k = ln / ln u ⇔ K j
K′
then k is always non-positive. This is because K ′ ≥ M (tj ) ≥ S(tj ).

Once we have set K ′ = S0uℓ for some integer ℓ and a similar form for
S(tj ), it is seen that the fixed strike lookback call value cX (S(tj ), K ′, tj )
is homogeneous in S(tj ), since the ratio of the payoff to the prevailing
asset price can be expressed as a power function in u. Homogeneity in
S(tj ) helps achieve dimension reduction. To this goal, we consider

cX (S(tj ), K ′, tj )
X(k, tj ) = .
S(tj )

The next move in the asset price may or may not result in the updating
of K ′. If there is no updating of K ′, then k is increased (decreased) by
one for an up-move (down-move) of the asset price.

73
• k ≤ −1 [S(tj ) ≤ K ′/u or K ′ ≥ S(tj )u so that it is not possible to have
an updated K ′ in the next time step.]
We adopt the usual backward induction procedure for the call value
normalized by S(tj ):

X(k, tj ) = [(1 − p)X(k − 1, tj+1)d + pX(k + 1, tj+1)u]e−r∆t.

When the asset price moves up (down) while K ′ does not change, the
S(tj )
index k increases (decreases) to k + 1 (k − 1) since k = ln ′
/ ln u.
K
• k = 0 [S(tj ) = K ′ = max(M (tj ), K)]
A downward move of the underlying asset price makes k to become
−1.
For an upward move, the underlying asset price exceeds the adjusted
strike K ′. The option holder is entitled to receive an extra payoff
equal to S(tj )(u − 1) at maturity. Also, the value of k remains to be
zero with the updating value of K ′ = S(tj+1) at tj+1.

74
Conditional on an upward move, the present value of this extra payment
is
S(tj )(u − 1)e−(N −j)r∆t.
The guaranteed payoff arising from updating of realized maximum of asset
price comes from the accumulation of these payments. The corresponding
binomial scheme at k = 0 is modified as

X(0, tj ) = [(1 − p)X(−1, tj+1)d + pX(0, tj+1)u]e−r∆t

+ p(u − 1)e−(N −j)r∆t.

Terminal condition

At maturity, the option value is zero since M (tN ) ≤ K ′ = max(M (tN ), K).
We then have X(k, tN ) = 0 for all values of k.

Though the terminal option value is set to be zero under the present
framework of adjusted strike, we have been accumulating the sum of the
present value of extra payments whenever a new updated K ′ is recorded.

75
Truncated tree representation

• When S(t0) ≥ K, the fixed strike lookback call option is sure to be

in-the-money. The truncated binomial tree starts at (0, 0).

• When S(t0) < K, we start the binomial tree at (k, 0) with k < 0.

76
European fixed strike currency lookback call (S = 100, K = 100, r̃d =
0.04, r̃f = 0.07 and T = 0.5). Here, N is the number of time steps in the
binomial tree.

• The rate of convergence is very slow. Even with 10, 000 time steps,
the numerical results barely achieve accuracy within 2% error.

77
1.2 Trinomial schemes

In a trinomial model, the asset price S is assumed to jump to either uS, mS

or dS after one time period △t, where u > m > d. We consider a trinomial
formula of option valuation of the form
p1V ∆t ∆t ∆t
u + p2 V m + p3V d
V = , R = er△t.
R
This is deduced from the risk neutral valuation principle: the current
option value is the discounted expectation of the terminal option value
under the risk neutral pricing measure.

There are 6 unknowns: p1, p2, p3, u, m and d. We take m = 1, u = 1/d.

We obtain 3 equations by

(i) equating mean, (ii) equating variance,

(iii) setting sum of probabilities = 1. We are left with one free parameter.

78
Discounted expectation approach

Under the assumption of the Geometric Brownian motion followed by the

continuous asset price process, we write

ln St+△t = ln St + ζ,
( )
σ2
where ζ is a normal random variable with mean r− △t and variance
2
σ 2△t. We approximate ζ by an approximate discrete random variable ζ a
with the following distribution


 v with probability p1
ζa = 0 with probability p2

 −v with probability p3
√
where v = λσ △t and λ ≥ 1. The corresponding values for u, m and d in
the trinomial scheme are: u = ev , m = 1 and d = e−v . This is because
St+∆t St+∆t
when ln assumes the value v, then assumes the value ev .
St St
Since we allow zero displacement with probability p2 > 0, we expect
stronger discrete move to the right or left (giving λ ≥ 1) when compared
to the binomial random walk. Also, we expect λ = 1 when p2 = 0.
79
To find the probability values p1, p2 and p3, the mean and variance of the
approximating discrete trinomial random walk variable ζ a are chosen to
be equal to those of ζ. These lead to
( )
σ2
E[ζ a] = v(p1 − p3) = r − △t
2

var(ζ a) = v 2(p1 + p3) − v 2(p1 − p3)2 = σ 2△t.

We see that v 2(p1 − p3)2 = O(∆t2). We may drop this term so that

v 2(p1 + p3) = σ 2△t,

while still maintaining O(∆t) accuracy.

S S
By considering the approximation of ln t+∆t instead of t+∆t , the alge-
St St
braic equations for solving p1, p2 and p3 involve only linear functions of
∆t rather than exponential functions of ∆t.

80
Lastly, the probabilities must be summed to one so that

p1 + p2 + p3 = 1.
We then solve together to obtain
σ 2 √
1 (r − 2 ) △t
p1 = +
2λ2 2λσ
1
p2 = 1 − 2
λ
σ 2 √
1 (r − 2 ) △t
p3 = 2
− ,
2λ 2λσ
here λ is a free parameter.

• In order that p2 ≥ 0, we must choose λ ≥ 1. Numerical experiments

√
indicate that the optimal choice of λ is 3 so that p2 = 2/3. Indeed,
Kwok and Lau (2001) show mathematically that the truncation error
√
is smallest when λ = 3.

• Trinomial schemes are first order accurate, where Vnum − Vexact =

K∆t. The different choices of λ amount to different values of K.
81
• Note that p2 = 0 when λ = 1, which reduces to the Cox-Ross-
Rubinstein binomial scheme. This illustrates an effective mean of
deriving the binomial/trinomial parameters using the discrete approx-
imation of the logarithm of the price ratio at successive time steps.
( )√
r− 2σ2 ∆t
1
• When λ = 1, we have p1 = + . This would agree with
2 2σ √
R−d
the Taylor expansion of p = σ
, u = 1/d = e ∆t up to O(∆t).
u−d

82
Multistate extension – Kamrad-Ritchken’s approach

• We assume the joint density of the prices of the two underlying assets
S1 and S2 to be bivariate lognormal.
• Let σi be the volatility of asset price Si, i = 1, 2 and ρ be the corre-
lation coefficient between the two lognormal diffusion processes.
△t
• Let Si and S i denote, respectively, the price of asset i at the current
time and one period △t later.
• Under the risk neutral measure, we have
△t
Si
ln = ζi, i = 1, 2,
Si
( )
σ 2
where ζi is a normal random variable with mean r − i △t and vari-
2
ance σi2△t.

83
The instantaneous correlation coefficient between ζ1 and ζ2 is ρ. The
joint bivariate normal processes {ζ1, ζ2} is approximated by a pair of joint
discrete random variables {ζ1a, ζ2a} with the following discrete distribution
{ 1 2}

ζ a1 ζ a2 probability
v1 v2 p1
v1 −v2 p2
−v1 −v2 p3
−v1 v2 p4
0 0 p5

√
where vi = λiσi △t, i = 1, 2. We first assume 2 free parameters λ1 and
λ2. Later, we argue that we must choose λ1 = λ2 for consistency.

The above form of the discrete distribution can be shown to be sufficient

to serve as the discrete approximation of the correlated diffusion processes
with drifts. It is redundant to include scenarios, like ζ1a = v1 and ζ2a = 0,
ζ1a = 0 and ζ2a = v2, etc.

84
Equating the corresponding means gives
( )
σ12
E[ζ1a] = v1(p1 + p2 − p3 − p4) = r− △t (i)
2
( )
σ22
E[ζ2a] = v2(p1 − p2 − p3 + p4) = r− △t. (ii)
2

By equating the variances and covariance to O(△t) accuracy, we have

var(ζ1a) = v12(p1 + p2 + p3 + p4) = σ12△t (iii)

var(ζ2a) = v22(p1 + p2 + p3 + p4) = σ22△t (iv)

E[ζ1aζ2a] = v1v2(p1 − p2 + p3 − p4) = σ1σ2ρ△t. (v)

In order that Eqs. (iii) and (iv) are consistent, we must set λ1 = λ2.

85
Writing λ = λ1 = λ2, we have the following four independent equations
for the five probability values
σ12 √
(r − 2 ) △t
p1 + p2 − p3 − p4 =
λσ1

σ22 √
(r − 2 ) △t
p1 − p2 − p3 + p4 =
λσ2
1
p1 + p2 + p3 + p4 = 2
λ
ρ
p1 − p2 + p3 − p4 = 2 .
λ
Since the probabilities must be summed to one, this gives the remaining
condition as
p1 + p2 + p3 + p4 + p5 = 1.

86
The solution of the above linear algebraic system of equations gives
   
√ σ12 σ22
1 1 △t  r − 2 r− 2  ρ
p1 =  +  +  + 
4 λ2 λ σ1 σ2 λ2

   
√ σ12 σ22
1 1 △t  r − 2 r− 2  ρ
p2 =  2+  −  − 2
4 λ λ σ1 σ2 λ

   
√ σ12 σ22
1 1 △t  r − 2 r− 2  ρ
p3 =  2+ − −  + 2
4 λ λ σ1 σ2 λ

   
√ σ12 σ22
1 1 △t  r − 2 r− 2  ρ
p4 =  2+ − +  − 2
4 λ λ σ1 σ2 λ

1
p5 = 1 − 2 , λ ≥ 1 is a free parameter.
λ

87
Two-state trinomial model

• For convenience, we write ui = evi , di = e−vi , i = 1, 2.

• Let V ∆t
u1u2 denote the option price at one time period later with asset
prices u1S1 and u2S2, and similar meaning for V ∆t u d , V ∆t and V ∆t .
d u d d
1 2 1 2 1 2
• We let V ∆t
0,0 denote the option price one period later with no jumps
in asset prices.
• The corresponding 5-point formula for the two-state trinomial model
based on the risk neutral valuation approach can be expressed as
△t △t △t △t
V = (p1Vu△t
1 u2
+ p2Vu d + p3Vd d + p4Vd u + p5V 0,0)/R.
1 2 1 2 1 2

• When λ = 1, we have p5 = 0 and the above 5-point formula reduces

to the 4-point formula.

88
1.3 Forward shooting grid methods (strongly path dependent op-
tions)

• For path dependent options, the option value also depends on the
path function Ft = F (S, t) defined specifically for the given nature
of path dependence, say, the minimum asset price realized along a
specific asset price path.
• Since option value depends also on Ft, we find the value of the path
dependent option at each node in the lattice tree for all alternative
values of Ft that can occur.
• The approach of appending an auxiliary state vector at each node
in the lattice tree to model the correlated evolution of Ft with St is
commonly called the forward shooting grid (F SG) method.

89
• Consider a trinomial tree whose probabilities of upward, zero and
downward jump of the asset price are denoted by pu, p0 and pd, re-
spectively.
• Let Vj,k
n denote the numerical option value of the exotic path depen-

dent option at the nth-time level (n time steps from the tip of the
tree). Also, j denotes the j upward jumps from the initial asset value
and k denotes the numbering index for the various possible values of
the augmented state variable Ft at the (n, j)th node.
• Let G denote the function that describes the correlated evolution of
Ft with St over the time interval ∆t, that is,

Ft+∆t = G(Ft, t, St+∆t).

90
• Let g(k, n, j) denote the grid function which is considered as the dis-
crete analog of the evolution function G. Here, k is the index for Ft,
n is the index for t and j is the index for St+∆t.

• The trinomial version of the FSG scheme can be represented as follows

[ ]
Vj,k = puVj+1,g(k,n,j+1) + p0Vj,g(k,n,j) + pdVj−1,g(k,n,j−1) e−r∆t,
n n+1 n+1 n+1

where e−r∆t is the discount factor over time interval ∆t.

• To price a specific path dependent option, the design of the FSG

algorithm requires the specification of the grid function g(k, n, j).
For notational convenience, if the grid function has no dependence
on t, we simply write it as g(k, j).

91
Cumulative Parisian feature of knock-out

• Let M denote the prespecified number of cumulative breaching occur-

rences that is required to activate knock-out, and let k be the integer
variable that counts the cumulative number of breaching occurrences
so far.

• Let B denote the down barrier associated with the knock-out feature.
Let xj denote the value of x = ln S that corresponds to j upward
moves in the trinomial tree. That is, xj = ln S0 + j∆x, where S0 is
the initial asset price and ∆x is the stepwidth of the state variable x.

• When n∆t happens to be a monitoring instant, the index k increases

its value by 1 if the asset price S falls on or below the barrier B, that
is, xj ≤ ln B.

92
Counting the number of time steps that xj falls below or at ln B

To incorporate the cumulative Parisian feature, the appropriate choice of

the grid function gcum(k, j) is defined by

gcum(k, j) = k + 1{xj ≤ln B}.

The forward shooting grid algorithm is exemplified by

 [p V n + p V n +p Vn ]e −r∆t

 u j+1,k 0 j,k d j−1,k


n−1 if n∆t is not a monitoring instant
Vj,k = −r∆t .

 [p u V n + p V n + p V n ]e

 j+1,gcum(k,j+1) 0 j,gcum (k,j) d j−1,gcum(k,j−1)

if n∆t is a monitoring instant

The number of breaching occurrences k is updated to gcum(k, j + 1) when

n
the updated asset price at the nth time level is Sj+1 [up move from Sjn−1 at
the (n − 1)th time level]. The knock-out condition is defined by Vj,Mn = 0.

93
Schematic diagram that illustrates the construction of the grid function
gcum(k, j) that models the cumulative Parisian feature. The down barrier
ln B is placed mid-way between two horizontal rows of trinomial nodes.
Here, the nth-time level is a monitoring instant. In this example, since
xj−1 < ln B, the forward shooting grid algorithm is
[ ]
Vj,k = puVj+1,k + p0Vj,k + pdVj−1,k+1 e−r∆t,
n−1 n n n k = 1, 2, · · · .

94
1. The pricing of options with the continuously monitored cumulative
Parisian feature is obtained by setting all time steps to be monitoring
instants.

2. The computational time required for pricing an option with the cu-
mulative Parisian feature requiring M breaching occurrences to knock
out is about M times that of an one-touch knock-out barrier option.

3. The size of the augmented state vector appended at each node grows
from zero at the tip of the trinomial tree to the maximum size of M
as we proceed the time marching in the trinomial calculations. At
maturity, we set


 V N = V (S N ), 0 ≤ k < M,
 j,k T j


 V N = 0, k = M.
j,M

95
4. Applications of the cumulative counting feature can also be found in
structured products, say, the coupons (as in reverse convertibles) are
accrued contingent on the underlying stock price lying within certain
range of values.

5. The consecutive Parisian feature counts the number of consecutive

breaching occurrences that the asset price stays in the knock-out
region. The count is reset to zero once the asset price moves out
from the knock-out region. Assuming B to be the down barrier, the
appropriate grid function gcon(k, j) in the FSG algorithm is given by

gcon(k, j) = (k + 1)1{xj ≤ln B}.

6. The consecutive counting feature can be found in the soft call pro-
vision in a convertible bond. In most convertible bond contracts,
the issuer is allowed to issue the notice of redemption conditional on
the underlying stock price staying above the preset hurdle price for a
prespecified number of trading days.

96
Call options with the strike reset feature

• Consider a call option with the strike reset feature where the option’s
strike price is reset to the prevailing asset price on a preset reset date
if the option is out-of-money on that date.

• Let ti, i = 1, 2, · · · , M , denote the ith reset date and Xi denote the
strike price specified on ti based on the above reset rule. Write X0 as
the strike price set at initiation, then Xi is given by

Xi = min(Xi−1, Sti ), i = 1, 2, · · · , M,
where Sti is the prevailing asset price on the reset date ti.

• Why does it become superfluous to set

Xi = min(Xi−1, Sti , X0), i = 1, 2, · · · , M ?

Since X1 = min(X0, St1 ), the information of the initial strike price
X0 has been embedded in the strike reset procedure. Suppose the
discrete realized minimum asset price has not reached as low as X0
after i time steps, then Xi = X0.

97
 • The strike price at expiry of this call option is not fixed since its value
depends on the realization of the asset price on the reset dates.

• When we apply the backward induction procedure in the trinomial

calculations, we encounter the difficulty in defining the terminal payoff
since the strike price can assume many possible values due to the reset
mechanism.

• These difficulties can be resolved easily using the FSG approach by

tracking the evolution of the asset price and the strike reset through
an appropriate choice of the grid function. The terminal payoff in the
FSG lattice tree is computed with respect to all possible values of k
that can be realized at maturity.

Remark If we do not impose the initial strike X0, then this strike reset
call option resembles the discretely monitored floating strike lookback call
option with terminal payoff: max(ST − Smin, 0).

98
• Suppose the original strike price X0 corresponds to the index k0, this
would mean X0 = S0uk0 . For convenience, we may choose the pro-
portional jump parameter u such that k0 is an integer. In terms of
these indexes, the grid function that models the correlated evolution
between the reset strike price and asset price is given by

greset(k, j) = min(k, j),

where k denotes the index that corresponds to the strike price reset in
the last reset date and j is the index that corresponds to the prevailing
asset price at the reset date.

• Since the strike price is reset only on a reset date, we perform the usual
trinomial calculations for those time levels that do not correspond
to a reset date while the augmented state vector of strike prices are
adjusted according to the grid function greset(k, j) for those time levels
that correspond to a reset date.

99
• The FSG algorithm for pricing the reset call option is given by

 [ ]

 p V n + p V n + p V n e−r∆t

 u j+1,k 0 j,k d j−1,k







 if n∆t ≠ ti for some i

n−1
Vj,k = [ ] .



 p V n + p V n + p V n e−r∆t ,

 u j+1,greset (k,j+1) 0 j,greset (k,j) d j−1,greset (k,j−1)






 if n∆t = ti for some i

• The payoff values along the terminal nodes at the N th time level in
the trinomial tree are given by
N = max(S uj − S uk , 0),
Vj,k j = −N, −N + 1, · · · , N,
0 0

and k assumes values that lie between k0 and the index corresponding
to the lowest asset price on the last reset date (since there is no reset
on maturity date). It is necessary to list all possible nodal values that
can be assumed by the reset strike.

100
Floating strike arithmetic averaging call

• To price an Asian option, we find the option value at each node for
all possible values of the path function F (S, t) that can occur at that
node.

• Unfortunately, the number of possible values for the averaging value

F at a binomial node for the arithmetic averaging option grows ex-
ponentially at 2n, where n is the number of time steps from the tip
of the binomial tree. (Why 2n? Since there are 2n possible realized
asset paths after n time steps and each path gives a unique arithmetic
averaging value.)

• Therefore, the binomial schemes that place no constraint on the num-

ber of possible F values at the binomial nodes would become compu-
tationally infeasible.

101
Illustration

Consider the following tree

62.99
56.12

50.00 50.00

44.55
39.69

There are 4 = 22 possible arithmetic averaging values after 2 time steps,

namely,
50.00 + 56.12 + 62.99 50.00 + 56.12 + 50.00
Auu = , Aud = ,
3 3
50.00 + 44.55 + 50.00 50.00 + 44.55 + 39.69
Adu = , Add = .
3 3

102
Note that these arithmetic averaging values do not coincide with the asset
prices at the nodes at the 2nd time level. Extending to a 3-step binomi-
al tree, there are 8 = 23 possible arithmetic averaging values, namely,
Auuu, Auud, Audu, · · · , Addd.

Geometric averaging values

• Two-step binomial tree

√
3
Guu = (S0)(S0u)(S0u2) = S0u,
√
S0(S0u−1)(S0u−2) = S0u−1,
3
Gdd =
√
Gud = 3
(S0)(S0u)(S0) = S0u1/3,
√
(S0)(S0u−1)(S0) = S0u−1/3.
3
Gdu =

103
• Three-step binomial tree
√
(S0)(S0u)(S0u2)(S0u3) = S0u1.5,
4
Guuu =
√
(S0)(S0u−1)(S0u−2)(S0u−3) = S0u−1.5,
4
Gddd =
√
4
Guud = (S0)(S0u)(S0u2)(S0u) = S0u,
Gudu = S0u0.5, Gduu = S0,
√
Gudd = (S0)(S0u)(S0)(S0u−1) = S0,
4

Gdud = S0u−0.5, Gddu = S0u−1.

There are 7 possible geometric averaging values after 3 time steps.

104
• A possible remedy is to restrict the possible values for F to a certain
set of predetermined values. The option value V (S, F, t) for other
values of F is obtained from the known values of V at predetermined
F values by an interpolation between the nodal values.

• The methods of interpolation include the nearest node interpolation,

linear (between 2 neighboring nodes) and quadratic interpolation (be-
tween 3 neighboring nodes).

• How to cope with the exponentially large number of possible values

assumed by taking the arithmetic averaging of the realized asset price
path? We limit the number of averaging values to some multiple of
the number of values assumed by the asset price (here, the multiple
is 1/ρ).

105
For a given time step ∆t, we fix the respective step width for the logarithm
of asset price and average to be
√
∆W = σ ∆t and ∆Y = ρ∆W, ρ < 1,
and define the possible values for St and At at the nth time step by

Sjn = S0ej∆W and An

k = S0 ek∆Y ,

where j and k are integers, and S0 is the asset price at the tip of the
binomial tree.

• We take 1/ρ to be an integer. The larger integer value chosen for 1/ρ,
the finer the quantification of the arithmetic averaging asset value.

106
 ( )
1
Quantification of arithmetic averaging asset value Here, = 3 is taken.
ρ

107
The continuous version of the arithmetic averaging state variable is de-
fined by
∫
1 t
At = Su du.
t 0

• The terminal payoff of the floating strike Asian call option is given by
max(ST − AT , 0), where AT is the arithmetic average of St over the
time period [0, T ].

• Similarly, the terminal payoffs of other related Asian options are

(i) Floating strike Asian put option: max(AT − ST , 0);

(ii) Fixed strike Asian call option: max(AT − X, 0), where X is the fixed
strike price.

108
Updating rule of At over successive discrete time points

Consider the following relation between At and St in differential form:

1
d(tAt) = St dt or dAt = (St − At) dt,
t
we approximate the above differential at time t + ∆t by adopting

(t + ∆t)[At+∆t − At] = (St+∆t − At+∆t)∆t,

so that
(t + ∆t)At + ∆t St+∆t
At+∆t = ≡ G(t, At, St+∆t).
t + 2∆t
This is the updating rule of At+∆t ∆t at the new time level t + ∆t based
on the old value At at the previous time level t and the updated asset
value St+∆t at the new time level t + ∆t.

Suppose t = n∆t, then

(n + 1)An + Sn+1
An+1 = .
n+2
Note that there are n + 1 asset prices recorded between 0 and n∆t.
109
Consider the binomial procedure at the (n, j)th node, suppose we have
n+1
an upward move in the asset price from Sjn to Sj+1 and let An+1 be
k+(j)
the corresponding updated value of At changing from An k when the asset
n+1
price moves up from Sjn to Sj+1 . Setting A0
0 = S0 and taking t = n∆t,
the equivalence of the above discrete updating rule is given by
n+1
(n + 1)An
k + S j+1
An+1
+ = . (a)
k (j) n+2
n+1
For a downward move in the asset price from Sjn to Sj−1 , An
k changes to
An+1
k− (j)
where
n+1
(n + 1)An
k + Sj−1
An+1
k− (j)
= . (b)
n+2
n+1 n+1 k ′ ∆Y
Note that Ak±(j) in general do not coincide with Ak′ = S0e , for
some integer k′.

n+1 k ± (j)∆Y n+1

Recall Ak±(j) = S0e and Sj±1 = S0e(j±1)∆W .

110
In terms of ∆W and ∆Y , after canceling the common factor S0, eqs. (a)
and (b) can be expressed as

k ± (j)∆Y (n + 1)ek∆Y + e(j±1)∆W

e = .
n+2
Accordingly, we compute the indexes k±(j) by
(n+1)ek∆Y +e(j±1)∆W
ln
g(n, k, j ± 1) = k±(j) = n+2
. (1)
∆Y

• We define the integers kf±loor such that An+1 are the largest possi-
kf±loor
ble An+1
k ′ values less than or equal to An+1
±
k (j)
. We then set k +
f loor =
floor(k+(j)) and kf−loor = floor(k−(j)), where floor(x) denotes the
largest integer less than or equal to x. Equation (1) corresponds to
n+1 n+1
the evolution of Ank to A ±
k (j)
depending on the updated value of S j±1
[in terms of the indexes k and k±(j)].

111
Restricting the size of the augmented state vector representing possible
averaging values

• What would be the possible range of k at the nth time step? We ob-
serve that the arithmetic averaging state variable At must lie between
the maximum asset value Snn and the minimum asset value S−n n , so k

must lie between − nρ ≤ k ≤ nρ . Unless ρ assumes a very small value,

the number of predetermined values for At is in general manageable.

• Consider Anℓ , where ℓ is in general a real number. We write ℓf loor =

floor(ℓ) and let ℓceil = ℓf loor + 1, then An
ℓ lies between An
ℓf loor and An .
ℓceil
Though the number of possible values of ℓ grows exponentially with
the number of time steps in the binomial tree, both ℓf loor and ℓceil at
n n
the nth time level assume an integer value lying between − and .
ρ ρ

112
Linear interpolation

• Let cn j,ℓ denote the numerical approximation to the Asian call value at
the (n, j)th node with the averaging state variable assuming the value
An n
ℓ , and similar notations for cj,ℓ and cn
j,ℓ .
f loor ceil

• For non-integer value ℓ, cn

j,ℓ is approximated through linear interpola-
tion using the call values cn
j,ℓ and cn
j,ℓ at the neighboring nodes.
f loor ceil
( )
cn
j,ℓ = c n
j,ℓf loor + ϵ ℓ c n
j,ℓceil − cn
j,ℓf loor
= ϵℓ c n
j,ℓceil + (1 − ϵ n
ℓ j,ℓf loor ,
)c
where
ℓ − ln Aℓ
ln An n
f loor
ϵℓ = .
∆Y
Here, ϵℓ is the fraction of one step width ∆Y between ln An
ceil(ℓ)
and
ln An
f loor(ℓ)
, where

An n
ℓ = Aℓf loor e
ϵℓ ∆Y .

113
• Here, ℓ is a real number lying between two consecutive integers,
floor(ℓ) and ceil(ℓ), where ceil(ℓ) = floor(ℓ) + 1.

• Numerical option values are available only at An

f loor(ℓ)
and A n
ceil(ℓ)
,
where the index k in Ank assumes an integer value [like floor(ℓ) or
ceil(ℓ)].

• We use the log distance between An floor(ℓ)

, An n
ℓ and Acell(ℓ) as reference
for linear interpolation. For ℓ to be non-integer, we approximate cn j,ℓ
by linear interpolation between cnj,f loor(ℓ)
and cnj,ceil(ℓ)
.

114
• By applying the above linear interpolation formula [taking ℓ to be
k+(j) and k−(j) successively], the FSG algorithm with linear interpo-
lation for pricing the floating strike arithmetic averaging call option is
given by
[ ]
cn = e−r∆t pcn+1 + (1 − p)cn+1
j,k j+1,k+(j) j−1,k− (j)
{ [ ]
≈ e−r∆t p ϵk+(j)cn+1 + + (1 − ϵk+(j))cn+1 +
j+1,k ceil j+1,k f loor
[ ]}
+ (1 − p) ϵk−(j)cn+1 − + (1 − ϵk−(j))cn+1 − , (2)
j−1,k ceil j−1,k f loor
n
n = N − 1, · · · , 0, j = −n, −n + 2, · · · , n, k is an integer between − and
ρ
n ±
, k (j) are given by Eq. (i) while
ρ

ln An+1
±
k (j)
− ln An+1
±
kf loor
ϵk±(j) = . (3)
∆Y

115
Terminal payoff of a floating strike Asian call option

The final condition is

cN
j,k = max(Sj
N − AN , 0)
k
= max(S0ej∆W − S0ek∆Y , 0), j = −N, −N + 2, · · · , N.

The upper (lower) bound of arithmetic averaging values can be deduced

by assuming upward (downward) moves of the stock price at all time
steps. We can then deduce the range of values that can be assumed by
k.

116
In summary, we compute the updated arithmetic average values based on
n+1
n, Ank and S j±1 .
n+1 n+1
An
k −→ A + when Sjn −→ Sj+1
k (j)
n+1 n −→ S n+1
An
k −→ A −
k (j)
when S j j−1

Note that k is an integer while k+(j) and k−(j) are in general non-integers.
Since the numerical call option values at the (n+1)th time step are known
at integer value of the index k′ for An+1
k′
, we use the interpolation scheme
to estimate cn+1
j,k±(j)
as follows:

cn+1±
j,k (j)
= ϵ ± cn+1
±
k (j) j,ceil(k (j)) + (1 − ϵ n+1
k (j) j,f loor(k±(j)) ,
± )c
where
ln An+1
k± (j)
− ln A n+1
f loor(k± (j))
ϵk±(j) = .
∆Y
Using the discounted expectation approach, we have
[ ]
cn = pc n+1
+ (1 − p)cn+1 e−r∆t .
j,k j+1,k+ (j) −
j−1,k (j)

117
• Recall that Sjn = S0uj = S0ej∆W and An k = S0 e
k∆Y = S ekρ∆W . Also,
0
n+1 n → S n+1 and An+1 when S n → S n+1.
An
k becomes A + when S j j+1 k−(j) j j−1
k (j)

• At each time step, we compute the numerical option values at all

possible integer values of k.

118
Remarks

1. When dealing with the discretely monitored Asian options, we only

update the values of k at a time step that corresponds to a monitoring
instant. At a time step that does not correspond to a monitoring
instant, we compute cn j,k using the binomial formula:

cn = [pc n+1
+ (1 − p)cn+1
]e−r∆t .
j,k j+1,k j−1,k

2. The range of averaging value of the asset price at a given time step
can be deduced by finding the largest possible averaging value (upward
move at every time step) and the smallest value (downward move at
every time step). Given the step width ∆Y for the averaging value,
(n) (n)
we can determine kmax and kmin that correspond to the upper bound
and lower bound on the averaging value at the nth time step.

119
Geometric averaging

Extension to Asian options on geometrical averaging of asset values

1
ln Gn = (ln S0 + ... + ln Sn)
n+1
1
ln Gn+1 = (ln S0 + ... + ln Sn + ln Sn+1),
n+2
so the evolution of Gn+1 in terms of n, Gn and Sn+1 is

(n + 2) ln Gn+1 − (n + 1) ln Gn = ln Sn+1
n+1 1
Gn+1 = (Gn) n+2 (S n+1 ) n+2 .
It is necessary to convert this correlated evolution function between Gt+∆t,
Gt and St+∆t in terms of the indexes that correspond to the discrete val-
ues of the three state variables.

120
Suppose we write
Gn
k = G 0 ek∆Y = S ekρ∆W
0 0

n+1
Sj±1 = S0e(j±1)∆W ,
we deduce that
k ± (j)ρ∆W n+1 1
e = (ekρ∆W ) n+2 (e(j±1)∆W ) n+2 .
This gives the grid function
± n+1 j±1 1
k (j) = g(n, k, j ± 1) = k + .
n+2 ρ n+2
In general, k±(j) would not assume integer values. The option value at
k±(j) at the (n + 1)th time step is obtained by linear interpolation at
f loor(k±(j)) and ceil(k±(j)). We have
[ ]
n+1 n+1 n+1 n+1
Vj±1,k ± (j) = Vj±1,f loor(k ± (j)) + ϵk ± (j) Vj±1,ceil(k ± (j)) − Vj±1,f loor(k ± (j))

where
ϵk±(j) = k±(j) − f loor(k±(j)).

121
Alpha quantile option

The α-quantile option takes the barrier level to be a stochastic state

variable that defines the terminal payoff.

For a given percentile α, 0 ≤ α ≤ 1, the α-quantile of {St}t∈[0,T ] is defined

as
{ ∫ }
1 T
Binf (T ; α) = inf B : 1{St≤B} dt ≥ α . (A)
T 0

We gradually lower the barrier B and eventually the percentage of time

that St stays at or below B just hits at α. In other words, Binf (T ; α) is the
barrier level such that the asset price St is at or below Binf (T ; α) exactly
α of the monitoring period. When α = 0.5, Binf (T ; 0.5) is the median
Smedian of the asset price process over the time period [0, T ].

122
• The asset price is below Smedian exactly half of the time period [0, T ].
• Binf (T ; 1) is the realized maximum asset price over [0, T ] since the
asset price is below this barrier level (infimum among all barrier levels)
100% of the time period.
123
For a European α-quantile call option, the terminal payoff is gven by

Vα(S, T ) = max(Binf (T ; α) − X, 0),

where X is the strike price.
• In the discrete trinomial tree model with N time steps, we write SjN
as the discrete terminal asset price at maturity, j = −N, −N + 1, ..., N .
We assume that the possible values taken by the stochastic variable
Binf are limited to Sj , j = −N, ..., N − 1, N ; Sj = S0uj , where u is the
up-jump parameter. One may adopt a finer resolution of the discrete
values that can be taken by Binf for better accuracy (say, allowing the
jump parameter of Binf to be ρu, where ρ < 1).
• The numerical approximate value of the continuously monitored Eu-
ropean α-quantile call option is given by
N
∑
Vα(S, 0) = e−rT P [Binf = Sj ] max(Sj − X, 0), Sj = S0uj .
j=−N
This is the summation of the state price of the event {Binf = Sj },
j = −N, . . . , N , multiplied by the corresponding terminal payoff at
Binf = Sj .
124
Binary cumulative options

bin (α, B) denote the value of a binary option that pays $1 at maturity
Let Vcum
T if the cumulative time staying at or below the down-barrier B is less
than α of the total life of the option, 0 ≤ α ≤ 1; otherwise the terminal
payoff of the option is zero. This option value is equivalent to the state
price of the following event:
∫
1 T
1{St≤B} dt < α.
T 0

For a fixed value of α, the payoff of this binary option is $1 (corresponding

to the occurrence of the above event) only if the specified down-barrier
B is below the realized value of Binf (T ; α). If otherwise, suppose B ≥
∫
1 T
Binf (T ; α), according to eq.(A), then 1{St≤B} dt ≥ α, a contradiction
T 0
to the fact that the above event occurs. The fair value of this binary
cumulative barrier option is e−rT P [Binf (T ; α) > B]. In other words, if B
is set to be too low such that Binf (T ; α) > B, this is equivalent to expire
in-the-money for the binary option Vcum bin (α, B).

125
In the discrete world of the trinomial tree, we choose B = Sj for some j.
We then have
bin (α, S ) = e−rT P [B
Vcum j inf > Sj ]

so that

e−rT P [Binf = Sj ] = e−rT {P [Binf > Sj−1] − P [Binf > Sj ]}

bin (α, S bin
= Vcum j−1 ) − Vcum(α, Sj ).

bin (α, B) is given by

The terminal payoff of Vcum

1 if 0 ≤ k < αN
N =
Vj,k ,
0 if k ≥ αN
where k counts the number of time steps in the total number of N time
steps that Sjn ≤ B and N is the total number of time steps in the lattice
tree calculations. Note that the terminal payoff is independent of j (the
index for the asset price) since the payoff of the binary cumulative barrier
option is independent of asset price.

126
Parameter values: α = 0.8, S = 100, X = 95, r = 0.05, q = 0, σ = 0.2
and T = 0.25.
127
Accumulators

• Entails the investor entering into a commitment to purchase a fixed

number of shares per day at a pre-agreed price (the “Accumulator
Price”). This Price is set (typically 10-20%) below the market price
of the shares at initiation. This is portrayed as the “discount” to the
market price of the shares.

Example

Citic Pacific entered into an Australian dollar accumulator as hedges “with

a view to minimizing the currency exposure of the company’s iron ore
mining project in Australia”. The company benefits from a strengthening
in the A$ above A$1 = US$0.87.

128
Citic Pacific’s )ϛ߬੎൲* bitter story

• Citic Pacific signed an accumulator that not only set the highest gains
but failed to include a floor for losses. The Australian dollar’s value
was rising when the contract was signed.

• After July, 2008, the AUD’s value against the USD declined, sliding
as low as 1 to 0.65. The firm also said its highest, marked-to-market
loss could reach HK$14.7 billion. Some analysts say if the AUD falls to
1 to 0.50 USD, the mark-to-market loss would rise to HK$26 billion.

• Citic Pacific shares fell 80% on the Hong Kong exchange to HK$5.06
a share on October 24, compared with HK$28.20 a share on July 2.

• The company was driven by a “mixture of greed and a gambling

mentality” to use the accumulator. Why not simply buy the less risky
currency futures to hedge the iron ore mining project?

129
Cap on upside gain

If the market price of the shares rises above a pre-specified level (“Knock-
Out price”) then the obligation to purchase shares ceases. This Price is
set (typically 2% to 5%) above the market price of the shares at initiation.

Intensifying downside losses (“I will kill you later”)

If the market price falls below the Accumulator Price (10-20% below
the market price at initiation), then the investor would be obligated to
purchase more shares. This is called the Step-Up feature. The Step-Up
factor can be 2 or up to 5.

• Margin is required to minimize counterparty risks. The investor gener-

ally benefits where the share prices remain relatively stable, preferably
between the Knock Out Price and the Accumulator Price.

130
Example of an accumulator on China Life Insurance Company

• Stock Price Movement of China Life Insurance Company Limited

(June 12, 2009 - July 13, 2009)

131
SGD-Equity Accumulator Structure

Underlying Shares: SEMBCORP INDUSTRIES LTD

Start Date: 05 November 2007

Final accumulation 03 November 2008

Date:

Maturity Date: 06 November 2008

(subject to adjustment if a Knock-Out Event
has occurred)

Strike Price: $4.7824

132
Knock-Out Price: $6.1425

Knock-Out Event: A Knock-Out Event occurs if the official clos-

ing price of the Underlying Share on any Sched-
uled Trading Day is greater than or equal to
the Knock-Out Price. Under such event, there
will be no further daily accumulation of Shares
from that day onward. The aggregate number
of shares accumulated will be settled on the Ear-
ly Termination Date, which is the third business
day following the occurrence of Early Termina-
tion Event.

133
Shares Accu- On each Scheduled Trading Day prior to the occur-
mulation: rence of Early Termination Event, the number of
shares accumulated will be
1,000 when Official Closing Price for the day is
higher than or equal to the Strike Price
2,000 when Official Closing Price for the day is
lower than the Strike Price

Monthly Set- The Shares accumulated for each Accumulation Peri-

tlement Date: od will be delivered to the investor on the third busi-
ness day following the end of each monthly Accumu-
lation Period

Total Number Up to the maximum of 500,000 shares (the worst

of Shares: scenario is 2, 000 shares purchased for 250 trading
days)

134
Accumulation Period and Delivery Schedule

12 accumulation periods in total

Accumulation Period Number of days Delivery Date

05 Nov 07 to 03 Dec 07 20 06 Dec 07

04 Dec 07 to 02 Jan 08 19 07 Jan 08
03 Jan 08 to 04 Feb 08 23 11 Feb 08
05 Feb 08 to 03 Mar 08 18 06 Mar 08
04 Mar 08 to 02 Apr 08 21 07 Apr 08
03 Apr 08 to 02 May 08 21 06 May 08
05 May 08 to 02 Jun 08 20 05 Jun 07
03 Jun 08 to 02 Jul 08 22 07 Jul 08
03 Jul 08 to 04 Aug 08 23 07 Aug 08
05 Aug 08 to 02 Sep 08 21 05 Sep 08
03 Sep 08 to 02 Oct 08 21 07 Oct 08
03 Oct 08 to 03 Nov 08 21 06 Nov 08

135
Let V (j, ℓ, i, k) denote the value of the accumulator at the j th accumulation
period, ℓth business date (ℓth time step if the time step is taken to be one
business day), ith stock price level, and k units of shares accumulated in
the j th period up to the ℓth day.

• 04 Dec 07 corresponds to the first day in the second accumulation

period, so ℓ = 1, j = 2; 02 Oct 08 corresponds to the 21st day in the
11th accumulation period, so ℓ = 21, j = 11.

• The last day of the j th accumulation period can be considered as the

0th day of the (j + 1)th accumulation period.

136
An accumulator as a portfolio of occupation time derivatives

• Without the “intensifying loss” feature, the product is like a portfo-

lio of forward contracts with the knock-out feature. Purchases are
conditional on survival until the date of accumulation of shares.

• The “intensifying loss” feature can be considered as a portfolio of for-

ward contracts with the “excursion time” feature. The accumulated
amount of shares depends on the total excursion time of the stock
price staying below the strike price, again conditional on survival until
the date of accumulation of share.

In other words, one has to count the number of days that the stock
price stays below the strike price, conditional on “no knock-out”. The
knock-out feature limits the upside gain of the accumulator investor.

137
1. Use the forward shooting grid technique to keep track of the total
number of shares to be purchased. The grid function is defined by

G(k, i) = k + 2, 0001{i≤Istrike} + 1, 0001{i>Istrike}.

Istrike + 1
We set Istrike such that S0u is the actual strike price to avoid
2

the ambiguity of determining whether 1, 000 or 2, 000 stocks to be

bought when the stock price is exactly equal to the strike price.
Suppose we take m time steps per each business day, m ≥ 1; and let
ℓ denote the number of time steps lapsed from the last settlement
date. For those time steps that do not correspond to the time of
stock accumulation, we have the usual binomial scheme:

V (j, ℓ, i, k) = e−r∆t[pV (j, ℓ + 1, i + 1, k) + (1 − p)V (j, ℓ + 1, i − 1, k)];

while at a time step that corresponds to stock accumulation, we have

V (j, ℓ, i, k) = e−r∆t[pV (j, ℓ + 1, i + 1, G(k, i + 1))

+ (1 − p)V (j, ℓ + 1, i − 1, G(k, i − 1)].

138
2. Jump conditions are applied across each settlement date (ending date
of an accumulation period).

• Right after the settlement of the accumulated shares, k is reset

to zero. Right before delivery, compute the value function for
all possible values of k. If there are ℓmax(j) days in the j th peri-
od, then k assumes values from 1, 000 × ℓmax(j), 1, 000 × [ℓmax(j) +
1], · · · , 2, 000 × ℓmax(j).

• The jump in the accumulator value across each settlement date is

the value of the accumulated units of stock on the settlement date.
Moving from the j th period to the (j + 1)th period, the value of the
accumulator is split into the continuation value of the accumulator
with k being reset to zero and the value of the stocks transacted.
The delivery of the accumulated stocks is done M time steps after
the end of the accumulation period, so

V (j, mℓmax(j), i, k) = V (j + 1, 0, i, 0) + k(Si − Xe−rM ∆t).

Here, M is the number of time steps between the settlement date
and delivery date.
139
Rebate upon knock-out at the upstream knock-out price

n+1
When Si+1 is above the knock-out price, knock-out event occurs and the
( )
n+1
value of the accumulator becomes g(k, i+1) Si+1 − Xe −rM ∆t . Suppose
the stocks are delivered 3 days after the knock-out date, then M = 3m,
where m is the number of time steps for each business date.

140
Decomposition of an accumulator under immediate settlement

Under the assumption of monitoring of the upper knock-out barrier on

each business day and immediate settlement of the accumulated stock,
one can decompose an accumulator into a portfolio of up-and-out barrier
call and put options. The accumulator survives up to ti if max Sτ < H.
0≤τ ≤ti−1

The payoff on the observation date ti is given by



 if max0≤τ ≤ti−1 Sτ ≥ H
0



Sti − K if max0≤τ ≤ti−1 Sτ < H and Sti ≥ K


2(St − K) if max0≤τ ≤ti−1 Si < H and Sti < K,
i

where K = strike price and H = upper knock-out level. Here, the realized
maxima of Sτ , 0 ≤ τ ≤ ti−1 is sampled on each business day.

• The delivery of one or two stocks is determined by Sti ≥ K or other-

wise, independent of whether knock-out occurs or not on ti.

141
n= total number of observation dates
cuo = up-and-out barrier call option
puo = up-and-out barrier put option
∑n
Fair value of an accumulator = i=1 [cuo (ti; K, H) − 2puo(ti ; K, H)].

• When Sti ≥ K, the ti-maturity put option is out-of-the-money and the

ti-maturity call option has the payoff Sti − K.

• When Sti < K, the call option is out-of-the-money and the put option
becomes in-the-money with payoff K − Sti . When the two put options
are in short position, the payoff is −2(K − Sti ) = 2(Sti − K).

The up-and-out barrier put and call price formulas are available under
continuous monitoring of the barrier. Approximation formulas of barri-
er options under discrete monitoring of the barrier can be obtained by
appropriate adjustment to the continuous monitoring counterpart.

142
Cautious note

In an actual contract, when the accumulator is not knocked out in an

accumulation period, the accumulated shares within the accumulation
period will be delivered 3 trading dates after the end of the accumulation
period. The accumulator may be knocked out prior to the end of the
accumulation period. When the accumulator is knocked out, the delivery
of the accumulated shares will be 3 trading days after the knock-out
date. This is unlike the above simplified assumption that the stocks
are delivered immediately on each business date. The discount factor in
the strike depends on the delivery date, which is uncertain due to the
uncertainty of the knock-out event.

Due to uncertainty of the delivery date of the accumulated shares, it is

necessary to use the forward shooting grid method that accounts for all
possible cases of delivery of the accumulated shares, either 3 days after
knocked out or 3 days after the end of the previous accumulation period.

143
Numerical studies on risk characteristics [taken from “Accumulator Pric-
ing” by K. Lam et al.
(2009)]

One-year tenor, 21 trading days in each month, n = 252, H = $105. The

initial stock price S0 is $100, quantity bought on each day is either 1 or 2
depending on the stock price staying above the down-region or otherwise.

• Since the accumulator parameters (H and K) are designed so that it

has a near zero-cost structure, the fair price for the sample accumu-
lator is typically small.

144
 %&'("&)*+,"-%"&..*/*)&0-(".-10(&.0,



 !
"#$ "#"  

 

 



      
      
      
      
      
     
     

The parameter values are: S0 = 100, H = 105, r = 0.03, q = 0.00. For

a zero-cost accumulator with monthly settlement at σ = 20%, the fair
discounted purchase price is shown to be 89.32 (interpolated between
K = 90 and K = 84 along the row of σ = 20%). At moderate level of
σ = 20% and S0 = $100, the accumulator contract should set K = 89.32
(slightly more than 10% discount) in order that the initial value is zero.

145
Implied Volatility

For given values of H and K, we find the volatility such that the value of
the accumulator is zero.

GS+

For a fixed value of K, the implied volatility value is higher for a high-
er barrier level H. For a fixed barrier level, the implied volatility is a
decreasing function of the strike price K.

146
• Suppose an investor anticipates a volatility of 25% in the next one
year. This investor will find the barrier-strike combination in the up-
per left corner (bold area in Table) favorable because the implied
volatilities in those cells have implied volatility larger than 25%.

• The investor should be compensated with a higher barrier level and/or

lower purchase price at a higher volatility level since she is shorting
two puts and long only one call with maturity on each business date.

147
Value at risk analysis

• Profit/loss distribution is highly asymmetric.

Probability distribution of profit and loss of the sample accumulator

148
• It has a long left tail meaning that an extreme loss for the investor is
possible.

• An extreme profit for the investor is unlikely as the distribution has

a short right tail. This is because the contract will be knocked out
once the stock price breaches the upper barrier H.

• For the sample accumulator contract analyzed, the lower 5-percentile

is −$2424.50. This means that at the maturity of the contract, there
is a 5% chance to run a loss more than $2424.50.

• For the seller of the contract, we can estimate his/her corresponding

loss using the same confidence level 0.95. Computation result shows
that the value at risk at maturity is $841.01 with 95% confidence.
Based on these two values at risk, we can conclude that the seller
runs a much smaller risk than the buyer.

149
Greek values calculations (S0 = 100, H = 105, K = 90, r = 0.03, q = 0,
σ = 0.2, n = 252)

#$$% &' ())*+*,(-&# )&.-#()-

 !"
     

     

 

      





/ "0 
     

     

 
      




1  2 
     

Gamma is the sensitivity of delta to stock price. Vega is the sensitivity

of contract value to volatility. Delta, gamma, and vega are all sizable
because an accumulator contract is composed of many option contracts
with varying expiration dates.

150
• There is an asymmetry in the delta and vega values. When the spot
price is low (say S = 88), the magnitude of delta and vega values are
much larger than those when the spot price is high (say S = 104).

• Delta values are decreasing function of S because gamma values re-

main at a negative level. Delta has a magnitude of 288.05 (delay
settlement) when S = 88, but its magnitude drops to −6.47 when
S = 104. This means that losing buyers will be more vulnerable to
price changes than winning buyers.

• Vega has a magnitude of 12201 when S = 88, but drops to a mag-

nitude of 2978 when S = 104 meaning that compared to winning
buyers, losing buyers are more vulnerable to volatility changes as well.
This may be one reason why some buyers of the contract become
very desperate when the market turns south.

• This asymmetry in risk exposure between the two parties is consistent

with the finding that the value at risk of the buyer is several times
that of the seller.

151