Theoretical Exam

Computational Finance (6 ECTS)

Computational Science
University of Amsterdam
 Figure 1: Two-period binomial tree

1 Binomial tree model

We consider a binomial tree model for the evolution of a stock S. Our assumptions are the
following:

1. After time ∆t the stock Sn+1 can have two values: Sn · u or Sn · d with 0 < d < u and
u · d = 1.

2. The risk-neutral probability of an up movement is denoted p.

3. The expected return of the asset is the risk-free rate: E (Sn+1 ) = Sn · er∆t .

4. For small values of ∆t, the variance of the stock price change is approximately Sn2 σ 2 ∆t.

1.1 Parameters p and u (15 points)

In this question you will derive expressions for p and u.
er∆t −d
a. Use assumptions 1-3 to show that p = u−d .
b. Use assumptions 1-3 and the result of (a) to show that
   
2 r∆t 1 2r∆t
Var (Sn+1 ) = Sn e u+ −1−e .
u

With assumption 4, the result of (b), and several Taylor expansions, it can be shown that
1
u+ ≈ σ 2 ∆t + 2
u
√
c. Solve the equation above for u and subsequently show that eσ ∆t is a good approxi-
mation of u for small ∆t.
1.2 Option pricing (15 points)
Now, consider a two-period binomial tree model. See Figure 1. Use this tree to price a
standard Lookback call option with payoff
 
max ST − min St , 0 ,
0≤t≤T

where St denotes the value of the stock at time t and T denotes the maturity. Let VLB (t)
denote the value of a standard Lookback call option on a non-dividend-paying stock S
with a maturity of one year. Let the one year interest rate be 3% and the current price
of the stock be e50. Furthermore, assume that the volatility is 20%. Use the expressions
for p, u and d = u1 as derived in questions (a)-(c). Complete the two-period tree for the
stock S and compute VLB (0), the fair price of a standard Lookback call option based on
the two-period model.

2 Monte Carlo methods

Consider a stock price that follows a geometric Brownian motion given by the SDE

dSt = rSt dt + σSt dWt , (1)

where r is the risk-free interest rate, σ is the (constant) volatility, t ∈ [0, T ] the time and S0 the
stock price at time 0. Assume that N equidistant (in time) values are observed in the interval
[0, T ].

2.1 European option (15 points)

a. Explain how you would compute the fair value for a European put option with expiry date
T (in years) and strike price K (in EUR) using Monte Carlo simulation.

b. The hedge parameter ∆ in Monte Carlo can be estimated by the bump-and-revalue method,
explain why using the same seed reduces the standard error of the estimation.

c. How can the standard error be estimated in a Monte Carlo simulation?

2.2 Digital option (15 points)

Consider a digital call option which pays 1 euro if the stock price at expiry is higher than the
strike and otherwise nothing.

a. Explain how to calculate the hedge parameter ∆ with the bump-and-revalue method,
explain why this is not working for a digital option.

b. Explain how the likelihood ratio method works for the estimation of ∆ of a digital call
option on a stock.
3 Finite difference methods
In 1900 (long before the famous publication of Black and Scholes), Louis Bachelier studied a
stock price model which in contemporary form has risk-neutral dynamics

dSt = rSt dt + σdWt (2)

with interest rate r, volatility σ, time t ∈ [0, T ] and W denoting a Brownian motion. Consider a
European call option, written on a stock S modeled by the SDE of (2), with pay-off at maturity
t = T given by V (S, T ) = (S − K)+ .

3.1 The Bachelier PDE (15 points)

a. Apply Itô’s lemma to derive the risk-neutral dynamics of the option price V and use this
to prove that the diffusion-coefficient of V is equal to σ ∂V
∂S .

Consider a portfolio with value Π(t) composed by selling one option and buying ∆(t) units of
the underlying stock S.

b. Compute the dynamics of Π and prove that its SDE is independent of W if and only if
∆(t) = ∂V
∂S (t).

c. According to the no-arbitrage principle, the following relation should hold: dΠ = rΠdt.
Use this relation and your results from (a.) and (b.) to show that the Bachelier PDE is
given by

∂V 1 ∂2V ∂V
+ σ 2 2 + rS = rV (3)
∂t 2 ∂S ∂S

3.2 The Crank-Nicolson scheme (15 points)

Under the transformation τ = T − t, (3) is equivalent to

∂V 1 ∂2V ∂V
− + σ 2 2 + rS = rV (4)
∂τ 2 ∂S ∂S
Let vjn = V (Sj , τn ), ∆S = Sj+1 − Sj and ∆τ = τn+1 − τn .

a. Derive the forward and backward approximations of the first derivative with respect to
time
vjn+1 − vjn vjn − vjn−1
 n  n
∂v ∂v
≈ , ≈
∂τ j ∆τ ∂τ j ∆τ
and the central approximation of the second derivative with respect to space
 2 n n
∂ v vj+1 − 2vjn + vj−1
n
≈
∂S 2 j ∆S 2

using Taylor expansions around vjn . Also provide the order of the error.
Consider the special case r = 0. In that case, the Crank-Nicolson scheme for the Bachelier PDE
(4) can be represented in matrix form as
 
 n k1
v1 0
 
A~v n+1 = B~v n + ~k n , ~v n =  ...  , ~k n =  ... 
   
n
 
vM 0
k2
   
a0 a1 b0 b1
a−1 a0 a1  b−1 b0 b1 
   
A=
 . .. . .. . .. 
, B =
 . . . . . . 




 . . . 

 a−1 a0 a1   b−1 b0 b1 
a−1 a0 b−1 b0

b. Use the results of (a.) to derive the matrix entries a−1 , a0 , a1 , b−1 , b0 and b1 .

c. The Dirichlet boundary conditions state

(
0 for S → 0
V (S, t) ≈ (5)
S for S → ∞

Use this to compute k1 and k2 . Also provide the initial condition of the scheme.

d. State two advantages of the Crank-Nicolson scheme in comparison to the explicit FTCS-
scheme.