You are on page 1of 26

A U T U M N    2 0 0 9

MONTE CARLO SIMULATIONS

Numerical Methods in Finance (Implementing Market Models)


COMPUTATIONAL FINANCE
MSc
©Finbarr Murphy 2007

Lecture Objectives
 Introduction to Monte Carlo simulations
COMPUTATIONAL FINANCE
MSc
©Finbarr Murphy 2007

Agenda
Page

1
Monte Carlo Methods 2

3
COMPUTATIONAL FINANCE
MSc

3
©Finbarr Murphy 2007

Monte Carlo Simulations

 Recall from FI-6051, Lecture 3.2

 The variable xt is said follow a generalized Wiener


process if
dxt = adt + bdz t
where a and b are constants, and dzt is the
continuous-time Wiener increment as before

 The parameter a is referred to as the drift rate of


the process
©Finbarr Murphy 2007

Monte Carlo Simulations

 The problem here is the constant expected return.


This should be expressed relative to the stock price,
not as a simple flat constant

 So, instead of μ, a constant, the expected drift rate


should be μS

 So, assuming no volatility

δS = µSδt
 or
dS
= µdt asδt → 0
S
©Finbarr Murphy 2007

Monte Carlo Simulations

 But there is volatility and this volatility is also


expressed as a percentage of the current stock price

 Or dS = µSdt + σSdz
dS
= µdt + σdz
S
 This continuous time model is known as Geometric
Brownian Motion (GBM)
 In discrete terms:

δS
= µ δt + σ ε δt
S
©Finbarr Murphy 2007

Monte Carlo Simulations

 We can simulate the path followed by S by dividing


time into a series of short intervals of length δt

 Look at the following example taken from Hull:

 A stock pays no dividends


 Has a volatility of 30% per annum
 Has an expected return of 15% per annum (CC)
 The process for the stock price is given by:

dS
= 0.15dt + 0.30dz
S
©Finbarr Murphy 2007

Monte Carlo Simulations

 If S is the stock price now and δS is the increase in


the stock price over the next small time interval,
then:

δS
= 0.15δt + 0.30ε δt
S
 Now, assume the time interval is one week, so:

δt = 1 / 52 = 0.0192
©Finbarr Murphy 2007

Monte Carlo Simulations

 Further, assume that the stock price now is €100


 So:
δS = 0.288 + 4.16ε
 Or, the price increase is a random draw from a
normal distribution with a mean €0.288 and a
standard deviation €4.16

 Let’s visualise this:


©Finbarr Murphy 2007

Monte Carlo Simulations

 The Standard Normal Distribution X->N(0,1)


©Finbarr Murphy 2007

Stock Price Monte Carlo Simulations

ε = -0.56
New Stock Price
€100 €100
+€0.288
+€4.16*-0.56
-------------------
€97.96
===========
1 Week €0.288
+ ε x €4.16
©Finbarr Murphy 2007

Stock Price Monte Carlo Simulations

ε = 0.29
New Stock Price
€97.96
+€0.288
+€4.16*0.29
-------------------
€99.45
€97.96 ===========

1 Week €0.288
+ ε x €4.16
©Finbarr Murphy 2007

Stock Price Monte Carlo Simulations

€100

€97.96

1 Week 1 Week 1 Week


©Finbarr Murphy 2007

Stock Price Monte Carlo Simulations

€100

1 Week 1 Week 1 Week 1 Week 1 Week 1 Week 1 Week


©Finbarr Murphy 2007

Monte Carlo Simulations

We are left with a


Geometric Brownian Motion (GBM)
Stock Price

stock price process

€100

1 Week 1 Week 1 Week 1 Week 1 Week 1 Week 1 Week


©Finbarr Murphy 2007

Monte Carlo Simulations

Notice how our GBM representation of stock price


movement can mimic actual share price movement
©Finbarr Murphy 2007

Monte Carlo Simulations

Notice the trend over a long period (20 years)


©Finbarr Murphy 2007

Monte Carlo Simulations

 Monte Carlo simulation was named for Monte Carlo,


Monaco, where the primary attractions are casinos
containing games of chance. Games of chance such
as roulette wheels, dice, and slot machines, exhibit
random behavior.

 Using a random draw from N(0,1) and the GBM we


can simulate a stock price movement
©Finbarr Murphy 2007

Monte Carlo Simulations

S (t + δt ) = S (t ) + µˆS (t )δt + σS (t )ε δt
 Where ε is a random sample drawn from a normal
distribution with mean 0 and standard deviation of 1

 In practice, we use dlnS rather than dS


 From Itô’s lemma, the above process becomes:

 σ 
2
d ln S =  µˆ − dt + σdz
 2 
©Finbarr Murphy 2007

Monte Carlo Simulations

 So we can construct a path using:

S (t + δt ) = S (t )e [ ( µˆ −σ 2 )δt +σ ε δt ]
2

 The following graph


illustrated a single MC
simulation for
 S0 = 100
 μ = 7.5%
 Δ = 25%
©Finbarr Murphy 2007

Monte Carlo Simulations

 What do 100 simulations look like?


©Finbarr Murphy 2007

Monte Carlo Simulations

 Monte Carlo valuation works by considering a random


path described by a GBM process
 At payoff, the option is valued [max(0,S-X)] for a call
option
 Another sample path is generated and the option
payoff is again computed
 This is repeated for many paths
 The average payoff is calculated
 The average payoff is discounted at the risk free rate
to estimate the current value of the derivative
©Finbarr Murphy 2007

Monte Carlo Simulations

 MatLab Code (see/run euroMCSimple.m)


©Finbarr Murphy 2007

Monte Carlo Simulations

 From euroMCSimple.m …
No simulations Option Price Std Dev Std Err
100 2.2917 2.9168 0.2917
500 2.3904 2.9585 0.1323
1000 2.7962 3.3691 0.1065
2500 2.7522 3.2909 0.0658
5000 2.6195 3.1714 0.0449
10000 2.7361 3.2949 0.0329
20000 2.6638 3.2694 0.0231
50000 2.6701 3.2693 0.0146
75000 2.6857 3.28 0.012
100000 2.69 3.2792 0.0104
©Finbarr Murphy 2007

Monte Carlo Simulations

 Note the convergence of the option price and the


standard error reduction as the number of
simulations is increased
 Of course, the greater the number of simulations,
the greater time taken.
 We need to find ways of reducing the computational
time
©Finbarr Murphy 2007

Recommended Texts
 Required/Recommended
 Clewlow, L. and Strickland, C. (1996) Implementing derivative
models, 1st ed., John Wiley and Sons Ltd.
— Chapter 4

 Additional/Useful
 Hull, J. (2005) Options, futures and other derivatives, 6th ed.,
Prentice Hall
— Chapters 17 P-410
COMPUTATIONAL FINANCE
MSc

26

You might also like