View
1
Download
0
Category
Preview:
Citation preview
© 2004 Brian Eales
Share Price Share Price MovementsMovements
Brian A. EalesApril 2004
Page1 © 2004 Brian Eales
Monte Carlo Simulation
Share Price MovementsShare Price Movements
In continuous time or( )1dzSdtSdS σ+µ=
( )2zStSS ∆σ+∆µ=∆In discrete (measurable) time
Page2 © 2004 Brian Eales
Monte Carlo Simulation
Where:
dS or ∆S represents the change in the stock price
µ represents the mean return on the stock (the drift factor)
σ represents the stock’s volatility
dt or ∆t represents the change in time over which the change in stock price is measured
Page3 © 2004 Brian Eales
Monte Carlo Simulation
The term dz or ∆z captures the uncertainty associated with the stock price movement.
Equations (1) and (2) are examples of a Generalized Wiener process.
Page4 © 2004 Brian Eales
Monte Carlo Simulation
Zero Uncertainty
100
100.1
100.2
100.3
100.4
100.5
100.6
100.7
100.8
100.9
101
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Time
Shar
e Pr
ice
Page5 © 2004 Brian Eales
Monte Carlo Simulation
Zero Drift
88
90
92
94
96
98
100
102
104
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Time
Shar
e Pr
ice
Page6 © 2004 Brian Eales
Monte Carlo Simulation
Generalized Wiener Process
90
92
94
96
98
100
102
104
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Time
Sha
re P
rice
Page7 © 2004 Brian Eales
Monte Carlo Simulation
All Three Evolutions
90
92
94
96
98
100
102
104
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Time
Shar
e Pr
ice
Drift only
Wiener Process
Generalized Wiener Process
Page8 © 2004 Brian Eales
Monte Carlo Simulation
How are the Evolutions Calculated?How are the Evolutions Calculated?
The tracks involving uncertainty have been calculated using Monte Carlo simulation.
A set of random numbers is generated to replace the ∆z term in equation 2.
Page9 © 2004 Brian Eales
Monte Carlo Simulation
In generating this track a further assumption is used:
where ε is normally distributed (0, 1)
ε ∼ N(0, 1) and
∆S/S ∼ N(µ∆t, σ(∆t)½)
( )3tz ∆ε=∆
Processes like this can be used to obtain prices for exotic options.
Page10 © 2004 Brian Eales
Monte Carlo Simulation
ExampleExampleSimulate the daily track of a share whose starting value is 100, return 10% p.a., and volatility 20%p.a.
00274.0day1t =−=∆
000274.0)00274.0)(01.0(returndaily ==
( ) 010468.000274.0)2.0(tvolatilitydaily ==∆σ=
Page11 © 2004 Brian Eales
Monte Carlo Simulation
Taking a N(0, 1) random number (0.11656), the next day’s simulated price is:
( ) ( )( )[ ] 10011656.0010468.0000274.0100SS ++=+∆
1494.100S1 =
1001494.0SSS 1 +==+∆
Repeating this for several evolutions gives the following tracks:
Page12 © 2004 Brian Eales
Monte Carlo Simulation
Share Price Share Price RandomEvolution Changes Numbers100.0000 0.149415 0.11656100.1494 -1.312036 -1.2776898.8374 0.2797973 0.24425799.1172 1.3515742 1.276474
100.4688 1.2878416 1.19835101.7566 1.8739939 1.733133
Page13 © 2004 Brian Eales
Monte Carlo Simulation
LogLogee share price evolutionsshare price evolutions
In the trinomial case trees were developed using loge evolution rather than the underlying’s value. In practice this is also done in simulations, too.
( ) ε∆σ+∆
σ−µ+=∆+ tt2
SlnttSln2
t
ε∆σ+∆
σ
−µ=∆+
tt2
2
SeSSor
Page14 © 2004 Brian Eales
Monte Carlo Simulation
Share Price Share Price Share Price RandomEvolution Evolution Changes Numbers
100 4.60517019 0.0011661 0.11656100.11668 4.60633629 -0.013429 -1.2776898.781215 4.59290745 0.0025028 0.24425799.028757 4.59541029 0.0133081 1.276474100.35545 4.60871836 0.0124903 1.19835101.61678 4.62120864 0.0180884 1.733133
Page15 © 2004 Brian Eales
Monte Carlo Simulation
Problems with the basicmodel
Problems associated with theapproach - constant variance,constant drift, slow.
Antithetic variables: can help tospeed up the process by averagingoutcomes of the mirror image tracks,i.e.
( )2
ppps−+ +
=
Page16 © 2004 Brian Eales
Monte Carlo Simulation
Antithetic Variables Approach
82
87
92
97
102
107
112
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Time
Und
erly
ing
Secu
rity
Pric
Original random numbers
Antithetic counterparts
Page17 © 2004 Brian Eales
Monte Carlo Simulation
Hull approach: bbcvasva pppp +−=
where:
pa is the estimated option variable,
pasv is the simulated stochasticvariance option variable,
pbcv is the simulated control variateoption variable,
pb is the known option variable value.
Page18 © 2004 Brian Eales
Monte Carlo Simulation
Clewlow and Strickland approach:
They use delta, gamma, and vega as control variates. Inthis way the hedge parameters are also obtained as adirect by product of the simulation. For a call option, forexample the simulation proceeds in the normal way butfor each loop the control variate (cv) is accumulated foruse in the expiration formula:
[ ] ( )C S E cvT T= − + ⋅max , 0 β
where: β is set at -1, cv is the control variate found as:
( )cv cv delta S S drif tt t t= + −− 1 0. *
Page19 © 2004 Brian Eales
Monte Carlo Simulation
The call option value at expirationwill be the discounted average of allthe payout values obtained from theM simulated tracks.
Bootstrapping:
Is a parameter free approach andpermits the past data tracks todetermine average future tracks.
All simulation methods can provide ameasure of variance reduction byusing the standard error formula: MX
σ=σ
Page20 © 2004 Brian Eales
Monte Carlo Simulation
Stochastic varianceStochastic variance
wStSS ∆σ+∆µ=∆
1_2
ztaln2
∆σθ+∆
−σβ−θσ=σ∆
Page21 © 2004 Brian Eales
Monte Carlo Simulation
where: ∆σ represents the change in volatility that occurs
over the time period under consideration, a_ represents the
mean reverting value for lnσ volatility to which the
volatility each day, σ, is adjusting at a rate β, θ is the
proportional volatility of the volatility ∆z1 is given by
∆ ∆z t1 1= ε where ε1 is a random drawing from the N(0,1)
distribution adjusted to reflect the volatility’s mean drift
and variance over the discretised time frame used in the
simulations.
Page22 © 2004 Brian Eales
Monte Carlo Simulation
t_
1_
∆
σ−σ=σ∆ λ
Following the work of Hofmann,
Platen, and Schweizer (1992) a third
equation was included to model daily
adjustments to a_. This model takes
the form:
Recommended