Open issues in equity derivatives modelling
Talk Outline
Equity derivatives at SG
A brief history of equity derivative products
Prehistory –
1997
History
1997 –
2003
Modern times 2003 –
Modelling issues, algor
ithmic issues
Risk m
easurem
ent and management
Conclusion
g
Equity derivatives at SG
SG regarded by industry particip
ants as No 1 in
equity derivatives
g
A brief history of equity derivative products
Prehistory –
1997
Products
Risks
Barrier options / Digitals
Skew
: level / dynamics (little)
Max / Min options
same
Asian options
Smile
Basket options
Correlation (level)
Vo
latility swaps
Smile, V
olOfV
ol
Simple cliquets
Forward smile
Models / algos
-Black S
choles / local vol
-PDE / straight Monte Carlo
gomi+ ⎟ ⎠⎞ ⎜ ⎝ ⎛ − ∑ K t S N i 1 + ⎟ ⎠⎞ ⎜ ⎝ ⎛ − ∑ K i T S N 1
()
+ ⎟ ⎠⎞ ⎜ ⎝ ⎛ − K t S t max K σ k S k S T k ˆ 2 1 ln 1 − ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ − ∑ + ⎞ ⎟ ⎟ ⎠ ⎜ ⎜ ⎝ ⎛ − K S S T T2 1A brief history of equity derivative products
History
-1 -1997 –
2005
Capital-guaranteed products di
stributed by retail networks
Everest 1997
5 y
ears
/ 12 s
tocks
¸
Emerald 2004
10 years / 20 stocks
Every year,the stock whose perfomance since
t = 0
is the largest gets f
rozen and
removed from t
he bas
ket, and its level is
floored a t 200% ot its initial value. ¸ 100% + maximum perormance of yearly ba
sket values since
t = 0
, floored at 0.
… and many, many, many, other variations
¸ tryin g to fin d closed -form fo rmu
las for specific exo
tic payo ffs n ow ir relevant and u seless g omi ⎟ ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎜ ⎝ ⎛ + j S j T S 0 min % 100
g
A brief history of equity derivative products
History
-2 1997 –
2005
Variance Swaps
3 months
¸
5 years
Pays realize d varianc e – usually measured using daily returns
stocks / indices
Napoleon
5 y
ears
/ 1 index
Every year, pays coupon reduced by worst
of 12 monthly performanc es of the index.
Accum
ulat
or
3 years / 1 i
ndex
At maturity pays the sum –
if it is positiv e – of the mont hly performances, cappe d and and floored. g omi + ⎞ ⎟ ⎟⎟ ⎠ ⎜⎜ ⎜ ⎝ ⎛ ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ − + 1 min k S k S k C + ⎞ ⎟ ⎟⎟ ⎠ ⎜⎜ ⎜ ⎝ ⎛ ⎟⎟ ⎟ ⎠ ⎞ ⎜⎜ ⎜ ⎝ ⎛ − ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ − −
∑
k k S k S %1 , %1, 1 1 min max T σ k S k S k 2 2 ˆ 1 ln − ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ −∑
Modern times
Corridor variance swaps
Daily varia nce only c ou nte d whe n un
derlying is inside given int
erval
Correlation swaps
Pays realize
d correlation over 3 yea
rs by stocks
of an index
Gap notes
Maturity = 1 year, a series of daily
puts on daily returns of
an
index
with strikes 85%, 90%
Options on realized variance
On indic
es, maturities: 3 months to 2 years
Timer options
Va nilla pa yo ff, paid wh en realized variance Qt reaches se t level:
Hybrids
Equities / Ra tes / Forex / Commodities A rbitrary payoffs g omi []∑
⎟ ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎜ ⎝ ⎛ ∆ − ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ − ∈ k H L k S t σ k S k S 2 2 , ˆ 1 ln 1[]
[
]
[
]
H L H L ,0 , , , , ∞ + + ⎞ ⎟ ⎟ ⎠ ⎜ ⎜ ⎝ ⎛ − ⎟⎟ ⎠ ⎞ ⎜⎜ ⎝ ⎛∑
− 2 2 1 ˆ ln 1 K τ τ τ σ S S T∑
= + ⎟⎟ ⎠ ⎞ ⎜⎜ ⎝ ⎛ = τ i i i τ S S Q 1 2 1 lnModelling issues
–
1
Why not just delta-hedge ?
Variance of residual P&
L too l
ar
ge
¸
use other options
¸
Options are hedged with options
Once we start using options
as hedging instr
uments
Less sensit
ivity to
h
isto
ric
al parameters,
mo
re sensi
st
ivity to
imp
lied
parameters
¸
Model the dynamics of implied parameters
Example of simple cliquet
g omi t 1 T 2 T 12 ˆσ
()
L,
,
ˆ
12r
σ
P
+ ⎞ ⎟ ⎟ ⎠ ⎜ ⎜ ⎝ ⎛ − 1 1 2 T T S S 200 8 2007 2006 2005 200 4 2003 2002 S m ile 3 m o is K = 9 5 - K = 1 0 5 0 1 2 3 4 5 6 7 2/5 /20 01 2 /5/2 0 0 2 2/5 /20 03 2/5 /2 0 0 4 2 /4 /20 05 2/4/2 0 0 6 2 /4 /20 07 2/4/2 0 0 8Modelling issues
–
2
How should calibration be
done ? Do we really
need to calibrate ?
• N ot compulsory: char ge a h edging cost. We hedg
e par ameter p by tr ad ing ins trument O so that s ensit ivity to p •M od el p ri ce P is adjusted so as to includ e h edging cost:
•
Then what is the po
int in calibrating ?
•
Ensures price factors in he
dging c
osts incurred at
t = 0
–
not future costs !
¸
Necessary to
calib
rate mo
del on
relevant set of
hedging instruments
¸
Useless if one is unabl
e to specify how to hedge the ex
otic with the he
dge instruments
g omi dp dO λ dp dP =()
(
)
()
()
()
Market Market Price p p P p O p O λ p P = ≈ − + = ˆ ˆModelling issues
–
3
Volatility risk –
m
odels
•
«
O
ld models
»
• L ocal volatility •H es to n •S A B R • M odels b as ed on process of inst antaneou s variance: •J u m p / L év y•
Challenge: Build models that give
control on joint dynamics of
implied volat
ilities and spot:
•First step: model dynamics of
curve of forward variances
•Next step: model dynamics of
the implied volatility surface
•
Dir ect mod el li ng of d ynamics of implied volatilities is a dead end
• L ow-dimens ional M arkov r epr es entat ion d esir abl e • H ow muc h freedom are we al lowed ? g omi V S dW dt dV dW S V dt dS ) ( + = + = K K E u ro st oxx 5 0 - m at 1 an 10 15 20 25 80 90 10 0 1 10 120 S m ile o rig in a l S m ile S = 1 0 5% S m ile S = 9 5 %
Modelling issues
–
4
Hybrids
•
E quities • Interest rates • Forex • C ommoditi es•Hyb
rid
mo
dels are no
t built
by simp
ly g
lueing together models f
or each asset
class
•
Passive hybrids: payoff in
volves one asset class only
•
Long-dat ed equit y, For ex options • Credit / Equity: convertibl
e bonds
•
Act
ive hyb
rids
:payof
f invo
lves all asset
classes
• R equir e st at e-of-the a rt mo d els fo r ea ch a sse t cla ss • Even local vol calibration for
equity smil
es not easy when
interest rates are stochastic
g
Modelling issues
–
5
Correlation –
how de we put to
gether correlation matrices ?
•
How do we bu ild th e l arg e corr el at ion ma trices n eed ed in hybrid model ling ? • Simpler qu est ion: imagine a 1-f actor s toch. vo l mo d el and a pa yo ff in vo lv in g 2 se cu ri ties • H ow do we s et the cr oss-corr el ations ? •E ve n s im p le r q u es ti on –h ow d o we me asure c orrela ti on s ? • E xa mple of Europ ean / Jap anes e stocks – n o over la p
Correlation –
how de we me
asure correlation risk ?
Corrrelation –
how to model correlation smile ?
g omi o C o C o C o C o C o C o C o C
Europ
e
Japan
?
?
Algorithmic issues
M
onte Carlo
•
How can
we speed
up pricing ?
•
Quasi-random numbers
•
Discretization of SDEs ?
•Callable / putable options
•Computing sensitivi
es to
•Initial conditions •Param
eter
s of dynamics (vol
atilities / correlations, etc..)
g
Conclusion
These are exciting times for
doing quantitative finance
Lots of new instrumen
ts /
p
roduct / algori
th
m
ic issues
Rich mathematical tool
box from which to pick
g