Professional Documents
Culture Documents
Arch Garch Guide 2
Arch Garch Guide 2
Chapter 10
1
Motivation:
• Until the early 80s econometrics had focused
almost solely on modeling the means of series,
i.e. their actual values. Recently however we
have focused increasingly on the importance of
volatility, its determinates and its effects on
mean values.
Definition of Volatility
• Suppose that Si is the value of a variable on day
i. The volatility per day is the standard
deviation of ln(Si /Si-1)
1 m
u un i
m i 1 4
Simplifications Usually Made in
Risk Management
• Define ui as (Si−Si-1)/Si-1
• Assume that the mean value of ui is zero
daily - short near 0
• Replace m-1 by m
(high frequent)
1 m
This gives s n = å
2
un-2 i
m i=1
5
Implied Volatilities
• Of the variables needed to price an
option the one that cannot be observed
directly is volatility
6
VIX Index: A Measure of the Implied
Volatility of the S&P 500
7
Linear Regression –the Workhorse
of Financial Modeling
• General form for the linear regression of Yt on
X1t , X 2 t ,..., X kt is
Yt 1 β 0 β1X1t β 2 X 2 t ... β k X kt ε t 1
2
ε t ~ N(0, σ )
Yt E t 1[Yt | X t 1 ] σε t
10
Variation in the Conditional
Variance
•To allow a non-constant conditional variance in the
model, multiply the white noise term by the
conditional standard deviation. This product is
added to the conditional mean as in the previous
slide.
Yt E t 1[Yt | X t 1 ] σ t ε t
15
Normal and Heavy-Tailed
Distribution
16
Are Daily Changes in Exchange
Rates Normally Distributed?
17
ARCH and GARCH
18
References
The classics:
• Engle, R.F. (1982), Autoregressive Conditional
Heteroskedasticity with Estimates of the Variance of
U.K.
Introduction/Reviews:
• Bollerslev T., Engle R. F. and D. B. Nelson (1994),
ARCH Models
i 1 i n i
m
2
n u 2
where
m
i 1
i 1
20
ARCH(q) Model
Engle(1982)
Auto-Regressive Conditional Heteroscedasticity
Yt βX t σ t ε t where ε t ~ N(0,1)
σ γVL i 1 α i (σ t i ε t i )
2 q 2
t
where
q
γ αi 1
i 1
22
EWMA Model
• In an exponentially weighted moving
average model, the weights assigned to
the u2 decline exponentially as we move
back through time
• This leads to (yesterday+
σ λσ
2
t
2
t 1 (1 λ)(Yt E t 1[Yt | X t 1 ]) 2
23
Attractions of EWMA
• Relatively little data needs to be stored
• We need only remember the current
estimate of the variance rate and the
most recent observation on the
innovation to the market variable
• Tracks volatility changes
• RiskMetrics uses = 0.94 for daily
volatility forecasting
24
GARCH(p,q)
Bollerslev (1986)
In empirical work with ARCH models high q
is often required, a more parsimonious
representation is the Generalised ARCH
model VL= average of LT variance
p q
σ t ω α i σ ε
2 2 2
t i t 1 β jσ 2
tj
i 1 j1
ω γVL
25
GARCH (1,1)
In GARCH (1,1) we assign some weight
to the long-run average variance rate
σ ω ασ
2
t
2
ε 2
t 1 t 1 βσ 2
t 1
ω γVL
ω
VL
1 α β
Since weights must sum to 1
+ + =1
26
Example
• Suppose
σ 2t 0.000002 0.13 σ 2t 1ε 2t 1 0.86 σ 2t 1
27
Example continued
• Suppose that the current estimate of the
volatility is 1.6% per day and the most
recent percentage change in the market
variable is 1%.
• The new variance rate is
0.000002 013
. 0.0001 086
. 0.000256 0.00023336
28
Independence vs. Zero
Correlation
• Independence implies zero correlation but not
vice versa
• GARCH processes are good examples
• Dependence of the conditional variance on the
past is the reason the process is not
independent
• Independence of the conditional mean on the
past is the reason that the process is
uncorrelated
Other Models
• Many other GARCH models have been
proposed
30
Variance Targeting
• One way of implementing GARCH(1,1)
that increases stability is by using
variance targeting
• We set the long-run average volatility
equal to the sample variance
• Only two other parameters then have to
be estimated
31
Maximum Likelihood Methods
• All parameters in GARCH models are
simply estimated by maximum likelihood
using the same basic likelihood function,
assuming normality
T
log( L) ( log( t ) t / t )
2 2 2
i 1
p(1 p) 9
33
Example 2
Estimate the variance of observations from a
normal distribution with mean zero
Observations are: u1 , u2 ,........, um
n 1 ui2
Maximize : exp
i 1 2v 2v
n
ui2
or : ln(v)
i 1 v
1 n 2
This gives : v ui
n i 1
35
Application to GARCH (1,1)
We choose parameters that maximize
n
u
2
i
ln(vi )
i 1 vi
36
Forecasting and Persistence with
GARCH(1,1) model
σ 2t 1 ω ασ 2tε 2t βσ 2t (1 α β) VL ασ 2tε 2t βσ 2t
σ 2t 1 VL α(σ 2t ε 2t VL ) β(σ 2t VL )
Taking expectation at time t
E t (σ 2t 1 VL ) (α β) E t [σ 2t VL ]
By repeated substitutions:
E t (σ 2t j ) VL (α β) j (σ 2t VL )