Estimating Volatilities and

Correlations

Standard Approach to Estimating
Volatility
Define o
n
as the volatility per day between
day n-1 and day n, as estimated at end of day
n-1
Define S
i
as the value of market variable at
end of day i
Define u
i
= ln(S
i
/S
i-1
)


o
n n i
i
m
n i
i
m
m
u u
u
m
u
2 2
1
1
1
1
1
=

( )
Simplifications Usually Made
Define u
i
as (S
i
-S
i-1
)/S
i-1
Assume that the mean value of u
i
is zero
Replace m-1 by m

This gives

o
n n i
i
m
m
u
2 2
1
1
=

=

Weighting Scheme
Instead of assigning equal weights to the
observations we can set

o o
o
n i n i
i
m
i
i
m
u
2 2
1
1
1
=
=

=
=

where
ARCH(m) Model
In an ARCH(m) model we also assign
some weight to the long-run variance rate,
V
L
:

=
=

= o +
o + = o
m
i
i
m
i
i n i L n
u V
1
1
2 2
1
where
EWMA Model
In an exponentially weighted moving
average model, the weights assigned to
the u
2
decline exponentially as we move
back through time
This leads to

2
1
2
1
2
) 1 (

+ o = o
n n n
u
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 market variable
Tracks volatility changes
RiskMetrics uses = 0.94 for daily
volatility forecasting
GARCH (1,1)
In GARCH (1,1) we assign some weight to
the long-run average variance rate



Since weights must sum to 1
+ o + | =1

2
1
2
1
2

|o + o + = o
n n L n
u V
Cont
Setting e = V the GARCH (1,1) model is


and


| o
e
=
1
L
V
2
1
2
1
2

|o + o + e = o
n n n
u
Illustration
Suppose


The long-run variance rate is 0.0002 so
that the long-run volatility per day is 1.4%
o o
n n n
u
2
1
2
1
2
0000002 013 086 = + +

. . .
cont
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

The new volatility is 1.53% per day
0000002 013 00001 086 0000256 000023336 . . . . . . + + =
GARCH (p,q)

o e o | o
n i n i j
j
q
i
p
n j
u
2 2
1 1
2
= + +

= =


Maximum Likelihood Methods
In maximum likelihood methods we
choose parameters that maximize the
likelihood of the observations occurring
Illustration-1
We observe that a certain event happens
one time in ten trials. What is our estimate
of the proportion of the time, p, that it
happens?
The probability of the event happening on
one particular trial and not on the others is

We maximize this to obtain a maximum
likelihood estimate. Result: p=0.1

9
) 1 ( p p
Illustration-2
Estimate the variance of observations from a
normal distribution with mean zero

[
=
=
=
=
(

|
|
.
|

\
|

t
m
i
i
m
i
i
m
i
i
u
m
v
v
u
v
v
u
v
1
2
1
2
1
2
1
) ln(
2
exp
2
1
: Result

: maximizing to equivalent is this logarithms Taking
: Maximize
Application to GARCH
We choose parameters that maximize

[
=
=
(

|
|
.
|

\
|

t
m
i
i
i
i
i
i
m
i
i
v
u
v
v
u
v
1
2
2
1
) ln(
2
exp
2
1
or
Excel Application
Start with trial values of e, o, and |
Update variances
Calculate

Use solver to search for values of e, o,
and | that maximize this objective
function
Important note: set up spreadsheet so
that you are searching for three numbers
that are the same order of magnitude

=
(

m
i
i
i
i
v
u
v
1
2
) ln(
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
How Good is the Model?
The Ljung-Box statistic tests for
autocorrelation
We compare the autocorrelation of the

u
i
2
with the autocorrelation of the u
i
2
/o
i
2

Forecasting Future Volatility



The variance rate for an option expiring on
day m is



) ( ) ( ] [
2 2
L n
k
L k n
V V E o | + o + = o
+
| |

=
+
o
1
0
2
1
m
k
k n
E
m
Forecasting Future Volatility
| |
|
|
.
|

\
|

+
| + o
=

L
aT
L
V V
aT
e
V
T
a
) 0 (
1
1
ln
252
is option day - a for annum per volatility The
Define
Volatility Term Structures
The GARCH (1,1) suggests that, when calculating
vega, we should shift the long maturity volatilities less
than the short maturity volatilities
Impact of 1% change in instantaneous volatility for
Japanese yen example:
Option Life
(days)
10 30 50 100 500
Volatility
increase (%)
0.84 0.61 0.46 0.27 0.06
Correlations and Covariances
Define x
i
=(X
i
-X
i-1
)/X
i-1
and y
i
=(Y
i
-Y
i-1
)/Y
i-1
Also
o
x,n
: daily vol of X calculated on day n-1
o
y,n
: daily vol of Y calculated on day n-1
cov
n
: covariance calculated on day n-1
The correlation is cov
n
/(o
u,n
o
v,n
)

Updating Correlations
We can use similar models to those for
volatilities
Under EWMA
cov
n
= cov
n-1
+(1-)x
n-1
y
n-1

Under GARCH (1,1)

1
1
1
cov cov

+ + =
n
n
n n
y x | o e
Positive Finite Definite Condition
A variance-covariance matrix, O, is
internally consistent if the positive semi-
definite condition


for all vectors w
w w
T
O > 0
Example
The variance covariance matrix



is not internally consistent
1 0 0 9
0 1 0 9
0 9 0 9 1
.
.
. .
|
\

|
.
|
|
|