Estimation in Financial Models - RiskLab
Estimation in Financial Models - RiskLab
Estimation in Financial Models - RiskLab
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Theorem 6 Depend<strong>in</strong>g on the value of we have the follow<strong>in</strong>g asymptotic<br />
behaviour for the normalized deviation q I n ()(^ n , ):<br />
q<br />
<br />
lim P I<br />
n,!1 n ()(^ n , ) z =<br />
8<br />
><<br />
>:<br />
(z); jj < 1;<br />
H (z); jj =1;<br />
Ch(z); jj > 1;<br />
where (z) is the standard normal distribution, Ch(z) is the Cauchy distribution<br />
and H (z) is the distribution of the random variable<br />
<br />
2 3 2<br />
where W denotes a Wiener process.<br />
W 2 (1) , 1<br />
R 1<br />
0<br />
W 2 (s)ds ;<br />
Hence, <strong>in</strong> the stationary case jj < 1 the normalized deviation q I n ()(^ n ,)<br />
is asymptotically normally distributed, whereas <strong>in</strong> the other cases it has<br />
as limit distribution the Cauchy distribution or the quite unexpected H <br />
distribution. The question arises wether we can reduce the number of limit<br />
distributions by modify<strong>in</strong>g the normaliz<strong>in</strong>g factor q I n (). Indeed, <strong>in</strong>stead of<br />
us<strong>in</strong>g the Fisher <strong>in</strong>formation I n () =EhMi n , choos<strong>in</strong>g the stochastic Fisher<br />
<strong>in</strong>formation hMi n as normaliz<strong>in</strong>g factor we obta<strong>in</strong><br />
Theorem 7<br />
lim<br />
n,!1 P <br />
q<br />
(<br />
(z); jj 6=1;<br />
hMi n (^ n , ) z =<br />
H (z); jj =1:<br />
Summariz<strong>in</strong>g: the MLE ^ n is (strongly) consistent and the normalized deviations<br />
q I n ()(^ n , ) and q hMi n (^ n , ) are asymptotically distributed as<br />
shown <strong>in</strong> both theorems.<br />
3.2.2 ARCH and GARCH models<br />
In the follow<strong>in</strong>g we concentrate on ARCH and GARCH models and especially<br />
on estimation <strong>in</strong> these models. We refer to the fundamental papers by<br />
Engle [22] and Bollerslev [12] and to Bollerslev, Chou and Kroner [13] and<br />
Bollerslev, Engle and Nelson [14].<br />
Conventional econometric time series models assume constantvariance. However,<br />
over a decade ago risk and uncerta<strong>in</strong>ty considerations lead to the development<br />
of new econometric time series models that allow for the model<strong>in</strong>g<br />
45