Applications of state space models in finance
Applications of state space models in finance
Applications of state space models in finance
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
62 5 Conditional heteroskedasticity <strong>models</strong><br />
have been proposed to capture asymmetric effects. Below, only the two most <strong>in</strong>fluential<br />
nonl<strong>in</strong>ear extensions are discussed. For a more comprehensive overview <strong>of</strong> the<br />
broad spectrum <strong>of</strong> nonl<strong>in</strong>ear GARCH specifications, see, for example, Hentschel (1995)<br />
or Franses and van Dijk (2000, ¢ 4.1.2).<br />
5.1.2.1 Exponential GARCH<br />
The first GARCH extension, <strong>in</strong> which the conditional volatility depends on both the<br />
size as well as the sign <strong>of</strong> lagged shocks, was proposed Nelson (1991). In its simplest<br />
specification with p = q = 1, the exponential GARCH (EGARCH) model can be written<br />
as<br />
log(ht) = ω + γ1zt−1 + ϑ1(|zt−1| − E(|zt−1|)) + δ1 log(ht−1). (5.13)<br />
Due to the model<strong>in</strong>g <strong>of</strong> ht <strong>in</strong> logarithms, no restrictions on the coefficients have to be<br />
imposed to ensure nonnegativity <strong>of</strong> the conditional volatility. Let the function g(zt) be<br />
def<strong>in</strong>ed as g(zt) := γzt + ϑ1(|zt| − E(|zt|)) where the two summands relate to the sign<br />
and to the magnitude effect. By rewrit<strong>in</strong>g it as<br />
g(zt) = (γ1 + ϑ1)ztI(zt > 0) + (γ1 − ϑ1)ztI(zt < 0) − ϑ1(E|z1|), (5.14)<br />
with I(·) be<strong>in</strong>g an <strong>in</strong>dicator function, it can be seen how asymmetric effects are <strong>in</strong>corporated:<br />
while the term (γ1 + ϑ1) is affected by positive shocks, negative shocks have an<br />
impact on (γ1 − ϑ1). Generally, positive shocks have a smaller effect on ht than negative<br />
shocks <strong>of</strong> equally sized positive shocks (cf. Engle and Ng 1993). This becomes clear by<br />
hav<strong>in</strong>g a look at the news impact curve (NIC) 13 for the EGARCH model, which is given<br />
by<br />
NIC � ɛt|ht = σ 2� ⎧ � �<br />
(γ1 ⎪⎨<br />
+ ϑ1)<br />
A exp ɛt , for ɛt > 0,<br />
σ<br />
= � �<br />
(5.15)<br />
(γ1 ⎪⎩<br />
− ϑ1)<br />
A exp ɛt , for ɛt < 0,<br />
σ<br />
with A = σ2δ1 �<br />
exp(ω − ϑ1 2/π), for parameter constellation γ1 < 0, 0 ≤ ϑ < 1 and<br />
ϑ1 + δ1 < 1. As the EGARCH model is not differentiable with respect to zt−1 at zero,<br />
its estimation is more difficult than that <strong>of</strong> alternative asymmetric <strong>models</strong>. Another<br />
problem is related to forecast<strong>in</strong>g. Usually, the researcher is <strong>in</strong>terested <strong>in</strong> forecast<strong>in</strong>g ht+l<br />
and not log ht+l. This requires a transformation that depends on the complete l-step<br />
ahead forecast distribution, f(yt+l|Ωt), which is generally not available <strong>in</strong> closed-form.<br />
5.1.2.2 GJR-GARCH<br />
Glosten et al. (1993) and Zakoian (1994) <strong>in</strong>dependently <strong>in</strong>troduced an alternative nonl<strong>in</strong>ear<br />
extension to capture asymmetric effects: the GJR-GARCH or threshold GARCH<br />
model. Accord<strong>in</strong>g to L<strong>in</strong>g and McAleer (2002), the GJR-GARCH model represents the<br />
13 Engle and Ng (1993) <strong>in</strong>troduced the NIC, def<strong>in</strong>ed as the functional relationship between<br />
the conditional variance and lagged shocks, as a measure <strong>of</strong> how the arrival <strong>of</strong> new <strong>in</strong>formation<br />
is reflected <strong>in</strong> volatility estimates. The NIC can be used to compare different GARCH <strong>models</strong>.