1
$\begingroup$

I'm reading this paper by Abosedra et. al (2006), where they study the volatility of US natural gas prices.

They report the estimation from a AR(1)-GARCH(1,1) model and a AR(1)-GJR-GARCH(1,1) model as in the image below.

Why do they also report $ \alpha+ \beta + \frac12 \lambda $? And how does this turn out to be exactly the same as $ \alpha + \beta $ in the AR(1)-GARCH(1,1) model?

Note: $\alpha, \beta, \gamma$ here are the ARCH, GARCH and Asymmetry (GJR model) parameters respectively. The equation(s) estimated is given below the table, if necessary.

enter image description here

Estimated equation(s):

           enter image description here

$\endgroup$
1
$\begingroup$

I am not sure if half the asymmetry parameter has any useful interpretation on its own, but it does in the context of $\alpha+\beta+\frac{1}{2}\lambda$. Judging from the table you have included, the authors probably want to measure persistence in volatility. As is clarified in the thread "Persistence in GJR-GARCH (asymmetric GARCH)", the persistence of GJR-GARCH(1,1) is $$\alpha+\beta+\frac{1}{2}\lambda$$ if the distribution of standardized innovations is symmetric. The persistence of vanilla GARCH(1,1) is $$\alpha+\beta.$$ You have both of these in the table.

$\endgroup$
2
  • $\begingroup$ Is there any significance to the fact that the persistence measures for both the models are exactly the same? $\endgroup$
    – WorldGov
    Mar 17 at 13:11
  • 1
    $\begingroup$ @WorldGov, this is an interesting finding. For those who care directly about persistence, this shows its measurement is robust to model choice between vanilla GARCH and GJR-GARCH. This holds for this specific time series but not necessarily in general. $\endgroup$ Mar 17 at 13:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.