Estimation of GARCH type models with IV as exogenous var

[15291057]

Hey Forum!

We are conducting a study on the informational content of implied volatility for the South African economy. In one part, we follow a study by Day and Lewis (1992), in which they include implied volatility (similar to the VIX index) as an exogenous variable in the GARCH model. My question is, if I specify this in Eviews, and include it in the "Variance Regressors" input box to specify it as a exogenous variable in the variance equation, should I use Implied Volatility as is, or should I be using the squared value of implied vol. (Basically, should I include VIX or VIX^2)

Thanks in advance!

[trubador]

I do not think there is one definite way here. If you are trying to estimate the conditional volatility with controlling the impact of global turmoil, then you can use the square of VIX since it is simply an annualized standard deviation. However, if you are interested in global developments only as a reflection of the change in the sentiment, then you can prefer VIX as a proxy. Please keep in mind that you can include the variable as a regressor in mean equation as well. I would experiment with alternative specifications and see which one fits the data better...