0
$\begingroup$

I am trying to estimate a DCC GARCH model with an external regressor in the DCC equation. However, the external regressor option in the rmgarch package is for the VAR option. Is there any way I can include the external regressor please?

$\endgroup$
10
  • $\begingroup$ I guess you can include external regressors when defining the univariate models via rugarch (which has to be done before defining and fitting the DCC part), can you not? $\endgroup$ Mar 26, 2023 at 13:57
  • $\begingroup$ In this case, I am looking to extract the dynamic correlation coefficients on the basis of the impact of the external variable. Whereas in ugarch I am specifying on the basis of one variable. $\endgroup$
    – Jyoti Nair
    Mar 26, 2023 at 20:04
  • $\begingroup$ So you want an external regressor in the DCC equation? I have not heard of an implementation that does that. Can you be more specific? $\endgroup$ Mar 27, 2023 at 5:13
  • $\begingroup$ Actually I will be extracting the dynamic covariances from the DCC model. I am trying to ascertain If, in the wake of the exogenous variable, there are changes in the covariances or to put simply is there any difference in the spillover in the presence of the exogenous variable. I also could not come across anywhere regarding adding an exogenous to a DCC Garch Model. $\endgroup$
    – Jyoti Nair
    Mar 27, 2023 at 8:40
  • $\begingroup$ DCC is not suitable for modelling spillovers, as the volatility of each series only depends on its own past via a univariate GARCH model. DCC only models contemporaneous relationships the between time series, while a spillover suggests time-lagged effects. Perhaps BEKK-GARCH would be a more relevant model for your purposes. $\endgroup$ Mar 27, 2023 at 9:29

1 Answer 1

0
$\begingroup$

I am not aware of a software implementation of the DCC model with an external regressor in the DCC equation. A simpler solution might be to obtain the fitted conditional correlations and covariances from a vanilla DCC model and then regress them on your external variable of interest. This is not nearly as elegant a solution, but much easier to implement.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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