Dear all,
I have run into some articles that perform a VAR model after performing a GARCH(1,1) model. However, they do not explain how they use the output of the GARCH model to estimate the VAR. In fact, they only say that:
"in order to estimate the VAR model, we have transformed the results obtained form the GARCH(1,1), creating a variance equation for each GARCH(1,1) model with the objective of obtaining the standard deviation of each series, thus allowing the creation of the VAR models"
Does anyone knows how to do this with a better explanation?
I appreciate all the help in advance.
Yours sincerely,
Afonso Rodrigues
Dear Pedro Afonso Leitão Ventura de Sousa Rodrigues
It is an interesting case. Would it be possible to share the paper details so that others can have a look?
Dear Akram Shavkatovich Hasanov
Unfortunately I am afraid that I do not have permission to share the papers I have mentioned. However, they are written in Portuguese so I guess it wouldn't be very helpful anyway.
Nonetheless, regarding my question, I have been told that I should perform the GARCH model in Eviews making sure it is well estimated (with the diagnostic tests). Then I should select proc - make GARCH varience series. Then I should perform the VAR model as usual, adding the GARCH varience series variable in the VAR estimation.
It is so confusing. The procedure should be the other way round. First you estimate a VAR model. You then apply multivariate GARCH (MGARCH) model to the vector of residuals obtained from the VAR model. The purposes of using VAR and GARCH are completely different: the former is used to estimate (or forecast) the conditional 'mean', while the latter is used to model the conditional 'variance'.
- Badges
- Science topic