The RATS Software Forum

Hi,
I want to estimate volatility spillover between spot and futures markets. I want to use VECM-MGARCH-BEKK model. I already estimated VECM model. After taking residuals from VECM model, can I directly enter the residuals into bivariate MGARCH BEKK estimation as variables? (I know there are codes combine VECM and MGARCH BEKK, but I want to do it in two steps). And if so, after estimating bivariate MGARCH BEKK model, does alpha coefficients (alpha (1,2) and (2,1)) mean one markets return shocks effect the other market volatility, and also beta coefficients ((1,2), (2,1)) mean one market volatility effect the other market volatility?
I need help, it is really urgent for me!
Thanks

ibrahim wrote:Hi,
I want to estimate volatility spillover between spot and futures markets. I want to use VECM-MGARCH-BEKK model. I already estimated VECM model. After taking residuals from VECM model, can I directly enter the residuals into bivariate MGARCH BEKK estimation as variables? (I know there are codes combine VECM and MGARCH BEKK, but I want to do it in two steps).

ibrahim wrote: And if so, after estimating bivariate MGARCH BEKK model, does alpha coefficients (alpha (1,2) and (2,1)) mean one markets return shocks effect the other market volatility, and also beta coefficients ((1,2), (2,1)) mean one market volatility effect the other market volatility?
I need help, it is really urgent for me!
Thanks

Yes, non-zeros in those locations imply the different types of spillover effects.

Hi Tom,
Thank you very much for quick reply. I appreciate it.
I understand that your "yes" reply covers both questions.
By the way what about interpreting vecm results. I see that many studies using cointegration-vecm analysis interpret vecm coefficients directly rather than ecm coefficient. By doing so, they also interpret lead-lag relations between variables. Does this methodology include any wrong doing?
Best,

Hi again,
I am not able to run basic var model in rats, although I examined the forum and also programming codes guide.
I am trying to run the below code

SYSTEM(MODEL=VARMODEL)
VARIABLES dresid01 dresid02
LAGS 1 to 12
DETERMINISTIC constant
END(SYSTEM)
estimate(noprint)

it gives the below error.
## VAR4. SYSTEM definition is incomplete or has empty model

This is so basic I know but I am trying to solve rats language. Sorry about that.

ibrahim wrote:Hi again,
I am not able to run basic var model in rats, although I examined the forum and also programming codes guide.
I am trying to run the below code

SYSTEM(MODEL=VARMODEL)
VARIABLES dresid01 dresid02
LAGS 1 to 12
DETERMINISTIC constant
END(SYSTEM)
estimate(noprint)

it gives the below error.
## VAR4. SYSTEM definition is incomplete or has empty model

This is so basic I know but I am trying to solve rats language. Sorry about that.

Are you sure that you're executing all those instructions? The only way I see that happening is if the VARIABLES line doesn't get executed.

ibrahim wrote:Hi Tom,
Thank you very much for quick reply. I appreciate it.
I understand that your "yes" reply covers both questions.
By the way what about interpreting vecm results. I see that many studies using cointegration-vecm analysis interpret vecm coefficients directly rather than ecm coefficient. By doing so, they also interpret lead-lag relations between variables. Does this methodology include any wrong doing?
Best,

I'm not sure I understand the difference between the VECM and ECM coefficients. Could you give an example?

I need help about above questions.
Any help is greatly appreciated.

Sorry, while I was writing for help, you also replied my quetions.
First,
I don't know why but the VAR code that I wrote is not running. (For all cases)
Second,
For VECM model, for instance a VECM model for two variables with two lags, there are two error correction term and also four lagged variables, right. If I interpret lagged variables directy for lead-lag relations, does it include any wrong doing? I am confused that in many study they are interpreted and listed in table. But I also know that for VAR model, it is especially used for variance decomposition or for impulse-response functions.
Tom, I need your help.
Thanks

ibrahim wrote:Sorry, while I was writing for help, you also replied my quetions.
First,
I don't know why but the VAR code that I wrote is not running. (For all cases)

You'd have to post the entire program (or e-mail to support@estima.com) so we could see what's wrong with the setup.

ibrahim wrote: Second,
For VECM model, for instance a VECM model for two variables with two lags, there are two error correction term and also four lagged variables, right. If I interpret lagged variables directy for lead-lag relations, does it include any wrong doing? I am confused that in many study they are interpreted and listed in table. But I also know that for VAR model, it is especially used for variance decomposition or for impulse-response functions.
Tom, I need your help.
Thanks

The coefficients on the lagged differences are almost impossible to interpret in isolation---the dynamics are dominated by the error correction terms and publishing the lagged difference coefficients is largely a waste of paper.

Thank you Tom,
By the way the VAR system error is solved.
I appreciate your help.
Ibrahim