The RATS Software Forum

Hello,
I am estimating a GARCH-BEKK model in RATS and one of the variables is treated as exogenous. In the RATS software, there is no option for exogenous variables when estimating the BEKK model since all variables in both the mean and the variance equations are dependent variables. Therefore, I would like to know if there is a way of including an exogenous variables in the GARCH-BEKK model in RATS for both the mean and the variance equations.
Thank you

I'm not sure what you're asking. Don't the REGRESSORS and XREGRESSORS options do what you want?

Hi Tom,
Many thanks for your reply. I am actually estimating a GARCH-BEKK model with two variables as endogenous and other two variables as exogenous. The endogenous variables in the model are DLGSECI and DLEXR whilst the exogenous variables are DLCOP and DLSP500. The aim is to estimate the spillover effects of the exogenous variables on the endogenous variables whilst restricting the effects of the endogenous variables on the exogenous variables. In estimating the BEKK, I put the endogenous variables in the dependent variables box. I also added the endogenous variables in the mean model variables box, whilst the exogenous variables are put in the variance shift variables box. The results of the model are presented below. As you can see, the estimated coefficients only show shock and volatility spillivers between the dependent variables. There are no estimates of volatility spillovers from the exogenous variables to the dependent variables. The exogenous variables and their coefficients estimates are shown at the bottom of the output results but I am not sure what relationships they are measuring. The model used regressors and xregressors as you adviced but I am not able to obtain the shock and volatility spillovers from the exogenous variables to the dependent variable. I would be glad if you could give me some more clues as to how to make these estimations.

GARCH(P=1,Q=1,MV=BEKK,ASYMMETRIC,XREGRESSORS,REGRESSORS,ITERS=300,PMETHOD=BFGS,PITERS=15) / DLEXR DLGSECI
# Constant DLEXR{1} DLGSECI{1}
# DLCOP DLSP500

MV-GARCH, BEKK - Estimation by BFGS
Convergence in 76 Iterations. Final criterion was 0.0000043 <= 0.0000100
Monthly Data From 1991:02 To 2015:12
Usable Observations 299
Log Likelihood 1313.2556

Variable Coeff Std Error T-Stat Signif
************************************************************************************
1. Constant 0.000776730 0.000419023 1.85367 0.06378619
2. DLEXR{1} 0.641655857 0.049218872 13.03679 0.00000000
3. DLGSECI{1} 0.002261733 0.006751566 0.33499 0.73762978
4. Constant 0.010508865 0.002384376 4.40739 0.00001046
5. DLEXR{1} 0.007080197 0.068625859 0.10317 0.91782727
6. DLGSECI{1} 0.525880001 0.047041218 11.17913 0.00000000
7. C(1,1) -0.000557549 0.000485294 -1.14889 0.25060185
8. C(2,1) 0.018187766 0.004929394 3.68966 0.00022456
9. C(2,2) 0.018938241 0.005104316 3.71024 0.00020706
10. A(1,1) 0.675805652 0.057590707 11.73463 0.00000000
11. A(1,2) 0.226891894 0.122642931 1.85002 0.06431065
12. A(2,1) -0.006607671 0.009875216 -0.66912 0.50342110
13. A(2,2) 0.814235431 0.086100651 9.45679 0.00000000
14. B(1,1) 0.842405496 0.018575218 45.35104 0.00000000
15. B(1,2) -0.002430261 0.119632371 -0.02031 0.98379256
16. B(2,1) -0.003657030 0.010841380 -0.33732 0.73587463
17. B(2,2) 0.020798660 0.062276057 0.33398 0.73839825
18. D(1,1) -0.040325202 0.161322234 -0.24997 0.80261303
19. D(1,2) -2.547173517 0.614809687 -4.14303 0.00003428
20. D(2,1) 0.001830267 0.015093126 0.12126 0.90348119
21. D(2,2) 0.767383584 0.215568451 3.55981 0.00037112
22. DLCOP 0.004437167 0.004728146 0.93846 0.34800904
23. DLCOP -0.109952548 0.038178470 -2.87996 0.00397723
24. DLCOP 0.079625425 0.047415180 1.67932 0.09308902
25. DLSP500 0.085711369 0.012889944 6.64948 0.00000000
26. DLSP500 -0.016920904 0.098827015 -0.17122 0.86405283
27. DLSP500 0.345576594 0.111872491 3.08902 0.00200817

My suggestion is that you figure out what the intended model actually looks like (what the mean models are and what the variance model is), as mathematical formulas.