Heteroskedasticity in VAR models

[ts highbury]

Hi,

Is there a way in Stata to test for heteroskedasticity in a VAR model, other than testing the residuals of each equation?

A follow up to this is is there are a way of correcting for heteroskedasticity, i.e., Newey-West standard errors for a VAR.

Again, I guess one could run separate regressions for each equation with Newey-West errors. Would be incorrect to do so?

However, I've noticed that the standard errors in VARs are slightly different than those in the separate equations (though the parameter estimates are the same). Does anyone know why this is?

Thanks

[Magdalena Go]

I would also be interested in this problem! Does anyone have a solution?

Thanks!

[Emad Shehata]

Check here
Post #2

http://www.statalist.org/forums/foru...roskedasticity

[Manh Hoang Ba]

Hi everyone,
I wrote a post-estimation command called varlmhet that does the job. You can install it with the command

ssc install varlmhet

. If that doesn't work, you can visit the following link: [STATA] Test for heteroskedasticity in time-series VAR and VEC model: varlmhet & veclmhet command – Hoàng Bá Mạnh – Kinh tế lượng

. webuse lutkepohl2, clear
(Quarterly SA West German macro data, Bil DM, from Lutkepohl 1993 Table E.1)

. tsset

Time variable: qtr, 1960q1 to 1982q4
        Delta: 1 quarter

. gen t=_n

. 
. *       VAR model
. qui var dln_inv dln_inc dln_consump, lags(1/2) exog(t)

. varlmhet

VAR Residual Heteroscedasticity Tests (Includes Cross Terms)
H0: Homoscedasticity
    Chi2(210) =   226.8771
    Prob > chi2 =      0.2019


. varlmhet, nocross

VAR Residual Heteroscedasticity Tests (Excludes Cross Terms)
H0: Homoscedasticity
    Chi2( 84) =    81.2221
    Prob > chi2 =      0.5656


. 
. *       VEC model
. qui vec ln_inv ln_inc ln_consump, lags(2)

. veclmhet
VEC Residual Heteroscedasticity Tests (Includes Cross Terms)
H0: Homoscedasticity
    Chi2(  84) =    82.1650
    Prob > chi2 =      0.5363


. veclmhet, nocross
VEC Residual Heteroscedasticity Tests (Excludes Cross Terms)
H0: Homoscedasticity
    Chi2(  48) =    52.1179
    Prob > chi2 =      0.3169