ICSS test on standardized GARCH residuals for parameter stability

[jimm306]

Hi,

I have daily stock returns and I am considering applying the ICSS test to standardized residuals from a fitted GARCH model. Would this be methodologically appropriate if the goal is to detect variance breaks and then examine whether the model parameters change across the regimes defined by those breaks?

Thanks

[TomDoan]

I'm not sure that that will provide any better information than you could get with just the standard @flux test. In particular, it would be unlikely to help with any specific "parameter stability" since it really wouldn't be testing for that, but just for a general process break.

The program below simulates a UV GARCH with a break in the variance intercept part way through. When I try that with different values for the shift, I never have found a case where the ICSS procedure finds a break when @flux doesn't.

Code: Select all

*
* UVGARCHSIM.RPF
* Simulation of a univariate (conditionally Normal) GARCH model
*
* GARCH parameters
*
compute cg=0.5
compute ag=.10
compute bg=.80
*
compute nobs=1000
*
all nobs
*
* Initialize to the ergodic variance
*
set uu = cg/(1-ag-bg)
set h  = uu
set u  = 0.0
*
* Simulate a GARCH process with a change to the variance intercept
*
set uu 2 nobs = h=cg+.5*(t>=500)+bg*h{1}+ag*uu{1},u=%ran(sqrt(h)),uu=u^2
set y 2 nobs  = u
*
graph(footer="Simulation of GARCH(1,1) model")
# u
*
garch(stdresids=ss,derives=dd) / y
@flux
# dd
set stdu = ss(1)
@icss(robust) stdu