Overflow error in GARCH-M

[kco5min]

Hi All,

I am runnig a AR - GARCH-M model of inflation and inflation uncertainty proxied by the errors of conditional mean ecuation.

The code is:


INFLATIE = C(1)*GARCH + C(2) + C(3)*INFLATIE(-1) + C(4)*DUMMY*INFLATIE(-1) + C(5)*INFLATIE(-2) + C(6)*INFLATIE(-4) + C(7)*INFLATIE(-6)

GARCH = C(8) + C(9)*RESID(-1)^2 + C(10)*GARCH(-1) + C(11)*INFLATIE(-1)

(inflatie=inflation)

The model runs perfect, but when i'm trying to conduct tests for errors anf coeffcients , Eviews says " Overflow". Why is this happening?

Can you help me please.

Thx in advance

[EViews Gareth]

Could you post the workfile?

[kco5min]

Thx Gareth for the fast reply.

Attached you will find my workfile. The Garch model is named grach-m.

Thx,

[trubador]

It seems your model is ill-defined. The problem arises from the fact that the variance specification does not fulfill the stationarity condition (i.e. c(9)+c(10)<1). I can offer two solutions:
i) I do not think it is a good idea to include the lagged dependent variable as an exogenous variable in the variance equation, so it might be a better idea to drop it.
ii) If you want to keep the exogenous variable in the variance equation, try including the standard deviation instead of variance as the garch-in-mean specification.

[kco5min]

Thx @trubador.

I will keep the lagged dependent variable of inflation because I am intending to test both influences bettwen inflation and its variance, in both directions.

I've changed the specifiaction to std dev and it works perfect.

Thx a lot :)

[kco5min]

Since i've used std dev instead of variance in the mean equation Eviews showed me all the test that i needed, but still I have the stationarity problem and I really dont know how to solve it.

The coefficient restrictions c(9) + c(10) = 1 showes a p-value of ,8990 even after I've included std dev in the conditional mean.

How can I repair this stationarity problem?

Thx in advance