Nobody seems to have answered your question. It is a long time I have used EViews and I don't have it on my computer but I could look in the users' manual. A few remarks:

• the coefficients C(6) to C(8) are not 0; they are just very small;
• you can multiply the dummies by a very small constant to make them appear bigger; this is because your variance is big;
• your model is not pure GARCH since there is RETRN(-1);
• did you specify starting values for the parameters, e.g. 0.6 for C(5)?
• can you provide more details on the command you entered;
• I would recommend you to work step by step, and check everything at each step: first GARCH, then one break dummy, etc.

Don't forget that with nonlinear models there is no guarantee that estimation works well if the starting values are not adequate.

@ Guy Mélard Thank you so much sir for coming forward to help me. Here are explanation that you have asked for.

• Sir as you mentioned the Coefficients C(6) and C(8) being 0 means they are very small but how should I interpret them with p value 1.
• my model is AR(1)- GARCH (1,1) model as suggested by related paper. But I am curious about does it make any difference while incorporating the dummy variable.
• Sorry sir, I didn't get what you mean by specifying start values for parameters.
• For the commands, first of all I have adopted ICSS methodology for identifying the break points in volatility. Thereafter, I've assigned 1 for the starting of break date until next break date comes such as for break point Mar 2007 , Mar 2008, April 2009 , i took 1 from the mar 2007 to mar 2008 and 0 else where. After that i have adopted the procedure for GARCH model as described in papers and put the 3 dummy variables as variance regressors in GARCH model.
• I would really appreciate your help if you can guide me through out the procedure. Further, if you recommend me any reading for better understanding of break dummy variable.

Note: Mentioned break date are just for example.

I am looing forward to your guidance

Dear Indu Bansal

1. You are right that the p-values indicates that the parameters are not significantly different from zero. But, EViews has made 0 iterations, so the result is wrong.

2. I am not affiliated to IHS Global so I don't know what their program does.

3. The starting values are specified by a param instructions. In the manual (I have manual II for version 8), I see the example

param c(1) 153 c(2) .68 c(3) .15

sets C(1)=153, C(2)=.68, and C(3)=.15.

I suppose you specified C(5) to 0.6 since optimization didn't work. Or perhaps EViews found that value using an automatic procedure.

4. I don't know about neither the ICSS methodology nor the "related paper" that you mentioned.

5. Please follow my suggestion, first fitting your AR(1)-GARCH(1,1) model without the dummy variables, then adding them one by one. At each step you can add a param instruction to specify the initial values.

I have the Francq-Zakoïan book in my office room but I went there yesterday and I am not expected to go back before one week. I don't know if they speak about that kind of model.

Best regards

Guy