Chan & Maheu, JBES 2002 (Jump GARCH model)

[prashantj]

Here it is. Thanks.
I would like to implement similar to this.
Kuttu, S.. (2017). Time-varying conditional discrete jumps in emerging African equity markets. Global Finance Journal, 32, 35–54. https://doi.org/10.1016/j.gfj.2016.06.004

[prashantj]

Here it is. Thanks.
I would like to implement similar to this.
Kuttu, S.. (2017). Time-varying conditional discrete jumps in emerging African equity markets. Global Finance Journal, 32, 35–54. https://doi.org/10.1016/j.gfj.2016.06.004

I am posting it to looking for the code to avoid messing up with the question related to Chan & Maheu.

[prashantj]

Hi Tom,

I was trying to apply Chan & Maheu model for my data. Interesting things happen. My ignorance! Using the following command, when I estimated I got the message that. Of course, I checked for missing values and found none.
maximize(parmset=meanparms+garchparms+jumpparms+poissonparms+pegomega,$
pmethod=simplex,piters=2,method=bfgs,$
title="Constant Intensity Jump GARCH, 610-866") logl 610 866

Message:
## NL6. NONLIN Parameter DELTA Has Not Been Initialized. Trying 0

Constant Intensity Jump GARCH, 610-866 - Estimation by BFGS
Convergence in 52 Iterations. Final criterion was 0.0000000 <= 0.0000100

Usable Observations 251
Skipped/Missing (from 257) 6
Function Value -429.4217

Then, I used Trace inside Maximize and I did not get the ##NL6 error message. I am attaching the output. Is there anything wrong with it? which one is correct?

Thanks for your help in advance,

Prashant

[prashantj]

Tom,

I tried to apply arji-garch to my dataset and I ended up with the following error in the output. I tried a lot but of no avail. I cannot figure out what is the source of the error. Could you please help? highly appreciated. I was trying to generate * Constant intensity jump model through 1984 (column 1 in table 3)"

## NL6. NONLIN Parameter ZETA0 Has Not Been Initialized. Trying 0
## SR10. Missing Values And/Or SMPL Options Leave No Usable Data Points

I am attaching the data file, program and output.

sp500.xlsx

(29.19 KiB) Downloaded 784 times

arjigarch.RPF

(6.55 KiB) Downloaded 735 times

arjigarch-output.RPF

(2.16 KiB) Downloaded 754 times

Thanks,
Prashant

[TomDoan]

1. The program doesn't use SP500.XLSX, but spbgci.xlsx. It might be intended to be the same file.
2. As you have that written, you can't use dates (which you are using later in the program). Given the data file, you should use mapped dates.

OPEN DATA "sp500.xlsx"
DATA(FORMAT=XLSX,ORG=COLUMNS,JULIAN=JULIAN) 1 866 SP500
CALENDAR(JULIAN=JULIAN)

3. Given the behavior of your data, the calculation of SSVAR doesn't work---the base GARCH model has a+b>1 so there is no steady state variance. SSVAR is only used for guess values, so an alternative value can be used. (Such as the variance of the GARCH residuals---just do a STATISTICS on the residuals to get that).

4. Your JUMPGARCH estimation starts at entry 2. But you using a two lag AR mean model (on returns) so you are losing 3 data points to lags.