Is ARCH test mandatory for VAR? If so, what lag length of the ARCH test should I use? The same as the lag length of my VAR or VEC model?
$\begingroup$
$\endgroup$
2
asked May 25, 2016 at 13:51
-
$\begingroup$ Lars later asked So, if neglected ARCH effects will result in INEFFICIENT, but at the same time CONSISTENT, estimates, can I still use these estimates? Tow which I answer: Yes, you can use them. $\endgroup$– Richard HardyCommented Jul 9, 2016 at 12:15
-
$\begingroup$ I was going through my old answers and noticed this one was not accepted. Do you perhaps need further clarification? $\endgroup$– Richard HardyCommented Feb 15, 2017 at 10:32
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Is ARCH test mandatory for VAR?
"Mandatory" is a strong word when it comes to statistical practice. I would say, ARCH test is advisable, as is any relevant diagnostic test.
- Extra diagnostic testing will not hurt. In addition, you may learn useful information about the suitability of your model.
- Neglected ARCH effects will result in inefficiency of OLS or GLS estimates for VAR. But OLS or GLS estimates will still be consistent regardless of presence or absence of ARCH effects.
[W]hat lag length of the ARCH test should I use?
Choice of lag length can be motivated by subject-matter knowledge. You should check all lags where you could plausibly expect ARCH effects. You do not need to check too long lags where ARCH effects could only be due to chance. Since ARCH-LM test is of portmanteau nature (it tests a group of lags at once), you could try a few different lag lenghts to better elicit possible ARCH effects. Diagnostic plots of autocorrelations of squared residuals could also be useful.
answered May 25, 2016 at 13:58