How to interpret results for Impulse, Granger

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

smrehman
Posts: 4
Joined: Sun Jul 30, 2017 6:26 am

How to interpret results for Impulse, Granger

Postby smrehman » Tue Aug 29, 2017 3:44 am

Hi!

Sorry, I understand if the following questions are basic, but I have been trying to do research, and don't have anywhere else to ask. I hope you can help me!

My question is about Impulse Response is:
(1) how do I interpret whether my impulse response is (in)significant?
(2) And how do I know if its at a 1% or % level?
(3) And how do I see whether it's a strong relationship or weak relationship?
(4) And how many periods is it generally accepted for a short-run andmedium-run period?

In case of Granger:
(4) how do I interpret whether my Granger Causality result is (in)significant?



I have attached some examples on Impulse response functions and Granger causality results:
a) response of real stock return(RSR) on oil price shock (ROP)for Germany
b) response of (RSR) on (ROP)for Japan
c) response of (RSR) on (ROP)for UK
d) response of RSR on industrial production
e) Granger causality p-value for each country


Thank you in advance!
Attachments
result.png
result.png (66.93 KiB) Viewed 3863 times

cgoes
Posts: 12
Joined: Thu Aug 24, 2017 3:10 pm

Re: How to interpret results for Impulse, Granger

Postby cgoes » Tue Aug 29, 2017 6:39 pm

(1) how do I interpret whether my impulse response is (in)significant?


If zero is outside the red-dashed band, it is statistically significant for a give confidence level.

(2) And how do I know if its at a 1% or % level?


It needs to be clear in the paper. If I recall the standard in Eviews is to show two standard error confidence bands (meanings it's slightly below 0.05).

(3) And how do I see whether it's a strong relationship or weak relationship?


That depends on the economic interpretation of responses. You can scale responses to be of a specific size.

(4) And how many periods is it generally accepted for a short-run andmedium-run period?


Since we`re usually working with stationary variables when running VARs, we assume that over the long run shocks go to zero and variables revert to their means or trends. But whether they revert at a fast or slow pace will depend on how many periods it takes for the process to be completed and whether you`re working with monthly/quarterly/yearly data.

smrehman
Posts: 4
Joined: Sun Jul 30, 2017 6:26 am

Re: How to interpret results for Impulse, Granger

Postby smrehman » Wed Aug 30, 2017 5:11 am

cgoes wrote:
(1) how do I interpret whether my impulse response is (in)significant?


If zero is outside the red-dashed band, it is statistically significant for a give confidence level.

(2) And how do I know if its at a 1% or % level?


It needs to be clear in the paper. If I recall the standard in Eviews is to show two standard error confidence bands (meanings it's slightly below 0.05).

(3) And how do I see whether it's a strong relationship or weak relationship?


That depends on the economic interpretation of responses. You can scale responses to be of a specific size.

(4) And how many periods is it generally accepted for a short-run andmedium-run period?


Since we`re usually working with stationary variables when running VARs, we assume that over the long run shocks go to zero and variables revert to their means or trends. But whether they revert at a fast or slow pace will depend on how many periods it takes for the process to be completed and whether you`re working with monthly/quarterly/yearly data.



thank you for your response!

in regardings to (1), so if I understand it correctly, all the graphs a b c d are insignificant?

for (2) yes, it's 2 standard errors. So if all my results are significant, it means that they are all significant at a 5% level (for example, a and c, but not b and d?)
I thought that significance was read off the y-axis?

cgoes
Posts: 12
Joined: Thu Aug 24, 2017 3:10 pm

Re: How to interpret results for Impulse, Granger

Postby cgoes » Wed Aug 30, 2017 10:54 pm

Yes, only (d) is significant at periods 1-4.


Return to “Econometric Discussions”

Who is online

Users browsing this forum: Google [Bot] and 24 guests