i run johansen cointegration test, it reveals the presence of long run economic relationship between the variables and made error correction also.
Now, I want to test the short run causality between the variables, how to run it in eviews.
thanks in advance
In a VAR object you can follow the sequence: View--> Lag structure --> Granger causality/Block exogeneity tests
If cointegration exists then (Granger-) causality also exists in at least one direction (but not necessarily the direction presumed in your model.) The video suggested by Ahmadi shows (towards the end) how to test for non-causality in the direction presumed in the model. Briefly, in an "X->Y" model, the non-causality hypothesis asserts that all terms involving lagged X should have zero coefficients. After you have estimated your VECM, the Eviews vector C contains all the estimated coefficients; inspect this to see which are attached to dX-terms in the dY-equation and use the "coefficients diagnostics" view to (Wald) test the restriction that their values can all be set to zero. Depending on what you mean by "short run causality" you might or might not include the Einclude the ECM coefficient in those that you restrict to zero.
please see this.
http://www.eviews.com/help/helpintro.html#page/content/demo-Specification_and_Hypothesis_Tests.html
This is well covered in Lutkepohl (2005), New introduction to multiple time series, Springer. The relevant sections are sections 6 and 7. Granger non-causality is implied by certain zero restrictions on the coefficients of the error correction terms or zero restrictions on the coefficients of the first difference terms. These test statistics that correspond to the standard wald test for stationary variables do not have standard distributions, See page 316 for details. I have not used eviews for a long time and do not know if eviews deals with this problem. If you can not find a specific mention of the problem in your eviews manual it is unlikely that standard eviews can provide the correct critical levels for the non-standard tests
i will suggest to watch this video on you tube or eviews help which i am going to share with you .
https://www.youtube.com/watch?v=49-C5h4E1dg
http://www.eviews.com/help/helpintro.html#page/content/demo-Specification_and_Hypothesis_Tests.html
Muhammad,
Your youtube video illustrates some interesting points but as far as I can see contains an econometric error. The t-statistic on the error correction term does not have the usual t-distribution but has a non-standard distribution, His approach makes strong assumptions about exogeneity that may not be satisfied.
The appropriate way to test for a zero loading on the error correction term is to use the Johansen procedure and test the restriction on the alpha loading matrix. The appropriate method is set out in your link to the EVIEWS manual under | Advanced Multivariate Analysis | Johansen Cointegration Test | Imposing Restrictions |
The youtube video example talks about Granger Causality between the euro-dollar exchange rate and a one-month oil furure price. The price of a futures contract is determined by the spot price plus the carrying price plus commission. If you find any other price mechanism you will make a fortune through arbitrage. If oil is traded in dollars then the spot price in euro is related to the dollar price. I would always insist that any Granger Causality test be based on sound economic theory and include all relevant variables. Clearly this video does not meed either of these requirements.
On the basis of the results in the Youtube video, any disequilibrium has a half-life of the order of 2o periods.
- Badges
- Science topic
- Similar topics
- Engineering
- Petroleum Engineering
- Oil