I am investigating the long run/short run effects of some dependent variables on exchange rate using the Johansen cointegration test and vector error correction model. I can get the short run relationships from the VECM estimations. What about the long run relationship from the cointegration test?
I chose 2 lags and have 3 cointegrating equations. The normalized cointegrating equation I have to look at is I guess the one which says "3 cointegrating equations" but 3 variables are normalized (with a value 1.000) including the dependent variable. Does this mean that the coefficients of the 2 independent variables is 0 when I write the long run cointegrating equation?
What are adjustment coefficients?
dear Josef Gemains
please see this Video
LINK:
http://www.uplooder.net/cgi-bin/dl.cgi?key=ed6c7f647efb7fe00b79f6f6e6ca444d
Dear Salah,
Thank you very much for your reply. I have seen the video and most of the guy's uploads on youtube but his examples involve few dependent variables and mostly 1 cointegrating equation. In such cases, it is relatively easy to read the tables and interpret the results. My problem is on how to read the tables, especially to write the long run equilibrium equation when there are more than 1 cointegrating equations. Do you have any idea or any other materials on this?
I appreciate your help!
Dear Josef
You do not need to report the results of all the cointegrating equations. You should only take the first equation where only dependent variable is normalized. The number of cointegrating equations only tell you that there is long run relationship.
The error correcting coefficients are your adjustment coefficients which are present below the normalized equation (the ones in differenced form).
The finest method of testing for unit roots is by using the method maximum likelihood estimator of Johansen (1988 and 1991) is a test for cointegration restrictions in a VAR representation. “Johansen” estimation is preserved in much detail in the book by Johansen (1995). This estimator also gives you asymptotically efficient estimates of the cointegrating vectors and of the adjustment parameters. “Johansen’s method” is the maximum likelihood estimator of the so-called reduced rank model. This is the core idea of the cointegration test yet it explores different issues of time series data analysis including AR (k).
The finest method of testing for unit roots is by using the method maximum likelihood estimator of Johansen (1988 and 1991) is a test for cointegration restrictions in a VAR representation. “Johansen” estimation is preserved in much detail in the book by Johansen (1995). This estimator also gives you asymptotically efficient estimates of the cointegrating vectors and of the adjustment parameters. “Johansen’s method” is the maximum likelihood estimator of the so-called reduced rank model. This is the core idea of the cointegration test yet it explores different issues of time series data analysis including AR (k).
You can test for the long run relationships by looking to the indidividual significativity of the coefficients of the error correction terms given when estimating the VECM.
Nabya Tehreem, thanks. The adjustment coefficients are all negative but some of them are not significant. Is this a problem? I read somewhere that the error correction term must be negative and significant. Is this only for the error correction term in the short-run model?
Dear Josef,
When the coefficient of the ECT is negative and significant, we say that there is a long run relationship but not short run. If the coefficient of ECT is negative but not significant, this implies the absence of long run causal relationship.
Short run causality is cheked by testing the significativity of the coefficients of the independent variables.
Dear Prof. Mounir,
Thank you! If I have 2 cointegrating equations, I have 2 error correction terms I guess. How can I spot them? Sorry for my trivial questions
Dear Josef,
You should use only one or you use the two but separatly.
When you have two co integrating equations, use the one which is normalized just for the dependent variable i.e the dependent variable's coefficient is one.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis