Determining cointegrating vectors with Johansen

[kristj20]

After running a cointegration test I get the results of the trace and eigen test. I see there are two cointegrating vectors, but how can one tell which are these two vectors just by looking at the output table? The help guide in Eviews doesn't seem to have any section on how to interpret results.

[doda]

same question

[trubador]

If you scroll down the output of Johansen Cointegration Test, you'll see the cointegration equations at the bottom. Another way is to estimate the VEC and change the number of cointegrating vectors (i.e. rank) from within the VAR Specification window under the Cointegration tab. Estimation output will also display the cointegration equations.

[kristj20]

If you scroll down the output of Johansen Cointegration Test, you'll see the cointegration equations at the bottom. Another way is to estimate the VEC and change the number of cointegrating vectors (i.e. rank) from within the VAR Specification window under the Cointegration tab. Estimation output will also display the cointegration equations.

I estimated the VEC and got one CE. In the eviews help guide it says that when estimating VEC, the number of CEs should be one lower than the number of explanatory variables? Why is this requirement so?

And in the case where one gets 2 CEs from Johansen, but applies VEC to only 1 CE (given the eviews guide rule about one less CE than total explanatory variables) , how does Eviews decide to which CE to apply the VEC estimation?

[alfernz]

In the eviews help guide it says that when estimating VEC, the number of CEs should be one lower than the number of explanatory variables? Why is this requirement so?

Because the EC matrix is not of reduced rank and thus a levels VAR is appropriate. The following is the usual rule of thumb:
rank(pi)=rank(X) <-- level VAR
0<rank(pi)<rank(X) <-- CVAR
rank(pi)=0 <-- VAR in first differences

More here: http://www.econ.ku.dk/okokj/papers/kjdhengii.pdf

HTH

Alan