Eviews VECM with a cointegrating rank of 2

Question

Using Eviews, I have performed a Johansen cointegration test on three variables, suggesting 2 cointegrating relationships. I am now trying to estimate a VECM model, but when following Brooks (2014) pg. 410, the author finds 3 cointegrating relationships, yet only declares 1 in the VECM menu. Would this not create an issue?

Further when estimating a VECM with 2 cointegrating equations declared I am returned an output similar to this

                 coint1             coint2
y1.l2  1.0000000000000000 0.0000000000000000
y2.l2  0.0000000000000000 1.0000000000000000
y3.l2 -0.7328533912672071 0.2951961665740829

whereas with 1 I recieve an output similar to this

          coint1            
y1.l2  1.000000000000000  
y2.l2 -4.732435772888009  
y3.l2 -2.129850289981805

The output recieved from inputing just 1 equation seems much more reasonable, but would I not be suffering from omitted variable bias?

Could you comment on your title also in the body? Do you find rank 2 in your analysis? This is not the same as 2 cointegrating relationships. — Richard Hardy, Mar 27 '20 at 18:45
Johansen's states 2 cointegrating equations, so yeah a rank of 2 — David Witts, Mar 28 '20 at 11:40

Eviews VECM with a cointegrating rank of 2

0 Answers0