VEC stability condition check

[nectar]

Roots of Characteristic Polynomial
Endogenous variables: LREER11 LPROD RDEBT TBG M1D OPEN LREM
Exogenous variables: C
Lag specification: 1 1
Date: 09/09/10 Time: 12:39

Root Modulus

1.162461 1.162461
-1.060304 1.060304
1.000000 1.000000
1.000000 1.000000
1.000000 1.000000
1.000000 1.000000
1.000000 1.000000
1.000000 1.000000
-0.486868 - 0.074955i 0.492604
-0.486868 + 0.074955i 0.492604
0.208337 0.208337
-0.122074 - 0.132795i 0.180379
-0.122074 + 0.132795i 0.180379
-0.020994 0.020994

VEC specification imposes 6 unit root(s).

=======================================================

Above are the characteristics root from the VECM estimation. There are 7 variables and 33 annual observations. The model has one cointegrating vector and therefore VEC specification shows there are 6 unit roots. Moreover, the model is specified with one lag in difference (which is two lags in level). I have a couple of question on reading the table correctly.

First, are the roots in the table given in the order as it appears in the group window? And, what about the roots of the lags, ie, do the first and second roots represent the two lags of the first variable and subsequently for the other variables?

Second, there are two roots outside the unit circle (on top of 6 unit roots imposed). This means the VECM is not stable, correct? But I wonder why the variables are listed in level when the VECM is estimated in difference? I presume this is so because the error correction part (cointegrated relation) is given in levels. What can done to ensure that the VECM meets stability condition?

Thank you in advance for your help.

anjan

[nectar]

Any help, plz!

How are the roots ordered for variable and its lag in the AR roots table?

Is the model unstable if there are roots greater than one (beyond k-r unit roots)?

Urgent, kind help, plz!

[mirza kaleemullah]

hi necter

i hope that u have addressed this issue. I am indulged in the same thing that VECM is not stable and every time in estimation same vecm, it gives new results.
anybody there plz?
regards

[aleconomista]

Hi everybody

I will give you one example for one long-run relationship (one cointegrating vector): just recover the error term derived from the long run equation, replicate the same VECM results meaning that you have to include the lagged residual as an exogenous variable: If there is stability condition, then the model is right.

I hope it will answer your question!!!