VECM Johansen test difficulties

[kris747]

Hello,

I found a peculiar problem in doing Johansen cointegration test in VECM. I tried to replicate the model of three variables based on the following paper:
http://economics.soc.uoc.gr/macro/8conf ... krusec.PDF

VECM includes the sample period from 1961:1 to 2001:4 of Canadian economy. The variables are real, log values of GDP, government spending on goods and services, and net taxes (sum of income tax, Contributions to social insurance plans,Taxes on production and imports, and Other current transfers from persons minus current transfers to persons and business). The same as in the paper. However, when I run Johansen test the results suggest only one cointegration instead of two.

I have two questions:
firstly, what I did wrong or what I missed in my specification?
secondly, how to distinguish from the data deterministic specification of test? When I use subsample 1961:1-1991:4, VECM with 3 lags, and 3rd specification of test I get full pi matrix, while if the test is specified with 4th option I get two cointegrated relation. How I know, which specification is correct?

I use EViews 8. I attach the dataset I created based on information from CANSIM database.
Thank you in advance for any suggestion:)