Johansen cointegration test

[lappenlappen]

Dear all,

I've searched the forum and I haven't been able to find the answer as to how I can estimate (and automate) a VECM once I have found cointegration through the Johansen procedure. I'm currently using Eviews 6 and the code I have so far is:

group importprices
importprices.drop resid
importprices.drop time
importprices.drop series01
table cointegration

for !i=1 to importprices.@count
%name = importprices.@seriesname(!i)
group g1
g1.add {%name} logforeignproxy loggdp logneu
g1.coint(s,4)
freeze(table2) g1.coint(s,4)
cointegration(1,1)="name"
cointegration(1,!i+1)=%name
cointegration(2,1)="trace intercept no trend"
cointegration(2,!i+1)=table2(12,4)
cointegration(3,1)="max-eig intercept no trend"
cointegration(3,!i+1)=table2(13,4)
d g1
d table2
next

So I get the output for both the Trace-statistic and maximum eigenvalue. In half the cases the Johansen procedure shows no cointegration and for less than 10% is the cointegrating rank greater or equal to 2.

Suppose the cointegrating rank is one. Is the Engle-Granger framework a valid approach in this case, assuming weak exogeneity of the regressors? I know this isn't ideal, but I have over 300 regressions and I don't know how to programme a VECM, whereas the Engle-Granger framework is quite straightforward. Any help is much appreciated.

Many thanks,
Seb

[lappenlappen]

Ok, scrap my last question. After some work, I realised it isn't actually that difficult to get the VECM up and running. However, how can I include contemporaneous first differences in the variables? Eviews only seems to give the option to include lags of first differences. But, theoretically, shouldn't you be able to have the contemporaneous first differences in there alongside the cointegrating equation and the lags? Am I missing something here?

Thanks,

Seb