Restricting the constant in a VECM

[ashtongate]

Anyone know how to do this? Tried a few tricks, but will not work.

[hum2004]

hi, i understand you want to test wheather the constant in the cointegration relation can be restricted. So you are looking wheather cointegration case tow or case one or case 3 or case one is the best description of the DPG. (the "cases" refer to the option available in VECM estimating in EVIEWS.

I do the following if i want to test case 2 against case 1:
estimate case two, take the loglikelyhood and call it LogL-UR
estimate then case one, take the loglikelyhood and call it LogL-R

The test statistic is=-2*(Logl-R-LogL-UR) and follows a chi-squared distribution (degrees of freedom = 1)

Compare the test statistic with the critical value.

to test wheather case two or case three is more appropriate i do the same as above, but take the loglikelyhood of the VECM estimated with case 3 and compare it with the loglikelyhood of the VECM estimated with case 2.

Applied econometric time series, Walter Enders page 400 (old eddition gives an example where you can varify the procedure)