Add observation, regular dated series

[stoddj]

OK, an even more simple-minded question than my last one, still unanswered (but it's a 'long' weekend in the US). So forget the panel data for now. I don't seem to even be able to get the benefit of added observations in a VEC model, using a regular dated series model. I want to go back 3 lags on first-differenced GDP, and see the effect on a Turnover variable I am treating as dependent within the VAR. This Turnover variable starts in 1948, but I can easily get GDP data back to 1944.

When I restructure the workfile and successfully update the GDP series, however, it makes no difference in the execution of my VEC. It still runs starting in 1952 , and not (as I had hoped) from 1948, to as to use all the observations in my Turnover series.

I must be messing up something pretty basic, but can't figure it out from the User's guides.

Thanks for any help, JS

[EViews Gareth]

Neither of your questions are particularly clear.

However, I'll see if I can take a stab at this one. When estimating a VAR, you need to have observations available for all series. Having more observations in a single series doesn't help at all if those observations are not available for all of the other series in the VAR.

[stoddj]

Ah -- I just saw it -- a 'duh' moment! It's the essentially symmetrical treatment of variables provided by the VAR (and VEC) procedures -- with a similar number of lags for each. How embarrassing!

This means I have to set up a VEC with different numbers of lags, in a stepwise linear fashion, if i want to get the benefit of the additional d.o.f. on the 'long' variable. Sorry if my initial questions were unclear, and thanks for your clarifying comment.