Unbalanced panel data structure.

[mjanota]

Dear all, im currently working on a unbalanced dataset to work with eviews for some simple regressions.
I currently work with 94 cross sections (countries), observed over 11 periods (years) of 5 different measurements/variables.

The problem I encounter is that once I include certain explanatory variables, the regression result window indicates that it is a balanced panel. If I leave either one of these two variables out, the panel regression result remains unbalanced. This is most likely because both of the two explanatory variables are not measured over the 11 year period, but are just measured for the first year and are included to reflect initial conditions for each country.

My question is, have I designed the datafile in a wrong way? How can I adjust these two variables to be included without having the entire panel to become balanced? I attached a picture of the different outputs when including/excluding the "problematic" variables and the series view of them compared to other variables.

Best regards
Marcus

[EViews Gareth]

It isn't 100% clear what you're trying to achieve with the "initial conditions", but it is probably the case that you need to ensure that each cross-section has a value for every time period, not just the first one.

[mjanota]

Thank you for the reply. Let me try to clarify.

I'm using "initial" levels of education and income to control for the stage of development each country has. I used them in my simple cross-sectional analysis where each country simply had 1 observation per variable, represented by its 11 period mean value. Now that I'm trying to move to panel data to exploit the time series dimension. But I'm quiet sure I'm doing something wrong either in the equation line or how I structured the data file. I plan to make use of both 2sls and GMM, the question is, can I specify in the estimation to just make use of the 1st period value, or is it possible to have the value constant over all 11 periods? These two "troublesome" variables are just part of a informational set.

[EViews Gareth]

As I said before, you need to ensure that each cross-section has a value for every time period, or, as you put it, the value is constant over all 11 periods.

Perhaps an easy way to do this is with the following code:

Code: Select all

x=@recode(x=na,x(-1),x)

where you replace "x" with the name of your series.