EViews.com

Here you go - hope you can tell me what the problem is..

What are your cross-sections?

Greenf is my dependent variable, Localiz & Urbaniz are my two independent variables.
The other ones are control variables.
For Eviews, the problem seems to be that it does not recognize my year-listings in the right manner.

No, what are your cross-sections. You say you have panel data. Therefore you have a time variable (year), and a cross-section variable. I want to know what your cross-section variable is.

These are my regions

The way a panel is defined is that each observation is uniquely identified by a time period and a cross-section. There is exactly one observation (row) for each combination of year and cross-section.

Your data doesn't have that. There are 30 observations that have the year "2003" and the cross-section "BE10 - Brussel". Thus you cannot structure your workfile as a panel.

Ok, thank you for your answer Gareth.
Do you have any idea which regression would be better?
Maybe regular OLS instead of pooled OLS, or a negative binomial (zero inflated?), or a conditional logit model?
Thanks again.

Hi Gareth,

My data in excel is like the first excel file you posted, but it is unbalanced. There time dimension is 31 years for each country but some years are missing for most countries. In this case, how do I enter my unbalanced panel into excel?

Would really appreciate if you could help me out as I am currently stuck with my dissertation.

regards,
Fida

EViews should read the unbalanced data from Excel in the same way it would read the balanced data. Just have rows in Excel for the observations that do exist.

Thank you so much for your reply Gareth. So for instance it should be done like this:

Country Year Variable 1 Vairable 2
Armenia 1989 xxxxx xxxx
Armenia 1992 xxxxx xxxx
Armenia 1993 xxxxx xxxx

As opposed to like this:

Country Year Variable1 Variable 2
Armenia 1989 xxxx xxxxx
Aremenia 1990 empty empty
Armenia 1991 empty empty
Armenia 1992 xxxx xxxxx
Armenia 1993 xxxx xxxxx


Am I correct in assuming that the first way is right?

Regards,
Fida

Yep.

Hi Gareth,


Thank you for your help, I was able to figure it out.

One more question:

The panel regressions I am currently running are "The regressions are estimated with fixed effects, using a GLS estimator", as cited by the paper. Would this mean I select fixed effects for both period and cross section?

And for the with GLS estimator bit, do I still use the Least Squares method? If so, what sort of GLS weights should i give: none, cross section, cross section SUR, period or period SUR

Regards,
Fida

Can't help with that.

Thats alright. Could you tell me if there are any robustness tests for (unbalanced) panel data models?

Hi Gareth,

My data is now in the format (as above) and has been recognised by Eviews as panel data. If I now want to lag a certain variable by say a year, would Eviews still recognise the unbalanced panel?

So for eg my data is in the format:
Year Country Spread Crisis Dummy Credit growth Dummy
1981 Albania 3 1 1
1984 Albania 2 1 0
1985 Albania 4 1 0
1986 Albania 2 0 1
1990 Albania 7 1 0

And if I want to lag the credit growth dummy by say two years, would it recongnise that there will be no value for 1984? Or would it simply transpose the 1981 value to 1985?

If lagging the variables works in this unbalanced form, could you also tell me the formula/way to lag this?

Thank you so much for your help,
Regards,
Fida