Panel data

[winger]

Hello,

I have some question about panel data in Eviews 7.1. I have a dataset with 33 cross sections (dependend variable) and 7 ( yearly) explanatory variables over a time period of 17 years (1996-2011). So a 528 obs balanced panel.

First question: How can eviews show the different coefficients of the explanatory variables over time and cross section?

Second question: How can i model an dummy period-variable in eviews, so that my range is split in two, and i can compare the coefficients in the first range and the second range?

Can you help me please?

Thanks in advance,

Winger.

[EViews Gareth]

I presume by the first question you mean fixed effects, which are available from the View menu of a panel equation.

You can use a dummy variable in a panel equation in the same way as a normal equation.

[winger]

Thank you for your quick responds,

I allready explored this fucntion, but with the view->fixed-ramdom effects ->cross section effects menu, i only see one column with effects on the cross-section variables. My question is, how can i see the (hopefully) changing coefficients of the explanatory variables in time?

By example: C(1) 1996: 0.333 C(1) 1997: 0.555 ....... C(1) 2011: -0.6

And so one for the different explanatory variables?

Again, thank you in advance

Winger

[EViews Gareth]

Ah, I see. You want cross-section dependent regressor coefficients too. We don't support that in Panels, but we do in Pools. You'll have to convert your panel into a pool.

[winger]

Ok,

Can i convert this from my panel data, because i opened my panel data as foreign data out of a excel file. I do not realy need all of the years to be accounted for, but i don need all the cross-sections. One coefficient before 2007 and one after, per cross-section would work to. Is this possible within panel data?


Thanks,

Winger

[EViews Gareth]

I'm not sure I follow your question, but converting from panel to pool is easy. Just click on Proc->Restructure Current Page->Unstack in new page.

In the first box of the dialog you enter the name of your cross-section identifier, and in the second box you enter the name of your time series identifier. Hit ok and you're done.

[EViews Glenn]

The one-coefficient-before-one-coefficient-after spec can done with panels (and more generally in all of EViews) simply by creating a dummy variable, say

Code: Select all

series after = @year > 2007

and instead of entering your specification as, say

Code: Select all

y c x

entering the interaction variables

Code: Select all

 y after (1-after) after*x (1-after)*x

The above spec assumes that you don't have fixed effects of any sort. If you do and don't want them interacted with the before-after period it's easy to adjust the spec. If you want an interaction of the fixed effects, it's a bit harder.

[d952]

Dear Eviews

I have some questions about applying Dynamic panel data model!
firsty, I would like to know that in a panel GMM estimation I should use the raw data or log of the data? I tried to use raw data since there is differenced option in the procedure, but my results look like the results of raw data and was not normal!

and my second question is that, if I choose the fixed effects option, my results do not show overidentified by Sargan test, in this case, is there any problem with my data?

I am looking forward to hear from you
Thank you in advance
Niaz

[d952]

by the way... I forgot to say that when I use fixed effects model, then the 2step procedure is near singular matrix, so I applied 1 step model and Sargan test reject the normality of the result!

[EViews Glenn]

I can't really help with either of those questions. Whether you should use the logs or levels of your data is really a question you need to answer about what specification you wish to estimate. And without knowing lots about your particular economic problem it's difficult to answer identification questions. I will note that you generally need a fair number of cross-sections and not too long panels for the DPD stuff to work well.