EViews.com

Hello,

This is the first time that I am dealing with panel data, I already read through this topic and I considered it to be very helpful, thank you for all your efforts!
However, my data is unbalanced and even though I read the post of distressedbydiso (dated 13th of February 2012), I am still in doubt and would like to be 100% sure before proceeding. Short version of my data:
Year - Country - Variable1 - Variable2
1995 - US - 512 - NA
1996 - US - 498 - 12
1995 - UK - 322 - xxx
1996 - UK - 340 - xxx

In this case, where only one of my variables for 1995 is missing for the US, should I delete 1995 - US from my dataset anyway or keep it?
Thank you in advance!

That's an econometric question, rather than an EViews one. EViews will read the unbalanced data in without any problems.

Dear Gareth,

For our master thesis we have to set up a bankruptcy prediction model, using financial variables of a sample of firms, and extend this with some non financial variables which are country specific.
We have an Excel file that is structured as your first format. ( We use Eviews 7)
Our sample consists of firms from different countries, 50 % of them are failed and the other 50 % are still healthy, we matched the failed and healthy ones for each country. We would like to create a prediction model that is able to predict which firms went bankrupt and which are healthy on the basis of our data from three years in the past.

In the attachment you can find our Excel file

Could you please have a look at our Excel sample and give some advice for how we could import this into Eviews ?
In this forum we found several similar examples but our seems to be a litte more complicated because of the distinction between
failed and healthy and the fact we have several firms from the same country.



Kind Regards,
Brecht

What are the dimensions of your panel? What are the dimension identifiers?

Dear Gareth,

Our dimensions are time and cross sections.

A little more information than that is required. What are the actual series containing the dimensions of the panel?

Dear Gareth,

What do you exactlly mean with the actual series containing the dimensions ?
We've only had a little basic statistics so our knowledge of Eviews and panel data isn't that high.

Kind Regards

Ignore EViews and statistics for the moment...

You state you have an Excel file containing panel data. You say that the dimensions of the panel are time and cross-sections. Which columns in the Excel file contain that information? I'm going to take a guess and say that "Year" contains the time information. Which column contains the cross-sections?

The Cross -Sections are our companies but we didn't incorporate the names of it,
the different companies are structured under the different countries with their data from 2005-2008 ...

When following the instructions on page 1 of this topic, I can import the sample and the variables into Eviews,
but when I try to structure my sample from unstructured to Dated Panel Eviews gives me the following
message :
' indices do not uniquely identify observations. 48 duplicate date(s) found. Create CELLID series to identify observations? Yes or NO

So the problem is that for every country we have more than one set of observations from our variables for each year because of the several companies ?

So is the solution to incorporate the names ?

kind regards

Right, EViews needs to be able to identify your observations. To do that, it needs information on the time periods and on the cross-sections. If your data file is missing that information, EViews can't interpret it as a panel properly.

We entered a new column in our Excel file to identify the companies with a number. Now we can successfully import our database into Eviews.

As already mentioned we have to create a model which predicts bankruptcy so makes a distinction between failed and non -failed.

For that we should use a binary logit model. We allready incorporated a column (status) with a 0 for Failed and a 1 for Non-Failed.

This is our estimation:
status c roa intco proma cura sora log(totas) gdpppp unra ltir cpi recra contime

Is our way of working correct ? Does Eviews know that 0 means Failed and 1 stands for Non Failed ?

Kind Regards

EViews doesn't care.

Dear Gareth,

We have been able to create a binary logit model and incorporate our sample into Eviews.

Because we had a lack of data for several countries we have grouped them into 4 groups which have the same legal background.

Now we would like to put these four groups into the same sample ( otherwise the number of observations is too small ) by using dummy variables.

So four groups means three dummy variables and we gave a 1 to all the companies coming from a country in that group.

Is this correct ?

Because when doing the estimation we get negative coefficients for the three groups and we only get a R squared of 7% compared to 70% when using groups separately..

Kind Regards

I cannot comment on econometrics rights or wrongs.