EViews.com

Hi,

I am trying to perform a panel regression but I’m having some trouble here and was hoping that you guys could help me.
Although I have been using EViews for only a couple of days now (please keep that in mind when you answer ;-)), I am familiar with the basic estimation of panel data. But I cannot figure out (or find anything in the user’s guide) how to solve the following problem:

In my data set I have a list of donor and recipient countries. My dependent variable are the credit claims of country j (recipient) to country i (donor). In addition I have some independent variables, most of them for either for i or j. Thus what I have looks something like this:
Claimsijt = (…) b1 x1it + b2 x2jt + b3 x3ijt (…)

I have only found references to panel data where there is only one set of individuals, i.e. only i OR j. How can I implement TWO sets?
If e.g. i = 1,2,3 and j = 4,5,6, how do I specify cross section identifiers?
Simply writing
_1
_2
_3
_4
_5
_6
won’t work (as I am not interested in a country’s own claims), will it?

I really hope someone can help me with this problem. I of course apologize if a similar question has already been answered (I could not find anything though).
Thank you very much in advance for answering!

(I’m using EViews 7)

Can someone tell me if that's even possible with EViews?
And as I've noted above, I'm not that famliar with EViews yet, so is it possible to do such a regression simply by using the pool object and specifying cross-section indentifiers?
I would really appreciate it if someone could help me :-) Thanks in advance again!

To be honest, I'm not quite certain what you are trying to do here. Could you please try to (re)explain in a bit more detail.

Sorry, I will try to explain again in more detail ;-)

For simplicity assume that I have 4 countries (although I have of course many more) I observe, let's call them A1, A2, B1 and B2.
A1 and A2 are donor countries, thus their banks lend credits to B1 and B2 (i.e. A1 lends to B1 & B2 and in addition A2 lends to B1 and B2. Just to avoid misunderstandings ;-)) which result in credit claims. Thus, my dependant variable is:
claimsijt, where i = B1,B2 and j=A1,A2. Consequently, I have TWO sets of individuals. (I am not interest in credit claims from country A1 to A2 or A2 to A1and also not in a country's banks' claims on domestic debtors, i.e. for example in A1's claims on A1)

I try to explain these credit claims with a set of independant variables. Among them are:
- dgdpit = gdp growth rate in countries B1,B2
- dgdpjt = gdp growth rate in countries A1,A2
- interestijt = real interest rate spread, i.e. interest rate in j - interest rate in i

Thus, my variables in my data set look like this:
- dgdp_A1
- dgdp_A2
- dgdp_B1
- dgdp_B2
- claims_A1_B1
- claims_A1_B2
- claims_A2_B1
- claims_A2_B2
- interest_A1_B1
- interest_A1_B2
- interest_A2_B1
- interest_A2_B2

What I would normally do with panel data is just create a pool object and type in all cross-section identifiers. But if I type
_A1
_A2
_B1
_B2,
then I would tell EViews that there is only ONE set of individuals. What I would do then is:
dependant variable: "claims??"
independant variable: "dgdp?" and "interest??"
where ? tells EViews to insert the cross-section identifiers. But that, of course, won't work. I would be telling EViews to check for A1's claims on A1, A2's claims on A2, B1's claims on A1 and so on and so on. And that is not what I am trying to do here.
Instead, I need to tell EViews that there are TWO sets of individuals, i=B1,B2 and j=A1,A2. I and do not know, how this is done or if it's even possible.


What I have tried to do is to specify the following cross-section identifiers:
_A1_B1
_A1_B2
_A2_B1
_A2_B2
to avoid to create to two sets of indivuals. Then, I simply typed:
dependant variable: "claims?"
independant variable: "dgdp?" and "interest?"
But that does of course not work either. What it would result in is that it works fine with the claims and the interest rate spreads but NOT with dgdp as I have data for every country's gdp growth rate.

So, I am at a loss here. I do not know how to do this correctly, do not know if it is even possible and do not know if it's maybe completely easy to do and nevertheless I cannot figure it out.

I hope that it's clearer now what I'm up to and that it's not too much text ;-)

It seems to me like a spatial model than a panel. It's a panel structure if only every recipient country has ONLY one creditor, more than one creditor makes it a spatial model. I'm not sure how to model that in EViews, but there was a recent discussion on that!

Mhh ... I have never hear of spatial models and I cannot find anything useful about it either. Do you maybe have a link?


In addition, I've been trying something else which I have a question about.
Since I could not estimate the model like explained above, I have simply estimated the following:

claimsit = (...) b1 dgdpit + b2 interestit

with i=B1,B2 for every country j=A1,A2. Thus, I have estimated a model for every donor country j.
By doing that I avoid having to have two sets of indivuals.
I have 2 questions though:

1) Should I use a fixed effects model or an SUR? I have looked at some of the error terms' correlation matrices and the error terms are not that highly correlated, but I get very different results with an SUR. Many more variables are significant and I usually get a much higher R².

2) In a paper I have read the authors say that an SUR requires a balanced data set which I don't have. Thus an SUR should significanlty reduce my number of obersavations. But it doesn't. This is an example of what I get:

Dependent Variable: DCLAIM_BG?
Method: Pooled EGLS (Cross-section SUR)
Included observations: 17 after adjustments
Cross-sections included: 11
Total pool (unbalanced) observations: 181
Linear estimation after one-step weighting matrix

It says 'unbalanced'. So, I do not seem to need a balanced panel, do I? (Also there's a box titled 'unbalanced SUR approximation')