Problem with fixed effects estimation 'near singular matrix'

[gizmo]

Hello Guys,

I'm working with panel data in EViews 6 to estimate exports of different countries. I tried to estimate a pooled system with 29 cross sections over 3 years using LS.
When I try to include period random effects I get the error message 'Positive or non-negative argument to function expected for estimate of random effects innovation variance'.
When I try to include period fixed effects I get the error message 'near singular matrix'. I checked for collinearity in my dataset and I also tried to exclude variables, but it would not work.
I would like to know how I can add the effects.

Attached, find my dataset


I'd be really glad, if you could help me.
Thanks for your time.

[EViews Glenn]

There are a number of odd things with your data and specification so let me try to offer some perspective on what I think is going on here...

First, the workfile is a dated panel with 29 cross-sections and three years of data.

From what I can see in with the pool object EXPORT you are trying to estimate a pool specification on this panel. Specifically, you create a set of identifiers for each of the 29 cross-sections and then estimate a model which uses the stacked data. This really isn't what you want to do. Estimating the pool has the effect of estimating the same equation for each of the 29 cross-sections using all 87 observations in the workfile. (This had me confused for a bit since you have residuals for each of the 29 cross-sections that are perfectly correlated, and I couldn't figure one why that might be.)

What you need to do is to estimate a standard equation (not a pool). with your random effects specification. You can use a command like

Code: Select all

LS(CX=R) EXPORTS DISTANCE REALGDP REALGDPPC COSTI DTRADE DBORDER DLANGUAGE DLAND DCURRENCY

to estimate the RE model.

You should also note that you're not going to be able to estimate the fixed effects model since at least a couple of your variables (DISTANCE, DLANGUAGE) do not vary within cross-sections and are not identified in a fixed effects specification.

[gizmo]

Thank you so much! You helped to clarify things a lot for me. I'm new to EViews and still learning, so that really helped!

Cheers!

[Rashmi]

Hi, I am also facing the same problem with my panel data. I have removed the dummy variable but still getting "near singular matrix' error while running fixed effect regression.
I guess one of my independent variable is not varying cross-sectional i.e. concentration value (HHI). I have to report this variable same for each banks in a year. It varies only by year not by banks.
Can any body suggest me how to fix this problem. I can not leave this variable since it is going to decide the relationship.

Thanks

[startz]

If you have a variable that is the same for all banks in a given year (HH), then you cannot have period fixed effects. If you have a variable that is the same in all years for a given bank, then you cannot have time fixed effects.

[Mya]

Hey, you faced this problem in 2010 and i have the same problem today, I do a model of exports and when i run fixed effect it doesn't work prbably because of the constant variables (distance and dorders) what is the solution? how can i solve this?