Different fixed effects in OLS estimation for panel data

[monica88]

Hi,

I do not have much experience with Eviews 7 and would like some help.

I have panel data for 6 countries over 80 quarters (Q1 1992 to Q4 2012) for the variables housing wealth, financial wealth and consumption. I would like to estimate the relationship between these using OLS. I am following the setup of another research paper, and they specify three models that they use:

Model 1: (A) Includes fixed effects (i.e. a set of dummy variables for each country).

Model 2: (A) Includes fixed effects (i.e. a set of dummy variables for each country) AND (B) country-specific time-trends.

Model 3: (A) Includes fixed effects (i.e. a set of dummy variables for each country) AND (C) year-specific fixed effects AND (D) seasonal (i.e. quarterly) fixed effects.

How do I do this in Eviews 7? I know I can choose between cross-section and period effects, but I don't know which they apply to (A, B, C, or D?) and how I get the other effects included in the estimation.

I hope someone can help me! Thank you in advance :)

[EViews Glenn]

If you are in a panel workfile

1. the built in cross-section fixed effects option

2. the built-in cross-section fixed effects option and the variable @EXPAND(ID)*@TREND
where id is the name of your cross-section ID series.

3. the built-in cross-section and period fixed effects options and the variable @EXPAND(@QUARTER). You'll have to use the option on @EXPAND to drop a quarter to avoid singularity.

There are different ways of doing this if you are using a pool object.

[monica88]

Thank you very much Glenn, this is very helpful! :)

Can I ask how I do it if I am using a pool object?

[EViews Glenn]

If you are in a pool workfile

1. the built in cross-section fixed effects option

2. the built-in cross-section fixed effects option and the put @TREND in the cross-section specific regressors box.

3. the built-in cross-section and period fixed effects options and the variable @EXPAND(@QUARTER). You'll have to use the option on @EXPAND to drop a quarter to avoid singularity.

[monica88]

Thank you very much :)

[monica88]

Hi again,

Sorry, I just had a problem with the last model (the built-in cross-section and period fixed effects options and the variable @EXPAND(@QUARTER). You'll have to use the option on @EXPAND to drop a quarter to avoid singularity).

I am using a panel workfile. When I tried this, I got the message: "Near singular matrix" :(

I first tried: logcons c loginc lognet logfin loghouse @expand(@quarter)
Then I tried: logcons c loginc lognet logfin loghouse @expand(@quarter,@dropfirst)

Have I not done it correctly?

Thank you for your help!

[EViews Glenn]

The period effects are always going to be correlated with your quarterly dummies since the latter are the same. I misspoke (mistyped) earlier. Since you are structured as a quarterly workfile, you'll want to expand the year variable to add your year dummies (I had thought you were structured as an annual). Note that you still may get singularities if there are any variables that don't have variation within one of you classifications.