EViews.com

Hello,
Its being a long time since I have done a Panel Data regression.

I have the following questions.

I have 3 variables information (one is the dependent: "Price") of 8 companies for a period of 3 years (4 quarters). (First comment: I think there is not enough data)

The paper that I am trying to replicate talks about "vectors containing firm, year and quarter effects".

I am using dummies (as a proxy of what they are doing) (second comment: I am not sure if that is a proper way of doing it)

My questions are the following:

1) If I regress "Price" and i include a dummy for each quarter and each year, but one (because i have a constant to capture that particular effect, lets say i do not include the dummy for the last year) Eviews alerts me that "near singular matrix". Whats the problem here?
2) If i regress "Price" and forget the dummies and go to "Panel Options" and select "Fixed" in "Period" from "Effects Specification". The Regression goes as planned, but I do not know if the "years" and the "quarters" are controlled. It is that correct?
3) Do I have to say something also about the effect of "cross section"?

Thank you for your time and thank you for any help

Best

1) Probably a dummy variable trap.
2) Not sure what you're asking, but period fixed effects means an effect for every period. If you have quarterly data, then it will be an effect for each quarter in your sample.
3) Hard for us to say.

1) Probably a dummy variable trap.
2) Not sure what you're asking, but period fixed effects means an effect for every period. If you have quarterly data, then it will be an effect for each quarter in your sample.
3) Hard for us to say.

Thank so much for your help!

Let me explain myself a little better. Sorry If I was not clear.

Agreed with the dummy trap. But I am not including one variable to avoid it. Like for example. If I have 4 quarters and 3 years.
I will include a dummy for the First, second and Third quarter (leaving out fourth) and 3 dummies for each year (Comment: Also I do not know If this is correct)

"If i regress "Price" and forget the dummies and go to "Panel Options" and select "Fixed" in "Period" from "Effects Specification". The Regression goes as planned, but I do not know if the "years" and the "quarters" are controlled. It is that correct?"

Here I am asking. I am using Eviews 7. When I go to "Estimate an Equation". I have the Option. That says "Panel Options". If I mark Fixed in "Periods". And I do not incluide any dummy. I am controlling the effect of each quarter and each year? I am asking this, because if it does that. I do not need to put any dummy.

Best Regards and thank you again

The fact that you are dropping a single dummy doesn't mean you don't still have perfect collinearity due to the dummy variables.

Period fixed effects will control for each period, yes.

The fact that you are dropping a single dummy doesn't mean you don't still have perfect collinearity due to the dummy variables.

Period fixed effects will control for each period, yes.

Great! Thank you again!

So to be sure (I am a little slow :))

If i put "Period fixed effects" I do not need to put any dummy to control "quarter or year".

Last question!

Given the fact that the data Cross Section ID are the Firms. I have to control also that by putting "Fixed" in "Cross-Section" at the "Effects Specification" in the "Panel Options"

Best Regards and thank you for the second time!

Odds are that the original article had a dummy for each year and seasonal dummies for the quarters, as opposed to a dummyfor each quarter in the sample.

Odds are that the original article had a dummy for each year and seasonal dummies for the quarters, as opposed to a dummyfor each quarter in the sample.

Hi Startz

What do you mean by "the original article had a dummy for each year and seasonal dummies for the quarters, as opposed to a dummyfor each quarter in the sample"

I understand that

Dummy for year 1 = 1 for each row of data that is in that period; 0 if not
Dummy for quarter 1 = 1 for each row of data that is in that period: 0 if not

You are saying that? Or other thing

Best and thank you!

One dummy for winter, one for spring, etc. As opposed to one dummy for winter 1990, one for winter 1991, etc.

One dummy for winter, one for spring, etc. As opposed to one dummy for winter 1990, one for winter 1991, etc.

Ahh ok. I am doing that also.

Can you help me with a question that I ask earlier?

If i put "Period fixed effects" I do not need to put any dummy to control "quarter or year".

Last question!

Given the fact that the data Cross Section ID are the Firms. I have to control also that by putting "Fixed" in "Cross-Section" at the "Effects Specification" in the "Panel Options"

Any idea?

Best and thank you for your help