FIXED EFFECT regressions

[lorenzoruggeri]

Hello,

I am suppose to evaluate the consequences of a merger on banks' performance, examining the main balance sheet indicators of costs, revenues, and profitability (dependent variables). I should use fixed effect regressions where explanatory variables are dummy variables that take the value of 1 either in the year of the merger (MERGEO), in the following three years (MERGE13), or in all years after the third (MERGEGT3).

I am wondering how I can estimate this kind of equations in Eviews and if is there any variable or information that I miss?

Thanks for helping!

[EViews Gareth]

That question is a little vague, but assuming you have a panel data workfile and wish to estimate a panel regression with fixed effects, simply use the "Panel Options" tab of the estimate dialog.

[lorenzoruggeri]

My workfile consists of almost 2000 banks and of a period of 12 years (hence 12 x 2000 = 24000 observations). To make an example, for 1 bank which engage in a deal in 2008 (dependent variable is SIZE and explanatory variables are dummies as previously discussed):

size(2012) = alpha + beta1 * (MERGEO = 0) + beta2 * (MERGE13 = 0) + beta3 * (MERGET3 = 1)
size(2011) = alpha + beta1 * (MERGEO = 0) + beta2 * (MERGE13 = 1) + beta3 * (MERGET3 = 0)
size(2010) = alpha + beta1 * (MERGEO = 0) + beta2 * (MERGE13 = 1) + beta3 * (MERGET3 = 0)
size(2009) = alpha + beta1 * (MERGEO = 0) + beta2 * (MERGE13 = 1) + beta3 * (MERGET3 = 0)
size(2008) = alpha + beta1 * (MERGEO = 1) + beta2 * (MERGE13 = 0) + beta3 * (MERGET3 = 0)


How can I "format" this file to be a panel data workfile?

I am using Eviews 7 and I couldn't find the "Panel Options" tab in the "equation estimation" window..did you mean somewhere else?

Sorry for the basic questions.

Thank you very much!

[EViews Gareth]

http://forums.eviews.com/viewtopic.php?f=7&t=74

[dianaur86]

Hi, I have data for 135 firms during 3 years. Some of my explanatory variables are dummies (e.g. if the company is multinational or not) and some others have the same value as well for the three years for each company (e.g. the percentage owned by a family for each firm).
I already put the data as a panel, and want to estimate the fixed effects. When I choose, quick, estimate equation I am writing the normal regression as :
ccc c fam tn cd bs lev rev list tech fp .
My first question is if I should include the constant or not? and my second question is that when I select Fixed in the panel options, it appears "near singular matrix". And when I take out all the dummies and variables that don´t change across the time for each company, it works. However I need this dummies and variables in my model and they are not related to each other.
Maybe I am doing some mistake in the specification of the equation, or the equation can be written exactly as in OLS?
How to include this variables that don´t change value for the companies across the time?

Thanks

[startz]

You can't have both company fixed effects and variables that don't change across time for a given company. Nothing to do with EViews, it's just a fundamental property of fixed effects.

[dianaur86]

Ok, now I understand. Thank you so much! I have spent so long time reading but I just get more confused! now I know that I cannot use fixed effects model!! :D ...
I am not an expert in econometrics and I am just getting familiarized with EViews, sorry! Just two more questions:
1)So I could just estimate my regression in a pooled data or in a panel data?
2)Actually, in an article they had a similar regression and they used weighted generalized least squares: "general least square with cross section weights. In this regression, the common intercept was calculated for all variables and assigned a weight". How can I do this in Eviews? In the equation estimation dialog, in the panel options, I see that I can select GLS weights: cross section weights. What does Eviews do more specifically, how it calculates the weights? is it the same weighted generalized least squares?
Thank you again!!

[startz]

Those are much harder questions. You may want to look at the help system to see exactly what EViews does.

[ANURADHA AGARWAL]

Hello Sir,
I have a quarterly data for 5 countries over a period of 15 years with 11 explanatory variables. I want to apply fixed effect model with dummy variables and i use @expand(@crossid) and get the result with none cross section effect. But In case of fixed cross effect specification it shows a near singular matrix.
Please tell me how can i do it.

[startz]

Do you have a constant term? Or are some of your variables already fixed for each cross-section?

[ANURADHA AGARWAL]

Thanks Sir for responding to me

Sir I have a same 11 explanatory variables for each countries and i am not including constant term in regression equation with dummy variable while taking fixed cross section effect but still it shows same result (near singular matrix).
Please tell me sir is it valid if i am following same process with none cross section effect and continue it. Otherwise please suggest me some other method to do it.

[startz]

Do any of your variables always have the same value for a given country?

[ANURADHA AGARWAL]

No sir, there is no variable having same value in my study.

[startz]

You might want to post your workfile, including the equation you wish to estimate.

[ANURADHA AGARWAL]

The regression equation is "cab adr ds fb ir lp nfa reer ri to tot expand@(crossid)". I want to apply LSDV model or the fixed effect model with dummy variables. Please help me.