difference in reg command and xtreg command with interaction dummies

[Jordy Snijders]

Hi,

When I run the following commands, normally, the coefficients of the reg command and xtreg command should be equal to eachother. When I run the understanding regressions, coefficients for both commands are not the same. Does anyone now how I can fix this? In other words, should both coefficients always be equal, or are there reasons why they are not the same?

reg Leverage_3 i.LF#c.Profitability_1 i.LF#c.Size i.LF#c.Growth_1 i.LF#c.Tangibility i.LF#c.Age i.Country i.NACEREV1, robust cluster (ID)

xtreg Leverage_3 i.LF#c.Profitability_1 i.LF#c.Size i.LF#c.Growth_1 i.LF#c.Tangibility i.LF#c.Age, fe robust (ID)

To make the regressions more clear. My sample consists of private family firms and public firms (LF = Legal Form). For my research, I study the impact of Profitability, Size, Growth, Tangibility and Age on Firm's Leverage. I control for Country fixed effects and Industry fixed effects (NACEREV1). For robustness, I run two regression to compare the outcomes of a simple Pooled OLS regression with the outcomes of a firm fixed effect regressions.

Thanks in advance!

Jordy

[Carlo Lazzaro]

Jordy:
welcome to this forrum.
What is the theoretical reason that support your belief to expect the same results from two different models?

[Jordy Snijders]

Hi Carlo,

Thanks for your reaction. That is the point that is unclear to me. When I compare my college sheets to some examples online. It is unclear to me whether the coefficients of a pooled OLS regression and the firm fixed effect regression should be equal to eachother or could differ from eachother. Therefore, I am doubting whether I have used the right regressions/commands and if my results are correct.

I hope that you can help me with this!

Kind regards,

Jordy

[Eric de Souza]

Without seeing your results it is impossible to diagnose what is happening.

webuse grunfeld
reg invest mvalue kstock i.company, robust cluster(company)
xtreg invest mvalue kstock, robust cluster(company) fe
xtreg invest mvalue kstock, robust fe

yield exactly the same coefficients

[Jordy Snijders]

Maybe this helps. In the annex you can find my results. My dataset is to big to share. If it is useful, I can provide my do file as well.

Thanks in advance!

[Eric de Souza]

This is not enough. But note that
in your first regression you have at the end

 i.Country i.NACEREV1, robust cluster (ID)

in your xtreg regression you have only

 , fe robust (ID)

The two models are not the same.

[Carlo Lazzaro]

Jordy:
please note that attaching spreadsheets is by no means the best way to share your datasets (the risk of downloading malware is often a good reason to ignore them); for the future, use -dataex-. Thanks.