Withinh group analyisis.

[Angeliki Skoura]

Hi Statalisters,

I need your valuable assistance.
I am troubled by the analysis within groups. More specifically, I want to see if there is a difference of income in the specific year of the status change (from 1 to 0) compared to the other years of the company's existence. I have also created a dummy variable indicating whether the firm is undergoing a reform and thus changing status.

Example of my data follows:

firm	year	status	dummy	income
1000	2005	1	1	6112
1000	2006	1	1	4402
1000	2007	1	1	5970
1000	2008	0	1	1079
1000	2009	0	1	3338
1000	2010	0	1	1215
1000	2011	0	1	1703
2000	2007	1	1	3068
2000	2008	1	1	1210
2000	2009	1	1	7445
2000	2010	0	1	3333
2000	2011	0	1	4521
2000	2012	0	1	9876
2000	2013	0	1	7576
3000	2007	0	0	1111
3000	2008	0	0	2222
3000	2009	0	0	3333
3000	2010	0	0	4444
3000	2011	0	0	5555
3000	2012	0	0	6666

Αny advice would be appreciated. Thank you in advance.

[Carlo Lazzaro]

Angeliki_
I would try:

xtset firm year
xtreg income i.status i.dummy i.year, fe

If you have => 30 firms, go -vce(cluster firm)- standard errors.

[Angeliki Skoura]

Thank you for your ideas , it was really helpful.
But I have one more question, more specifically (1) do the p-values show if there is a statistically significant difference per year in terms of income when status =1 or (2) does it compare the income per company when the status in that company is equal to 1 with the other years when the particular company had status equal to 0 ?
If not how do you think I could make the second argument (2) ?
I attach the results below:

Thanks for your attention!

Best wishes,
Angeliki

[Angeliki Skoura]

Hi everyone, I still have not figured this out. Anyone has an idea?

Thank you in advance.

Best,
Angeliki

[Carlo Lazzaro]

Angeliki:
assuming that your modell is correctly specified (somethiing I'm pretty queasy about, given a sigle predictor plus -i.Year-):
1) please consider that the -fe- estimator focuses on within (the same panel) variations. Between panels variatins are not consdered;
2) when -i.status- switch from 0 to 1, the dependent variable increases, on average and othe things being equal, by 22176.11, regardless of the effect of -i.Year- that is measured separately;
3) -i-Year-represents the contribution of time, on average and other things being equal, to within panel variations of the dependent variable.

[Angeliki Skoura]

Thanks for your help in this matter, it was extremely helpful to me.

Best wishes,
Angeliki