Random effect model - controlling for time

[Mathias Rechsteiner]

Hy folks,

i have the following panel data:

ID	Birthyear	Year of Assessment	Fahter played football	Math	Gender	Age
1	1980	1988	0	50	0	8
1	1980	1990	0	55	0	10
1	1980	1992	0	60	0	12
2	1990	1998	1	70	0	8
2	1990	2000	1	75	0	10
2	1990	2002	1	80	0	12
3	1985	1993	0	60	1	8
...	...	...	...	...	...	...

and would like to know, whether the parental sporting activity is positively related to the childrens school performance. In the observed country, the rate of parents playing football increased a lot over time, as well as the math marks are increasing over the ages and year of assessments. So i was thinking using the following model

xtset ID Birthyear
xtreg Math fatherplayedfb gender, re

The question raises, whether i should control for time, as the rate and marks are increasing?

xtset ID Birthyear
xtreg Math Fatherplayfb gender YearofAssessment, re

or can you recommend any other suitbale model?

thank you very much for your aprreciate adivce.

best wishes

Mathias

[Joro Kolev]

If you had actually tried what you are saying

. xtset id birthyear
repeated time values within panel
r(451);

you would have already noticed that there is a problem, and the problem is that for each id, the birthyear is constant, or at least should be constant if you do not have data errors. We all get born only once, that is, everybody has only one birthdate.

So probably what should be done here is

. xtset id yearofassessment
       panel variable:  id (unbalanced)
        time variable:  yearofasses~t, 1988 to 2002, but with gaps
                delta:  1 unit

And yes, you should include year fixed effects.

[Mathias Rechsteiner]

thank you very much for your prompt advice. So you recommend:

xtset id YEAROFASSESSMENT
xtreg Math Fatherplayfb gender, YEAROFASSESSMENT age, re

or is it wrong to use it twice in the "xtset" and "xtreg"?

Best wishes

[Carlo Lazzaro]

Mathias:
if you're interested in time fixed effect considering -YEAROFASSESSMENT- as a categorical variable, your code should become:

xtreg Math Fatherplayfb gender i.YEAROFASSESSMENT age, re

Conversely, if you want to investigate the existence of turning point in -YEAROFASSESSMENT-, your code should become:

xtreg Math Fatherplayfb gender c.YEAROFASSESSMENT##c.YEAROFASSESSMENT age, re

I cannot say whether or not you will find collinearity issues between -YEAROFASSESSMENT- and -age-.