FE regression uneven space

[Jules Herz]

Hey everyone,

I have a problem running a FE regression in Stata due to the data structure I have. I have observations for several countries and times, but some years appear multiple times per country. Here is a stylized example:

Country	Year	Var
A	2020	1
A	2020	2
A	2021	1
B	2020	4
B	2021	7

If I set "xtset Country Year", it says that multiple time values appear (which I get). My idea was to assign different dates to observations that took place in the same year, e.g. 2020/01/01, 2020/01/02 and so on to have unique dates. This solves the multiple time value problem but I cannot run a FE regression, probably because the distances between the time points are not even.

Do you have an idea how I can run a FE regression with my data set?

Best,
Jules

[Andrew Musau]

As long as you are not using time-series operators, such as lags, leads, and differences, you can xtset your data with the panel-identifier only.

xtset Country

See https://journals.sagepub.com/doi/abs...6867X231162020 for some discussion.

[Jules Herz]

Thank you for your quick response. This certainly solves the problem of multiple time values. But I want to use country fixed effects as well as time fixed effects in my regression model. Is this possible with setting just one panel-identifier as you described?

[Andrew Musau]

Yes. The time variable is only needed if you are using time-series operators, as I stated.

xtset Country
xtreg y x controls i.Year, fe robust

[Jules Herz]

Thank you! When I run the regression now as you suggested, Stata shows me "option FE not allowed r(196)". Without the "fe" specification and with "i.Year" it works:

xtreg y x controls i.Year, robust

But is this now a regression with time and country fixed effects or just with time FE?

[Carlo Lazzaro]

Jules:
you actually coded -xtreg- wity the -re- specification (which is the default specification for -xtreg-).

[Jules Herz]

Carlo Lazzaro: If xtreg is the random effects specification, how do I get FE for country and time? Would this work:

xtreg y x controls i.Year i.Country, robust

[Carlo Lazzaro]

Jules:
1) -xtreg <depvar> <predictors> <controls>, re = -xtreg <depvar> <predictors> <controls>, re;
2) the categorical predictors you typed will give back their contributions to variation in the <depvar> between panel units (and not within each panel, as the -fe- estimator is expected to do);
3) just out of curiosity: how can it be that Stata tells you that -fe- option is not allowed?

[Jules Herz]

1)+2) Thanks for clarification. But how do I run then a within-panel FE regression if the command is not allowed?
3) I currently have no idea why Stata does not let me run a FE regression. It also shows the same error if i type , RE instead of FE

[Andrew Musau]

What happens if you run

webuse grunfeld, clear
xtreg invest mvalue kstock, fe
which xtreg

? Copy and paste the output here.

[Jules Herz]

This is the output I get:
[ATTACH=CONFIG]n1735085[/ATTACH]

[Carlo Lazzaro]

Jules:
you probably need to update your copy of Stata:

. webuse grunfeld, clear

. 
. xtreg invest mvalue kstock, fe

Fixed-effects (within) regression               Number of obs     =        200
Group variable: company                         Number of groups  =         10

R-squared:                                      Obs per group:
     Within  = 0.7668                                         min =         20
     Between = 0.8194                                         avg =       20.0
     Overall = 0.8060                                         max =         20

                                                F(2,188)          =     309.01
corr(u_i, Xb) = -0.1517                         Prob > F          =     0.0000

------------------------------------------------------------------------------
      invest | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
      mvalue |   .1101238   .0118567     9.29   0.000     .0867345    .1335131
      kstock |   .3100653   .0173545    17.87   0.000     .2758308    .3442999
       _cons |  -58.74393   12.45369    -4.72   0.000    -83.31086     -34.177
-------------+----------------------------------------------------------------
     sigma_u |  85.732501
     sigma_e |  52.767964
         rho |  .72525012   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(9, 188) = 49.18                     Prob > F = 0.0000

. 
. which xtreg
C:\Program Files\Stata17\ado\base\x\xtreg.ado
*! version 1.9.1  10feb2021

.