Fixed effect - time & country fixed vs. time-country fixed

[su lee]

Hi, I am running two fixed effect regressions
- one with time fixed effect and country fixed effect
- one with time-country fixed effect (time and country pair)

The results - signs of coefficients are opposite in first and in second case. I understood that time-country fixed fair is more demanding. But I am confused why signs are different.
How should I interpret them?

Thank you very much for the comments!

[Carlo Lazzaro]

Su:
from your description, your models are different: hence, no wonder that they produce conflicting results (by the way, I would stick with the first one).
In your second code, you broke the panel data structure of your data and created a -newid- instead, based on original id a time pairs.
If this reply were not helpful, please. as per FAQ, provide what you typed and what Stata gave you back, along with an example/excerpt of your data via -dataex-. Thanks.

[su lee]

Dear Carlo,

Thank you very much for the reply.
My regression is the same as both models have the same structure except fixed effect. What I want to do is to variate the fixed effect : one with time FE and country FE and second one with time-country pair FE.
Could you please review my models?

xtset country year
xtreg depvar var1 var2 i.year, fe cluster(country)

xtset country-year
xtreg depvar var1 var2 i.year, fe cluster(country-year)

would it be any other way to diversify country-year FE?

Thank you so much!

[Carlo Lazzaro]

Su:
I do not think that "diversifying" country-year FE helps.
As already replied, I would stick with your fisrt code that makes methodological sense.
As per the folowing toy-example, you will see that, as expected, your models are not identical in the results they give (please aloso note the difference in the number of groups):

. use "http://www.stata-press.com/data/r15/nlswork.dta"
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. xtset idcode year
       panel variable:  idcode (unbalanced)
        time variable:  year, 68 to 88, but with gaps
                delta:  1 unit

. xtreg ln_wage age, fe vce(cluster idcode)

Fixed-effects (within) regression               Number of obs     =     28,510
Group variable: idcode                          Number of groups  =      4,710

R-sq:                                           Obs per group:
     within  = 0.1026                                         min =          1
     between = 0.0877                                         avg =        6.1
     overall = 0.0774                                         max =         15

                                                F(1,4709)         =     884.05
corr(u_i, Xb)  = 0.0314                         Prob > F          =     0.0000

                             (Std. Err. adjusted for 4,710 clusters in idcode)
------------------------------------------------------------------------------
             |               Robust
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0181349   .0006099    29.73   0.000     .0169392    .0193306
       _cons |   1.148214   .0177153    64.81   0.000     1.113483    1.182944
-------------+----------------------------------------------------------------
     sigma_u |  .40635023
     sigma_e |  .30349389
         rho |  .64192015   (fraction of variance due to u_i)
------------------------------------------------------------------------------

. egen idcode_year=group(idcode year)

. xtset idcode_year
       panel variable:  idcode_year (balanced)

. xtreg ln_wage age, fe vce( cluster idcode_year )
note: age omitted because of collinearity

Fixed-effects (within) regression               Number of obs     =     28,510
Group variable: idcode_year                     Number of groups  =     28,510

R-sq:                                           Obs per group:
     within  =      .                                         min =          1
     between =      .                                         avg =        1.0
     overall =      .                                         max =          1

                                                F(0,28509)        =          .
corr(u_i, Xb)  =      .                         Prob > F          =          .

                       (Std. Err. adjusted for 28,510 clusters in idcode_year)
------------------------------------------------------------------------------
             |               Robust
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |          0  (omitted)
       _cons |   1.674944          .        .       .            .           .
-------------+----------------------------------------------------------------
     sigma_u |  .47810418
     sigma_e |          .
         rho |          .   (fraction of variance due to u_i)
------------------------------------------------------------------------------

.

Again, If this reply were not helpful, please. as per FAQ, provide interested listers with an example/excerpt of your data via -dataex-. Thanks.