including time dummy variables make all other variables insignificant

[Junaid Ahmed]

Dear Statalist,

I am working on bilateral remittances data over the period 2011-2017 for a set of developing economies, using a fixed-effect model with time dummy make several variables insignificant.

xtset pairid year

xtreg lcost2 lmig_st exc_stab exc_stab_hos lll_hos lll rural_pop i.year, fe vce(robust)

Do we may also take the interaction of host and home countries with time dummy variables, for example, i.year#i.host

KIndly guide me.

Best regards,
Junaid

[Dario Maimone Ansaldo Patti]

The thing I would do first is to checjif the time dummies are significant or not.

testparm i.year

if they are not significant, you can drop them.

[Junaid Ahmed]

xtreg lcost2 $xlist0

 i.year , fe cluster(pairid) 

Fixed-effects (within) regression               Number of obs      =       731
Group variable: pairid                          Number of groups   =       206

R-sq:  within  = 0.2813                         Obs per group: min =         1
between = 0.0023                                        avg =       3.5
overall = 0.0000                                        max =         6

F(10,205)          =     11.63
corr(u_i, Xb)  = -0.6580                        Prob > F           =    0.0000

(Std. Err. adjusted for 206 clusters in pairid)

Robust
lcost2       Coef.   Std. Err.      t    P>t     [95% Conf. Interval]

lmig_st     .069615   .0503821     1.38   0.169    -.0297185    .1689484
exc_stab    .0873111   .0568756     1.54   0.126     -.024825    .1994472
exc_stab_hos    .2272386   .0836498     2.72   0.007     .0623144    .3921628
lll_hos   -.5768892   .2106512    -2.74   0.007    -.9922097   -.1615686
lll    .3598963   .1269386     2.84   0.005     .1096237    .6101689
             
year 
2012     .0539734   .0179348     3.01   0.003     .0186131    .0893336
2013    -.0141409    .018561    -0.76   0.447    -.0507358     .022454
2014     -.116838    .029934    -3.90   0.000     -.175856   -.0578199
2015    -.1314768   .0330942    -3.97   0.000    -.1967254   -.0662282
2016    -.1824544   .0307185    -5.94   0.000    -.2430192   -.1218897
             
_cons     2.18509   1.460114     1.50   0.136     -.693675    5.063855

sigma_u   .57456225
sigma_e   .15322647
rho   .93360198   (fraction of variance due to u_i)


. 
end of do-file

. testparm i.year

( 1)  2012.year = 0
( 2)  2013.year = 0
( 3)  2014.year = 0
( 4)  2015.year = 0
( 5)  2016.year = 0

F(  5,   205) =   12.98
Prob > F =    0.0000 

[Junaid Ahmed]

Dear Dario Maimone Ansaldo Patti , thanks for it. Please check the estimation results above

[Dario Maimone Ansaldo Patti]

Difficult to read as you should put the table under [CODE] [\CODE]. Said that it seems that time variables are strictly significant. Hence you should keep them in your model