Why is a constant omitted?

[Julia Park]

Hi,
I have a quick question.
I ran a multilevel model below:

quietly xtset election
xtreg evaluation i.economy##c.soph partisanship income female married educ age i.country

However, the result shows that the constant is "omitted."
If I change the model specification by dropping one of the control variables, the constant is not omitted.

Why is the constant omitted in that model specification?
What would be the possible reasons when a constant is automatically omitted in Stata?
If it is omitted, is the estimates of the model biased and problematic?

I would appreciate any of your advice. Thank you!

[Steve Samuels]

You will have a better chance of getting a useful answer if you follow the FAQ guidelines (Section 12) and show the commands and the results of those commands. Put both between CODE delimiters as the FAQ requests.

[Julia Park]

Thank you for your advice. Please see below the results I got. You can see the constant is 0 (omitted).
Does it mean the constant is simply 0? or not necessarily 0 and omitted?
When and why do we get omitted constant?
Does it mean this model specification is problematic so that the estimates are biased?

I will be looking forward to hearing from you and other colleagues. Thank you very much. -Julia

. quietly xtset election
. xtreg evaluation i.econ##c.soph partyid income female married unemployed educ age i.co
> untry if country!=6430 & ug2pr==1

Random-effects GLS regression                   Number of obs      =     15478
Group variable: election1                       Number of groups   =        17

R-sq:  within  = 0.0966                         Obs per group: min =       252
       between = 1.0000                                        avg =     910.5
       overall = 0.1741                                        max =      1861

                                                Wald chi2(27)      =         .
corr(u_i, X)   = 0 (assumed)                    Prob > chi2        =         .

----------------------------------------------------------------------------------
         evaluation |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------+----------------------------------------------------------------
       1.econsal |  -.0394243   .0290655    -1.36   0.175    -.0963917    .0175432
            soph |   -.023869   .0075498    -3.16   0.002    -.0386663   -.0090718
                 |
  econsal#c.soph |
              1  |   .0493219   .0144529     3.41   0.001     .0209947    .0776491
                 |
           partyid |   .6356366   .0158746    40.04   0.000     .6045229    .6667503
          income |   .0091996   .0050462     1.82   0.068    -.0006908      .01909
          female |  -.0163028   .0121463    -1.34   0.180    -.0401091    .0075035
         married |   .0048474   .0136449     0.36   0.722    -.0218961     .031591
      unemployed |  -.0542336   .0296942    -1.83   0.068    -.1124332    .0039659
            educ |  -.0043813    .004217    -1.04   0.299    -.0126464    .0038838
             age |  -.0005712   .0004043    -1.41   0.158    -.0013636    .0002211
                 |
         country |
  0080. ALBANIA  |          0  (empty)
0360. AUSTRALIA  |   2.399828    .041077    58.42   0.000     2.319319    2.480338
   0760. BRAZIL  |          0  (empty)
   1240. CANADA  |   2.061612   .0427605    48.21   0.000     1.977803    2.145421
1580. TAIWAN ..  |    1.93494   .0405569    47.71   0.000      1.85545     2.01443
  2460. FINALND  |   2.211667   .0424087    52.15   0.000     2.128547    2.294786
   2500. FRANCE  |   2.115946   .0454603    46.54   0.000     2.026845    2.205047
  2760. GERMANY  |          0  (empty)
  3480. HUNGARY  |   2.460903   .0423674    58.08   0.000     2.377864    2.543941
           3720  |   2.633887   .0380259    69.27   0.000     2.559358    2.708416
   3760. ISRAEL  |          0  (empty)
    3800. ITALY  |   1.998325   .0485233    41.18   0.000     1.903221    2.093429
    3920. JAPAN  |   1.960844   .0503252    38.96   0.000     1.862209     2.05948
4100. REPUBLI..  |          0  (empty)
   4840. MEXICO  |          0  (empty)
5280. NETHERL..  |   1.939427   .0419305    46.25   0.000     1.857245    2.021609
5540. NEW ZEA..  |   2.219475   .0429933    51.62   0.000      2.13521     2.30374
   5780. NORWAY  |   2.305931   .0398395    57.88   0.000     2.227847    2.384015
     6040. PERU  |          0  (empty)
6080. PHILIPP..  |   2.260111   .0442556    51.07   0.000     2.173372    2.346851
   6160. POLAND  |          0  (empty)
 6200. PORTUGAL  |          0  (empty)
  6420. ROMANIA  |          0  (empty)
6430. RUSSIAN..  |          0  (empty)
 7050. SLOVENIA  |   1.939896   .0582978    33.28   0.000     1.825634    2.054157
    7240. SPAIN  |   1.959489   .0422664    46.36   0.000     1.876649     2.04233
   7520. SWEDEN  |          0  (empty)
8260. UNITED ..  |   2.123826   .0459088    46.26   0.000     2.033846    2.213806
8400. UNITED ..  |   2.418703   .0446524    54.17   0.000     2.331186     2.50622
                 |
           _cons |          0  (omitted)
-----------------+----------------------------------------------------------------
         sigma_u |          0
         sigma_e |  .74031103
             rho |          0   (fraction of variance due to u_i)
----------------------------------------------------------------------------------

[Carlo Lazzaro]

Julia:
have you ruled out already that:
-you haven't missing values for the omitted coefficients;
- the -if- conditions do not play some joke?

[Joao Santos Silva]

Dear Julia:

Following up on Carlo's comments, it looks as if you do not have (or are not using because of the -if-) data for some of the countries, and that implies that the dummies for these countries cannot be estimated. If you do not have data for the country being excluded (the base category), then the constant cannot be estimated either. This, however, is not a problem. I suggest that you do one of three things:

a) Change the country that is chosen as the base category; make sure it is one for which the dummy was estimated.

b) Before the estimation, drop the observations for the countries for which you have no data.

c) Create the country dummies before the estimation and then explicitly include them in the model; that is, do not use the i. notation for the variable country.

Best regards,

Joao

[Julia Park]

Thank you! I will try one of the three.
Best,
Julia