xtlogit, re gives constant with large magnitude

[Fadhili Sooi]

My xtlogit, re regression gives constant estimate of 40.23.. (and is significant) while my model covariates have estimates below 1. Is this a sign of not including all variables or not using the right model? Does the magnitude of constant matter in these models?

[Carlo Lazzaro]

Fadhili:
welcome to this forum.
Nobody, I suppose, can reply positively to your query unless you share what you typed and what Stata gave you back (as per FAQ). Thanks.

[Fadhili Sooi]

Thank you Carlo. So, briefly about what i am doing. I am running a regression to identify the factors that increase likelihood SMEs to participate in linkages with large firms (dependent variable). I am focusing on 4 types of linkages (as identified through literature) which are all dummy variables (1- for SMEs participate in linkage; 0- SME does not participate). This is what I typed in my stata (ver 14.2)

xtlogit tech1 shotrawm ltotprod train lexper association export foreignown capacity2 private i.year i.region i.sector, intpoints (120)

tech1 (technology linkage) shotrwam (shortage of raw material); train (whether or not firm provide training); association (firm is an association member or not); private (firm is in private sector or not); export (a firm exports her output or not); foreignown (a firm has foreign ownership); capacity2 (a firm operates below 80% capacity)--- these are all dummies

lexper is log of age of the firm; ltotprod is the log of total production

The stata output

 xtlogit tech1 shotrawm ltotprod train lexper association export foreignown capacity2 pri
> vate i.year i.region i.sector, intpoints (120)

Fitting comparison model:

Iteration 0:   log likelihood = -2737.1107 
Iteration 1:   log likelihood = -2259.3343 
Iteration 2:   log likelihood = -2252.2398 
Iteration 3:   log likelihood = -2252.1895 
Iteration 4:   log likelihood = -2252.1894 

Fitting full model:

tau =  0.0     log likelihood = -2252.1894
tau =  0.1     log likelihood = -2225.4985
tau =  0.2     log likelihood = -2196.5444
tau =  0.3     log likelihood = -2165.1468
tau =  0.4     log likelihood =  -2131.136
tau =  0.5     log likelihood = -2094.3711
tau =  0.6     log likelihood = -2054.7963
tau =  0.7     log likelihood = -2012.6291
tau =  0.8     log likelihood = -1969.1241

Iteration 0:   log likelihood = -2012.6291 
Iteration 1:   log likelihood = -2002.5601  (not concave)
Iteration 2:   log likelihood =  -1787.106 
Iteration 3:   log likelihood = -1736.2235 
Iteration 4:   log likelihood = -1726.7616 
Iteration 5:   log likelihood = -1725.8781 
Iteration 6:   log likelihood = -1725.8697 
Iteration 7:   log likelihood = -1725.8697 

Random-effects logistic regression              Number of obs     =      4,094
Group variable: id                              Number of groups  =      2,457

Random effects u_i ~ Gaussian                   Obs per group:
                                                              min =          1
                                                              avg =        1.7
                                                              max =          2

Integration method: mvaghermite                 Integration pts.  =        120

                                                Wald chi2(38)     =     158.16
Log likelihood  = -1725.8697                    Prob > chi2       =     0.0000

--------------------------------------------------------------------------------
         tech1 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------+----------------------------------------------------------------
      shotrawm |   .8140508    .422711     1.93   0.054    -.0144475    1.642549
      ltotprod |   .9097748   .1431753     6.35   0.000     .6291563    1.190393
         train |   2.337424   .4257947     5.49   0.000     1.502882    3.171967
        lexper |   .8892272   .2076141     4.28   0.000     .4823111    1.296143
   association |   4.423817   .5444025     8.13   0.000     3.356808    5.490826
        export |   .4651273   .9389973     0.50   0.620    -1.375274    2.305528
    foreignown |  -.7693068   .8725729    -0.88   0.378    -2.479518    .9409046
     capacity2 |   .3488401   .3641837     0.96   0.338    -.3649468    1.062627
       private |   .3545232    .850531     0.42   0.677    -1.312487    2.021533
               |
          year |
         2016  |  -.2214501   .1865467    -1.19   0.235    -.5870749    .1441748
               |
        region |
       Arusha  |  -2.199402   1.697751    -1.30   0.195    -5.526933    1.128128
  Kilimanjaro  |  -3.936974   1.773016    -2.22   0.026    -7.412022   -.4619257
        Tanga  |  -1.470954   2.049484    -0.72   0.473    -5.487869    2.545961
     Morogoro  |   1.970796   1.585128     1.24   0.214    -1.135998     5.07759
        Pwani  |  -.6852578   1.944615    -0.35   0.725    -4.496633    3.126117
Dar es Salaam  |  -3.844667   1.545016    -2.49   0.013    -6.872843    -.816491
        Lindi  |   .1250497   2.599072     0.05   0.962    -4.969037    5.219136
       Mtwara  |  -6.042631   2.964485    -2.04   0.042    -11.85292   -.2323476
       Ruvuma  |   5.828933   2.394214     2.43   0.015     1.136359    10.52151
       Iringa  |   1.671959   2.580392     0.65   0.517    -3.385516    6.729434
        Mbeya  |   .6942266   1.816674     0.38   0.702     -2.86639    4.254843
      Singida  |   7.329189   1.901211     3.86   0.000     3.602885    11.05549
       Tabora  |   .4233345   3.266224     0.13   0.897    -5.978348    6.825017
        Rukwa  |    2.40628   2.125091     1.13   0.258     -1.75882    6.571381
       Kigoma  |   .3052764   1.887672     0.16   0.872    -3.394494    4.005047
    Shinyanga  |   4.722757   1.959367     2.41   0.016     .8824689    8.563045
       Kagera  |  -2.374264   2.212152    -1.07   0.283    -6.710003    1.961474
       Mwanza  |  -2.897224   1.886793    -1.54   0.125     -6.59527    .8008224
         Mara  |  -3.089262    1.88827    -1.64   0.102    -6.790203    .6116789
      Manyara  |   2.272483   1.683396     1.35   0.177    -1.026913    5.571879
       Njombe  |   .6088471   2.994191     0.20   0.839     -5.25966    6.477354
       Katavi  |  -6.427104   2.811699    -2.29   0.022    -11.93793    -.916275
       Simiyu  |  -4.190653   2.369187    -1.77   0.077    -8.834175    .4528682
        Geita  |  -1.726489   2.983888    -0.58   0.563    -7.574802    4.121824
       Songwe  |  -.8637094    2.45628    -0.35   0.725     -5.67793    3.950511
               |
        sector |
            2  |  -1.271559   .8749467    -1.45   0.146    -2.986423    .4433053
            3  |  -4.757185   2.458771    -1.93   0.053    -9.576288    .0619184
            4  |  -1.860722    1.44389    -1.29   0.198    -4.690695     .969251
               |
         _cons |  -17.78459   2.819769    -6.31   0.000    -23.31124   -12.25794
---------------+----------------------------------------------------------------
      /lnsig2u |    4.13783   .1610976                      3.822085    4.453576
---------------+----------------------------------------------------------------
       sigma_u |    7.91623    .637643                      6.760132    9.270042
           rho |   .9501207   .0076346                      .9328451    .9631278
--------------------------------------------------------------------------------
LR test of rho=0: chibar2(01) = 1052.64                Prob >= chibar2 = 0.000

I have posted output for one type of linkage that is technology linkage (tech1) although the problem I have described has occurred with the rest of linkages (competition, forward and backward linkage) too.

The panel data set is confidential so I can not share it here. It is collected from firms in the industrial sector annually between 2008-2016 except for 2013. My concern is the size of the significant constant term (-17.78) which I find to be abit too large compared to other coefficients. In other cases (competition, forward and backward) the constant estimate goes up to 40. Could it be that I have not included some important variables or is it an issue with this model??

NB: If anyone can flag any other issue from that output please let me know.

I am still new with panel data analysis but I hope, i'll be able to follow well the replies