st: Multicollinearity

[Cordula Stolberg <cs32@sussex.ac.uk>]

Dear Statalisters,

I have a question related to multicollinearity. I run the following 
regression:

xi: ivreg2 lfrag lfdi (lspec = lexp liit76 liit54) i.dapp_cz i.dapp 
i.dapp_slvk i.dag i.dneg i.dneg_slvk i.dgo i.dcz i.dhun i.dslvk

The problem is that Stata -no matter whether I use -ivreg2-, -reg-, or 
-xtreg- includes the variable "i.dneg_slvk" always twice and then drops it 
to avoid multicollinearity (see below):

Instrumental variables (2SLS) regression
----------------------------------------

                                                     Number of obs =     40
                                                     F( 12,    26) =   2.43
                                                     Prob > F      = 0.0270
Total (centered) SS     =  335.3994166                Centered R2   = 0.3195
Total (uncentered) SS   =  3741.564996                Uncentered R2 = 0.9390
Residual SS             =  228.2362191                Root MSE      = 2.4

---------------------------------------------------------------------------
---
      lfrag |      Coef.   Std. Err.      z    P>|z|     [95% Conf.

-------------+-------------------------------------------------------------
---
      lspec |  -3.313096   5.880005    -0.56   0.573    -14.83769
       lfdi |   64407.86   37018.25     1.74   0.082    -8146.577
_Idapp_cz_1 |  -6.691789   3.708458    -1.80   0.071    -13.96023
   _Idappa1 |   -7.56607   3.602525    -2.10   0.036    -14.62689
_Idapp_slv~1 |  -7.471712   3.489523    -2.14   0.032    -14.31105
    _Idag_1 |  -4.624783   2.560676    -1.81   0.071    -9.643615
   _Idneg_1 |  -3.974766   2.381558    -1.67   0.095    -8.642534
_Idneg_slv~1 |  -2.901205   2.515054    -1.15   0.249     -7.83062
_Idneg_slv~1 |  (dropped)
    _Idgo_1 |  -3.072955   2.099523    -1.46   0.143    -7.187944
    _Idcz_1 |   3.790734   5.606013     0.68   0.499    -7.196849    1
   _Idhun_1 |  -.9256889   2.651573    -0.35   0.727    -6.122677
  _Idslvk_1 |   7.115247   8.029922     0.89   0.376     -8.62311
      _cons |  -246131.1   141481.5    -1.74   0.082    -523429.9    3
---------------------------------------------------------------------------
---
Sargan statistic (overidentification test of all instruments):         1.655
                                                  Chi-sq(1) P-val = 0.19830
---------------------------------------------------------------------------
---
Collinearities detected among instruments: 2 instrument(s) dropped
Instrumented:  lspec
Instruments:   lfdi _Idapp_cz_1 _Idappa1 _Idapp_slvk_1 _Idag_1 _Idneg_1
              _Idneg_slvk_1 _Idneg_slvk_1 _Idgo_1 _Idcz_1 _Idhun_1 
_Idslvk_1
              lexp liit76 liit54
---------------------------------------------------------------------------
---


Does anyone know why Stata does this and how I can avoid it? Any 
suggestions would be much appreciated.

Thanks,
Cordula







*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/