Instrumental-variables regression automatically includes too many instruments

[Tom regt]

Dear Members,

I would like to perform an instrumental-variables regression in which I would like to use LOGameanPOP, LOG2ameanPOP, LOGameanGDP and LOG2ameanGDP as instruments for the endogenous variables LOGhhi and LOGroutetotalpassengers. However, when I perform the regression and look at the results, I notice that the exogenous variables (LCCdummy, ULCCdummy, LOGLCCmshare and LOGULCCmshare) are also included as instruments. I do not why this happens or how I could solve this. I am using Stata/MP 14 (64-bit). The results are shown below:

xtivreg lwaprice LCCdummy ULCCdummy LOGLCCmshare LOGULCCmshare (LOGhhi LOGroutetotalpassengers = LOGameanPOP LOG2ameanPOP LOGameanGDP LOG2ameanGDP), fe vce(robust)

Fixed-effects (within) IV regression            Number of obs     =     35,711
Group variable: id                              Number of groups  =      1,843

R-sq:                                           Obs per group:
     within  =      .                                         min =          1
     between = 0.1438                                         avg =       19.4
     overall = 0.1079                                         max =         40


                                                Wald chi2(6)      =      23.52
corr(u_i, Xb)  = -0.9406                        Prob > chi2       =     0.0006

                                            (Std. Err. adjusted for 1,843 clusters in id)
-----------------------------------------------------------------------------------------
                        |               Robust
               lwaprice |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
------------------------+----------------------------------------------------------------
                 LOGhhi |  -.2798624   .4070153    -0.69   0.492    -1.077598    .5178729
LOGroutetotalpassengers |   1.540179   .2670391     5.77   0.000     1.016792    2.063566
               LCCdummy |   .4970906     .11264     4.41   0.000     .2763202     .717861
              ULCCdummy |   .3053469   .1067737     2.86   0.004     .0960744    .5146195
           LOGLCCmshare |  -.2274299   .0558618    -4.07   0.000    -.3369171   -.1179427
          LOGULCCmshare |  -.1975084   .0697084    -2.83   0.005    -.3341343   -.0608825
                  _cons |  -9.886094   3.805169    -2.60   0.009    -17.34409   -2.428099
------------------------+----------------------------------------------------------------
                sigma_u |  2.2243066
                sigma_e |  .60856989
                    rho |  .93035644   (fraction of variance due to u_i)
-----------------------------------------------------------------------------------------
Instrumented:   LOGhhi LOGroutetotalpassengers
Instruments:    LCCdummy ULCCdummy LOGLCCmshare LOGULCCmshare LOGameanPOP LOG2ameanPOP
                LOGameanGDP LOG2ameanGDP

Is there anyone who can help me out?
Kind regards,
Tom

[Sebastian Kripfganz]

Exogenous variables are instrumented by themselves. They are always included in the set of instruments. Please consult any introductory econometrics textbook for further information.