Announcement

Collapse
No announcement yet.
X
Collapse
  • Page of 1
  • Filter
  • Time
  • Show
Clear All
new posts
  • #1

    Endogeneity test in probit model

    Hope you are all doing well.

    I am doing a probit model and I want to check whether there is any endogeneity in my model. The basic regression model that I have run is as follows:


    [probit vocationaltraining hhsize gender literacy uneduc_adults mfs_use loan totalsavings remitt totalassets mobile1 barisal chittagong khulna rajshahi rangpur sylhet members_in_agri agrishock distfacility3, robust
    margins, dydx(*) post

    outreg2 using F:\UDOY\Stata\405_AE1_Paper\AE1resultsSca, word]

    My instructor suggested me that the variable remitt (it means remittance) might be endogenous. So, she told me to check that. Here remitt is a binary variable which is 1 if the HH receives remittance and 0 otherwise.

    Now I want to do the Hausman Endogeneity test. What is the code to do that and how to understand the result of Hausman Endogeneity test whether endogeneity exists or not


    Thank You all.
    Tags: None
  • #2
    You're gonna need an instrument or two to do any test (the hard part).

    Could do ivprobit and it will provide a test of exogeneity.
    Could use ivreg2 and estimate a linear probability model and use that for a test too.

    Also,
    Code:
    search probexog





    Comment

    • #3
      George Ford Thank you for your suggestion. Actually I am not able to produce any instrument till now. I am trying.

      I have tried the command ivprobit:

      Code:
      ivprobit vocationaltraining hhsize gender literacy uneduc_adults mfs_use loan totalsavings remitt totalassets mobile1 barisal chittagong khulna rajshahi rangpur sylhet members_in_agri agrishock distfacility3, robust
      The outcome message shows in red mark that:
      Code:
      no endogenous variables; use probit instead.
      Would you please explain what does this actually mean?

      I also tried the the ivreg2:

      Code:
      ivreg2 vocationaltraining hhsize gender literacy uneduc_adults mfs_use loan totalsavings remitt totalassets mobile1 barisal chittagong khulna rajshahi rangpur sylhet members_in_agri agrishock distfacility3, robust
      The outcome shows:

      Code:
      OLS estimation
--------------

Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity

                                                      Number of obs =    18366
                                                      F( 19, 18346) =    16.95
                                                      Prob > F      =   0.0000
Total (centered) SS     =  1329.624796                Centered R2   =   0.0201
Total (uncentered) SS   =         1443                Uncentered R2 =   0.0971
Residual SS             =  1302.939403                Root MSE      =    .2664

---------------------------------------------------------------------------------
                |               Robust
vocationaltra~g |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
----------------+----------------------------------------------------------------
         hhsize |  -.0029057   .0012877    -2.26   0.024    -.0054297   -.0003818
         gender |  -.0165139   .0063534    -2.60   0.009    -.0289663   -.0040616
       literacy |   .0004154   .0021108     0.20   0.844    -.0037216    .0045524
  uneduc_adults |   -.014285   .0027984    -5.10   0.000    -.0197698   -.0088003
        mfs_use |   .0104733   .0041394     2.53   0.011     .0023603    .0185863
           loan |    .007121   .0044495     1.60   0.110    -.0015998    .0158417
   totalsavings |  -3.93e-09   1.22e-08    -0.32   0.746    -2.78e-08    1.99e-08
         remitt |  -.0081815   .0045388    -1.80   0.071    -.0170774    .0007144
    totalassets |   1.86e-10   7.02e-10     0.27   0.791    -1.19e-09    1.56e-09
        mobile1 |   .0199803   .0023931     8.35   0.000     .0152899    .0246707
        barisal |  -.0105475   .0079833    -1.32   0.186    -.0261944    .0050995
     chittagong |   -.027286   .0057229    -4.77   0.000    -.0385027   -.0160693
         khulna |  -.0036871   .0072383    -0.51   0.610    -.0178739    .0104997
       rajshahi |   .0106332   .0072828     1.46   0.144    -.0036407    .0249072
        rangpur |  -.0000258   .0073128    -0.00   0.997    -.0143586     .014307
         sylhet |   .0249522   .0069462     3.59   0.000     .0113379    .0385666
members_in_agri |  -.0300347   .0040368    -7.44   0.000    -.0379467   -.0221226
      agrishock |   .0139598    .006699     2.08   0.037     .0008299    .0270897
  distfacility3 |   .0001684   .0007209     0.23   0.815    -.0012446    .0015814
          _cons |   .0862148   .0106432     8.10   0.000     .0653545    .1070751
---------------------------------------------------------------------------------
Included instruments: hhsize gender literacy uneduc_adults mfs_use loan
                      totalsavings remitt totalassets mobile1 barisal chittagong
                      khulna rajshahi rangpur sylhet members_in_agri agrishock
                      distfacility3
------------------------------------------------------------------------------
      Would you please explain me in detail whether my regression has any endogeneity base on these results. Thanks for your cooperation

      Comment

      • #4
        You need an instrument, something correlated with remitt but that does not affect the dependent variable except through remitt.

        To see it work, try this

        Code:
         
 ivprobit vocationaltraining hhsize gender literacy uneduc_adults mfs_use loan mobile1 barisal chittagong khulna rajshahi rangpur sylhet members_in_agri agrishock distfacility3 (remitt = totalsavings totalassets), robust
        It's not legit (presumably), but will provide a result.

        Comment

        • #5
          This discussion uses IV for remittances.

          HTML Code:
          https://www.wider.unu.edu/sites/default/files/Events/PDF/Slides/MigrationConf-slides-Agwu.pdf

          Comment

          • #6
            I am able to find an instrument. I will like to use ownership of nonfarm activities as an instrument for remitt. My instrumental variable is nonfarm (where nonfarm= 1 means HH owns any nonfarm activities and 0 otherwise). Now, what is the command to run the probit model with instrument.

            Comment

            • #7
              I have used the following command as you mentioned earlier

              Code:
              ivprobit vocationaltraining hhsize gender literacy uneduc_adults mfs_use loan scatosavings scatoassets mobile1 barisal chittagong khulna rajshahi rangpur sylhet members_in_agri agrishock distfacility3 (remitt = nonfarm)
              The output is:

              Code:
              Fitting exogenous probit model

Iteration 0:   log likelihood = -5055.4365  
Iteration 1:   log likelihood = -4866.5472  
Iteration 2:   log likelihood = -4863.6472  
Iteration 3:   log likelihood =  -4863.646  
Iteration 4:   log likelihood =  -4863.646  

Fitting full model

Iteration 0:   log likelihood = -15261.642  
Iteration 1:   log likelihood = -15261.636  
Iteration 2:   log likelihood = -15261.636  

Probit model with endogenous regressors         Number of obs     =     18,366
                                                Wald chi2(19)     =     639.41
Log likelihood = -15261.636                     Prob > chi2       =     0.0000

-----------------------------------------------------------------------------------------------------
                                    |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
------------------------------------+----------------------------------------------------------------
                             remitt |   -1.28141   .1897747    -6.75   0.000    -1.653362   -.9094587
                             hhsize |  -.0279564   .0082156    -3.40   0.001    -.0440587   -.0118541
                             gender |  -.6905124   .0935798    -7.38   0.000    -.8739254   -.5070994
                           literacy |   .0020287    .014152     0.14   0.886    -.0257087    .0297661
                      uneduc_adults |  -.0774123   .0206575    -3.75   0.000    -.1179003   -.0369242
                            mfs_use |   .1333084   .0269564     4.95   0.000     .0804749     .186142
                               loan |  -.0089358   .0309157    -0.29   0.773    -.0695294    .0516577
                       scatosavings |  -.0001013   .0000806    -1.26   0.209    -.0002593    .0000566
                        scatoassets |   .0000142   4.15e-06     3.42   0.001     6.07e-06    .0000223
                            mobile1 |   .1266633   .0122999    10.30   0.000     .1025558    .1507707
                            barisal |   .0330197   .0558682     0.59   0.555      -.07648    .1425194
                         chittagong |  -.0158528   .0533031    -0.30   0.766    -.1203249    .0886194
                             khulna |  -.1027929   .0479139    -2.15   0.032    -.1967025   -.0088833
                           rajshahi |  -.0473593   .0483026    -0.98   0.327    -.1420306    .0473121
                            rangpur |  -.0878123   .0490751    -1.79   0.074    -.1839976    .0083731
                             sylhet |   .0757864   .0410683     1.85   0.065     -.004706    .1562787
                    members_in_agri |  -.1568515   .0299381    -5.24   0.000    -.2155291    -.098174
                          agrishock |   .0538937   .0397037     1.36   0.175    -.0239241    .1317114
                      distfacility3 |   .0002176   .0028504     0.08   0.939    -.0053692    .0058044
                              _cons |  -.2700579    .219464    -1.23   0.218    -.7001994    .1600836
------------------------------------+----------------------------------------------------------------
 corr(e.remitt,e.vocationaltraining)|   .5293574   .0823616                       .349594     .671529
                        sd(e.remitt)|   .4262266   .0022239                        .42189    .4306077
-----------------------------------------------------------------------------------------------------
Instrumented:  remitt
Instruments:   hhsize gender literacy uneduc_adults mfs_use loan scatosavings scatoassets mobile1
               barisal chittagong khulna rajshahi rangpur sylhet members_in_agri agrishock
               distfacility3 nonfarm
-----------------------------------------------------------------------------------------------------
Wald test of exogeneity (corr = 0): chi2(1) = 26.52       Prob > chi2 = 0.0000
              What does it mean?
              One more thing to know, why can't I use robust option at the end while using instrument

              Comment

              • #8
                Using nonfarm is consistent with what others have done. You reject exogeneity, so endogenous. That result, however, depends on what you use as instruments.

                Do these results make sense to you?

                Comment

                • #9
                  use , vce(robust)

                  Comment

                  • #10
                    You'll need to test your instrument. Read this.
                    HTML Code:
                    https://www.stata-journal.com/sjpdf.html?articlenum=st0171
                    For starters, run the model with the -twostep- option and not robust, then run -weakiv-. You don't want to reject.

                    I'm sure there are a variety of instrument tests available. (rivtest, for example).

                    Comment

                    • #11
                      Originally posted by George Ford View Post
                      Using nonfarm is consistent with what others have done. You reject exogeneity, so endogenous. That result, however, depends on what you use as instruments.

                      Do these results make sense to you?
                      Hi, I wanted to know if it is endogenous (since we reject the null of the Wald test) then doing IV is the right thing? and then we follow up with weakivtest?

                      Comment

                      • #12
                        yes. if endogenous, you must do something.

                        Comment

                        Previous Next
                        Working...
                        X