2SLS , if the first stage is probit model and the second stage is probit model

[hussein bataineh]

Hi please i need to apply 2SLS where my first stage is probit and the second stage is probit . i used these code and im not sure if they are correct


asdoc ivregress 2sls hard_final_Exact_new Firm_Size_w ROA_w Leverage_w Market_book_four_w Non_pension_CFO_w STD_CFO_w Board_Independence_w BoardSize_w Gender_Diversity_w Fund_Status_w FUNDING_RATIO_w Platn_Size_w CSR_Committee SustainabilityScore_w i.year i.ff_12 (csopresence1 = CSO_Percentage) , first robust cluster (id) replace nest drop( i.year i.ff_12 ) dec(4) save(msss)


estat first, forcenonrobust
estat endogenous csopresence1

[Harald Tauchmann]

Dear Hussein,
I can't say much about your code without knowing anything about your data. However, when you use ivregress 2sls, you are not estimating a probit model, but a linear probability model. If both hard_final_Exact_new Firm_Size_w and csopresence1 are binary (this is what I think you mean by 'first and second stage is probit') and you want to estimate a probit model using a two-step procedure, then control function estimation is probably what you should be using. This is implemented by cfprobit (official Stata, available for StataNow users only) and cfbinout probit (available from ssc). ivprobit (without option twostep) implements joint ML estimation. Best wishes, Harald

[Harald Tauchmann]

Sorry, you should use biprobit instead of ivprobit to estimate the model by joint maximum likelihood.

[George Ford]

biprobit will work, or cfprobit if you have v18.

[hussein bataineh]

hi thanks for help actually my hard_final_Exact_new and csopresence1 are binary

are using ivprobit correct as it good results

does using these command is right then

ivprobit hard_final_Exact_new Firm_Size_w ROA_w Leverage_w Market_book_four_w Non_pension_CFO_w STD_CFO_w Board_Independence_w BoardSize_w Gender_Diversity_w Fund_Status_w FUNDING_RATIO_w Platn_Size_ CSR_Committee SustainabilityScore_w i.year i.ff_12 (csopresence1 = CSO_Percentage), twostep

[George Ford]

help ivprobit ("Both estimators assume that the endogenous covariates are continuous and so are not appropriate for use with discrete endogenous covariates.")

[hussein bataineh]

please i run biprobit . I'm not sure ,how we can present first stage and related tests such as Endogeneity test and F-statistics .

. biprobit (hard_final_Exact_new = Firm_Size_w ROA_w Leverage_w Market_book_four_w Non_pension_CFO_w STD_CFO_w Board_Independence_w BoardSize_w Gender_Diversity_w Fund_Status_w FUNDING_RATIO_w Platn_Size_ CSR_Committee Sustain
> abilityScore_w i.year i.ff_12 csopresence1)  (csopresence1 = CSO_Percentage), robust

Fitting comparison equation 1:

Iteration 0:  Log pseudolikelihood = -358.54602  
Iteration 1:  Log pseudolikelihood =   -329.264  
Iteration 2:  Log pseudolikelihood = -326.78258  
Iteration 3:  Log pseudolikelihood = -326.74249  
Iteration 4:  Log pseudolikelihood = -326.74237  
Iteration 5:  Log pseudolikelihood = -326.74237  

Fitting comparison equation 2:

Iteration 0:  Log pseudolikelihood = -1925.9727  
Iteration 1:  Log pseudolikelihood = -1705.2604  
Iteration 2:  Log pseudolikelihood = -1704.8319  
Iteration 3:  Log pseudolikelihood = -1704.8319  

Comparison:   Log pseudolikelihood = -2031.5742

Fitting full model:

Iteration 0:  Log pseudolikelihood = -2031.5742  (not concave)
Iteration 1:  Log pseudolikelihood = -2031.5152  (backed up)
Iteration 2:  Log pseudolikelihood = -2030.9562  
Iteration 3:  Log pseudolikelihood = -2030.1775  
Iteration 4:  Log pseudolikelihood = -2030.1553  
Iteration 5:  Log pseudolikelihood = -2030.1553  

Seemingly unrelated bivariate probit                    Number of obs =  3,167
                                                        Wald chi2(43) = 563.40
Log pseudolikelihood = -2030.1553                       Prob > chi2   = 0.0000

---------------------------------------------------------------------------------------
                      |               Robust
                      | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
----------------------+----------------------------------------------------------------
hard_final_Exact_new  |
          Firm_Size_w |  -.1346719   .0745992    -1.81   0.071    -.2808836    .0115398
                ROA_w |   1.946918   1.329454     1.46   0.143    -.6587632      4.5526
           Leverage_w |  -.7282058    .378328    -1.92   0.054    -1.469715    .0133033
   Market_book_four_w |  -.0006798   .0085561    -0.08   0.937    -.0174495    .0160899
    Non_pension_CFO_w |   -4.35556   1.799502    -2.42   0.016    -7.882518   -.8286009
            STD_CFO_w |   .8498084   2.625005     0.32   0.746    -4.295107    5.994723
 Board_Independence_w |   .0038096   .0050628     0.75   0.452    -.0061134    .0137326
          BoardSize_w |  -.0212048   .0238728    -0.89   0.374    -.0679947    .0255851
   Gender_Diversity_w |  -.0038736    .006345    -0.61   0.542    -.0163096    .0085624
        Fund_Status_w |  -3.688263   1.946652    -1.89   0.058     -7.50363    .1271043
      FUNDING_RATIO_w |   -.548908   .3974585    -1.38   0.167    -1.327912    .2300963
         Platn_Size_w |    .029275   .0661141     0.44   0.658    -.1003064    .1588563
        CSR_Committee |   .1171023    .126457     0.93   0.354    -.1307489    .3649535
SustainabilityScore_w |   .0023581   .0034981     0.67   0.500     -.004498    .0092143
                      |
                 year |
                2007  |  -.1893796   .4824649    -0.39   0.695    -1.134993    .7562342
                2008  |   .3703923   .4019114     0.92   0.357    -.4173396    1.158124
                2009  |   .5153293   .3911218     1.32   0.188    -.2512553    1.281914
                2010  |   .3656944   .4108336     0.89   0.373    -.4395247    1.170914
                2011  |   .0708561    .428661     0.17   0.869    -.7693042    .9110163
                2012  |   .3657888   .4044102     0.90   0.366    -.4268405    1.158418
                2013  |   .2803688   .4104458     0.68   0.495    -.5240901    1.084828
                2014  |    .360477   .4095485     0.88   0.379    -.4422233    1.163177
                2015  |   .3537351   .4137397     0.85   0.393    -.4571797     1.16465
                2016  |   .0784589    .423008     0.19   0.853    -.7506216    .9075393
                2017  |   .1262666   .4213481     0.30   0.764    -.6995605    .9520937
                2018  |   .3577145   .4153034     0.86   0.389    -.4562652    1.171694
                2019  |  -.0183752   .4474583    -0.04   0.967    -.8953773    .8586269
                2020  |   .2577183   .4336884     0.59   0.552    -.5922953    1.107732
                2021  |  -.1472962   .4668719    -0.32   0.752    -1.062348    .7677558
                2022  |   .1538622    .541226     0.28   0.776    -.9069212    1.214646
                      |
                ff_12 |
                   2  |   .1149295   .3525047     0.33   0.744     -.575967     .805826
                   3  |   .0547361   .2282032     0.24   0.810    -.3925339    .5020061
                   4  |   .0371616   .3826385     0.10   0.923     -.712796    .7871192
                   5  |  -.0179077   .2321963    -0.08   0.939    -.4730041    .4371887
                   6  |   .2384004    .231927     1.03   0.304    -.2161682     .692969
                   7  |   .6368554   .3566165     1.79   0.074       -.0621    1.335811
                   8  |   .0149633    .273617     0.05   0.956    -.5213161    .5512428
                   9  |   .6683979   .2549403     2.62   0.009     .1687241    1.168072
                  10  |   .1631866   .2537802     0.64   0.520    -.3342135    .6605867
                  11  |   .6306283    .234109     2.69   0.007     .1717832    1.089474
                  12  |   .1210401   .2556985     0.47   0.636    -.3801198       .6222
                      |
         csopresence1 |   1.244748   .4130107     3.01   0.003     .4352615    2.054234
                _cons |  -1.053993   .6796596    -1.55   0.121    -2.386101    .2781156
----------------------+----------------------------------------------------------------
csopresence1          |
       CSO_Percentage |   2.942302   .1429113    20.59   0.000     2.662201    3.222403
                _cons |  -1.364117   .0474742   -28.73   0.000    -1.457165   -1.271069
----------------------+----------------------------------------------------------------
              /athrho |   -.581801     .25471    -2.28   0.022    -1.081023   -.0825787
----------------------+----------------------------------------------------------------
                  rho |  -.5239732   .1847799                     -.7935783   -.0823915
---------------------------------------------------------------------------------------
Wald test of rho=0: chi2(1) = 5.21744                     Prob > chi2 = 0.0224

[George Ford]

Here's a shot, though I may be corrected.

At present, I don't think non-linear models have the standard set of tests for instruments. The last line indicates endogeneity. With only one exclusion, you wouldn't get the Sargan test in any case.

In IV linear regression, the rule of thumb is F>10. On the exclusion, you have chi2 = 20.59^2 = 423.95 [or testparm CSO_Percentage, equation(csopresence1)], so the "equivalent F" would be (1+1)*423.95 > 10, so passes easily.

You might think about some of the Xs being in the second equation (e.g., the FE).

[George Ford]

Actually, you want chi2 > 20. (k+1)*10.

[hussein bataineh]

Thanks professor George Ford I'm not sure how can i present all results of the first stage

[George Ford]

what results are you talking about.

[hussein bataineh]

What I mean I want to present the whole regression for the first stage, as I need to present the first stage and the second stage.

[George Ford]

Your first stage is just the csopresence results (CSO_Percentage and _cons). A little thin in my opinion.