LR test not significantly different from zero

[Sergio Ponguane]

Hi everyone

I ran

 heckman saved age female account_fin inc_q emp_in educ, select(account_mob = age female inc_q emp_in educ  account_fin mobileown
> er)

and I got this

Iteration 0:   log likelihood = -731.84756  
Iteration 1:   log likelihood = -731.84703  
Iteration 2:   log likelihood = -731.84703  

Heckman selection model                         Number of obs      =       998
(regression model with sample selection)        Censored obs       =       576
                                                Uncensored obs     =       422

                                                Wald chi2(6)       =     25.04
Log likelihood =  -731.847                      Prob > chi2        =    0.0003

------------------------------------------------------------------------------
             |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
saved        |
         age |  -.0051578   .0019571    -2.64   0.008    -.0089936    -.001322
      female |    .010929   .0427027     0.26   0.798    -.0727668    .0946248
 account_fin |   .1966438   .0570488     3.45   0.001     .0848302    .3084574
       inc_q |  -.0226961   .0187744    -1.21   0.227    -.0594933    .0141011
      emp_in |  -.0806902   .0491241    -1.64   0.100    -.1769717    .0155912
        educ |   .0449475    .036255     1.24   0.215     -.026111    .1160061
       _cons |   .8413716   .1777263     4.73   0.000     .4930345    1.189709
-------------+----------------------------------------------------------------
account_mob  |
         age |   -.015169   .0034509    -4.40   0.000    -.0219327   -.0084053
      female |   .1981766   .0925461     2.14   0.032     .0167896    .3795636
       inc_q |   .1759306   .0332975     5.28   0.000     .1106687    .2411924
      emp_in |  -.3053182   .0973598    -3.14   0.002      -.49614   -.1144965
        educ |   .2918701   .0637277     4.58   0.000     .1669661     .416774
 account_fin |   .5520342   .0966461     5.71   0.000     .3626113    .7414572
 mobileowner |  -.9522857   .1073431    -8.87   0.000    -1.162674    -.741897
       _cons |   .2710259    .320518     0.85   0.398    -.3571779    .8992296
-------------+----------------------------------------------------------------
     /athrho |   .0130454    .192169     0.07   0.946    -.3635989    .3896896
    /lnsigma |  -.8678992   .0344477   -25.19   0.000    -.9354155   -.8003829
-------------+----------------------------------------------------------------
         rho |   .0130446   .1921363                     -.3483801    .3710926
       sigma |   .4198326   .0144623                      .3924228    .4491569
      lambda |   .0054766   .0806727                      -.152639    .1635921
------------------------------------------------------------------------------
LR test of indep. eqns. (rho = 0):   chi2(1) =     0.00   Prob > chi2 = 0.9463

I am concerned with the LR test of indep.eqns which turns NOT to be significantly different from zero, because I fail to reject the null hypothesis. Does this mean that there is no self selection bias? Or can I assume that the selection and outcome equations are independent?

[Andrew Musau]

Does this mean that there is no self selection bias?

Assuming that your exclusion restriction is valid, the implication is so. You can run OLS on the continuous sample and obtain unbiased results.

[Sergio Ponguane]

Many thanks Andrew Musau