Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

 From "Clifton Chow" To statalist@hsphsun2.harvard.edu Subject st: st: Biprobit for Endogenous Binary Regressor-Opposite Sign Interpretation? Date Thu, 14 Jun 2012 00:20:57 -0500

```I am wondering if anyone has run a biprobit to correct for an endogenous binary regressor and got an opposite sign from the same regressor in an ordinary probit?  My main equation is a probit predicting Work(LFP).  However, one of my regressors, a policy instrument that is binary, is endogenous, so I ran a biprobit with 2 instruments (Ivprobit is not advisable when the endogenous regressor is binary).  As you can see below, the coefficient on XBinEndog is negative, but if I treated it as an exogenous binary variable in an ordinary probit, it is positive as hypothesized.  I am concerned as the Wald test on rho is significant and not sure how the switch in sign can be interpreted.

Any suggestions would be great.  I have checked the archives on biprobit, and I can't find any similar posting.

Thanks

Main Probit Eq.
Work        |
X1         |  -.3708311   .1433871    -2.59   0.010    -.6518647   -.0897974
X2         |   -.174675   .1370149    -1.27   0.202    -.4432192    .0938692
X3         | -2653913   .1184114    -2.24   0.025    -.4974733   -.0333093
XBinEndg  |  -.8087503   .3468027    -2.33   0.020    -1.488471   -.1290294
-----------------------------------------------------------------------------------------------------------------
XBinEndg   |
Z1           |   1.204996   .2139061     5.63   0.000     .7857481    1.624245
Z2           |   .6297623   .1320375     4.77   0.000     .3709736    .8885509
_cons |  -1.171011   .0843702   -13.88   0.000    -1.336374   -1.005649
-------------+----------------------------------------------------------------
/athrho |   .5757204   .2832209     2.03   0.042     .0206177    1.130823
-------------+----------------------------------------------------------------
rho |    .519548    .206771                      .0206148    .8113008
------------------------------------------------------------------------------
Wald test of rho=0:                 chi2(1) =  4.13212    Prob > chi2 = 0.0421

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```