I'm trying to estimate the AT on my treated group using psmatch2. I don't get why the probit regression Stata runs at the beginning of the command psmatch2 is different from the one I run manually, on the same set of covariates...

Code:
probit in_program $parsimonius if followup == 0 & treatment == 1;
Output

Code:
Probit regression                               Number of obs     =        149
                                                LR chi2(6)        =      25.08
                                                Prob > chi2       =     0.0003
Log likelihood = -90.331981                     Pseudo R2         =     0.1219

------------------------------------------------------------------------------------------
              in_program |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------------------+----------------------------------------------------------------
  ran_total_assets_value |   1.01e-08   1.24e-08     0.81   0.418    -1.43e-08    3.45e-08
     ran_schooling_years |   .0587416   .0331389     1.77   0.076    -.0062094    .1236926
        ran_age_business |   .0390573   .0138214     2.83   0.005     .0119678    .0661468
                 ran_age |  -.0171452   .0107971    -1.59   0.112    -.0383071    .0040167
        ran_indigenous_d |          0  (omitted)
     ran_profit_margin_d |   1.331083   .4680356     2.84   0.004     .4137504    2.248416
ran_total_assets_value_d |  -1.147386   .5011412    -2.29   0.022    -2.129605   -.1651678
                   _cons |  -.6480671   .7562728    -0.86   0.391    -2.130334    .8342002
------------------------------------------------------------------------------------------
And

Code:
psmatch2 in_program $parsimonius if followup == 0 & treatment == 1, kernel k(epan)
         out(prodt1);
Output:

Code:
Probit regression                               Number of obs     =         86
                                                LR chi2(5)        =      14.07
                                                Prob > chi2       =     0.0151
Log likelihood = -52.573699                     Pseudo R2         =     0.1180

------------------------------------------------------------------------------------------
              in_program |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------------------+----------------------------------------------------------------
  ran_total_assets_value |   5.42e-08   3.90e-08     1.39   0.164    -2.22e-08    1.31e-07
     ran_schooling_years |   -.000173   .0427222    -0.00   0.997    -.0839071     .083561
        ran_age_business |   .0375623   .0176681     2.13   0.034     .0029335    .0721911
                 ran_age |  -.0221552   .0167707    -1.32   0.186     -.055025    .0107147
        ran_indigenous_d |          0  (omitted)
     ran_profit_margin_d |          0  (omitted)
ran_total_assets_value_d |  -1.094862   .7255817    -1.51   0.131    -2.516976    .3272519
                   _cons |    .419883   1.088111     0.39   0.700    -1.712775    2.552541
------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------
        Variable     Sample |    Treated     Controls   Difference         S.E.   T-stat
----------------------------+-----------------------------------------------------------
          prodt1  Unmatched | .498623773  -.082308616   .580932389   .194825868     2.98
                        ATT | .483210399  -.057348028   .540558426   .222870184     2.43
----------------------------+-----------------------------------------------------------