Propensity scores & matching pscore r(2000)

[Denise Vella]

Hi I'm practicing using propensity scores and propensity score matching on dummy date before using it on my real data.

However when I come to use pscore, the following error comes up, could you let me know why ? I checked and the data is binary 0 and 1.
//Estimation of the propensity score

outcome = msmoke > 0 predicts data perfectly
r(2000);

end of do-file

with regards to data:
I used this data:

webuse cattaneo2

drop mmarried mhisp fhisp foreign alcohol deadkids fage fedu nprenatal monthslb mrace frace prenatal birthmonth lbweight fbaby prenatal1

Therefore remaining with the following variables:
bweight mage medu order msmoke mbsmoke

//Creating macros

global treatment mbsmoke //intervention/treatment

. global ylist bweight //outcome

. global xlist mage medu msmoke //variables to match on t


//Creating a regression but now controlling for cofounding
. reg $ylist $treatment $xlist


//Creating a regression but now controlling for cofounding
. reg $ylist $treatment $xlist

MY problem lies here:

//Estimate propensity scores
. pscore $treatment $xlist,pscore(myscore)blockid(myblock)comsup

[David Radwin]

The user-written program pscore (Becker & Ichino, available from net install st0026_2.pkg) uses probit (default) or logit to estimate propensity scores. But neither estimation method can accommodate covariates (msmoke) that perfectly predict the outcome (mbsmoke). Substantively, that is because mbsmoke = 0 if msmoke = 0 and mbsmoke = 1 otherwise.

So you can't use msmoke as a covariate in this situation. The following may help show why.

. tab mbsmoke msmoke

      1 if |
    mother |     Cigarettes smoked during pregnancy
    smoked |   0 daily  1–5 daily  6–10 dail  11+ daily |     Total
-----------+--------------------------------------------+----------
 Nonsmoker |     3,778          0          0          0 |     3,778 
    Smoker |         0        200        337        327 |       864 
-----------+--------------------------------------------+----------
     Total |     3,778        200        337        327 |     4,642 

. logit mbsmoke msmoke 

outcome = msmoke > 0 predicts data perfectly
r(2000);

. probit mbsmoke msmoke 

outcome = msmoke > 0 predicts data perfectly
r(2000);