Hi all,
I am trying to come up with control firms for treatment firms in the sample. It is a cross-sectional data, and I want one control firm (matched within the same industry code and closest propensity score) for each treatment firm. So, one-on-one matching for a treatment-control pair.
If firm B can act as a control firm for firm A (based on the criteria, i.e. same industry code and closest propensity), then it should not act as a control for any other firm. However, the next best alternative should be used as a control firm, i.e. within the same industry code, but the next best closest propensity score.
firm_id Treatment_indicator Industry_code Pscore
1111 0 11 0.35
1112 1 11 0.44
1113 0 12 0.60
1114 0 15 0.80
1115 1 17 0.56
After creating two separate files for treatment and control firms, I am using this code (I got it from previous posts):
data treatment_firms conrol_firms;
set have;
if treatment_indicator = 0 then output control_firms;
if treatment_indicator = 1 then output treatment_firms;
run;
proc sql;
create table like_each_other as
select O.*, T.firm_id as Tfirm_id, abs(O.pscore-T.pscore) as P_Diff
from control_firms as O inner join treatment_firms as T
on o.industry_code=t.industry_code
order by tfirm_id;
quit;
proc means data=like_each_other noprint;
by Tfirm_id;
output out=close_match idgroup(min(P_diff) out[1] (firm_id)=);
run;
Can you please help in coming up with the correct codes? These codes are working on 'with replacement' basis, so, one control firm is chosen as a 'control firm' for another treatment firm as well. I want one firm to act as a control for only one treatment firm.
Regards,
Aman
