One-to-One Matching in STATA

[Moksh Garg]

Dear community

I intend to do one-to-one matching in STATA based on two variables and a condition. There are firms (each classified as family and non-family using a dummy variable), however family firms occupy 70% of the entire dataset. Therefore, I am trying to create a matched sample using non-family firms as the baseline, findind one-to-one match on certain parameters for subsequent analysis.

The two parameters include - Year and Industry Code. Further, I am including a conditional logic, which says that the net sales differential between the matched pair should not exceed 5 percent.

Using the existing joinby command, I am generating many-to-many matches and that too not unqiue for a given family/non-family firm. Therefore, I wanted someone to advise about how should I proceed with my analysis?

Below is the sample data for you to visualize the problem at hand:

Company Code	Firm	Year	Industry Code	Dummy (Family/Non-Family)	Net Sales
11	ABC	1997	8	1	219
11	PQR	1998	8	1	300.2999878
11	XYZ	1997	8	0	225.8999939

[Tirupam Goel]

You can create a dummy variable based on which you want to do exact matching:
egen exactmatch = group(A B ...)

Then you can create a propensity score (if you want)
logit treat $predictors
predict propscore

Finally you can match based on the propensity score within levels of the exactmatch dummy
gen weight = .
levelsof exactmatch, local(gr)
foreach j of local gr {
psmatch2 treat if exactmatch == `j', kernel pscore(propscore) com cal(0.05)
replace weight = _weight if weight == .
}