How to specify family-level option for psmatch2 (my data is at the child-level)

[Denise Tan]

Hi everyone, first post so pardon me if I'm making a rookie mistake.

I want to compare the difference in children's education based on whether they are adopted or not. My dataset from PSID has adoptive and birth families.

I use psmatch2 to match adoptive families (treated) to birth families (control) based on parent characteristics. Thus, the matching should be done at the family-level. (This is my code, if relevant:

psmatch2 adopted old_mother old_father father_rich mother_zerochildren, outcome(child_collegedummy) neighbor(1) common ate ties

)

However, my data is at the child-level (1 obs per child), hence each familyID can take up several observations, depending on how many children there are per family.

Is there a way to specify in psmatch2 to match at the family-level? I thought of restructuring my dataset to become 1 obs per family but it'll lead to some complications.

Would really appreciate if anyone could advise. Many thanks!

[Mike Lacy]

Given that your interest is in analyzing children, i.e., individuals, I think your problem is conceptual, not computational.

You are not matching at the family level, but rather you are *matching individual children* using family-level variables to characterize each child, such as "child's mother's age, number of children in child's family, etc." In that case, with the variables calculated for each child, there should be no problem in getting psmatch2 to match individuals on these variables.

Now, perhaps you are thinking that all the children in one family should have matches that come from the same other family, or something like that. If so, I don't have any relevant experience, and so won't comment.