Matching Similar Observations

[Philipp Luig]

Hello,

I want to run a difference in difference model and need to construct a control group. I have approximately 200 observations for the treatment group (companies) and I want to match each of them with the most similar company from a different group of companies. I want to do so by finding the company that is most similar with regard to a few key metrics, industry code, total assets, number of employees etc. Is there an easy method how I can perform this matching process, without doing so manually for each observation, or would you recommend an entirely different approach to constructing the control group?

Many thanks in advance.

Best regards,

Philipp

[JJ Kovach]

Have you looked into the -teffects or -psmatch or -psmatch2 commands? I have used -teffects for multiple studies to identify a control group for a treatment group based on similar pre-treatment characteristics.

[George Ford]

cem using a k2k match would work. (ssc install cem)

kmatch offer several options and is faster than cem. a nn match might be sensible given your task.