Beginners help needed - How to run Nearest Neighbor Matching (nnmatch) without outcome variable value?

[Chris Hulstenberg]

Hello,

At first, I am very beginner of using Stata.

I am currently working on my thesis where I will use the Diff-in-Diff method to analyze whether specific publications (treatment) affect stock prices of firms which are named in these publications.

At first, I need to find matches. I thought of Nearest neighbor matching (nnmatch). I prepared my dataset as follow:

- Treatment: Treated (t=1) n=16 , Untreated (t=0) n > 45000
- ISIN = Identifier
- Matching variables: Industry, Total assets, FTE (Full-Time Employees), RoA (Return on assets)

My procedure:
1) I generated numeric values for the 4 matching variables: e.g., encode Industry, generate (Industry_n)


2) I used "teffects nnmatch (Industry_n Total_assets_n FTE_n RoA_n) (Treatment), biasadj(Total_assets_n FTE_n RoA_n) vce(robust)"


Unfortunately, this line of code always crashes my Stata as it breaks down.


In general, I have to main questions:
1) In literature, it is always ref. to the average treatment effects. I do not have the values of my outcome variables yet. I thought I could find matches solely based on matching variables. I have a sample of 45k possible control matching firms. I cannot calculate all of them. So, do I need a outcome variable to succesfully find matches?
2) If not, how do I need to adjust my code in order to get matches regardless of 1-to-1 or 1-to-many?

I really appreciate your support as I am very new to Stata and perhaps, I miss some basic understand.

So, Thank you very, very much!!!
Chris

[Mike Lacy]

If you want just to identify matching observations, you could install and use one of the many community-contributed Stata packages for matching. Here are two that might work for you:

ssc describe ultimatch
ssc describe kmatch

If you're not familiar with using community-contributed packages in Stata, try -help ssc install- and click on "View complete PDF manual entry."

[Chris Hulstenberg]

That's great! Thank you, Mike!

I used the following code:
ultimatch industry_nb assets_nb fte_nb roa_nb, treated(treatment) exact(industry_nb) report (industry_nb assets_nb fte_nb roa_nb)


But I am struggling to find an output which states: Treatment observation 1 ----> (matched) Contro, observation 1

Could you also help here? I think it should only be an attend or short line of code.

Thank you very much!

[Chris Hulstenberg]

Or does anyone else can help me here?

I only need to display the matching pairs but I fail to find the right line or part of code.

I am very thankful for a short notice/support!!

Thank you, Chris

[Mike Lacy]

I haven't used -ultimatch- myself, but -help ultimatch- indicates that the variable _match that it creates contains the same value for observations (treatment or control) that are part of the same matched set. After running -ultimatch-, you might try sorting your data set to produce a convenient ordering, and then using -browse- or -list- to see if this what you want, e.g., something like this:

ultimatch .....
sort _match
browse _match Treatment