[Tahseen Hasan]

Dear @Jesse Wur,

It was as you said, I had a few duplicates under a certain type of ID (cusip) but once I switched to another one (gvkey) the issue disappeared. I was able to reshape my dataset to wide but

[Tahseen Hasan]

Dear Jesse Wursten ,

It was as you said, I had a few duplicates under a certain type of ID (cusip) but once I switched to another one (gvkey) the issue disappeared. I was able to reshape my dataset to wide but when dealing with 20+ years, I am having a difficult time figuring out how to use the codes for my panel dataset because my treatment effects were administered at different years. Is it a necessary condition for me to reshape it to Wide format or can I still perform DID inference on a Long format?

For example, ID 111 received a treatment in year 2000, and they also received a treatment in year 2003. However, ID 555 received their treatment in 2005 only. In my panel dataset I have over 50,000 observations.

So my question is, after I have I reshaped it to wide format, should I run the teffects code for each individual year?

1. teffects nnmatch (Dependent2000 X12000, X22000) (Treatment2000), ematch(sic200),.......

2. teffects nnmatch (Dependent 2001 X12001 X22001) (Treatment2001), ematch(sic201)..........

3. 2002, 2003... 2017,



If that would prove to be too complicated for my kind of panel data set-up, where treatment occurs at random years, then is there any way to proceed with a long dataset?



In the past I have successfully managed to use the user created nnmatch formula which automatically drops observations that don't have exact matches. The only problem is that it takes 6-7+ hours to complete:

nnmatch Dependent Treatment X1 X2, exact(sic fyear) bias(X1 X2) keep(nnmatch)

Also just for reference, is there any way to store the matching results for teffects just like there is for the user nnmatch code?



Dataset structure (long):

ID	Year	Income	Treat
111	2000	10	1
111	2001	40	0
111	2002	90	0
111	2003	100	1
111	2004	120	0
111	2005	190	0
333	2000	10	1
333	2001	45	1
333	2002	90	0
333	2003	110	1
333	2004	160	0
333	2005	240	1
333	2006	290	0
333	2007	380	0
555	2000	10	0
555	2001	20	1
555	2002	85	0
555	2003	195	0
555	2004	215	0

[Jesse Wursten]

The dataset being in wide or long format should not have an effect on whether it matters that the treatments happen in different time periods. At this point, it can be useful to think long and hard on what exactly you want to and how that translates into matrices of data. For info on created the matched sample, see this article.

[kjiyengar]

Tahseen,
Does your code

 
 nnmatch Dependent Treatment X1 X2, exact(sic fyear) bias(X1 X2) keep(nnmatch)

spit out the control firms identity? If not, how can we get the control firms?