How to run plot parallel trends lines and event study diagram in STATA 15?

[Prateek Mishra]

Hi, I am new to STATA. My data has 6 periods, 1 in pre-treatment (1983). The dependent variable is LIT, and the others are covariates. Under14 is the treatment term (1, if 6 <= AGE <= 13, 0 if 14 <= AGE <= 17) and post1986 (1, if YEAR > 1986, 0 otherwise) is the post-term. The problem is I have no data for the treatment year 1986.
URBAN: 1 for urban, 0 for rural
SEX: 1 for male, 0 for female.
NCHILD: number of children in a family.
FAMSIZE: number of members in the family.
STATE: geographical states of India.
LIT: literacy rate (1, if literate, 0 otherwise).
Periods: 1983, 1987, 1993, 1999, 2004, 2009.
Control group - 14-17 years and, treatment group - 6-13 years.
This is an example of my data, which has more than 900k observations.

YEAR	STATE	URBAN	FAMSIZE	AGE	SEX	LIT	Post1986	Under14	NCHILD
1983	J&K	0	6	8	0	0	0	1	0
1987	Andhra Pradesh	0	9	10	0	0	1	1	1
1990	Delhi	0	23	16	1	0	0	1	0
2004	West Bengal	1	23	6	0	0	0	1	0

The treatment was implemented in a single period, 1986, and therefore not staggered. Also, if it is helpful, my DiD model was meant to be a TWFE model, but I had to drop the time-fixed effects due to multicollinearity with the Post1986 term.
I have already tried several ways of creating the plots, also in RStudio, but I think I am getting them wrong. If anyone could please help me with the syntax, it would be really helpful. Thank you!

[Clyde Schechter]

You cannot assess parallel trends with this data design because you have only one pre-intervention observation time. So whatever you are getting, it is necessarily wrong.

Also, if it is helpful, my DiD model was meant to be a TWFE model, but I had to drop the time-fixed effects due to multicollinearity with the Post1986 term.

Yes, this always happens if the intervention begins at the same time in all treated units. It is not a problem. Don't waste any time or energy worrying about it.

[Prateek Mishra]

Clyde Schechter Sir, thank you so much for your response. I apologise for replying this late.

It is not a problem. Don't waste any time or energy worrying about it.

Yes, I understand that since my treatment isn't staggered across periods, there's nothing really I can do about it.

You cannot assess parallel trends with this data design because you have only one pre-intervention observation time.

Actually, this work pertains to my master's thesis, and being totally new to causal inference and its related topics, I made the error of not cross-checking from the beginning itself while selecting the research area. Honestly, this data at hand is the only source, and none other is available anywhere. I had a very detailed chat with my thesis advisor about it, who said there are certain techniques to bypass this issue of working with a single pre-treatment period.

Ever since then, I have been looking for literature and other sources all over the web that might have used such techniques but have been unsuccessful so far. I understand this might seem very naive of me to ask for such help, but could you guide me in this regard, as in what possible techniques could there be to tackle this situation? I am totally stuck at the moment. Thank you!

[Clyde Schechter]

I'm not aware of any techniques that bypass this issue. I can't say for sure whether that is due to their non-existence or represents a gap in my knowledge. But either way, I'm afraid I have nothing to offer you on this. Perhaps another Forum member following the thread knows something helpful and will chime in. Or, you should press your thesis advisor for more specific advice.

[Prateek Mishra]

Sure, I will discuss this issue with them. Thank you so much for your feedback!