Diff in Diff with different times

[Mina Sami]

Dear Statalisters,

I am studying the following question: "What is the impact of distributing dividends on sales while the firm is experiencing a loss"?
I have the following information

Firm	year	Distribute=1	Sales
F1	2000	1	1000
F1	2001	1	2000
F1	2002	0	500
F2	2000	1	1000
F2	2001	0	2500
F2	2002	1	3000
F3	2000	1	2500
F3	2001	1	2550
F3	2002	1	3000

The question here: can I use Diff in Diff in this case? Please note that here I cannot insert a time from which the program starts as there is not fixed year of program. This totally depends on the firm and the year of treatment. Please note also that some firms have always 1 (such as the third firm).
If can I use diff in diff, what command can fit my data so I have robust results?

[Clyde Schechter]

This data design is not suitable for a diff-in-diff analysis. You have neither a treatment nor a control group, and the distribution variable goes on and off haphazardly over time. You can still analyze this data just using the distribution variable, and probably including fixed effects for both firm and year. But it won't be a diff-in-diff analysis, and its identification of causal effect will not be as strong as that of a diff-in-diff analysis.