Test Parallel Trends Assumption Using Leads and Lags

[Asaf Yancu]

Dear Statalists,
I run a Difference in differences regression to explore treatment effect happened in 2016.
I generate leads and lags to test the parallel trends assumption.
This is a graph of the treatment coefficents by year:

Does the parallel trends assumption hold in this case?
If not, how can I interpret this graph?

Many Thanks!

[Maxence Morlet]

Could you show us the regression equation that generated this graph? Do you have panel data? In this case, you would regress the outcome on unit and time fixed effects, and an interaction between the treated dummy and the post indicator, rescaled to be the number of periods relative to treatment. Is that the coefficient you're showing?

On a side note, you should take a look at the community contributed command

xtevent

if you have panel data.