Difference-in-differences graph & parallel trends

[Olivier Van Coppenolle]

Hello everyone,
For our masters thesis we are doing an Difference in difference analysis which gives following results:

To test the validity of the model we did a parallel trends test and a granger test to see whether the parallel trends assumption was met. As you can see this estat ptrends is good as we can’t reject the H0 saying that parallel trends are met. However when we plot the means of the ROA of both the treatment and control groups we see different trends before the treatment (Brexit in 2016).
Anyone knows how this comes or is this graphically enough to suggest parallel trends ( look at 2013-2014)?



Thanks in advance!
Olivier

[George Ford]

I don't know anything about yoru data or your thesis, but that's a strange pattern for the controls (huge change). You sure that's legit?

Looks like the control are more affected by Brexit and the treated.