Hi all,

I seek your advise about parallel trend graph in a staggered diff-in-diff setting. I am examining staggered adoption of a policy change where the policy was adopted by around 20 U.S. states in different years over the period of 2000-2019, while rest of the states never adopted this policy.

The main benefit of this policy adoption is that, financially distressed firms which are domiciled in these states benefit as this policy helps improve their financial stability, as compared to firms which are not domiciled in these 20 states.

My question is that I would like to plot parallel trend (PT) graph and I am not sure if I should use full sample (all firms in affected states vs all firms in non-affected states) to plot these graphs or use only sample of financially distressed firms (only financially distressed firms in affected states vs only financially distressed firms in non-affected states)? In other words, in creating the PT graph should I still focus on full sample which includes financially distressed as well as financially sound firms, or only financially distressed firms? The question comes because the policy only affects financially distressed firms which are domiciled in states adopting the policy.

For main results, I run following regression:
(outcome var)i,s,t = (pre-event financial distress indicator * PolicyChange indicator)i, policy + (pre-event financial distress indicator)i, policy + (PolicyChange indicator)i,policy + Controls + firm-state fixed effect + year fixed effect + error

where i = firm, s = state, t = year
pre-event financial distress indicator = 1 if a firm is financially distressed in the pre-event year, and 0 otherwise
PolicyChange indicator = 1 for the year a state has adopted the policy and following years, and 0 for previous years

I hope that my question is clear, but I'm happy to explain more! Thanks in advance for your helpful suggestions.

Best,
Anil