Hi,

I need to run a difference in difference, but I'm not sure how to construct a formula for this. The problem is that the timing of events affecting the treatment group is not uniform, like in traditional diff.in diff model.
Firm Year (SR)Special report issued Y /N (notworking) Auditor assessment of internal control system working Y- working (0); N-not working (1) (notfixed) If determined that system was not working in prior year is working this year (Yes, 0) or not working again (N, 1)
A 2009 1 0 0
A 2010 1 1 0
A 2011 0 0 1
B 2009 0 1 0
B 2010 0 0 1
B 2011 1 0 0
The table above is a panel data for different firms and I would like to find out the impact of firms issuing special reports in (1) firm-years when auditors have determined that firms' controls are not working properly and (2) during years when the auditors in previous year assessed the controls as ineffective and this year the firm receives the same assessment from auditors. The formula would need to look something like this: Future earnings= SR+ notworking+ SR* notworking + controls, where the interaction variable SR*notworking would be the variable of interest. The same would need to be repeated for firms that receive the same assessment again: Future earnings= SR+ notfixed+ SR* notfixed + controls. I would be comparing the results to control group that has been matched using propensity score. I have already done the PS matching, but I'm not sure about the formulas. Are there any better ways to assess the effect of SR* notworking on future earnings, when compared to the (PS matched) control group of firms? It as suggested to me that will need to create pseudo- treatment years for control groups, but I'm not sure how to do that.