Hi everyone,

I have a research project trying to investigate the impact of a policy shock, and the shock occurred at different times for different individuals -- the staggered rollout pattern makes me think about using generalized DiD framework, Y(it) = a X(it)+..., where X(it) is the dummy which is 1 after the policy shock.

However, what is special about my dataset is that the treatment is not always "on" -- after the first policy occurs, the policy may disappear in some years. I am wondering is it still suitable to use it as a quasi-experiment in the generalized DiD analysis? Or are there more suitable models to analyze this type of dataset?

The panel data looks like
Company Year Treatment Dependent Variable
1 2010 1
1 2011 0
1 2012 1
2 2009 0
2 2010 1
2 2011 1
2 2012 0
3 2011 1
3 2012 0