Difference-in-differences with time varying differentiation.

[Sergio Garcia Bernal]

I have a data panel with information on natural disasters occurred in different municipalities and at different points in time. A municipality can have zero, one or more than one disaster during the period of time. For example, a municipality X suffered an earthquake in t, t+2 and t+5.

I want to measure the effect of the disaster on educational outcomes such as dropout rate. This by using a difference-in-differences model with time varying differentiation.

I need help in testing the parallel trends assumption. Any advice on how to do it given the setting? I’ll appreciate it a lot

year	municipality	disasters
2011	A	24
2012	A	10
2013	A	32
2014	A	27
2015	A	21
2016	A	20
2017	A	15
2011	B	1
2012	B	1
2013	B	0
2014	B	0
2015	B	3
2016	B	1
2017	B	0

[Jared Greathouse]

You have a continuous treatment?

[Sergio Garcia Bernal]

I have the two kind of treatment, and I can use either of one.
One: discrete in which 1 is treated, 0 otherwise.
Second: close to continuous as I have the intensity: 1 if only one episode, 2 for two episodes and so on.