Dear Statalist Members,
I have been delving into the intricacies of difference-in-differences analysis, particularly with multiple time periods, and I wanted to extend my gratitude to contributors like Clément de Chaisemartin, Jeff Wooldridge, and FernandoRios for their invaluable insights.
However, I have encountered a puzzling scenario in my understanding.
Let's consider a dataset spanning from 2003 to 2009, where a treatment was implemented for a specific firm in 2004 (denoted as 1), absent in 2005, 2006, and 2007 (denoted as 0 for each year), and reintroduced in 2008 (1) and 2009 (1).
xthdidregress fails to consider this situation. csdid and jwdid also look at the first treatment effect at different time periods. From my understanding and from Jeff Wooldridge , the staggered_did_exit strategy, although applicable, is not effective nor looks at repeated treatment effects at multiple time periods.
Thus, I am struggling with effectively analyzing the treatment effect in such a situation, especially when it occurs intermittently within the cohort.
I would greatly appreciate any guidance or suggestions to navigate this challenge.
Thank you for your assistance.
I have been delving into the intricacies of difference-in-differences analysis, particularly with multiple time periods, and I wanted to extend my gratitude to contributors like Clément de Chaisemartin, Jeff Wooldridge, and FernandoRios for their invaluable insights.
However, I have encountered a puzzling scenario in my understanding.
Let's consider a dataset spanning from 2003 to 2009, where a treatment was implemented for a specific firm in 2004 (denoted as 1), absent in 2005, 2006, and 2007 (denoted as 0 for each year), and reintroduced in 2008 (1) and 2009 (1).
xthdidregress fails to consider this situation. csdid and jwdid also look at the first treatment effect at different time periods. From my understanding and from Jeff Wooldridge , the staggered_did_exit strategy, although applicable, is not effective nor looks at repeated treatment effects at multiple time periods.
Thus, I am struggling with effectively analyzing the treatment effect in such a situation, especially when it occurs intermittently within the cohort.
I would greatly appreciate any guidance or suggestions to navigate this challenge.
Thank you for your assistance.
