I am doing an impact evaluation of a program (P) by an organization (O) on poor households' outcome variable (Y). The baseline survey of this was conducted on 2007. After 2007 treatment was assigned randomly at branch level. There are 4 follow up surveys 2009, 2011, 2014 and 2018. However, after 2011 the organization started to support households from control branches. To estimate the impact, I have used the following regression-
I find that the treatment has strong positive impact on outcome (Y) after 7 years of intervention (2014. But after 11 years (2018) the coefficient is positive but insignificant. Why might this be happening? I assume that after 2011 the treatment in control branches has created a spillover effect for which the result is insignificant in 2018. This seems logical to me as the treatment after 2011 was not at branch level, rather at individual level. How can I control for the spillover effect in my analysis in this context?
Code:
reg Y i.treatment##i.year $control, cluster (branchid) robust
