Non-Linear Difference in Differences

[Chinyere Oguh]

I am currently trying to understand the literature on non linear difference in differences. I am investigating the impact of implementing an employment welfare scheme on the likelihood of trust in the government. For this analysis, I only have two panel waves of data, one in 2005 and one in 2012 which serve as pre treatment and post treatment periods. The data also contains whether an individual participated in the scheme or not.

I am currently struggling with figuring out exactly how to do this and how to interpret the coefficients.

Another thing that is affecting me is the fact that the scheme was implemented in rounds where some districts received treatment in 2006, some in 2007 and some in 2008, which would create heterogenous treatment effects depending on the time at which each group is treated.

If anyone at all could possibly help me I would greatly appreciate it!!!!!

Kind regards

[Chinyere Oguh]

Bumping so that someone can help please.

[Jared Greathouse]

I don't understand the problem. You have one pre period, T_1=1, and one post period, T_2=1, and 2 total periods, T_1 U T_2=2. Right?

The best you can do here is just standard DID. You can't even get at heterogeneous effects, unless I've misunderstood your setup.

[Chinyere Oguh]

Hey Jared, I understand how you may be confused.

I have control variables that accounts for the year when the scheme was implemented for certain areas within the country where the event took place.

[Jared Greathouse]

Very well. Use dataex to show me how your data look, then.

[Chinyere Oguh]

I have actually managed to do the regression, but I was wondering whether a staggered treatment model is unideal when the dependent variable is binary, or can it just be interpreted in a similar way.