Difference-in-Difference estimation for panel data

[Richa jain]

Dear all,
I am tryin to use Fuzzy DID technique for multiple periods and groups. While running the code, few years Wald_DID is not estimated.
Could anyone suggest me why this happens? when does DID is not estimated by the model?

[Jared Greathouse]

You'd need to read Chaisemartin and d'Haultfoeuille's paper (I think in review of economic studies) on this. I don't know off the top of my head and I definitely can't say since you don't present your data and code, but they likely explain this in the help or the paper

[Richa jain]

I am following Chaisemartin and d'Haultfoeuille's paper and code only for my data. It is written in the paper that Wald DID may not be computed when there may not be groups where the treatment is perfectly stable between two consecutive periods.
Could you please explain me the meaning of this line?

[Jared Greathouse]

I'd advise you to email him (seriously, he helped me a lot with my masters thesis that used one of his methods), but if I understand what they're saying here, they're saying that there needs to be at least one or two untreated units in the control group for any given analysis. They write

This theorem states that with multiple groups and two periods of time, treatment effects for switchers are identified if there is at least one group in which the treatment rate is stable over time.

They further write on that page that the treatment assignment must be independent of time, saying

This requires that the distribution of groups be stable over time. This will automatically be satisfied if the data are a balanced panel and G is time invariant. With repeated cross sections or cohort data, this assumption might fail to hold.

So. if I'm re-reading this correctly, this property might not hold for your purposes.


Now, I didn't use their fuzzy DD command; so, whatever the issue is, it's almost certainly specific to your particular situation. And if I'm being honest with you, I'm (as far as I know) maybe one of the few regular Statalist contributors (including folks like FernandoRios ) who even works with these kinds of methods on a regular basis, and people like Fernando know way more about this than I do. So, to "really" understand what the issue is, your best bet is to email Chaisemartin.

[Richa jain]

Thank you so much.Will email him for the help!