Whether we need to show all coefficients of all variables after testing DiD?

[Phuc Nguyen]

I am curious about whether it is a need to show all coefficients of all variables after testing DiD.

Yi,t=αi+βt+γDi,t

while i and t are unit and time fixed effects. γ is the coefficient of variables of interest D.

We all know that the most important variable in DiD is γ. It tells us the DiD, for example, the impact of anti-corruption laws on asset growth.

However, in one package did_imputation following Borusyak, 2021, there is no way to got the coefficients of other variables apart from the γ. I am wondering whether we need to report the coefficients of other variables in DiD setting? If yes, how should we get other coefficients from did_imputation package?

[Phuc Nguyen]

In another word, how to show the coefficients of all independent variables after running this package, sorry I cannot edit the original post.

[Justin Niakamal]

According to the helpfile this is a missing feature, which I interpret to mean it could be added in a future update.

Missing Features

   - Report the coefs on continuous controls from Step 1
    - Save imputed Y(0) in addition to treatment effects
    - Making hbalance work with autosample
    - Throw an error if imputation is not possible with complicated controls
    - Allow to use treatment effect averages for SE computation which rely even on observations outside the estimation sample for the current wtr
    - Estimation when treatment switches on and off
    - More general interactions between FEs and continuous controls than with timecontrols and unitcontrols

[Phuc Nguyen]

According to the helpfile this is a missing feature, which I interpret to mean it could be added in a future update.

Missing Features

 - Report the coefs on continuous controls from Step 1
- Save imputed Y(0) in addition to treatment effects
- Making hbalance work with autosample
- Throw an error if imputation is not possible with complicated controls
- Allow to use treatment effect averages for SE computation which rely even on observations outside the estimation sample for the current wtr
- Estimation when treatment switches on and off
- More general interactions between FEs and continuous controls than with timecontrols and unitcontrols

Thanks Justin Niakamal , I emailed the author and he said that we can extract the coefficients if we want based on the code progress, I am wondering if you can give a hint about this case?

[Justin Niakamal]

It sounds like you'll need to look at the source code for did_imputation and save r(table) where reghdfe is called.

[Phuc Nguyen]

It sounds like you'll need to look at the source code for did_imputation and save r(table) where reghdfe is called.

Thanks Justin Niakamal

Whether we can get the coefficients by regress:

reg y x i.unit i.time ?

[Justin Niakamal]

I'm not sure I fully understand the question but it sounds like you want

reg y x i.unit i.time
mat list r(table)