How to interpret honestdid results?

[Matthew Williams]

Dear all,

I am lost at interpreting results from honestdid command (from SSC), I would appreciate it if anyone who is familiar with this technique/command and could help.
A bit information on my data and set up as follows:
- I have a pooled repeated cross-sectional data
- I evaluate the effect of a policy that affects on individuals aged < 18 years old (younger cohorts) and those aged >=18 (older cohorts) are not affected by the policy.
- I conducted an event study and it suggests that the parallel trends hold because regression coefficients of older cohorts are not statistically significant.
- As a robustness check, I conducted honestdid, however, I am unsure how to interpret its results.

I use the following code to employ honestdid (see more here: https://github.com/mcaceresb/stata-honestdid#honestdid)
reghdfe is from SSC

reghdfe health age1_post-age48_post age1-age48 post $fe, abs($fe) vce(cluster region)

honestdid, pre(1/14) post(16/41) mvec(0.5(0.5)2)
local plotopts xtitle(Mbar) ytitle(95% Robust CI)
honestdid, cached coefplot `plotopts'

Event study graph


honestdid graph

|    M    |   lb   |   ub   |
| ------- | ------ | ------ |
|       . | -0.012 |  0.015 | (Original)
|  0.5000 | -0.026 |  0.029 |
|  1.0000 | -0.042 |  0.046 |
|  1.5000 | -0.060 |  0.064 |
|  2.0000 | -0.080 |  0.083 |
(method = C-LF, Delta = DeltaRM, alpha = 0.050)

[Matthew Williams]

I have just posted this thread but don't know why it is not appeared in the first page of the topic list, so this response is just to make my post visible to everyone. I am sorry for this inconvenience.

[Jared Greathouse]

The issue isn't that it's not appearing, the issue is that you didn't provide a data example or even talk about why this DD is different from the other difference-in-differences estimators. The difference is, there are different DD estimators, all with different assumptions and different parameterizations, so linking to the paper and giving a data example so that we can use your code and work with it on our computer will make all the difference...