Wild-t bootstrap & staggered DiD: Trouble implementing boottest when using wooldid

[Jessica Marnie]

Hello,

I have a dataset containing individual level data recorded annually in 8 states over a 6 year time-period. I am attempting to estimate the ATT of a treatment that is staggered in timing and have been using the useful package 'wooldid' to do so.

As I have a small number of states, I wish to use wild-t bootstrap standard errors clustered at the state level. For this I have been trying to implement "boottest" - the package associated with Roodman et al. 'Fast & Wild: Bootstrap inference in Stata using boottest'.

I have been running the following code to get the ATT estimates:

wooldid abortions_count state wave treat_wooldid, cluster(state)

with the resulting output:

Main Results (Full Estimation Sample): 

             |  estimate         se          t          p    lb_95ci    ub_95ci      relyr 
-------------+----------------------------------------------------------------------------
main         |                                                                            
         att |  .0023808    .001131   2.105007    .073327  -.0002936   .0050553          .

Then, when following up with the postestimation command:

boottest treat_wooldid, bootcluster(state)

I encounter an error, and no results are produced:

variable main_att not found

My suspicion is that the stored values required by boottest are either named differently by the wooldid package, or are not produced. However, I am not sure how to check or remedy this.

Any advice would be greatly appreciated. Thank you!

Using: Stata SE 18 on Mac