I am using a diff-in-diff methodology on a cross-sectional data.

I am first using bootstrap clustering with 1000 replication to get the bootstrap standard error.

Now for asymptomatic refinements t value and also because the cluster group is small. I am employing, wild cluster bootstrap t (p-value) for the post-estimation test using the boottest command.

Now, most of the applied papers I came across doesn't use bootstrap standard errors in the first step but cluster-robust standard errors (eg: cluster(states))

I am contemplating with my step 1 vs others to use. Is my step 1 correct?