Should I use month(time variable) for clustering?

[Prateek Manikpuri]

Hi,

I am trying to see whether household savings have changed after implementation of a policy in treated states. For this I am thinking of running a diff-in-diff regression, comparing the border districts. I have monthly data for household savings of 2 years, 1 year pre-period and 1 year post-period. Should I use only district variable for clustering or should I use both district and month?

If yes, then which of the two commands should I use -

reghdfe savings interaction treatment post, absorb(district month) cluster(district month)

or

egen district_month = group(district month)
reghdfe savings interaction treatment post, absorb(district month) cluster(district_month)

Thank you.