How to compare treated units with never treated units in a panel fixed effects regression?

[Anupam Ghosh]

Hi Statalisters,

I am trying to gauge the effect of a treatment on the treated group and I want to use solely the never treated group for comparison. However, I don't know how can I do this using the Reghdfe function. I would greatly appreciate any help

clear
input float(LIndex_Crime_p Minor_Hurr0_4 Minor_Hurr5_10 Major_Hurr0_4 Major_Hurr5_10 Treated_Ever)
5.238985 0 0 0 0 0
5.011518 0 0 0 0 0
4.943139 0 0 0 0 0
5.035996 0 0 0 0 0
4.733276 0 0 0 0 0
4.5950246 0 0 0 0 0
5.061726 0 0 0 0 0
4.910241 0 0 0 0 0
5.052951 0 0 0 0 0
4.910607 0 0 0 0 0
5.060119 0 0 0 0 0
5.374332 0 0 0 0 0
5.403593 0 0 0 0 0
5.264794 0 0 0 0 0
4.98415 0 0 0 0 0
4.91075 0 0 0 0 0
4.964687 0 0 0 0 0
4.909596 0 0 0 0 0
4.769395 0 0 0 0 0
4.850884 0 0 0 0 0
4.890217 0 0 0 0 0
4.811014 0 0 0 0 0
4.2335343 0 0 0 0 0
4.791189 0 0 0 0 0
4.832047 0 0 0 0 0
4.742643 0 0 0 0 0
4.727606 0 0 0 0 0
4.7475553 0 0 0 0 0
4.706063 0 0 0 0 0
4.43231 0 0 0 0 0
4.3844333 0 0 0 0 0
4.7082186 0 0 0 0 0
4.806615 0 0 0 0 0
4.772902 0 0 0 0 0
4.778645 0 0 0 0 0
4.4632854 0 0 0 0 0
4.6437855 0 0 0 0 0
4.585012 0 0 0 0 0
5.098518 0 0 0 0 0
4.654015 0 0 0 0 0
3.83578 0 0 0 0 0
5.13114 0 0 0 0 1
5.067639 0 0 0 0 1
5.018124 0 0 0 0 1
4.948152 0 0 0 0 1
4.989182 0 0 0 0 1
4.92171 0 0 0 0 1
4.820203 0 0 0 0 1
4.843768 0 0 0 0 1
4.85526 0 0 0 0 1
4.797133 0 0 0 0 1
4.69439 0 0 0 0 1
4.6795115 0 0 0 0 1
4.5897694 0 0 0 0 1
4.810523 0 0 0 0 1
4.59656 0 0 0 0 1
4.6465926 1 0 0 0 1
4.563173 1 0 0 0 1
4.7138095 1 0 0 0 1
4.5374784 1 0 0 0 1
4.629023 1 0 0 0 1
4.058429 1 1 0 0 1
3.5422964 1 1 0 0 1
4.093278 1 1 0 0 1
4.306723 0 1 0 0 1
4.104986 0 1 1 0 1
3.5019445 1 1 1 0 1
3.7638135 1 1 1 0 1
4.4924383 1 1 1 0 1
3.974254 1 1 1 0 1
4.124094 1 0 0 1 1
3.992968 0 1 0 1 1
4.333003 0 1 0 1 1
4.275995 0 1 0 1 1
4.178552 0 1 0 1 1
4.0757422 0 1 0 1 1
3.9907436 0 1 0 0 1
4.0082874 0 0 0 0 1
4.0844755 0 0 0 0 1
4.585639 0 0 0 0 1
4.0777307 0 0 0 0 1
3.253058 0 0 0 0 1
1.6949587 0 0 0 0 0
1.231052 0 0 0 0 0
1.44198 0 0 0 0 0
1.1040821 0 0 0 0 0
. 0 0 0 0 0
. 0 0 0 0 0
. 0 0 0 0 0
. 0 0 0 0 0
. 0 0 0 0 0
1.2143424 0 0 0 0 0
.5789477 0 0 0 0 0
.5622102 0 0 0 0 0
2.0859761 0 0 0 0 0
2.5692904 0 0 0 0 0
2.886828 0 0 0 0 0
2.754909 0 0 0 0 0
1.8406652 0 0 0 0 0
1.837195 0 0 0 0 0
end
[/CODE]

qui reghdfe LIndex_Crime_p Minor_Hurr0_4 Minor_Hurr5_10 Major_Hurr0_4 Major_Hurr5_10, absorb(i.fips i.year#c.Share_Police_1980 i.year#c.Share_Black i.year#c.Share_Male i.year#c.Share_Unemp_1980 i.year#c.Share_BPL_1980 i.year#c.Log_Pop_Density_1980 i.year#c.Log_CoastDist_1980 i.year#c.Log_AvgWages_1980) vce(robust)

[George Ford]

are all treated at the same time? if not, you'll have to use a different method.

[Anupam Ghosh]

no they are not. There is variation in treatment timing.

[George Ford]

If the treatments are staggered, you may need csdid, jwdid, or some such suited to staggered treatments.

looks like you are looking at hurricanes, so treatments come and go. are you looking at short term effects?

[Anupam Ghosh]

Both short and long term effects