Event Study: Sample restrictions & Minimum amount of observations per relative time period

[Kathrin Me]

Hi all,

I aim at estimating the effect of job loss on the labor supply of the spouse using an event study. In my control group, there are only not yet treated units. When only looking at spouses who did not work in the period prior to the job loss, I use

gen helo = .
replace helo = 1 if working_p == 0 & $event == -1 
bys pid: egen maxhelo = max(helo)
keep if maxhelo == 1

before running the regression. working_p denotes the probability to work for the partner, and event is the relative time period. Unfortunately, I get somewhat unrealistic results, when only looking at spouses who did not work prior to the job loss. Is there a mistake in how I approach to only look at non-working spouses (in the period prior to the job loss)?

Alternatively, I thought it might be due to the low number of observations per relative time period?

Relative time period	Treated Observations	Control Observations
-3	372	4,269
-2	503	4,138
-1 (normalized to 0)	0	7,392
0	707	3,934
1	524	4,117
2	433	4,208
3	349	4,292
4 5	310 249	4,331 4,392

Could the effect be driven by outliers?

Thanks already now for your help!

Best,
Kathrin