Event study - multiple cumulative abnormal return windows

[Nelleke vanderLei]

Dear all,

I am doing an event study and I calculated my abnormal returns for each day in my event window. Now I would like to have cumulative abnormal returns, I do know how to get total cumulative abnormal returns but I would like to get cumulative abnormal returns for day 0 and 1, for day 0 + 1 + 2, day 0 + 1 + 2 + 3 etc.
How can I get this? I did generate the separate days, see my data example.
I hope this is clear and I hope you can help me. Thanks!

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(group_id days abnormal_return ar0 ar1 ar2 ar3 ar4 ar5)
1 -5 -.8582855 . . . . . .
1 -4 .16700104 . . . . . .
1 -3 .1703391 . . . . . .
1 -2 1.9902632 . . . . . .
1 -1 -1.768956 . . . . . .
1 0 3.237539 3.237539 . . . . .
1 1 -3.343374 . -3.343374 . . . .
1 2 3.76568 . . 3.76568 . . .
1 3 -2.884671 . . . -2.884671 . .
1 4 -2.3570244 . . . . -2.3570244 .
1 5 -1.8902892 . . . . . -1.8902892
2 -5 -2.621948 . . . . . .
2 -4 -4.3372827 . . . . . .
2 -3 -3.202744 . . . . . .
2 -2 .15659374 . . . . . .
2 -1 2.008903 . . . . . .
2 0 -2.81355 -2.81355 . . . . .
2 1 -1.2368627 . -1.2368627 . . . .
2 2 2.4508095 . . 2.4508095 . . .
2 3 -1.8451455 . . . -1.8451455 . .
2 4 .9693156 . . . . .9693156 .
2 5 1.8602867 . . . . . 1.8602867
3 -5 2.3462427 . . . . . .
3 -4 .7954761 . . . . . .
3 -3 2.719074 . . . . . .
3 -2 -.6082731 . . . . . .
3 -1 .14762443 . . . . . .
3 0 1.0986892 1.0986892 . . . . .
3 1 1.5341017 . 1.5341017 . . . .
3 2 -1.6110423 . . -1.6110423 . . .
3 3 -.8279147 . . . -.8279147 . .
3 4 -2.262677 . . . . -2.262677 .
3 5 .923644 . . . . . .923644
5 -5 -7.321589 . . . . . .
5 -4 7.963714 . . . . . .
5 -3 -1.246004 . . . . . .
5 -2 -1.5393357 . . . . . .
5 -1 4.6231594 . . . . . .
5 0 .4869067 .4869067 . . . . .
5 1 -.9067889 . -.9067889 . . . .
5 2 -1.686836 . . -1.686836 . . .
5 3 .1129115 . . . .1129115 . .
5 4 -.43044415 . . . . -.43044415 .
5 5 -.10070435 . . . . . -.10070435
end
[/CODE]

[Carlo Lazzaro]

Nelleke:
does the following code do what you want?

bysort group_id: gen x=sum(abnormal_return) if days>=0

[Nelleke vanderLei]

Carlo, thank you. I just saw that I did not copy my example correctly. Well the idea is correct, but I want separate variables for it. Thus:
1 variable that sums abnormal return on day 0 and day 1
1 variable that sums abnormal return on day 0 and day 1 and day 2
1 variable that sums abnormal return on day 0 and day 1 and day 2 and day 3
1 variable that sums abnormal return on day 0 and day 1 and day 2 and day 3 and day 4 and day 5
1 variable that sums abnormal return on day 0 and day 1 and day 2 and day 3 and day 4 and day 5

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(group_id days abnormal_return ar0 ar1 ar2 ar3 ar4 ar5)
 1 -1  -1.768956          0          0          .          .          .          .
 1  2    3.76568          0          0    3.76568          .          .          .
 1  3  -2.884671          0          0          .  -2.884671          .          .
 1 -5  -.8582855          0          0          .          .          .          .
 1 -3   .1703391          0          0          .          .          .          .
 1 -2  1.9902632          0          0          .          .          .          .
 1  1  -3.343374          0  -3.343374          .          .          .          .
 1 -4  .16700104          0          0          .          .          .          .
 1  4 -2.3570244          0          0          .          . -2.3570244          .
 1  5 -1.8902892          0          0          .          .          . -1.8902892
 1  0   3.237539   3.237539          0          .          .          .          .
 2  2  2.4508095          0          0  2.4508095          .          .          .
 2  4   .9693156          0          0          .          .   .9693156          .
 2  1 -1.2368627          0 -1.2368627          .          .          .          .
 2 -4 -4.3372827          0          0          .          .          .          .
 2  5  1.8602867          0          0          .          .          .  1.8602867
 2 -1   2.008903          0          0          .          .          .          .
 2 -5  -2.621948          0          0          .          .          .          .
 2 -2  .15659374          0          0          .          .          .          .
 2  3 -1.8451455          0          0          . -1.8451455          .          .
 2 -3  -3.202744          0          0          .          .          .          .
 2  0   -2.81355   -2.81355          0          .          .          .          .
 3 -3   2.719074          0          0          .          .          .          .
 3 -1  .14762443          0          0          .          .          .          .
 3  5    .923644          0          0          .          .          .    .923644
 3  3  -.8279147          0          0          .  -.8279147          .          .
 3 -4   .7954761          0          0          .          .          .          .
 3 -5  2.3462427          0          0          .          .          .          .
 3  4  -2.262677          0          0          .          .  -2.262677          .
 3  2 -1.6110423          0          0 -1.6110423          .          .          .
 3  1  1.5341017          0  1.5341017          .          .          .          .
 3 -2  -.6082731          0          0          .          .          .          .
 3  0  1.0986892  1.0986892          0          .          .          .          .
end

[Nelleke vanderLei]

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(group_id days abnormal_return x)
 1 -1  -1.768956          .
 1  2    3.76568    3.76568
 1  3  -2.884671   .8810096
 1 -5  -.8582855          .
 1 -3   .1703391          .
 1 -2  1.9902632          .
 1  1  -3.343374 -2.4623644
 1 -4  .16700104          .
 1  4 -2.3570244  -4.819389
 1  5 -1.8902892  -6.709678
 1  0   3.237539 -3.4721394
 2  2  2.4508095  2.4508095
 2  4   .9693156   3.420125
 2  1 -1.2368627  2.1832623
 2 -4 -4.3372827          .
 2  5  1.8602867   4.043549
 2 -1   2.008903          .
 2 -5  -2.621948          .
 2 -2  .15659374          .
 2  3 -1.8451455  2.1984036
 2 -3  -3.202744          .
 2  0   -2.81355  -.6151461
 3 -3   2.719074          .
 3 -1  .14762443          .
 3  5    .923644    .923644
 3  3  -.8279147  .09572935
 3 -4   .7954761          .
 3 -5  2.3462427          .
 3  4  -2.262677 -2.1669476
 3  2 -1.6110423   -3.77799
 3  1  1.5341017 -2.2438881
 3 -2  -.6082731          .
 3  0  1.0986892  -1.145199
end

I just checked it and the code by Carlo, unfortunately, does not do what I want. For example, for day 0 the variable x should return the same value as the simple abnormal return on day 0. For day 1, I would expect the sum of the abnormal return of day 0 and day 1 but that is also not the case. What am I missing?

[Carlo Lazzaro]

Nelleke:
you may want to try the following code:

input float(group_id days abnormal_return x)
 1 -1  -1.768956          .
 1  2    3.76568    3.76568
 1  3  -2.884671   .8810096
 1 -5  -.8582855          .
 1 -3   .1703391          .
 1 -2  1.9902632          .
 1  1  -3.343374 -2.4623644
 1 -4  .16700104          .
 1  4 -2.3570244  -4.819389
 1  5 -1.8902892  -6.709678
 1  0   3.237539 -3.4721394
 2  2  2.4508095  2.4508095
 2  4   .9693156   3.420125
 2  1 -1.2368627  2.1832623
 2 -4 -4.3372827          .
 2  5  1.8602867   4.043549
 2 -1   2.008903          .
 2 -5  -2.621948          .
 2 -2  .15659374          .
 2  3 -1.8451455  2.1984036
 2 -3  -3.202744          .
 2  0   -2.81355  -.6151461
 3 -3   2.719074          .
 3 -1  .14762443          .
 3  5    .923644    .923644
 3  3  -.8279147  .09572935
 3 -4   .7954761          .
 3 -5  2.3462427          .
 3  4  -2.262677 -2.1669476
 3  2 -1.6110423   -3.77799
 3  1  1.5341017 -2.2438881
 3 -2  -.6082731          .
 3  0  1.0986892  -1.145199
end
sort group_id days
bysort group_id: g flag=1 if days>=0
bysort group_id: replace flag=sum(flag) if flag!=.
forval i = 1(1)6  {
bysort group_id: g count`i'=sum( abnormal_return ) if flag<=`i'
}
forval i = 1(1)6  {
bysort group_id: replace count`i'=. if flag[_n]!=`i'
}

[Nelleke vanderLei]

Carlo, it worked! Thanks a lot.