Event Study Graph: Do I need an indicator variable for each pre and post treatment period (except t-1)?

[Vera Mikhailovna]

 * Example generated by -dataex-. For more info, type help dataex clear input float(id monthly independent sales TreatZero lead2 lead3 lead4 lead5 lead6 lead7_backwards lag1 lag2 lag3 lag4 lag5 lag6 lead1) 1 672 0  249512 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 673 0  177712 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 674 0  109524 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 675 0   20776 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 676 0  846471 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 677 0  328806 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 678 0   46470 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 679 0  394758 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 680 0  301179 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 681 0  756129 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 682 0  116117 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 683 0  374293 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 684 0  432423 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 685 0  364780 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 686 0  797174 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 687 0  400569 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 688 0  126897 0 0 0 0 0 0 0 0 1 0 0 0 0 0 2 672 1   65104 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 673 1   77133 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 674 1   76200 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 675 1  218342 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 676 1   39265 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 677 1    6649 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 678 1   41677 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 679 1  156277 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 680 1   98535 0 0 0 0 0 1 0 0 0 0 0 0 0 0 2 681 1    3920 0 0 0 0 1 0 0 0 0 0 0 0 0 0 2 682 1  165573 0 0 0 1 0 0 0 0 0 0 0 0 0 0 2 683 1   73413 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 684 1   97216 0 1 0 0 0 0 0 0 0 0 0 0 0 0 2 685 1  106015 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 686 1   33066 1 0 0 0 0 0 0 0 0 0 0 0 0 0 2 687 1   54207 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 688 1  118173 0 0 0 0 0 0 0 0 1 0 0 0 0 0 3 672 0  737203 0 0 0 0 0 0 1 0 0 0 0 0 0 0 3 673 0  306725 0 0 0 0 0 0 1 0 0 0 0 0 0 0 3 674 0  198990 0 0 0 0 0 0 1 0 0 0 0 0 0 0 3 675 0 1054751 0 0 0 0 0 0 1 0 0 0 0 0 0 0 3 676 0 1886147 0 0 0 0 0 1 0 0 0 0 0 0 0 0 3 677 0 1142545 0 0 0 0 1 0 0 0 0 0 0 0 0 0 3 678 0 1277825 0 0 0 1 0 0 0 0 0 0 0 0 0 0 3 679 0  397706 0 0 1 0 0 0 0 0 0 0 0 0 0 0 3 680 0 1354199 0 1 0 0 0 0 0 0 0 0 0 0 0 0 3 681 0 1348788 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 682 0  914274 1 0 0 0 0 0 0 0 0 0 0 0 0 0 3 683 0  805134 0 0 0 0 0 0 0 1 0 0 0 0 0 0 3 684 0  769588 0 0 0 0 0 0 0 0 1 0 0 0 0 0 3 685 0  292174 0 0 0 0 0 0 0 0 0 1 0 0 0 0 3 686 0 1236297 0 0 0 0 0 0 0 0 0 0 1 0 0 0 3 687 0   58338 0 0 0 0 0 0 0 0 0 0 0 1 0 0 3 688 0 1681455 0 0 0 0 0 0 0 0 0 0 0 0 1 0 4 672 1   82611 0 0 0 0 0 0 1 0 0 0 0 0 0 0 4 673 1  190401 0 0 0 0 0 0 1 0 0 0 0 0 0 0 4 674 1  122867 0 0 0 0 0 0 1 0 0 0 0 0 0 0 4 675 1  111444 0 0 0 0 0 0 1 0 0 0 0 0 0 0 4 676 1   44781 0 0 0 0 0 1 0 0 0 0 0 0 0 0 4 677 1  158895 0 0 0 0 1 0 0 0 0 0 0 0 0 0 4 678 1   71693 0 0 0 1 0 0 0 0 0 0 0 0 0 0 4 679 1   62140 0 0 1 0 0 0 0 0 0 0 0 0 0 0 4 680 1  321720 0 1 0 0 0 0 0 0 0 0 0 0 0 0 4 681 1  188944 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 682 1  179921 1 0 0 0 0 0 0 0 0 0 0 0 0 0 4 683 1  159214 0 0 0 0 0 0 0 1 0 0 0 0 0 0 4 684 1  118173 0 0 0 0 0 0 0 0 1 0 0 0 0 0 4 685 1  246030 0 0 0 0 0 0 0 0 0 1 0 0 0 0 4 686 1   83191 0 0 0 0 0 0 0 0 0 0 1 0 0 0 4 687 1  100867 0 0 0 0 0 0 0 0 0 0 0 1 0 0 4 688 1   42409 0 0 0 0 0 0 0 0 0 0 0 0 1 0 5 672 0   32247 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 673 0    9993 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 674 0   44384 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 675 0   28284 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 676 0    6873 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 677 0   35780 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 678 0     226 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 679 0   41062 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 680 0   34161 0 0 0 0 0 0 1 0 0 0 0 0 0 0 5 681 0    5773 0 0 0 0 0 1 0 0 0 0 0 0 0 0 5 682 0   12586 0 0 0 0 1 0 0 0 0 0 0 0 0 0 5 683 0   22660 0 0 0 1 0 0 0 0 0 0 0 0 0 0 5 684 0   40637 0 0 1 0 0 0 0 0 0 0 0 0 0 0 5 685 0   40881 0 1 0 0 0 0 0 0 0 0 0 0 0 0 5 686 0    3560 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 687 0    9365 1 0 0 0 0 0 0 0 0 0 0 0 0 0 5 688 0     852 0 0 0 0 0 0 0 1 0 0 0 0 0 0 6 672 0   94715 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 673 0    2692 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 674 0  123457 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 675 0  724462 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 676 0  871857 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 677 0   16821 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 678 0  499244 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 679 0  441009 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 680 0  429921 0 0 0 0 0 0 1 0 0 0 0 0 0 0 6 681 0  156341 0 0 0 0 0 1 0 0 0 0 0 0 0 0 6 682 0  461273 0 0 0 0 1 0 0 0 0 0 0 0 0 0 6 683 0  325237 0 0 0 1 0 0 0 0 0 0 0 0 0 0 6 684 0  302210 0 0 1 0 0 0 0 0 0 0 0 0 0 0 6 685 0  332281 0 1 0 0 0 0 0 0 0 0 0 0 0 0 6 686 0  298871 0 0 0 0 0 0 0 0 0 0 0 0 0 0 end format %tm monthly

I have a staggered diff in diff setting. In the dataset, lead values correspond to the indicator variables for pre-treatment values, whereas lag values correspond to the indicator variables for post-treatment values. The dependent variable is sales. This is my code

xtset id monthly  xtreg sales lead7_backwards lead6 lead5 lead4 lead3 lead2 lead1 TreatZero lag1 lag2 lag3 lag4 lag5 lag6 i.monthly, fe vce(cluster id)  coefplot, vertical omitted keep(lead6 lead5 lead4 lead3 lead2 lead1 TreatZero lag1 lag2 lag3 lag4 lag5 lag6) ciopts(recast(rcap)) yline(0) msymbol(d)

My question is, do I need to include lead7_backwards in the regression? Or instead, I need to run:

xtreg sales lead6 lead5 lead4 lead3 lead2 lead1 TreatZero lag1 lag2 lag3 lag4 lag5 lag6 i.monthly, fe vce(cluster id)

[Jared Greathouse]

I don't recommend you use this approach. Use did_multiplegt by Chaisemartin and D'Haultfœuille or a similar command

[Vera Mikhailovna]

Thank you for reply Jared. For robustness, I use the csdid and flexpaneldid packages. Indeed, I examined the ado files of several did packages and discovered that while some of them use all available time periods in the regressions that underpin the graphs, others use only a limited number of pre- and post-treatment periods. This is a little confusing...

[Jared Greathouse]

Not confusing, just reflects the recent changes in difference-in-differences estimation. Relative event time and heterogenous treatment effects has become quite an important topic of discussion among econometricians

[Vera Mikhailovna]

Then I should include lead7_backwards or not in the regression, if not confusing

[Jared Greathouse]

You can. I can't really tell from your data example though if it makes sense, though