Dear all,

I am running an event study with 649 firm announcements about digital innovations.

With the help of that forum I successfully implemented the standard market model: R(i,t) = alpha(i)+beta(i)*Rm(t) + e(i,t) where R(i,t) is the normal return of a firm and Rm(t) is the return of the market portfolio:

forvalues i=1(1) 649 {
l id_regress id if id_regress==`i' & dif==0
quietly reg actual_ri_return actual_sp_500_cdax_perf_return if id_regress==`i' & estimation_window==1
predict p if id_regress==`i'
replace predicted_return = p if id_regress==`i'
drop p

}

(I am using Stata version 15.)

I plan to implement an event study with risk effects following the work from Dewan, Sanjeev; Ren, Fei, December 2007, Risk and Return of Information Technology Initiatives: Evidence from Electronic Commerce Announcements, Information Systems Research, Vol. 18, No.4 (https://pubsonline.informs.org/doi/a...isre.1070.0120).

They introduce on page 374 the following model: R(i,t) = alpha(i)+alpha´(i)*D(t)+beta(i)*Rm(t)+beta´(i)*R m( t)*D(t)+tau(i)*D(0)+e(i,t).
The dummy variable D(t) is set to zero before the event window and 1 after the event window to make the change of the parameters alpha and beta visible. The dummy variable D(0) is set to one only during the event window to estimate the abnormal return of the firm through the parameter tau.

The idea is to estimate the parameters through Zellner's seemingly unrelated regression (sureg).

Does somebody know if there exists a specific command for that estimation or if I just have to rewrite the "regression line".
My main problem is the implementation of the parameter tau(i) that should only be estimated during the event window itself.

Thank you very much in advance!