Creating Event Plots for Multiple treated units, Multiple time periods and Heterogenous treatments.

[Anupam Ghosh]

I have a US county panel (T=30) where different counties are treated by varying intensities of natural disasters at different points in time.

I generate the lags and leads manually and then use them in the regression. However, in the current procedure, I am unable to plot the null coefficient of the omitted period (t=-1) and also I am unable to get a horizontal/vertical line. Hence, it would be really helpful if someone can suggest a better piece of code.

Also in an alternative treatment specification, I have a continuous treatment which denotes the Z score of the disaster intensity. Hence any suggestions to that front would also be quite helpful.

************************************************** ********************
by fips (year), sort: gen MajLag0 = inrange(Hurr_Lag0, 48, 299)
by fips (year), sort: gen MajLag1 = inrange(Hurr_Lag1, 48, 299)
by fips (year), sort: gen MajLag2 = inrange(Hurr_Lag2, 48, 299)
by fips (year), sort: gen MajLag3 = inrange(Hurr_Lag3, 48, 299)
by fips (year), sort: gen MajLag4 = inrange(Hurr_Lag4, 48, 299)
by fips (year), sort: gen MajLag5 = inrange(Hurr_Lag5, 48, 299)
by fips (year), sort: gen MajLag6 = inrange(Hurr_Lag6, 48, 299)
by fips (year), sort: gen MajLag7 = inrange(Hurr_Lag7, 48, 299)
by fips (year), sort: gen MajLag8 = inrange(Hurr_Lag8, 48, 299)
by fips (year), sort: gen MajLag9 = inrange(Hurr_Lag9, 48, 299)
by fips (year), sort: gen MajLag10 = inrange(Hurr_Lag10, 48, 299)

************************* **************************

//by fips (year), sort: gen MajLead1 = 0 //drop t-1
by fips (year), sort: gen MajLead2 = inrange(Hurr_Lead2, 48, 299)
by fips (year), sort: gen MajLead3 = inrange(Hurr_Lead3, 48, 299)
by fips (year), sort: gen MajLead4 = inrange(Hurr_Lead4, 48, 299)
by fips (year), sort: gen MajLead5 = inrange(Hurr_Lead5, 48, 299)
by fips (year), sort: gen MajLead6 = inrange(Hurr_Lead6, 48, 299)
by fips (year), sort: gen MajLead7 = inrange(Hurr_Lead7, 48, 299)
by fips (year), sort: gen MajLead8 = inrange(Hurr_Lead8, 48, 299)
by fips (year), sort: gen MajLead9 = inrange(Hurr_Lead9, 48, 299)
by fips (year), sort: gen MajLead10 = inrange(Hurr_Lead10, 48, 299)

local treat_vars "MajLead6 MajLead5 MajLead4 MajLead3 MajLead2 MajLag0 MajLag1 MajLag2 MajLag3 MajLag4 MajLag5 MajLag6 " // Omitted MajLag1 (baseline t = -1)

local dep_vars "Violent_Percap Property_Percap Total_Crime_Percap "

local i = 1
foreach dep_var in `dep_vars' {
qui xtreg `dep_var' `treat_vars' i.year i.state_var##c.year, fe vce(cluster fips)
estimates store Maj_Reg`i'
local i = `i' + 1
}

coefplot (Maj_Reg1, label("Violent Crime") pstyle(p4) ) ///
|| (Maj_Reg2, label("Property Crime") pstyle(p4) ) ///
, keep(MajLead* MajLag*) vertical ciopts(recast(rcap)) drop(_cons) yline(0) msize(1) msymbol(D) byopts(yrescale row(1)) ///
coeflabels(MajLead6 = "-6" ///
MajLead5 = "-5" ///
MajLead4 = "-4" ///
MajLead3 = "-3" ///
MajLead2 = "-2" ///
(0) = "-1" /// Manually set t = -1 as zero
MajLag0 = "0" ///
MajLag1 = "+1" ///This is t = 0 (current period)
MajLag2 = "+2" ///
MajLag3 = "+3" ///
MajLag4 = "+4" ///
MajLag5 = "+5" ///
MajLag6 = "+6" ///
) ///
legend(off) xlabel(, angle(45)) xtitle({stSerif: Year}, size(3) color(gs5)) ///
ylabel(, angle(45)) ytitle({stSerif: Coefficient }, size(3) color(gs5)) ///
subtitle(, size(3) margin(none) justification(center) color(gs5) bcolor(white) bmargin(top_bottom))