Generating and plotting of an event window relative to a shock.

[Andres Salman]

Dear all,

I am just getting started with some event studies tools in Stata. I hope you can help me clarify some things.
I am struggling particularly with the generation of an event_window variable (relative to the time of the event).

In the esplot package (@Dylan Balla-Elliott), the event_window is defined as follows.

event_indicator = <current_time> == <time of event>
event_time = <current_time> - <time of event>

But I am not sure if I am using the correct definition.
Here is my data example, with a time variable, a continuous variable, and a set of event indicator dummies (which are basically random shocks).

* Example generated by -dataex-. For more info, type help dataex
clear
input str7 modate float epeu_lvl byte(cop_shock unpri_reg_shock eu_reg_shock) float tid
"2004/1"   75.34063 0 0 0   1
"2004/2"   76.99823 0 0 0   2
"2004/3"  125.02164 0 0 0   3
"2004/4"  109.83804 0 0 0   4
"2004/5"  114.84982 0 0 0   5
"2004/6"   99.84531 0 0 0   6
"2004/7"   115.9254 0 0 0   7
"2004/8"    77.3424 0 0 0   8
"2004/9"   89.59677 0 0 0   9
"2004/10" 120.00146 0 0 0  10
"2004/11" 127.93832 0 0 0  11
"2004/12"  83.33497 1 0 1  12
"2005/1"   58.94662 0 0 0  13
"2005/2"   74.97708 0 0 0  14
"2005/3"   81.45479 0 0 0  15
"2005/4"   89.07868 0 0 0  16
"2005/5"   99.44091 0 0 0  17
"2005/6"   99.41497 0 0 0  18
"2005/7"   85.08384 0 0 0  19
"2005/8"   82.83349 0 0 0  20
"2005/9"  160.47383 0 0 0  21
"2005/10"  71.51886 0 0 0  22
"2005/11"  95.44765 0 0 0  23
"2005/12"  61.47662 1 0 1  24
"2006/1"   83.96114 0 0 0  25
"2006/2"   60.63415 0 0 0  26
"2006/3"   79.82993 0 0 0  27
"2006/4"   89.04356 0 0 0  28
"2006/5"   82.44514 0 0 0  29
"2006/6"   89.85152 0 0 0  30
"2006/7"   82.00437 0 0 0  31
"2006/8"   58.86663 0 0 0  32
"2006/9"   76.82971 0 0 0  33
"2006/10"   71.2218 0 0 0  34
"2006/11"  73.84509 1 0 0  35
"2006/12"  74.91799 0 0 0  36
"2007/1"   62.33881 0 0 0  37
"2007/2"   58.51786 0 0 0  38
"2007/3"   71.11645 0 0 0  39
"2007/4"   65.16531 0 0 0  40
"2007/5"   54.99327 0 0 0  41
"2007/6"   60.84606 0 0 0  42
"2007/7"   47.69234 0 0 0  43
"2007/8"   94.66286 0 0 0  44
"2007/9"   166.7332 0 0 0  45
"2007/10"  96.88046 0 0 0  46
"2007/11"  97.73734 0 0 0  47
"2007/12"  98.01473 1 0 1  48
"2008/1"  160.25905 0 0 1  49
"2008/2"  128.78455 0 0 0  50
"2008/3"  139.87073 0 0 0  51
"2008/4"   96.74758 0 0 0  52
"2008/5"   76.82344 0 0 0  53
"2008/6"  106.42784 0 0 0  54
"2008/7"   87.93302 0 0 0  55
"2008/8"   92.29639 0 0 0  56
"2008/9"   156.0435 0 0 0  57
"2008/10"  216.5918 0 0 0  58
"2008/11" 156.77446 1 0 0  59
"2008/12" 136.78456 0 0 0  60
"2009/1"  159.99384 0 0 0  61
"2009/2"  139.69698 0 0 0  62
"2009/3"  133.46071 0 0 0  63
"2009/4"   119.9992 0 0 1  64
"2009/5"   122.9601 0 0 0  65
"2009/6"  113.23891 0 0 0  66
"2009/7"   95.94823 0 0 0  67
"2009/8"   91.37744 0 0 0  68
"2009/9"   104.3236 0 0 0  69
"2009/10" 105.04014 0 0 0  70
"2009/11" 133.00749 1 0 1  71
"2009/12"  115.2626 0 0 1  72
"2010/1"  142.00356 0 0 0  73
"2010/2"  136.73906 0 0 0  74
"2010/3"   137.8383 0 0 0  75
"2010/4"  152.78447 0 0 0  76
"2010/5"  203.30525 0 0 0  77
"2010/6"  171.40266 0 0 1  78
"2010/7"  186.55524 0 0 0  79
"2010/8"  172.81606 0 0 0  80
"2010/9"  161.69014 0 0 0  81
"2010/10"  186.1411 0 1 0  82
"2010/11" 172.68817 1 0 0  83
"2010/12"   183.076 0 0 0  84
"2011/1"  143.03174 0 0 0  85
"2011/2"  122.44579 0 0 0  86
"2011/3"   154.4015 0 0 0  87
"2011/4"   145.5086 0 0 0  88
"2011/5"  134.21507 0 0 1  89
"2011/6"   168.2959 0 0 0  90
"2011/7"  183.40234 0 0 0  91
"2011/8"  230.29893 0 0 0  92
"2011/9"  280.05814 0 0 0  93
"2011/10" 241.75185 0 0 0  94
"2011/11" 304.60022 1 0 0  95
"2011/12"  228.8716 0 0 0  96
"2012/1"  216.73445 0 0 0  97
"2012/2"  193.44435 0 0 0  98
"2012/3"   177.4927 0 0 0  99
"2012/4"  216.99586 0 0 0 100
end

At glance I thought on the following loop to generate event_window. But some questions arise about how to handle the variable with two sequential shocks (i.e in 2009/11 and 2009/12 for eu_reg_shock). Or where two or more shocks are included in the time window. If the window is too large, it will be problematic, I assume.

g event_time =0 if eu_reg_shock==1
foreach v of numlist 1/5 {
    replace event_time = `v' if eu_reg_shock[_n+`v']==1
    replace event_time = -`v' if eu_reg_shock[_n-`v']==1
}

Further on, I want to analyze if these shocks affect the continuous variable before and after. So I plotted in the following way.

graph bar (mean) epeu_lvl, over(event_time)

Ideally, I need to normalize the continuous variable (with mean of one) before the shock.
Here is on plot from Scott R. Baker Nicholas Bloom Stephen J. Terry (2022), that I wish to replicate for these three shocks but I am stuck in the normalization part.

References:
Scott R. Baker Nicholas Bloom Stephen J. Terry (2022). https://www.nber.org/papers/w27167
Dylan Balla-Elliott. https://dballaelliott.github.io/esplot/index.html

[Andres Salman]

Cross posted : https://stackoverflow.com/questions/...a-shock-repost