Event study

[Juliana Cunha]

Hi All, Please, could anyone help me with coding for an event study. The dataex follows below:

* Example generated by -dataex-. For more info, type help dataex
clear
input float(id ill dt_notific crimedate) str7 crime_type float(crime_dummy crime_count crime_dummy_post crime_count_post age)
1001 0     .     . ""        0 0 0 0 16
1002 0     . 20980 "Robbery" 1 2 0 0 16
1002 0     . 21087 "Theft"   1 2 0 0 16
1003 1 20834 20860 "Theft"   1 1 1 1 17
1004 0     .     . ""        0 0 0 0 17
1005 1 20929 20855 "Assault" 1 2 1 1 17
1005 1 20929 20940 "Robbery" 1 2 1 1 17
1006 0     .     . ""        0 0 0 0 15
1007 0     . 21189 "Fraud"   1 1 0 0 15
1008 1 21094 21063 "Theft"   1 1 0 0 17
1009 0     .     . ""        0 0 0 0 16
1010 1 21227 21258 "Drug"    1 1 1 1 15
1011 0     .     . ""        0 0 0 0 15
1012 0     .     . ""        0 0 0 0 15
1013 0     .     . ""        0 0 0 0 16
1014 1 21411     . ""        0 0 1 1 17
1015 0     . 21220 "Theft"   1 1 0 0 18
1016 0     .     . ""        0 0 0 0 16
1017 1 21498 21519 "Robbery" 1 2 1 2 16
1017 1 21498 21628 "Fraud"   1 2 1 2 16
1018 0     .     . ""        0 0 0 0 15
1019 0     .     . ""        0 0 0 0 15
1020 0     . 20888 "Theft"   1 2 0 0 16
1020 0     . 21664 "Assault" 1 2 0 0 16
1021 1 21613     . ""        0 0 1 1 15
1022 0     .     . ""        0 0 0 0 18
1023 0     .     . ""        0 0 0 0 17
1024 1 21710     . ""        0 0 1 1 18
1025 0     .     . ""        0 0 0 0 16
end
format %td dt_notific
format %td crimedate

I want to estimate the causal effect of getting ill (ill - dt_notific for date of notification of the illness) on subsequent crime commitment.
Some individuals commit crimes before and/or after getting sick.
My outcomes of interest are:

1. crime_dummy_post – indicator for committing any crime after dengue, and
2. crime_count_post – number of crimes after dengue.

How should I construct an event-study for this in STATA?
Many thanks in advance!

[Joseph Coveney]

How should I construct an event-study for this in STATA?

If your data snippet is representative, then I don't think that you can.

You don't have any elapsed time (duration of period under observation) except for those who were charged with a crime after having become ill—you don't have a censoring date in your dataset for those who became ill but were never charged with a crime. And those youths who never became ill never became at risk—you don't have elapsed time at all for them regardless of whether they were charged with a crime.

[Juliana Cunha]

Hi Joseph, thank you for answering. What if the data looks like this?

* Example generated by -dataex-. For more info, type help dataex
clear
input float(id ill dt_notific crimedate) str7 crime_type float(crime_dummy age)
1001 0     .     . ""        0 16
1001 0     . 20980 "Robbery" 1 16
1001 0     .     . ""        0 16
1001 0     .     . ""        0 16
1002 0     .     . ""        0 16
1002 0     . 21087 "Theft"   1 16
1002 0     .     . ""        0 16
1003 1 20834     . ""        0 17
1003 1 20834 20860 "Theft"   1 17
1003 1 20834     . ""        0 17
1003 1 20834     . ""        0 17
1004 0     .     . ""        0 17
1004 0     . 20929 "Assault" 1 17
1004 0     .     . ""        0 17
1005 1 20929     . ""        0 17
1005 1 20929 20950 "Robbery" 1 17
1005 1 20929     . ""        0 17
1006 0     .     . ""        0 15
1006 0     . 21189 "Fraud"   1 15
1006 0     .     . ""        0 15
1007 1 21094     . ""        0 17
1007 1 21094 21098 "Theft"   1 17
1007 1 21094     . ""        0 17
1008 0     .     . ""        0 15
1008 0     .     . ""        0 15
1008 0     . 21227 "Drug"    1 15
1009 0     .     . ""        0 15
1009 0     .     . ""        0 15
1010 1 21411     . ""        0 17
1010 1 21411 21420 "Theft"   1 17
1010 1 21411     . ""        0 17
1011 0     .     . ""        0 18
1011 0     . 21498 "Robbery" 1 18
1011 0     .     . ""        0 18
1012 1 21498     . ""        0 16
1012 1 21498 21628 "Fraud"   1 16
1012 1 21498     . ""        0 16
end
format %td dt_notific
format %td crimedate

[Joseph Coveney]

What if the data looks like this?

I don't see that you've added any substantive data.

About the best that I can see is for some kind of cross tabulation of the ill and crime_dummy variables. If you want to add age as a covariate, then perhaps fit a logistic regression model. Both would be one observation per person. Neither would "estimate the causal effect of getting ill . . . on subsequent crime commitment."

[Juliana Cunha]

I think that what I meant was how to use csdid - Callaway and Santana method for my data set layout. thank you!