generate a dummy varaible based on whetehr the other dummy has the same or more ==1 than ==0 over the sample period

[Alice Yang]

Hello

I have a panel data with firm and year variables. I also have a dummy variable, litigation, which equals 1 if there are 1 or more litigations for a firm in a year and 0 if there is no litigation for the firm in that year. I want to create a new dummy variable, abovemedian_litigation, which equals 1 if the firm has 50% or more of the times with litigation==1 than litigation==0 over the sample period. How can I do this? Thanks a lot for any suggestions!

* Example generated by -dataex-. To install: ssc install dataex
clear
input float year double firm float litigation
2010  2 1
2011  2 1
2012  2 1
2013  2 1
2014  2 0
2015  2 0
2016  2 1
2017  2 1
2018  2 1
2019  2 1
2010  4 0
2011  4 0
2012  4 1
2013  4 1
2014  4 1
2015  4 0
2016  4 1
2017  4 0
2018  4 1
2019  4 1
2010  5 1
2011  5 1
2012  5 1
2013  5 1
2014  5 1
2015  5 0
2016  5 1
2017  5 1
2018  5 1
2019  5 1
2011  7 0
2012  7 0
2013  7 1
2014  7 0
2015  7 1
2016  7 0
2017  7 0
2018  7 0
2019  7 1

[Øyvind Snilsberg]

egen wanted = max(sum(litigation)/_N>=.5), by(firm)

[Alice Yang]

Hi Øyvind,

Thanks a lot for your quick reply and help. I tried your code but there may be something wrong? Such as firm 2, there are 8 1s and 2 0s for litigation, so the firm has more years with litigation than without, then the wanted variable should be coded 1 for firm 2 from 2010-2019, however if I use the code above, it gives 0 instead. Can you please kindly have another look at the code? Thanks!

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(year firm litigation wanted)
2010 2 1 0
2011 2 1 0
2012 2 1 0
2013 2 1 0
2014 2 0 0
2015 2 0 0
2016 2 1 0
2017 2 1 0
2018 2 1 0
2019 2 1 0
2010 4 0 0
2011 4 0 0
2012 4 1 0
2013 4 1 0
2014 4 1 0
2015 4 0 0
2016 4 1 0
2017 4 0 0
2018 4 1 0
2019 4 1 0
2010 5 1 1
2011 5 1 1
2012 5 1 1
2013 5 1 1
2014 5 1 1
2015 5 0 1
2016 5 1 1
2017 5 1 1
2018 5 1 1
2019 5 1 1
2011 7 0 1
2012 7 0 1
2013 7 1 1
2014 7 0 1
2015 7 1 1
2016 7 0 1
2017 7 0 1
2018 7 0 1
2019 7 1 1
end

[Nick Cox]

. bysort firm : egen median = median(litigation)

. gen wanted = median == 1 

. tabdisp firm, c(wanted)

----------------------
     firm |     wanted
----------+-----------
        2 |          1
        4 |          1
        5 |          1
        7 |          0
----------------------

[Alice Yang]

Hi Nick,

Thanks for your help! I noticed that if there are equal numbers of 1s and 0s in litigation for a firm in the sample period, this method will assign wanted 0. I'd like wanted to be 1 if there are 50% or more of the 1s in litigation, how can the code be revised? Thanks!

[Nick Cox]

You're right. The median will be 0.5 with equal numbers of 1s and 0s.

median >= 0.5

[Alice Yang]

Hi again,

I have another question regarding this post. So in #1, for each firm, I want to account for whether more than 50% of the years the firm has litigation==1 than litigation==0 over the entire sample period. Now I have a new variable, eventyear, to indicate the year a event happens to a firm. If a firm is in the treatment group, the eventyear variable will indicate the year it has been treated. Different firms are being treated at different years, and a firm can only be treated once during the sample period. Other firms, namely those in the control group, will never be treated and have no value for the eventyear variable.

I want to create a dummy variable to indicate:
- For treatment firms, account for the years prior to the eventyear only, whether more than 50% of the years the firm has litigation==1 than litigation==0.
- For control firms, same as in #1, account for all years in the sample period, whether more than 50% of the years the firm has litigation==1 than litigation==0.

How can the code be revised? Thakns a lot for helping me again!

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(year firm litigation eventyear)
2010 2 1 2013
2011 2 1 2013
2012 2 1 2013
2013 2 1 2013
2014 2 0 2013
2015 2 0 2013
2016 2 1 2013
2017 2 1 2013
2018 2 1 2013
2019 2 1 2013
2010 4 0    .
2011 4 0    .
2012 4 1    .
2013 4 1    .
2014 4 1    .
2015 4 0    .
2016 4 1    .
2017 4 0    .
2018 4 1    .
2019 4 1    .
2010 5 1 2015
2011 5 1 2015
2012 5 1 2015
2013 5 1 2015
2014 5 1 2015
2015 5 0 2015
2016 5 1 2015
2017 5 1 2015
2018 5 1 2015
2019 5 1 2015
2011 7 0    .
2012 7 0    .
2013 7 1    .
2014 7 0    .
2015 7 1    .
2016 7 0    .
2017 7 0    .
2018 7 0    .
2019 7 1    .
end

[Alice Yang]

Anyone can kindly help me? Thanks!