Panel data, if one observation is true for a participant generate a variable that is true for all observation of that participant

[Hugo Cooke]

Dear Statlist,

A simple question this morning I think but one I have been struggling to answer (although I have found one very clunky solution). I am using panel data with employment status and I have been trying to generate a dummy variable of 1 for all observation of participants who have at least one instance of job loss during the duration of the panel.

The solution i have found is using the command below and just repeating it until there are no more changes. It's a solution but I would really appreciate a cleaner one.

gen jobloss_anypoint = 1 if jobloss == 1
replace jobloss_anypoint = 1 if jobloss[_n - 1] == 1 & _nomem_encr == _nomem_encr[_n - 1]
replace jobloss_anypoint = 1 if jobloss[_n + 1] == 1 & _nomem_encr == _nomem_encr[_n + 1]

* Example generated by -dataex-. For more info, type help dataex
clear
input double _nomem_encr float jobloss int _Year float jobloss_anypoint
800009 0 2008 .
800009 0 2009 .
800009 0 2010 .
800009 0 2011 .
800009 0 2012 .
800009 0 2013 .
800009 0 2014 .
800009 0 2015 .
800009 0 2016 .
800009 0 2017 .
800009 0 2018 .
800009 0 2019 .
800015 0 2008 .
800015 0 2009 .
800015 0 2010 .
800015 0 2011 .
800015 0 2012 .
800015 0 2013 .
800015 0 2014 .
800015 0 2015 .
800015 0 2016 .
800015 0 2017 .
800015 0 2018 .
800015 0 2019 .
800042 0 2008 .
800042 0 2009 .
800042 0 2010 .
800042 0 2011 .
800042 0 2012 .
800042 0 2013 .
800042 0 2014 .
800042 0 2015 .
800042 0 2016 .
800042 0 2017 .
800042 0 2018 .
800042 0 2019 .
800057 0 2008 .
800057 0 2009 .
800057 0 2010 .
800057 0 2011 .
800057 0 2012 .
800057 0 2013 .
800057 0 2014 .
800057 0 2015 .
800057 0 2016 .
800057 0 2017 .
800057 0 2018 .
800057 0 2019 .
800073 0 2015 .
800073 0 2016 .
800073 0 2017 .
800073 1 2018 1
800073 0 2019 .
800085 0 2008 .
800085 0 2009 .
800085 0 2010 .
800085 0 2011 .
800085 0 2012 .
800085 0 2013 .
800085 0 2014 .
800085 0 2015 .
800085 0 2016 .
800085 0 2017 .
800085 0 2018 .
800085 0 2019 .
800100 0 2008 .
800100 0 2009 .
800100 0 2010 .
800100 0 2011 .
800100 0 2012 .
800100 0 2013 .
800100 0 2014 .
800100 0 2015 .
800100 0 2016 .
800100 1 2017 1
800100 0 2018 .
800100 0 2019 .
800119 0 2008 .
800119 0 2009 .
800119 0 2010 .
800119 0 2011 .
800119 0 2012 .
800119 0 2013 .
800119 0 2014 .
800119 0 2015 .
800125 0 2008 .
800125 0 2009 .
800125 0 2010 .
800125 0 2011 .
800125 0 2012 .
800125 0 2013 .
800125 0 2014 .
800131 0 2008 .
800131 0 2009 .
800131 0 2010 .
800131 0 2011 .
800131 0 2012 .
800131 0 2013 .
800131 0 2014 .
800131 0 2015 .
end
format %ty _Year

Thanks for the help.

Kind Regards,

Hugo

[Carlo Lazzaro]

Hugo:
you may want to try:

. bysort _nomem_encr: egen wanted=sum (jobloss)

. replace jobloss_anypoint=wanted

. drop wanted

[Hugo Cooke]

Hey Carlo,

Thanks so much for the quick reply, if I can be cheeky and ask for a second (related) question? I have also been trying to differentiate between attributes prior to jobloss and post. In effect creating a further dummy variable which is 1 for individuals post a jobloss (regardless of employement status) and 0 prior? So in the example of data above individual 800100 would be assigned 0 up to 2016 and 1 from 2017 to 2019.

Kind Regards,

Hugo

[Hugo Cooke]

Actually Carlo, I think my original code does what I am asking for. Apologies.

[Carlo Lazzaro]

Hugo:
as far as your last question is concerned, you may want to try:

. bysort _nomem_encr: gen wanted2=jobloss

bysort _nomem_encr: replace wanted2 = wanted2[_n-1] if wanted2[_n-1]==1

[Nick Cox]

#1 #2 See also FAQ https://www.stata.com/support/faqs/d...ble-recording/