*Eitan Hersh
*Using the Validation variables in the 2006 CCES
*October 2008
*Contact: hersh@fas.harvard.edu


gen  report_reg = v3004 
recode report_reg 1=1 2=0 3=.
      *This changes the "I don't knows" to missing values, and creates a dummy variable where 1 means that R reported being registered.
label var report_reg "Reported Registrants"

gen  report_vote = v4004
recode report_vote 2=0 3=. 
      *Again, I omit the "I don't knows" and create a dummy variable for reporting R voted. 
label var report_vote "Reported Voters"

gen valid_vote = 1 if g2006 == "Y"
recode valid_vote . =0
label var valid_vote "Validated Voters"

gen  valid_reg = matched
label var valid_reg "Validated Registrants"

gen  state = v1002
egen quality = mode(g2006sta), by(state)
     *For the most part, all respondents in the same state have the same designation for record quality, which is variable "g2006sta." 
     *In many states, there are a few respondents who are marked as a different quality than the modal value. 
     *Treat all respondents according to the modal quality value in their state, as I have done here.

drop if state == "WI"
     *Although the modal quality value for Wisconsin is a 3, almost all respondents have a missing value for g2006sta. 
     *Leave Wisconsin out of the sample (for now).

gen  weight = v1001
     *Use a-weights.

drop if v3086 == 2
     *Drop non-citizen immigrants

tab report_reg valid_reg if quality ==3 [aw=weight], row col
tab report_vote valid_vote if quality ==3 [aw=weight], row col