*****Prepare data for calculation of EITC, send to TAXSIM, and createdataset for IV

*****Setup

*Baseline without state

**001 is wage income only

**002 is wage plus unearned income

**012, 011 include state taxes and credits

foreach j in 012 011 211 311 411 511 611 711 {

***KEY to datasets***

*hat=0: real data, hat=4: baseline predicted, hat=2,3 needed for Table 4,hat=5,6,7 needed for Table 5

*state=0: no state tax information calculated, state=1: state taxesincluded

*spec=1: excludes unearned income, spec2: includes unearned income

local hat=substr("`j'",1,1)

local state=substr("`j'",2,1)

local spec=substr("`j'",3,1)

use main, replace

sort idchild year

*IMPORTANT NOTE: all income variables are in 1979 dollars, and need to beconverted to nominal dollars before sending through taxsim

*(convert to nominal dollars by multiplying by cpily)

replace earnincrimp=earnincrimp*cpily

replace earnincsimp=earnincsimp*cpily

replace earnincrsimp=earnincrsimp*cpily

replace unearnincrsimp=unearnincrsimp*cpily

replace nontaxincrsimp=nontaxincrsimp*cpily

gen pretaxinc=earnincrsimp+unearnincrsimp+nontaxincrsimp

*****As needed, make variable changes or sample restrictions

*note: one year already subtracted off to reflect that NLSY reports laggedincome

gen trend=year-1986

***Set state variable to missing if necessary (for state=0 samples)

replace state=0 if `state'==0

*****Sample restrictions

gen samp=1

replace samp = (year==1985|year==1987|year==1989|year==1991|year==1993|year==1995|year==1997|year==1999)& (piamatsn!=. | piarecsn!=. | piarersn!=.) if `hat'==2 | `hat'==3 |`hat'==4 | `hat'==5

replace samp = samp |(year==1988|year==1990|year==1992|year==1994|year==1996|year==1998) if `hat'==6| `hat'==7

sort idchild year

xtset idchild year, yearly delta(1)

*create variables for SIV and to define sample

gen test=(2.295/(10000*cpily))*earnincrsimp

gen x0=L2.test==0

gen x1=L2.test

gen x2=x1^2

gen x3=x1^3

gen x4=x1^4

gen x5=x1^5

gen getdiv02=marrlyimp==0 & (L1.marrlyimp==1|L2.marrlyimp==1)

gen getmarr02=marrlyimp==1 & (L1.marrlyimp==0|L2.marrlyimp==0)

gen flag1 = (2.295/(10000*cpily))*pretaxinc<10 &(2.295/(10000*L2.cpily))*L2.pretaxinc<10

gen flag = year>=1989 & !getmarr02 & !getdiv02 & flag1 if`hat'==2 | `hat'==3 | `hat'==4 | `hat'==6 | `hat'==7

replace flag = year>=1987 & !getmarr02 & !getdiv02 & flag1if `hat'==5

keep if samp

*drop poor oversample

keep if ((samprandom&sampnm)|sampnmblack|sampnmhisp) | (year==1994|year==1996|year==1998)

*****Get predicted income

*Need to have earned income (for eitc credit) and unearned income (forlocation on eitc schedule)

gen earninc=.

gen unearninc=.

*Use actual data (for hat=0 samples)

replace earninc=earnincrsimp if `hat'==0

replace unearninc=unearnincrsimp if `hat'==0

*year trend

if `hat'==2 {

  reg test x0 x1 x2 x3 x4 x5 trend ifsamp & flag

  predict earninchat if samp &flag

  replaceearninc=earninchat*((10000*cpily)/2.295) if samp & flag

}

*year dummies

if `hat'==3 {

  reg test x0 x1 x2 x3 x4 x5 yy* ifsamp & flag

  predict earninchat if samp &flag

  replaceearninc=earninchat*((10000*cpily)/2.295) if samp & flag

}

*baseline

if `hat'==4 {

  reg test x0 x1 x2 x3 x4 x5 if samp& flag

  predict earninchat if samp &flag

  replaceearninc=earninchat*((10000*cpily)/2.295) if samp & flag

}

*for Table 5

if `hat'==5 {

  gen f0=F2.test==0

  gen f1=F2.test

  gen f2=f1^2

  gen f3=f1^3

  gen f4=f1^4

  gen f5=f1^5

  reg test f0 f1 f2 f3 f4 f5 if samp& flag

  predict earninchat if samp &flag

  replaceearninc=earninchat*((10000*cpily)/2.295) if samp & flag

}

*for Table 5

if `hat'==6 {

  gen f0=L1.test==0

  gen f1=L1.test

  gen f2=f1^2

  gen f3=f1^3

  gen f4=f1^4

  gen f5=f1^5

  reg test f0 f1 f2 f3 f4 f5 if samp& flag & (year==1987|year==1989|year==1991|year==1993|year==1995|year==1997|year==1999)& (piamatsn!=. | piarecsn!=. | piarersn!=.)

  predict earninchat if samp &flag

  replaceearninc=earninchat*((10000*cpily)/2.295) if samp & flag

}

*for Table 5

if `hat'==7 {

  gen f0=F1.test==0

  gen f1=F1.test

  gen f2=f1^2

  gen f3=f1^3

  gen f4=f1^4

  gen f5=f1^5

  reg test f0 f1 f2 f3 f4 f5 if samp& flag &(year==1987|year==1989|year==1991|year==1993|year==1995|year==1997|year==1999)& (piamatsn!=. | piarecsn!=. | piarersn!=.)

  predict earninchat if samp &flag

  replaceearninc=earninchat*((10000*cpily)/2.295) if samp & flag

}

gen oldyear=year

*****Replace pwages and ui with predicted values earninc and unearninc andsave pre-eitc dataset

replace swages=0

replace pwages=earninc

replace ui=unearninc

*Set non-wage income (unearnincrsimp, which we are feeding through taxsimas "ui") equal to zero if necessary (for spec=1 samples)

replace ui=0 if `spec'==1

drop earninc unearninc

*Save pre-eitc master dataset, using real data, for later merging in futureprogram

if "`j'"=="012" {

  save preeitcinput, replace

}

keep idchild cpily year state mstat depx agex pwages swages dividendsotherprop pensions gssi transfers rentpaid proptax otheritem childcare uidepchild mortgage stcg ltcg oldyear

*Note: don't send idchild to taxsim if using state identifiers

*****Send data to TAXSIM and get after tax income and EITC

taxsim9, replace full

replace year=oldyear

sort idchild year

save taxsimout`j', replace

}

*****Now rename taxsim variables and merge taxsim datasets together

*****Rename variables from taxsim output

local datasets 012 011 211 311 411 511 611 711

foreach i of local datasets {

  use taxsimout`i', replace

  rename v10 fedagi

  rename v25 eitc

  rename v28 fedinctax

  rename v39 stateeitc

  rename v40 statetotcredit

  keep idchild year fedagi eitcfedinctax stateeitc statetotcredit siitax cpily pwages otherprop

  *Put dollar values back in realterms (2000 dollars, so divide by cpily and multiply by 2.295) and divide allmonetary variables by 10,000

  local vars "fedagi eitcfedinctax stateeitc statetotcredit siitax pwages"

    foreach var of varlist `vars' {

    qui replace `var' =`var'*(2.295/(10000*cpily))

  }

  *Create income variables

  geninc=fedagi-fedinctax-siitax+statetotcredit

  gen incnotax=pwages+eitc

  gen tax=fedagi-eitc

  *rename variables with appropriateextension

  local vars "fedagi eitcfedinctax stateeitc statetotcredit siitax pwages inc incnotax tax"

  foreach var of varlist `vars' {

    rename `var' `var'`i'

  }

  save taxsim`i', replace

}

*****Merge all datasets together

*Start with preeitcinput for real data

use preeitcinput, replace

sort idchild year

foreach dataset of local datasets {

  merge idchild year usingtaxsim`dataset'

  tab _merge

  drop _merge

  sort idchild year

}

*Now for additional variables, put back in real terms (into 2000 dollars,so divide by cpily and multiply by 2.295) and divide all monetary variables by10,000

  local vars "ui otherpropnontaxincrsimp earnincrsimp unearnincrsimp"

    foreach var of varlist `vars' {

    replace `var' =`var'*(2.295/(10000*cpily))

  }

*Now for some variables, put into year 2000 dollars in real terms (convertfrom 1979 dollars to 2000 dollars, so multiply by 2.295) and divide allmonetary variables by 10,000

  local vars "totweadimp2"

    foreach var of varlist `vars' {

    replace `var' =`var'/(10000*(1/2.295))

  }

save taxsim-merged, replace

求大神解决，谢谢！！！