Datasets:

chuxuan
/

REPRO-Bench

clear all
set more off
set maxvar 10000
set mem 500m

*Set path
***

*cd ---- specify path to \data_replication\psid\

local procdata  "./proc_data/"
local rawdata  "./raw_data/"

cd proc_data

*load income data
forval x=2015(2)2017{
use pX`x'.dta, clear
ren aftertax_income_defl_equiv inc`x'
ren age_head age`x'
ren ed_head edu`x'
ren marital_status_head mstat`x'
ren fam_ID_year fam_ID`x'
ren size_fam size_fam`x'
drop fam_ID_1968 year_survey year mstat //drop variables that are not useful
save pkX`x'.dta,replace
}

*Join datasets to form a panel
use pkX2015.dta, clear
forval x=2017(2)2017{
joinby hid using pkX`x'.dta, unmatched(both)
drop if _merge!=3
drop _merge
local i=`x'-2
di "`i'"
gen b=age`x'-age`i'
tab b  
drop if b!=2 & b!=1 & b!=3   // drop people that are inconsistent about their age (1/2/3 are consistent due to the differences in month of interview)
drop b
gen a=edu`x'-edu`i'
tab a 
drop if a<0     // drop people that are inconsistent about their education (to adjust for any errors, see 2009 family codebook release info)
drop a
}

*Reshape the data for better use
reshape long inc age edu mstat fam_ID size_fam, i(hid) j(year) 



drop if inc<0


drop if year<2015
drop if year>2017
drop if age<25
drop if age>85

keep fam_ID year hid inc equiv age
save income_panel1_temp.dta, replace

cd ..
*###############################################################################
*Prepare the refinancing data for each year
*###############################################################################
forval x=2015(2)2017{
clear
*###############################################################################
*Prepare the wealth data
*###############################################################################

*Convert family files to stata format
cd raw_data
do FAM`x'ER.do

*Adjust wealth components
do wealth_comp`x'.do
gen year=`x'

drop ER*

cd ..
joinby fam_ID year using `procdata'income_panel1_temp.dta, unmatched(both)
drop if _merge!=3
drop _merge 

save `procdata'ref_temp_`x'.dta,replace

}

cd proc_data

*###############################################################################
*Generate Panel and calculate Share of refinancing
*###############################################################################

use ref_temp_2015.dta, clear
forval x=2017(2)2017{
append using ref_temp_`x'.dta
}

sort hid year
xtset hid year

*sum up both mortgages
gen tot_mtg_debt=m1_principal+m2_principal

*generate difference in mortgage balances between survey waves
gen past_mtg_balance=L2.tot_mtg_debt
gen diff_mtg_balance=tot_mtg_debt-past_mtg_balance
replace diff_mtg_balance=. if past_mtg_balance==0    //make sure to account for people that had no mtg and got one (went from 0 to positive mtg balance)

*generate refinancing dummies
*1. any increase in mtg balance
gen refi_status1=1 if diff_mtg_balance>0 & diff_mtg_balance!=.
replace refi_status1=0 if refi_status1!=1
*2. 5% increase in mtg balance (results are the same in terms of nr of observations between option 1. and option 2.)
gen pct_increase_mtg_balance=diff_mtg_balance/past_mtg_balance
gen refi_status2=1 if diff_mtg_balance>=.05 & diff_mtg_balance!=.
replace refi_status2=0 if refi_status2!=1

****************************************************************************************************************
*compute moments
****************************************************************************************************************
*create moments to be reported
gen liq_assets=(stocks+checking_savings-other_debt)/equiv
gen income = inc   //already equivalized
gen liq_assets_to_inc = liq_assets/income
gen h_net_worth_share = (house_val - m1_principal-m2_principal)/wealth
gen wlth = wealth/equiv
gen ltv_m=tot_mtg_debt/house_val
gen h_value=house_val/equiv
*gen age_head=age
gen mtg_payments_total = (mtg_paym1+mtg_paym2)/equiv
gen pti = mtg_payments_total/inc
*save file
save temp_refi_data.dta, replace





*************************************************************************
* OTHER REFI MOMENTS
*************************************************************************
*^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
*2017
*^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
use temp_refi_data, clear
keep if year==2015 | year==2017
*move refi status one step behind
gen refit1=f2.refi_status2
gen refit2=f2.turnover1
gen refit=1 if refit1==1 & refit2!=1
replace refit=0 if refit!=1
*move back increase in mtg balance
gen mtg_delta=f2.pct_increase_mtg_balance
gen mtg_delta1=f2.diff_mtg_balance/house_val
*keep only orginal year
keep if year==2015
*winsorize data
winsor2 liq_assets income liq_assets_to_inc h_net_worth_share wlth ltv_m h_value age_head pti mtg_delta mtg_delta1, replace cuts(1 99)


preserve
collapse (mean) refit 
gen res=round(refit, 0.01)/2
replace res=round(res, 0.01)
gen title="fraction of people refinance"
save moment1.dta, replace
restore


forval x=1(1)1{
use moment`x'.dta, clear
keep res title
duplicates drop
save moment`x'.dta, replace
}

use moment1.dta, clear

order title res

cd ..

cd output

outsheet using tabA3_refi.csv, replace c

	clear all
	set more off
	set maxvar 10000
	set mem 500m

	*Set path
	***

	*cd ---- specify path to \data_replication\psid\

	local procdata "./proc_data/"
	local rawdata "./raw_data/"

	cd proc_data

	*load income data
	forval x=2015(2)2017{
	use pX`x'.dta, clear
	ren aftertax_income_defl_equiv inc`x'
	ren age_head age`x'
	ren ed_head edu`x'
	ren marital_status_head mstat`x'
	ren fam_ID_year fam_ID`x'
	ren size_fam size_fam`x'
	drop fam_ID_1968 year_survey year mstat //drop variables that are not useful
	save pkX`x'.dta,replace
	}

	*Join datasets to form a panel
	use pkX2015.dta, clear
	forval x=2017(2)2017{
	joinby hid using pkX`x'.dta, unmatched(both)
	drop if _merge!=3
	drop _merge
	local i=`x'-2
	di "`i'"
	gen b=age`x'-age`i'
	tab b
	drop if b!=2 & b!=1 & b!=3 // drop people that are inconsistent about their age (1/2/3 are consistent due to the differences in month of interview)
	drop b
	gen a=edu`x'-edu`i'
	tab a
	drop if a<0 // drop people that are inconsistent about their education (to adjust for any errors, see 2009 family codebook release info)
	drop a
	}

	*Reshape the data for better use
	reshape long inc age edu mstat fam_ID size_fam, i(hid) j(year)



	drop if inc<0


	drop if year<2015
	drop if year>2017
	drop if age<25
	drop if age>85

	keep fam_ID year hid inc equiv age
	save income_panel1_temp.dta, replace

	cd ..
	*###############################################################################
	*Prepare the refinancing data for each year
	*###############################################################################
	forval x=2015(2)2017{
	clear
	*###############################################################################
	*Prepare the wealth data
	*###############################################################################

	*Convert family files to stata format
	cd raw_data
	do FAM`x'ER.do

	*Adjust wealth components
	do wealth_comp`x'.do
	gen year=`x'

	drop ER*

	cd ..
	joinby fam_ID year using `procdata'income_panel1_temp.dta, unmatched(both)
	drop if _merge!=3
	drop _merge

	save `procdata'ref_temp_`x'.dta,replace

	}

	cd proc_data

	*###############################################################################
	*Generate Panel and calculate Share of refinancing
	*###############################################################################

	use ref_temp_2015.dta, clear
	forval x=2017(2)2017{
	append using ref_temp_`x'.dta
	}

	sort hid year
	xtset hid year

	*sum up both mortgages
	gen tot_mtg_debt=m1_principal+m2_principal

	*generate difference in mortgage balances between survey waves
	gen past_mtg_balance=L2.tot_mtg_debt
	gen diff_mtg_balance=tot_mtg_debt-past_mtg_balance
	replace diff_mtg_balance=. if past_mtg_balance==0 //make sure to account for people that had no mtg and got one (went from 0 to positive mtg balance)

	*generate refinancing dummies
	*1. any increase in mtg balance
	gen refi_status1=1 if diff_mtg_balance>0 & diff_mtg_balance!=.
	replace refi_status1=0 if refi_status1!=1
	*2. 5% increase in mtg balance (results are the same in terms of nr of observations between option 1. and option 2.)
	gen pct_increase_mtg_balance=diff_mtg_balance/past_mtg_balance
	gen refi_status2=1 if diff_mtg_balance>=.05 & diff_mtg_balance!=.
	replace refi_status2=0 if refi_status2!=1

	****************************************************************************************************************
	*compute moments
	****************************************************************************************************************
	*create moments to be reported
	gen liq_assets=(stocks+checking_savings-other_debt)/equiv
	gen income = inc //already equivalized
	gen liq_assets_to_inc = liq_assets/income
	gen h_net_worth_share = (house_val - m1_principal-m2_principal)/wealth
	gen wlth = wealth/equiv
	gen ltv_m=tot_mtg_debt/house_val
	gen h_value=house_val/equiv
	*gen age_head=age
	gen mtg_payments_total = (mtg_paym1+mtg_paym2)/equiv
	gen pti = mtg_payments_total/inc
	*save file
	save temp_refi_data.dta, replace





	*************************************************************************
	* OTHER REFI MOMENTS
	*************************************************************************
	*^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	*2017
	*^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	use temp_refi_data, clear
	keep if year==2015 \| year==2017
	*move refi status one step behind
	gen refit1=f2.refi_status2
	gen refit2=f2.turnover1
	gen refit=1 if refit1==1 & refit2!=1
	replace refit=0 if refit!=1
	*move back increase in mtg balance
	gen mtg_delta=f2.pct_increase_mtg_balance
	gen mtg_delta1=f2.diff_mtg_balance/house_val
	*keep only orginal year
	keep if year==2015
	*winsorize data
	winsor2 liq_assets income liq_assets_to_inc h_net_worth_share wlth ltv_m h_value age_head pti mtg_delta mtg_delta1, replace cuts(1 99)


	preserve
	collapse (mean) refit
	gen res=round(refit, 0.01)/2
	replace res=round(res, 0.01)
	gen title="fraction of people refinance"
	save moment1.dta, replace
	restore


	forval x=1(1)1{
	use moment`x'.dta, clear
	keep res title
	duplicates drop
	save moment`x'.dta, replace
	}

	use moment1.dta, clear

	order title res

	cd ..

	cd output

	outsheet using tabA3_refi.csv, replace c