cls
**==========================================
** Program Name: Construct GINI index
** Author: Yilin Chen
** Date: 2023-10-01
**------------------------------------------
** Inputs/Ouputs:
* Data files used: 
  * <https://raw.githubusercontent.com/eileenCHEN-9/PhD_2022_2025/main/data/county_predicted.dta>

* Data files created as intermediate product:
*===========================================
cd "/Users/yilinchen/Documents/PhD/thesis/PhD_2022_2025/data"
* ssc install ineqdeco, replace

** 1. Setup
clear all
macro drop _all
capture log close
set more off
set maxvar 32567
version 15

** 2. Import county level dataset
use county_predicted.dta, clear

*Drop Hongkong, Macau, Taiwan and Sansha city in South China Sea (no population)
drop if province_id == 810000 | province_id == 820000 | province_id == 710000
drop if city_id == 460300

*keep if year>2014
*keep if year<2016

*Drop cities with only one county 
egen count_county = count(county_id), by(city_id year)
drop if count_county == 1

*drop if missing(pred_gdppc_county)

*Generate different per capita GDP
gen pred_gdppc_county=exp(lg_gdppc_predicted)
gen totallightpc_county=total_meanlight
label variable pred_gdppc_county "Predicted GDP per capita using NTL(county)"
label variable totallightpc_county "Mean of lights (county)"
label variable count_county "Number of counties per city"

summarize
describe

** 3. Calculations of regional inequality measures using predicted GDP
*GINIW
gen GINIW_pred_GDP_pc = .
egen group = group(city_id year)  

su group, meanonly
qui forval i = 1/`r(max)' {
  qui count if group == `i' & !missing(pred_gdppc_county)
  if r(N) > 0 {
    qui ineqdeco pred_gdppc_county [aw=total_population] if group == `i' & !missing(pred_gdppc_county)
    replace GINIW_pred_GDP_pc = r(gini) if group == `i'
  }
}

drop group

*Generalized Entropy class -1 
gen GE_m1W_pred_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
  qui count if group == `i' & !missing(pred_gdppc_county)
  if r(N) > 0 {
  qui ineqdeco pred_gdppc_county [aw=total_population]  if group == `i'
replace GE_m1W_pred_GDP_pc = r(gem1) if group == `i'	
}
}

drop group

*Generalized Entropy class 0 (mean logarithmic deviation)
gen GE_0W_pred_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(pred_gdppc_county)
  if r(N) > 0 {
qui ineqdeco pred_gdppc_county [aw=total_population]  if group == `i'
replace GE_0W_pred_GDP_pc = r(ge0) if group == `i'	
  }
} 

drop group

*Generalized Entropy class  1 (Theil index)
gen GE_1W_pred_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(pred_gdppc_county)
  if r(N) > 0 {
qui ineqdeco pred_gdppc_county [aw=total_population]  if group == `i'
replace GE_1W_pred_GDP_pc = r(ge1) if group == `i'	
  }
} 

drop group

*COVW GE(2) is half the square of the coefficient of variation.
gen GE_2w_pred_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(pred_gdppc_county)
  if r(N) > 0 {
qui ineqdeco pred_gdppc_county [aw=total_population]  if group == `i'
replace GE_2w_pred_GDP_pc = r(ge2) if group == `i'	
  }
} 
*foreach var of varlist _Icounty_id_* {
 *   drop `var'
*}
drop group
gen COVW_pred_GDP_pc=sqrt(GE_2w_pred_GDP_pc*2)
drop GE_2w_pred_GDP_pc		

** 4. Calculations of regional inequality measures using lights

*replace totallightpc_county = . if totallightpc_county == 0

*GINIW
gen GINIW_light_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(totallightpc_county)
  if r(N) > 0 {
qui ineqdeco totallightpc_county [aw=total_population]  if group == `i'
replace GINIW_light_pc = r(gini) if group == `i'
  }	
} 

drop group

*Generalized Entropy class -1 
gen GE_m1W_light_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(totallightpc_county)
  if r(N) > 0 {
qui ineqdeco totallightpc_county [aw=total_population]  if group == `i'
replace GE_m1W_light_pc = r(gem1) if group == `i'	
  }
}

drop group

*Generalized Entropy class  0 (mean logarithmic deviation)
gen GE_0W_light_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(totallightpc_county)
  if r(N) > 0 {
qui ineqdeco totallightpc_county [aw=total_population]  if group == `i'
replace GE_0W_light_pc = r(ge0) if group == `i'	
  }
} 

drop group

*Generalized Entropy class  1 (Theil index)
gen GE_1W_light_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(totallightpc_county)
  if r(N) > 0 {
qui ineqdeco totallightpc_county [aw=total_population]  if group == `i'
replace GE_1W_light_pc = r(ge1) if group == `i'	
  }
} 

drop group

*COVW GE(2) is half the square of the coefficient of variation.
gen GE_2w_light_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(totallightpc_county)
  if r(N) > 0 {
qui ineqdeco totallightpc_county [aw=total_population]  if group == `i'
replace GE_2w_light_pc = r(ge2) if group == `i'	
  }
} 

drop group
gen COVW_light_pc=sqrt(GE_2w_light_pc*2)
drop GE_2w_light_pc		

** 5. Calculations of regional inequality measures using observed GDP

replace county_rgdppc = . if county_rgdppc == 0

*GINIW
gen GINIW_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(county_rgdppc)
  if r(N) > 0 {
qui ineqdeco county_rgdppc [aw=total_population]  if group == `i'
replace GINIW_GDP_pc = r(gini) if group == `i'
  }	
} 

drop group

*Generalized Entropy class -1 
gen GE_m1W_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(county_rgdppc)
  if r(N) > 0 {
qui ineqdeco county_rgdppc [aw=total_population]  if group == `i'
replace GE_m1W_GDP_pc = r(gem1) if group == `i'	
  }
}

drop group

*Generalized Entropy class  0 (mean logarithmic deviation)
gen GE_0W_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(county_rgdppc)
  if r(N) > 0 {
qui ineqdeco county_rgdppc [aw=total_population]  if group == `i'
replace GE_0W_GDP_pc = r(ge0) if group == `i'	
  }
} 

drop group

*Generalized Entropy class  1 (Theil index)
gen GE_1W_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(county_rgdppc)
  if r(N) > 0 {
qui ineqdeco county_rgdppc [aw=total_population]  if group == `i'
replace GE_1W_GDP_pc = r(ge1) if group == `i'	
  }
} 

drop group

*COVW GE(2) is half the square of the coefficient of variation.
gen GE_2w_GDP_pc=.
egen group = group(city_id year)
su group, meanonly
qui forval i = 1/`r(max)' {
	qui count if group == `i' & !missing(county_rgdppc)
  if r(N) > 0 {
qui ineqdeco county_rgdppc [aw=total_population]  if group == `i'
replace GE_2w_GDP_pc = r(ge2) if group == `i'	
  }
} 

drop group
gen COVW_GDP_pc=sqrt(GE_2w_GDP_pc*2)
drop GE_2w_GDP_pc	

*There are negative inequality measurement due the missing values, so we replace them with . (missing)	
replace GINIW_light_pc = . if GINIW_light_pc < 0
replace GE_m1W_light_pc = . if GE_m1W_light_pc < 0
replace GE_0W_light_pc = . if GE_0W_light_pc < 0
replace GE_1W_light_pc = . if GE_1W_light_pc < 0
replace COVW_light_pc = . if COVW_light_pc < 0

replace GINIW_GDP_pc = . if GINIW_GDP_pc < 0
replace GE_m1W_GDP_pc = . if GE_m1W_GDP_pc < 0
replace GE_0W_GDP_pc = . if GE_0W_GDP_pc < 0
replace GE_1W_GDP_pc = . if GE_1W_GDP_pc < 0
replace COVW_GDP_pc = . if COVW_GDP_pc < 0

*Create new id by city by year
gen id_t_j = string(year) + city

*Aggregate city-level GINI index
collapse (first) city_id year GINIW_pred_GDP_pc - COVW_GDP_pc, by(id_t_j)
sort city_id year 		
drop id_t_j
					
save "/Users/yilinchen/Documents/PhD/thesis/PhD_2022_2025/data/City_Inequality_Data.dta", replace

*Collapse to cross section 	
collapse (mean) GINIW_pred_GDP_pc - COVW_GDP_pc, by(city_id)						
describe
summarize

*Correlations		
pwcorr  GINIW_pred_GDP_pc 		GINIW_GDP_pc 		GINIW_light_pc			
pwcorr  GE_m1W_pred_GDP_pc 		GE_m1W_GDP_pc 		GE_m1W_light_pc
pwcorr  GE_0W_pred_GDP_pc 		GE_0W_GDP_pc 		GE_0W_light_pc
pwcorr  GE_1W_pred_GDP_pc 		GE_1W_GDP_pc 		GE_1W_light_pc
pwcorr  COVW_pred_GDP_pc 		COVW_GDP_pc 		COVW_light_pc