****these lines just once to build the large dataset    
****these lines just once to build the large dataset => saved as "$mydata/prosoc/tmplis15"   

*create the extract dataset from lis* 
*we have more variables than needed but ...*
*we keep a set of 16 advanced countries on the time span 2010-today*

lissyuse, pvars(dname hid ppopwgt age sex relation) ///    
hvars(hid cname year iso2 hpopwgt  dhi hifactor nhhmem region_c ) ///    
lis iso2(at au be ca de dk es fi fr il it lu nl no uk us) from(2010) to(2020)                        
   
  
save  "$mydata/prosoc/tmplis15"  , replace              
****these lines just once to build the large dataset      
****these lines just once to build the large dataset      
       
       
****since the extract "$mydata/prosoc/tmplis15" already exists we can begin with this...     
  
use  "$mydata/prosoc/tmplis15"  , replace              
  
*useful variables such as country-year code, equivalised pre and post distribution household HH income...   
gen ccyyyy=iso2+string(year)       
gen gdhi=.       
gen ghif=.       
gen edhi=dhi/sqrt(nhhmem)       
gen ehif=hif/sqrt(nhhmem)       
  
  
* we keep only the latest year for each country ...   
bysort iso2: egen maxy=max(year)      
keep if year==maxy      
drop maxy       
  
*tricks: here is a series of standard treatment for LIS key figures compatibility...   
  
*drop "bad" disposable incomes    
drop if dhi==. | dhi==0    
* select only if there is a weight     
gen hwgt=hpopwgt    
drop if hwgt==. | hwgt==0     

* trick: create person weight as hwgt times number of household member     
* trick: this weight is non-zero and works "almost all the time" as frequency weight   
* BE CAREFUL = ONLY DESCRIPTIVE WEIGHT, NOT TO BE USED FOR REGRESSIONS, MODELS, etc..
generate wt=int(hwgt*nhhmem *100)    

*keep head of household HH, in "non elderly HH"   
keep if relation==1000       
keep if age<65      
              
*for each country-year we compute edhi equivalised disposable hh income  
*same thing for ghif, = factor income = before redistribution  
*trick: recode ccyyyy from 1 to max, max= total nb of countries-years...  
  
encode ccyyyy, gen(nname)       
su nname       
local max=r(max)       
       
*trick: we make use of the lis-key-figures top and bottom codings...  
*gini is evaluated through the SJ+PVK ineqdec0 command ...   
*edhi income gini is then saved in a variable=  
  
forvalues i=1/`max' {    
  
quietly {  
sum edhi [w=wt] if nname==`i'    
generate botlin=0.01*_result(3) if nname==`i'    
replace edhi=botlin if edhi<botlin &  nname==`i'    
quietly sum dhi [w=wt]  if nname==`i', de    
generate toplin=10*_result(10)  if nname==`i'    
replace edhi=(toplin/(nhhmem^0.5)) if dhi>toplin  & nname==`i'    
drop botlin toplin    
       
ineqdec0 edhi [w=wt] if nname==`i'       
replace gdhi= r(gini) if nname==`i'        
ineqdec0 ehif [w=wt] if nname==`i'       
replace ghif= r(gini) if nname==`i'        
}  
}       
      
*we then collapse by ccyyyy  
collapse gdhi ghif, by(ccyyyy)      
     
*we then graph the histograms  
*(plenty of tricks to order the ccyyyy)  
  
graph hbar ghif gdhi , over(ccyy, sort(2) descending) scale(.6)  blabel(bar, position(outside) format(%5.2f))  legend(off)       
  
*important: this command to have the graph on lissy...  
graphexportpdf $mypdf/graphteststata.pdf, replace