*==============================================================================
* FILE: 1F-Variables
* Researcher: Julian B. Adam
* Version: July 2018
*
* Project: Works Councils reform act
*
* Data used: Establishment panel 98-04
*
* Description: This program harmonizes the industry classifications
*              99 and 00. It generates variables necessary for the econometric
*              analysis.
*
*==============================================================================



clear
set more off
set linesize 120

global main	= "W:\BetriebsratStata\Daten\StataIABcompil_erw_panel"

use "$orig\hvpi_dta.dta", clear
recast int jahr
save "$data\hvpi_dta.dta", replace


use "$data\panel.dta", clear
cap log close
log using "$log\1F-Variables.log", replace
label lang de



drop if jahr==2014

*-------------------------------------------------------------------------------
* merge with dataset containing harmonized consumer price index (Eurostat)

merge m:1 jahr using "$orig\hvpi_dta.dta"



*------------------------------------------------------------------------------
* change industry classification to avoid small case numbers, all services in (7)


*transforming missings
mvdecode _all, mv(-9)
mvdecode _all, mv(-8)

*vorleist takes values>100, set all values >100 to 100
replace vorleist=100 if vorleist>100
label var jahr "year"

*redefining industries, harmonization
gen bran99=.
replace bran99=1 if bran_n99==1|bran_n99==2
replace bran99=2 if bran_n99>2 & bran_n99<17
replace bran99=3 if bran_n99==17|bran_n99==18
replace bran99=4 if bran_n99==19
replace bran99=5 if bran_n99==20
replace bran99=6 if bran_n99==21|bran_n99==22
replace bran99=7 if bran_n99>22 & bran_n99<36
replace bran99=8 if bran_n99>35
replace bran99=. if bran_n99==.

label var bran99 "industry until 1999, 8-classification"

gen bran00=.
replace bran00=1 if bran_n00==1|bran_n99==2
replace bran00=2 if bran_n00>2 & bran_n00<19
replace bran00=3 if bran_n00==19|bran_n00==20
replace bran00=4 if bran_n00==21|bran_n00==22|bran_n00==23
replace bran00=5 if bran_n00==24|bran_n00==25
replace bran00=6 if bran_n00==26|bran_n00==27
replace bran00=7 if bran_n00>27 & bran_n00<39
replace bran00=8 if bran_n00>38
replace bran00=. if bran_n00==.

label var bran00 "industry from 2000, 8-classification"

*merging the 2 variables
replace bran99=0 if bran99==.
replace bran00=0 if bran00==.
gen bran=bran99+bran00
replace bran=. if bran==0
label var bran "industry, 8-classification"

*------------------------------------------------------------------------------
* Generating treatment indicators; post if year> 2001
gen post=0
replace post=1 if jahr>2001
label var post "year after 2001"

* generating indicator whether establishment has size of treatment for
* increase of the council
gen incr_counc=0
replace incr_counc=1 if gesamt>100 & gesamt<151
replace incr_counc=1 if gesamt>200 & gesamt<301
replace incr_counc=1 if gesamt>400 & gesamt<601
replace incr_counc=1 if gesamt>700 & gesamt<1001
replace incr_counc=1 if gesamt>1500 & gesamt<2001
replace incr_counc=1 if gesamt>2500 & gesamt<3001
replace incr_counc=1 if gesamt>3500 & gesamt<4001
replace incr_counc=1 if gesamt>4500 & gesamt<5001
replace incr_counc=1 if gesamt>4000 & gesamt<9001
replace incr_counc=1 if gesamt>9000

label var incr_counc "increase WoCo 0/1"

gen incr_counc_number=.
replace incr_counc_number=2 if incr_counc==1 & gesamt<6001
replace incr_counc_number=4 if incr_counc==1 & gesamt>6000
replace incr_counc_number=0 if incr_counc==0
gen incr_counc_number_empl_prop=incr_counc_number/gesamt
label var incr_counc "increase WoCo, number"
label var incr_counc_number_empl "increase WoCo, rel. to no. of employees"

* generating variable: change in number of councilors relative to previous
* council size
gen incr_counc_prop=0
replace incr_counc_prop=2/5 if gesamt>100 & gesamt<151
replace incr_counc_prop=2/7 if gesamt>200 & gesamt<301
replace incr_counc_prop=2/9 if gesamt>400 & gesamt<601
replace incr_counc_prop=2/11 if gesamt>700 & gesamt<1001
replace incr_counc_prop=2/15 if gesamt>1500 & gesamt<2001
replace incr_counc_prop=2/19 if gesamt>2500 & gesamt<3001
replace incr_counc_prop=2/23 if gesamt>3500 & gesamt<4001
replace incr_counc_prop=2/27 if gesamt>4500 & gesamt<5001
replace incr_counc_prop=2/29 if gesamt>5000 & gesamt<6001
replace incr_counc_prop=4/29 if gesamt>6000 & gesamt<7001
replace incr_counc_prop=4/31 if gesamt>7000 & gesamt<9001
replace incr_counc_prop=4/33 if gesamt>9000 & gesamt<12001
replace incr_counc_prop=4/35 if gesamt>12000 & gesamt<15001
replace incr_counc_prop=4/37 if gesamt>15000 & gesamt<18001
replace incr_counc_prop=4/39 if gesamt>18000 & gesamt<21001
replace incr_counc_prop=4/41 if gesamt>21000 & gesamt<24001
replace incr_counc_prop=4/45 if gesamt>24000 & gesamt<27001
replace incr_counc_prop=4/47 if gesamt>27000 & gesamt<30001
replace incr_counc_prop=4/49 if gesamt>30000 & gesamt<33001
replace incr_counc_prop=4/51 if gesamt>36000 & gesamt<39001
replace incr_counc_prop=4/53 if gesamt>39000 & gesamt<42001
replace incr_counc_prop=4/55 if gesamt>42000 & gesamt<45001
replace incr_counc_prop=4/57 if gesamt>45000 & gesamt<48001
replace incr_counc_prop=4/59 if gesamt>48000 & gesamt<51001
label var incr_counc_prop "increase WoCo, prop."


*defining size classes based on no. of employees
gen treat_group=.
replace treat_group=1 if gesamt<101
replace treat_group=2 if gesamt>100 & gesamt<151
replace treat_group=3 if gesamt>150 & gesamt<201
replace treat_group=4 if gesamt>200 & gesamt<301
replace treat_group=5 if gesamt>300 & gesamt<401
replace treat_group=6 if gesamt>400 & gesamt<501
replace treat_group=7 if gesamt>500 & gesamt<601
replace treat_group=8 if gesamt>600 & gesamt<701
replace treat_group=9 if gesamt>700 & gesamt<901
replace treat_group=10 if gesamt>900 & gesamt<1001
replace treat_group=11 if gesamt>1000 & gesamt<1501
replace treat_group=12 if gesamt>1500 & gesamt<2001
replace treat_group=13 if gesamt>2000 & gesamt<2501
label var treat_group "size class (current)"

*above 9000 employees, there was no change in number of councilors.



*variable tarif (collective agreement) is not coded 0/1
*shall take value 1 if sectoral or companywide collective agreement
gen tarif_1=.
replace tarif_1=1 if tarif==1 |tarif==2
replace tarif_1=0 if tarif==3
label var tarif_1 "collective agreement 1/0"

*generate gross pay per capita
gen pc_pay=.
replace pc_pay=lohn/gesamt
label var pc_pay "avg. pay per capita June"

* set as panel data
xtset idnum jahr

* deflating wage sum, base year: 2001, deflate by HVPI EuroStat

gen rlohn=lohn/hvpi  if jahr>1997 & jahr<2005
label var rlohn "real wage cost p.c."

*generation pay per capita in real terms
gen rpc_pay=rlohn/gesamt
label var rpc_pay "avg. real wage per worker in June"

*generation female quota
gen frau_quota=ges_frau/gesamt
label var frau_quota "Frauenquote"

*XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX


gen sales_pc=geschvol/gesamt
label var sales_pc "sales per employee"
gen lgeschvol=ln(geschvol)

gen kuend_ratio=kuend_an/gesamt
label var kuend_ratio "quit ratio"

sort idnum jahr

*XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX


* generate a variable that denotes sample
* only establishment that do not change WoCo status
bysort idnum: egen betrratsum=sum(betrrat)
bysort idnum: egen betrratmean=mean(betrrat)
gen sample=0
replace sample=1 if betrratmean==1 & betrrat!=.& jahr<2005 & geschart<4 & treat_group<5 ///
& treat_group>2 & gesamt!=. & tarif_1!=. & geschvol!=. & bran!=. & kuend_an!=. ///
& geschart!=. & treat_group!=. & gesamt>150 & gesamt<251

gen sample_treat=.
replace sample_treat=0 if sample==1 & treat_group==3
replace sample_treat=1 if sample==1 & treat_group==4
label var sample_treat "treated/ untreated in sample 0/1"


* generating indicator for treatment group and treatment dummy: establishments
* with 200-300 establishments after 2001
gen treat4=.
replace treat4=1 if treat_group==4
replace treat4=0 if treat_group!=4
gen treatpost=treat4*post
gen treatposttarif=treat4*post*tarif_1

*productivity measure: value added per worker
gen va=geschvol-vorleist
label var va "value added total"

gen va_pc=va/gesamt
label var va_pc "value added per worker"

gen lva_pc=ln(va_pc)

gen lrlohn=ln(rlohn)

bysort sample_treat: sum betrrat west gesamt geschvol lgeschvol va_pc lva_pc ///
rlohn lrlohn tarif_1 invest kuend_an entlass frau einvern befrist azubi ertlag ///
uebtarif kuend_ratio

bysort sample:sum betrrat west gesamt geschvol lgeschvol va_pc lva_pc ///
rlohn lrlohn tarif_1 invest kuend_an entlass frau einvern befrist azubi ertlag ///
uebtarif kuend_ratio

bysort sample_treat: count

*###############################################################################

* manually include dummies for spec.(7) and (8)
quietly tabulate jahr, generate(j)
rename j1 j98
rename j2 j99
rename j3 j00
rename j4 j01
rename j5 j02
rename j6 j03
rename j7 j04
rename j8 j05

gen j98treat=j98*treat4
gen j99treat=j99*treat4
gen j00treat=j00*treat4
gen j01treat=j01*treat4
gen j02treat=j02*treat4
gen j03treat=j03*treat4
gen j04treat=j04*treat4

gen j98treattarif=j98*treat4*tarif_1
gen j99treattarif=j99*treat4*tarif_1
gen j00treattarif=j00*treat4*tarif_1
gen j01treattarif=j01*treat4*tarif_1
gen j02treattarif=j02*treat4*tarif_1
gen j03treattarif=j03*treat4*tarif_1
gen j04treattarif=j04*treat4*tarif_1

*###############################################################################
* for robustness check: do firms changing collective agreement status make
* a difference?
bysort idnum: egen tarif_change=mean(tarif) if tarif!=.
gen TV=.
replace TV=1 if tarif_change==2
replace TV=1 if tarif_change==1
replace TV=0 if tarif_change==3
gen treatposttarif_a=treat4*post*TV

* variable for checking whether est. changed coll. agr. status
bysort idnum: egen tarif_1_check=mean(tarif_1)
tab tarif_1_check if sample==1

sum kuend_an ertlag gesamt tarif_1 west invest einvern rpc_pay frau_quota sales_pc if sample==1 & tarif_1_check==1 | sample==1 & tarif_1_check==0
sum kuend_an ertlag gesamt tarif_1 west invest einvern rpc_pay frau_quota sales_pc if sample==1 & tarif_1_check>0 & tarif_1_check<1

*###############################################################################
* left-censoring: how many est. have 0 quits?

tab kuend_an if sample==1

*###############################################################################
* rescale variable rpc_pay

gen rpc_pay_1000=rpc_pay/1000

save "$data\panel_1_F.dta", replace
log cl