# Libraries ----------------- library(ggplot2) library(cowplot) library(RColorBrewer) library(reshape2) library(forcats) library(knitr) library(kableExtra) library(rlang) library(wesanderson) library(haven) library(margins) library(broom) library(stargazer) library(scales) library(dplyr) library(tidyr) library(xlsx) detach(package:mfx) detach(package:MASS) #### 1 Preparing data #### a<-readRDS("gi.rds") #### 1a Independent variables #### #age table(a$age) a$age<-ifelse(a$age<0, NA, a$age) #Age a<-a %>% mutate(age=ifelse(age<=2, TRUE, FALSE)) table(a$country, a$age) # gender, factor a<-a %>% mutate(female=case_when(.$gender==1 ~ 1, #female .$gender==0 ~ 0, #male .$gender==2 ~ 2, #prefer not to say & missings .$gender==-99 ~ 2)) %>% mutate(female = factor(female)) # partner table(a$partner, useNA = "always") # Todo later in general file: na of natlang_partner in INDIA (4 cases) a<-a %>% mutate(natlang_partner2=ifelse(is.na(natlang_partner1), natlang_partner3, natlang_partner2), natlang_partner3=ifelse(is.na(natlang_partner1), "-9", natlang_partner3), natlang_partner1=ifelse(is.na(natlang_partner1), "Hindu", natlang_partner1)) a<-a %>% mutate(natlang_partner3=ifelse(is.na(natlang_partner2), "-9", natlang_partner3), natlang_partner2=ifelse(is.na(natlang_partner2), "English", natlang_partner2)) a<-a %>% mutate(natlang_partner3=ifelse(is.na(natlang_partner3), "-9", natlang_partner3)) a<-a %>% mutate(part= ifelse(partner<0, NA, ifelse(partner==2, 0, ifelse(natlang_partner1=="German" | natlang_partner2=="German" | natlang_partner3=="German", 1, 2)))) with(a[a$natlang_partner1=="German",], table(a$natlang_partner2, a$natlang_partner3)) table(a$natlang_partner1, a$partner) table(a$natlang_partner2, a$partner) table(a$natlang_partner3, a$partner) table(a$partner, a$part, useNA = "always") # children table(a$children) a<-a %>% mutate(children_d=case_when(.$children<0 ~ NA, .$children==0 ~ FALSE, .$children>0 ~ TRUE), children_d_na=case_when(.$children==-999 ~ NA, .$children==-99 ~ FALSE, .$children==0 ~ FALSE, .$children>0 ~ TRUE)) table(a$children, a$children_d_na, useNA = "always") # risk/patience/applic_gen/comm_skills table(a$risk, useNA="always") table(a$patient, useNA="always") table(a$applic_gen, useNA="always") table(a$mig_phase, useNA="always") table(a$mig_decision, useNA="always") na <- function(x) (ifelse(x<0, NA, x)) a<-a %>% mutate_at(c("risk", "patient", "applic_gen", "mig_decision"), na) # degree table(a$deg_participant) a<-a %>% mutate(degree=deg_participant) %>% mutate(degree=replace(degree, .$deg_participant==5 & .$occupation_participant==0, 0), #pupils degree=replace(degree, .$deg_participant< 0 & .$occupation_participant==0, 0), #pupils degree=replace(degree, .$deg_participant==5 & .$occupation_participant==1, 2), #students degree=replace(degree, .$deg_participant< 0 & .$occupation_participant==1, 2)) %>% #students mutate(degree=replace(degree, .$degree==-999 | .$degree==-99, NA)) #Nas table(a$degree, a$deg_participant, useNA = "always") # labour market status table(a$occupation_participant, a$empstat, useNA="always") a<-a %>% mutate(empstat=case_when(.$occupation_participant<0 ~ 3, .$occupation_participant==12 ~ 3, .$occupation_participant==11 ~ 3, .$occupation_participant==0 ~ 1, .$occupation_participant==1 ~ 1, .$occupation_participant==2 ~ 1, .$occupation_participant==3 ~ 1, .$occupation_participant==4 ~ 2, .$occupation_participant==5 ~ 2, .$occupation_participant==6 ~ 2, .$occupation_participant==7 ~ 2, .$occupation_participant==8 ~ 3, .$occupation_participant==9 ~ 3, .$occupation_participant==10 ~3)) # English speaker (dummy) a<-a %>% mutate(english=case_when((lang_1name == "English" & lang_1level >= 4) ~ 1, (lang_2name == "English" & lang_2level >= 4) ~ 1, (lang_3name == "English" & lang_3level >= 4) ~ 1, (lang_1name == "English" & lang_1level < 4) ~ 0, (lang_2name == "English" & lang_2level < 4) ~ 0, (lang_3name == "English" & lang_3level < 4) ~ 0, (lang_1name != "English" & lang_2name != "English" & lang_3name != "English") ~ 0, natlang_participant1 == "English" ~ 1, natlang_participant2 == "English" ~ 1, natlang_participant3 == "English" ~ 1, natlang_participant4 == "English" ~ 1 )) a$english <- as.logical(a$english) # International applicability of skills a<-a %>% mutate(applic_gen_d=case_when(applic_gen <= 3 ~ 0, applic_gen > 3 ~ 1)) a$applic_gen_d <- as.logical(a$applic_gen_d) a<-a %>% mutate(occupationac=case_when((empstat == 2 & comm_skills >=5 & applic_gen >=4 ~ 2), #in occupation, high app, high comm. (empstat == 2 & comm_skills < 5 & comm_skills >= 1 & applic_gen >=4 ~ 1), #in occupation, high app, low comm. (empstat == 2 & comm_skills >= 1 & applic_gen <4 ~ 0), #in occupation, low app. (empstat == 1 ~ 3), #in education (empstat == 3 ~ 4) #other occ./n/a )) a$occupationac <- as.factor(a$occupationac) # (Main) Reason for migration a <- a %>% mutate(mainrea_educ = ifelse((mig_phase>=1 & mreason_mig1 == 0), 1, ifelse((mig_phase<0| mreason_mig1<0), 2, 0)), mainrea_labor = ifelse((mig_phase>=1 & (mreason_mig1 >= 1 & mreason_mig1 <= 6)), 1, ifelse((mig_phase < 0 | mreason_mig1 < 0), 2, 0)), mainrea_other = ifelse((mig_phase>=1 & (mreason_mig1 >= 7 & mreason_mig1 <= 14)), 1, ifelse((mig_phase < 0 | mreason_mig1 < 0), 2, 0)), mainrea_educlab = ifelse((mainrea_educ == TRUE | mainrea_labor == TRUE), 1, ifelse((mig_phase < 0 | mreason_mig1 < 0), 2, 0)) ) # Add further independent variables here indv<- a %>% select(country, institute, id_uni, age, female, part, children_d_na, risk, patient, degree, occupationac, english, mainrea_educ, mainrea_labor, mainrea_other, mainrea_educlab, mreason_mig1, mig_phase) # Check for share of NAs apply(indv[,4:ncol(indv)], 2, function(x) round(mean(is.na(x)), digits = 3)) table(indv$applic_gen, indv$country, useNA = "always") table(indv$applic_gen, indv$country, useNA = "always") indv<-indv %>% mutate(compl=ifelse(complete.cases(indv), TRUE, FALSE)) table(indv$compl) # Dataset of independent variables with only complete cases indv_c<-indv[indv$compl==TRUE,] #### 1b Dependent variables #### ##In IDN, there was additionally the item "Vocational Training". We merge that to education, study, training, phd. table(a$mreason_lang) a$mreason_lang[a$mreason_lang==-14]<- -99 a$mreason_lang[a$mreason_lang==17]<-0 table(a$reason_stud, a$reason_train) a$reason_stud[a$reason_train==TRUE]<-TRUE a<-select(a, -reason_train) # For surveys in JPN, BIH, GBR and POL instead of having the two categories "(Possible) move to a German-speaking country for professional reasons" (mreason_) # (mreason_lang=9, reason_move_prof=TRUE) and "(Possible) move to a German-speaking country for other reasons" (mreason_lang=10, reason_move_other=TRUE) # we have "(Possible) move to a German-speaking country" (mreason_lang=16, reason_move=TRUE) # Here we deal with that issue in order to split up the single response into the more fine-grained ones. table(a$mreason_lang) # Variable that indicates that observation was done in JPN, BIH, GBR and POL a<-a %>% mutate(early=ifelse(country=="JPN" | country=="BIH" | country=="GBR" | country=="POL", TRUE, FALSE) ) # First, for those we have to decide whether their migration intention is for professional reasons or other reason (early==TRUE & mreason_lang==16) # we create a variable mig_reason_ger that indicates their main reason to migrate (mreason_mig1-3) to their highest preferred German-speaking # destination country (mig_1-3). #Then we recode mreason_lang in such a way, that those who indicate to migrate professional reasons (mig_reason_ger>=0 & mig_reason_ger<7) # as mreason_lang=9, and those who indicate to migrate non-professional reasons (mig_reason_ger>6 & mig_reason_ger<16) as mreason_lang=10. # Those who have no German speaking country among their first three preferred destination countries (mig_reason_ger=-9999) we code as mreason_lang==-999 #(and drop them from our observations (in analysis they are dropped anyway then.)) a <- a %>% mutate(mig_mreason_ger= ifelse(early==TRUE & mreason_lang==16 & (mig_1=="DEU" | mig_1=="CHE" | mig_1=="AUT"), mreason_mig1, ifelse( early==TRUE & mreason_lang==16 & (mig_2=="DEU" | mig_2=="CHE" | mig_2=="AUT"), mreason_mig2, ifelse( early==TRUE & mreason_lang==16 & (mig_3=="DEU" | mig_3=="CHE" | mig_3=="AUT"), mreason_mig3, ifelse( early==TRUE & mreason_lang==16, -9999, 9999)))), mreason_lang= ifelse(mig_mreason_ger>=0 & mig_mreason_ger<7,9, ifelse( mig_mreason_ger>6 & mig_mreason_ger<16, 10, ifelse( mig_mreason_ger<0, -999, mreason_lang))), mig_reason_ger= ifelse(early==TRUE & reason_move==TRUE & (mig_1=="DEU" | mig_1=="CHE" | mig_1=="AUT"), mreason_mig1, ifelse( early==TRUE & reason_move==TRUE & (mig_2=="DEU" | mig_2=="CHE" | mig_2=="AUT"), mreason_mig2, ifelse( early==TRUE & reason_move==TRUE & (mig_3=="DEU" | mig_3=="CHE" | mig_3=="AUT"), mreason_mig3, ifelse( early==TRUE & reason_move==TRUE, -9999, 9999)))), reason_move_prof=ifelse(mig_reason_ger>=0 & mig_reason_ger<7,TRUE, ifelse( mig_reason_ger>6 & mig_reason_ger<16, FALSE, ifelse( mig_reason_ger<0, NA, reason_move_prof))), reason_move_other=ifelse(mig_reason_ger>=0 & mig_reason_ger<7,FALSE, ifelse( mig_reason_ger>6 & mig_reason_ger<16, TRUE, ifelse( mig_reason_ger<0, NA, reason_move_other))) ) %>% filter(mig_mreason_ger>=0 & !(mig_reason_ger<0 & mreason_lang<0)) %>% ### second condition maximizes number of observations conditional on later types, Needs rethinking if analysing all reasons) select(-mig_mreason_ger, -mig_reason_ger) #table(a$mig_reason_ger, a$reason_move_other) table(a$mig_mreason_ger, a$mig_reason_ger) table(a$reason_move, a$reason_move_other, useNA = "always") table(a$mreason_lang, useNA="always") table(a[a$reason_visa==TRUE,]$reason_move_other,a[a$reason_visa==TRUE,]$reason_move_prof) ## Similiarily we deal with "Requirement for visa" a <- a %>% mutate(mig_mreason_ger= ifelse(mreason_lang==11 & (mig_1=="DEU" | mig_1=="CHE" | mig_1=="AUT"), mreason_mig1, ifelse(mreason_lang==11 & (mig_2=="DEU" | mig_2=="CHE" | mig_2=="AUT"), mreason_mig2, ifelse(mreason_lang==11 & (mig_3=="DEU" | mig_3=="CHE" | mig_3=="AUT"), mreason_mig3, ifelse(mreason_lang==11, -9999, 9999)))), mreason_lang= ifelse(mig_mreason_ger>=0 & mig_mreason_ger<7,9, ifelse(mig_mreason_ger>6 & mig_mreason_ger<16, 10, ifelse(mig_mreason_ger<0, -999, mreason_lang))), mig_reason_ger= ifelse(reason_visa==TRUE & reason_move_other==FALSE & reason_move_prof==FALSE & (mig_1=="DEU" | mig_1=="CHE" | mig_1=="AUT"), mreason_mig1, ifelse(reason_visa==TRUE & reason_move_other==FALSE & reason_move_prof==FALSE & (mig_2=="DEU" | mig_2=="CHE" | mig_2=="AUT"), mreason_mig2, ifelse(reason_visa==TRUE & reason_move_other==FALSE & reason_move_prof==FALSE & (mig_3=="DEU" | mig_3=="CHE" | mig_3=="AUT"), mreason_mig3, ifelse(reason_visa==TRUE & reason_move_other==FALSE & reason_move_prof==FALSE , -9999, 9999)))), reason_move_prof= ifelse(mig_reason_ger>=0 & mig_reason_ger<7,TRUE, ifelse(mig_reason_ger>6 & mig_reason_ger<16, FALSE, ifelse(mig_reason_ger<0, NA, reason_move_prof))), reason_move_other= ifelse(mig_reason_ger>=0 & mig_reason_ger<7,FALSE, ifelse(mig_reason_ger>6 & mig_reason_ger<16, TRUE, ifelse(mig_reason_ger<0, NA, reason_move_other)))) %>% filter(mig_mreason_ger>=0 & !(mig_reason_ger<0 & mreason_lang<0)) %>% ### second condition maximizes number of observations conditional on later types, Needs rethinking if analysing all reasons) select(-mig_mreason_ger, -mig_reason_ger, -reason_visa) #deal with mreason_lang<0 test<-a %>% select(country, mreason_lang) %>% mutate(answered=ifelse(mreason_lang>=0, TRUE, FALSE)) %>% group_by(country) %>% mutate(mean_na=1-mean(answered)) %>% ungroup() %>% select(-mreason_lang, -answered) %>% unique() table(a$country, a$reason_oth) ### For now we drop some of "Other", i.e. main reason = Other or NA if reason= other table(a$mreason_lang, useNA = "always") table(a$reason_other, a$mreason_lang) with(a[a$mreason_lang==15,], table(country)) a <- a %>% filter(mreason_lang!="15") #table(a$mreason_lang) # Convert numeric to factor and change labels a$mreason_lang <- factor(a$mreason_lang, levels = c(-999, -99, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 13, 14)) #15 deleted a$mreason_lang <- fct_recode(a$mreason_lang, "No answer" = "-999", "No answer" = "-99" , "Education" = "0" , "Company communication" = "1" , "Trading partners" = "2" , "Customers" = "3" , "Income" = "4" , "Employer" = "5" , "Labour market" = "6" , "Family" = "7" , "Friends" = "8" , "Educational and labour migration" = "9" , "Other migration" = "10" , "Holiday" = "12" , "Culture" = "13" , "Interest" = "14" ) table(a$mreason_lang, a$reason_move, useNA = "always") ###create type consumption vs. investment ### a<- a %>% mutate(type= case_when(.$mreason_lang=="Education" ~ TRUE, .$mreason_lang=="Company communication" ~ TRUE, .$mreason_lang=="Trading partners" ~ TRUE, .$mreason_lang=="Customers" ~ TRUE, .$mreason_lang=="Income"~ TRUE, .$mreason_lang=="Employer"~ TRUE, .$mreason_lang== "Labour market"~ TRUE, .$mreason_lang=="Family" ~ FALSE, .$mreason_lang=="Friends" ~ FALSE, .$mreason_lang=="Educational and labour migration"~ TRUE, .$mreason_lang== "Other migration"~ FALSE, .$mreason_lang=="Holiday"~ FALSE, .$mreason_lang=="Culture"~ FALSE, .$mreason_lang=="Interest"~ FALSE, .$mreason_lang=="No answer"~ NA), reason_cons=ifelse(reason_partner==TRUE | reason_soc==TRUE | reason_move_other==TRUE | reason_holi==TRUE | reason_cult==TRUE | reason_inter==TRUE, TRUE, FALSE) , reason_inv=ifelse(reason_stud==TRUE | reason_trade==TRUE | reason_comp==TRUE | reason_cust==TRUE | reason_inc==TRUE | reason_requ==TRUE | reason_lab==TRUE | reason_move_prof==TRUE , TRUE, FALSE), type=ifelse(is.na(type) & reason_cons==TRUE & reason_inv==FALSE, FALSE, type), type=ifelse(is.na(type) & reason_cons==FALSE & reason_inv==TRUE, TRUE, type) ) with(a[is.na(a$type),], table(reason_cons, reason_inv)) table(a$reason_cons, a$reason_inv) a<-a %>% mutate(reason_om=ifelse(reason_cons==TRUE & reason_inv==TRUE, TRUE, FALSE)) table(a$reason_om) table(a$type, a$mreason_lang, useNA="always") ### create more detailed type ### table(a$type2, useNA = "always") is.factor(a$type2) a<- a %>% mutate(type2= case_when(.$mreason_lang=="Education" ~ "Education", .$mreason_lang=="Company communication" ~ "Domestic labour market", .$mreason_lang=="Trading partners" ~ "Domestic labour market", .$mreason_lang=="Customers" ~ "Domestic labour market", .$mreason_lang=="Income"~ "Domestic labour market", .$mreason_lang=="Employer"~ "Domestic labour market", .$mreason_lang=="Labour market"~ "Domestic labour market", .$mreason_lang=="Family" ~ "Personal reasons", .$mreason_lang=="Friends" ~ "Personal reasons", .$mreason_lang=="Educational and labour migration"~ "Educational and labour migration", .$mreason_lang=="Other migration"~ "Personal reasons", .$mreason_lang=="Holiday"~ "Cultural interest", .$mreason_lang=="Culture"~ "Cultural interest", .$mreason_lang=="Interest"~ "Cultural interest", .$mreason_lang=="No answer"~ "No answer"), type2=ifelse(type2=="No answer", NA, type2), reason_culture =ifelse(reason_holi==TRUE | reason_cult==TRUE | reason_inter==TRUE, TRUE, FALSE) , reason_education =ifelse(reason_stud==TRUE , TRUE, FALSE), reason_domestic =ifelse(reason_trade==TRUE | reason_comp==TRUE | reason_cust==TRUE | reason_inc==TRUE | reason_requ==TRUE | reason_lab==TRUE , TRUE, FALSE), reason_foreign =ifelse(reason_move_prof==TRUE , TRUE, FALSE), reason_social =ifelse(reason_partner==TRUE | reason_soc==TRUE | reason_move_other==TRUE , TRUE, FALSE) , type2=ifelse(is.na(type2) & reason_culture==TRUE & reason_education==FALSE & reason_domestic==FALSE & reason_foreign==FALSE & reason_social==FALSE , "Cultural interest", type2), type2=ifelse(is.na(type2) & reason_education==TRUE & reason_culture==FALSE & reason_domestic==FALSE & reason_foreign==FALSE & reason_social==FALSE, "Education", type2), type2=ifelse(is.na(type2) & reason_domestic==TRUE & reason_education==FALSE & reason_culture==FALSE & reason_foreign==FALSE & reason_social==FALSE, "Domestic labour market", type2), type2=ifelse(is.na(type2) & reason_foreign==TRUE & reason_education==FALSE & reason_domestic==FALSE & reason_culture==FALSE & reason_social==FALSE, "Educational and labour migration", type2), type2=ifelse(is.na(type2) & reason_social==TRUE & reason_education==FALSE & reason_domestic==FALSE & reason_foreign==FALSE & reason_culture==FALSE, "Personal reasons", type2)) #change order of factors a$type2 <- factor(a$type2,levels=c( "Education", "Domestic labour market", "Educational and labour migration", "Personal reasons", "Cultural interest", "No answer")) levels(a$type2) table(a$type2, a$type, useNA = "always") table(a$type2, a$mreason_lang, useNA = "always") ##Analyse missings dv<-select(a, id_uni, type, type2, langjob, reason_culture, reason_education, reason_domestic, reason_foreign, reason_social, reason_cons, reason_inv) table(dv$langjob, useNA="always") apply(dv, 2, function(x) mean(is.na(x))) apply(dv, 2, function(x) sum(is.na(x))) table(a$reason_cons, a$reason_inv) # Other reasons ab<-select(a, starts_with("reason")) apply(ab, 2, function(x) mean(is.na(x))) ## Dataset of DV with NAs dv<-select(a, id_uni, type, type2, langjob, mreason_lang, starts_with("reason")) dv$langjob<-ifelse(dv$langjob < 0, NA, dv$langjob) dv<-dv %>% mutate(compl=ifelse(complete.cases(dv), TRUE, FALSE)) apply(dv, 2, function(x) mean(is.na(x))) table(dv$compl) mean(dv$compl) dv<-select(dv, -compl) #before remove uncomplete cases, we join ind. var. table(dv$langjob, useNA="always") #### Country variables #### table(indv$country) indv_c<-indv_c %>% mutate(ling=case_when(.$country=="BIH" ~ 0, .$country=="CZE" ~ 0, .$country=="ESP" ~ 0, .$country=="GBR" ~ 1, .$country=="IDN"~ 0, .$country=="IND"~ 1, .$country=="ITA"~ 0, .$country=="JPN" ~ 0, .$country=="KOR" ~ 0, .$country=="MEX"~ 0, .$country=="NLD"~ 1, .$country=="POL"~ 0, .$country=="ROU"~0, .$country=="UKR"~ 0), europe=case_when(.$country=="BIH" ~ 1, .$country=="CZE" ~ 0, .$country=="ESP" ~ 0, .$country=="GBR" ~ 0, .$country=="IDN"~ 2, .$country=="IND"~ 2, .$country=="ITA"~ 0, .$country=="JPN" ~ 2, .$country=="KOR" ~ 2, .$country=="MEX"~ 2, .$country=="NLD"~ 0, .$country=="POL"~ 0, .$country=="ROU"~0, .$country=="UKR"~ 1), income=case_when(.$country=="BIH" ~ 1, .$country=="CZE" ~ 2, .$country=="ESP" ~ 2, .$country=="GBR" ~ 2, .$country=="IDN"~ 0, .$country=="IND"~ 0, .$country=="ITA"~ 2, .$country=="JPN" ~ 2, .$country=="KOR" ~ 2, .$country=="MEX"~ 1, .$country=="NLD"~ 2, .$country=="POL"~ 2, .$country=="ROU"~1, .$country=="UKR"~ 0), cultclose=case_when(.$country=="BIH" ~ 1, .$country=="CZE" ~ 1, .$country=="ESP" ~ 0, .$country=="GBR" ~ 0, .$country=="IDN"~ 1, .$country=="IND"~ 0, .$country=="ITA"~ 1, .$country=="JPN" ~ 1, .$country=="KOR" ~ 1, .$country=="MEX"~ 0, .$country=="NLD"~ 0, .$country=="POL"~ 0, .$country=="ROU"~0, .$country=="UKR"~ 1)) ### Cultural distance library(reshape2) library(readxl) d <- read_excel('cultdist.xlsx') # drop some columns d <- d[, c(2,9:10)] # Distance to Germany d$ltodist <- abs(d$ltowvs-82.871537) d$ivrdist <- abs(d$ivr-40.401786) d %>% rename(country = country.iso3) -> d indv_c <- left_join(indv_c, d) rm(d) # Normalize between 0 and 1 indv_c$ltodist <- (indv_c$ltodist/max(indv_c$ltodist, na.rm = TRUE)) indv_c$ivrdist <- (indv_c$ivrdist/max(indv_c$ivrdist, na.rm = TRUE)) #### 1c Independent and dependent variables #### ## Ind. only complete cases, dep. v with missings est<-left_join(indv_c, dv) saveRDS(indv, "indv.RDS") apply(est, 2, function(x) mean(is.na(x))) ##Here we only filter out cases that have neither type, type2 nor langjob, not on entirely complete cases est<-est %>% mutate(complete=ifelse(complete.cases(est), TRUE, FALSE), one_m= ifelse( !is.na(type) | !is.na(type2) | !is.na(langjob), TRUE, FALSE))%>% filter(one_m==TRUE) %>% select(-compl) table(est$one_m, est$complete) table(est$country, est$complete) apply(est[est$country=="KOR",], 2, function(x) mean(is.na(x))) saveRDS(est, "est.RDS") table(est$type2, est$type, useNA = "always") table(est$complete)