Leaving for more or settling for less: Analyses (log hourly wages)

rm(list=ls())

1 Packages

library(tidyverse)
library(lme4)

2 Input

load(file="H:/processed_data/df_mmfc.rda")

load(file="H:/processed_data/df_mmfc_r2.rda")

load(file="H:/processed_data/df_noother.rda")

3 Non-linear effects of time (in year since Phd)

Creating variables for t^2, t^3 up to t^6

df_mmfc$t2 <- df_mmfc$t^2
df_mmfc$t3 <- df_mmfc$t^3
df_mmfc$t4 <- df_mmfc$t^4
df_mmfc$t5 <- df_mmfc$t^5
df_mmfc$t6 <- df_mmfc$t^6

4 Check linearity t and log real pay

linearmod <- lm(log_realpay ~ t, data=df_mmfc)

ggplot(broom::augment(linearmod), aes(t, log_realpay)) +
 geom_point() +
 stat_smooth(method="lm", se=FALSE) +
 geom_segment(aes(xend = t, yend = .fitted), color="red", size=0.3)

par(mfrow = c(2,2))
plot(linearmod)

lm1 <- lm(log_realpay ~ t, data=df_mmfc)
lm2 <- lm(log_realpay ~ t + t2, data=df_mmfc)
lm3 <- lm(log_realpay ~ t + t2 + t3, data=df_mmfc)
lm4 <- lm(log_realpay ~ t + t2 + t3 + t4, data=df_mmfc)
lm5 <- lm(log_realpay ~ t + t2 + t3 + t4 + t5, data=df_mmfc)
lm6 <- lm(log_realpay ~ t + t2 + t3 + t4 + t5 + t6, data=df_mmfc)

5 Some releveling

levels(as.factor(df_mmfc$phd_disci))
summary(as.factor(df_mmfc$phd_disci))

df_mmfc$phd_disci <- factor(df_mmfc$phd_disci, levels=c("Health sciences", "Social sciences", "Natural sciences and mathematics", "Engineering", "Humanities"))

df_mmfc <- df_mmfc %>% 
  mutate(gender = ifelse(gender==1, "men", "women"))

df_mmfc$gender <- factor(df_mmfc$gender, levels=c("men", "women"))


df_mmfc$temporary_emp <- haven::zap_labels(df_mmfc$temporary_emp)

6 Overall - pay

# Null model: random intercept
M0 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# Compute ICC
icc_ip <- as.data.frame(VarCorr(M0))[4]
icc <- icc_ip[1,1] / sum(icc_ip)

# M1: Transition effect and controls
M1 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
M2 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + as.factor(phd_disci)*as.factor(gender) + as.factor(child_u5)*as.factor(gender) + child_u5_b*as.factor(gender) + log_hrs*as.factor(gender) + log_hrs_b*as.factor(gender) + as.factor(temporary_emp)*as.factor(gender) + temporary_emp_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + phd_disci*t*as.factor(gender) + child_u5_b*t*as.factor(gender) + log_hrs_b*t*as.factor(gender) + temporary_emp_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

7 By gender: pay

Creating separate datasets for men and women, for the separated analyses

df_men <- df_mmfc %>% filter(gender=="men")
df_wom <- df_mmfc %>% filter(gender=="women")

7.1 MEN

# Null model: random intercept
M0m <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# Compute ICC
icc_ip <- as.data.frame(VarCorr(M0m))[4]
icc_m <- icc_ip[1,1] / sum(icc_ip)

# M1: Transition effect and controls
M1m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
M2m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

7.2 WOMEN

# Null model: random intercept
M0w <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

# Compute ICC
icc_ip <- as.data.frame(VarCorr(M0w))[4]
icc_w <- icc_ip[1,1] / sum(icc_ip)

# M1: Transition effect and controls
M1w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
M2w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

8 Adding mechanisms step-by-step

8.1 Discipline

M2_disci <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + as.factor(phd_disci)*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + phd_disci*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# men
M2m_disci <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# women
M2w_disci <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

8.2 Child < 5

M2_child <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + as.factor(phd_disci)*as.factor(gender) + as.factor(child_u5)*as.factor(gender) + child_u5_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + phd_disci*t*as.factor(gender) + child_u5_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# men
M2m_child <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# women
M2w_child <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

8.3 Hours

M2_hrs <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + as.factor(phd_disci)*as.factor(gender) + as.factor(child_u5)*as.factor(gender) + child_u5_b*as.factor(gender) + log_hrs*as.factor(gender) + log_hrs_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + phd_disci*t*as.factor(gender) + child_u5_b*t*as.factor(gender) + log_hrs_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# men
M2m_hrs <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# women
M2w_hrs <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

8.4 Temporary employment

(These are the original models)

9 Robustness checks

9.1 Max 5 years post-PhD

df_mmfc %>%
  filter(t<6) -> df_mmfc_r5


df_men_r5 <- df_mmfc_r5 %>% filter(gender=="men")
df_wom_r5 <- df_mmfc_r5 %>% filter(gender=="women")

Men

# MAX 5 YEARS POST-PHD
# M0
R5_M0m <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men_r5)

# M1
R5_M1m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men_r5)

# M2
R5_M2m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men_r5)

Women

# MAX 5 YEARS POST-PHD
# M0
R5_M0w <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom_r5)

# M1
R5_M1w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom_r5)

# M2
R5_M2w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom_r5)

Overall results

# MAX 5 YEARS POST-PHD
# M0
R5_M0 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc_r5)

# M1
R5_M1 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc_r5)

# M2
R5_M2 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + as.factor(phd_disci)*as.factor(gender) + as.factor(child_u5)*as.factor(gender) + child_u5_b*as.factor(gender) + log_hrs*as.factor(gender) + log_hrs_b*as.factor(gender) + as.factor(temporary_emp)*as.factor(gender) + temporary_emp_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + phd_disci*t*as.factor(gender) + child_u5_b*t*as.factor(gender) + log_hrs_b*t*as.factor(gender) + temporary_emp_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc_r5)

9.2 Transition * t^2 / t^3

# OVERALL 
# Null model: random intercept
R7_M0 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# M1: Transition effect and controls
R7_M1 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + as.factor(trans_lt)*t2*as.factor(gender) + trans_lt_b*t2*as.factor(gender) + as.factor(trans_lt)*t3*as.factor(gender) + trans_lt_b*t3*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
R7_M2 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + as.factor(phd_disci)*as.factor(gender) + as.factor(child_u5)*as.factor(gender) + child_u5_b*as.factor(gender) + log_hrs*as.factor(gender) + log_hrs_b*as.factor(gender) + as.factor(temporary_emp)*as.factor(gender) + temporary_emp_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(otherjob)*as.factor(gender) + otherjob_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + as.factor(trans_lt)*t2*as.factor(gender) + trans_lt_b*t2*as.factor(gender) + as.factor(trans_lt)*t3*as.factor(gender) + trans_lt_b*t3*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + otherjob_b*t*as.factor(gender) + phd_disci*t*as.factor(gender) + child_u5_b*t*as.factor(gender) + log_hrs_b*t*as.factor(gender) + temporary_emp_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_mmfc)

# BY GENDER

# MEN
# Null model: random intercept
R7_M0m <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# M1: Transition effect and controls
R7_M1m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + as.factor(trans_lt)*t2 + trans_lt_b*t2 + as.factor(trans_lt)*t3 + trans_lt_b*t3 + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
R7_M2m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t  + as.factor(trans_lt)*t2 + trans_lt_b*t2 + as.factor(trans_lt)*t3 + trans_lt_b*t3 + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men)

# WOMEN


# Null model: random intercept
R7_M0w <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

# M1: Transition effect and controls
R7_M1w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + as.factor(trans_lt)*t2 + trans_lt_b*t2 + as.factor(trans_lt)*t3 + trans_lt_b*t3 + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
R7_M2w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(otherjob) + otherjob_b + as.factor(trans_lt)*t + trans_lt_b*t + as.factor(trans_lt)*t2 + trans_lt_b*t2 + as.factor(trans_lt)*t3 + trans_lt_b*t3 + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + otherjob_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom)

9.3 No other job

df_noother$t2 <- df_noother$t^2
df_noother$t3 <- df_noother$t^3
df_noother$t4 <- df_noother$t^4
df_noother$t5 <- df_noother$t^5
df_noother$t6 <- df_noother$t^6

df_noother$phd_disci <- factor(df_noother$phd_disci, levels=c("Health sciences", "Social sciences", "Natural sciences and mathematics", "Engineering", "Humanities"))

df_noother$temporary_emp <- haven::zap_labels(df_noother$temporary_emp)

df_noother <- df_noother %>% 
  mutate(gender = ifelse(gender==1, "men", "women"))

df_noother$gender <- factor(df_noother$gender, levels=c("men", "women"))

df_noother %>% filter(gender=="men") -> df_men_noo
df_noother %>% filter(gender=="women") -> df_wom_noo


# OVERALL 
# Null model: random intercept
R9_M0 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_noother)


# M1: Transition effect and controls
R9_M1 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) + as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_noother)


# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
R9_M2 <- lmer(log_realpay ~ 1 + t*as.factor(gender) + t2*as.factor(gender) + t3*as.factor(gender) + as.factor(trans_st)*as.factor(gender) + as.factor(trans_lt)*as.factor(gender) + trans_lt_b*as.factor(gender) + as.factor(phd_disci)*as.factor(gender) + as.factor(child_u5)*as.factor(gender) + child_u5_b*as.factor(gender) + log_hrs*as.factor(gender) + log_hrs_b*as.factor(gender) + as.factor(temporary_emp)*as.factor(gender) + temporary_emp_b*as.factor(gender) + phd_coh*as.factor(gender) + phd_sat*as.factor(gender) + as.factor(partnered)*as.factor(gender) + partnered_b*as.factor(gender) + as.factor(sector_gov)*as.factor(gender) + as.factor(sector_nonpr)*as.factor(gender) + sector_gov_b*as.factor(gender) + sector_nonpr_b*as.factor(gender) + break_job*as.factor(gender) + break_job_b*as.factor(gender) + abroad_time*as.factor(gender) + abroad_time_b*as.factor(gender) +  as.factor(trans_lt)*t*as.factor(gender) + trans_lt_b*t*as.factor(gender) + phd_coh*t*as.factor(gender) + phd_sat*t*as.factor(gender) + partnered_b*t*as.factor(gender) + sector_gov_b*t*as.factor(gender) + sector_nonpr_b*t*as.factor(gender) + break_job_b*t*as.factor(gender) + abroad_time_b*t*as.factor(gender) + phd_disci*t*as.factor(gender) + child_u5_b*t*as.factor(gender) + log_hrs_b*t*as.factor(gender) + temporary_emp_b*t*as.factor(gender) + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_noother)


# BY GENDER

# MEN

# Null model: random intercept
R9_M0m <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men_noo)


# M1: Transition effect and controls
R9_M1m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men_noo)


# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
R9_M2m <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(trans_lt)*t + trans_lt_b*t  + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_men_noo)


# WOMEN


# Null model: random intercept
R9_M0w <- lmer(log_realpay ~ 1 + t + t2 + t3 + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom_noo)


# M1: Transition effect and controls
R9_M1w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom_noo)


# M2: Adding explanatory variables (hours worked, discipline, parental status, temporary contract)
R9_M2w <- lmer(log_realpay ~ 1 + t + t2 + t3 + as.factor(trans_st) + as.factor(trans_lt) + trans_lt_b + as.factor(phd_disci) + as.factor(child_u5) + child_u5_b + log_hrs + log_hrs_b + as.factor(temporary_emp) + temporary_emp_b + phd_coh + phd_sat + as.factor(partnered) + partnered_b + as.factor(sector_gov) + as.factor(sector_nonpr) + sector_gov_b + sector_nonpr_b + break_job + break_job_b + abroad_time + abroad_time_b + as.factor(trans_lt)*t + trans_lt_b*t + phd_coh*t + phd_sat*t + partnered_b*t + sector_gov_b*t + sector_nonpr_b*t + break_job_b*t + abroad_time_b*t +  phd_disci*t + child_u5_b*t + log_hrs_b*t + temporary_emp_b*t + (1 + t + t2 | RINPERSOON) + (-1 + t3 | RINPERSOON), data=df_wom_noo)

---

Copyright © 2025