|
# Code for checking the helpbuy questions on W12 hhresp file
|
|
|
|
|
|
# Libraries (install first as required)
|
|
packages <- c("tidyverse", "haven", "lubridate", "descr")
|
|
|
|
lapply(packages, library, character.only = T)
|
|
|
|
rm(packages)
|
|
|
|
|
|
|
|
|
|
# Load W12 hhresp from computer (adapt file path below first)
|
|
UKHLS_W12_hhresp <- read_dta(file = "./Datasets/SN6614/UKDA-6614-stata/stata/stata13_se/ukhls/l_hhresp.dta") %>%
|
|
mutate_if(haven::is.labelled, haven::as_factor)
|
|
|
|
|
|
|
|
|
|
# Retain the variables we need for checking the issue
|
|
UKHLS_test <- select(UKHLS_W12_hhresp,
|
|
l_hidp, l_ivfho, l_origadd,
|
|
l_intdatem, l_intdatey,
|
|
l_hholdmodedv,
|
|
l_hsownd,
|
|
starts_with("l_helpbuy"))
|
|
|
|
|
|
|
|
|
|
|
|
# Helpbuy questions are routed to owners only - so drop everyone else
|
|
table(UKHLS_test$l_hsownd, UKHLS_test$l_helpbuy1, useNA = "ifany")
|
|
|
|
|
|
UKHLS_test <- filter(UKHLS_test,
|
|
l_hsownd %in% c("Owned outright",
|
|
"Owned/being bought on mortgage",
|
|
"Shared ownership (part-owned part-rented"))
|
|
|
|
|
|
|
|
# No obvious tenure patterning in inapplicable codes on helpbuy variables
|
|
table(UKHLS_test$l_hsownd, UKHLS_test$l_helpbuy1, useNA = "ifany")
|
|
# Note - number of inapplicable/refusal/DK is constant across helpbuy variables
|
|
|
|
|
|
|
|
# Generate interview date indicator to explore whether dates are part of the
|
|
# missing data puzzle
|
|
UKHLS_test <- mutate(UKHLS_test,
|
|
intdate = my(paste(as.character(l_intdatem),
|
|
as.character(l_intdatey),
|
|
sep = " ")))
|
|
|
|
|
|
|
|
# Examine helpbuy pattern by interview time
|
|
crosstab(UKHLS_test$intdate, UKHLS_test$l_helpbuy1, plot = F, prop.r = T)
|
|
# Note - can see the big jump in 'inapplicable' codes at Jan 2021. Prior to Jan '21
|
|
# around 20% were inapplicable each month but from Jan this is 70+%
|
|
|
|
|
|
|
|
# Now examine whether interview mode has a bearing on this
|
|
table(UKHLS_test$l_hholdmodedv, useNA ="ifany")
|
|
|
|
crosstab(UKHLS_test$l_hholdmodedv, UKHLS_test$l_helpbuy1, prop.r = T,
|
|
plot = F) # Overall higher NA from CAWI, middle CATI, lowest CAPI
|
|
|
|
capi <- filter(UKHLS_test, l_hholdmodedv == "capi")
|
|
|
|
crosstab(capi$intdate, capi$l_helpbuy1, prop.r = T, plot = F)
|
|
# Low n CAPI in this wave due to COVID
|
|
|
|
cati <- filter(UKHLS_test, l_hholdmodedv == "cati")
|
|
|
|
crosstab(cati$intdate, cati$l_helpbuy1, prop.r = T, plot = F)
|
|
# Percentage of CATI respondents with inapplicable values is low until Feb 21
|
|
# then increases hugely
|
|
|
|
cawi <- filter(UKHLS_test, l_hholdmodedv == "cawi")
|
|
|
|
crosstab(cawi$intdate, cawi$l_helpbuy1, prop.r = T, plot = F)
|
|
# Sudden jump in inapplicables from ~20% in 2020 to >80% in January 2021.
|
|
|
|
# Overall conclusion - through 2020 most owner respondents did report values
|
|
# on helpbuy. However, from 2021, across CATI and CAWI we see a switch to
|
|
# much higher % inapplicable. The reason for this is unclear.
|
|
|
|
|
|
|