1
|
# Code for checking the helpbuy questions on W12 hhresp file
|
2
|
|
3
|
|
4
|
# Libraries (install first as required)
|
5
|
packages <- c("tidyverse", "haven", "lubridate", "descr")
|
6
|
|
7
|
lapply(packages, library, character.only = T)
|
8
|
|
9
|
rm(packages)
|
10
|
|
11
|
|
12
|
|
13
|
|
14
|
# Load W12 hhresp from computer (adapt file path below first)
|
15
|
UKHLS_W12_hhresp <- read_dta(file = "./Datasets/SN6614/UKDA-6614-stata/stata/stata13_se/ukhls/l_hhresp.dta") %>%
|
16
|
mutate_if(haven::is.labelled, haven::as_factor)
|
17
|
|
18
|
|
19
|
|
20
|
|
21
|
# Retain the variables we need for checking the issue
|
22
|
UKHLS_test <- select(UKHLS_W12_hhresp,
|
23
|
l_hidp, l_ivfho, l_origadd,
|
24
|
l_intdatem, l_intdatey,
|
25
|
l_hholdmodedv,
|
26
|
l_hsownd,
|
27
|
starts_with("l_helpbuy"))
|
28
|
|
29
|
|
30
|
|
31
|
|
32
|
|
33
|
# Helpbuy questions are routed to owners only - so drop everyone else
|
34
|
table(UKHLS_test$l_hsownd, UKHLS_test$l_helpbuy1, useNA = "ifany")
|
35
|
|
36
|
|
37
|
UKHLS_test <- filter(UKHLS_test,
|
38
|
l_hsownd %in% c("Owned outright",
|
39
|
"Owned/being bought on mortgage",
|
40
|
"Shared ownership (part-owned part-rented"))
|
41
|
|
42
|
|
43
|
|
44
|
# No obvious tenure patterning in inapplicable codes on helpbuy variables
|
45
|
table(UKHLS_test$l_hsownd, UKHLS_test$l_helpbuy1, useNA = "ifany")
|
46
|
# Note - number of inapplicable/refusal/DK is constant across helpbuy variables
|
47
|
|
48
|
|
49
|
|
50
|
# Generate interview date indicator to explore whether dates are part of the
|
51
|
# missing data puzzle
|
52
|
UKHLS_test <- mutate(UKHLS_test,
|
53
|
intdate = my(paste(as.character(l_intdatem),
|
54
|
as.character(l_intdatey),
|
55
|
sep = " ")))
|
56
|
|
57
|
|
58
|
|
59
|
# Examine helpbuy pattern by interview time
|
60
|
crosstab(UKHLS_test$intdate, UKHLS_test$l_helpbuy1, plot = F, prop.r = T)
|
61
|
# Note - can see the big jump in 'inapplicable' codes at Jan 2021. Prior to Jan '21
|
62
|
# around 20% were inapplicable each month but from Jan this is 70+%
|
63
|
|
64
|
|
65
|
|
66
|
# Now examine whether interview mode has a bearing on this
|
67
|
table(UKHLS_test$l_hholdmodedv, useNA ="ifany")
|
68
|
|
69
|
crosstab(UKHLS_test$l_hholdmodedv, UKHLS_test$l_helpbuy1, prop.r = T,
|
70
|
plot = F) # Overall higher NA from CAWI, middle CATI, lowest CAPI
|
71
|
|
72
|
capi <- filter(UKHLS_test, l_hholdmodedv == "capi")
|
73
|
|
74
|
crosstab(capi$intdate, capi$l_helpbuy1, prop.r = T, plot = F)
|
75
|
# Low n CAPI in this wave due to COVID
|
76
|
|
77
|
cati <- filter(UKHLS_test, l_hholdmodedv == "cati")
|
78
|
|
79
|
crosstab(cati$intdate, cati$l_helpbuy1, prop.r = T, plot = F)
|
80
|
# Percentage of CATI respondents with inapplicable values is low until Feb 21
|
81
|
# then increases hugely
|
82
|
|
83
|
cawi <- filter(UKHLS_test, l_hholdmodedv == "cawi")
|
84
|
|
85
|
crosstab(cawi$intdate, cawi$l_helpbuy1, prop.r = T, plot = F)
|
86
|
# Sudden jump in inapplicables from ~20% in 2020 to >80% in January 2021.
|
87
|
|
88
|
# Overall conclusion - through 2020 most owner respondents did report values
|
89
|
# on helpbuy. However, from 2021, across CATI and CAWI we see a switch to
|
90
|
# much higher % inapplicable. The reason for this is unclear.
|
91
|
|
92
|
|
93
|
|