In the 1997 Child Development Supplement, time
diary data were obtained directly
from the children or from the children with input from the caregiver, or, for very
young children, solely from the caregiver report. Because school day activities
are often not reported in detail, a school day diary was asked of the activities
of the child during the school day (any out of home care setting or formal
educational setting where a ‘teacher’ could be identified). These data are
for the time diary reports of the caregiver/child.
For each
child, a random school day (Monday through Friday) and a random weekend day
(Saturday or Sunday) was recorded in a diary. The diaries were originally mailed
out and the difficulties respondents had in completing them led to follow-up
telephone interviews in which trained interviewers recorded most of the diary
answers given by the caregiver/child.
As with
previous time use data collected at the Survey Research Center of the University
of Michigan, the diary is a chronology of events starting at midnight. There are
typically anywhere from 15-40 entries, so the data need to be aggregated into
amounts of time by type of activity across all of the entries. There are
numerous three digit codes, so for a given child on a given day it is often the
case that there was no time devoted to many of the disaggregated, three digit
categories. This aspect of time diaries – their unbiasedness (Juster and
Stafford, “The Allocation of Time: Empirical Findings, Behavioral Models and
Problems of Measurement,” Journal of Economic Literature, June, 1991,
(Table5) ) but low reliability for certain activities which are not regular
daily activities has been discussed by Graham Kalton (Graham Kalton,
“Sample Design Issues in Time-Diary Studies,” in
Time Goods and Well-Being, F.T. Juster and F.P. Stafford eds.,
Institute for Social Research, University of Michigan, 1985).
Despite this
issue of reliability, especially for narrow 3-digit time use codes, we have
created files which aggregated, by day of the week, the hours and minutes (HH:MM)
in time format, for the child’s weekday – with an indicator as to which
weekday this was, and with an indicator of whether this child completed a
Saturday or Sunday diary. To view the 3-digit codes please refer to the Coding
Manual (pages 21-43) or our experimental use of Tree Viewer of
Activity Codes. Please direct any comments on this graphics documentation format to PSID
help.
In 1981-1982 Eccles, Juster, and Stafford
implemented a pilot project sponsored by the Foundation for Child Development
(see Timmer, Susan
G., Eccles, Jacquelynne and Kerth O’Bien, “How Children Use Time,” in
Juster and Stafford, 1985) that used a sample of one weekday and one weekend day. In that pilot study, a great deal of
time was simply not possible to allocate (See Juster and Stafford, 1991, p. 480,
Table 4. Please note the 23.2 hours (of a possible 168 per week) of ‘Not
Allocated’ time for the primary school students in the U.S. data) and much of
the remaining data was very difficult to code. In 1997 the interviewers were
given detailed instructions on the importance of not letting respondents give
vague answers on the time diary.
We learned
from the 1981-1982 pilot that the 1997 data are far superior. Also, by 1997 new
computer-based time uses had emerged. For
these reasons the current aggregations make no attempt to aggregate in line with
the 1981-82 codes, though the 1997 time use codes have evolved from the codes
set out in the 1965 multinational time use study of urban adults.
Users of the
1997 time diary data are urged to aggregate the detailed codes into categories
which they find meaningful for their research purposes – we have created rectangular files which are easy to aggregate.
For example, one may want to add 549 (Homework) to 501 (lessons in
computers, learning how to use a computer) and to 504 (Using the computer for
homework, ….) to create a ‘study’ variable.
Even a bit more broadly one may wish to add 512 (Computer work, getting
computer programs to work, …) as a ‘study’ variable. These files allow
such aggregations to be easily created by specifying the string of time
use codes which one wishes to aggregate.
The problem
of days of the week - especially
whether a Saturday or Sunday – is
more difficult to work with. Obviously it would have been better to have another
school day (weekday) diary and both a Saturday and Sunday diary as in the
1975-76 time use study (Juster and Stafford, 1985). But the respondent and
financial burden would simply have been too great. Since Saturday and Sunday
time uses are generally quite different for everyone – including children (Timmer,
O’Brien, and Eccles, 1985), analysts may want to work with the sub-samples of
children with a Saturday (or Sunday) diary separately.
In a statistical model one would at a minimum want to indicate the type
of weekend day for which the diary was collected. A variable indicating whether the weekend day was a Saturday or a
Sunday is provided on the file.
As discovered
with the 1981-82 pilot study, the time diary codes apply quite well to children
of all ages as well as to adults. Thus the active ‘reading books’ by adults
(code 939) is more likely to be ‘being read to’ (Code 943) for young
children – but adults can also be read to, and young children can read some
things (reading to a child is code 238). So it’s primarily the mix of time uses which changes across age
categories, not the codes. This is very important, since the time use codes
apply across all ages – including adults.
As the children age into the 5-18 year age range in 2002-03 (our next
planned diaries for children) similar codes and aggregations will apply.
