Specialization or Gender Roles: The Effect of Different Paid Work Models on The Division of Household Labor

Specialization or Gender Roles: The Effect of Different Paid Work Models on
The Division of Household Labor.

Abstract: Despite greater equality in public life, the home remains a bastion of
gender stratification, especially in the division of household labor. This paper looks
at how two different employment arrangements in the household affect patterns of the
division of household labor. The first involves gender roles and is operationalized
through overlap of shifts, based on Harriet Presser’s prior work in the subject. The
second is based on Gary Becker’s specialization thesis and is operationalized through
which partner works more. Both theses are analyzed using the UK 2000 Time Use
Data, with five major findings presented in the text. The first is that a gender gap
remains in the amount and type of work done in the household, with women doing
more unpaid labor in the home. Second, the distribution of work schedules for the
UK is calculated using 7-day work diaries for the first time, adding interesting and
useful information to a growing body of research on the 24-hour work force. The
third major finding is that a model based on variations shift work (derived from
Presser, 2003) fits the data. The fourth and most important finding is that a model
based on which partner works more better explains the division of household labor.
Finally, the two models are found to work slightly better when integrated, opening up
many avenues for future research. The research presented in this paper will be useful
in helping alleviate stratification in the household as it points to what circumstances
foster a more equitable division of labor in the home.







Michael R Corey
September 15, 2007
NOT FOR CITATION

---

1. Introduction
Despite gains in equality in the workplace, the division of labor remains an arena for
lingering stratification in the amount and type of household tasks performed by
partners in married couples. This project uses 2000/1 UK Time Use Data to establish
if a significant gap in housework exists by gender in dual-earner British couples. It
then measures the work patterns of these couples, and uses that data to analyze two
differing theories of how differences in paid work affects the household division of
labor.

This paper begins with a discussion of major theories surrounding the gender gap in
household labor and the pattern of work scheduling in the UK (section 2). The first is
centered on the assignment of women’s roles. This includes theories of gender roles
(Major, 1993; Hochschild, 1994) resource gaps and gender identification (Coverman,
1988). The second set of theories is concerned with the specialization hypothesis
(Becker, 1981). This literature review and discussion is followed by a description of
the data set and methodology used to address this question (section 3).

The next section presents the first major finding- that a gap in household labor still
exists in the UK, though it has shrunk over time similar to the gap in the US. This
fourth section of the paper also maps the schedules of the UK population, and finds
that UK arrangements differ from those in the US. This is the second major finding,
because it is the first time 7-day work diaries have been combined with questionnaire
data to map schedule distribution in the UK. This finding also opens up future
avenues for research into both the feminization of the labor market and integration of
gaps in household labor. The fourth section also operationalizes the two theoretical
Page 1 of 33
NOT FOR CITATION

---

hypothesis to test the gender gap in housework using bivariate analysis. This is
performed without controls, and provides a point of departure to begin the section on
modeling.

The fifth section tests the two main hypothesis with the addition of control variables
for alternative hypothesis including human capital, class and life course differences.
These models show that both hypothesis are partially supported, with the
specialization model fitting the data better. Aside from that finding the other major
finding is that the two independent variables for paid work fit the data only slightly
better when used together in a model. This suggests future areas for research and
theory which are included in the discussion.

[Space Left Intentionally Blank]
Page 2 of 33
NOT FOR CITATION

---

2. Literature Review
There is a substantial body of research on the division of labor in the household. The
ubiquitous finding is that women do a greater share of the housework no matter the
employment arrangements. It is also known that tasks are gendered, with women’s
tasks taking more time (cf. Coltrane, 2000). The major research question is thus not if
there is a division of labor in the household, but how does it vary between households.
This is a crucial question because only by finding variations in the patterns of the
division of household labor can factors leading to a more equitable division be
identified and more widely adopted.

2.1 Prior Research
The research presented below integrates different traditions of research on the work-
life balance. The first is Presser’s substantial body of work, which culminates in
Working in a 24/7 Economy (2003). This forms the basis of the categorizations used
in this paper, especially the classification of scheduling. Presser’s work is important
as it is becoming a standard reference for studies of the 24/7 economy and has worked
to bring the topic to the attention of academics and policy makers. Her work is based
in questions of gender relations and expectations, and she cites three hypothesis,
centering on a resource gap, traditionalism of sex role attitudes, and lack available
time-- all three being inversely related to the amount of housework done (Presser,
2003). This paper looks at the second and third issues the most. Traditionalism of
roles stresses desynchronization and attunement to gender roles and both are
operationalized through the overlap of shifts. The former is obvious, and the latter
comes from research stating that men perform more work when they are home
without their wives (Hochschild, 1997).
Page 3 of 33
NOT FOR CITATION

---

Coltrane (2000) lays out similar the central tenets of household labor research in his
meta-analysis of division of household labor studies in the 1990s. These theories
echo Presser as centering on relative resources, socialization-gender role attitudes and
time availability-constraints. He further notes that human capital is operationalized
through control variables for education while class can reflect resource-bargaining
power. All of these factors are considered in the models used below.

A gendered view of the household is based on the theory that wives do more
housework because they do female tasks, which are more plentiful and take longer.
These theories rest on learned behaviors in the home and societal categorization in the
wider world. Hochschild (1989) describes the second shift that wives in dual-earner
relationships endure. Her ethnographic work highlights the situations at a household
level which create this phenomena. Again and again an underlying reason comes
down to fulfilling expectations learned in the household. It is exactly these
expectations that Major (1993) discusses as gender differences in comparison to
standards. There are clear normative, feasibility and self (habitual) comparisons that
propagate the divisions of household labor. This is also considered by West and
Zimmerman (1987) who discuss gender as both a biological and social construction.
This all leads to the idea that women do women’s work, and more work, in the home
because it is expected and that is how it has always been done.

Gershuny et. al. (2005) and Breen and Cooke (2005) both look at the changes in the
division of household labor over time. Gershuny and his coauthors find a “lagged
adaptation” in the division of household labor following changes in the patterns of
Page 4 of 33
NOT FOR CITATION

---

paid labor. Breen and Cooke find that more equitable divisions of labor in the
household result in large part for a willingness of the male partner to adapt. This
latter idea is partially tested below with the inclusion of a gender identification scale
that measures an individual’s willingness to perform different tasks.

The role of Becker’s (1981) theories on the efficiency of a division of labor in the
household are central to the argument that a gap in human capital makes an imbalance
of housework desirable to the household. It also purports that specialization is
beneficial to the household, and in classical economic terms the stratification in the
household should break down as the number of work hours are equalized. The
research below will show that this model fits the data better, as the gap narrows but
does not disappear for dual earner couples.

2.2 Time Use Studies
The work-life balance has been a popular study for people using time use data for
decades. From the first major studies (Szalai, 1972; Young and Wilmot, 1973) the
balance of different activities has been studied by looking at the whole day. Time, of
course, is a truly zero sum game when taken on the micro level, and while some
multi-tasking can be accomplished (such as checking personal e-mail at work) the
balance of paid work and house work can not be multi-tasked unless working from
home is considered (which it is not in most of the studies cited as well as in this one).
As will be seen later in this paper, time use estimates are consistently lower than
questionnaire estimates for time spent on housework. These estimates are consistent
with older UK and US time-use estimates (Gershuny and Robinson, 1998; Bianchi et.
al, 2000). This is also consistent with the measurement gap discussed by Robinson
Page 5 of 33
NOT FOR CITATION

---

and Gershuny (1994). Robinson and Bostrom (1994) further discuss the
overestimation of the workweek by survey methods at length, coming to the
conclusion that people are either ‘blocking’ (my term) out the time to include breaks
and commuting time, or overestimating for social reasons (acceptance, the cult of
busyness). This would help explain some of the variation the housework estimates as
well, as it is highly conceivable that a respondent would lump the time spent even on
the individual process of cooking, eating, doing the dishes and cleaning to include a
fair amount of non-active time (i.e. the interruption of a phone call, or a break to listen
to the radio or watch TV). The time diaries provide a smaller unit of measurement
which allows more detailed analysis and allows for a more accurate summation by
researchers, rather than a single summation of a week’s activity by the respondent1.
One of the contributions the research below makes is in adding to the accurate
information available from time use studies into both work schedules and household
labor. The following two sub sections lay out the main hypotheses tested below and
comment on some of their expected results.

2.3 Gender Roles Hypothesis
The first hypothesis tested below concerns gender roles. Using Presser (2003) as a
basis for the first set of models, the theory that husbands do more work when home
alone is tested. This is operationalized by looking at how differences in work
schedules within a couple affect divisions of household labor A second operalization
of gender roles is the gender ID scale used as a control. In this research,
unfortunately, questions are only asked about willingness to complete various

1 Time Diaries have seen increasing favor in US social science research, with the US Bureau
of Labor Statistics recently creating the American Time Use Survey (ATUS) to complement
the existing American Heritage Time Use Survey (AHTUS)
Page 6 of 33
NOT FOR CITATION

---

household tasks, instead of a full battery of questions about gender identification.
While this differs from the information in Presser’s model, it does allow some
vantage point into how much individual willingness to complete household tasks
matters to their execution, which taps directly into the above discussion about the
work of Breen and Cooke (2005).

2.4 The Specialization in the Household Hypothesis
Following on Becker’s specialization theory, the second hypothesis tests what effect a
differential in amount of time worked has on the division of labor in the household.
This thesis is operationalized through calculating the work differential (husband’s
hours last week - wife’s hours last week), to create a competing measure of paid
work. Looking directly at the division of labor in the household without considering
shifts is more directly related to the specialization thesis in terms of bargaining within
the household and proves substantively and statistically more significant both with
and without controls. The research below will first calculate mean statistics for the
division of labor in the household and data on the work life balance. It will then test
the two hypothesis above using bivariate analysis and models which include controls.
The results will be presented and discussed, especially regarding their significance to
the current body of research and as a starting point for future research.

[Space Left Intentionally Blank]
Page 7 of 33
NOT FOR CITATION

---

3. Data and Methodology
3.1 The Data Set
This paper uses the UK 2000/1 Time Use Study to calculate time spent in paid work
and household labor. The data set used for this research is the UK 2000/1 Time Use
Survey. This is a specially drawn probability sample which was undertaken by the
UK Office of National Statistics. The survey lasted from summer 2000 through fall
2001. Every member of a selected household over 8 years of age was asked a battery
of questions as well as to provided with a 7 day work diary as well as two 24-hours
time use diaries one each for a weekday and weekend. The work diaries were used to
record if the respondent was working or not in 15-minute intervals. The 24 hours
diaries were used to record 144 ten minute time slots with main activity, secondary
activity, other person present and location recoded for each time slot. The day started
at 4:00 am and ends at 3:59 am the next day. According to the Office of National
Statistics (ONS), the response rate for the household questionnaire was 61% (Office
of National Statistics, 2004). Within those households, 81% of individuals took part
in the individual questionnaire, with 73% of all possible diaries being completed. The
net response rate was 45%, and 21000 24-hour diaries are available for analysis.
From this sample, a sub sample of dual earner couples was selected. A further
specification was that they have each completed the 7-day work diary and at least one
24-hour diary, including 3 distinct necessary activities (work, sleep, eating, personal
care, leisure) and have no more than 90 minutes of missing data. In total this left
1368 valid respondents with 2583 diaries for analysis.2



2 There are some minor issues with missing data which may cause the N values presented
here to not always add up to 1368- this wil be fixed in future versions of the paper.
Page 8 of 33
NOT FOR CITATION

---

3.2 Counting Time in Housework
The first task in arranging the data was to count the hours spent in housework for each
individual. This was done for 9 separate categories of housework as defined by
Presser (2003). This includes female (cooking, dishes, cleaning the house and tending
to clothes), male (work on the physical or outdoor home, and car repair) and neutral
tasks (shopping for household goods, paying bills, driving household members). The
primary activity is considered first, then the secondary activity (if the primary activity
is not a household task) with no time slot counted twice.

As discussed in the literature review, these estimates proved to be much lower than
similar questionnaire estimates. They were, however, directly in line with prior US
and UK estimates. Table 1 illustrates this point by comparing the UK 2000 estimates
with Bianchi et. al. (2000).

TABLE 1- AVERAGE HOURS SPENT ON HOUSEWORK
BY GENDER.
Year
Men
Women
Average
US 1965
4.9
30
17.45
US 1975
7.2
23.7
15.45
US 1985
9.8
19.7
14.75
US 1995
10
17.5
13.75
UK 2000
10.1
18.4
14.25
Sources: 1965-1995, Bianchi et. al. (2000); 2000 author.

The means for specific types of housework are presented in Table 2 below (section 4),
and widely used in the results below. All bivariate analyses of household division of
labor are weighted results, using the ONS supplied gross diary weight, which is meant
to be representative of the UK population. Husband’s share of female tasks and total
minutes spent on female tasks were both calculated from this information and used a
dependent variables in the models presented in section 5.

Page 9 of 33
NOT FOR CITATION

---