tasks which have traditionally been their responsibility.€¦ · useful supplement to other...
TRANSCRIPT
The Within-Household Distribution of Subjective Well-being
By
Shelley Phipps
Peter Burton
Lars Osberg
Department of EconomicsDalhousie UniversityHalifax, Nova Scotia
CanadaB3H 3J5
Phone: 902-494-2026Fax: 902-494-6917
E-mail:[email protected] [email protected]
July 7, 1996
We would like to acknowledge the excellent research assistance of Michael Rushe and thefinancial support of the Social Science and Humanities Research Council of Canada.
2
An exception is Apps (1994) who incorporates measures of housework as well as labour1
supply in a model of family well-being.
This paper examines the distribution of well-being within households, as indicated by
subjective self-reports of happiness/satisfaction. In particular, we study differences between men
and women in their reported degree of satisfaction with time for personal interests. In looking at
self-reported assessments of well-being, this paper diverges from much of the economics
literature. Although a central assumption underlying the economic explanation of human
behaviour is maximization of utility, economists in fact rarely examine directly self-reports of
utility or individual satisfaction. Clark and Oswald (1994, p. 648) argue that
The probable reason is that economists have traditionally been hostile to the notion thatutility can be measured. A different attitude is found among psychologists (who might bethought to be better qualified than economists to judge such things). Thousands of papersin the psychology literature are concerned with the statistical analysis of subjective utilityinformation.
Moreover, the sources of utility which economists have typically considered are rather
limited. Although labour economists, for example, are accustomed to working with models which
specify utility as a function of income (or goods consumed) and leisure time, most studies of the
distribution of well-being within households (e.g., Browning, et.al., 1994; Hoddinott and Haddad,
1992; Phipps and Burton, 1996) have focussed just on income (or expenditures). 1
Analysis of well-being that considers only income is unsatisfactory, from many
perspectives. Writers in the feminist tradition (e.g., McDaniel, 1994) have, for example, expressed
concern that although increased labour-force participation by wives has produced higher levels of
family money income, such trends have also generated a `double work day’ for many women --
some wives who now work for pay outside the home also return home to perform most of the
3
If gender stereotypes are commonplace, and “caring” is assigned the label of a feminine,2
unmanly activity, are men and women equally likely to report that they are engaged in “childcare”?
tasks which have traditionally been their responsibility.
Assessing the distribution of well-being within the household evidently requires some
assessment of the distribution of both material resources and available time - yet the
categorization of household activities as “work” or “leisure” is highly problematic. Although time
budget studies (e.g., Marshall, 1993) document how much time husbands and wives report
spending at various activities (e.g., working for pay, cooking, child care, yardwork), it is far from
clear how to allocate hours that are spent in simultaneous activities (e.g. cooking while chatting)
or whether particular activities should be seen as leisure or as part of household production (e.g.,
cooking) or whether men and women label their activities similarly . 2
For these reasons, this paper looks directly at self-reports of “happiness” and
“satisfaction” with available time. It is hoped that the use of subjective reports of satisfaction with
time can give some indication of the circumstances under which men or women feel constrained
for time, and can therefore indicate some of the determinants of the distribution of well-being.
We examine, for example, the impact of increased income and increased hours of paid labour on
reported well-being and whether the presence of young children affects a husband’s level of
satisfaction with time in the same manner as the wife’s.
We recognize that although we have criticized the use of inputs (money income, time use
data) as indicators of well-being, our own use of direct survey evidence on subjective satisfaction
also has deficiencies. “Satisfaction” with an outcome is, after all, dependent both on an
individual’s aspirations and the degree to which the realized outcome approaches that aspiration.
4
An approach which avoids this problem is the `functionings’ idea of Sen (1993). While3
we might prefer to assess well-being of husbands and wives in terms of their capabilities orfunctionings, the data requirements are more substantial than for the approach adopted here.
A subjective approach is relatively common, on the other hand, in the economics poverty4
literature (e.g., Hagenaars, 1986) and has also been used to study the unemployed (Clark andOswald, 1994). Subjective data have been widely used by sociologists.
Our data do not enable us to separately identify the influence of differences in aspirations and
differences in outcomes. However, we do feel that the use of survey-based self reports can be a 3
useful supplement to other evidence on the distribution of well-being within households. To our
knowledge, it is novel in the economics literature to use subjective measures to explore the
distribution of well-being within the family. 4
Section 2 of the paper discusses the data and some of the psychological literature on
`happiness/satisfaction.’ Section 3 examines the connections between reports of over-all
happiness and the traditional economic variables of income and labour supply as well as the links
between happiness and reported satisfaction with time for self. We argue that reported
satisfaction with time for self provides a better proxy for `leisure’ than the residual of total time
minus hours of paid labour supply.
Section 4 of the paper then pursues the main theme of the paper -- gender differences in
reported satisfaction with time for self. In particular, we examine differences between
husbands/wives engaged in paid employment, the factors influencing the probability that an
individual will report himself/herself satisfied with time and how these factors differ for men and
women. Section 5 attempts to assess the relative quantitative importance of time and money, for
personal satisfaction. Since greater satisfaction can be produced by either more income or less
working time, we ask: `How much would a person’s probability of being very satisfied with time
5
available for self increase if we gave them a 20 percent increase in income (labour supply
constant)?’ ‘How much would their satisfaction with time available for self increase with a 20
percent reduction in hours of paid work (income constant)?
Section 6 concludes.
2. Measuring Happiness -- The Canadian Data and the Existing Literature
If we are to attempt to understand the relative `happiness’ of Canadian husbands and
wives, it seems important first to attempt some definition. Psychologists have pointed out the
many dimensions of `happiness.’ `Happiness’ can be thought of as `joy’ -- an emotional
experience, as in “Do you `feel happy’? Are you in `a good mood’? Do you feel that you mainly
have `pleasant experiences’?” Alternatively, we might think of `happiness’ as akin to
`satisfaction’ -- a more cognitive concept describing an individual’s state of mind when he or she
has reasonably assessed personal circumstances, as in “To what extent do you feel that your
personal aspirations have been met? To what extent do you feel that you are getting what you
want from life (Argyle, 1987; Veenhoven, 1989)?” This second conception of happiness is
probably the most relevant for our work, given both the way economists traditionally think about
utility as the outcome of rational choice and given the way our survey questions have been asked.
Ideally, we would measure the frequency, duration and intensity of happiness.
Psychological evidence indicates that frequency and duration of happy and unhappy spells are
inversely related while the intensity of these experiences is positively related (i.e., individuals who
feel most intensely happy are also the ones who are likely to feel most intensely unhappy, when
things go wrong). Unfortunately, such sophisticated measurement is not possible with the data
6
we have available to us, which simply asks about perceived happiness at the time of the survey.
But, is it really possible to measure happiness - and how should it be done? There do not
appear to be any physical attributes consistently associated with happiness (e.g., brain waves,
blood pressure). Nor is observed behaviour always reliable -- `active, friendly and outgoing’
people are more likely to be happy but many quiet, shy and sedentary individuals are also happy
(Veenhoven, 1989). Thus, many psychologists conclude that self-reports provide the best means
of assessing happiness, despite the many criticisms levelled against such subjective data.
Veenhoven (1989) summarizes several of the most important of these.
First, people appear capable of answering questions about whether or not they are happy
and about how happy they are. Apparently, about 80 percent of the US population think about
how happy they are at least once a week. Further, interviewers find that respondents answer
quickly when asked about happiness and non-response is low. Finally, the temporal stability of
answers is very high. This is not to say that individuals are not influenced by such random factors
as the weather or the behaviour/appearance of the interviewer (in fact, more people are likely to
report themselves happy on a sunny than a rainy day), but error introduced in this way is believed
to be random (Veenhoven, 1989).
A second important criticism of subjective data on happiness is that individuals appear to
over-state their relative happiness -- most individuals believe themselves to be happier than
average!! Yet, no evidence can be found that people are lying when responding to these surveys
(Argyle, 1987) and patterns of response are reasonable - e.g. people living in absolutely miserable
conditions (e.g., very low income, poor sanitation) are less likely to claim to feel happy. Thus,
Veenhoven (1989, p. 14) concludes that:
7
“people tend to enjoy their lives once conditions are not too bad. From an adaptivebiological point of view this does not seem strange. Nature is unlikely to have burdenedus with characteristic unhappiness . . . . Like “health,” happiness would seem to be anormal condition.” However, “the prevalence of satisfaction with life-as-a-whole does notwash away the multitude of suffering and complaints.”
However, Annas (1993) argues that there is a gender dimension to the issue:
“For people’s desires can be in large part formed by the circumstances and options thatthey perceive as being open to them. . . .In societies in which the options open to them arefewer than those open to men, it has always been a common adaptive strategy for womento adjust their desires to what they can realistically expect. So examining the actualdesires of women may lead us to the conclusion that women on the whole get what theywant. . . . But it cannot, of course, be right that the happiest women should turn out to bethose whose horizons are so limited that they cannot even conceive of alternatives. Andeven apart from this issue, we have only to look at the way that women’s desires haveexpanded with the expansion of the alternatives open to them to see that women’s actualdesires cannot settle this issue for us.”
The point is, of course, more general -- anyone living in reduced circumstances with little
hope of improvement may adjust expectations and learn to be happy with their lot in life.
However, the feminist point raised by Annas is particularly relevant for our research which
compares self-reports of happiness/satisfaction between men and women. If women have learned
to expect less, they may report themselves to be as happy as do their husbands even if, to an
outsider, they appear to have absolutely less (in terms of time, money, etc.). As a consequence,
the results obtained from our research provide only one piece of a much larger puzzle.
DATA
The data used for this research are drawn from the Statistics Canada General Social
Survey (GSS) -- Cycle 5 (Family and Friends). The target population for the survey was all
persons 15 years and older, with the exception of residents of the Yukon and Northwest
Territories and full-time residents of institutions. The original micro-data sample includes 13,495
8
Respondents living in either legally married or common-law relationships with or without5
children are included.
We exclude as well a small number of observations with missing information.6
households. However, for this research we have selected respondents between the ages of 25 and
55 who were living in a husband/wife relationship at the time of the survey (January through
March of 1990) -- 1551 men and 1113 women. We focus on individuals living in husband/wife5
relationships since we are particularly interested in the within-household distribution of well-being
(no same-sex couples are identifiable in the data). We choose to study `prime-age’ men and
women as much of our work focusses on individuals who are labour-force participants. (We also
exclude those who are retired or disabled.) Finally, we exclude households with more than two
income-earners (e.g., a teenager with earnings) as we feel that this introduces an additional
pooling/sharing dynamic which we prefer to avoid at this stage.6
The major advantage of the General Social Survey for the purposes of our research is that
in addition to providing fairly standard socioeconomic information, it asks respondents a variety
of questions about their levels of satisfaction with particular aspects of life and with life over-all.
Specifically, respondents are asked ‘Would you describe yourself as: 1) very happy?; 2) somewhat
happy?; 3) somewhat unhappy?; 4) very unhappy?’ As well, respondents are asked the two-stage
question: `Are you satisfied with the amount of time you have available for other interests? If
satisfied, are you very or somewhat satisfied? If not satisfied, are you very or somewhat
dissatisfied?’
Appendix A provides means and standard deviations of all variables by gender and labour-
market status of respondent. Notice that, consistent with the psychological literature, an
9
overwhelming number of respondents report themselves to be happy or very happy with life in
general (98.7 percent of men; 97.6 percent of women). A majority of both men and women are
`very happy’ with life in general (61 percent of men; 65.6 percent of women).
Levels of satisfaction with particular dimensions of life are not quite so high. Only 69.5
percent of men (64.5 percent of women) are satisfied or very satisfied with the time they have
available to pursue their own interests. Only 33 percent of men (28 percent of women) are very
satisfied.
The major disadvantage of the data set (besides the fact that it is cross-sectional rather
than longitudinal) is that we only have information on either the husband or the wife. Hence, all
our inferences must be based on averages for all husbands, compared to all wives. Another
limitation of the data is that while respondent’s earnings are reported as a continuous variable,
total household income is reported only by category, with a top-code at $80,000. Since much of
the literature on well-being within households focusses on an individual’s share of household
income as an important measure of `power’ within the family, we have converted the categorical
income data to continuous by reporting values at mid-points and including a dummy variable for
households with incomes above the $80,000 cut-off.
3. Does Income and Labour Supply determine Happiness?
Labour economists traditionally specify utility functions which depend on both income and
leisure time. However, since most micro-data surveys do not actually define “leisure” or provide
information about hours of leisure available, the standard proxy for leisure is constructed as a
10
If leisure hours are defined to be equal to total time available minus hours of paid7
employment, all hours of unemployment are thereby assumed to be identical to leisure. As anumber of articles note, this can be a very misleading assumption - see Osberg and Phipps(1993).
Recall that 98.7 percent of husbands and 97.6 percent of wives reported themselves to be8
`happy,’ where happy includes both very happy and somewhat happy. Hence, we feel that it ismost useful to study individuals who reports themselves as `very happy.’
residual - total time available less time spent in paid employment. While economists are unused7
to data providing self-reports of happiness/satisfaction (and there are problems with such reports
as discussed above), “happiness” and “utility” are very much the same idea and it would seem that
there should be some link between our notions of `utility’ and self-reports of happiness. In this
section we therefore examine the standard labour economics hypothesis that income and hours of
paid labour supply are important determinants of happiness.
Table 1 reports the results of simple OLS regressions which use the respondent’s personal
happiness index as dependent variable (reported happiness is greater as the index value increases)
as well as the results of probit regressions where the dichotomous dependent variable takes the
value one when the individual reports himself/herself to be `very happy’ with life in general. [8
Since these regressions use individual well-being as the dependent variable, they are consistent
with the perspective of collective models of family behaviour (Chiappori, et.al., 1992) which
assume that individuals within families have different preferences and experiences, but they are
only consistent with the perspective of traditional unitary models of the family (e.g., Becker,
1974; 1981) if everyone in the family is equally well-off.]
Regressions are run separately for husbands and wives, and we report three different
specifications. If we take the standard neo-classical labour supply model as our starting point,
utility is assumed to be a positive function of income and a negative function of hours of labour
11
We believe that the `needs’ adjusted equivalent income concept is more likely to be9
linked with happiness than would be total unadjusted household income. Household equivalentincome is calculated using the OECD scale [1.0:0.7:0.5:0.5: etc] .
Total annual hours of labour supply are calculated for each spouse as the product of10
total weeks of work and usual weekly hours. These are added for husband and wife to obtaintotal household hours.
supply -hence we would expect people to be happier the higher their household income and the
fewer the hours of paid work that they have to do. The first specification, (a), includes as
explanatory variables total household equivalent income and total household hours supplied to9
the paid labour market. A greater total commitment of hours to paid employment is, in the10
neo-classical formulation, expected to reduce well-being (income held constant) because it
reduces time available for leisure.
The second specification, (b), separates total weekly hours and total weeks of labour
supply since it has been suggested (Marshall, 1993) that most household responsibilities (cooking
meals, caring for children) cannot be deferred until another week - hence the crucial time crunch is
that of weekly hours of work.
The third specification, (c), recognizes that an individual’s personal well-being may be
affected differently by increased hours of spouse’s paid employment than by increased hours of
own paid employment. We therefore include as regressors both the total weekly hours and total
weeks of household labour supply and `respondent’s share’ of hours and `respondent’s share’ of
weeks. While unitary models of the family traditionally recognize that increases in his
employment may not have the same impact on household utility as increases in her employment,
the explanation provided for this is that husband and wife have different specializations and
receive different returns from another hour’s paid versus unpaid work. Total household income
12
Note that a unitary model would predict that increases in the wife’s employment should11
have the same impact on happiness for both husband and wife. Similarly increases in thehusband’s employment should have the same impact on each member of the household althoughin this case labour market conditions may mean that the household receives more income.
It is important to note that these regressions are not presented as `explanations’ of12
happiness -- they achieve appallingly bad fits. Rather, we are interested only in examiningwhether there are statistically significant links between the traditional economic variables and thesubjective happiness indicators.
Our results on the importance of income in determining over-all happiness appear to13
contradict those current in the psychology literature where income is generally found to have asmall or insignificant relation with happiness. Income relative to peers rather than absoluteincome is regarded as the more important variable (Argyle and Martin, 1989). We intend toexplore this point in future research.
may increase more (and household production may fall less) when he increases paid employment
than when she does. However, unitary models do not recognize that his happiness might be11
affected in a different way from hers when he takes on more paid employment versus when she
takes on more paid employment. `Collective’ (Chiappori et.al., 1992) models of family behaviour,
on the other hand, take as their basic premise the idea that individual family members have
different tastes and experiences. It is thus quite consistent with such models that changes in her
behaviour may affect her happiness in a different way than they affect his happiness.
The first very clear message to take from Table 1 is that household equivalent income is a
significant and positive determinant of over-all happiness. This is true for both husbands and12
wives, for both the probit and OLS regressions and for all specifications. Thus, the first basic
premise of the neo-classical models receives strong support .13
The second important point to take from Table 1 is that the labour supply variables are not
13
Total household hours, in specification (a) and total weeks, in specification (b) are14
marginally significant (and negative) for men.
statistically significant determinants of happiness in any regression for men or for women. 14
Contrary to the predictions of standards models, more hours of paid employment does not mean
lower levels of happiness, given a particular level of income.
This second finding is not consistent with the standard view of labour economics that paid
work, at the margin, is a source of disutility. However, there may be a number of reasons for this
finding. First, many people regard having a job as a `good thing.’, and some individuals find their
work intrinsically satisfying. Argyle (1987) reports that in the UK, 32 percent of respondents
would carry on with their present jobs even if it were not financially necessary to do so. Another
65 percent would want to work `at something.’ Argyle also reports on the basis of 15 national
surveys for the US that 51.8 percent of respondents were `very satisfied’ with their jobs. Even if
the work itself is not intrinsically satisfying, psychologists argue that a paid job helps to satisfy
many important human needs. For example, work provides a source of identity, a source of
relationships outside the nuclear family, a means of structuring time, an opportunity to develop
skills and a sense of purpose in life (Furnham, 1991). Thus, a labour-supply model which assumes
that income increases utility while labour supply decreases utility seems far too simple - and may
in fact be misleading, since we do not find the labour-supply variables to be
Table 1Income and Labour Supply as Determinants of Happiness by Gender
Independent Husband WifeVariable
OLS Index of Happiness Probit OLS Index of HappinessVery Happy
(a) (b) (c) (a) (b) (c) (a) (b) (c) (a)
Household 0.471E-05 0.546E-05 0.527E-05 0.122E-04 0.140E-04 0.136E-04 0.551E-05 0.498E-05 0.512E-05 0.129E-04Equivelant (2.390) (2.739) (2.633) (2.576) (2.923) (2.836) (2.769) (2.486) (2.545) (2.707)Income
Total Household -0.182E-04 ------- ------- -0.457E-04 ------- ------- -0.125E-04 ------- ------- -0.144E-04Hours of paid (-1.487) (-1.565) (-1.029) (-0.501)Employment
Total Weekly ------- -0.343E-03 -0.219 ------- -0.120E-02 -0.902E-03 ------- -0.668E-03 -0.54E-03 -------Hours of Paid (-0.441) (-0.267) (-0.643) (-0.459) (-0.898) (-0.708)Employment
Respondent’s ------- ------- -0.439E-01 ------- ------- -0.592 ------- ------- -0.82E-01 -------Share of Total (-0.374) (-0.210) (-0.736)Weekly Hours ofPaid Employment
Total Weeks of ------- -0.109E-02 -0.525E-03 ------- -0.249E-02 -0.139 ------- 0.170E-03 0.711E-04 -------Paid Employment (-1.584) (-0.634) (-1.151) (-0.699) (0.257) (0.971E-01)
Respondent’s ------- ------- 0.128 ------- ------- 0.235 ------- ------- 0.691E-01 -------Share of Total (1.056) (0.809) (0.621Weeks of PaidEmployment
Constant -1.440 -1.401 -1.505 0.189 0.292 0.755E-01 -1.431 -1.431 -1.430 0.208(-33.69) (-27.66) (-13.23) (1.844) (2.310) (0.278) (-35.94) (-30.11) (-29.98) (2.210)
% Observations at ------- ------- ------- 0.595 0.595 0.595 ------- ------- ------- 0.643One
% Right ------- ------- ------- 0.596 0.597 0.596 ------- ------- ------- 0.645Observations
Maddala R- ------- ------- ------- 0.428E-02 0.613E-02 0.674E-02 ------- ------- ------- 0.461E-02Squared
Cragg-Uhler R- ------- ------- ------- 0.578E-02 0.827E-02 0.909E-02 ------- ------- ------- 0.633E-02Squared
Adjusted R- 0.003 0.004 0.003 ------- ------- ------- 0.003 0.003 0.002 -------Squared
Sample Size 1619 1619 1619 1619 1619 1619 1764 1764 1764 1764
*T-ratios in parenthesisIndex of Happiness: 1 = Very Unhappy; 2 = Somewhat Unhappy; 3 = Somewhat Happy; 4 = Very Happy
In fairness, we must acknowledge that our measures of labour supply are not ideal. It15
would be preferable to have an exact measure of annual hours rather than one we have estimated. Nonetheless, measures such as the one used here are common in the literature.
statistically significant determinants of happiness. 15
Second, measures of time spent in paid employment are also a bad proxy for the
theoretical idea of `leisure time,’ given that time outside paid employment may be spent in many
ways others than leisure (e.g., engaging in household production, or involuntary unemployment).
This is most obviously true for individuals who are full-time home-makers. Table 2 repeats the
analysis of Table 1 (without the various disaggregations) replacing labour supply with a dummy
variable indicating whether or not the respondent felt `satisfied with time for own interests.’ We
argue that this is a much better proxy for leisure time and that, in fact, this self-assessment has
advantages over time use data in which it is the investigator, rather than the respondent, who
decides whether to code activities as “ leisure” or not.
As textbooks on “leisure studies” note, defining the field of leisure studies is difficult,
because there is practically no activity that some people do for leisure that other people do not
also do for money. And in addition to the general difficulty of labelling types of activities as
“work” or “leisure”, there is the problem of context. For example, taking one’s two year old child
to the park on a sunny day when he is well-behaved may be seen by many perents as “leisure”.
When it is cold and miserable and he takes a temper tantrum, it may be regarded more as child
care “work”. Some individuals may regard cooking as recreation; others may regard it as a chore
(and this, too, may vary from day to day).
It is certainly clear from Table 2 that `being satisfied with time for own interests’ is a
strongly significant and positive determinant of happiness. Household equivalent income also
16
Table 2Income and Time Satisfaction as Determinants of Happiness by Gender
Independent Husband WifeVariable
Probit OLS Index of Probit OLS IndexVery Happy Happiness Very of
Happy Happiness
Household Equivelant 0.777E-05 0.283E-05 0.159E-04 0.610E-05Income (1.797) (1.601) (3.701) (3.541)
Dummy=1 if Satisfied With 0.422 0.182 0.565 0.252Time for Other Interests (6.266) (6.520) (8.519) (9.189)
Constant -0.173 -1.592 -0.268 -1.656(-1.934) (-43.245) (-2.925) (-44.205)
% Observations at One 0.595 -------- 0.643 --------
% Right Predictions 0.608 -------- 0.656 --------
Maddala R-Squared 0.267E-01 -------- 0.448E-01 --------
Cragg-Uhler R-Squared 0.361E-01 -------- 0.615E-01 --------
Adjusted R-Squared ----------- 0.027 ----------- 0.048
Sample Size 1619 1619 1764 1764
*T-ratios in parentheses
Index of Happiness: 1 = Very Unhappy; 2 = Somewhat Unhappy; 3 = Somewhat Happy; 4 =Very Happy.
17
remains positive and significant in this specification. We therefore argue that the variable “satisfied
with time for other interests” is a relatively good proxy for leisure time - the point of this section
is to justify the emphasis placed on `satisfaction with time’ in the rest of the paper since this
variable is strongly linked with over-all assessments of well-being -- much more so than the more
traditional economic construct of labour supply.
4. Do Husbands have more leisure than Wives?
This section explores gender differences in levels of satisfaction with time and in
determinants of satisfaction with time. Table 3 presents basic descriptive information about
responses to the `satisfaction with time’ question for husbands and wives, by major activity (paid
employment, looking for a paid job and keeping house). For the vast majority of both husbands
and wives, the main activity is a paid job (97 percent of men; 65 percent of women). A significant
minority (33.5 percent) of wives keep house, but very few men (only 8; less than 0.5 percent of
the sample) list keeping house as their major activity. Thus, we do not have statistically reliable
results for homemaker husbands.
Homemaker wives are in general satisfied with time available for themselves -- 80 percent
are satisfied; 45 percent are very satisfied. Nonetheless, 20 percent report themselves to be
dissatisfied (5 percent are very dissatisfied). [In general, respondents are more willing to identify
themselves as dissatisfied with “time available for themselves” than to report themselves as
unhappy with life over-all (only 1 percent of all husbands; 2 percent of all wives) - a finding which
increases our confidence in the use of this variable.].
18
Sample means are statistically different at the 1 percent level.16
Table 3Male/Female Differences in Satisfaction with Time for Self by Major Activity
Degree ofSatisfaction
Main Activity of Respondent
Husband Wife
Paid Job Paid Job Looking Keeping Paid Job Looking Keepingand and Spouse For Paid House For Paid House
Spouse in Keeping Job JobPaid Job House
Very 0.066 0.073 0.077 --------- 0.084 0.077 0.047Dissatisfied
Somewhat 0.243 0.226 0.179 * 0.271 0.077 0.155Dissatisfied
Somewhat 0.367 0.371 0.308 * 0.364 0.5 0.347Satisfied
Very Satisfied 0.324 0.331 0.436 * 0.280 0.346 0.450
Satisfied - --------- --------- --------- 0.001 ---------- 0.002 Degree NotStated
Sample 1551 39 8 1113 26 574
Husbands and wives with paid employment are significantly less satisfied with time
available for themselves than are homemakers. Wives with paid jobs are less satisfied with time16
for themselves than are husbands with paid jobs. Thirty-five percent of wives; 30 percent of
husbands who work outside the home report themselves to be dissatisfied. Looked at another
way, only 33 percent of husbands; 28 percent of wives report themselves to be very satisfied with
time available for themselves. These differences are statistically significant at the 1 percent level -
but whether they are “large” is an assessment we will leave to readers. Thus, while very few
respondents are willing to report themselves to be unhappy with life over-all, it does seem that
availability of time is a trouble spot - one which has the potential for generating inequities within
19
the home.
An interesting point illustrated by Table 3 is that husbands with wives who are full-time
homemakers are not more satisfied with time available for themselves than are husbands with
wives who work outside the home. An explanation for this finding could be that the division of
household duties continues to fall along gender lines. If husbands do not significantly increase
their share of domestic responsibilities when their wives take on paid employment, then there is no
particular reason why they should feel less satisfied with time than when their wives worked full-
time at home.
Which factors, then, increase the probability that an individual will be satisfied with time
available for self? Are the same factors significant for men and women? Table 4 reports the
results of OLS regressions using the index of time satisfaction as dependent variable; Table 5
reports probit results where the dependent variable is a dummy equal to one if the individual is
satisfied with time (including both somewhat and very satisfied with time); Table 6 reports probit
results where the dependent variable is a dummy equal to one if the individual is very satisfied
with time.
Explanatory variables include income, labour supply, number and ages of children present
in the family, level of education of the respondent, and age of the respondent. We include the
child variables with the idea that children present in the family, particularly younger children, will
increase the amount of work which must be done within the home. We include both the
continuous variable `number of children’ and a dummy to indicate presence of any children since
we think that not only does each additional young child cause an increase in parental work load,
there is also a discrete difference between families with and without children (Browning, 1992).
20
It is, of course, possible to have both a preschooler and a teenager present in the family.17
Experimentation indicated that the categorical variables were more useful than a18
continuous variable, even with a quadratic included.
We also include two dummy variables to indicate, respectively, presence of a pre-schooler and
presence of a teenager.17
We include 5 dummy variables to indicate highest level of education completed by the
respondent (high-school education is the base), because there is some evidence (Marshall, 1993)
that individuals with higher levels of education are more likely to share household chores. As a
result, we (and in particular, the male co-authors of this paper) consider it highly probable that
men with higher levels of education are less satisfied with time for themselves.
We include two categorical variables for age of respondent (35-44 years and 45 to 54
years with 25 to 34 being the base) in an effort to capture possible life-cycle effects in18
satisfaction. The psychological literature appears to suggest, for example, that happiness in
general increases as an individual ages and expectations diminish (Argyle, 1987). [It is also
possible, however, that these variables measure cohort effects as people of different ages grow up
with different expectations and opportunities.]
We expect income to affect satisfaction with time since additional income, other things
constant, provides people with a greater opportunity to purchase time-saving services or capital
goods which can save time for themselves (e.g. dishwashers or clothes washers and dryers). More
money might also allow people to purchase help with the care of their children (e.g., a live-in
nanny), help with household chores (e.g., daily or weekly maid services), the ability to live in a
more expensive neighbourhood close to the office or the ability to eat out. We include both total
21
household income and respondent’s share of that income, since earlier work (Phipps and Burton,
1996) indicates that patterns of expenditures may depend very much on whose income is
available. For example, child care expenditures increased only with wife’s income even if both
husband and wife have full-time full-year paid employment.
Finally, we include a series of labour supply variables, since working more hours in the
paid labour market will reduce total hours available for other pursuits, very likely reducing time
available for self. As with the previous regressions, we include the labour supply variables in
three different ways. Column (a) includes a single indicator of total annual household hours.
Column (b) disaggregates the separate effects of total household weeks and total household hours
per week. Column (c) adds the respondent’s share of annual weeks and weekly hours. Our
discussion of results focuses on Table 6, which reports our probit analysis of the factors
increasing the probability of being `very satisfied with time for self,’ - differences with other
regressions are noted where appropriate.
We do not find it surprising that the presence of a pre-school child in the household
reduces the probability of being very satisfied with time for both men and women. Men without
any children are more likely to be satisfied with time (though this variable is only significant at
22
Table 4OLS Estimates of Index of Satisfaction With Time for Self by Gender
Main Activity=Paid Work
Independent Husband WifeVariable
(a) (b) (c) (a) (b) (c)
Paid Employment Hours -0.172E-03 ---------- ---------- -0.112E-03 ----------- ------------- Total Household (-7.556) (-3.947)- Resp. ShareWeekly hrs.- Total Weeks
Total Weekly Hours of Paid _______ -0.703E-02 -0.117E-01 ------------ -0.790E-02 -0.945E-02Employment (-4.967) (-7.570) (-4.781) (-5.493)
Respondent’s Share of Total ------------ ----------- -1.295 ------------ ----------- -0.886Weekly Hours of Paid (-5.990) (-3.360)Employment
Total Weeks of Paid _______ -0.742E-03 -0.640E-02 ----------- 0.247E-02 -0.373E-03Employment (-0.572) (-3.145) (1.341) (-0.181E-01)
Respondent’s Share of Total_______ ----------- -0.669E-01 ----------- ------------ 0.742E-01Weeks of Paid Employment (-0.245) (0.259)
Number of Children -0.130E-01 -0.167E-01 0.735E-02 0.322E-01 0.390E-01 0.269E-01(-0.379) (-0.481) (0.215) (0.678) (0.823) (0.569)
Dummy=1 if No Children 0.145 0.122 0.130 0.499 0.498 0.511Present (1.303) (1.091) (1.183) (3.274) (3.272) (3.380)
Dummy=1 if Preschooler -0.175 -0.166 -0.148 -0.981E-01 -0.106 -0.107Present (-2.455) (-2.314) (-2.092) (-1.041) (-1.129) (-1.148)
Dummy=1 if Teenager 0.769E-01 0.733E-01 0.579E-01 0.232 0.234 0.259Present (0.979) (0.925) (0.742) (2.112) (2.135) (2.375)
Dummy=1 if -0.216 -0.216 -0.246 0.810E-02 0.217E-01 0.403E-01Master/Doctorate (-2.020) (-2.005) (-2.316) (0.522E-01) (0.139) (0.260)
Dummy=1 if -0.147 -0.152 -0.157 -0.921E-01 -0.757E-01 -0.954E-01Bachelor/Undergraduate (-2.062) (-2.117) (-2.229) (-1.175) (-0.967) (-1.221)
Dummy=1 if Diploma from -0.164 -0.163 -0.177 -0.192E-01 -0.285E-01 -0.450E-01College (-1.845) (-1.814) (-2.001) (-0.232) (-0.346) (-0.547)
Dummy=1 if Diploma from -0.799E-01 -0.782E-01 -0.963E-01 -0.186 -0.189 -0.198Trade School (-1.242) (-1.207) (-1.508) (-2.154) (-2.201) (-2.307)
Dummy=1 if Less than High -0.858E-02 -0.826E-02 0.200E-01 0.161E-01 0.122E-01 0.503E-01School (-0.585E-01) (-0.559E-01) (0.137) (0.643E-01) (0.490E-01) (0.203)
Total Household Income 0.385E-05 0.272E-05 0.335E-05 -0.256E-05 -0.399E-05 -0.302E-05(2.285) (1.584) (1.979) (-1.330) (-2.050) (-1.549)
Respondent’s Share of Total -0.243 -0.233 -0.113 -0.311 -0.298 -0.168Income (-2.917) (-2.696) (-1.296) (-2.694) (-2.556) (-1.378)
23
Table 4OLS Estimates of Index of Satisfaction With Time for Self by Gender
Main Activity=Paid Work
Dummy=1 if 0.160 0.163 0.162 -0.978E-02 0.194E-01 0.116E-02Household Income is (1.888) (1.893) (1.915) (-0.996E-01) (0.197) (0.118E-01)Greater than $80,000
Dummy=1 if Age 35-44 0.412E-01 0.486E-01 0.486E-01 -0.283E-01 -0.315E-01 -0.289E-01(0.704) (0.824) (0.837) (-0.432) (-0.483) (-0.445)
Dummy=1 if Age 45-54 0.154 0.161 0.187 0.315 0.317 0.311(2.247) (2.326) (2.738) (4.081) (4.099) (4.030)
Constant 3.439 3.473 5.043 3.150 3.167 3.846(21.566) (19.694) (14.496) (15.605) (13.030) (11.170)
Adjusted R-Squared 0.051 0.039 0.069 0.08 0.085 0.095
Sample Size 1551 1551 1551 1112 1112 1112
*T-ratios in parentheses
Index of Satisfaction: 1 = Very Dissatisfied; 2 = Somewhat Dissatisfied; 3 = Somewhat Satisfied; 4 = Very Satisfied.
24
Table 5Probit Estimates of the Probability of Being Satisfied With Time for Other Interests
Main Activity=Paid Work
Independent Husband WifeVariable
(a) (b) (c) (a) (b) (c)
Total -0.279E-03 ---------- ---------- -0.985E-04 ----------- ------------Household (-7.984) (-2.326)Hours of PaidEmployment
Total Weekly ---------- -0.103E-01 -0.183E-01 ------------ -0.893E-02 -0.111E-01Hours of Paid (-4.924) (-7.508) (-3.584) (-4.137)Employment
Respondent’s ------------ ------------ -1.944 ------------ ----------- -1.559Share of Total (-5.717) (-3.684)Weekly Hoursof PaidEmployment
Total Weeks of _______ -0.269E-02 -0.154E-01 ----------- 0.485E-02 0.197E-02Paid (-1.408) (-4.258) (1.748) (0.629)Employment
Respondent’s _______ ------------ -0.632 ----------- ------------ 0.648Share of Total (-1.350) (1.509)Weeks of PaidEmployment
Number of -0.315E-01 -0.410E-01 -0.235E-02 0.112 0.122 0.104Children (-0.615) (-0.805) (-0.448E-01) (1.574) (1.711) (1.447)
Dummy=1 if 0.198 0.156 0.161 0.815 0.823 0.839No Children (1.203) (0.957) (0.971) (3.608) (3.306) (3.674)Present
Dummy=1if -0.240 -0.229 -0.215 0.315E-02 -0.553E-02 -0.680E-02Preschooler (-2.294) (-2.195) (-2.028) (0.229E-01) (-0.402E-01) (-0.493E-01)Present
Dummy=1 if 0.149 0.138 0.114 0.293 0.298 0.332Teenager (1.291) (1.202) (0.979) (1.847) (1.874) (2.073)Present
Dummy=1 if -0.291 -0.287 -0.337 -0.181 -0.155 -0.137Master/Doctora (-1.842) (-1.824) (-2.113) (-0.806) (-0.685) (-0.603)te
Dummy=1 if -0.120 -0.124 -0.136 -0.197 -0.176 -0.198Bachelor/Under (-1.131) (-1.170) (-1.270) (-1.708) (-1.519) (-1.685)graduate
Dummy=1 if -0.198 -0.191 -0.228 0.318E-01 0.182E-01 -0.760E-02Diploma from (-1.512) (-1.465) (-1.732) (0.260) (0.149) (-0.617E-01)College
25
Table 5Probit Estimates of the Probability of Being Satisfied With Time for Other Interests
Main Activity=Paid Work
Dummy=1 if -0.152 -0.149 -0.172 -0.213 -0.221 -0.227Diploma from (-1.595) (-1.576) (-1.796) (-1.674) (-1.725) (-1.765)Trade School
Dummy=1 if 0.766E-01 0.808E-01 0.899E-01 -0.197 -0.182 -0.894E-01Less than High (0.345) (0.364) (0.409) (0.526) (-0.476) (-0.231)School
Total 0.745E-05 0.606E-05 0.732E-05 -0.743E-05 -0.939E-05 -0.785E-05Household (2.965) (2.407) (2.858) (-2.575) (-3.194) (-2.632)Income
Respondent’s -0.417 -0.407 -0.219 -0.311 -0.293 -0.153Share of Total (-3.137) (-3.025) (-1.592) (-1.792) (-1.683) (-0.822)Income
Dummy=1 if 0.177 0.163 0.175 0.128 0.171 0.147Household (1.363) (1.265) (1.338) (0.888) (1.174) (1.004)Income isGreater than $80,000
Dummy=1 if -0.262E-01 -0.110E-01 -0.208E-01 0.138E-01 0.957E-02 0.121E-01Age 35-44 (-0.305) (-0.129) (-0.240) (0.144) (0.995E-01) (0.126)
Dummy=1 if 0.142 0.148 0.188 0.458 0.469 0.458Age 45-54 (1.356) (1.419) (1.777) (3.755) (3.811) (3.674)
Constant 1.335 1.419 4.538 0.621 0.562 1.298(5.439) (5.296) (7.147) (2.079) (1.533) (2.433)
% Observations 0.676 0.676 0.676 0.638 0.638 0.638at One
% Right 0.682 0.685 0.696 0.659 0.654 0.662Predictions
Mandalla R- 0.061 0.047 0.08 0.071 0.077 0.089Squared
Cragg-Uhler R- 0.082 0.066 0.111 0.097 0.106 0.122Squared
Sample Size 1551 1551 1551 1113 1113 1113
*T-ratios in parentheses
26
Table 6Probit Estimates of the Probability of Being Very Satisfied With Time for Other Interests
Main Activity=Paid Work
Independent Husband WifeVariable
(a) (b) (c) (a) (b) (c)
Total Household Hours of Paid -0.171E-03 ---------- ---------- -0.202E- ----------- ------------Employment (-4.865) 03
(-4.378)
Total Weekly Hours of Paid ----------- -0.692E-02 -0.129E-01 ------------ -0.116E-01 -0.142E-01Employment (-3.252) (-5.181) (-4.330) (-4.933)
Respondent’s Share of Total ------------ ----------- -1.686 ------------ ----------- -1.024Weekly Hours of Paid (-5.027) (-2.493)Employment
Total Weeks of Paid _______ -0.496E-03 -0.480E-02 ----------- 0.676E-03 -0.372E-02Employment (-0.258) (-1.618) (0.237) (-1.159)
Respondent’s Share of Total _______ ----------- 0.324 ----------- ------------ -0.203Weeks of Paid Employment (0.803) (-0.451)
Number of Children 0.197E-01 0.165E-01 0.406E-01 -0.385E- -0.265E-01 -0.489E-01(0.387) (0.324) (0.788) 01 (-0.348) (-0.633)
(-0.503)
Dummy=1 if No Children 0.291 0.262 0.274 0.185 0.183 0.200Present (1.749) (1.580) (1.643) (0.751) (0.746) (0.807)
Dummy=1 if Preschooler -0.231 -0.218 -0.207 -0.278 -0.282 -0.292Present (-2.137) (-2.019) (-1.896) (-1.716) (-1.747) (-1.798)
Dummy=1 if Teenager Present0.174 0.166 0.150 0.111 0.110 0.148(1.462) (1.395) (1.252) (0.616) (0.611) (0.815)
Dummy=1 if Master/Doctorate -0.227 -0.225 -0.273 0.216 0.217 0.229(-1.409) (-1.402) (-1.670) (0.893) (0.892) (0.938)
Dummy=1 if -0.405 -0.409 -0.426 0.202E-01 0.363E-01 0.162E-02Bachelor/Undergraduate (-3.675) (-3.706) (-3.828) (0.162) (0.289) (0.127E-01)
Dummy=1 if Diploma from -0.234 -0.231 -0.257 0.125E-01 -0.212E-02 -0.207E-01College (-1.713) (-1.696) (-1.872) (0.949E- (-0.161E-01) (-0.157)
01)
Dummy=1 if Diploma from -0.116 -0.112 -0.140 -0.242 -0.238 -0.253Trade School (-1.212) (-1.181) (-1.460) (-1.725) (-1.697) (-1.803)
Dummy=1 if Less than High -0.342 -0.340 -0.298 0.279 0.258 0.294School (-1.525) (-1.521) (-1.329) (0.742) (0.678) (0.777)
Total Household Income 0.113E-05 0.177E-06 0.563E-06 -0.242E- -0.420E-05 -0.295E-05(0.450) (0.696E-01) (0.219) 05 (-1.384) (-0.957)
(-0.807)
27
Table 6Probit Estimates of the Probability of Being Very Satisfied With Time for Other Interests
Main Activity=Paid Work
Respondent’s Share of Total -0.254 -0.234 -0.103 -0.531 -0.528 -0.327Income (-1.908) (-1.724) (-0.738) (-2.991) (-2.944) (-1.703)
Dummy=1 if Household 0.312 0.316 0.321 -0.129 -0.911E-01 -0.124Income is Greater than (2.473) (2.501) (2.517) (-0.796) (-0.563) (-0.760)$80,000
Dummy=1 if Age 35-44 0.140 0.145 0.143 -0.132 -0.136 -0.130(1.587) (1.645) (1.617) (-1.248) (-1.285) (-1.228)
Dummy=1 if Age 45-54 0.301 0.309 0.335 0.406 0.405 0.409(2.985) (3.051) (3.288) (3.544) (3.512) (3.510)
Constant -0.134E-01 -0.354E-02 1.550 0.382 0.531 1.608(-0.559E-01) (-0.135E-01) (2.997) (1.180) (1.383) (2.936)
% Observations at One 0.311 0.311 0.311 0.262 0.262 0.262
% Right Predictions 0.701 0.692 0.67 0.726 0.73 0.738
Mandalla R-Squared 0.048 0.042 0.061 0.073 0.074 0.083
Cragg-Uhler R-Squared 0.067 0.059 0.086 0.107 0.109 0.121
Sample Size 1551 1551 1551 1112 1112 1112
*T-ratios in parentheses
28
Results for the other two regressions, reported in Tables 4 and 5, respectively, differ19
somewhat for wives. In both cases, having any children significantly reduces the probability ofwomen being satisfied with time for self; having a pre-schooler is not statistically different fromhave a young school-aged child; having a teen-ager increases the probability of being satisfiedwith time. For men, the pre-schooler dummy remains the only significant child variable.
It may also be that men with higher levels of education are not so much “kitchen20
drones” as avid readers of existential philosophy, who view the all too rapid passing of life ingeneral more bleakly. However, in the psychological literature most studies find individuals withhigher levels of education to have a greater level of satisfaction with life over-all (Argyle, 1987, p.96).
quite low levels), but the same is not true for women. Notably, controlling for preschoolers and
the presence of any children, the number of children in the family and presence of a teen-ager are
not statistically significant.19
Education appears to be more important for men than for women in determining the
likelihood of being satisfied with time. Education is in almost every case insignificant for women.
(The exception is that women with diplomas from trade schools are less likely to be satisfied with
time for self than are women with high-school level education.) For men, on the other hand,
additional education reduces the probability of being satisfied with time for self. Since we are
controlling for labour supply, the explanation for this finding is not that men with more education
work longer hours in the paid labour market. Rather, this may indicate that men with higher
levels of education are more likely to share household responsibilities with their wives, which
would have the effect of reducing time available for self relative to other men working the same
number of hours in the paid labour market but with lower levels of education. 20
However, husbands and wives appear to disagree, on average, about exactly how much
each is contributing to inside home maintenance - although both agree that wives do more
(though men do more work outside the home). Wives with paid jobs report that they do 76
29
percent of meal preparation, 69 percent of meal clean-up, 77 percent of cleaning and laundry and
21 percent of outside maintenance. Husbands, on the other hand, report that they do 39 percent
of meal preparation (a total of 118 percent, not counting possible contributions of other
household members!), 44 percent of meal clean-up (a total of 113 percent), 33 percent of laundry
and cleaning (a total of 110 percent) and 76 percent of outside maintenance (in contrast, a total of
97 percent).
With respect to the effects of education on contributions to home and yard work, women
with more education claim to do a smaller share of meal clean-up and laundry/cleaning than do
less well-educated women (which isn’t true for cooking or outside maintenance). Men with higher
levels of education report themselves as doing a significantly larger share of inside work and a
significantly smaller share of outside maintenance than less educated men.
Why do women with higher education, who report that their spouses are more likely to
share domestic duties with them, not report being more satisfied with time for themselves than
are less-educated women? One possible explanation lies in expectations. It could be that more
well-educated women have very high expectations of their spouses. In this case, even if their
husbands do more home work than other men, they may still not live up to their wives’
expectations, so these women may not feel happier with available time than others.
For both men and women, the probability of being very satisfied with time for self
increases with age (though for women, there is no statistically significant difference between
women aged 25 to 34 and those aged 35 to 44). Individuals most likely to be very satisfied with
time for self, controlling for the presence and age of children, are those aged 45 to 54. (Recall
that older individuals have been excluded from the sample.)
30
Tables 4, 5 and 6 differ in their implications for the impact of household income on
satisfaction with time. In Table 6, the over-all level of household income does not affect the
probability that either husbands or wives will be satisfied with time for self, suggesting that
purchase of substitute services cannot entirely relieve the time crunch for more affluent couples.
However, making a larger contribution to household income reduces the probability of being very
satisfied with time in all specifications for men and women, and the effect is usually statistically
significant. Since we have already controlled for total household income and for hours of labour
supply, the explanation may lie in a greater intensity of work or in greater aspirations for a share
of household leisure time which people feel would be “fair” as they bring in more of the
household’s income. The magnitude (and the significance) of this effect is greater for women.
For men, contribution to household income is less important than the dummy variable indicating
that household income is higher than $80,000 annually. Since men, on average, report that they
contribute more to household income than women, men living in high-income households are
more likely to do so because of their own high earnings. For men, there may be some solace in
being a high-income earner, over and above the share of family income or the level of family
income - which one could interpret as a status effect.
It should also be noted that average household income reported by male and female
respondents included in our sample is nearly identical at about $51,000. Fourteen percent of men
live in households with incomes above $80,000 annually; 12 percent of women live in high-
income households.) However, while the total incomes reported from both sides are consistent,
the estimation of shares is not. The average share of total household income which is reported to
be contributed by husbands is 72 percent, while the average share of income contributed by wives
31
The phenomenon of mutually inconsistent estimates of relative21
contributions/entitlements, within a clearly defined total, has also been noted by Woolley andMarshall (1994), who report that among young Canadian couples 90% of the males and 80% ofthe females felt that the other partner had more closet space.
is 42 percent. 21
It is perhaps not surprising that the total of perceived shares of housework seems to be
more than 100%, since very few households keep any formal record of household tasks performed
and it may be human nature to exaggerate somewhat the importance of one’s own contribution.
However, it is surprising that husbands and wives can, on average, essentially agree on the total
level of household income (which must come from one or the other), yet disagree about where it
comes from.
Finally, we find, not surprisingly, that the greater the total number of hours supplied to the
paid labour market annually, the lower the probability of either husbands or wives reporting
themselves to be very satisfied with time for self (see specification a). When total annual hours
are disaggregated into weekly hours and total annual weeks, we find that it is total weekly hours
which is the significant (and negative) factor, confirming the ideas of Marshall (1993). Finally,
when we take account of the respondent’s own personal labour supply, we again find that it is
weekly hours which matter most and that the respondent’s own weekly hours worked are very
important in determining the probability that he or she is personally satisfied with time available
for self.
32
5. Simulation Results
Our probit analysis of the probability of being very satisfied with time for self can be used
to assess the relative quantitative importance of two important variables: income and labour
supply. We evaluate the change in the probability of being very satisfied with time for a husband
and wife with average sample characteristics if: 1) the household had an additional $10,000 of
annual income (about 20 percent of mean income) - with no change in hours worked; 2) hours of
paid employment were reduced by 15 hours per week (about 20 percent of mean hours) - with no
reduction in income. As well, since we have emphasized that it may matter who receives the
income or who supplies the hours of paid labour, we calculate the difference it makes whether the
increase in income is received entirely by the respondent, entirely by the spouse or 50/50. We
follow the same procedure for the reduction in hours of paid employment.
Results of the simulation exercise are reported in Table 7. The first point which is very
clear is that an additional $10,000 of income (holding labour supply constant) will not be of much
help in relieving the time crunch, regardless of who receives the money. (In fact, as Table 6
reports, Household Income is statistically insignificant as a predictor of the probability of being
very satisfied with time for self.) This suggests that money is of only limited usefulness in
purchasing services for work traditionally done within the home.
A reduction in hours of labour supply (holding income constant), on the other hand, is
more beneficial. The probability of the average husband being satisfied with time for himself
increases from 24.5 percent to 36.3 percent if his paid employment time falls by 15 hours per
week. The probability of the average wife being very satisfied with time available for herself
33
Table 7Change in the Probability of Being Very Satisfied with Time For Other Interests
Initial Probability
Average Respondent
Husband Wife
0.245 0.212
$10,000 more for household(work hours constant)*
All earned by respondent
50/50
All earned by spouse
0* -0.170*
0.003* -0.012*
0.006* -0.003*
Fifteen less hours of paid employment for householdper week (income constant)
All received by respondent
50/50
All received by spouse
0.118 0.111
0.039 0.066
-0.03 0.03
*[Note: Relevant coefficient not statistically significantly different from zero].
increases from 21.2 percent to 32.3 percent if she reduces her labour supply or to 24.2 percent if
her husband reduces his labour supply. In both cases, the impact on personal time satsifaction is
greater if the respondent is the one who works fewer hours in the paid labour market, but there is
still an important effect if the spouse works fewer hours, suggesting that the one who reduces
labour supply uses the freed-up time in part for personal leisure but in part for work within the
home.
34
6. Conclusions
This paper is an initial attempt to see what can be learned about the distribution of well-
being within families from subjective data on happiness/satisfaction. In particular, we study the
possibility that continued traditional gender roles with respect to the division of work done within
the home have generated inequities as more and more women work outside the home. While we
hope that this research adds a little to our understanding of what goes on within families, we want
to emphasize what we feel to be a major limitation of this approach -- subjective assessments of
personal well-being will depend upon personal expectations which are likely to be conditioned be
social norms and personal experience. Thus, it is possible for example, that women may report
themselves to be as happy as men because they have learned to expect less. With this very
important qualification in mind, we would summarize our findings to date as follows:
1. Income is an important determinant of over-all happiness, but standard measures of labour
supply are not. A better proxy for available leisure time than time not spent in paid labour
is the degree to which an individual feels satisfied with time for self.
2. Women who work outside the home are about 5% less likely to be satisfied with time
available for themselves than men who work outside the home. Home-maker wives are
more likely to be satisfied with time than either men or women with paid employment,
however their husbands are not more likely to be satisfied with time.
3. Pre-schoolers in the family reduce satisfaction with time for self; satisfaction with time
increase with age. This is true for both men and women.
4. The effect of education is not the same for both men and women. Education appears to
play a much larger role in determining male satisfaction with time for self. Further,
35
looking at education makes the important point that with subjective data, responses given
by husbands and wives may differ. Men with more education say that they do a much
larger share of work within the home than other men. And, men with more education are
less satisfied with the time they have for themselves. However, the wives of men with
more education do not agree that their husbands work harder at home. These women are
not more satisfied than other wives with time for themselves.
5.3. Additional income, labour supply held constant, is not particularly useful for alleviating the
time crunch. Thus, it would seem that the purchase of substitutes for home production is
limited. Fewer hours of labour supply, income held constant, makes a relatively much
more important impact on satisfaction with available time.
36
Appendix AMean and Standard Deviation of Variables by Activity and Gender
Variable Husband Wife
Dummy=1 if Happy Job JobFull Sample Main Activity: Paid Full Sample Main Activity: Paid+
0.98561 0.98711 0.97782 0.97574(0.11914) (0.11286) (0.14732) (0.15392)
Dummy=1 if Very Happy 0.60889 0.61057 0.65791 0.65588(0.48815) (0.48778) (0.47455) (0.47529)
Index of Happiness -1.4080 -1.4049 -1.3666 -1.3711(0.52963) (0.52654) (0.53487) (0.54124)
Dummy=1 if Satisfied With Time 0.69650 0.69504 0.69936 0.64510(0.45992) (0.46054) (0.45867) (0.47870)
Dummy=1 if Very Satisfied With 0.33417 0.33011 0.33800 0.28032Time (0.47185) (0.47040) (0.47317) (0.44936)
Index of Satisfaction With Time 2.9637 2.9581 2.9685 2.8437(0.91529) (0.91379) (0.92546) (0.93050)
Total Household Hours of Paid 3394.9 3440 3291.9 3813.8Employment (1173) (1141.7) (1224.2) (1017.1)
Total Weekly Hours of Paid 72.038 72.201 70.185 78.970Employment (21.395) (21.1) (22.728) (17.906)
Respondent’s Share of Total Weekly 0.67065 0.67520 0.36729 0.48212Hours of Paid Employment (0.21020) (0.20522) (0.23117) (0.14297)
Total Weeks of Paid Employment 81.421 82.502 81.206 94.261(25.225) (24.271) (26.505) (17.093)
Respondent’s Share of Total Weeks 0.64885 0.65612 0.38987 0.51852of Paid Employment (0.22874) (0.22038) (0.24065) (0.13302)
Household Equivalent Income 16801 17011 15877 17999(7351.1) (7302.9) (7404.4) (7340.8)
Number of Childern 1.4237 1.4223 1.3654 1.1204(1.1253) (1.1275) (1.1360) (1.0762)
Dummy=1 if No Childern 0.29287 0.29594 0.32283 0.40970Present (0.45522) (0.45661) (0.46769) (0.49200)
Dummy=1 if Preschooler 0.20526 0.20245 0.17688 0.14286Present (.40402) (0.40196) (0.38168) (0.35008)
Dummy=1 if Teenager 0.56258 0.55964 0.53707 0.49865Present (0.49622) (0.49659) (0.49877) (0.50022)
Dummy=1 if Master/Doctorate 0.50688E-01 0.21921E-01 0.19848E-01 0.50314E-01(0.21943) (0.14647) (0.13952) (0.21869)
37
Appendix AMean and Standard Deviation of Variables by Activity and Gender
Dummy=1 if 0.13579 0.13926 0.13835 0.13567Bachelor/Undergraduate (0.34268) (0.34633) (0.34537) (0.34259)
Dummy=1 if Diploma from College 0.70713E-01 0.99291E-01 0.12142 0.61995E-01(0.25643) (0.29915) (0.32672) (0.24125)
Dummy=1 if Diploma from Trade 0.17334 0.47711E-01 0.11617 0.97934E-01School (0.37866) (0.21322) (0.32052) (0.29736)
Dummy=1 if Less than High School 0.27534E-01 0.21277E-01 0.16346E-01 0.36837E-01(0.16369) (0.14435) (0.12684) (0.18845)
Total Household Income 50452 51090 47083 51473(19204) (18998) (19174) (18296)
Respondent’s Share of Total Income 0.71896 0.72312 0.32581 0.42108(0.28041) (0.27813) (0.30259) (0.24599)
Dummy=1 if Household Income is 0.13517 0.13926 0.90485E-01 0.11590Greater than $80,000 (0.34201) (0.34633) (0.28696) (0.32025)
Dummy=1 if Age 35-44 0.38235 0.38685 0.30706 0.31087(0.48611) (0.48719) (0.46141) (0.46306)
Dummy=1 if Age 45-54 0.17960 0.18182 0.15236 0.15454(0.38397) (0.38582) (0.35948) (0.36163)
Sample Size 1598 1551 1713 1113
*Standard Deviation in parenthesesFull Sample: Respondents whose Main Activity is either Paid Employment, Looking for Work, or Keeping House+
38
References
Annas, Julia. 1993. “Women and the Quality of Life: Two Norms or One?” in MarthaNussbaum and Amartya Sen (eds.). The Quality of Life. Oxford: Clarendon Press, pp279-296.
Apps, Patricia. 1994. “Female Labour Supply, Housework and Family Welfare,” in TheMeasurement of Household Welfare. Richard Blundell, Ian Preston and Ian Walker(eds.). Cambridge, University Press, 140-214.
Argyle, Michael. 1987. The Psychology of Happiness. London: Methuen and Co., Ltd.
Argyle, Michael and Maryanne Martin. 1989. “The Psychological Causes of Happiness” inStrack, Fritz, Michael Argyle and Norbert Schwarz (eds.). Subjective Well-Being: AnInterdisciplinary Perspective. Oxford: Pergamon Press, 77-100.
Becker, Gary. 1974. “A Theory of Marriage,” in Economics of the Family, T.W. Schultz (ed.). Chicago: University of Chicago Press.
Becker, Gary. 1981. A Treatise on the Family. Cambridge: Harvard University Press.
Browning, Martin. 1992. “Children and Household Economic Behavior,” The Journal ofEconomic Literature, 30, 1434-1475.
Browning, Martin, Francois Bourguignon, Pierre-Andre Chiappori and Valerie Lechene. 1994. “Incomes and Outcomes: A Structural Model of Intra-Household Allocation” Journal ofPolitical Economy. 102:6, 1067-96.
Chiappori, Pierre-Andre, Lawrence Haddad, John Hoddinott and Ravi Kanbur. 1992. “Unitaryversus Collective Models of the Household: Time to Shift the Burden of Proof” Paperpresented at the American Economic Association Meetings, Annaheim, California, January5-7, 1992.
Clark, Andrew and Andrew Oswald. 1994. “Unhappiness and Unemployment.” The EconomicJournal. 104, 648-659.
Furnham, Adrian. 1991. “Work and Leisure Satisfaction” in Subjective Well-Being: AnInterdisciplinary Perspective. Strack, Fritz, Michael Argyle and Norbert Schwarz (eds.). Oxford: Pergamon Press, 235-259.
Hagenaars, Aldi. 1986. The Perception of Poverty. Amerstdam: North-Holland Publishing Co.
39
Hoddinott, John and Haddad, Lawrence. 1992. “Does Female Income Share InfluenceHousehold Expenditure Patterns?” University of Oxford and the International FoodPolicy Research Institute. Xerox.
Marshall, Katherine. 1993. “Employed parents and the division of housework.” Perspectives onLabour and Income. Statistics Canada. Cat. No. 75-001E., 5:3, 23-30.
Osberg, Lars and Shelley Phipps. 1993. “Labour-Supply with Quantity Constraints: Estimatesfrom a Large Sample of Canadian Workers” Oxford Economic Papers, 45, 269-291.
Phipps, Shelley and Peter Burton. 1996. “What’s Mine is Yours? The Influence of Male andFemale Incomes on Patterns of Household Expenditures” Department of Economics,Dalhousie University. Xerox.
Sen, Amartya. 1993. “Capability and Well-Being” in Martha Nussbaum and Amartya Sen (eds.). The Quality of Life. Oxford: Clarendon Press, 30-53.
Strack, Fritz, Michael Argyle and Norbert Schwarz (eds.). 1991. Subjective Well-Being: AnInterdisciplinary Perspective. Oxford: Pergamon Press.
Veenhoven, Ruut. 1989. “Questions on happiness: classical topics, modern answers, blind spots.” in Subjective Well-Being: An Interdisciplinary Perspective. Strack, Fritz, Michael Argyleand Norbert Schwarz (eds.). Oxford: Pergamon Press, 7-26.
Woolley, Frances and Judith Marshall. 1994. “Measuring Inequality within the Household” TheReview of Income and Wealth. Sseries 40:4, 415-432.