TIME SPENT WORKING: Paid work, housework and the gender difference in psychological distress

A common finding in many countries is that women and men use their time differently and that they engage to different degrees in roles connected to gainful employment and housework (e.g., Gershuny 2000; Eurostat 2004). It is also often found that women experience more psychological distress than men do (e.g., Karasek et al.1987; Frankenhaeuser et al.1989; McDonough and Walters 2001). The present study examines a possible link between these two findings. The study focuses on Sweden, which is a particularly interesting case as many women work relatively long hours in the labour market while, at the same time, the division of housework is still rather traditional (Eurostat 2004). Therefore, a relatively large proportion of Swedish women may have long working hours in both paid and unpaid work. Gender differences in the amount of time spent doing each type of work and in the total time consumed by work are, thus, possible causes of the gender gap in psychological distress in Sweden.

Previous studies have found that the association between paid working time and psychological distress is negative, but curvilinear, whereas the association between housework hours is either positive or non-significant (Bird and Fremont 1991; Glass and Fujimoto 1994; Voydanoff and Donnelly 1999; Gähler and Rudolphi 2004). Regarding the combination of paid work and housework, the role enhancement theory predicts beneficial health effects of dividing one's time between different roles (e.g., Sieber 1974), whereas the role strain theory instead predicts negative outcomes (e.g., Moen 1992). The role context theory states that role conditions modify the relationships between roles and psychological distress (ibid.).

The present study investigates the connection between the gender difference in psychological distress and time spent doing paid work and housework, taking into account interactions between paid working time and time spent on housework as well as the context of the family and worker roles, i.e., working conditions, housework environment and division of labour in the household. The study utilizes data from the nationally representative Swedish Level-of-living Survey, which comprises information on hours spent on paid work and housework as well as family situation, working conditions and psychological distress. The data include information collected directly from the respondents' partners, which facilitates an analysis of the division of labour within the respondents' households. Importantly, the data also include a panel, making it possible to control for previously reported psychological distress.

In the present study, roles are defined as social positions that involve certain obligations and normative behavioural expectations. They give guidance in life as well as circumscribe the individual's behaviour (Thoits 1983; Simon 1997). Naturally, the obligations attached to roles take time and may also function like stressors. Each family or worker role may therefore entail both positive and negative consequences. The same holds for having multiple roles, i.e., being engaged in several social positions at a particular point in life.

Many of the empirical studies of the associations between psychological distress and paid work and housework use US data, but the results of the Swedish studies that exist tend to correspond quite well to the results of the US studies. Looking first at direct associations between time spent on family and worker roles, paid working hours show a negative but curvilinear relationship to psychological distress in Swedish as well as other studies. This curvilinearity means that the association turns positive at long working hours, and thus, paid working hours are associated with decreasing psychological distress up to a point, after which additional hours are instead associated with increasing distress (Bird and Fremont 1991; Glass and Fujimoto 1994; Voydanoff and Donnelly 1999; Gähler and Rudolphi 2004). US studies have found that hours of housework increase distress (Bird and Fremont 1991; Glass and Fujimoto 1994). However, using Swedish data, Gähler and Rudolphi (2004) found that time spent on housework was not related to psychological distress. A US study found that number of hours devoted to childcare was not related to psychological distress (Bird and Fremont 1991). Focusing on housework, Bird (1999) found that hours of housework and the division of housework accounted for more than a fourth of the gender difference in depression among US women and men. A first aim of the present study is to investigate whether the associations between psychological distress and paid working hours as well as housework hours account for some of the gender difference in psychological distress in Sweden.

On average, Swedish women and men spend the same total number of hours on family and worker roles (Statistics Sweden 2003; Eurostat 2004). There are, to my knowledge, only a few studies that explicitly investigate the implications of this total role time. Krantz et al. (2005) found no significant relationship between common symptoms of distress and having a long total time devoted to paid and unpaid work among white-collar workers in Sweden. Yet the risk of experiencing symptoms of distress was higher for women working long hours in both paid work and housework than for other women, indicating that a long total role time is negative only when it is associated with role conflict. Similar results have been reported by Coverman (1989) and Pleck (1985). These results may be explained by the role strain theory, which states that multiple roles increase psychological distress by causing role overload or role conflict (e.g., Moen 1992). On the other hand, the role enhancement theory states that multiple roles decrease psychological distress by providing several sources of, for example, social integration, social capital, self-knowledge and self-realization (e.g., Sieber 1974). Among empirical studies counting the number of roles possessed by an individual, support for the role enhancement theory is strong, making it plausible that there is something inherently positive in the mere quantity of roles (e.g., Thoits 1983 ,1986). A second aim of the present study is to assess which of these contrasting theories best describes the associations between paid working hours, housework hours, total role time and psychological distress in a nationally representative sample of cohabiting/married women and men. A further aim is to investigate whether differences between the genders in the combination of paid work and housework can account for any of the gender difference in psychological distress.

Though women and men generally react very similarly to similar types of roles and role combinations (Thoits 1986; Karasek et al.1987), some studies indicate that, in Sweden as well as in the US, paid work benefits men more than women and combining work and family is more distressing for women than for men (Thoits 1992; Lundberg et al. 1994). Similarly, though men on average work longer hours in paid work than women do, long working hours may increase psychological distress more among women (Glass and Fujimoto 1994; Gähler and Rudolphi 2004; though see Krantz et al.2005). The role context theory poses an explanation for such findings, as it stresses the importance of the conditions under which roles are performed, e.g., how demanding they are or how much social support they provide (e.g., Moen 1989). Swedish women have less control over their paid work and are more likely than men are to have stressful jobs, i.e., jobs with high psychological demands but low level of control (Lundberg et al.1994; le Grand et al.2001; de Smet et al.2005). In addition, most roles included in a US study implied more stress for women than for men (Thoits 1992). In Swedish as well as other empirical studies, conditions in the workplace and the household have been shown to moderate the negative association between distress and parenthood (Gähler and Rudolphi 2004) and paid working time (Glass and Fujimoto 1994), respectively.

The balance model specifically discusses the division of labour in the household as a contextual factor influencing psychological distress. According to this model, multiple roles benefit life control and self-esteem, while the equal division of housework provides a buffer opportunity when demands from paid work are high (Härenstam and Bejerot 2001). Several studies show that Swedish women perform the largest share of the housework in their households, while men perform more of the paid work (e.g., Ahrne and Roman 1997; Nordenmark 2002). In a Swedish study, the division of work between partners was beneficial for both women and men only when both were gainfully employed and shared the housework and financial responsibilities (Härenstam and Bejerot 2001). A final aim of the present study is to investigate whether working conditions and the housework environment modifies the associations between psychological distress and paid working hours and housework hours, respectively. The main emphasis is on psychological demands at work and the division of paid and unpaid work in the household.

Five hypotheses can be derived from the results of previous research:

The data used in the present study come from the Swedish Level-of-Living Survey (LNU), which has been conducted five times starting in 1968, each time with a nationally representative sample of approximately 0.1 percent of the Swedish population aged 18–75. The survey includes a panel with respondents who have participated in two or more of the survey waves. The respondents were interviewed in their homes, the interviews covering many areas of life such as family, social relations, housework, employment and health. In 1991, the original sample was 6,710 of which 5,306 participated, resulting in a response rate of 79.1 percent. In 2000, 6,711 individuals were asked to partake and 5,142 responded, giving a response rate of 76.6 percent. A group of 3,502 main respondents reported living with a partner, and of these partners, 2,737 responded to a postal partner questionnaire (Partner-LNU 2000), giving a response rate of 78.2 percent among the partners. This questionnaire was filled out by the partner respondents themselves and comprised the same main areas as the interviews. The present study analyses a sample of employed main respondents who participated in both 1991 and 2000 and lived with an employed, responding partner in 2000. The sample thus consists only of main respondents, while the partner questionnaire is used to add information to the main respondent data. Non-employed are excluded partly because of the design of the questions, and partly because of the very small amount of homemakers among the non-employed. Of the 2,737 main respondents with responding partners, 2,268 (82.9 percent) participated in both 1991 and 2000. Among these, 1,376 (60.7 percent) were employed and had an employed partner in 2000. After excluding 99 main respondents (7.2 percent) who had missing values on the dependent or any of the independent variables, the sample consists of 1,277 main respondents, 563 females (44.1 percent) and 714 males (55.9 percent).

A lagged dependent variable approach (LDV) is used for the OLS regressions. This means that psychological distress in 1991 is included in all analyses of psychological distress in 2000. This way, differences in psychological distress appearing before the central variables are measured is controlled for. Johnson (2005) and Allison (1990), among others, have argued for the superiority of change score analysis (CS) over LDV when analysing two-wave panel data. With CS, the change in the dependent variable between two time points is regressed on changes in all independent variables. Although CS may be the better choice methodologically here, I have chosen to use LDV because of the lack of some central variables in the 1991 data related to maintenance work and child care. I will, however, report the results from CS and LDV analyses of a restricted sample with information on all variables that exist in both the 1991 and the 2000 data.¹

An index of psychological distress is constructed from questions about the experience and severity of physical and psychological symptoms during the past 12 months. The symptoms used are general tiredness, insomnia, nervous trouble (anxiety, uneasiness, anguish), depression/deep dejection and overexertion. The answer ‘No’ has been coded 0, ‘Yes, mild’ 1 and ‘Yes, severe’ 3. Thus, an index value of 0 indicates low and a value of 15 indicates high psychological distress (details can be found in Table 1). A similar index, albeit not including depression/deep dejection and overexertion, shows good reliability comparable to measures of self-rated general health, aches, diseases and functional abilities (Lundberg and Manderbacka 1996).

Descriptions of all variables used in the analyses can be found in Table 1. Below are further points that need to be made about some of the variables.

Household work is measured by two variables: housework measuring routine, everyday activities traditionally performed by women, and maintenance measuring more infrequent activities traditionally performed by men. The respondents estimated the total number of hours spent on each of three fields of housework activities and one field of maintenance activities an average week in their household, and how many of these hours are spent doing these activities by themselves. Partner respondents estimated their own and their partners' (i.e., the main respondents') hours. Controls show that the indicators have rather high reliability and that there seems not to be any noteworthy systematic bias in the reported hours according to gender.²

Several studies have found that respondents over-report their housework hours with stylized estimates like these, as compared to time diary based estimates, and that women over-report more than men do (for a review, see Kan 2008). However, Scandinavian studies have found only small or even inverted differences, and only small gender differences in the level of over- or under-reporting (Bonke 2005 analysing Danish data; Kitterød and Lyngstad 2005 analysing Norwegian data). One reason for this may be the egalitarian gender attitudes in Scandinavia (Kitterød and Lyngstad 2005). Because Sweden has similar work patterns and gender attitudes as other Scandinavian countries, stylized estimates of hours of household work should be rather reliable. The validity of the estimates should be improved by the fact that respondents estimated the average time spent on four specified fields of household work rather than the total time spent on any type of household work.

The variable childcare occasions is based on questions posed only to individuals living with children aged 0–12 (see Table 1 for details). Therefore, all models control for the presence of children of this age in the household.

For share of housework and share of paid work (respondent's hours divided by respondent's plus partner's hours), information about the partner's hours is reported by the partners themselves. The relations between indicators of share of housework, share of paid work and share of childcare occasions on the one hand, and psychological distress on the other, are linear. Other functional forms have been tested, but as the results were the same, the simpler form was chosen.

Psychological distress 1991 (see Table 1) is included as a control variable in all models together with age, which is a continuous variable.

As an introduction, Table 1 displays clear gender differences in weekly time spent on housework, maintenance work and paid work. On average, women spend eight and a half hours more on housework than men do, which is more than twice as much, while men spend six and a half hours more on paid work and two and a half hours more on maintenance. Hence, the average total amount of time spent on these activities is the same for women and men, about 54 hours/week. However, women's total time spent on paid work, housework and maintenance is significantly longer than men's among full-time working respondents (57.5 for women as compared to 55.4 for men) and especially among full-time working parents of 0–12 year olds (59.2 for women as compared to 56.4 for men; these figures are not shown in Table 1). In addition to this, mothers perform childcare activities on five more occasions per week than do fathers. Therefore, it is plausible that the actual time that women spend on the family and worker roles is longer than that spent by men, something that has been indicated in studies including hours of childcare in measures of unpaid work (Lundberg et al.1994; Krantz et al.2005).

There is a significant average gender difference in psychological distress of 0.48 index points, which corresponds to about a quarter of a standard deviation. Albeit not large, this difference is not negligible. There are also gender differences in all of the items that constitute this index.³

The results of a series of OLS regressions are seen in Table 2. These analyses show changes between models in the association between gender and psychological distress controlling for family roles, worker roles and/or role conditions.

As seen in Table 2, Model 1, age and previous psychological distress account for part of the gender difference, which has decreased from 0.48 to 0.37 but is still highly significant. Model 2 adds time spent on housework and maintenance and childcare occasions, which account for nearly 40 percent of the remaining gender difference, which, furthermore, is no longer statistically significant. Additional analyses (not shown) reveal that it is above all the inclusion of hours of housework in the model that causes the reduction. Model 3 shows that hours of paid work do not reduce the gender gap, but rather increases it. The gender difference is reduced again in the additive Model 4 but does not change further in Model 5, which introduces an interaction between hours of housework and hours of paid work. Model 6 adds the division of housework and paid work between partners and psychological demands at work. Although all three variables contribute to the reduction of the gender difference in this model, psychological demands contribute the most (analysis not shown).⁴ Compared to Model 1, the gender difference is reduced by 45 percent and rendered insignificant when hours of housework and paid work, an interaction between these types of work and, finally, role conditions are taken into account.

Hypothesis I stated that part of the gender difference in psychological distress is accounted for by men's longer hours of paid work and women's longer hours of housework. As hours of housework but not hours of paid work account for part of the gender difference, Hypothesis I is partly supported.

Table 3 shows the results of the separate analyses of women and men. Model 1 for women suggests that neither time spent on the family roles nor time spent on the worker role is significantly related to women's psychological distress,⁵ but a different pattern appears in Model 2 where the interaction between hours of housework and hours of paid work is included.⁶ While additional hours of housework or paid work are associated with less distress, spending many hours on one of these roles appears to decrease the possible beneficial effect of the other role (shown by the positive interaction term). The statement in Hypothesis I that there is a negative, but curvilinear, association between paid work and psychological distress is thus partly supported for women. The statement that there is a positive association between hours of housework and psychological distress is not supported. Instead, psychological distress decreases with additional hours of housework, though less so for longer hours, as shown by the positive quadratic term. As dividing one's time between paid work and housework appears to be less beneficial when a great deal of time is spent on one or both roles, Hypothesis II, predicting such a positive interaction between the role variables, is supported for women, whereas there is no support for Hypothesis III, which stated that there should be a negative interaction.

Table 3, Model 3 adds share of housework, share of paid work and psychological demands at work, and shows that only psychological demands at work are significantly, and positively, associated with women's psychological distress. Neither share of housework nor share of paid work is related to women's psychological distress, and this is true also when hours of housework and paid work are excluded from the model (not shown in Table 3). There is no significant association between share of childcare occasions and psychological distress among women living with 0–12 year olds (additional analyses not shown in Table 3). Hypothesis V, stating that psychological distress would be high among people performing an unequal share of housework, paid work or childcare, is hence not supported for women.

Table 3 exhibits a pattern among men that is quite different from that among women. None of the models shows any significant relationships between hours of housework or paid work and psychological distress, and there is only a tendency for lower psychological distress with additional hours of maintenance.

Men who live with 0–12 year olds are more psychologically distressed than other men. As childcare occasions and time spent on housework and paid work are controlled for, this association is not caused by the practical childcare activities that these men attend to (at least not those analysed in the present study) or the additional time spent on housework or paid work that may be the consequence of having children. The association is thus likely to be caused by aspects of living with children or being a father that is not taken into account in the present analysis.

As seen in Model 3 for men, having a demanding job is associated with higher psychological distress, but there are no significant associations between men's share of housework or paid work and psychological distress. This holds even when hours of housework and paid work are excluded from the model (additional analysis not shown). Share of childcare occasions is not significantly associated with psychological distress among men living with 0–12 year olds (analysis not shown). Thus, among men, there is no support for any of the hypotheses regarding the relationships between family and worker roles and psychological distress.

To test Hypothesis IV, that role conditions modify the relationships between psychological distress and paid work and housework, respectively, a number of interactions between role conditions and hours of housework and paid work has been tested for both women and men, but none proved to be significant (e.g., between paid working hours and psychological demands at work, control at work and inconvenient working hours, respectively, and between hours of housework and housing problems such as noise from neighbours and traffic, poor building maintenance and poor indoor air quality).

It is possible that people may work more in the labour market and household and be less distressed the more physically healthy they are (a kind of healthy worker effect). However, including physical health (self-reported circulatory problems and pain) does not result in any noteworthy changes of the associations observed for either women or men. Economic factors (hourly wage and disposable household income) have also been added to the models without altering the results (analyses not shown).

The explained variance is small in all models for both women and men. This could mean that time spent on paid work and housework is not very important to psychological distress. However, the explained variance is generally small in studies of psychological distress, probably owing to the complexity of this phenomenon (for Swedish studies, see Gähler and Rudolphi 2004; Nordenmark 2004). Although time spent on paid work and housework may not account for a great deal of the variation in psychological distress in the working Swedish population, it does, however, account for a substantial part of the gender difference in distress.

To test whether CS analysis renders similar results as the LDV analysis used here, I have analysed a smaller sample with information on all variables from both 1991 and 2000, except maintenance and child care occasions. With this sample, and excluding maintenance and child care occasions, CS gives the same results as LDV among men. Among women, however, there are some differences. First, there is no association between paid working hours and distress with CS, whereas there is a weak negative association with LDV when the interaction between hours of housework and paid work is included (similar to the analyses above). This is not surprising, as CS tends to give less significant estimates than LDV. However, this may mean that the association found in this study is caused by selection of less distressed individuals into long paid working hours. Second, the CS analysis indicates that distress increases with increasing housework hours among women, but mainly among those with already fairly long housework hours. The LDV analysis shows no significant association between housework hours and distress in this sample. Hence, in the above analyses, the possibly detrimental aspect of housework hours should perhaps be stressed, rather than the favourable aspect seen at shorter hours.

Figure 1 shows the estimated level of psychological distress as the hours of housework increases for women who have different hours of paid work (estimates from Table 3, Model 2 for women). To give a sense of the magnitude of the displayed differences in psychological distress, the maximum difference, between women who do paid work for 20 hours/week and 45 hours/week, respectively, and housework for 30 hours/week, is somewhat smaller than a standard deviation on psychological distress for women. Figure 1 shows that among women who do approximately as much housework as the average woman (15 hours/week), there is no estimated difference in distress between women with different paid working hours. Among women who do a small amount of housework, those with longer paid working hours are less distressed than those with shorter paid working hours are, whereas among women who do more housework than the average woman, those with shorter paid working hours are less distressed than others. For all women, there is a point at which the association between psychological distress and hours of housework goes from negative to positive, and the position of this inflection point depends mainly on the total role time, i.e., the time devoted to paid work plus housework. If the total role time exceeds about 50–51 hours/week (somewhat less than the average of 53 hours/week for women), more housework results in increased psychological distress.

Regression analyses (not shown) of psychological distress and the total role time, measured as weekly hours of housework plus weekly hours of paid work,⁷ show that with control for age, psychological distress in 1991, the presence of 0–12 year olds, hours of maintenance and childcare occasions, the estimated inflection point is at 52 hours/week among women. The number of women who have a total role time that exceeds this critical point is not negligible. This applies to 50 percent of all women in the sample and to 70 percent of the women who are employed full-time (40 hours/week or more). Among men, there is no significant association between total role time and psychological distress.

An analysis of the pooled sample of women and men shows no reduction in the gender difference in distress when including total role time (not shown). Hence, total role time has a negative but curvilinear association with psychological distress among women, but Hypothesis II, stating that total role time contributes to the gender difference in psychological distress, is not supported.

Table 3 displayed that women's psychological distress is not related to the shares of housework and paid work they perform. However, inequality in the division of housework and paid work may be an underlying cause of the long total role time of some women, which does indeed relate to their level of distress. Figure 2 shows the distribution of the total role time between paid work and housework among respondents with a long total role time and among their live-in partners. The left part of Figure 2 shows that among respondents with a total role time of 51–60 hours/week, the differences between their and their partner's total role time are not large (compare adjacent bars). In the group of respondents with a total role time of more than 60 hours/week (the right part of Figure 2), the differences between women and men who live together are larger. Female respondents in this category work as many hours in the labour market as their partners do, but they perform nearly four times as many hours of housework (compare the fifth and sixth bar). Hence, if men were to perform half instead of a fifth of the housework in these households, women and men would have the same total role time. It looks like the long total role time of these women is indeed caused by inequality in the division of housework. Among men with a total role time of more than 60 hours/week, the main reason for their high workload is instead their paid work, which amounts to nearly 58 hours/week on average. While these men's partners perform the majority of the housework, they work less than standard full-time in the labour market, which contributes to the difference in total role time in these couples (compare the seventh and eighth bar).

The analyses presented here have shown that hours of paid work, and possibly also hours of housework, are associated with low psychological distress among women, but women's psychological distress increases when they spend a long time on housework or a long time on housework plus paid work. Hence, hours of housework, paid working time and total role time are associated with women's psychological distress. However, only hours of housework contribute to the gender difference in distress in Sweden. Women spend twice as many hours/week on housework compared to men, and this accounts for 40 percent of the gender difference in psychological distress.

Although spending additional hours on paid work or housework may be beneficial among women, the association with either type of work is dampened and even reversed when time spent on the other type of work increases, and this is mainly caused by the resulting increase in total role time. Among full-time workers, and especially full-time working parents, women have a longer total role time than men do, and the total role time of many working women in Sweden is of such length that it may cause psychological distress. This lends support to the role strain theory, which predicts that combining paid work and housework is detrimental as it increases role overload and role conflict, rather than to the role enhancement theory, predicting beneficial effects of dividing ones time between different roles. However, according to the results of the present study, the negative aspects of the combination of paid work and housework have more to do with the resulting total role time than with the fact that different types of work are combined.

Division of housework is likely to be one cause of the long total role time experienced by some women. In households where the woman has a total role time of more than 60 hours/week, she usually works more than full-time in the labour market while the housework is very unevenly divided between her and her partner, leaving the woman with half of the paid work and nearly four-fifths of the housework performed in her household.

According to the role context theory and the balance model, the conditions under which roles are performed modify the associations between roles and psychological distress. These theories are not supported by the results of the present study, as there are no significant interactions between role conditions and time spent on the roles. However, the psychological demands placed on women in their paid work contribute directly to women's high levels of psychological distress compared to men's.

Sweden is known for its egalitarian policies and gender relations, but the still quite traditional division of housework contributes to the gender difference in psychological distress and appears to result in a long total role time for some women. Policymakers in Sweden, as well as in other countries, have tried to solve women's problems of reconciling paid work and family by introducing tax reductions for household services purchased in the market. However, the present results indicate that these policymakers fail to see the bigger picture. Women with a long total role time do not do a large share of the housework in their household because their partner has less time to do it. On the contrary, they do a large share despite the fact that they work as many hours in paid employment as their partner does. Therefore, tax reductions for household services may not help women or even households cope with the dual-earner arrangement. Instead, they encourage men to continue to avoid doing their share of the unpaid work in their home. To efficiently tackle the problem of women's high workload, policymakers need to deal with men's unequal contribution to housework, and this means seriously considering gender norms, gender inequality and power relations within relationships.

Katarina Boye is a researcher in sociology at the Swedish Institute for Social Research (SOFI) at Stockholm University, Sweden. Her research focuses on gender inequality in the family and in the labour market, particularly the causes and consequences of gender differences in time spent on paid work and housework and the variation across countries in these factors.