Education and the Division of Household Labor in Dual-Earner Families

This article is the first report of a study on policies and the division of paid and unpaid work in families in Finland. The article examines the division ofhousehold labor and its determinants in Finnish dual-eamer families. The main objective is to examine whether education has any impact on the division ofunpaid work and men's participation in it controlling för other variables. It was found, that among women, rising educational levels, non-traditional attitudes and younger age cohort had a negative impact on time spent on housework, while among men only reduced time in employment and nontraditional attitudes increased the contribution at home. While both men and women with higher education and non-traditional attitudes were more likely to perceive their relative division as more equal, an analysis of the absolute number of hours spent on housework seems to support the notion that more equal distribution of tasks at home is more or less a result of younger and educated women doing less housework. The data comes from a survey conducted in 1998, in which 2,500 Finnish men and women were questioned about time use, employment, attitudes about gender roles, work and family, andreconciliation ofwork and family. The Finnish study is part ofa Europeanresearch project which studies the division oflabor in families in different cultural, political and societal settings.


Introduction
In Finland women form almost half ofthe labor force, and the labor force participation rate among married women is among the highest in Europe.Although the double bread-winnermodel is well establishedin families,and women's incomeis increasinglyneeded to maintain the living standards of the families, women still continue to perform the majority of the household and child-care tasks while men provide a helping hand.Compared with past decades, women are spending less time on household tasks while men have increased their contribution somewhat, but changes in the distribution of unpaid labor within families have been small.Increasing participation in paid labor and small changes in the distribution of unpaid work have meant that especially women's total work load has increased.Dual-earner families, especially those with children, feel questions concerning the distribution of housework most acutely.The reconciliation of work and family affects both men and women in these families.Changes in working life, such as flexible working hours, pressures to work overtime, part-time working and temporary employment contracts affect increasinglythe organization of daily life.The division ofhousehold tasks is also an important source of distress in unions.It is found to affect marital satisfaction (Pifia & Bengtson 1993) and cause conflicts and disagreements between partners (Reuna 1998;Blair 1993;Kluwer et al. 1996).
It seems that Finnish men spend a little more time on household tasks than menin Central or Southern European countries (Jacobs 2000;Miettinen 1997).Due to different measurement techniques the differences in time used för housework among men are not definitive.What is clear, though, is that in all countries women tend to contribute 2-3 times the amount men do.During the past decades, women's labor force participation has increased steadily in all countries and women are more educated than before, and often better educated than men.Family structure is changing: the average number of children in families has declined and the number ofpeople living alone is increasing.It is often expected that younger generations with more education and, perhaps, more individualistic life styles, are able to negotiate family roles more freely.The general ideological climate supports working women, and equality, not only in workplaces but also in the family.Why, then, in spite of these substantial changes affectingfamilylife,have familialdivisionof work andmen's participationin household chores been so resistant to change?When time use data from Finnish families from 1979 and from 1987/88 were compared, the division ofhousework between men and women had become more equal in all types of families during the intervening ten years.Men participated more in all families and women's contribution had decreased in families with 0-2 children.In spite of the change, men continued to do less than two-thirds ofthe work women did even in families without children in which the distribution was the most equal (Niemi & Pääkkönen 1989).Similar trends can be observed from studies using relative measures (Haavio-Mannila et al. 1984;Melkas 1998).Yet,relatively little is known ofthe factors which contributeto the division ofhousework at the family level in Finland.There may exist important differences in the values conceming family and work, and in the lifestyle preferences between social classes, or persons from different educational backgrounds, which affect, för example, men's participation in household chores.The objective of this article is to examine the division of housework and its determinants in Finnish dual-eamer families using recent data, and more specifically, to test the hypothesis that an increased level of education will positively affect the sharing of household duties.

Theoretical background
In general, there are three theoretical approaches which are commonly used to explain the division oflabor in families and men's participation in unpaid work.These include relative resources, time availability and gender role theories.In the relative resources approach the föcus ison the power relation between partners (relative power in household decision making, as first described by Blood & Wolfe 1960).Household work is seen as something unpleasant which both partners try to avoid.Both parties are using their relative resources to bargain a reduced share of chores.Eamings, occupational status, educational status, or relative measures of these variables, are used as indicators of power differentials within a relationship.
New home economics theory resembles the relative resources approach in that it also föcuses on the relative resources each partner brings into the relationship.Resources are not considered as expressions of relative power to be used in a bargaining process, but as indicators of comparative advantage in the market orin the household.According to the new home economics approach, in order to maximize household utility function partners will allocate their time between paid and unpaid labor according to the marginal utility derived from employment and housework (Becker 1993).A partner with a comparative advantage in the market will specialize or allocate more time in the market, and a partner with a comparative advantage in the household will specialize in household work.Relative eamings are often used as an indicator of differences in the comparative advantage in the market.Thus, the theory assumes, the smaller the relative advantage (or eamings gap between partners) a man has in the market, the more time his wife will allocate to the market and the more time he will use för housework.
Studies of the impact of eamings have produced relati vely consistent results: the smaller the gap between partners' eamings, the more equal is the division of housework (för example, Farkas 1976;Kamo 1988;Blair & Lichter 1991;Presser 1994).Other studies using occupational status or educational level to measure relative power have been less consistent in their results, which can partly be due to different operationalizations of the variables in question (see Shelton & John 1996, för the overview of studies using various measures of relative power ).
The time availability approach, on the other hand, föcuses on the supply and demand sides ofhousework (demand/response capability, by Coverman 1985, or work status, by Kamo 1988).In every household there is a certain amount ofhousework that needs to be done, and distribution of the tasks among household members is determined according to who has the most available time to do the work.Employment reduces possibilities to contribute to household work, and creates pressures för others to 'fill in the gap'.On the supply side, differences in the absolute hours in employment, or employment status, as well as in the flexible working hours between the couple, are used to explain differences in the division of labor.On the demand side, number of children is generally used to indicate pressure to increase time spent on household work.Most studies find that employment decreases both men's and women's contribution to household labor,but the relationship between women's employment and men's participation in housework is not clear (för example, Ross 1987;Kamo 1988;Blair & Lichter 1991;also Kunzler 1996,för the overview of studies which have used employment or hours in employment as an independent variable).Availabletime has also been operationalized as partners' overlapping hours in employment to better capture 'real' available time.In fact, in her study Presser (1994) föund that the number of hours husbands were not employed while their partners were working did predict husbands' participation in "feminine" tasks.
The third approach, or gender role theory, explains the division oflabor by differences in gender role attitudes.The theory assumes that men and women are socialized early in childhood or during adolescent years to appropriate gender behavior and thus later in their life express gendered personalities and preferences.According to theory, a person with more egalitarian, or liberal, non-traditional attitudes about gender roles will divide household tasks more equally (Stafford et al. 1977).Although a number of studies have föund support för this hypothesis, the results are not always consistent,probably caused partly by variation in measures of sex-role orientation or normative attitudes about gender behavior (för example, Kamo 1988;Brayfield 1992;Presser 1994; see also Shelton &John 1996, andKiinzler 1996, för the overview of studies using gender role attitudes to explain division of housework).From 1990s on, a gender constructionist perspective has gained more attention.It emerged as a critique of rational choice theories on household labor (Berk 1985;Ferree 1990 and1991).It tries to understand household workings from a gender construction perspective, where gender, like other products, is produced.Perförming certain household tasks provides an opportunity to behave according to as appropriate sex category, or 'do gender', to prove that one is a member of a sex category.The gender constructionistperspective has been used mainly in qualitative studies on household labor,and so far, only a few studies which have tried to explain the division of household tasks or men's time spent on housework have included it in the analysis (för example, South & Spitze 1994; Twiggs et al. 1999).The need to move beyond rationalistic explanations of the allocation ofhousehold labor and to better understand its social and cultural embeddedness are evident in recent research into the division of household labor and how it is linked to, för example, culturally different understandings of care work, kin relations, life course and family interaction factors (see Coltrane 2000, för an extensive overview ofresearch into household labor during the l 990s ).
Depending on the theoretical point of view, educational level has been used to measure relative power in a relationship, or differences in subcultural attitudes and life-style preferences.Its effect on housework behavior has thus been sometimes difficult to interpret.Many studies have föund a positive association between men's educational level and their participation in household work (Farkas 1976;Rexroat & Shehan 1987;Ross 1987;Kamo 1988;Brines 1994;Presser 1994;South & Spitze 1994), and a negative association between education and time used on household tasks among women (Rexroat & Shehan 1987;Bergen 1991;Brines 1994;South & Spitze 1994).In another study, the effect of education was föund to be negative also among men (Coverman 1985).In fact, in many cases, highly educated women reduced their contribution, thus creating a more equal distribution of household tasks, while highly educated menincreased their contribution only a little or not at all.Yet, other studies have shown that education has no relation to participation in household tasks among women (Ross 1987), or neither among men nor women (Barnett & Baruch 1987).Farkas ( 1976) and Presser (1994) have also studied the effect ofrelative education on the husbands' participation in housework.Contrary to the relative resources hypothesis, relative educational level did not have any significant effect on the division of housework in either of these studies.
It is possible that education captures life-style preferences or values which are not measured by common sex role attitude factors.The importance assigned to family, or conversely, to paid work, för example, are often föund to vary by education or social class.These may affect the meanings given to housework, and thus the willingness to participate in it.In Ross (1987) and Pittman & Blanchard (1996) it was föund that education played a role in the household division oflabor irrespective of sex-role attitudes.Twiggs et al. (1999) have argued that it may be more införmative to look at household chores separately, since participation in them and the characteristics of those who participate in them may be different depending on how sex-typed a task is, similar to segregation in the labor market.They föund that normative, or non-rational factors, such as attitudes and education, determined husbands' participation in different household chores more than practical factors.Educational level was significant only in washing the dishes, which was the least sex-segregated task.This may indicate that education affects which men are more likely to cross the boundary to participate at all.
The present study will investigate the division ofhousework and its determinants among Finnish dual-earner couples.First, on the hasis ofrelative resources and time availability approaches, the study will examine to what extent selected rational or pragmatic factors are able to explain the variation in the participation in household labor in the Finnish data.The second aim ofthis article is to find whether education hasan independent effect on the division of work when other variables, including factors measuring gender role attitudes, are controlled för.It is hypothesized that educational level will affectboth men's and women's participation, with a higher educationallevel implying a more equal division.Since a more equal distribution of housework between partners can be achieved by men increasing their contribution or women decreasing theirs, or both, the models will be tested för both sexes separately.Another likely hypothesis is that the more equal distribution of tasks is solely a result of more educated women devoting less time to housework.
The third question to be addressed in this article is more related to methodological issues.Research on household labor has used both absolute time use data and relative measures.Contradictory results in the factors affecting the division may be partly explained by different measurement techniques of the dependent variable.Since in the survey the division ofhousework was inquired both with absolutetime expenditure and direct questions concerning the relative division of tasks, it is possible to analyze these separately.Thus, the article examine whether the results from the analysis of absolute time use will be similar to those from the analysis of relative division.If the findings are inconsistent, more attention needs to be directed at the measurement technique and its effect on the results.

Data and method
The analysis is based on data collected by a postal survey conducted in 1998.In the survey, 2,500 persons aged 25 to 50 and living in households of two or more people were asked to report their time use, relative distribution of household tasks as well as their attitudes and opinions about gender roles, values concerning family and work, and problems related to the reconciliation of work and family.The questionnaire included questions about the employment and educational background ofboth the respondent and his/her partner.The overall response rate was 57 percent.Generally,women, older persons, married people, and those with a higher vocational education were somewhat over-represented in the data.In this analysis, a subsample of married or cohabiting persons living in dual-earner couples (N=993) is used.Table 1presents a summary of some background variables and variables used in the analysis from this subsample."Dual-earner" is here understood to mean that both partners are currently working.Generally, women with a child less than one year of age were excluded from the subsample, since in most cases the mother was still on maternity/parental leave and thus currently not working.Due to error in the questionnaire, there were more missing data in the question on education, and consequently,the size of the subsample is lower there.Participation in housework will be studied by using analysis of covariance, ANCOVA (of SPSS 10.0).It examines mean scores and their deviation from the overall mean of groups, adjusted för independent variables.Analysis of covariance, and related multiple classification analysis, will provide tests of significance, and predicted and adjusted means and deviations from the overall mean in absolute time use and relative distribution of tasks för categories of independent factors, as well as coefficients and tests of significance för covariates.

Dependent variable
The dependent variable, division af household labor, is measured by using absolute time expenditure in certain tasks, and by using self-assessments ofrelative division of labor between partners.For the purpose of this article, 1focus on the time spent on and the division of five typical household task, namely preparing meals, laying the table and washing dishes, doing laundry and ironing, cleaning the house, and shopping.These tasks are often described as feminine since in most cases women still perform the majority of them.As they are also time intensive and time restrictive in the sense that they have to be done daily or regularly, changes in the distribution of these tasks have apparent effects on the lives of family members.It is also clearthat a marked change in the allocation ofthese traditionally feminine tasks implies changes in the roles of men and women in the family.
Respondents were asked to report their time use, in hours and minutes, in a number of activities during a typical working day and on weekends.A weekly time expenditure in the five household tasks was calculated by summing up the hours spent on a working day (multiplied by 5) and during the weekend.Daily time use tumed out to be a rather demanding question and resulted in some missing data, especially in reporting the time use on weekends.Men seemed to suffer from this underreporting more than women, and estimates of hours spent in household tasks may thus be slightly too low among men.In the preliminary analysis, it was found that a clearly larger proportion among the lowest educated men had reported zero hours för the time spent on all nine household tasks specified in the questionnaire.When these were treated as missing data, and relevant cases excluded from the analysis, the results did not change markedly.Comparison of self-reported estimates from surveys and time use diaries has shown, on the other hand, that the reported time spent on housework is higher in the surveys than in time diary data, especially with more undefined or abstract tasks or tasks occurring frequently (Niemi 1993;Marini & Shelton 1993).Naturally, the problems related to survey data on time use are present in this study as well.
Division oflabor is also examined by using a relative measure.Respondents were asked to indicate the relative share of each family member and other persons in completing certain tasks.Again, 1will focus on the division oflabor between men and women in the five aforementioned tasks.Respondents' relative share is calculated as a ratio of her/his own share to the combined share.Although it was possible to indicate whether other family members or other persons participated in the tasks, in most cases the tasks were done totally or mostly by partners themselves.Children participated on average only a little.Other persons besides family members participated in these five tasks only very seldom or not at all.Since this is a subjective estimation of the relative division of labor between partners, it is not affected by the actual hours families spend doing these tasks.Thus, differences between families in their preferences and skills in doing household tasks, and ability to buy services outside the family, are controlled för.In addition, the relative measure provides information on the division of work between 'real partners'.A survey with individuals as respondents only provides absolute time use data of this individual, not of his/her partner.Proportional time spent on household activities can then be measured only by comparing mean hours men spent on a certain task to mean hours ofwomen.
A relative measure of participation in housework will provide information about how partners perceive this division.Subjective estimates of relative division are perhaps more prone to overestimation of one's own share than seemingly neutral estimates of absolute time use.Assessments of the reliability ofrelative measures are scarce.In a study using both a time diary and relative measures, a correlation of 0.34 to 0.37 between a person's self-estimates and data from the time diary was found (Layte 1999).Among working couples, the correlation was 0.45.The same study reported a correlation of 0.61 (0.77 among working couples) between female and male partners' estimates.When the relative division of tasks between partners was compared between female and male respondents in our data, it was found that both groups tended to exaggerate their own share at the expense of the other party.Differences between men's and women's accounts varied around 5 percentage points.There is no reason to expect that this overestimation varies systematicall y between diff erent groups, and respondents' own accounts of their share are used to measure the division oflabor in the household.

Independent variables
Educational level Vocational education.The respondent's and his/her partner's educational levels were classified into three categories.No vocational education, vocational school, or vocational courses were combined into one category.The other two categories were vocational institute and university degree.In general, the three categories refer also to the length of vocational education.Vocational school lasts on average 2-3 years, vocational institutes about 4 years, and university studies about 5-7 years, or even longer.To test the hypothesis that relative education hasan effect upon the division of housework, a variable with three categories (lower, the same, higher educational level than the partner 's) was created.Also, a four-category variable was created (neither, either or both partners having at least vocational institute level education) in order to examine whether the effect ofrespondents' education would differ depending on their partner's education.

Other independent variables
Earnings.Relative earnings are measured as the ratio of a respondent's earnings (measured in Finnish marks/month) to family earnings (both partners' earnings combined).It is expected, that with higher relative resources, e.g. higher relative earnings, the partner in question will be able to negotiate a lower share of household tasks.
Number af children under 10years af age.Number of small children in the family is used to measure the demand factor of housework.According to the time availability approach, increasing the number of small children will increase the participation in household duties.
Weeklyworking time.Respondents' weekly working time measures the supply factor.
Weekly working time is measured in minutes.It is expected that increased working time will decrease possibilities to participate at home.

Age group.
Although measured by year ofbirth, this variable will be used as a categorical variable in the models.In the preliminary descriptive analysis it was found that age group, or year of birth, had a curvilinear effect upon housework.Age group will measure cohort eff ect, that is, the assumption that younger cohorts devote less time to housework than older cohorts.

Attitude factors
Respondents' gender role ideologies were measured with several attitude questions which were then factor analyzed.Two gender role factors, which in the preliminary analysis turned out to have significant effects upon the division ofhousehold labor, will be used in the analysis.
Womans role equalsfamily role.This orientation was created from five items ("Very career-oriented men should not have a working wife", "Paid work means more to a man than to a woman", "Raising children is, in general, more rewarding to a woman than having a career", "Women are not as committed to their work as men", "Women are less suited to the competition in working life than men"), with the eigenvalue 2.05, and Cronbach's alpha .64.
Men should increase their participation at home.This was created from two items ("Men should take a daily active role in all aspects of their children's upbringing", "Men should do half the housework"), with the eigenvalue 1.98 and Cronbach's alpha .48.It is expected that person with more traditional attitudes (agreeing with 'woman's role equals family role', or disagreeing with 'men should increase their participation at home') will also behave along traditional lines, e.g.men devote less time to housework and womenmore.

Division of work between men and women
As expected, women spent more time on household tasks than men.Since the difference between average hours in employment was small (48.5 hours/week among men, 44.5 hours/week among women) women's total weekly workload (when paid work and all housework, except child care, are combined) exceeded that of men by almost eight hours.When the time used för the five household tasks is compared between men and women (Table 2), women spent 2.5 times as many hours on the five household tasks as men did (20 .4hours versus 8.2 hours ).The most time-consuming task among all women was meal preparation, while men spent the most time on meal preparation and shopping.Men with higher vocational education tended to spend a little more time on housework than men with no vocational education or with education from a vocational school.There seemed to be an increase in the time spent on every task, suggesting that if more educated men increase their contribution to housework, they are likely to do all kinds of tasks.Education has an opposite effect on women: those with a higher education tended to spend less time on housework.The decrease was most visible in meal preparation.The gendered division ofhousework is reproduced in the accounts of the relative division of tasks (Table 3).Both men and women with a higher education report a more equal distribution of tasks.Again, men with higher education tended to participate more in all tasks.In the most gender-specific task, doing the laundry, more than 30 percent of men with lower education did not participate at all, while among men with a university level education, the proportion of persons who did not participate at all was about 10 percent.Education had a similar effect in meal preparation, though not as clear.In the other tasks, the proportions of those who did not participate at all varied only a little between education groups (table not presented here).The two measures correlated significantly among both men and women (r=.418**, r=.249**), e.g.persons who reported more time spent in absolute hours on household tasks also indicated a greater relative share för themselves.Correlation was higher among men than women.It is possible that women tend to overestimate their share in these feminine tasks and thus the association between actual time used för these tasks and the reportedrelative share is lower among them.According to Niemi (1984), social desirability, or the idea that certain tasks belong to women, causes women with more traditional attitudes towards gender roles to overestimate their role in completing these chores.This social desirability didnot seem to cause underestimationamongmenin her study.

Analysis
Tables 4 and 5 show the results of analysis of variance för two dependent variables.In Table 4 the dependent variable is the absolute time spent on five household tasks, in Table 5, it is the relative division of the tasks.The tables present only one model, separately för men and women.In addition,two other models with variables measuring relative educational level were tested.More descriptive tables of the mean time spent on housework and relative division by chosen variables are included in the appendix in Tables 1 and 2.
In Table 4 (page 54) it can be seen that only two of the variables chosen för the analysis have a significant effect upon the variation in the number of hours men devote to housework, but that they are clearly better predictors of women's contribution.Weekly working time has a predicted relationship with the time used för household tasks among men, but not among women: when employed time increases, men allocate less time to housework (the coefficients show the average decrease, or increase, in minutes, in the time used för housework when the covariate, here weekly working time, increases by one unit).The other variable measuring situational factors, or time availability/ demands för housework, the number of children under 10 years of age is not a significant determinant of men's time use on housework.However, it clearly affects women's time use, thus giving support to the notion that children increase especially women's workload in the family (every additional child increases women's housework by approximately 99 minutes).Contrary to the relative resources theory, relative earnings do not affect the number of hours spent on housework.
Of the two variables describing attitudes towards gender roles, the attitude towards men's role in a family has a significant effect upon housework among men.Men who have more traditional attitudes and who believe that men's role in the family should not be increased are also less likely to devote more time to housework.In accordance, women who feel that women's role does not only equal family role, used less time on household tasks.Only the attitude factor measuring normative expectations towards the respondents' own sex was a significant determinant of the number ofhours devoted to housework.Thus, the participation of men in housework is clearly associated with men's ideas of their role in the family.Even if Finns seem to support equality and are fairly modern in their sex-role attitudes (see för example Ervasti 2000 andMelkas 1998), attitudes towards the role of the opposite sex are not reflected in their behavior, at least when it comes to the division ofroutine household tasks.Since this is a crosssectional study, it is not possible to determine whether behavior föllows the attitude or the reverse.The cohort effect, or the assumption that younger cohorts spent less time doing housework than older cohorts was significant among women, but not among men.In fact, the highest number of hours used för housework was föund among the second oldest (among women), orin the middle age groups (among men).
The level of vocational education failed to significantly affect housework among men, when other variables were controlled för.Although higher educational levels seemed to be positively associated with time used för housework (men with a vocational school education spent an average of 7.6 hours per week on housework while men with a university degree spent an average of 8.8 hours, when the effects of other factors were controlled för), the variation within educational groups was greaterthan between the Method: ANCOVA, experimental with covariates entered first groups.On the other hand, education did affect women's contribution in the predicted way.Womenwith a higher vocational education spent significantly less time on housework than women with a lower education.No significant interaction effect between factors was found in the data.This means that the effect of education is similar in all age groups, and that it plays a role among women also in the younger cohorts in determining the amount of time they spend on housework.
When we turn to relative division ofhousework, we get a somewhat differentpicture of the factors affecting the division (Table 5, page 56).Of the covariates, only attitude factors have a significant effect on the relative division among both men and women.The number of children, relative eamings, or weekly working time do not affect the way respondents perceive their division oflabor.The cohort, or age group factor, does not have a significant effect upon the relative division either.
Attitude factors are very important in explaining the relative division.Among women, both attitude factors had a predicted effect upon the division.Women with less traditional attitudes about women's role did a smaller share ofthe housework, and women with more traditional attitudes about men's role in the family increased their share.Sincethe negotiation over the division oflabor at home involves both partners, it is also possible that women who hold more traditional attitudes about men's role in the family and in child care may not be willing to increase men' s role in certaintasks.Among men, only the factor measuring attitudes towards men's role had a significant and predicted effect (more traditional men had a reduced share).
Even with the inclusion of attitude factors, vocational education still has predictive power on the relative division both among men and women.Among both sexes, increasing education implied a more equal distribution of tasks.On the hasis of the analysis of absolute time use, the more equal distribution of housework in the households with educated women seems to be the result of these women doing less housework.In a Finnish study on gender role attitudes (Ervasti 2000) it was found that perceptions of inequality between the sexes in the family and in working life diminished with education, e.g.men and women with lower education were more conscious about gender inequalities especially in the family than persons with higher education, thus giving support to the findings from this study.
Normative factors appeared to be better predictors of the perceived division ofhousework than pragmatic or rational factors (working time, number of children, or eamings).Although attitude factors and education did have a significant effect on absolute time use, and thus may contribute to more equal distribution of work, it is possible that they are indicators of a 'familymyth'.According to Hochschild (1989), familymyth is created when there exists a discrepancy between the actual division of work between partners and the expected, or preferred division.In her study she found that some couples with non-traditional attitudes about gender roles and division of housework presented, or perceived their division as more equal than it actually was.Persons with higher education and more non-traditional attitudes are more likely to value equality highly and thus report the division of tasks as being more equal to confirm their ideals ofthe familial division of work.It is also possible that even a small increase in men's participation may be judged as substantial by their partners, depending on the wives' initial ideas and norms about gender behavior.Another possible explanation is that Table 5.An ANCOVA model of the relative division of five household tasks, with selected independentvariables housework thus failed to be supported by this data.Neither relative eamings nor relative educational level had a significant effect upon participation in housework.
In the preliminary analysis it was föund that wives' vocational education explained more of the variation in the number ofhours men used för housework than men's own educational level.Thus, a model which included a variable measuring both partners' vocational education was analyzed to test whether the effect of the respondents' own education depended on their partner's education.In other words, are women with a higher vocational education able to invoke more help from their husbands, measured both by absolutenumber ofhours or by relative division, irrespective oftheir husbands' education?Table 6 (page 58) seems to provide at least some evidence för this.In the table, only the unadjusted and adjusted means för the factor measuring both partners' education are presented (the effects of control variables did not change markedly from those presented in Tables 4 and 5).The relative division was most equal (the wives' share was smallest and the husbands' greatest) in the two categories in which the wife had at least vocational institute education.Contrary to the power perspective, the husbands' share actually increased (and the wives' decreased) when their level of education was higher than that of their wives (from 25.5% to 30.3%), and was the highest (33.1%) in the category where both partners had a higher education.However, the effect of the variable measuring both partners' education on the relative division of housework did not get statistical support from absolute time use.In other words, although the time use among men supported the notions made on the hasis of the relative division,the effect failedto be statisticallysignificant (it had an almost significanteffect (p=0.055)upon women's time use).
The models analyzed in this article were able to explain only 7 to 13 percent of the variation in the number of hours used för housework.Selected variables predicted more women's time use than men's.In addition to attitude factors explaining the time spent on housework, the number of small children had a significant effect on women's work.On the other hand, men's contribution was affected more by the supply factor than by other factors.Increased hours in employment reduced men's participation in housework.Vocational education had a predicted effect upon women's time use, but failed to have any significant effect on men's time use.Thus, the second assumption that education affects only women's behavior did get support from this data, even with the inclusion of a sex-role factor.It is possible that educational level captures life-style preferences or values which were not measured by the attitude factors used in these models.
The variables chosen för the analysis were able to explain more of the variance in the relative division of tasks (14 and 16 percent).However, only the normative factors (attitudes and education) had a significant effect upon the perceived division, and, perhaps somewhat surprisingly,none of the rational or pragmatic factors did.Since the two analyses yielded somewhat different results, it is possible that some of the inconsistencies between findings from different studies may thus be attributed to different measurement techniques of the dependent variable.

Discussion
The gender division of household labor has proven to be very resistant to change.Findings in this article are similar to those reported by Niemi in her article ( 1984) based on time budget study conducted in Finland in 1979.Education and age cohort are only able to explain women's participation in housework, but not men's.The effect of age seems to capture more a life course effect than a cohort effect.Although it may be that younger generations are socialized into reduced standards of housekeeping or that they are more willing to use new technologies or buy outside services, and that it can be expected that in the future women's time use on housework will still decrease, the similar findings of the time spent on housework in age groups in this article and in the 1979 data seem to suggest that life course events are more important in determining women's time use than the birth cohort.Birth of a child, and number of small children in the family increase both partners' contributions, but the increase is more marked among women.In studies from other countries it has been föund that marriage or union förmation alone marks a difference in the time use and the division ofhousehold tasks: women's contributions increase and men's decrease (South & Spitze 1994;Jacobs et al. 2000).
Although the available time, when it was measured as weekly working time, did not have any significant effect upon women's time use, it is possible that in a future analysis we need to refine the concept ofavailable time.Partners' overlapping hours in employment, or shift working may be better predictors ofboth men's and women's participation in housework than simply the total time in employment, as already pointed out by Presser (1994).Increase of part-time work and flexible working hours may thus in the future have important effects in the familial division ofboth paid and unpaid work.It is also possible, that other commitments than paid work reduce the real available time.Some ofthese may be related to education, för example, the need to bring work from the work place to home, working extra hours, participating in other activities (meetings, social clubs, etc.) may be more common among educated persons, thus requiring more contributions at home from the partners.
In this article, division of work was examined by föcusing on only five household tasks.
It can be argued that inclusion of other tasks, such as child care, might have produced another picture.There is some evidence that gendered allocation of work at home is less accentuated in tasks related, för example, to child care.Indeed this was the case also with the Finnish data.The ratio of men's time use on child careto that ofwomen's was 60-70 percent in households with children under 10, when their share ofthe time used on routine household tasks varied between 35 to 50 in the same households.
Since the normative factors turned out to have an impact on the division of work in the family both in the analysis of absolute time expenditure and relative division, it is evident that more research upon gender role attitudes and norms affecting the division of household chores is needed.Men in general seem to hold more traditional attitudes towards family and gender roles (Ervasti 2000;Melkas 1998), but since the majority of them feel that their role in the family should be increased, it is possible that not only the expectations towards men's role in the family but also their behavior can be expected to change gradually.The need to include attitude or normative factors in future studies is also implicitly supported by the findings that different factors affected men's and women's time use.Why, för example, do children affect women's time use, but not men's, or why does time in employment decrease men's participation, but not women's?The fact that wives' education affected husbands' time use more than the husbands' own educational level, and that women with more non-traditional ideas about men's role in families did a lower share of housework seems to suggest that wives' attitudes, along with their work commitments, may be important factors in explaining the husbands' participation.
The Finnish study was part ofa European project entitled 'Policies and division of paid and unpaid work in the family' in which a number of European countries agreed to conduct a survey along similar lines.In the future, it may be hoped that the possibility to use comparable data from countries with a different cultural and political context will provide more insights into the factors that determine participation in housework.

Table 2 .
Absolute time spent on five household tasks (hours per week)

Table 3 .
Relative distribution of tasks (proportion based on own account)

Table 4 .
An ANCOVA model of the absolute time expenditure (time is measured in minutes/week) on five household tasks, with selected independent variables Note: In the factors age group and vocational education, average minutes ofhousework are transformed into hours to make it easier to understand the results.In the grand mean and covariates, average time/ change in the time use is measured in minutes/week.

Table 6 .
The effect ofboth partners' education level on the number ofhours spent on five household tasks and on the relative division of tasks