Gendered division of housework in Germany – The role of self-employment, relative resources and gender role orientation

The purpose of this paper is to investigate the division of housework within couples in Germany by taking employment status, relative resources and gender role orientation into consideration. We use a large scale primary data collection that deliberately oversampled self-employed and included questions on role orientation. While selfemployment and work autonomy was related to a lower share of housework for men, rather the opposite was true for women. Furthermore, the results indicate that the relative resources and bargaining theory and the time budget approach seem to be less relevant for female self-employed compared to their employed counterparts. Our data allowed for a direct control of the gender role orientation and shed more light on the relationship between relative resources and the share of housework. A traditional role orientation was found to be highly significant for the share of housework for men and women but did not moderate the effect of relative resources. Thereby our study supports the distinct effect of gender role orientation. This can be seen as an important contribution to the ongoing discussion where relative resources are interpreted in the light of gender role orientation.


Introduction
Gender inequality regarding the division of housework appears to be very persistent throughout a context of changing labor markets.The rising involvement of women in higher education and paid employment increases the pluralization and individualization of partnerships and families (Blossfeld/Timm 2003).Compared to women's rising involvement in paid work, however, their high share of unpaid work and their main responsibility for the household remains unchanged (Peuckert 2002).This paper analyses primary data collected in Germany and aims to address two research gaps.
First, we investigate the effect of self-employment on the share of housework.Little is known about this relationship and self-employed are often not discussed in previous studies.Our primary data allows us to investigate this special group since self-employed were deliberately oversampled in our study.Furthermore, we are interested in working conditions such as work autonomy.This is motivated by the statement that selfemployment leads to higher flexibility and autonomy of paid work and at the same time allows for freedom in the arrangement of the private life sphere (Lauxen-Ulbrich/Leicht 2003;McManus 2001).How this arrangement is made under flexible working conditions is rarely discussed in previous research.It is suggested that -due to the remaining female connotation of childcare and other household tasks -the risk increases for women that a higher flexibility at paid work leads to a higher responsibility for unpaid work instead of freedom from work (Henninger/Gottschall 2005;Wimbauer 2010).
Second, our dataset allows for exploring different explanatory approaches for the division of housework.It hereby fills an important gap, i.e. by including a direct measure for preferences and norms regarding gender roles, which was seen as limitation in previous studies (Kühhirt 2012).Additionally, the dataset includes a variety of work and family characteristics of the respondent and their partner to account for economic explanations.
The institutional context and country specific gender cultures were found to play an important role when it comes to gender equality in time use and the share of household duties performed by men and women (Geist 2005;Hofäcker et al. 2013;Knudsen/Waerness 2008;Treas 2010;van der Lippe et al. 2011).Comparing different welfare regimes (Esping-Andersen 1990), namely the "liberal", the "conservative" and the "social-democratic" welfare states, the gender gap in housework was found to be highest in conservative countries and lowest in egalitarian Scandinavian countries.Hence, Germany as a conservative country is an interesting case to study in terms of housework and traditional role orientation.Furthermore, the institutional context in Germany rather fosters women's secondary role in paid employment.Institutional care for children under three years is still scarce, with 80% of this age group not being enrolled in formal childcare at the time of the survey, in 2010 (Eurostat 2014).Part-time work is a common feature to compensate for the lack of (fulltime) childcare and to reconcile work and family responsibilities.In 2010, a comparatively high share of 45.5% of German women indicated to work part-time, while only 9.7% of men did so (Eurostat 2014b).Therefore, paid work is strongly gendered in this country context and the division of unpaid work can be expected to be equally gendered.Given the context of a rather rigid labor market in Germany, combined with limited institutional childcare options, self-employment is put forward as possible solution to deal with reconciliation problems of work and family life.The noticeable stronger increase of female self-employ-ment (780,000 to 1.3 mm) between 1991 and 2010 compared to male self-employment (2.3 mm to 2.9 mm) in Germany (Statistischesb Bundesamt 2012a) might be related to the need for self-determined flexibility and autonomy at work especially for women.Hence, women in Germany might choose self-employment for a better reconciliation of paid and unpaid work which might affect their share of housework.
The following section shall give a short overview on relevant theories and recent studies about the division of labor within couples, followed by a section on the effect of self-employment.We then illustrate the current situation for self-employed and employed individuals regarding their time use with data from the German Time Use Survey.This section is followed by multivariate regressions with our primary data to give a holistic view on the division of housework between partners.

Explanatory approaches for the division of housework
There are several prominent explanatory approaches for the gendered division of housework.The normative approach explains gender differences by the individual gender role orientation: a traditional role orientation leads to the female carer and the male breadwinner role (Fenstermaker 2002).The relative resources and bargaining theory (Lundberg/ Pollak 1996) assumes that partners bargain their share of household tasks, dependent on their resources at the labor market.According to this theory, women with high labor market qualification have high bargaining power to reduce their household tasks.However, their bargaining power depends on their relative recourses compared to their partners.Women can be disadvantaged in this regard since they are usually younger than their male partners (Skopek et al. 2011) which leaves them with less years of work experience.Additionally, differences in education can be found: women tend to search for an equally or higher educated partner, while the opposite is true for men (Blossfeld/Timm 2003).These structural characteristics of couples already disadvantage women in their bargaining power, leaving them with the higher responsibility for the household.Esping-Andersen (2009) identifies women's relative wage as most important influence on their own housework time and their husbands' time.Regardless of any other resources, the time budget approach focuses on the time resources (Hill/Kopp 1995): the partner with more hours in paid work has to do/does less unpaid work.This approach is also generally gender neutral, but indicates a gendered pattern as well.Women, especially in Germany, work more often part-time than men which is partly due to their higher family responsibilities.
While the normative approach argues from a perspective of gender roles and the bargaining approach does so from an economic perspective, a combination of both perspectives might be relevant.This means that high relative resources only lead to a reduction of housework for women if they have a progressive role orientation.Our dataset allows for directly testing the moderating effect of the gender role orientation on the other approaches which was not done in previous studies.
Hypothesis 1a: We expect to find a moderating effect of the gender role orientation on the relationship between relative resources/working hours and the division of housework.
Connecting gender roles to the resource theory, previous studies refer to a so-called deviance neutralization hypothesis (Brines 1994;Greenstein 2000) to explain the finding that women with higher labor market resources than their male partner still have a higher share of household tasks, compared to women with lower or equal resources (Bittman et al. 2003;Evertsons/Nermo 2004;Schneider 2011).This hypothesis describes that the atypical arrangement of a female breadwinner is compensated by adopting rather traditional roles in the family sphere, following the "doing gender" approach.Support for this u-shaped relationship between relative income and share of housework was found for women in the USA (Evertsons/Nermo 2004; Schneider 2011) and Australia (Bittman et al. 2003), and to some degree in Germany (Haberkern 2007).On a macro level, it was argued that in countries with high gender inequality, women's high relative earnings are considered to be more "deviant", compared to countries with lower gender inequality (Gupta et al. 2010).Accordingly, one would expect to find an indication for this neutralization hypothesis in Germany, where income inequalities among women are high and high relative earnings of women can be seen as "deviant".
Hypothesis 1b: We expect to find evidence for the deviance neutralization hypothesis for women in Germany.
Evidence for the deviance neutralization hypothesis was recently criticized by Sullivan (2011) who argues that previous findings on this hypothesis simply pictured low income households and refers to their more traditional gender roles.This critizism can be tested with our dataset by including gender role orientation.

Hypothesis 1c:
The deviance neutralization hypothesis should only be relevant for women with a traditional gender role orientation.

Gender arrangements in the context of self-employment
The effect of self-employment on the division of tasks at home has rarely been addressed in previous research.The few findings are often rather descriptive, follow a qualitative approach, focus on childcare or are studied in a different institutional context.Results from liberal countries like the USA, the United Kingdom and Australia indicate that selfemployment of either men or women rather fosters traditional division of labor (Baines et al. 2003;Craig et al. 2012;Bell/LaValle 2003;Gurley-Calvez et al. 2009).The effect for men's self-employment is usually explained by high work commitment and work load, while the effect for women's self-employment is interpreted by their job motivation to have more time for their children and for a better reconciliation.In a European comparison on childcare, Hildebrand and Williams (2003) found a similar effect for men in all investigated countries: self-employed men did less childcare than their employed counterparts.In contrast, the effect for women varied strongly between countries; for Germany they did not find significant differences in time spent on childcare between employed and self-employed women.However, the study of these authors includes only few control variables and no additional explaining variables.In contrast, a study on Sweden suggests that female self-employed spent more time on market work and are more likely to divide housework equally with their partner compared to female wage-employed (Mångs 2011).This indicates that the institutional and cultural context can be important for the relationship between self-employment and the division of housework.
Acknowledging that self-employment is a rather heterogeneous type of employment including solo self-employed individuals and high growth business owners (Carroll/ Mosakowski 1987), work characteristics are important to take into consideration.Selfemployed often have longer average working hours than employees.For 2011, the German Microcensus reveals that self-employed men (44.2 hours/week) and women (31.8 hours/ week) worked remarkably longer than employees (men: 35.4 hours/week; women: 26.7 hours/week) (Statistisches Bundesamt 2012b).Following the time budget approach, this should lead to a lower share of housework for men and women in self-employment.However, women are also more likely to be solo self-employed than men in Germany and many other European countries and women's businesses have a slightly lower likelihood to survive (Arum/Müller 2004).Furthermore, as previous literature suggests, the motivation to become self-employed has a gendered component.Achieving a good work-life balance was found to be specifically attractive for women in their choice for an entrepreneurial career (Orhan/Scott 2001;Mattis 2004) and a large share of women is motivated by the aspect of higher autonomy (McKie et al. 2013).Hence, it could be argued that women who need or want to take over a high share of housework become self-employed or choose jobs with higher autonomy.Hypothesis 2a: Thus, we posit that self-employment is related to a higher share of housework for women but to a lower share for men.
Given the lack of previous studies in Germany, it is useful to examine data from the German Time Use Survey 2001/02 for a better understanding of self-employment with regard to housework.The data shows clear gender differences regarding the time spent on housework tasks in partnerships.Women spent on average more time (226 minutes per day) on these tasks compared to men (132 minutes per day).A differentiation by type of employment shows only minor differences between women but strong differences between men.Employed men spent 49 minutes more than selfemployed men on household tasks.The same is true for the subsample of full-time working men (working more than 38 hours per week).The difference between both employment types is a bit lower for the subsample with an academic degree (26 minutes difference).For women, the differences between both employment types are marginal.Looking at the gender gap for both types, accordingly, the differences between men and women are much greater in self-employment than in dependent employment.Generally, gender gaps are lower in the two subsamples of full-time workers and those with an academic degree.An interesting switch can be observed for full-time working women.While selfemployed women otherwise tend to do rather more housework than employed women, this relationship changes in the full-time sample: in this category, self-employed women spent somewhat less time on housework than their employed counterparts.In conclusion, this data provides a first impression on the total amount of time spent on housework in Germany and points out that the gender gap in self-employment appears to be higher than in paid employment; mainly owing to the differences between men.This appears to be less the case for individuals working full-time which highlights the importance of working hours for the relationship between self-employment and the division of housework.Little is known about the dynamics of explanatory approaches for the share of housework with regard to self-employment.To test the relationship of the three explanatory approaches with the employment type, we include interaction effects for each approach.
Hypothesis 2b: The negative correlation between (a) absolute working hours, (b) relative income, and (c) progressive role orientation with the share of housework is stronger for self-employed women, compared to employed women.

Data and method
To analyze the division of tasks within couples we use a primary data collection.The sample was collected by an online questionnaire that was distributed via different career networks.Some of these networks were targeted on self-employed or on women, to increase the share of these two groups.Therefore, our online sample contains proportionally many women, self-employed and higher educated individuals.An additional random sample from Computer Assisted Telephone Interviews (CATI; n=721) expanded the online sample (n=1645) and is used as a control for a sample selection bias of the online sample.
For the following analysis, we exclude individuals without partners and those who had missings on one of the used variables.The division of housework was captured with a question on the own share of the total housework.Answers about the own share (compared to their partners) could be given on a scale from 0-100% in 10% steps.The share of the partner was thereby automatically calculated.Hence, any help from a third person was deliberately excluded.To use the share rather than the total amount of tasks as a measure was also suggested by Geist (2005), since a direct question on the share could provide a more realistic picture of the actual allocation of tasks, compared to accumulated time-use tasks.A disadvantage of this kind of measure could be an overestimation of the own share.Six categories captured the job position: self-employed, public servant, helping family member, employee at the partner's firm, employee, blue-collar worker, and trainee.For indicating if someone is self-employed, we use a dummy-variable.Autonomy at work was asked with an item "work autonomy" which could be rated on a 5-point Likert-type scale on how far the respondent achieved work autonomy in their current position.Weekly working hours were asked as "actual hours worked" and -if applicable -"contractual working hours".To compare self-employed and employees, we use the actual working hours.Income was asked as the average monthly (net) income for self-employed and employees.Respondents could classify themselves into one of fifteen income groups.For the analysis, the midpoints of the classes are used.The analysis is restricted to persons who reported their own or their partner's income to be 10.000 euro at a maximum because of the small number of cases in higher income groups.The educational level is coded as dichotomous variable to differentiate between individuals with and without an academic degree.The respondent's age is computed from the year of birth and is measured in years.A dummy-variable for children indicates if children are present in the household.Role orientation was captured by four items (Cronbachs alpha= .810) on a 5-point scale and is merged to an index by taking the average of their sum.Thus, the index has a minimum value of 1 and a maximum value of 5.These items were: "It would be good if there were part time positions also for men so that they can care more for children and household", "I think it is good when men interrupt their career in order to care for children so that women can continue working", "I think it is good to share housework equally between me and my partner", "I think it is good to share family obligations (like childcare and parent-teacher conferences) equally between me and my partner".
A higher index value indicates a more traditional role orientation.Some of the measured variables are used in the analysis as the value of the respondents relative to their partners' value.As a restriction, it has to be mentioned that the information for the partner is given by the respondent which may limit the accuracy.
Differences in working hours are metric variables, where negative values indicate that the partner has more working hours whereas positive values mean the same for the respondent.The difference in educational degree is a dummy-variable indicating if respondents have an academic degree while their partners do not.Share of household income is the respondent's contribution to the household income in percent, transformed to a scale ranging from -5 to +5.The lower endpoint means that only the partner contributes to the household income whereas respondents at the upper endpoint earn all of the household income themselves.In our analysis, both partners have an income, therefore all values lay between those extremes.A one-unit difference on that scale means a difference of 10 percentage points in the share of household income.To capture additional effects of income differences and gender role orientation, a categorial variable was included with the value 1 if respondent has a lower income than their partner, value 2 if they have a higher income and a traditional role orientation and value 3 if they have a higher income but no traditional role orientation.A traditional role orientation was defined by having an index of 2 and higher.Last, to account for level differences in the dependent variable caused by the survey design or mode of data collection, we include a dummy variable in the regression models indicating whether a respondent belongs to either the CATI or the online sample.
The share of household tasks is investigated by multivariate, ordinary least square (OLS) regressions in two separate models for men and women, and gender differences were tested by a Wald test.

Results
Even in our sample with many highly educated women, the division of housework is rather traditional.Women do a significantly higher share (64%) of household tasks than men (36%) do.

Table 2:
Linear regression on housework (Focus: Relative resources)  Results from our first model (M1) show that high relative income is related to a lower share of housework for men and women.However, there is no indication for a u-shaped relationship and our hypothesis H1b does not get affirmed.Hypothesis H1c could not be directly tested due to these unexpected results.Following Sullivan, we tested in M2 if the relationship of women's higher income on their share of housework is related to their traditional role orientation.Women who earn more than their partner do less housework, irrespective of their traditional role orientation.Again, this indicates that there is no deviance neutralization and that even women with traditional role orientation do less housework when they have higher earnings than their partner.Model 3 and 4 test the moderating effect of traditional role orientation on the effect of relative resources and working hours.Traditional role orientation is strongly related to the share of housework: men take over a lower share of housework and women a higher share.High relative resources and high relative working hours are related to a lower share of housework for men and women, which is in line with previous research.Controlling for role orientation, these effects stay fairly the same (not shown).The interaction effects of the share of household income and the role orientation working hour differences are not significant (M4).Hence, a higher share of household income and higher working hours are related to a lower share of housework for men and women, irrespective of their role orientation.H1a can be rejected.
Table 3: Linear regression on housework (Focus: Self-employment) Results from table 3 shed more light on self-employment with regard to the division of housework.Without controls, self-employed men have a significantly lower share of housework compared to employed men, while this is not the case for women (M5).However, the effects are not significant after controls in the final model (table 4).In this model (M9), a Wald test shows that the difference between men and women regarding selfemployment is not significant (χ² (1) =1.18, p=0,2783).Hypothesis H2a can be rejected.
Our results suggest that differences in the type of employment are due to differences in working conditions, i.e. work autonomy and working hours.
High working hours and high relative income are related to a lower share of housework for women.However, an interaction effect with the type of employment reveals that this relationship is less strong for self-employed women compared to employed women.Self-employed women with a high share of income (M6) or high working hours (M7) take over a relatively higher share of housework compared to their employed counterparts.Hence, the relative resources and bargaining theory and the time budget approach seem to be less applicable for women in self-employment.In the final model, these interactions are not significant anymore and are therefore excluded.The relationship of traditional role orientation and the share of housework seems to be independent of the employment type.Thus, the normative approach is also relevant in self-employment.Hypothesis H2b is not supported by the results.The final model (table 4) reveals another interesting aspect with regard to specific working conditions.It was argued that self-employment is related to higher work autonomy, which in turn might affect housework.The results show that high work autonomy is related to a significantly lower share of housework for men.Rather the opposite is true for women and the gender difference is significant (χ² (1) =8.32, p=0,0039).Hence, these re-sults give evidence for the hypothesis that work autonomy is used differently by men and women with regard to their share of housework which is in line with previous research on self-employment and contributes to the open questions on the role of working conditions in this type of employment.

Discussion
Our first aim of this study was to shed light on the relationship between self-employment or work autonomy and the share of household tasks for men and women.We find a gendered effect which can be interpreted as evidence that men with high job autonomy (which is more prevalent in self-employment) might use it for a higher involvement at their work sphere and reduce their housework.Women, on the other hand, are often motivated to choose self-employment for a better reconciliation of work and family responsibilities.Different to men, they do not take over a lower share of housework when they are self-employed.This result is in line with previous research about self-employment from other countries (Baines et al. 2003;Bell/LaValle 2003;Craig et al. 2012;Gurley-Calvez et al. 2009;Hildebrand/Williams 2003) stating a more traditional division of housework among couples with at least one self-employed partner.In addition to previous literature, our study shows that -at least for men -work autonomy has a stronger correlation with the division of housework than the actual type of employment.Thereby, our results shed some more light on Hildebrand and Williams' (2003) call for future research regarding the role of flexibility in self-employment.Furthermore, our study contributes to previous literature by testing different explanatory approaches with regard to employment type.Some indications can be found that the time budget approach and the relative resources and bargaining theory are less relevant for self-employed women.Hence, women in selfemployment might benefit less from higher income and high working hours with regard to their bargaining power.Another interpretation could be that highly work-oriented women with high domestic demands might choose self-employment to combine high working hours with a relatively high share of housework.A longitudinal approach could help to shed more light on this issue.In contrast to results from Sweden (Mångs 2011), self-employed women in Germany do not seem to divide housework more equally with their partners than employed women.This indicates that the relationship between women's employment status on the division of housework can depend on the institutional and cultural context, including different opportunities and motivations to become selfemployed.In countries, where the reconciliation of work and family is easier in dependent employment, self-employed women might be less motivated by reconciliation issues but rather by career aspects.Depending on this career choice motivation, self-employed women might rather work full-time which is related to less time spent on housework, compared to dependent employed women (see figure 1).The effect of men's employment status, on the other hand, seems to be more homogeneous across countries.On a descriptive level, our study confirms the persistent finding that self-employed men do less housework than employed men.Hence, at least for men, self-employment as a job type is possibly impaired by particularly high working hours, high work commitment, and other work characteristics that are related to a lower share of housework.
The second purpose of our analysis was to investigate the gender-role orientation and relative resources with regard to the division of housework.Generally, we find a rather traditional picture on the division of housework.Nevertheless, high relative work resources seem to be used by men and women to bargain a lower share of housework.In contrast to previous research and theories that suggest a u-shaped relationship between relative income and housework for women (Evertsons/Nermo 2004 for the USA; Schneider 2011; Bittman et al. 2003), we do not find a u-shaped effect which is surprising given the rather conservative context of Germany.Our results, however, shed some more light on the critique by Sullivan (2011) in terms of the importance of gender role orientation.Higher relative income is related to a lower share of housework for women irrespective of their gender role orientation.Concluding, gender role orientation plays an important role for the division of housework.However, it does not seem to moderate the relationship between relative resources/working hours and the share of housework and it is independent of the employment type.Thus, our study contributes to the underresearched aspect of gender role orientation with its strong and discrete relationship to the share of housework.
There are several limitations to this study.In contrast to some previous studies, we measure the share of housework and not the amount of time.Therefore, our results are not directly comparable to those studies, as suggested by Schulz and Grunow (2007).This limitation might explain gradual differences in the finding of a u-shaped relation between relative income and housework in Germany.Whereas Haberkern (2007) finds that women with higher positive income differences do more and more housework, our findings do not suggest this u-shape relation.Furthermore, our sample for the multivariate analysis might not be representative for Germany in general, since self-employed and highly educated individuals are overrepresented.Last, a longitudinal approach would be beneficial for understanding the mechanisms behind the division of housework.In particular, future research could profit from data on changes from dependent employment to selfemployment including the underlying motivation to explain gender differences in the division of housework.Specifically under a changing family context like the birth of a child, the impact of self-employment on housework division might differ and should be considered in further investigations.

Table 1 :
Sample description

Table 4 :
Linear regression on housework(Final model)