DYADIC ANALYSES OF THEDIVISION OF DOMESTIC LABOR:COMPETENCES,PREFERENCES AND MEN’S GOODWILL

The persistence of a distinctly gendered division of domestic labor in Western societies remains puzzling. Beyond standard economic and normative explanations, more recent approaches emphasize aﬀective, cognitive, and incorporated aspects of housework and the production, reproduction, and negotiation of gendered expectations via social interactions. However, the relevant indicators for these more implicit mechanisms are not routinely included in social surveys. Based on a unique set of items and a representative sample of heterosexual couples (N=1,396) from pairfam data (German Family Panel) (wave 10), we contribute to this research by analyzing the mutual eﬀects of both partners’ enjoyment, quality standards, and reciprocally perceived competencies on the division of domestic labor. An actor-partner interdependence model (APIM) is applied, which is adequate methodologically to model the partner dyad. Both an actor’s own and their partner’s assessments of competences and preferences – and particularly men’s attributes and perceptions – prove to be powerful and statistically signiﬁcant predictors of the division of domestic labor. The results contribute quantitative evidence on processes of doing and undoing gender in context.

unpleasant chores by opting out of housework.Consequently, both approaches are genderblind and focus on structural constraints and socioeconomic determinants, which happen to be dissimilarly distributed among men and women, for explaining gender differences in the division of domestic labor.Normative beliefs and inherent enjoyment and preferences for housework (i.e. the direct utility of housework) are either neglected, as in Becker's family economic approach, or housework is presumed to be onerous (creating disutility) and generally avoided by those with the power to do so, as assumed by the resource and bargaining approaches (Ahlander & Bahr, 1995;Stratton, 2012).
Second, sociocultural gender theories emphasize the role of societal norms and gender beliefs in the division of domestic labor above and beyond structural determinants.Women's disproportionate contribution to housework is seen as the outcome of persisting gender roles in society which assign men and women to different tasks, with household labor being regarded as a female duty, and employment outside the home as a male duty (Parsons & Bales, 1955).These gender ideologies are internalized and incorporated by processes of socialization, leading to stable gender differences in preferences and competences for different tasks which are assumed to persist throughout life (Adams & Coltrane, 2005).Empirical research in this field has mainly focused on general assessments of gender ideology, for example the effects of traditional versus egalitarian gender role attitudes for explaining gender differences in the division of domestic labor (D.L. Carlson & Lynch, 2013;Davis & Wills, 2014).
To date, a great deal of empirical work has been carried out providing extensive evidence for both theoretical strands (Coltrane, 2000;Davis & Wills, 2014).However, even when taking into account both gender ideology and economic factors, the lopsided division of domestic labor prevails.Neither approach can claim dominance and both approaches are limited in their explanatory power (Aassve et al., 2014;Geist & Ruppanner, 2018).Thus, much of the persisting gender inequality remains unexplained and puzzlingthe more so as socioeconomic inequalities within couples have decreased and egalitarian gender role attitudes increasingly prevail, in particular among younger cohorts (Goldscheider et al., 2015).
Against this backdrop, alternative explanations have increasingly entered the literature on family life, pointing to additional, more implicit mechanisms that translate into gender differences in the division of domestic labor.Constructivist 'doing gender' approaches emphasize the contextual and performative nature of gender and the symbolic meaning of housework.Performing housework is not a mere technical activity in reaction to objective needs but also a way of establishing and displaying gender identity ('doing gender'), emerging from deeply inscribed cultural expectations about gender-appropriate behavior and practices (Berk, 1985;Ridgeway, 2011;West & Zimmerman, 1987).Thereby, gender itself and the symbolic meaning of housework are actively produced, reproduced, and negotiated via social interactions.In this perspective, gendered labor division does not automatically reflect abstract societal gender roles but needs to be understood and analyzed in the social and interactional context of the actual relationship (Berk, 1985;Brines, 1994;Ridgeway, 2011).Moreover, as egalitarian attitudes become more widespread, more men and women actively strive to overcome traditional gender roles by re-interpreting and disentangling the symbolic associations of housework and gender, thereby initiating and enacting processes of 'undoing gender' (Deutsch, 2007;Rezeanu, 2015).
The usefulness of abstract normative concepts and global gender role attitudes for explaining gendered labor division is also questioned by another strand of literature: Originating from the work of Ferree (1991) and Berk (1985), these approaches question the implicit assumptions of traditional theories and focus on how exactly (macro-level) social norms translate into individual motives for performing housework, and how variability in housework arrangementsdespite the structural similarities of couplescan be explained.
Numerous studies analyze the affective (the actual enjoyment of tasks), cognitive-evaluative (quality standards and perception of responsibilities), and incorporated aspects of housework (competences in housework) (Ferree, 1991;Grote et al., 2002;, 1990;Pittman et al., 1999;Poortman & van der Lippe, 2009;Robinson & Milkie, 1998;Thébaud et al., 2019).However, most of the work focuses on perceptions of fairness and satisfaction with housework, rather than the division of domestic labor per se.Empirically, qualitative evidence dominates in both strands (M.W. Carlson & Hans, 2020;Coltrane, 1989;Tichenor, 2005;van Hooff, 2011), albeit this often represents a starting point for quantitative analysis.However, quantitative testing of gender performativity and the more implicit motives for housework itself appears challenging.
There is an almost total lack of appropriate indicators; where they are available, it is only in unrepresentative surveys and with small sample sizes (Berk, 1985;Ferree, 1991;Pittman et al., 1999;Poortman & van der Lippe, 2009;Robinson & Milkie, 1998).
In this context, we believe that investigating the affective, cognitive, and incorporated motives for performing housework has the potential to enhance our understanding of gendered labor division in the household beyond economic and normative explanations, allowing us to draw a more nuanced picture.In particular, we think that housework arrangements need to be analyzed in the interactional context of the actual partnership where symbolic meanings of gender and housework are enacted.We thereby focus on the gendered domestic labor division as indicated by the share of housework each partner performs.We use the term housework to refer to the routine, stereotypically female, and time-intensive tasks of cooking, doing laundry, and cleaning.Moreover, we focus on heterosexual couples sharing a household, as we are particularly interested in gendered processes of household labor allocation.We contribute to the literature in several ways: First, we focus on specific motives, associated with housework as reflected in genderspecific enjoyment, quality standards, and (perceived) competences with respect to household labor.To this end, we make use of a unique set of variables specifically included in the representative German panel cohort study pairfam for the purpose of this study, allowing us to consider all three aspects together.This approach enables a more direct measuring of the meaning of housework and underlying subjective reasons.
Second, we analyze both partners' information within a dyadic framework to account for the contextual nature of gender and housework.Within this framework, direct and reciprocal effects of partners' characteristics on housework arrangements are examined simultaneously.In this way, we explicitly take into account that doing gender and undoing gender is a "team performance" (Tichenor, 2005, p. 194).Gendered expectations are demonstrated to oneself and one's partner (and others in general), resulting in a performative interaction within which both partners' behaviors adapt to one another.
Third, with regard to competences, the data includes not only both partners' assessments of their own housework competences, but also both partners' assessments of the other partner's competences.We thereby explicitly exploit the theoretical potential of divergences between individuals' own and their partner's perceptions of competences.In this way, we also provide an alternative quantitative empirical approach to 'doing gender' theories with respect to the division of domestic labor 1 .
Fourth, our study also approaches the dyadic nature of couple data methodologically.
Studies on the division of (domestic) labor are mainly based on the information of one household member, which is not necessarily a comprehensive representation of each individual's contribution (Geist, 2010).Further, studies of couples' housework share which do include both partners' information are mostly based on simple regression models, violating the assumption of independence.We thus propose actor-partner interdependence models (APIM) (Kenny et al., 2006) which allow for the dyad to be the unit of analyses and for a methodologically adequate separation of individual and partner effects.
Our study is based on the German pairfam data, a household study that includes the birth cohorts 1971-1973, 1981-1983, and 1991-1993.In Germany, the gender gap in housework is relatively large compared to other Western societies.Based on German time-use data from 2012/13, the time women spend doing housework (cleaning, doing laundry, and cooking) is 2.4 times higher than the time men spend on these tasks (Federal Statistical Office [Statistisches Bundesamt], 2015).

Enjoyment of Housework in Context
With respect to the role of enjoyment, we explicitly acknowledge that housework can have intrinsic value and might be associated with feelings of pleasure, motivating actors to engage in more household tasks.We therefore expect that the level of enjoyment of housework will be positively associated with an individual's involvement in housework.However, we expect that women will, on average, show higher levels of enjoyment for housework than their male partners.We derive this assumption not only from socialization theories but also from constructivist gender display approaches which emphasize that doing housework is also a way of doing gender (Coltrane, 1989).By performing housework, women also produce utility in the form of gender identity and reassure themselves and others that they are fulfilling female ideals, such as being an attractive partner and/or a good mother (Poortman & van der Lippe, 2009;Robinson & Milkie, 1998).In addition, since housework is also an essential part of family care work and crucial for establishing family bonds, housework might be perceived more as expression of love and care for family members than a task to be avoided (Poortman & van der Lippe, 2009;Robinson & Milkie, 1998;Tichenor, 2005).In contrast, masculine identities are associated with disliking tasks that are regarded as typically feminine such as cleaning, doing laundry, tidying up, and decorating the home (Coltrane, 1989).Thus, despite a general shift from traditional to more gender-egalitarian attitudes, gendered expectations are still present, although they may be more likely to operate by way of gender-specific enjoyment levels.This is also supported by more recent studies, which show that women enjoy doing housework more than men (Grote et al., 2002;Poortman & van der Lippe, 2009;Stratton, 2012).
Furthermore, since interactionist perspectives emphasize that the production and negotiation of gendered expectations and the symbolic meaning of housework is a team performance accomplished via social interactions, the interplay of both partner's attitudes needs to be taken into account.Estimating separated regressions for men and women based on a (non-representative) sample of employees in organizations, Poortman and van der Lippe (2009) find that each partners' individual enjoyment increases the amount of time that partner spends on housework individually.Although an increase in an individual's reported enjoyment of housework correlates with an increase in the time they themselves spend cleaning, the respective effect on their partner's time spent on housework is, surprisingly, not significant.
Using UK Time Use Survey data, Stratton (2012) finds that women's enjoyment of housework in particular is of significance for both men's and women's contribution to housework, corroborating the relevance of social interactions.Given the fact that the gender asymmetrical division of houseworkwith women being both normatively and empirically responsible for houseworkis still the common and societally accepted way of organizing household labor, we assume in our study that men's enjoyment of domestic work will have a relatively stronger effect on achieving a more egalitarian division of unpaid labor, in particular as perceived by the female partner.In such instances, any deliberate (i.e., preferential) deviation by men from traditional gender norms will make a difference in the outcome.Male preferences for housework have also been described as symbolizing the alternative masculinities emerging from more recent discourses about gender equality, with the intention being to transgress traditional gender oppositions and develop ways of 'undoing gender' (Deutsch, 2007;Rezeanu, 2015).From this reasoning, we derive the following hypotheses: H1a: On average, women exhibit stronger enjoyment for housework than men.
H1b: Individual enjoyment of housework increases an individual's housework share among both men and women.H1c: However, comparing the relative impact of enjoyment within the couple, male enjoyment of housework will have a stronger impact on the partner's respective housework shares than female enjoyment of housework.

Quality Standards in Context
Gender-specific standards of cleanliness are another more subtle mechanism that has been used to explain the different amounts of housework men and women engage in.Socialization approaches argue that behavior, enjoyment, and attitudes towards the division of labor are developed by exposure to gender-specific adult behaviors and gendered expectations during childhood which are then reinforced throughout adulthood (Cordero-Coma & Esping-Andersen, 2018;Gupta, 2006).Such influences and expectations are not only related to skills and responsibilities, but also to standards of cleanliness which become internalized and deeply incorporated: The violation of norms of cleanliness is often associated with feelings and physical reactions of disgust and shame (Oakley, 2018, chapter 6 on standards and routines of housework).Due to gender-specific socialization, men and women attach different values to cleanliness in the home and perceive different levels of cleanliness as appropriate.Moreover, from the perspective of 'doing gender' approaches, developing and adhering to individual standards of cleanliness also becomes part of gender performance (Hochschild & Machung, 2003;Thébaud et al., 2019).The finding that women set higher quality standards for cleanliness has been established in quantitative (Poortman & van der Lippe, 2009) and in qualitative studies (van Hooff, 2011).
An experimental study by Thébaud et al. (2019) reveals that male and female participants evaluated mess similarly and saw the same urgency to clean up.However, compared to men, rooms occupied by women required a higher level of cleanliness to be perceived as clean; furthermore, women were judged more critically, revealing stronger social expectations for women to comply with norms of cleanliness.In contrast, although mess was perceived negatively for men, too, expectations of cleanliness were relatively lower for men.
Thus, empirically women are found to also feel more responsible for the household (Poortman & van der Lippe, 2009;Thébaud et al., 2019).Bianchi et al. (2000, 195) conclude that the cleanliness of the home is a reflection on a woman's competence as a "wife and mother", but not on a man's competence as a "husband and father".As Hochschild and Machung (2003) have elaborated, men use divergences in standards as an excuse to avoid housework.
Accordingly, when investigating partner effects, wives' higher standards were found to decrease their husbands' involvement (Ferree, 1991), but no partner effects were found in the studies by Poortman and van der Lippe (2009) and Pittman et al. (1999).With this in mind, and in line with existing evidence, we expect women to set higher standards of cleanliness than men.Although standards of cleanliness will increase both partners' motivation for doing housework, due to feelings of necessity (for empirical evidence see Ferree, 1991 andPittman et al., 1999), the effect will be stronger for women due to gendered expectations about responsibilities for a clean home.Thus, from an interactional perspective we expect thateven in the presence of a male partner with high standards of cleanlinessthe effect on housework involvement will be stronger for women, because women are held more accountable for a clean and comfortable home in very fundamental ways.This leads us to the following hypotheses: H2a: On average women exhibit higher standards of cleanliness than men.
H2b: Individual standards of cleanliness increase an individual's housework share among both men and women.H2c: However, comparing the relative impact of standards of cleanliness within the couple, male standards of cleanliness will have a stronger impact on the partner's respective housework shares than female standards of cleanliness.

Competences in Context
To perform housework and create a comfortable home, technical, organizational, and aesthetical knowledge and competences are needed.Competences and proficiencies in housework are thus emphasized by both economic and socialization approaches when providing explanations for the gendered division of domestic labor.It is assumed that, through gender-specific socialization, women incorporate knowledge and skills associated with housework which make them more productive in the household sphere (emphasized by economic approaches) and make them feel naturally dedicated to responsibilities in the home (emphasized by socialization approaches).Confirming the relevance of socialization, empirical studies show that girls and boys who do more housework during childhood also tend to engage in more housework later in life (Schulz, 2020).
However, from a constructivist point of view, it appears difficult to assess competences 'objectively', because perceptions of competences are themselves framed by cultural beliefs.
These beliefs are rooted in the historical segregation between unpaid domestic (reproductive) and paid market (productive) labor, which assigned women to the household sphere inside the home and men to the market sphere outside the home.In this process, competences in housework became attributed to the "essential nature" of women (Bock & Duden, 2007[1980]; Gorman-Murray, 2008).Thus, women are perceived as having higher competence levelsby mere attribution and assumption.Such beliefs are then also enacted by the actors themselves in the ongoing performance and display of gender (West & Zimmerman, 1987).For example, by demonstrating their own incompetence in household work and emphasizing women's competence, men can avoid doing household chores and thereby reassure dominant cultural constructions of 'masculinity' ("strategy of incompetence", M. W. Carlson & Hans, 2020;Ashraf, 2009).In contrast, women's sense of responsibility and displays of higher competence compared to men is also a way to reproduce feminine identity and establish superiority and empowerment in the domestic field (for an overview see Rezeanu, 2015).Grote et al. (2002) investigate perceived fairness in household labor division, and support this view by showing that women rate men's competence lower than men rate their own competence, and men rate their partners' competence higher than women rate their own competence.Moreover, men are shown to be more involved in housework when women consider them to be competent, which points to women's ability to engage in "gatekeeping" and agency with respect to housework (Allen & Hawkins, 1999).Moreover, a qualitative study by van Hooff (2011) indicates that men's postulated incompetence and women's postulated competence (and their higher standards for cleanliness in housework) are used as a justification for the unequal division of domestic labor.We therefore derive the following hypotheses: H3a: On average, women rate their competence higher than men do.H3b: Individual competences increase an individual's housework share among both men and women.
H3c: However, comparing the relative impact of competence within the couple, male competence in housework will have a stronger impact on the partner's respective housework shares than female competence.
As outlined above, perceptions of a partner's competence are also culturally framed and may be part of strategic enactments of doing and undoing gender.We therefore propose that, alongside the actor's assessment of their own competence, also deserving of special attention is an actor's perception of their partner's competence, as well as how these perceptions interact.
In the interactional context, the assumptions of the 'natural' competence of women and 'natural' incompetence of men are displayed by women who consider men to be less skilled than men themselves.Conversely, men rate women as more skilled than women themselves.
Moreover, we assume that especially women's perceptions of male competence will explain deviations from traditional housework arrangements, because any perception on the part of a woman of higher male domestic competence may reflect an enactment of egalitarian gender concepts and orientations, and thus the abandonment of "gatekeeping positions" (Allen & Hawkins, 1999).It should be noted that none of the standard theories about household labor division generates systematic predictions for either the influence of actors' assessment of their partners' competences, or gender differences between these assessments, on the division of household labor.
H4a: On average, women's rating of their male partner's competence is lower than men's own self-rated competence, but men's rating of their female partner's competence is higher than women's own self-rated competence.
H4b: An actor's positive rating of their partner's competence will decrease their own housework share.
H4c: An actor's positive rating of their partner's competence will have a stronger impact on the actor's own housework share than a positive rating of their own competence.H4d: Comparing the relative impact of reciprocal evaluation of competences within the couple, women's evaluation of male competences has a stronger impact on the partner's respective housework shares than men's evaluation of female competences.

Method Data, Sample and Variables
We use pairfam, a household panel survey funded by the German Research Foundation (Huinink et al., 2011).Pairfam is based on a random sample drawn from a German population register of more than 12,000 persons from three birth cohorts (born in 1971-1973, 1981-1983, and 1991-1993), in which respondents' partners are also interviewed.For our analytical sample, we use the 10 th wave, conducted in 2017/18, release 10.0 (Brüderl et al., 2019), since the information on enjoyment, quality standards, and competences required for our study were collected in this specific year.Thus, the value of the pairfam wave used here lies in the availability of our focal independent variables, which are not included in time use surveys and other representative data sets on labor division.Out of our sample of 1,799 couples (married and unmarried), we restrict our sample to couples living together (with or without children).
Couples who share their household with other adults (n=167; 9 percent) are excluded.As we are explicitly interested in the gender-specific dynamics, only heterosexual couples are included (homosexual couples: n=19; 1 percent).After including our 22 independent variables our analytical sample contains 1,396 couples with 2,792 individuals.

Dependent variable.
As dependent variable we use pairfam's measure of share of housework.Participants were asked to rate the item: "To what extent do you and your partner share housework (cleaning, doing laundry, cooking)?" on a 5-point response scale : '(almost) completely my partner' (1), 'for the most part my partner' (2), 'split about 50/50' (3), 'for the most part me' (4), '(almost) completely me' (5).As such, the direct assessment of housework share in pairfam has advantages over alternative relative measures used in the literature based on partners' absolute hours spent on housework.The reason is that reported hours spent on housework are likely to induce measurement error due to recall bias and the simultaneity of tasks 2 .Moreover, since both partners answer this question, we have two measures of the contribution to housework per householdmeasured separately for each partner.Unlike other studies, which rely on the information of only one partner and cannot accommodate different perceptions of partners, we do not need to presume which partner's perception of housework share is more reliable (Press & Townsley, 1998).Instead, we explicitly take into account different perceptions by applying APIMs, which allow us to consider both dependent variables simultaneously.Thus, the resulting regression model will yield two sets of estimates of covariates: One for women's perception of their contribution to housework, and one for men's perception of their contribution to housework.For computational reasons and to ease interpretation, both genders' item scores were recoded to give the following gender sensitive scale ranging from: '(almost) completely the man' (1), 'for the most part the man' (2), 'split about 50/50' (3), 'for the most part the woman' (4), '(almost) completely the woman' (5).

Independent variables.
Participants were asked about their enjoyment of housework with the item "How much do you enjoy doing housework, meaning cleaning, laundry, and tidying up?", ranging on an 11-point scale from 'not at all' (0) to 'very much' (10).Standards for a tidy and clean home were assessed with the item "How important are tidiness and cleanliness in your home to you?" using an 11-point scale ranging from 'it's enough if the basics are done' (0) to 'housework should be done with great care' (10).Self-rated competence and an individual's rating of their partner's competence was retrieved by the two items: "How do you rate your skills in housework?"and "How do you rate your partner's skills in housework?" on 11-point scales ranging from 'not very well' (0) to 'very well' (10).

Control variables.
Our investigation is focused on exploring the particular relevance of enjoyment, competences, and quality standards for understanding and explaining domestic labor division beyond the explanations already offered by structural and normative approaches.
We are, thus, interested in the marginal effect of these variablesnet of the socioeconomic influences and gender role attitudes which are deemed important by the traditional approaches.
Among socioeconomic determinants we include individual monthly net income, age, and education (in years of schooling) for each partner separately.Furthermore, following normative approaches, two items measuring gender role attitudes were included.Further confounders controlled in our models are: the region (eastern/western Germany), marital status, relationship duration, and age of the youngest child in the household.Supplementary table A1 gives a more detailed description of our variables and measures.

Analytical Procedure
In order to analyze partners' relative contributions to housework, we apply actor-partner interdependence models (APIM) which analyze data from both partners' perceptions of housework simultaneously and helps distinguish between actor and partner effects (Kenny et al., 2006).The application of APIMs is supported by the empirical evidence for a positive nonindependence between spouses (ICC=0.45).Thus, the model reports two sets of estimates: one for women's and one for men's perception of the division of domestic labor.The availability of information from both partners also enables us to determine interaction processes.For example, a woman's competence in housework may lead to a higher engagement and perception of engagement in housework (actor effect), but will also be affected by her partner's competence (partner effect).Both effects (here: effects of competences) are controlled for by the degree of similarity (or dissimilarity) between partners.The following APIM is specified as a random-intercept linear multilevel regression model, with each partner as level 1 and the couple as level 2. This way, the mutual effects of both partners' characteristics on the reported division of domestic labor can be estimated by explicitly considering the interdependent nature of partners in a relationship.APIMs were estimated using Stata (version 16.1).

Descriptive Findings and Bivariate Results
Most couples in our analytical sample are from western Germany (65 percent), married (75 percent), and living together with dependent children (75 percent).The average relationship length is 13 years.Average years of schooling amount to 14 years, which in Germany reflects a high-school degree (Abitur) or vocational training.Men's mean earnings are more than double women's mean earnings (2,406 Euro and 1,163 Euro per month).Men are on average 40 years old and women 37 years old (Supplementary table A2 gives a more detailed description of the variables and summary statistics.).
Concerning our dependent variable -partners' relative contribution to houseworkmen and women on average agree that women contribute a greater share of the housework than men (values larger than 3 indicate that more housework is performed by women), although women rate their contribution to housework slightly, but significantly, higher than men rate women's contribution (table 1).
Looking at our more direct measures of motives and attitudes towards housework, indisputable gender differences come to light.Women enjoy doing housework more than men (5.5 vs. 4.8), and also value a tidy and clean home more than men (7.3 vs. 6.7).Accordingly, women perceive themselves to be more competent in housework than men do (7.4 vs. 6.5) which is in line with our hypotheses 1a, 2a and 3a.Moreover, as assumed in hypothesis 4a, men rate their partner's housework competences (8.0) even higher than women rate their own competences (6.5), while women rate their partner's housework competences (6.2) even lower than men rate their own competences (7.4).

Determinants of Contribution to Housework
Table 2 presents the results of the actor-partner independence model (APIM).As outlined before, in the APIM, men's and women's respective perceptions of their own contribution to housework are estimated simultaneously in one model.The estimated coefficients in the left and right columns of table 2 display, respectively, the effect on the division of housework as perceived by men or women.Furthermore, for all individual characteristics, information of both partners is included.In the left (women's) column of table 2, the actor effect refers to the effect of the woman's characteristics on her perception of the division of housework, and the partner effect yields the effect of her male partner's characteristics on the woman's perception of her contribution to housework.Conversely, in the men's part of the model (table 2, right side) the actor effect reflects the influence of the man's variables on his perception of the division of labor between him and his partner, and the partner effects refer to the impact of female characteristics on his perception of the division of housework.Since, in our dependent variable, higher values reflect increasing female contribution to housework, effects of independent variables can be interpreted as effects on women's housework share.Overall, positive coefficients indicate an increase in (perceived) women's contribution to housework and negative coefficients denote a decrease in (perceived) women's contribution to housework.
As we distinguish between male and female perceptions of the division of domestic labor, all hypotheses need to be tested for men and women separately.In the upper part of table 2, standard measures derived from economic and normative theories are included: age, education, and income, as well as gender role attitudes concerning female roles in the family and male housework participation.However, because our study focuses on the additional explanatory power of preferences, competences, and quality standards, these standard variables merely function as controls and will not be interpreted substantially.In order to examine specific hypotheses, we also conducted Wald tests on the differences between the coefficients of the actor and partner effects (third column: _b[actor]-_b [partner]).
[Table 2 about here] The role of enjoyment in context.Strikingly, even in the presence of variables controlling for socioeconomic characteristics and global gender role attitudes, direct measures of housework enjoyment levels and competences for both actor and partner turn out to be significant predictors for the division of housework.A woman's enjoyment of cleaning and tidying has a positive effect (actor effect: 0.02) on her perceived contribution to housework, whereas her partner's enjoyment of housework reduces her perceived contribution to housework even more (partner effect: -0,04) (left column -women's assessment).When looking at men's contribution to housework as perceived by him (right column -men's assessment), his enjoyment reduces women's contribution by 0.08 (actor effect).Enjoyment of housework on the part of his female partner leads to an increase in women's contribution by only 0.02 (partner effect), which is the same size as the actor effect among women.Thus, considering and comparing the absolute value of actor and partner coefficients, i.e. comparing the relative impact of enjoyments level within the couple, male enjoyment for doing housework is more influential than comparable female enjoyment according to both his and her perception of the housework shareas hypothesized in H1c.Moreover, the differences in absolute values between actor and partner effects are significant for both men's and women's parts of the model (|0.021 + | and |0.064 *** |).We can accept hypothesis H1b, that individual enjoyment increases housework share for both men and women, and we can also accept H1cthat, within the couple, men's enjoyment is more influential for a couple's labor divisionas perceived by both partners.
The role of quality standards in context.Contrary to our hypotheses, the standards of cleanliness and tidiness show neither actor nor partner effects (H2b), nor differences in effects between male and female characteristics when comparing the coefficients within the couple (H2c).However, this can be explained by the correlation between enjoyment and competences on the one hand and quality standards on the other (bivariate correlation 0.36 and 0.50).People who like cleaning and are competent in housework are also more likely to set higher quality standards and vice versa.Thus, quality standards lead to the expected effects when enjoyment and competences are not controlled for (model not reported).
The role of self-rated and perceived partner's competence in context.In accordance with H3b, an actor's self-rated competence leads to significantly higher perceived subjective engagement in housework among both men (|0.5|) and women (0.7) (actor effect: own competence).Moreover, when looking at partner effects, we find that higher self-rated competence of the male partner reduces a women's perception of her housework participation by 0.03 scale points (women's partner effect: own competence); Contrarily, self-rated competent female partners have no statistically significant effect on men's perceptions of their housework share (men's partner effect: own competence).However, for both men and women, an actor's own competence exerts a stronger effect on their perceptions of labor division than partner's competence (difference in actor and partner effect in the women's column: |0.035 + | and in the men's column: |0.036 + |)).Thus, the expectation regarding the relative impact of competences withing the couple expressed in H3cthat men's self-rated competence is more influential than women's self-rated competence in determining a couple's labor divisionis only confirmed with respect to men's perceptions of the labor division (right column).As H3c is only supported with respect to men's perceptions of labor division, we reject H3c.To summarize these results in other words: For both partners own competence is more important than partner's competence for the perception of one's housework share.Thereby, men's evaluation of their involvement in housework does not depend at all on how competent their female partners feel, whereas women's perception of their relative contribution to housework is partially influenced by how competent their male partners feel.Now turning to H4b, which focuses on the effects of how men and women assess their partner's competences, we find that a positive rating of partners' housework competences leads to lower levels of a respondent's own engagement in housework, as perceived by both men (actor effect: partner's competence 0.11) and women (actor effect: partner's competence -0.11).A positive rating of one's partner's competences even has acomparatively speakingsignificantly stronger impact than respondents' own self-rated competencies (men's actor effect, own competence: -0,05; women's actor effect, own competence: 0,07), supporting H4c (difference for women: 0.05**, difference for men: 0.05**; not reported).H4d compares the relative size of coefficients within the couple and expects women's evaluations of men's competences (actor effect, i.e., 'how she perceives him') to have a stronger effect on a couple's labor division than men's evaluations of women's competences (partner effect: 'how he perceives her').This does indeed appear to be true for housework involvement as perceived by women: The absolute value of the actor effect 0.11 is significantly larger than the absolute value of the partner effect 0.7 [difference in actor and partner effect: 0.040*].However, when looking at housework involvement as perceived by men, we find the opposite is also true: For men's subjective perception of the division of domestic labor, a man's evaluation of his female partner's competences (men's actor effect: 0.11) is significantly more influential than the female partner's evaluation of his own competences on his reported contribution to housework (men's partner effect: -0.04) (difference in actor and partner effect: 0.070***).Thus, an actor's own subjective contribution to housework among both men and women is more strongly determined by how the actor perceives his or her partner's competences than how their own competence is perceived by the partner.Thus, H4d is only supported in the part of the model reflecting contribution to housework as perceived by women; we therefore reject H4d.
Interestingly, when comparing the men's and women's parts of the model (i.e., their respective columns), we observe that partner effects are, in sum, more important for women's evaluation of their contribution to housework than for men's evaluation of their contribution.For women's perceptions of housework participation, men's characteristics matter more than vice versa.
Finally, we compared explained variances both before including enjoyment, quality standards, and competences in the model (not reported; R 2 = 0.18) and after including these variables.As shown in Table 2, this leads to a drastic increase in the explained variance, (R 2 = 0.32) suggesting that these variables represent important predictors of the division of domestic labor.

Discussion and Conclusion
Although extensive research confirms the relevance of the dominant economic and normative explanations of gendered domestic labor division, much of the persisting gender gap in housework remains unexplained and puzzling.The purpose of this study was to reveal some of the more subtle mechanisms of gendered labor division by analyzing the contribution of partners' (mutually perceived) enjoyment, competences, and quality standards with respect to housework.We explicitly integrate into the explanatory framework interactional perspectives derived from gender performativity approaches, which emphasize the ongoing accomplishment and production of gender in everyday interactions.Drawing on the 10 th wave of pairfam, we use items on participants' direct ratings of their enjoyment, quality standards, and mutual perceptions of their own and their partner's housework competences.Moreover, we employ actor-partner independence models, which account for the mutual dependency of partners' characteristics regarding the share of housework performed by each partner.
First, as expected, we find significant gender differences in perceived competences, quality standards, and enjoyment with respect to housework, with women expressing, on average, more enjoyment of housework and more stringent quality standards than men.These results support the idea that gender norms and gendered socialization still translate into genderspecific affective, evaluative, and incorporated motives for housework, despite the general change toward more egalitarian attitudes.Second, after applying APIMs which take into account both partners' socioeconomic characteristics and gender role attitudes, as well as confounding demographic variables, the main results are as follows: We find that enjoyment and (perceived) competences exert significant influences on an individual's housework share, and that these variables explain a large amount of variation.Including competences and enjoyment of housework in the model increases explained variance from 18 percent to 32 percent, strongly supporting the relevance of these additional indicators for explaining gender inequality in labor division.Thus, it seems that, beyond the influence of general beliefs, gendered labor division emanates from gender-specific competencies and preferences with regard to housework at a personal level.
With respect to partner effects, intended to capture the contextual and interactional nature of gender performance, we find significant independent influences for an actor's partner's enjoyment and competences on the actor's own rating of their housework share.We particularly find that men's enjoyment exerts a stronger impact on women's perceived contribution to housework than women's own enjoyment.This finding can be explained by the fact that in particular men's deliberate deviation from standard, taken-for-granted gender expectations (e.g. about 'natural' inclinations and preferences for this kind of work) will initiate processes of increasing gender equality and 'undoing gender'.Contrary to our expectations and the results on enjoyment, we find no gender-specific effects with respect to partner's competences and perceptions of partner's competences, when examining the variables separately.For both men and women, an actor's own competence weighs significantly higher than the partner's competence in the subjective evaluation of the division of domestic labor.The same holds true for respondents' ratings of their partners' competences: For both men and women, an actor's assessment of their partner's competences is more important in determining the actor's personal share of housework than, conversely, the partner's evaluation of said actor's competence.Moreover, interestingly, when comparing an actor's self-rated competence to an actor's competence as rated by the partner, our estimates reveal that the partner's perception of the actor's competence has a significantly stronger impact on labor division than an actor's own self-rated competence.
However, gender effects do appear when looking at our competence variables jointly.
It appears that men's competence ratings, both of himself and of his partner, have a stronger effect on his and his partner's perceived contribution to housework than the respective women's ratings.In particular, women's assessments of their housework share are influenced by men's perceptions of competence, both how competent their male partner's feel and by how competent those male partners evaluate their female partner.Considering, in addition, the results on enjoyment, male characteristics tend to exert more and stronger effects on relative participation in housework overall.Given the normative and empirical standard of women doing the majority of the housework, this means that men's traits are more decisive for an equal division of housework than women's.
In a nutshell, our results suggest that women's and men's contributions to housework are shared more equally if and when a male partner expresses a genuine interest in engaging in housework and when he is willing to ascriberelatively speakingmore competence to himself and less to the partner.However, conversely, we also find that women need to be willing to at least accept and acknowledge their male partners' contributions and competences with respect to housework.It thus seems that women's gatekeeping mechanisms are not limited to childcare (Allen & Hawkins, 1999), but also extend to the field of housework.As such, achieving equality in the division of domestic labor can best be thought of as a 'team performance' of undoing gender, with the momentum coming from the male partner.Thereby, the assessment and ascription of competences itself is part of displaying and 'doing' gender, and explicitly involves the ascription of competences to the partner.Despite these novel and insightful results, some limitations of the study have to be noted that could motivate further research.Firstly, although pairfam is a panel study, we could not take advantage of its longitudinal structure because the items on enjoyment, competences and quality standards for housework are only included every two years from wave ten (used here) 3 .Thus, we could not control for time-constant unobserved heterogeneity, but we were able to include the most relevant controls known from the extensive literature on household labor division.Even more importantly, by including enjoyment, quality standards, and competences as variables in our models, we actually explicitly investigate those factors which in most analyses are assumed to be latent and unobserved.Consequently, however, the potential dynamics of these items over time, e.g. over the duration of a relationship, could not be analyzed, and must be for the object of future research.As such, some uncertainty remains regarding the direction of causality, as present involvement in housework might also affect future enjoyment and competences.However, we believe that this problem is more pronounced in approaches which use absolute levels of time spent on housework as the dependent variable (instead of a relative measure, like proportionate share, as here) and that the initial and prevailing labor division is still likely to be driven by differences in partners' (perceived) competences and enjoyment.
Our analysis indicates, ultimately, that further insights might be gleaned from research examining the complex interaction between male and female characteristics, because sociologically valuable information is bound up in mutual assessments.In particular, there is a lack of theory for systematically explaining the variations in partners' assessments of themselves and one another, which warrants further research.

Notes
1 Previous empirical strategies have mostly relied on examining the functional form of the relationship between partners' resources and housework arrangements; however, these approaches suffer from methodological fallacies and provide inconclusive evidence (for a review and discussion see Hook, 2017). 2 Studies have shown that it is difficult for respondents to adequately recall and estimate the average time spent on activities scattered throughout the day like housework on a typical day (Juster, Ono, and Stafford 2003;Kan and Pudney 2007), especially as housework is likely to be performed simultaneously alongside other activities like childcare.3 Wave 12 was recently released, but is not suitable for our study as data collection took place in the midst of the pandemic, which imposed severe restrictions on household decisions regarding housework share and needs to be analyzed separately theoretically and methodologically.

Table 2 :
Actor-Partner Interdependence Models with Share of Housework as Dependent Variable Data are from pairfam, wave 10 (2017/18), authors own calculation.