The Allocation of Housework in Same- and Different-Sex Partnerships: Recent Evidence from the U.S

Research on the division of housework among same-sex partners is limited. This is because gender-cultural theories – which emphasize the significance of gender identity and motivate many studies on the topic – are implicitly assumed to be less relevant in this case. Attending to admonitions that the division of housework in same-sex households is not free from gendering processes and practices, in this study we use the high-quality data of the American Time Use Survey (ATUS; 2003-2019), to compare the association between housework and relative earnings across partnership types. Since gender-cultural theories are based on the performance of gender identity by different-sex partners, we utilize the differences between same- and different-sex partners to better understand the effect of gender-cultural determinants on the division of housework. Our comparison of the relation across partnership-types validates the power of gender-cultural mechanisms in different-sex partners, provides a better assessment of the differences in housework patterns between different types of households, and serves as a novel quantitative test of gender-cultural mechanisms in same-sex partners.

The abundance of research examining the unequal division of paid and unpaid work between spouses focuses almost entirely on different-sex couples.A dominant stream in this extensive literature tests the explanatory power of economic theories versus gender-cultural theories (for review, see Lachance-Grzela & Bouchard, 2010) Economic theories suggest a spousal tradeoff in the allocation of unpaid and paid work between the spouses, so that the partner with the higher labor market resources invests in paid work, while the other invests more in unpaid work.Whereas the resulting allocation is obviously gendered, the focus of these theories is on the economic efficiency inherent in the exchange between paid and unpaid work.Gender-cultural theories, in contrast, point explicitly at gendering processes as significant determinants of housework allocation.According to this view, the exchange of paid work and housework is conducted within a socially gendered context in which women and men are "doing gender" by differentially investing in housework and paid work, respectively.In this regard, both paid work and unpaid work serve to confirm the assigned gender roles and reinforce the gendered identities of women and men in different-sex households.
The vast literature on the subject provides evidence in support of both theories (e.g., Bianchi et al., 2012;Hook, 2017;Kan, 2008a;Procher et al., 2018).However, whereas the evidence in support of gender-cultural theories is plentiful and robust in different-sex partners, direct evidence to these theories in same-sex partners is missing.This is due, at least partly, to the implicit assumption in this body of research that the allocation of housework between same-sex partners is "gender empty" (Blumstein & Schwartz, 1983;Rawsthorne & Costello, 2010), and that therefore, if there is a division of labor between them, it must result from an economic tradeoff between paid and unpaid work (Bauer, 2016).This assumption, however, has been challenged by scholars arguing that same-sex partners "[are] not exempt from the social demand for the construction of gendered selves" (Moore, 2008, p. 339), and that they too construct a feminine or masculine Amit Lazarus and Hadas Mandel have equally contributed in this work.
identity via housework performance (Oerton, 1997).That said, we found no comparative quantitative empirical examination of gender-cultural theories in same-sex partnerships.
In this study, we aim to fill this lacuna by quantitatively testing the explanatory power of gender-cultural theories also in same-sex partners.We do this by using models designed to be sensitive to the predictions of both theories, as explained below.We use the American Time Use Survey (ATUS), collected between 2003 and 2019.This dataset has several advantages over other data sources (like the ISSP or PSID), as it is based on time diary estimates of housework (see Kan, 2008b), and is the only U.S. data that differentiates between routine, time-inflexible housework tasks (like cooking and cleaning), and the less burdensome, non-routine tasks (like yard work and vehicle maintenance) (Coltrane, 2000(Coltrane, , p. 1210-11)-11).Given their culturally prescribed femininity, routine housework tasks present a more obvious test of the gender-cultural perspective.Therefore, in our comparisons across partnership types we use both a composite measure that includes both routine and non-routine tasks, and a more specific measure that includes only routine housework.Our analysis provides a novel and direct test of gender-cultural mechanisms in same-sex partners, a better assessment of the differences in housework allocation between different types of households, and an opportunity to reexamine the power of gender-cultural determinants in different-sex partners.

The Division of Housework: Economic Versus Gender-Cultural Theories
Research on the gendered division of housework has developed several contending theories to explain the inter-partner gaps in housework based on evidence from different-sex partners.The most prevalent theories are premised on an economic rationale.It may be the attempt to maximize the household's economic potential, or, conversely, the individual's desire to avoid household labor, but all economic based models suggest that the partner with the higher relative labor market resources (usually men) will devote more time to paid work, whereas the partner with fewer advantages in the market (usually women) will devote more time to household tasks (Becker, 1985;Kan, 2008a;Lachance-Grzela & Bouchard, 2010).
Three main mechanisms are highlighted in this theoretical corpus.The first mechanism, 'specialization', emphasizes a tradeoff between spouses in paid and unpaid work as part of their effort to maximize the efficiency of their joint household unit.The second mechanism is the 'relative resources' that underscore a negotiation or exchange between selfserving partners in which each attempts to capitalize on his/ her resources to avoid housework (Blood & Wolfe, 1960;Brines, 1994).The third is 'time availability' that emphasizes available time (i.e., time not in paid labor) as a 'relative resource', accordingly the more hours a partner spends in the labor market the less time s/he will spend on housework (Bianchi et al., 2000;Cunningham, 2007).
The three mechanisms align with and complement one another.In different-sex couples, the lower overall earning potential of women in the labor market makes it more efficient for them to do housework and for men to invest more of their time in the labor market (Becker, 1981).This dynamic is strengthened over time; the longer both partners play these roles, the more they specialize in them and the more costly it becomes to defect from them (Browning et al., 2014, p. 67-69).This dynamic maintains and stiffens the differing relative resources and time allocation of the partners.Under this situation, in different-sex partnerships, the lower labor market resources of women lower their bargaining power and thus their ability to exchange paid work with housework.
Although the specific mechanism differs between the economic theories, an economic rationale that involves a 'spousal tradeoff' between paid and unpaid work is common across them, where more paid work of one partner is related to less unpaid work of the other.In principle under these theories, both men and women can be either providers or dependents, contingent on their personal abilities and comparative advantages in the labor market.Individuals who work for pay for more hours, be they men or women, are more specialized in paid labor, less economically dependent on their partners, and have more time constraints.In practice, the unequal gender allocation of housework is explained via men's relative advantage in the labor market that "grants them the right" to do less housework.
In sharp contrast to this reasoning, gender-cultural theories put gender at the forefront of their explanation for the division of housework.According to these theories, social expectations regarding gendered roles in society ('gender roles'), as well as men and women's continuous construction of their gender identity ('doing gender'), determine the amount and type of housework carried out by each partner.Since the performance of housework (especially routine tasks) is socially constructed as "feminine" and of paid work as "masculine," such behaviors, when engaged by women and men respectively, either mirror the implementation of gender roles or result from men and women "doing gender."The latter refers to behaviors and activities that confirm and reinforce the gendered identities of both partners (West & Zimmerman, 1987).
In different-sex partnerships, both "gender roles" and "doing gender" can help explain the recurring and wellestablished finding that women do most of the housework, even if the amount of time they invest in paid work and their relative status or economic contribution is the same as, or even higher than, their partner's (Fuwa, 2004;Mandel et al., 2020;Miller, 2020).The notion of "gender deviance neutralization" (Bittman et al., 2003) is a specific case of the "doing gender" dynamic.According to this idea, in different-sex partnerships, when one partner "violates" his/her preassigned gender role in one sphere, s/he will compensate for this violation by intensifying a gender-appropriate behavior in another sphere (Brines, 1994;Schneider, 2012;Sullivan, 2011).Thus, a woman who is the sole or main breadwinner might compensate for this "deviation" from "appropriate" gender roles by increasing the amount of housework she undertakes.Concomitantly, her partner, might reduce his share of housework to "preserve" his precarious manhood (Brines, 1994), or display his masculine gender identity (Brines, 1994;Greenstein, 2000;Schneider, 2011).These findings serve as evidence for the explanatory power of the gender-cultural perspective (e.g., Mandel et al., 2020).
However, in contrast to the plentiful evidence in support of gender-cultural and economic theories (e.g., Aassve et al., 2014;Bianchi et al., 2012;Procher et al., 2018;Sabbah-Karkabi, 2021), the debate regarding the specific notion of gender deviance neutralization is on-going, despite assertions it should be concluded (Sullivan, 2011).The debate is fueled by a myriad of empirical findings that often contradict one another.This is partly due to variation in social contexts (i.e., different countries or periods), which has driven cross country and time series studies (e.g., Fuwa, 2004;Mandel & Lazarus, 2021).For example, evidence supporting gender deviance neutralization was found in some more ideologically conservative countries but not in others (e.g., Bittman et al., 2003;Mandel et al., 2020).However, even within the same cultural context the conclusions are inconsistent.For example, within the U.S., Brines (1993Brines ( , 1994) ) and Greenstein (2000) observed gender deviance neutralization only among U.S men, while others found evidence for it only among women (Evertsson & Nermo, 2004;Schneider, 2011Schneider, , 2012;;and see Procher et al., 2018 for the case of Germany).Others have refuted gender deviance neutralization as an explanation for the unequal division of housework in the U.S altogether, arguing that the findings suffer from methodological misspecifications or biases (e.g., Gupta, 2007;Gupta & Ash, 2008;Killewald & Gough, 2010).

The Doing of Gender in Same-Sex Partners
In contrast to the abundance of research on the division of labor and housework in different-sex partners, studies on same-sex partners are scarce.This scarcity may be due to the still persistent heteronormative assumptions regarding family structure that lead to a dearth of large-scale data on a more diverse range of families (Goldberg, 2013;Oerton, 1997;Sutphin, 2010).From the limited research that has examined the division of paid work and house-work among same-sex partners, earlier studies proposed that such partnerships are "gender empty" -i.e., free from gendering processes and practices (e.g., Blumstein & Schwartz, 1983;Kurdek, 1993) -and thus surmised that gender-cultural theories are less relevant for understanding these partnership types.
Quantitative comparisons between different-and samesex partnerships indeed show that the latter have a more balanced and equal division of both paid work and housework (e.g., Hofmarcher & Plug, 2022; for a review, see Goldberg, 2013).This more equitable divide of spheres is attributed, first and foremost, to the similarity of sex within the couple, which is expected to reduce differences in the behaviors that are linked to gender identity and social expectations associated with sex and gender roles (Goldberg et al., 2012).Same-sex partners also face more akin labor market opportunities and encounter similar gender-based experiences (e.g., discrimination, social tracking), which results in greater homogeny in labor market-related resources, creating also a more equal division of housework (Bauer, 2016).Overall, these greater similarities between the two partners may mitigate the economic and social pressure to specialize in a certain sphere and thus contribute to a more equal division of labor in same-sex partnerships (Bauer, 2016).
However, several scholars have challenged the conclusion that the more equal division of housework in same-sex couples is due to an absence of gendering processes in role assignment, claiming that this is an "over-simplified" conclusion (Moore, 2008;Oerton, 1997).Rather, these scholars argue that gendering processes can and do play a role in same-sex partnerships, as they too construct feminine or masculine identities via gender role differentiation denoting the allocation of paid and unpaid work in same-sex partners as "gender-full" (Oerton, 1997).These studies emphasize occurrences of unequal division of labor among same-sex partners and its intensification as their relationships grow older, especially if they become parents (Bauer, 2016;Goldberg et al., 2012;Hofmarcher & Plug, 2022).Research focusing on lesbian couples, for example, has found that after childbirth lesbian mothers tend to specialize and divide work, as biological mothers become prone to doing more unpaid work and less paid work.Explanations for this finding have pointed to social-structures -like maternity leave and social expectations that enhance specialization -and to biological-physiological needs that push the biological mother to take up more unpaid work (Downing & Goldberg, 2011;Gabb, 2005;Goldberg & Perry-Jenkins, 2007;Patterson, 1995).Gay partners also become less egalitarian as the relationship matures or when they become parents (Sutphin, 2010).In their case, however, the underpinning of the explanations offered are solely economic.For example, the increase in housework tasks due to the presence of children boosts the economic advantages of specialization (Browning et al., 2014;Carrington, 1999).
The literature discussed above stresses the significance of gender identity as a potential mechanism determining the division of labor also in same-sex partnerships.However, evidence for gender identity is not direct, emphasizing the unequal division of labor and the tendency to specialize also among same-sex partners, but such specialization does not stand in contrast to economic based theories.
In general, quantitative evidence for the validity of the economic theory within same-sex partners is not uniform.On the one hand, Hofmarcher and Plug (2022) and Sutphin (2010) support the economic rational showing that samesex respondents with higher earnings than their partner undertake a lower percentage of housework tasks or less housework time.Similarly, Bauer (2016) reported that in all three partnership types -gay, lesbian, and heterosexual -a more equal income contribution of the partners is related to a greater number of tasks shared equally between the partners (see also Solomon et al., 2005 andGoldberg et al., 2012 when examining "routine" housework).Other findings, in contrast, challenge the validity of economic theories in same-sex partnerships.For example, Civettini (2015) found no effect of income proportion on the division of both routine and non-routine housework in same-sex households, after considering the time spent in paid work.Similarly, Kurdek (1993) found only a weak relationship between personal power (educational level, employment, and income) and housework investment in same-sex households.Finally, examining "non-routine tasks," both Solomon et al. (2005) and Goldberg et al. (2012) did not find any support for the claim that a higher gap in "relative income" increases inequality in housework division.

Overview of Current Study and Hypotheses
Same-sex partnerships tend to be more egalitarian in their allocation of housework than different-sex partnerships (Goldberg, 2013).Whether this more equitable division of housework relates to an absence of gender-cultural processes has yet to be answered.In this study, we therefore test the possibility that gender does affect the allocation of housework also in same-sex partnerships -a hypothesis that has not yet been directly tested in quantitative research.
To test this prediction, we embrace a common analytical approach used in research on the division of housework among different-sex partners (e.g., Bittman et al., 2003;Brines, 1994;Mandel et al., 2020).In this body of research, the relationship between housework and paid work is examined under two opposing assumptions.First, under the economic based assumption, more housework is expected to be associated with less relative economic contribution to the household, and vice versa.Therefore, a negative and linear relationship between housework and paid work is expected.
Second, under the gender-cultural based assumption, doing (or not doing) housework serves to perform and reassert the gender identity of each partner.In this case, not only that women are expected to do more housework than men even when they are the main (or sole) providers (sex role), but these women do more housework relative to women who earn equal to their partners, an indicator of gender deviance neutralization.Thus, a curvilinear relationship between housework and relative economic contribution is expected.

Hypothesis 1
The association between housework and relative economic contribution is expected to be negative and linear, supporting an economic theory of housework allocation.Since the curvilinear relationship between housework and relative economic contribution supports gender-cultural processes, in the absence of these processes the relation is expected to be linear.

Hypothesis 2a
The association between housework and relative economic contribution is expected to be curvilinear.

Hypothesis 2b
If gender-cultural processes affect the allocation of housework the curvilinear relation between relative economic contribution and housework is expected to be especially evident when examining routine housework tasks.This is because routine tasks are more clearly culturally perceived as feminine tasks, and thus present a more obvious test of the gender-cultural theory.

Dataset and Participants
We use the integrated extract builder (ATUS-X) of the American Time Use Survey database (ATUS) (Hofferth et al., 2020), pooling all available years (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019).The ATUS is the first ongoing federally administered survey on time use in the United States.Its sample is based on a subset of households participating in the eighth rotation of the Current Population Survey (CPS).One household member, aged 15 or older, is randomly chosen 2 to 5 months after the CPS interview to fill a time usage log of all daily activities, including housework, done in the past 24 hours (see more in BLS, 2020).The information regarding the partner of the respondent's earning was obtained from the CPS survey.Our analytical sample included 73,777 married or cohabiting respondents aged 25 or older, who either themselves or their partners reported a positive earning.Since earning data of the self-employed is not available in ATUS, our sample includes only salaried or wage workers.
In the case of same-sex partners, sample size is always of concern given that survey designs of large scale representative national samples rarely account for their relatively small population size by over-sampling, as often done with other minority groups.Thus, most previous quantitative studies of same-sex partners use no more than a few dozen cases (e.g., Bauer, 2016;Kurdek, 2007) or non-representative samples (e.g., Solomon et al., 2005).In this respect, to the best of our knowleage, ATUS is the best available option (see also Hofmarcher & Plug, 2022), enabling us to utilize a representative sample of US population, and reach an adequately generalizable sample of 283 lesbian and 227 gay respondents (Hair et al., 2019).The proportion of same-sex (married or cohabiting) partners relative to different-sex partners in our sample (0.7%) is lower than the one estimated using the American Community Survey (see: https:// www.census.gov/ data/ tables/ time-series/ demo/ same-sex-coupl es/ ssc-housechara cteri stics.html).This is mostly due to a lower proportion in ATUS sample itself, rather than to our analytical sample selection.

Dependent Variable: Housework Time
Our dependent variable, housework time, is the sum of minutes per day (top coded at the 99.5 percentile of 660 daily minutes) that the respondent from each household reported, including both routine ("female-typed") and non-routine ("male-typed") housework.More specifically, we include eight aggregate activity categories, of which the first three are routine tasks and the remaining five are non-routine tasks (see also Hook, 2017): 1) general housework (i.e., cleaning, laundry, storing, and maintaining textiles); 2) food and drink preparation, presentation, and cleanup; 3) grocery shopping; 4) interior maintenance; 5) exterior maintenance; 6) lawn, garden, and houseplants care; 7) vehicle maintenance; 8) appliances, tools and toys set-up and maintenance.

Independent Variables
We define two focal independent variables: the respondent's economic contribution and type of partnership/sexual orientation (hereinafter "partnership type").The latter categorizes respondents into four groups: 1) men with a woman partner (different-sex men); 2) women with a man partner (differentsex women); 3) men with a man partner (same-sex men); 4) women with a woman partner (same-sex women).
We compute the former using the following equation: The numerator is the gap between the partners' usual weekly earnings, including usual overtime pay, and the denominator is the sum of both partners' weekly earnings.The index ranges from 0, which means that the partner is the sole breadwinner, crosses 1 where both partners contribute equally to the household's income, and ends at 2 where the respondent is the sole breadwinner.All earnings were top coded at a value of $2884.61 by the U.S. Census Bureau and adjusted for inflation using 2019 as the base rate.While earnings data of the respondent are surveyed as part of the ATUS questionnaire, earnings data of the partner are not.Therefore, as done in previous research using this dataset (e.g., Hook, 2017;Schneider, 2011), we draw the partners' earnings from the CPS interview conducted 2 to 5 months prior to the ATUS survey.

Control Variables
We included possible confounder variables that might be related to the two focal independent variables and housework, but cannot be considered part of the mechanism of influence.These include year of survey using a series of 16 dummy variables (0 = 2003), a dummy for whether the time use data refers to a weekday (Monday through Friday = 0) or a weekend (Saturday or Sunday = 1), and a dummy for a national holiday (holiday = 1), as in holidays and weekends people tend to do more housework.We also included race/ ethnicity as defined by five categories: White (the reference category), Black, Hispanic, Asian-Pacific Islander or Native American, and mixed ethnicity, and a control for age of the respondent, ranging between 25 and 80 (top coded by the Census Bureau in some of the years), and age 2 (as housework is expected to increase up to a point and then decrease again; see Batalova & Cohen, 2002).
To the augmented model we also added variables that may succeed gender or sexual orientation, and thus partnership type or economic contribution.These variables include two household composition variables that are important determinants of housework -whether the respondent is married or cohabitating (cohabitating respondents coded as 1) and the number of children under 18 in the household (top coded at 6).We also included educational status measures: First, educational attainment which is measured using the highest degree out of five levels: no degree (the reference category), high school diploma or GED, associate degree, bachelor's degree, and a master's degree or higher.Second, a dummy variable that differentiates between respondents who are currently enrolled in part-or full-time programs in high school or in a college/university (enrolled = 1).
Finally, we added family income as a proxy for the ability to outsource housework and household size, which might create differences in housework time between partnership types and economic contribution levels.It is measured as annual monetary income of all household members 15 or older, from all sources as reported in the CPS survey.For 5% of cases the measure of family income was not available.In these cases, we therefore imputed it by multiplying the partners' weekly earnings by 52 weeks.Using only complete cases hardly changes our results (available

Analytical Approach
To examine the effect of economic contribution on housework time, we use a multivariable OLS model, controlling for possible confounding variables to eliminate possible alternative mechanisms.A quadric term of relative economic contribution is also added to the equation to examine the possibility that the relationship between housework time and economic contribution is curvilinear, which, as discussed above, is hypothesized by gender-cultural theories.Interactions of partnership type with both the coefficient of economic contribution squared, and the coefficient of economic contribution, are added to all models to estimate whether and how their effect varies by partnership type.We then use these models to graph predictions of housework time by economic contribution level and partnership type to demonstrate the effects in all groups, once after controlling for possible confounders, and once after including all control variables (as detailed above).The risk of overfitting our model due to the smaller samples of same-sex respondents prevents us from using highly flexible methods (e.g., linear or cubic splines).However, we do test the robustness of our analysis by inserting our economic contribution variable into the model using three dummies, segmenting it into four quartiles, rather than modeling it using a quadratic coefficient as done in our main models.Predictions attained using this model are similar to those attained with the quadratic specification (see Table S1 in the online supplement).

Descriptive Differences in the Division of Labor among Partnership Types
We start our analysis by presenting the distribution of our two key variables -economic contribution and housework time -within each of the four partnership type groups, after dividing them into quartiles.Figures 1a and 1b present the former and the latter, respectively.Figure 1a clearly supports previous findings that same-sex partnerships are more egalitarian than different-sex partnerships in their economic contribution to the household (e.g., Bauer, 2016;Solomon et al., 2005).Indeed, while around 50% of different-sex men and women are highly economically in/dependent on their partners (i.e., contribute less than a quarter or more than three quarters of their joined income), the economic allocation among same-sex men and women is closer to equal, as only about a third (37% and 28%, respectively) reported that they are highly economically in/dependent on their partners.While the pattern for different-sex women is a mirror image of the pattern for different-sex men, the patterns for both same-sex men and women are remarkably similar.The same is true for housework distribution: about two thirds of all women in different-sex partnerships do more than an hour of housework a day compared to only about a third of men in different-sex partnerships.Put differently, in different-sex partnerships, the odds of women preforming housework more than one hour per day are three times higher than men (p < .01).Again, while the pattern for men and women in different-sex partnerships are nearly mirror images, the similarity between all partnership types, except women in different-sex partnerships, is striking.Indeed, while differences in the relative risk of being in any one of the housework quartiles between all partnership types, except women in different-sex partnerships, are not significant, the differences between women in different-sex partnerships and the other groups are significant in all cases (p < .05;results are available upon request).For example, we can see that more than a third of women in differentsex partnerships (34%) devote more than 2.7 hours a day to housework compared to less than a fifth (17% or less) of the respondents in the other three groups.It also follows that while the distribution of housework is very different between women in same-and different-sex partnerships, it is very similar between men in same-and different-sex partnerships.This similarity in outcome, however, may stem from different causes.While men in same-sex partnerships tend to have a lower housework load because they have less dependent children in the household, and because they tend to outsource housework (Bauer, 2016;Goldberg, 2013), men in different-sex partnerships do less housework because the women do most of it (Sayer, 2016).

Comparing the relation between economic contribution and housework
In this section we explore and compare the relationship between housework and economic contribution within the different partnership types using multivariable modeling.We start with a depiction of the relation in all groups with the potential confounding variables included as controls (Table 2, Model 1).Models with and without confounders produce very similar results, and therefore we present only the former.Such similarity means that the controls included in Model 1 have very little effect on the relation between economic dependency and housework within partnership types.We then examined the same relation after controlling for variables that could also be considered as mechanisms through which partnership type or economic dependency affect housework: family structure, education, and total family income (Table 2, Model 2 As noted above, we add our key independent variable -economic contribution -to the equation as a quadratic polynomial: the respondent's relative economic contribution and its square.The two coefficients of the polynomial are allowed to interact with partnership type.To simplify the comparison and demonstrate the relationship within all four groups, Fig. 2  As can be seen in Model 1, the control variables act as expected.We find that people tend to do more housework during holidays and weekends.Respondents who are older in age (up to around age 62) or Latino report doing more housework, and respondents who are Black report doing less.
As for the effect of economic contribution depicted in Fig. 2a, the most obvious distinction is between women in different-sex partnerships and the three other partnership types.Except for women in different-sex partnerships, the association between economic contribution and housework is very similar in all three groups, with no significant differences observed between the groups in the two coefficients of economic contribution.Furthermore, for the most part, the association follows the prediction of economic theories of a tradeoff between housework and economic contribution: economically dependent respondents do around twice as much housework than breadwinner respondents, supporting Hypothesis 1 in these groups.Further, this association is especially linear in the two same-sex groups.In these groups the negative correlation persists throughout the entire range of the economic contribution distribution, as expected when the relation is close to linear.These findings further strengthen the support of Hypothesis 1 in same-sex partnerships groups.In the case of men in different-sex partnerships, the relation is negative but then decelerates more substantially than in the case of same-sex partnerships, and finally flattens out in households in which the men respondents earn around 93% of the earnings or more.This finding is unexpected under both hypotheses.
In contrast to the other three groups, the association between housework and economic contribution for women in different-sex partnerships is strongly and clearly curvilinear, supporting Hypothesis 2a.That is, in the segment of the distribution in which women earn less than their partners, the correlation between economic contribution and housework is strongly negative, in accord with an economic tradeoff between partners.However, the strong quadratic effect, which is highly significant (p < .01),causes the negative relation between the variables to diminish, plateauing when women earn around 70% of the joined earnings, and then reverses to a strong positive relation, supporting the argument of gender-cultural theories.
To concretize this finding, while the gap in time spent on housework per day between men and women in differentsex partnerships in which both contribute equally to the family earnings (index = 1) is about 24 minutes (men = 81 minutes, women = 105 minutes), this gap is twice as large (grows to around 50 minutes) in different-sex partnerships where women or men are the sole breadwinners (index = 2).That is, being the sole breadwinners, men, as expected, do less housework than men who contribute less to their household (only 67 minutes), but women who are the sole breadwinners do even more (118 minutes) than women who contribute equally to the partners' earning (105 minutes, as noted above).Contrarily, women in samesex partnerships tend to devote around 76 minutes per day when their contribution is equal to their partners' and less (around 60 minutes) when they are the sole breadwinners, very similar to men in different-sex partnerships.
To further test the explanatory power of economic versus gender-cultural theories, Model 2 in Table 2 provides the results after controlling for the possibility that variation between economic dependency levels, and between partnership types, involves additional factors.Specifically, we added family structure variables (cohabitation and number of children), education, and total family income to the model.As expected, higher education, family income and cohabitation were associated with less housework, while a higher number of children was associated with more housework.In results that are available upon request, we also added the family structure variables separately from education and total family income and find that while the latter variables have almost no effect on the predictions, the predictions of the two same-sex groups increase a bit after family structure variables are inserted to the equations, indicating that members of same-sex partnerships are doing less housework because they tend to have less children and more cohabitation.This increase, however, is evenly spread throughout the economic contribution scale and thus the relation remains roughly linear.
The most important finding in Model 2 is that cross-group variation in the association between economic contribution and housework hardly changes, and that the minor differences that did occur brought the association among the three groups, excluding women in different-sex partnerships, to be more similar and closer to linearity.This can clearly be seen in Fig. 2b, which presents the predictions of housework time by economic contribution level, after accounting for all controls.
The association between housework and economic contribution is still very much curvilinear among women in different-sex partnerships (quadratic coefficient is significant (p < .01),with a positive slope for women that earn around 70% of the economic contribution or more, and, to a much lesser extent, also among men in different-sex partnerships (quadratic coefficient is significant (p < .01),but in this case the negative relation persists throughout, with a slightly negative slope for sole breadwinner men.In same-sex partnerships the predicted negative association is even closer to linear, with a more substantial negative slope also in the case of sole breadwinners.Additionally, the squared coefficients of the two groups are not significant (supporting Hypothesis 1).Furthermore, differences between men and women in the two same-sex groups in the coefficient of economic contribution and its square are also not significant.To further illustrate, breadwinner partners in same-sex partnerships do half (or less) of the housework done by their fully dependent partners.In the case of different-sex partnerships, breadwinner men do around an hour of housework, while breadwinner women do around double that amount.

Comparing the Relation between Economic Contribution and Routine Housework
To further understand the findings above, we narrow our focus to housework tasks that are more burdensome and routine.These tasks account for most of the overall housework load and have been empirically shown to be predominantly performed by women in different-sex partnerships and are culturally perceived as more feminine.Based on the rationale of "gender display," in different-sex partnerships, routine housework are the tasks through which women "do gender" whereas men "do gender" by avoiding routine housework.Therefore, in different-sex households we expect women to do more and men to do much less of these housework tasks, especially when they are the sole breadwinners.
Assuming that the gender determinant is less relevant in the case of same-sex partnerships, we expect the linear correlation to persist in their case.In other words, we expect the differences between same and different-sex households to be even more evident.Following Hook (2017), we define cleaning, laundry, maintaining and repairing textiles, food and drink preparation and clean up, and grocery shopping as routine tasks.Since the inclusion of "grocery shopping" as a "routine task" is not as widely agreed upon (e.g., Schneider, 2011), as a robustness check we excluded it from the measure and re-estimated all models.Results (available upon request) are very similar to the findings presented in Fig. 3.
Model 3 and 4 in Table 2 display the results when the dependent variable is routine housework, the former controlling for confounders and the latter including all controls.Figures 3a, b  in Models 3 and 4 are set to their mean when calculating predictions charted in Fig. 3.When we compare the prediction estimates in Fig. 3 to the ones presented in Fig. 2, we can see those tasks pertaining to routine housework account for most of the total time spent on housework.Furthermore, this is especially true for women in different-sex partnerships.Taking Models 1 and 3 as an example, we can see that routine housework accounts for around 90% (182 minutes out of a total 202 minutes a day) of the time devoted to housework by women in different-sex partnerships that are fully economically dependent on their spouse (index = 0).For fully economically dependent men in different-sex partnerships, on the other hand, routine housework tasks account for only about 55% of the total time invested in housework (75 minutes out of a total 134 minutes a day).
Moving to a cross-group comparison, we first note that the exclusion of non-routine ("male-typed") tasks from the measure of housework strengthen the picture showed before -as predicated by Hypothesis 2b.When focusing on feminine housework, differences in the pattern of the association between the same and the different sex groups become greater.Among men and women in different-sex partnerships, the greater differences between the genders found in Model 4 (compered to Model 2) indicate that gender identity indeed explains the curvilinear association.Men and women in same-sex partnerships, however, report the same amount of routine housework and the association is very close to linearity for both genders, supporting Hypothesis 1.That is, among same-sex partnerships, through the entire distribution of economic contribution, the larger the economic contribution of one partner, the less routine housework he/she tends to do, as predicted by economic theories.

Discussion
In the present study, we compare the explanatory power of economic and gender-cultural theories in different-sex and same-sex partnerships.To our knowledge this is the first study to quantitatively examine gender-cultural theories in samesex partnerships.This examination provides a more thorough comparison of differences in housework allocation between partnership types, and -no less important -revalidate the power of gender-cultural theories in explaining housework allocation in different-sex partnerships.
Our findings reveal large differences between same and different-sex partnerships.Specifically, members of samesex partnerships not only divide paid and unpaid work more equally (as also found by others -e.g., Bauer, 2016), but even when paid work is not equally divided, their housework division is aligned with economic theory.As can be seen in Fig. 1a and b, both economic contribution to the household and the number of hours each partner devotes to housework are much more equally divided between the partners in same-sex partnerships.Furthermore, as shown in Table 2 (and Figs. 2 and 3), in same-sex partnerships the negative relationship between housework and paid work is very similar (with no statistically significant differences between the two groups) and almost perfectly linear: the more a partner contributes to the household economically, the less he/she devotes to housework.This holds true also when examining only "feminine" routine housework.
In contrast, the relationship between housework and paid work in different-sex partnerships is more complex and follows economic theory only in part (see also Bittman et al., 2003;Mandel et al., 2020).As shown in Figure 1, different-sex partners tend to divide labor and housework along gender lines with women taking on more of the family housework and men more of the family income.Furthermore, among women in different-sex partnerships that adhere to the normative gender role (i.e., earn less than their partner (71% of women, as seen in Fig. 1a)), and even among women that earn slightly more than the men, economic theory prevails: the more they contribute to the household's economy, the less housework they do.However, when the partners violate their gender roles, i.e., when women are the main breadwinners (about 13% of differentsex partnerships), women still do a substantial amount of housework more than predicted under economic theory, and they do even more housework when they are the sole breadwinners compared to woman contributing less relative earning.According to the idea of gender-deviance neutralization, this pattern results from the need of these women to compensate for their "deviation" from prescribed gender roles and is consistent with other research focusing on U.S women, as well as women in other countries (e.g., Mandel et al., 2020;Procher et al., 2018;Schneider, 2011).
The relationship between economic contribution and housework performance among different-sex men is closer to linearity than among different-sex women, but at a certain point on the economic contribution distribution it becomes less negative.This means that men who account for most of the partners' earnings in different-sex households do more housework than expected under gender-cultural theories, countering notions of "precarious manhood" or gender-deviance neutralization (e.g., Brines, 1994).It also counters the linear association expected under economic theory.Based on this pattern, and on the linear form of the association when looking at routine housework, we deduce that some male-typed tasks, like gardening or car maintenance (and a smaller portion of routine tasks), are 'reserved' for sole breadwinner men.That is, while according to gender-deviance neutralization economically dependent men may preserve their manhood by avoiding housework, our findings show that sole breadwinner men may preserve their manhood by doing housework.

Limitations and Future Research Directions
While the ATUS is the only time diary source that we know of with large enough samples to enable a reliable quantitative analysis of housework of same-sex partners, one of its drawbacks is that it attains time diary information only from a single respondent in each household sampled.This means that we cannot run a dyadic analysis that examines differences in housework within partners, which could account for possible household level differences between partnership types or economic dependency levels that might affect housework time.We aimed to mitigate this problem by controlling for household and individual level variables, but it is possible we did not account for all such differences.If relevant dual or multiple respondent time diary data that is derived from a representative sample and includes a large enough sample of same-sex respondents becomes available in the future, examining this issue while using a dyadic approach should be of interest to further improve estimation.These data would enable researchers to add a relative measure of housework to the absolute measures used here, and in effect, compare gaps in housework between partners across partnership types.
The amount of housework outsourced is an important variable when examining the allocation of housework.While our control for family income from all sources partly capture the ability to outsource, it is not a direct measure of this concept.If ability to outsource varies across partnership types or economic contribution levels even after controlling for family income, it may explain differences in housework predictions seen in our results.An analysis exploring the role of outsourcing in same-sex partnerships, as seen in the research literature on different-sex partnerships (e.g., Gupta 2007;Brandon, 1999), could be valuable in future research on the subject if suitable data including such measures is compiled.
Although our findings for same-sex partnerships coincide with past research supporting economic theory as an explanation for the division of housework (e.g., Bauer, 2016;Sutphin, 2010), they do not refute the argument that same-sex partners "do gender" as well (Oerton, 1997).This is because the evidence that housework in same-sex partnerships is negatively related to the economic contribution only supports the economic trade-off dynamic over the gender-cultural dynamic in same-sex partners.It is possible that among same-sex partners the allocation of labor per se is not free of gendering processes, and that the doing-gender dynamic in same sex partners takes on different appearances and nuances that cannot be revealed with our quantitative method of analysis.Thus, complimentary qualitative research is needed to provide a deeper analysis of those dynamics.

Practice Implications
Our findings stress the power of gender-cultural mechanisms by demonstrating that women in different-sex partnerships devotes significantly more of her time to housework compared to women in same-sex partnerships or men in either type of partnership.The comparison between women in same-and different-sex couples highlight not only the importance of gender identity in determining the allocation of household's tasks, but also the inability of economic theories to predict housework among either highly dependent or independent women in different sex couples.The gender identity of differentsex women, especially in these households, may cause them to increase their housework production, relative to all other respondents.
It follows then that family-work policies which aim to reconcile women's paid and unpaid work, and drive women towards more committed and lucrative positions in the labor market, may mitigate some of the gender gap in housework, but only up to a certain point.In order to eradicate gender gaps in different-sex households beyond this point, cultural change in gender role perceptions is necessary -one that will disentangle womanhood and manhood from housework and/or widen the range of what is considered appropriate for women and men within the home.This would enable women to do less housework tasks, and perhaps pave the way for men to do more.Additionally, it is important to advance policies that strengthen men's commitment to the household, such as the "use it or lose it" parental leave policy, aimed at increasing men's attachment to the home, and not only supporting women's attachment to the labor market.

Conclusion
The findings presented in this research indicate that in same-sex households the allocation of labor between the two spheres is more equally divided than in different-sex households.Furthermore, when same-sex partners do diverge in economic contribution to the household, an economic tradeoff between housework and paid work ensues so the more one partner contributes to the earnings relative to her/his partner, the less housework he/she tends to do.This pattern of labor allocation stands in stark contrast to the curvilinear relations between economic contribution and housework observed for women in different-sex partnerships.Women living in different-sex households are more likely to contribute less economically to the household, and in such situations do substantially more housework than any other respondent.Furthermore, increasing the involvement of women in different-sex partnerships in paid labor, beyond a certain point, only serves to further increase their involvement in housework, and thus the gender gap in unpaid work.The differences between women in different partnership types found in this study, which could not have been uncovered in a comparison based on linear models alone, revalidates the explanatory power of gender identity in different-sex partnerships.Gender processes continue to determine the experience of women in the home, showing that the exchange between paid work and housework is conducted within a socially gendered context, even in the liberal context of the U.S., with more women than ever fully participating in paid labor.

Fig. 1
Fig. 1 (a) Distribution of Economic Contribution Quartiles by Partnership Type.(b) Distribution of Housework Time Quartiles by Partnership Type displays predictions of housework time by economic contribution level and partnership type based on Model 1 (Panel A), and Model 2 (Panel B).Both Fig. 2a and b present predictions that are conditional on economic contribution level and partnership type, while fixing all control variables in each model to their grand means.

Fig. 2
Fig. 2 Prediction of Minutes Spent on All Housework, by Economic Contribution, Partnership Type, and Model Specification Fig. 3 Prediction of Minutes Spent on Routine Housework, by Economic Contribution, Partnership Type, and Model Specificationf

Table 1
Means and Distributions of All Variables Used in the Analysis, by Partnership Type Table1displays the means or distributions of all variables included in the analyses and sample sizes by partnership type.

Table 2
OLS Regressions on Minutes of Housework Performed a Day Note.**p < .05;***p < .01;Square brackets indicate reference category if variable is categorical, or range if variable is continuous; Weighted using probability and replication weights; Age, family income, and number of children are mean centered