Gender Roles, Comparative Advantages and the Life Course: The Division of Domestic Labor in Same-Sex and Different-Sex Couples

Abstract

Lesbian and gay couples by definition cannot establish sex-specific divisions of domestic tasks, at least not literally. Previous research has shown that high levels of equality characterize domestic work arrangements in same-sex couples. This study scrutinizes explanations for this. The theoretical background stems from family economics, from the theory of relative resources and from gender role and life-course considerations. The empirical analysis is based upon the Generations and Gender Survey from Austria, Belgium, France, the Netherlands, Norway, Sweden and Australia. Results show that same-sex partners engage in more tasks equally, that their housework-sharing pattern is less segregated and that the partners’ workload is more equally balanced. Comparative advantages, life-course differences and gender roles contribute to the explanation of higher levels of equality in same-sex compared to different-sex relationships.

Introduction

Division of domestic labor has garnered much attention in family research, commonly addressing how women and men share tasks at home, how they specialize in certain household activities (Coltrane 2000) and how couples balance the workload from paid and unpaid work (Sullivan 2000). Previous work has focused on how the demand for housework varies over a couple’s life course and has found that the establishment of domestic work arrangements can best be understood as long-term processes (Baxter et al. 2008, 2010; Grunow et al. 2012). Previous research has highlighted the economic rationale of the division of labor (e.g., Becker 1981), finding its causes in imbalanced bargaining power (e.g., Davies and Carrier 1999) or referring to gender norms as predictors of housework patterns (e.g., Martínez et al. 2010; Poortman and van der Lippe 2009).

This article contributes to our knowledge of how gay and lesbian couples divide domestic tasks compared to different-sex couples and applies general theories in explaining the higher levels of equality in same-sex households. This article’s ambition is to explain the (empirical) differences in domestic work arrangements of same- and different-sex couples and to scrutinize theoretical claims within a subgroup that provides particularly strong test conditions: Although gender norms and role expectations may also possess relevant to same-sex couples (c.f. Oerton 1997), these couples cannot establish sex-specific housework arrangements per se. Same-sex couples may have gendered living arrangements (Biblarz and Savci 2010), but a “natural” starting point and social norms do not exist to guide the initial household arrangement.

This article focuses on three research questions: (1) Do same-sex couples, as theoretically expected, divide their housework more equally than different-sex couples? (2) What characteristics could explain the potentially more egalitarian division of domestic tasks within same-sex couples? (3) Do different- and same-sex couples differ in their responsiveness of housework to these characteristics (equivalence of mechanisms)?

Related Research

The first comparative and in large part quantitative analysis of same-sex and heterosexual couples dates back to the early 1980s when Blumstein and Schwartz (1983) included more than 1700 same-sex relationships in their work on “American Couples.” The authors worked to develop new perspectives on general partnership processes by applying traditional research questions to newly emerging types of couples. With regard to housework, they found that lesbian couples especially strove for total equality, but that gay couples also showed a housework pattern in which both partners engaged actively in domestic tasks. Ten years later, Kurdek (1993) examined the housework strategies employed by gay (N = 95) and lesbian couples (N = 61) and also compared their behavior to different-sex couples (N = 145). His aim was to analyze strategies for assigning roles and tasks based on criteria other than gender (p. 127). He concluded that same-sex couples showed a more balanced housework pattern than heterosexual couples and that their housework was not significantly correlated with education, income or employment. The overall workload from household duties was more equally distributed within same-sex than within different-sex childless married couples. In gay couples, each partner specialized in certain household tasks but neither partner did all the housework (as opposed to a profound percentage of women in different-sex couples). Lesbian couples gained a balanced housework schedule by applying a strategy best characterized as performing activities either jointly or alternately. With another non-random sample of 43 gay men and 36 lesbian women, Kurdek (2007) replicated parts of his earlier study and reached very similar results.

Solomon et al. (2004, 2005) compared the housework of same-sex married respondents (N = 335) to that of their (similarly socialized) different-sex married sisters or brothers (N = 413) and to non-married gay and lesbian couples (N = 238). Married lesbian women reported sharing household tasks more equally than their sisters in traditional marriages. The finding for gay men was similar, as they also took a larger share in housework than their heterosexual brothers. Chan et al. (1998) found that female same-sex parents (N = 30) showed more equality in housework as well as in childcare tasks than do heterosexual parents (N = 16, all couples made use of donor insemination). Exploiting a sample of lesbian couples (N = 26), Patterson (1995) reported that biological mothers were more involved in childcare than their partners, who, in turn, spent more time in the labor market. Adoptive families headed by lesbian (N = 50) and gay (N = 58) couples reported sharing childcare labor more equally than those headed by heterosexual (N = 100) couples (Farr and Patterson 2013). The amount of childcare tasks carried out in same-sex relationships also depends on a parent’s status relative to the other parent: lower education, lower earnings and lower working hours are positively related to involvement in childcare activities (Sutphin 2013). McWhirter and Mattison (1984) found effects of partnership duration: The longer a gay couple had been together, the higher was the probability of clear-cut household duties.

Although previous research has successfully addressed a variety of different theoretical mechanisms, further analyses would profit from more direct test strategies. One should also test whether the mechanisms at work are identical in both same- and different-sex couples because previous research focused on norms of equality that could be dominant in same-sex couples and could suppress alternative effects (e.g., Kurdek 1993, 2004). It is noticeable, finally, that almost no previous study has made use of large-scale surveys based on random sampling and that almost all surveys or qualitative interviews were carried out in North America.

Theory

Previous research has applied different strands of several theories to the division of domestic labor. This section briefly summarizes the main theoretical arguments.

Family Economics

As proposed in the work of Becker (1981), partners’ allocation of time is determined by comparative advantages. The theory implies that “the resources of members of a household […] should be allocated to various activities according to their comparative or relative efficiencies” (Becker 1981: 16). Such advantages result from previous investments in human capital, i.e., educational enrollment, previous labor market participation or the acquisition of specific household skills. Differences in efficiencies are in principle not determined by biological differences (Becker 1981: 24). Thus, family economic models are gender neutral, but a person’s later experiences are in part consequences of parents’ (sex-specific) investment behavior, of intrinsic differences between the sexes (e.g., pregnancy and nursing), and of discriminatory labor markets. As a result, wage rates for women are usually lower and women are more productive in the household. Thus, most heterosexual couples face comparative advantages: The household member with comparative advantages in the household would allocate time to this life domain; the member with higher efficiency in the labor market (i.e., the person with the higher wage rate) would allocate time to employment. Same-sex partners do not necessarily face such clear relative efficiencies because both partners have been socialized similarly. Becker hence argues that same-sex households “are unable to profit from the sexual difference in comparative advantages” (Becker 1981: 22–23). For different- and same-sex couples, however, investments within the partnership can create and strengthen comparative advantages (Browning et al. 2014: 26). The claimed mechanisms should thus also prove valid within same-sex couples, even though they may initially divide labor less efficiently.

Relative Resource Theory

Resource-oriented (bargaining) approaches differ markedly from joint utility models. Blood and Wolfe (1960) argue that, unlike Becker’s joint utility approach, individual preferences might differ from the household rationality when one assumes that both partners perceive household duties as “humdrum and menial in nature” (p. 73). The basic idea therefore is that both partners try to reduce their share of housework and that the more powerful partner can delegate them to the less powerful. Power thus translates into a decrease in a person’s share of housework (Bittman et al. 2003). Relative resource theory makes no gender-specific forecasts and is thus applicable to same- and different-sex partnerships. Although family economics and resource-oriented (bargaining) approaches differ in the suggested mechanisms, predictions from both theories are very similar and hypotheses therefore often do not compete. This is because the variables constituting “power” are often identical to the variables indicating gains from specialization (Iyigun and Walsh 2007; Loving et al. 2004). When education and income are taken into account, both theories predict that the partner with lower earning capacity would do most of the housework (Evertsson and Nermo 2007).

Gender Roles and Doing Gender

Whereas the predictions resulting from the theories discussed so far are independent of gender, a large body of literature has developed normative explanations for the division of paid and unpaid work. Such approaches state that humans act in a certain way either because they are male/female, or because sex is a trait linked to societal expectations. Alternative theories claim that men and women show certain behaviors in order to shape their male/female identity.

The first perspective, that individuals take specific action because of gender, corresponds to theories of social roles. In this model, human action is characterized not by free choices between alternatives but by automatic practices due to norm-compliant behavioral routines (Lindenberg 1985). The second mechanism refers to the “doing-gender approach” (West and Zimmermann 1987). It views gender as a socially constructed achievement and not as a norm anchored in society. Thus, domestic work “may be integral to producing gender; it may be an essential part of being accountably feminine or masculine” (Kroska 2004: 906). Both the role-oriented approach and doing gender are in principle applicable to same-sex relationships: If social norms of housework division affect same-sex relationships, or if these partners need to construct gender through housework, one expects gender to influence the division of domestic labor in gay and lesbian couples as well.

Life-Course Considerations

Life-course considerations focus on the demand side of housework. The establishment of a certain housework arrangement is understood as a long-term process (e.g., Schulz 2010), in which the type and amount of housework vary according to different states within a “family cycle model” (Höpflinger and Charles 1990). It is usually assumed that the demand for housework increases with the presence of children in the household (e.g., South and Spitze 1994; Baxter et al. 2008) and that the division of domestic labor starts out rather undifferentiated and becomes more segregated in later years: “Domestic role sharing is more likely to occur in the early stages of the family life cycle” (Haas 1981: 962). Further, the transition to parenthood is accompanied by more traditional housework arrangements. As outlined above, family economic theory also predicts an increasing differentiation and a decrease in housework sharing over a couple’s life course caused by specific investments within the partnership. Specific biological differences (e.g., breastfeeding) and caring skills are relevant only in households with children.

Hypotheses

What are the expectations as to how same-sex couples differ from households led by two heterosexual partners with respect to the variables relevant in family economics and relative resource theory? It has been shown that same-sex relationships (in the USA) are educationally less homogamous (Schwartz and Graf 2009), but the correlation between partners’ incomes could still be stronger, there being no gender wage gap. Family economics and relative resource theory both predict that relative resources/specialization gains (here: education and income) explain how partners divide housework. If same-sex partners hold more similar amounts of resources, they will share their housework more equally.

With regard to the impact of gender, previous work has emphasized that even same-sex mates make use of gender as an organizing principle. Gay and lesbian partners are thus “not exempt from the social demand for the construction of gendered selves” (Moore 2008: 339). On the other hand, same-sex couples face a “lack of cultural or historical models of same-sex relationships” or of “institutional support and cultural guidelines” (Heaphy et al. 2002: 250). The impact of gender roles should be weaker in same-sex relationships because social expectations (norms) and expected sanctions do not provide a (natural) starting point for the establishment of a gendered division of labor. Furthermore, in cases where same-sex couples make use of queer examples in order to develop understandings of how to organize housework, they might choose the equal task sharing as a point of origin.

Whereas the previous hypotheses refer to the total of all incurred household tasks, I now derive expectations of how same- and different-sex couples allocate specific activities to each partner. If gender defines which partner holds responsibility for certain household tasks, doing one typically male task should strongly predict the engagement of the other partner in a typically female task. Hence, I expect heterosexual couples to organize their housework more closely along the dimensions of typically male/female tasks, even if such tasks are rather simple and require no advanced skills.

With regard to differences in life-course stages, previous work shows that the presence of children leads to less egalitarian, more segregated housework-sharing patterns in heterosexual relationships. The housework-sharing pattern of same-sex couples could thus be more balanced merely because gay and lesbian couples are less likely to live together with children. Controlling for the presence of children in the household should reduce the differences between same- and different-sex couples. The effect of children on the division of domestic labor should be similar across groups.

Furthermore, even without the influence of socially defined gender roles, couples might tend to specialize over time. The availability of same-sex couples offers the possibility of estimating such effects minus the influence of gender roles. Initial gains from specialization are rather low for same-sex couples. This suggests that same-sex cohabiters start sharing housework at a level of high equality, whereas heterosexual relationships establish a more traditional housework-sharing pattern directly upon moving in together. Partnership duration may thus have a stronger effect on segregating household tasks in same-sex relationships.

Data and Methods

Data

Analyses rely on data from seven Generations and Gender Surveys collected in Austria, Belgium, France, the Netherlands, Norway, Sweden and Australia between 2002 and 2013. Data are accessible for scientific use upon registration via http://www.ggp-i.org/. Although the survey covers additional countries, these are not used because there are no same-sex couples in the data (Bulgaria, Estonia, Georgia, Hungary, Italy, Romania and the Russian Federation) or because household tasks are measured on different scales (Germany). One respondent per household provides all information, including information on partner’s characteristics (see for a discussion of the data quality in GGS wave 1 Vergauwen et al. 2015).

Division of Household Tasks

Household tasks covered by the questionnaire differ by country, but all seven surveys included four major activities: “preparing daily meals,” “shopping for food,” “vacuum cleaning the house” and “doing small repairs in and around the house.” Respondents indicated whether those tasks were done always by the respondent (1) or by the partner (5), usually by the respondent (2) or by the partner (4), or equally often by the respondent and his/her partner (3). I follow Kurdek’s approach and calculate three indices, facilitating separate analyses of housework-sharing patterns. All scales range, after standardization, from 0 to 1. The way values have been recoded is outlined in Table 1.

Table 1.

Original and recoded values for the housework-sharing indices

	Categories
	Always respondent	Usually respondent	r and p equally	Usually partner	Always partner
Indices
Original coding	1	2	3	4	5
Equality index	0	0	1	0	0
Segregation index	2	1	0	1	2
Balance index r	3	2	1	1	1
Balance index p	1	1	1	2	3

Equality Index

Equality index counts the number of tasks shared about equally between the partners (e.g., via turn taking or doing tasks together), divided by the number of all tasks incurred. Kurdek’s “equality index” measures the percentage of tasks shared equally.

Segregation Index

The segregation index measures the degree of specialization by counting the number and extensiveness of household tasks usually or always done by either one partner or the other. It takes the value 0 if partners do not specialize (do all tasks together). It takes the value 1 if partners do not share tasks, irrespective of the distribution of tasks to both household members.

Balance Index

Neither the segregation index nor the equality index takes into account in how balanced a manner the partners share the overall workload from unpaid labor. The balance index reaches high values if either a high percentage of tasks are shared or the number and extent of tasks done primarily by one partner are fairly equal to that done primarily by the other. An index capturing each partner’s engagement in household activities first had to be calculated for the respondent and his/her partner separately. Then, the score for the respondent is divided by the score for the partner and standardized to the range from 0 to 1. The resulting balance score takes the value 0 if either partner does all the work. It takes the value 1 if the workload is balanced equally.

Type of Partnership/Sexual Orientation

The GGS questionnaire asks the respondent for the partner’s sex. Respondents living together with partners of the opposite sex constitute heterosexual couples. The other couples are coded gay or lesbian.

Respondent’s and Partner’s Education

Educational measures in the GGS data were provided on the International Standard Classification of Education scale (ISCED). I distinguish between low, medium and high education. Low education captures ISCED 0, 1 and 2, medium ISCED 3 and 4 and high education comprises ISCED 5 and 6.

Couple’s Educational Constellation

If the respondent and partner have the same educational level (low, medium and high, as defined above), the educational constellation is coded as homogamous (educ. homogam); otherwise, it is coded as heterogamous (educ. heterogam). The educational constellation is a measure of gains from specialization as well as of bargaining power.

Household Income

With the exception of Norwegian and Swedish data, the GGS questionnaire asks for the net income from various sources separately for the respondent and for the partner. Data on income refer to gross earnings in Norway as well as Sweden. In those two countries, data for these variables were collected from administrative registers on an annual basis. Here, I have calculated the monthly gross income. In order to account for the different measurements in the seven countries, the calculated monthly household income is centered to the country-specific means and standardized.

Income Ratio

In order to yield a relative measure of income, I calculated the ratio of respondent’s and partner’s monthly income. The minimum income (either respondent’s or partner’s income) has been divided by the maximum income. Where both partners contribute equally to the household income, the variable takes the value 1. It is 0 if only one partner has an income.

Partnership Duration

Partnership duration measures the years that have passed since the partners started cohabiting.

Children in the Household

This dummy variable takes the value 1 if at least one child aged 0–18 lives in the household, regardless of partners’ biological parenthood.

Age and Country

As further control variables, all statistical models control for effects of age (in decades, centered, in linear and squared form) and country.

Results

Descriptive Findings

The differences between gay, lesbian and heterosexual couples with regard to the variables used in this analysis are summarized in Table 2. The table’s upper section displays the sample composition: Although this study uses a large-scale survey and pools data from seven countries, the number of respondents living together with partners of the same-sex is small (N _hetero = 27,415; N _gay = 89; N _lesbian = 93; N _total = 27,597). Thus, only 0.66 % of cohabiting couples in the GGS data are gay or lesbian. Even though statistics on the number of cohabiting same-sex couples are scarce, this group is probably underrepresented in the data. This is a common concern in related research exploiting survey as well as census data (Black et al. 2007). Furthermore, each different-sex couple for which one partner has a sex miscode will appear as a same-sex couple. On the one hand, such miscodes likely increase the number of same-sex couples. On the other hand, such mistakes potentially reduce differences between the groups and thus lead to rather conservative estimates.

Table 2.

Descriptive statistics

Couple type	Hetero		Gay		Lesbian		Total
	N	%	N	%	N	%	N	%
Country
France	4365	98.91	32	0.73	16	0.36	4413	100.00
Netherlands	3062	98.68	18	0.58	23	0.74	3103	100.00
Norway	8074	99.73	12	0.15	10	0.12	8096	100.00
Austria	2548	99.77	3	0.12	3	0.12	2554	100.00
Belgium	2668	98.71	14	0.52	21	0.78	2703	100.00
Australia	2714	99.38	6	0.22	11	0.40	2731	100.00
Sweden	3984	99.67	4	0.10	9	0.23	3997	100.00
Total	27,415	99.34	89	0.32	93	0.34	27,597	100.00
Housework items
Preparing meals
Always r	5262	19.19	12	13.48	13	13.98	5287	19.16
Usually r	6499	23.71	20	22.47	21	22.58	6540	23.70
r/p equally	6294	22.96	25	28.09	38	40.86	6357	23.04
Usually p	6399	23.34	15	16.85	17	18.28	6431	23.30
Always p	2961	10.80	17	19.10	4	4.30	2983	10.81
Total	27,415	100.00	89	100.00	93	100.00	27,597	100.00
Repairs
Always r	6628	24.18	34	38.20	14	15.05	6676	24.19
Usually r	5364	19.57	18	20.22	17	18.28	5399	19.56
r/p equally	4043	14.75	19	21.35	28	30.11	4090	14.82
Usually p	6325	23.07	7	7.87	22	23.66	6354	23.02
Always p	5055	18.44	11	12.36	12	12.90	5078	18.40
Total	27,415	100.00	89	100.00	93	100.00	27,597	100.00
Shopping for food
Always r	4100	14.96	8	8.99	13	13.98	4121	14.93
Usually r	5460	19.92	17	19.10	15	16.13	5492	19.90
r/p equally	11,434	41.71	39	43.82	48	51.61	11,521	41.75
Usually p	4510	16.45	14	15.73	15	16.13	4539	16.45
Always p	1911	6.97	11	12.36	2	2.15	1924	6.97
Total	27,415	100.00	89	100.00	93	100.00	27,597	100.00
Vacuum cleaning
Always r	4682	17.08	10	11.24	11	11.83	4703	17.04
Usually r	6179	22.54	14	15.73	24	25.81	6217	22.53
r/p equally	8599	31.37	35	39.33	38	40.86	8672	31.42
Usually p	5642	20.58	15	16.85	18	19.35	5675	20.56
Always p	2313	8.44	15	16.85	2	2.15	2330	8.44
Total	27,415	100.00	89	100.00	93	100.00	27,597	100.00

Housework indices	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Equality index	0.28	0.25	0.33	0.27	0.41	0.31	0.28	0.25
Segregation index	0.51	0.23	0.50	0.28	0.39	0.25	0.51	0.23
Balance index	0.87	0.19	0.84	0.23	0.87	0.20	0.87	0.19

Education	N	%	N	%	N	%	N	%
Level
Low	5948	21.70	15	16.85	15	16.13	5978	21.66
Medium	12,370	45.12	39	43.82	28	30.11	12,437	45.07
High	9097	33.18	35	39.33	50	53.76	9182	33.27
Homogamy	15,524	56.63	50	56.18	54	58.06	15,628	56.63

	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Income
Income ratio	0.57	0.30	0.66	0.28	0.62	0.29	0.57	0.30
Life course
Age	46.92	14.16	48.33	13.80	43.73	12.92	46.91	14.16
Partnership duration	20.96	14.50	17.46	14.62	12.60	12.39	20.92	14.51

	N	%	N	%	N	%	N	%
Living with children	13,444	49.04	15	16.85	30	32.26	13,489	48.88

Besides the country-specific sample composition, the descriptive Table 2 shows that gay and lesbian respondents more often share the four housework tasks equally with their partners. Thus, the equality index shows higher and the segregation index lower values for same-sex couples. By trend, lesbian couples share their housework more equally than gay couples. Heterosexual couples share on average 28 % of the tasks, gay men 33 % and lesbian women 41 %. With regard to the explanatory variables, same-sex respondents have more often reached higher educational levels than heterosexual respondents. Furthermore, the sexual orientation appears not to correlate with a couple’s educational constellation (homogamy varies only between 56 and 58 %). Both partners’ contributions to the household income, however, are more similar in same-sex couples. In heterosexual couples, the partner with the lower income earns on average 57 % of the income of the partner with the higher earnings. In same-sex households, these values rise to 62 % (lesbian) and 66 % (gay). With regard to household income level (centered and standardized for each country), same-sex couples yield higher incomes than the average heterosexual couple (not given in Table 2). For couples living in the Eurozone, the average monthly net household income is 3120€ for different-sex, 3400€ for gay and 3760€ for lesbian couples.

As expected, the presence of children in the household is most common in different-sex couples (49 %), but a considerable proportion of same-sex couples also take on parental responsibilities. In total, 17 % of gay and 32 % of lesbian couples in the analytic sample live together with at least one child.

Same-sex couples also differ markedly from heterosexual unions with respect to a second life-course characteristic under investigation: The average partnership duration is 21 years for heterosexual couples. Gay and lesbian couples live together, on average, since 17 and 13 years, respectively. This finding is in line with dissolution rates being higher among same-sex than among different-sex couples (Lau 2012), i.e., same-sex couples have more but shorter relationships over the life course. Hence, gay and lesbian respondents are on average older when entering a new relationship (Andersson et al. 2006). The third life-course characteristic taken into account is respondent’s age: It is 47 years for heterosexual, 48 for gay and 44 for lesbian survey participants. Same- and different-sex couples thus differ regarding characteristics theoretically linked to the distribution of household tasks. These couples also differ regarding some of the measures used as further control variables in the following models (educational level, household income, age and country of residence).

Multivariate Findings: Explaining Housework-Sharing Patterns

Figure 1 shows how the three different varieties of couples share housework. The plotted coefficients refer to heterosexual couples as the reference category. I analyze differences on the three housework scales (columns in Fig. 1) with three OLS regression model specifications (rows in Fig. 1): Model 1 does not include any controls and thus displays bivariate differences. Model 2 adds controls for age (linear and squared) and country dummy variables. Specification 3 includes all predictors used in model 2 and adds respondent’s education, educational constellation, household income, income ratio (i.e., partner’s relative income), partnership duration (up to the third polynomial) and children (yes/no). The Appendix 1 comprises the detailed estimation results (Tables 5, 6, 7, 8, 9).

Fig. 1.

Selected regression coefficients (according to Table 5)

Table 5.

OLS regression models producing Fig. 1

	Equality index			Segregation index			Balance index
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Sexual orientation (ref = hetero)
Gay	0.055* (0.027)	0.083** (0.026)	0.050+ (0.026)	−0.012 (0.025)	−0.061* (0.023)	−0.028 (0.023)	−0.032 (0.020)	−0.013 (0.019)	−0.024 (0.019)
Lesbian	0.132*** (0.026)	0.150*** (0.026)	0.117*** (0.025)	−0.120*** (0.024)	−0.139*** (0.023)	−0.104*** (0.023)	−0.000 (0.019)	0.012 (0.019)	0.002 (0.019)
Country (ref = France)
Netherlands		−0.015* (0.006)	−0.013* (0.006)		−0.087*** (0.005)	−0.086*** (0.005)		−0.004 (0.004)	−0.003 (0.004)
Norway		0.096*** (0.005)	0.103*** (0.005)		−0.171*** (0.004)	−0.172*** (0.004)		0.067*** (0.003)	0.066*** (0.003)
Austria		−0.007 (0.006)	−0.002 (0.006)		−0.039*** (0.006)	−0.043*** (0.006)		−0.000 (0.005)	0.000 (0.005)
Belgium		0.000 (0.006)	−0.005 (0.006)		−0.021*** (0.005)	−0.013* (0.005)		0.008+ (0.004)	0.005 (0.004)
Australia		−0.009 (0.006)	−0.013* (0.006)		−0.109*** (0.005)	−0.103*** (0.005)		−0.004 (0.004)	−0.004 (0.004)
Sweden		0.081*** (0.005)	0.078*** (0.005)		−0.205*** (0.005)	−0.193*** (0.005)		0.067*** (0.004)	0.061*** (0.004)
Age		−0.002*** (0.000)	−0.001* (0.000)		0.003*** (0.000)	0.001*** (0.000)		−0.001*** (0.000)	−0.000** (0.000)
Age2		0.000*** (0.000)	0.000** (0.000)		−0.000*** (0.000)	−0.000* (0.000)		0.000*** (0.000)	0.000* (0.000)
Resp.’s education (ref = low)
Medium			−0.010* (0.004)			−0.012*** (0.004)			0.017*** (0.003)
High			−0.005 (0.004)			−0.039*** (0.004)			0.025*** (0.003)
Educ. constellation (ref = homogam)
Heterogam			−0.008** (0.003)			0.006* (0.003)			0.000 (0.002)
Household income			0.003+ (0.002)			−0.006*** (0.002)			0.004** (0.002)
Household income2			−0.000 (0.000)			0.000+ (0.000)			−0.000*** (0.000)
Income ratio (min/max)			0.066*** (0.005)			−0.069*** (0.004)			0.034*** (0.004)
Partnership duration			−0.003*** (0.000)			0.003*** (0.000)			−0.000 (0.000)
Partnership duration2			0.000*** (0.000)			−0.000*** (0.000)			0.000*** (0.000)
Partnership duration3			−0.000* (0.000)			0.000* (0.000)			−0.000** (0.000)
Children in household			−0.039*** (0.004)			0.026*** (0.004)			−0.005 (0.003)
Constant	0.277*** (0.002)	0.221*** (0.004)	0.243*** (0.006)	0.512*** (0.001)	0.627*** (0.004)	0.634*** (0.005)	0.870*** (0.001)	0.836*** (0.003)	0.819*** (0.004)
Observations	27,597	27,597	27,597	27,597	27,597	27,597	27,597	27,597	27,597
R 2	0.001	0.047	0.069	0.001	0.118	0.145	0.000	0.034	0.041

Standard errors in parentheses

⁺ p < .1; * p < .05; ** p < .01; *** p < .001

Table 6.

OLS regression models producing Fig. 2, upper part

	(1) Equality index		(2) segregation index		(3) Balance index
Sexual orientation (ref = hetero)
Gay	0.077*	(0.035)	−0.048	(0.031)	0.006	(0.026)
Lesbian	0.074*	(0.033)	−0.070*	(0.030)	−0.014	(0.025)
Resp.’s education (ref = low)
Medium	−0.010*	(0.004)	−0.012***	(0.004)	0.017***	(0.003)
High	−0.005	(0.004)	−0.039***	(0.004)	0.025***	(0.003)
Educ. constellation (ref = homogam)
Heterogam	−0.008**	(0.003)	0.006*	(0.003)	0.000	(0.002)
Gay × heterogam	−0.063	(0.052)	0.046	(0.046)	−0.069+	(0.039)
Lesbian × heterogam	0.101*	(0.051)	−0.083+	(0.046)	0.038	(0.038)
Country (ref = France)
Netherlands	−0.014*	(0.006)	−0.086***	(0.005)	−0.003	(0.004)
Norway	0.103***	(0.005)	−0.172***	(0.004)	0.066***	(0.003)
Austria	−0.002	(0.006)	−0.043***	(0.006)	0.000	(0.005)
Belgium	−0.005	(0.006)	−0.013*	(0.005)	0.005	(0.004)
Australia	−0.013*	(0.006)	−0.103***	(0.005)	−0.004	(0.004)
Sweden	0.079***	(0.005)	−0.193***	(0.005)	0.061***	(0.004)
Age	−0.001*	(0.000)	0.001***	(0.000)	−0.000**	(0.000)
Age2	0.000**	(0.000)	−0.000*	(0.000)	0.000*	(0.000)
Household income	0.003+	(0.002)	−0.006***	(0.002)	0.004**	(0.002)
Household income2	−0.000	(0.000)	0.000+	(0.000)	−0.000***	(0.000)
Income ratio (min/max)	0.066***	(0.005)	−0.069***	(0.004)	0.034***	(0.004)
Partnership duration	−0.003***	(0.000)	0.003***	(0.000)	−0.000	(0.000)
Partnership duration2	0.000***	(0.000)	−0.000***	(0.000)	0.000***	(0.000)
Partnership duration3	−0.000*	(0.000)	0.000*	(0.000)	−0.000**	(0.000)
Children in household	−0.039***	(0.004)	0.026***	(0.004)	−0.005+	(0.003)
Constant	0.243***	(0.006)	0.634***	(0.005)	0.819***	(0.004)
Observations	27,597		27,597		27,597
R 2	0.069		0.146		0.042

Standard errors in parentheses

⁺ p < .1; * p < .05; ** p < .01; *** p < .001

Table 7.

OLS regression models producing Fig. 2, lower part

	(1) Equality index		(2) Segregation index		(3) Balance index
Sexual orientation (ref = hetero)
Gay	0.047+	(0.027)	−0.026	(0.024)	−0.030	(0.020)
Lesbian	0.115***	(0.026)	−0.105***	(0.023)	0.011	(0.019)
Income ratio (min/max)	0.066***	(0.005)	−0.069***	(0.004)	0.034***	(0.004)
Gay × income ratio	0.034	(0.093)	−0.021	(0.083)	0.069	(0.069)
Lesbian × income ratio	0.041	(0.087)	0.005	(0.077)	−0.193**	(0.065)
Country (ref = France)
Netherlands	−0.013*	(0.006)	−0.086***	(0.005)	−0.003	(0.004)
Norway	0.103***	(0.005)	−0.172***	(0.004)	0.066***	(0.003)
Austria	−0.002	(0.006)	−0.043***	(0.006)	0.000	(0.005)
Belgium	−0.005	(0.006)	−0.013*	(0.005)	0.005	(0.004)
Australia	−0.013*	(0.006)	−0.103***	(0.005)	−0.004	(0.004)
Sweden	0.078***	(0.005)	−0.193***	(0.005)	0.061***	(0.004)
Age	−0.001*	(0.000)	0.001***	(0.000)	−0.000**	(0.000)
Age2	0.000**	(0.000)	−0.000*	(0.000)	0.000*	(0.000)
Resp.’s education (ref = low)
medium	−0.010*	(0.004)	−0.012***	(0.004)	0.017***	(0.003)
High	−0.005	(0.004)	−0.039***	(0.004)	0.025***	(0.003)
Educ. constellation (ref = _homogam)
Heterogam	−0.008**	(0.003)	0.006*	(0.003)	0.000	(0.002)
Household income	0.003+	(0.002)	−0.006***	(0.002)	0.004**	(0.002)
Household income2	−0.000	(0.000)	0.000+	(0.000)	−0.000***	(0.000)
Partnership duration	−0.003***	(0.000)	0.003***	(0.000)	−0.000	(0.000)
Partnership duration2	0.000***	(0.000)	−0.000***	(0.000)	0.000**	(0.000)
Partnership duration3	−0.000*	(0.000)	0.000*	(0.000)	−0.000**	(0.000)
Children in household	−0.039***	(0.004)	0.026***	(0.004)	−0.005+	(0.003)
Constant	0.243***	(0.006)	0.634***	(0.005)	0.819***	(0.004)
Observations	27,597		27,597		27,597
R 2	0.069		0.145		0.042

Standard errors in parentheses

⁺ p < .1; * p < .05; ** p < .01; *** p < .001

Table 8.

OLS regression models producing Fig. 3

	(1) Equality index		(2) Segregation index		(3) Balance index
Sexual orientation (ref = hetero)
Gay	0.031	(0.048)	−0.010	(0.043)	−0.063+	(0.036)
Lesbian	0.069	(0.053)	−0.077	(0.047)	−0.008	(0.040)
Partnership duration	−0.003***	(0.000)	0.003***	(0.000)	−0.000	(0.000)
Partnership duration2	0.000***	(0.000)	−0.000***	(0.000)	0.000**	(0.000)
Partnership duration3	−0.000*	(0.000)	0.000*	(0.000)	−0.000**	(0.000)
Gay × partnership duration	−0.014**	(0.004)	0.016***	(0.004)	−0.004	(0.003)
Gay × partnership duration2	−0.000	(0.000)	0.000	(0.000)	0.000	(0.000)
Gay × partnership duration3	0.000*	(0.000)	−0.000**	(0.000)	0.000	(0.000)
Lesbian × partnership duration	−0.009+	(0.005)	0.006	(0.005)	−0.001	(0.004)
Lesbian × partnership duration2	0.000	(0.000)	0.000	(0.000)	−0.000	(0.000)
Lesbian × partnership duration3	0.000	(0.000)	−0.000	(0.000)	−0.000	(0.000)
Children in household	−0.038***	(0.004)	0.025***	(0.004)	−0.005+	(0.003)
Gay × children	0.037	(0.071)	−0.026	(0.063)	0.116*	(0.053)
Lesbian × children	−0.096+	(0.056)	0.052	(0.049)	−0.017	(0.042)
Country (ref = France)
Netherlands	−0.014*	(0.006)	−0.086***	(0.005)	−0.002	(0.004)
Norway	0.103***	(0.005)	−0.172***	(0.004)	0.066***	(0.003)
Austria	−0.002	(0.006)	−0.043***	(0.006)	0.001	(0.005)
Belgium	−0.005	(0.006)	−0.013*	(0.005)	0.006	(0.004)
Australia	−0.014*	(0.006)	−0.103***	(0.005)	−0.004	(0.004)
Sweden	0.078***	(0.005)	−0.193***	(0.005)	0.061***	(0.004)
Age	−0.001*	(0.000)	0.001***	(0.000)	−0.000**	(0.000)
Age2	0.000**	(0.000)	−0.000*	(0.000)	0.000*	(0.000)
Resp.’s education (ref = low)
Medium	−0.010**	(0.004)	−0.012***	(0.004)	0.017***	(0.003)
High	−0.005	(0.004)	−0.039***	(0.004)	0.025***	(0.003)
Educ. constellation (ref = homogam)
Heterogam	−0.008**	(0.003)	0.006*	(0.003)	−0.000	(0.002)
Household income	0.003+	(0.002)	−0.007***	(0.002)	0.005**	(0.002)
Household income2	−0.000	(0.000)	0.000+	(0.000)	−0.000***	(0.000)
Income ratio (min/max)	0.066***	(0.005)	−0.069***	(0.004)	0.034***	(0.004)
Constant	0.243***	(0.006)	0.634***	(0.005)	0.819***	(0.004)
Observations	27,597		27,597		27,597
R 2	0.069		0.146		0.042

Standard errors in parentheses

⁺ p < .1, * p < .05, ** p < .01, *** p < .001

Table 9.

Ordered logistic regression models, average marginal effects in Tables 3 and 4

	Model 1 on		Model 2 on
	Cooking daily meals		Shopping for food
Sexual orientation (ref = hetero)
Same-sex	0.268	(0.254)	−0.075	(0.214)
Repairs (ref = r/p equally)
Repairs mostly r	1.448***	(0.036)
Repairs mostly p	−0.858***	(0.036)
Same-sex × repairs mostly r	−1.088***	(0.326)
Same-sex × repairs mostly p	0.440	(0.361)
Vacuum cleaning (ref = r/p equally)
Vacuuming mostly r			−1.081***	(0.028)
Vacuuming mostly p			0.926***	(0.030)
Same-sex × vacuuming mostly r			0.262	(0.330)
Same-sex × vacuuming mostly p			0.088	(0.343)
Country (ref = France)
Netherlands	−0.058	(0.049)	−0.042	(0.046)
Norway	0.003	(0.038)	0.039	(0.036)
Austria	−0.215***	(0.052)	−0.011	(0.049)
Belgium	0.012	(0.050)	0.135**	(0.047)
Australia	−0.068	(0.050)	0.019	(0.047)
Sweden	0.007	(0.045)	−0.072+	(0.043)
Age	0.009***	(0.002)	0.012***	(0.002)
Age2	0.000***	(0.000)	−0.000	(0.000)
Resp.’s education (ref = low)
Medium	0.151***	(0.034)	0.065*	(0.031)
High	0.220***	(0.037)	0.018	(0.035)
Educ. constellation (ref = homogam)
Heterogam	0.037	(0.025)	0.026	(0.023)
Household income	0.024	(0.017)	0.032*	(0.016)
Household income2	−0.002+	(0.001)	−0.001	(0.001)
Income ratio (min/max)	0.031	(0.041)	0.006	(0.039)
Partnership duration	−0.001	(0.002)	−0.003	(0.002)
Partnership duration2	0.000	(0.000)	0.000***	(0.000)
Partnership duration3	−0.000**	(0.000)	−0.000**	(0.000)
Children in household	−0.013	(0.034)	−0.032	(0.032)
Constant cut 1	0.073	(0.057)	−0.801***	(0.051)
Constant cut 2	1.307***	(0.057)	1.306***	(0.051)
Observations	27,597		27,597
Pseudo R 2	0.138		0.086

Standard errors in parentheses

⁺ p < .1; * p < .05; ** p < .01; *** p < .001

With regard to the equality index, respondents in heterosexual relationships report an average index value of 0.28. Thus, 28 % of the household tasks under consideration are shared equally, via either turn taking or doing tasks jointly. Gay couples share 33 % (β = 0.06) and lesbian couples share 40 % (β = 0.13). This finding is in line with previous research (Kurdek 1993, 2007), reporting that lesbian couples emphasize equality, especially by performing certain activities jointly. Gay and lesbian couples’ equality scores differ significantly from the value reported by heterosexual respondents, as the 90 percent confidence intervals in Fig. 1 do not include the null value. Throughout the result section, the ten percent significance level is reported due to the small number of same-sex couples (see the Appendix 1 tables for standard errors and p values). Whereas lesbian couples show higher values on the equality score than gay couples, they in turn show lower values on the segregation index. Lesbian couples differ significantly from the reference category. Gay couples also report lower segregation scores, but the effect is insignificant. Focusing on overall housework balance, the two same-sex groups do not differ markedly from the reference category.

Adding control variables (model 2) and theoretically relevant predictors (model 3) increases the models’ explanatory power (equality: R ² from .1 to 4.7 to 6.9 %; segregation: R ² from .1 to 11.8 to 14.5 %; balance: R ² from .00 to 3.4 to 4.1 %). Comparing models 2 and 3, some differences among the three groups diminish and are in part insignificant, but other differences remain almost unchanged. The variables included in specification 3 do not sufficiently explain why sexual orientation and housework patterns correlate. The next section thus aims at testing whether all types of couples react similarly to specialization gains and to life-course characteristics. The following models include interaction effects between selected predictors and the type of partnership. Due to the small number of cases in the same-sex groups, I calculated three separate model specifications with interaction effects of educational constellation (Sect. 5.3), income ratio (Sect. 5.4) and the two life-course characteristics (Sect. 5.5). All other explanatory and control variables are included (main effects) but without interactions with the couple’s sexual orientation.

Educational Homogamy and Heterogamy

Family economic theory and considerations of relative resources predict that educational homogamy leads to more egalitarian housework-sharing patterns, i.e., more tasks are shared equally and/or the overall workload from unpaid work is more equally balanced. The upper section of Fig. 2 shows conditional effects of educational heterogamy (reference: homogamy) on the three indices (columns). For respondents in different-sex relationships, the effect of educational heterogamy is significant but small, reducing the number of shared tasks by about 1 % point. For gay men, the conditional effect is also negative (−7 %) and slightly insignificant, whereas lesbian women increase equality in housework sharing if the partners differ on educational characteristics (+9 %, statistically significant). With regard to segregation, results are similar: The effect is positive (as expected) for different-sex (significant) and gay (insignificant), but not for lesbian partnerships. On the overall balance score, educational heterogamy has a reducing effect only within gay households (significant at the 10 percent level). Replicating the models with a reduced set of covariates (i.e., not controlling for income and relative income) does not change the results markedly.

Fig. 2.

Conditional effects of educational heterogamy and income ratio (according to Tables 6, 7)

Partners’ Relative Income (Income Ratio)

The partners’ relative income provides a more direct measure of relative resources and specialization gains than the educational constellation. Although this is an advantage, one should keep in mind that a reduced income can also be the consequence and not the cause of doing housework. The lower section of Fig. 2 shows conditional correlations of the income ratio: Higher values (i.e., partners contribute to the household income more equally) correlate positively with the equality index. Effect sizes are similar for all three groups. Couples in which both partners contribute equal amounts to the household income share tasks 6–10 % more equally than couples in which only one partner has an income. With respect to the segregation index, results are again similar. Effect sizes do not differ by sexual orientation; the insignificance of effects for same-sex couples results from the rather small number of cases. On overall balance, the effect of an equal contribution to the household income is positive for different-sex (significant) and gay couples (insignificant). In lesbian partnerships, the effect is negative and thus does not support the relative resource hypothesis.

Life-Course Characteristics (Presence of Children and Partnership Duration)

The descriptive findings have shown that gay and lesbian couples’ average partnership duration is shorter than that for heterosexual partners. Also, the proportion of partners living without children is higher in same-sex relationships. The regression model includes partnership duration as a first, second and third polynomial. The nonlinear profiles, conditional on sexual orientation, are shown in Fig. 3: Couples cohabiting for a longer time span report lower equality and higher segregation scores, indicating that partners specialize over the course of their relationship. The effect on balance is clearly negative in lesbian couples. Averaging effect sizes of the polynomials to a single coefficient (AME) permits analysis of the statistical significances (middle section of Fig. 3) and enables visual comparisons between effect sizes (Bauer 2015). Overall, the results support family economic theory’s predictions of specialization gains arising over the course of a relationship and that all types of couples would rationally make use of specialization in their household.

Fig. 3.

Conditional effects of life-course characteristics (according to Table 8)

Living together with at least one child under the age of 18 (biological parenthood is known only for parts of the sample and could thus not be considered) correlates with lower equality and higher segregation values in both different-sex and lesbian partnerships. In gay couples, children have no effect on equality and segregation. This hints at possible differences between gay fathers and lesbian mothers: Whereas gay couples may more often adopt a child or live together with a child from a previous heterosexual relationship, lesbian mothers may make use of donor insemination. The latter results in biological motherhood and in the possibility of breastfeeding, which, in family economic terms, would be considered a relative advantage for the biological mother.

Gender Norms and Gendered Housework Skills

The previous analyses took into account how a household’s partners differ on characteristics relevant to the labor market (education and resultant income). Hence, the analyses considered gains from specialization that should lead to divisions of paid and unpaid labor. Family economic theory also states, however, that housework skills matter and that gender-specific abilities predict which partner will engage in which activity in the household. The final analyses will hence provide a more detailed perspective on correlations among the four distinct household tasks. Because same-sex couples rarely leave a task completely to either partner, the categories “always” and “mostly” are collapsed, which leads to a scale with 3 instead of 5 values. Otherwise, ordered logistic regression models could not be properly estimated. Summarizing gay and lesbian couples to same-sex couples simplifies the interpretation of results. Separate analysis with either gay or lesbian couples lead to similar results. These are briefly discussed below and available upon request.

Because I do not have any measures for partners’ skill levels in the household sector, this analysis has to rely on plausibility assumptions: I assume that, on average, women have better skills in “preparing daily meals” than men, and that men, in turn, are more experienced in “repairs in and around the house.” While sex-specific advantages in cooking and repairs result from long-term training (e.g., sex-specific courses in school, training by parents), “shopping for food” and “vacuum cleaning the house” demand lower skill levels. The latter activities are rather simple and nothing obvious precludes women and men managing such tasks with equal success. Hence, I assume comparative advantages to exist with regard to repairs and cooking in heterosexual couples, but not with regard to vacuuming and shopping. In homosexual couples, partners’ relative efficiencies should—on average—not differ with regard to any of the four tasks. As the analyses aim to disentangle effects of relative efficiencies from effects caused by norm-compliant behavior, I also explicate assumptions referring to the prevalence of gender norms. I assume that daily cooking, shopping for food and vacuum cleaning are typically female tasks. That is, there are societal expectations that women take responsibility for these activities. In turn, society expects men to engage in small repairs. Within same-sex households, corresponding societal expectations do not exist. Further descriptive analyses show that men in different-sex relationships engage more frequently in tasks requiring lower skill levels (vacuuming and shopping for food) than in tasks requiring skills typically obtained as a girl (preparing meals). Men predominantly carry out repairs. In same-sex couples, the two comparatively simple tasks (vacuuming and shopping) are more often carried out together or alternately than are the more complex (cooking and repairs).

My analytic strategy is to scrutinize the relationship between tasks under three conditions: (a) comparative advantages and gender roles simultaneously contribute to the correlation (negative correlation between repairs and cooking in different-sex couples), (b) comparative advantages do not exist and correlations should be due only to gender norms (positive correlation between shopping and vacuuming in different-sex partnerships) and (c) neither comparative advantages nor gender norms make significant differences (assumed on average for same-sex couples). Tables 3 and 4 display the results in the form of average marginal effects (AME), estimated from ordered logistic regression models (Table 9), controlling for all covariates as in the previous analyses. Table 3 shows the results for condition a, and Table 4 shows for condition b. Both tables provide information for same-sex couples under condition c.

Table 3.

Marginal effects of repairs on preparing meals (according to Table 9, model 1)

Change in probability that meals are prepared		If repairs in and around the house are done
		Mostly by r (%)	Equally	Mostly by p (%)
Mostly by r	Different-sex	−29.7***	Ref.	20.5***
	Same-sex	−8.3	Ref.	10.3
Equally	Different-sex	−3.7***	Ref.	−8.0***
	Same-sex	0.2	Ref.	−2.3
Mostly by p	Different-sex	33.4***	Ref.	−12.5***
	Same-sex	8.0***	Ref.	−8.0

Table 4.

Marginal effects of vacuum cleaning on shopping for food (according to Table 9, model 2)

Change in probability that shopping for food is done		If vacuum cleaning the house is done
		Mostly by r (%)	Equally	Mostly by p (%)
Mostly by r	Different-sex	25.5***	Ref.	−15.1***
	Same-sex	19.3***	Ref.	−16.8**
Equally	Different-sex	−11.8***	Ref.	−4.8***
	Same-sex	−8.6***	Ref.	−4.7
Mostly by p	Different-sex	−13.8***	Ref.	19.9***
	Same-sex	−10.8***	Ref.	21.5**

The tables’ rows contain the categories of the two variables used as dependent variables (preparing meals in Table 3 and shopping in Table 4) in the first column. The third, fourth and fifth columns show the categories of the variables used as independent variables (repairs in Table 3 and vacuuming in Table 4) in the first line. The AME can be interpreted as the change in the predicted probability for a specific category of the dependent variable conditional on the values of the independent variables and is measured in percentage points (%-p.). The significance tests refer to the reported reference category. In Table 3, the probability that a heterosexual respondent mostly prepares the meals is reduced by 29.7 %-p. if the respondent is responsible for repairs (row 2, column 3) compared to a situation in which repairs are equally shared between the partners. If the partner carries out repairs in the house, this increases the probability that the respondent holds responsibility for preparing meals by 20.5 %-p. compared to the reference category (row 2, column 5). For same-sex couples, these effects are smaller: Repairs by the respondent reduce the probability of his/her engagement in cooking only by 8.3 %-p. and repairs by the partner increase the probability that the respondent holds responsibility for meals only by 10.3 %-p. Compared to the interval [−29.7 %-p.; 20.5 %-p.] in different-sex couples, the effect (span of the interval) is 31.6 %-p. weaker [−8.3 %-p., 10.3 %-p.]. The reverse condition (row 4, columns 3 and 5) leads to very similar results. The difference in effect size is |(−12.5 to 33.4) − (−8.0 to 8.0)| = 29.9 %-p. I regard the average difference of (31.6 + 29.9)/2 = 30.8 %-p. in Table 3 as a composite effect of gains from specialization and gender norms. Both are presumably present in heterosexual but not in homosexual couples when analyzing cooking and repairs. This composite effect of norms and specialization gains appears to be considerably large. Alternative models with either gay or lesbian couples in comparison with heterosexual couples (not shown) lead to almost identical results. Gay and lesbian couples differ from each other, on average, by 1 %-p. only.

In Table 4, shopping and vacuuming are positively related: In general, the partner who shops for food also cleans the floor. The probability that shopping is done mostly by the respondent increases by 25.5 %-p. if the respondent vacuums at home compared to equal sharing of vacuum cleaning. If the partner always or usually vacuums, the respondent is −15.1 %-p. less likely to be primarily responsible for buying groceries. The effect size for different-sex couples is |(−15.1 to 25.5)| = 40.6 %-p. in row 2 and |(−13.8 to 19.9)| = 33.7 %-p. in row 4. For same-sex couples, effect sizes are estimated at |(−16.8 to 19.3)| = 36.1 %-p. in row 2 and |(−10.8 to 21.5)| = 32.3 %-p. in row 4. The differences in effect sizes between different- and same-sex couples are only 40.6–36.1 = 4.5 %-p. (line 2) and 33.7–32.3 = 1.4 %-p. (line 4), respectively. The average effect size in Table 4 is (4.5 + 1.4)/2 = 3 %-p. Separate models with either gay or lesbian couples (not shown) reveal some differences between the two same-sex groups. Although lesbian couples appear to differ from heterosexual couples by almost 16.4 %-p., the corresponding interaction effects in the regression model are insignificant.

Taken together, these analyses indicate that specialization gains and gender norms lead to differences between same-sex couples’ and heterosexual couples’ housework patterns. It appears that gender norms matter for different-sex couples, but that in themselves—as in Table 4—they do not lead to considerable different correlations. When both gains and norms are at work (Table 3), the difference in the effect takes an average value of 30.8 %-p. (composite effect). In cases to gender norms alone should have relevance (Table 4), the reported differences are much smaller and take an average value of 3 %-p. One could thus speculate that effects caused by specialization gains are stronger than effects caused by gender norms. The latter interpretation, however, requires assuming that the impact of gender norms is equally strong in cases in which specialization gains are relevant (repairs and cooking) and in those in which such gains do not exist (cleaning and shopping for food). The interpretation also requires the assumption that repairs and preparing meals are skill intensive tasks, stating that cooking usually consists of more than popping a frozen meal into the microwave.

Further analyses with one skill intensive and one simple task (not shown) lead to average differences between heterosexual and same-sex couples that are larger than in the scenario of two simple but smaller than in the scenario of two difficult tasks. In analyses with one difficult and one simple task, same-sex and different-sex couples differ from each other by 25 %-p. (vacuuming and repairs), 23 %-p. (vacuuming and cooking), 11 %-p. (shopping for food and repairs) and 7 %-p. (shopping for food and cooking). The latter difference is rather small, probably because the two activities are closely related. Shopping for groceries needed in the kitchen appears to be efficiently combinable with the preparation of meals, regardless of a couple’s sexual orientation.

Summary and Discussion

This article aimed to compare the distribution of housework in same- and different-sex partnerships. In line with previous research on same-sex couples’ organization of daily domestic tasks, I find that gay and lesbian partners apply a pattern characterized by higher levels of task sharing and lower levels of segregation. This study has exploited social theories aiming to provide general explanations of why partners divide domestic tasks rather than engaging in all tasks equally. With regard to the educational constellation, I find heterogamy relates to a more segregated and less balanced division of housework in gay and in different-sex living arrangements. The income ratio, an alternative variable indicating differences in partners’ financial resources, shows the expected correlations: The more equally partners contribute to household income, the higher the number of shared tasks, the lower the obtained segregation score, and the higher the overall balance (although the latter does not pertain for lesbian couples). Of course, the respondent’s and the partner’s income might be not causes but consequences of a specific division of labor. Because educational decisions are often made before a partnership begins, endogeneity is less problematic in reference to educational heterogamy as an indicator of comparative advantages in the labor market. Differences between same- and different-sex couples’ housework patterns decrease but do not disappear when the models control for relative resources.

My analyses further scrutinized effects of co-residing children and partnership duration. In lesbian and in heterosexual partnerships, children traditionalize the domestic work pattern: fewer tasks are carried out jointly and one partner becomes more involved in housework than the other. Partnership duration is found to have negative effects on equality and balance and a positive effect on segregation. With regard to all three dependent variables, the effect is strongest in lesbian couples. This finding supports predictions from family economics. While comparative advantages for same-sex couples, especially for couples with two women trained in housework, might be small in the beginning, specific investments can strengthen specialization gains in the long run (Becker 1981: 22–23).

Resources, specialization gains and life-course stages together contribute to reducing differences between same- and different-sex couples’ housework-sharing patterns. Because the differences do not fully disappear, one could assume that unobserved societal expectations lead to a more “traditional” division of housework in different-sex partnerships. The analysis of correlations among four household tasks aimed at disentangling effects of skills and norms. Correlations in same-sex couples differ from those observed in heterosexual couples. The differences are rather small in the scenario in which comparative advantages (skill level differentials) between women and men do not exist and rather large in the scenario in which the male and the female partners’ (assumed) efficiencies diverge. Previous research has argued that the capacity to form and maintain a household is an important dimension of gender inequality. Most studies have thus focused on employment characteristics as indicators of gender inequality with regard to this capacity (Neyer et al. 2013: 264). This article’s finding that skills in the household play an important role in the division of domestic work has so far not been garnered much attention. Gendered socialization experiences may thus not only translate into abstract role models and societal expectations but also into tangible skills within the domestic sphere. These skills constitute comparative advantages within different-sex couples. Same-sex couples do not face such clear comparative advantages at the beginning of their relationship but may specialize in certain tasks in the long run.

Although this article contributes to explaining the division of housework, some shortcomings bear discussing. One major concern refers to the identification of same-sex couples in the GGS data. Firstly, it is likely that the number of same-sex cohabiters in the population is higher than in the analytic sample. Unit and item non-response might cause selectivity in a way that only gay and lesbian couples with certain characteristics, e.g., living openly as same-sex couple, take part in the survey. Secondly, some units identified as same-sex couples might in fact be different-sex couples with sex miscodes for one of the partners. Though the number of such coding errors is unknown, it is likely not very large because same- and different-sex couples differ in their housework-sharing patterns in a way, which is similarly reported in previous research. Without coding errors, the differences among couple types might thus be even larger.

A next shortcoming refers to the measuring the division of domestic tasks. The number of available housework tasks is rather small and diaries could provide more accurate information than the scales used in the GGS. As the amount of time spent for certain tasks was unknown, I implicitly assumed each task to require an equal amount of time and excluded respondents outsourcing housework. In addition, pooling data from different countries collected in different survey years leads to cultural heterogeneity. For instance, the availability of gay and lesbian marriage or alternative civil unions and of the possibility of adopting children differs by both time and country. Country-fixed effects partly absorb cultural heterogeneity, but data restrictions preclude explicitly controlling for these differences in multilevel models. These could be used for estimating country-specific effects if the number of same-sex couples was larger. Even though gender roles, cultural norms and the legal system could vary by geographic region, the countries in the analytic sample share similar historical developments with regard to a predominance of the male breadwinner model before 1970, followed by women’s liberation movement which resulted in a more egalitarian division of labor both at home and on the labor market (Weziak-Bialowolska 2015). The countries studied in this article also share characteristics with regard to family and divorce laws which affect the allocation of time and resources within households (Reinhold et al. 2013; Kneip et al. 2014). Besides the country of residence, other factors such as religion and life styles may also be sources of cultural heterogeneity for which the analyses in this article do not account.

Various robustness checks have thus been conducted: Excluding countries does not change results systematically, nor does excluding influential observations (multidimensional outliers) or couples with respondents aged 65 or older. Further, applying different multiple imputation techniques instead of using a sample with non-missing information leads to almost identical results. Because standard software does not estimate average marginal effects with multiple imputations, the analyses presented here base on a sample of couples with complete information.

The discussed shortcomings offer opportunities for further research. The analysis of lesbian and gay couples appeals because it offers opportunities to test general theories. Thus, the aim of this article was not (primarily) to provide information about a rather small, in the data probably underrepresented group of people. The analysis of same-sex couples provides hard test conditions under which it is likely to falsify family economic predictions because one can separate comparative advantages from correlated societal expectations. Further research would require not only dyadic, but also longitudinal data to address appropriately the dynamic shifts in housework sharing over the couples’ life courses and reveal causal mechanisms instead of pure correlations.

Appendix 1

See Tables 5, 6, 7, 8 and 9

Compliance with Ethical Standards

Conflict of interest

The author declares that he has no conflict of interest.