This is an open access article distributed under the terms of the Creative Commons
Attribution License (Women in mixed-gender (i.e., heterosexual) couples do more household labor (including housework and childcare) compared to men in the U.S. and other Western nations, regardless of labor-market work, income, available time, gender ideology, skill, and preference (see multinational surveys by Davis, Greenstein, & Marks, 2007; and a U.S. regional survey by Erickson, 2005). Studies within this literature variously focus 1) exclusively on maintenance of one’s home and welfare of the entire family (e.g., cooking, cleaning, laundry, repairs); 2) mostly on the home and family, but including care of children (e.g., feeding, dressing, playing with); and 3) exclusively on childcare. Most of what we review falls within category (1), whereas our study falls within (2). For completeness, we label any studies that are within (2) or (3) accordingly; all other studies are within (1). Men’s contributions to household labor have risen in recent years, but only slightly. In 2016, a nationally representative survey found U.S. men to perform 36% of household chores and childcare (Livingston & Parker, 2019). In comparison, estimates reported 10–15 years prior in a U.S. national sample (Bianchi, Milkie, Sayer, & Robinson, 2000), a Swedish national sample (Evertsson & Nermo, 2007), and a collection of 28 nations including the U.S., Australia, and several European and Latin American countries (Davis et al., 2007) were that men did 30–35% of the housework and women 65–70% of it. Eliminating inequitable and stereotypical task allocations based on gender (and other social categories) requires as comprehensive an understanding of the inequitable phenomenon as possible and, therefore, lends importance and urgency to studies such as the present one. We contribute to this quest by applying ambivalent-sexism theory (Glick & Fiske, 1996, 1997, 2001) in novel and unique ways to understand gender disparities in housework.
Literature on Predicting Men’s and Women’s Housework Contributions
Much research regarding mixed-gender spouses’ or partners’ division of household labor examines factors possibly increasing or decreasing gendered allocations of housework. These include gender ideology/attitudes, perceived fairness of housework allocations, and marital quality (Cunningham, 2005, in a U.S. regional survey; Lavee & Katz, 2002, in Israeli villages and towns). According to gender-ideology theorizing, women and men with more egalitarian attitudes will more equally divide household labor than those with more traditional attitudes (Cunningham, 2005, 2007; Erickson, 2005). Most findings support this proposition for men, but the association between women’s gender-role attitudes and gendered behavior (i.e., time spent doing household labor) is usually weak (Cunningham, 2005, 2007; Erickson, 2005).
The present research aims to 1) improve upon the currently weak associations in the literature between gender-role attitudes and couples’ housework division; and 2) do so by assessing gender-related attitudes through an ambivalent sexism perspective (Glick & Fiske, 1996, 1997, 2001). Sexism measures previously emphasized only hostile sexism (Eagly & Mladinic, 1989, 1994; Glick & Fiske, 1996, 1997). Hostile sexism refers to hostility and negative stereotypes toward women, beliefs that men are more competent than women and deserving of higher status and power, and support for traditional women’s roles (Glick & Fiske, 1996). Solely focusing on hostile sexism can be misleading, however, as it misses a second form of sexism known as benevolent sexism. Benevolent sexism refers to seemingly positive attitudes toward women that nevertheless convey an underlying paternalism, patronization, and view of women as subordinate. Benevolent sexism also characterizes women in terms of purity, decency, and nurturance, which can diminish their sense of competence (Dardenne, Dumont, & Bollier, 2007) and convey they belong in the domestic sphere (Glick & Fiske, 1996) rather than in occupations requiring toughness, competitiveness, and/or physical stamina. Hostile and benevolent sexism are continuous dimensions, conceptually separate from each other but correlated empirically (from .37 to .74 in Glick & Fiske, 1996).
Differentiating hostile and benevolent sexism should further our understanding of how attitudes affect behaviors in gender-related contexts. In particular, various lines of argument exist for why the combination (i.e., statistical interaction) of high benevolent sexism and high hostile sexism should more strongly predict men’s and women’s housework allocation than either alone. First, men who subscribe to benevolent-sexist attitudes toward women would seem likely to urge women toward traditionally feminine roles and occupations. If some benevolent-sexist men also held hostile-sexist attitudes (which question women’s competence), this combination could intensify men’s efforts to restrict women to traditional roles, including housework. Second, Hammond, Overall, and Cross (2016) argued that men’s hostile sexism alone would not likely promote women’s traditional role performance because it would likely “foster women’s resentment of men’s societal power” (p. 215). Hence, men would need also to exhibit benevolent sexism to “temper this hostility… by characterizing women who adopt traditional caregiving roles as warm, wonderful, and in need of men’s protection” (p. 215). Likewise, women who internalize benevolent sexism toward their own gender would likely accept traditional roles, with this tendency possibly intensified if these women also internalized hostile sexism. One source of women’s internalized benevolent sexism—subtle modeling of this attitude by women’s male partners in mixed-sex relationships—can help us understand why benevolent sexism toward women may function similarly in women as in men (Hammond et al., 2016; samples from New Zealand, the U.S., and Canada). Further, part of channeling a group toward some roles involves directing its members away from others. Moya, Glick, Expósito, De Lemus, and Hart (2007) found, in experiments at Spanish universities, that women’s endorsement of benevolent sexism was positively associated with acceptance of role restrictions couched in protective terms. For example, when told that they should not drive long distances because it could be physically draining or that they should decline a law-school internship because it required interaction with criminal convicts, women high in benevolent sexism (relative to those low) more readily accepted the restrictions (Moya et al., 2007). Conceivably, then, internalization of benevolent sexism could divert women from non-traditional occupations (or even labor-force employment altogether) and toward greater household responsibilities.
Only two studies, to our knowledge, have investigated direct links from hostile and benevolent sexism to gendered housework allocation. Gaunt and Pinho (2018), studying British mothers of young children (≤ age 6), found mothers’ hostile (but not benevolent) sexism toward women to predict greater childcare work by the women than by co-resident biological fathers (i.e., husband or partner). This result was mediated by mothers’ gatekeeping, a construct reflecting disparagement of men’s caretaking, maternal identity linked to others’ evaluation of one’s home and children, and a sense of gendered domains (e.g., that only women enjoy maintaining the home). del Prado Silván-Ferrero and López (2007) investigated associations of ambivalent sexism with housework in Spanish high school students. Tasks were divided via exploratory factor analysis into highly gender-typed (e.g., vacuuming, dusting, cleaning bathrooms) and sex-neutral (e.g., cleaning one’s room, setting and clearing the dinner table). For our purposes, the main finding was a positive correlation in girls between benevolent sexism toward women and performance of gender-typed housework. (del Prado Silván-Ferrero & López, 2007 and Gaunt & Pinho, 2018, included measures of hostility and benevolence toward men, as well as toward women. We did not measure hostility and benevolence toward men, as we considered these beyond our scope.)
Chen, Fiske, and Lee (2009) developed a measure of gender-role ideology in marriage, one item of which assesses extent of respondents’ agreement that women not doing housework is irresponsible. In Chinese and U.S. university students, Chen et al. found the gendered-housework subscale to correlate positively with hostile (in Chinese men and U.S. men and women) and benevolent sexism (in U.S. women). That one study (Gaunt & Pinho, 2018) found hostile sexism (indirectly) to predict childcare work, but another (del Prado Silván-Ferrero & López, 2007) found benevolent sexism to predict housework presents an inconsistent, though interesting, pattern. None of these housework-related studies examined statistical interactions between hostile and benevolent sexism, which would allow testing how combinations of the two (high in both, high hostile/low benevolent, etc.) predict housework. Whereas some might argue that ambivalent implies simultaneously holding two different views (detectable via interaction), only one study to our knowledge has analyzed interactions between benevolent and hostile sexism (LaFrance & Woodzicka, 1998, on a topic unrelated to housework). Based on our arguments about the combination of benevolent and hostile sexism combining to intensify men’s and women’s beliefs that women should fulfill traditional roles, we test such interactions.
Current Study
To improve prediction of mixed-gender couples’ chore allocation, we go beyond traditional hostile sexism and test the multifaceted notion of ambivalent sexism (Glick & Fiske, 2001). Our investigation of only mixed-gender couples stems from the longstanding question of husbands’ and wives’ division of housework. Our use of interactions within multiple-regression allowed us to distinguish what we considered ambivalent sexism (high on both), unidimensional hostile and benevolent sexism, and non-sexist beliefs (low on both). Failure to test for interactions precludes potentially rich findings, such as one sexism dimension qualifying or accentuating the association between the other sexism dimension and housework. Our references to “high” and “low” scorers do not represent a categorical typology, but instead, for simplicity, represent relatively high or low continuous scores on the sexism measures. Our analyses also include numerous demographic and socioeconomic control variables, many of which have been associated in prior research with women’s and men’s housework. For example, women’s earnings (negatively), number of children (positively), age of youngest child (negatively), and husbands’ labor-force hours (positively) have been associated with women’s housework hours (Killewald & Gough, 2010, in a representative U.S. sample). Number of children (positively), age of youngest child (negatively), husbands’ labor-force hours (negatively), and wives’ labor-force hours (positively) predicted husbands’ housework hours (Killewald & Gough, 2010).
Based on the above ideas, we hypothesize that hostile and benevolent sexism will interact statistically (H1) so that:
-
H1a: When respondents (men or women) are low in both hostile and benevolent sexism, women will perform the fewest hours and smallest ratio (though still likely over 50%) of household labor, compared to when respondents are high in one or both forms of sexism.
-
H1b: When respondents (men or women) are high in both hostile and benevolent sexism, women will perform the greatest hours and largest ratio of household labor.
-
H1c: When respondents’ attitudes fit either of the other combinations (high hostile and low benevolent sexism, or the reverse), women’s and men’s hours and proportions will fall in between the levels expected in Hypotheses 1a and 1b.
Method
Sample and Procedures
The sample included N = 249 individuals (i.e., only one member per couple), who by our inclusion criteria were in mixed-gender marriages. Participants were also required to be at least 18 years old and married at least 1 year. Upon receiving Institutional Review Board approval, we publicized the study via two sources. One was Amazon.com’s Mechanical Turk (MTurk) system, which yielded 160 participants. MTurk allows businesses and researchers (“requesters”) conducting surveys, consumer preference tests, etc., to reach individuals interested in completing such tests for monetary compensation (“workers”). Upon completion of MTurk tasks, requesters can rate workers’ thoroughness, whereas workers can rate requesters’ fairness (e.g., regarding pay). Requesters can limit studies to workers with certain characteristics (e.g., home nation, languages spoken, history of high-performance ratings). Our study was open only to U.S. workers and, to exclude malicious users, to those with performance/accuracy rates > 95%. Despite concerns over MTurk workers’ generalizability to larger populations, comparisons of them with community and undergraduate samples has shown that many findings replicate across groups (Goodman, Cryder, & Cheema, 2013). The other recruitment source was a southwestern U.S. state university’s daily e-mail announcement to faculty, staff, and students, yielding 89 participants. We used two recruitment avenues to increase sample size and broaden participants’ demographic characteristics (e.g., older age-groups, education below a Bachelor’s degree) beyond those most observable in a university setting. Eligible individuals who visited the MTurk website or saw the university announcement could follow a computer link to the online survey. MTurk respondents were compensated $0.50 each, whereas university participants could enter a drawing for one of two $25 Amazon gift cards (funding was limited, as this was a student’s dissertation study without grant support). Characteristics of the overall sample and of the MTurk and university subsamples appear in Table 1. The MTurk subsample unexpectedly showed better gender balance than did the university subsample. Also, nearly half of MTurk participants had below a Bachelor’s degree, whereas only one-fifth of university-based respondents did. In the university setting, the 1-year marriage requirement likely attracted greater numbers of graduate students than undergraduates, as well as faculty and staff with advanced degrees.
Table 1
Sample Demographic Characteristics
| Variable / Category | Overall sample | MTurk | University |
|---|---|---|---|
| Gender | |||
| Men | 100 (40.2) | 77 (48.1) | 23 (25.8) |
| Women | 149 (59.8) | 83 (51.9) | 66 (74.2) |
| Age | |||
| 18–24 | 23 (9.2) | 17 (10.6) | 6 (6.7) |
| 25–34 | 105 (42.2) | 66 (41.3) | 39 (43.8) |
| 35–44 | 66 (26.5) | 48 (30.0) | 18 (20.2) |
| 45–54 | 31 (12.4) | 18 (11.3) | 13 (14.6) |
| 55–64 | 21 (8.4) | 9 (5.6) | 12 (13.5) |
| 65+ | 3 (1.2) | 2 (1.3) | 1 (1.1) |
| Ethnicitya | |||
| African-American/Black | 12 (4.8) | 11 (6.9) | 1 (1.1) |
| Asian-American | 14 (5.6) | 10 (6.3) | 4 (4.5) |
| European-American/White | 191 (76.7) | 134 (83.8) | 57 (64.0) |
| Hispanic | 13 (5.2) | 4 (2.5) | 9 (10.1) |
| Not Specified | 17 (6.8) | 1 (0.6) | 16 (18.0) |
| Education | |||
| High school or less | 20 (8.0) | 19 (11.9) | 1 (1.1) |
| Some college or 2-year | 70 (28.1) | 54 (33.8) | 16 (17.9) |
| Bachelor’s | 88 (35.3) | 65 (40.6) | 23 (25.8) |
| Master’s or higher | 71 (28.5) | 22 (13.8) | 49 (55.0) |
| Children | |||
| Yes (one or more) | 152 (61.0) | 96 (60.0) | 56 (62.9) |
Note. MTurk = Amazon Mechanical Turk. Frequencies appear outside of parentheses and percentages of overall sample and of MTurk and university participants in different subgroups appear inside parentheses.
aIn the university subsample, one respondent each (1.1%) described themselves as Arab-American and Native American.
Measures
Housework
Participants received a list of eight household tasks (cooking/meal clean up, housecleaning, laundry, childcare/child transportation/child activities, family management [planning, maintaining, scheduling activities and tasks], home repairs/gardening/lawn care, and paying bills/financial matters), along with an “other” option. Instructions asked participants to: “Please indicate the number of hours per week, on average, you and your spouse complete each of the activities below” (with one space to report one’s own hours on a given task and another to report on the spouse’s). Prior research has shown questionnaire measures of work hours to correlate moderately highly (r = .45 and .46 for women and men; Kan, 2008) with daily-diary measures, which are considered more accurate. These measures yielded 1) respondents’ self-reported average number of weekly hours they themselves spent on overall household labor (self-housework hours) and 2) the average number of weekly hours the focal respondent reported their spouse spent on household labor (spouse housework hours). These two variables then yielded respondents’ proportional housework (self-proportion = self-reported own housework hours/total of self-reported own hours plus hours attributed to one’s spouse; Greenstein, 2000).
Sexism
Participants completed the 22-item Ambivalent Sexism Inventory (Glick & Fiske, 1996) to measure hostile and benevolent sexism (each 11 items). Responses were on a 6-point Likert-type scale from 1 (disagree strongly) to 6 (agree strongly). Sample hostile-sexism items are: “Women seek to gain power by getting control over men,” and “A wife should not be significantly more successful in her career than her husband.” Sample benevolent-sexism items are: “A good woman should be set on a pedestal by her man,” and “Many women have a quality of purity that few men possess.” Alpha reliability coefficients ranged from .89 to .92 for the four scales (men’s and women’s benevolent and hostile sexism).
Demographics/Control Variables
Demographic control variables included: age (1 = 18–24, 2 = 25–34, 3 = 35–44, 4 = 45–54, 5 = 55–64, 6 = 65 or older), highest educational attainment (1 = elementary, 2 = some high school, 3 = high school, 4 = some college, 5 = Associate’s degree/2-year technical degree/vocational school, 6 = Bachelor’s, 7 = Master’s, 8 = doctorate/professional degree), race/ethnicity (see Table 1 for categories), and U.S. region of residence (Northeast, Midwest, South, West). The latter variable controlled for U.S. political differences as, for example, the Northeast tends to be most liberal and the South most conservative. The survey also inquired into the number of weekly hours the respondent and his/her spouse (assessed in separate items) worked for pay outside the home and the respondent’s current household income (1 = $0–24,999, 2 = $25,000–49,999, 3 = $50,000–74,999, 4 = $75,000–99,999, 5 = $100,000–149,999, 6 = $150,000+). Finally, the survey inquired into age at marriage, length of marriage, parenthood status (yes/no), and number of children, if parents.
Results
Analysis Plan
Analyses consisted of correlations and multiple-regression analyses. A sample of roughly 250 yields nearly 90% statistical power to detect a small-medium correlation of .20 as significant with p < .05 (Zhang & Yuan, 2018). Estimating statistical power to detect interactions in regression analyses is complex. Whereas, in experiments, the investigator can ensure the proper balance of cell sizes across combinations of conditions, survey and field researchers lack control over this aspect. In the present study, we could not know in advance if the number of participants with relatively high scores on both hostile and benevolent sexism would be comparable to the number with relatively high scores on hostile sexism and relatively low scores on benevolent sexism, as well as the number with other combinations of high and low scores (McClelland & Judd, 1993). For this and other reasons, note McClelland and Judd, interactions in survey/field studies typically account for only 1–3% of the variance in dependent measures. In fact, McClelland and Judd conclude that “enormous samples are required to have the statistical power of optimally designed experiments for detecting interactions” (p. 387). Further, estimating the power of an interaction requires specifying in advance not only an estimate of the interaction effect’s magnitude, but also estimates of the main effects for the predictors and of the correlation between the predictors (Baranger, 2019a). As a guideline, an interaction with magnitude half as large as the associated main effects will require four times the sample size needed to detect the main effect (Baranger, 2019b). For our study, with bivariate associations estimated to be around r = .20, an interaction equal in magnitude to r = .10 would require nearly 1,000 participants (4 X 249). Such a sample size was not feasible for us and, hence, we acknowledge that our study has well below the conventional 80% power for detecting regression-based interactions.
Preliminary Analyses
Table 2 presents means and standard deviations of, and correlations among, the major variables, for men and women. Relative to the scale maximum (6) and midpoint (3.5) for benevolent and hostile sexism, men’s and women’s means were low-moderate. Men reported greater benevolent sexism toward women (M = 3.44) than did women, M = 3.08; t(247) = 2.94, p = .004; there was no gender difference in hostile sexism, t(247) = 0.27, p = .79, with both means around 2.70.
Table 2
Descriptives and Correlations of Major Variables by Gender
| Variable | 1. | 2. | 3. | 4. | 5. |
|---|---|---|---|---|---|
| 1. Hostile sexism | - | .32*** | .07 | .00 | 2.72 (1.18) |
| 2. Benevolent sexism | .34*** | - | .17* | .16 | 3.08 (0.96) |
| 3. Housework self (hours) | .21* | .16 | - | .46*** | 35.30 (32.39) |
| 4. Housework self (ratio) | .22* | .04 | .18 | - | 0.64 (0.19) |
| 5. M (SD) | 2.76 (1.11) | 3.44 (0.94) | 25.57 (24.44) | 0.50 (0.15) | - |
Note. Men’s correlations and descriptives appear below the diagonal, women’s appear above the diagonal.
*p < .05. ***p ≤ .001 (two tailed).
Women reported greater average hours of (self) housework (M = 35.30) than did men, M = 25.57; t(240.2) = 2.68, p = .008. Further, based on respondents’ reported own and spouse’s housework hours, women performed a greater proportion of housework than did men, t(235.8) = 6.38, p < .001. Interestingly, men reported doing, on average, an equal amount of housework as their wives (i.e., .50/.50), whereas women reported doing nearly two-thirds of it (i.e., .64). Examining each task individually, men’s and women’s mean reported (self) proportions differed substantially (p < .001) for all tasks except paying bills and “other” (which had only 26 responses). Although men’s and women’s mean proportions do not have to sum to 100% in our sample (which contains one spouse per couple), they nearly did for home repairs. Men estimated they made 74.5% of repairs, whereas women estimated making 28.4%. Women’s proportions exceeded men’s in the remaining five areas: cooking, 67.4% for women versus 47.2% for men; cleaning, 74.0% versus 42.9%; laundry, 76.2% versus 41.2%; childcare, 71.0% versus 48.2%; and family management (planning, scheduling, etc.), 77.4% versus 51.7%. Benevolent and hostile sexism were correlated with each other in the low-mid .30s in men and women.
Primary Analyses
Four multiple-regression analyses constituted the primary analyses, with (self) number of housework hours and (self) proportion of housework serving as dependent variables, separately for women and men. Because ordinary regression deals with continuous dependent and predictor variables, it offers the best way, in our view, to test statistical interactions between continuous predictors (i.e., hostile and benevolent sexism). For predicting wives’ self-reported housework hours, there were several significant results. Four control variables reached statistical significance (Table 3; main-effects-only model). Wives with children performed more housework hours than those without. Greater household income was associated with fewer hours of wives’ housework. Also, for each additional hour wives worked outside the home, they did nearly 1 hour fewer of housework. Wives living in the Midwest performed fewer hours of housework than those living in the West (referent category). Because main effects should be interpreted only with other main effects and not with interactions in the same model (Cohen, 1978), we see that benevolent and hostile sexism did not significantly predict women’s own hours of housework.
Table 3
Multiple-Regression Results for Women’s Housework (Unstandardized)
| Predictor | Dependent variable
|
|||
|---|---|---|---|---|
| Women’s (self) hours
|
Women’s proportion
|
|||
| Main-effects model | Interaction model | Main-effects model | Interaction model | |
| Constant | 74.0** | 47.3 | 0.45** | 0.27 |
| Age | −9.7 | −11.5 | 0.03 | 0.02 |
| Education | 3.0 | 3.4 | 0.00 | 0.00 |
| Household income | −3.9* | −3.3 | 0.01 | 0.01 |
| Age at marriage | 0.9 | 1.1 | −0.00 | −0.00 |
| Length of marriage | 0.1 | 0.3 | −0.00 | −0.00 |
| Race (1 = White, 0 = Non-White) | 6.7 | 6.0 | 0.07* | 0.06* |
| Number of children | −0.4 | −0.5 | 0.02 | 0.02 |
| Parental status (2 = No, 1 = Yes) | −30.5*** | −30.8*** | −0.02 | −0.02 |
| Resides in Northeast (dummy) | 2.6 | 1.0 | 0.10 | 0.09 |
| Resides in Midwest (dummy) | −17.7* | −17.0* | 0.08 | 0.08 |
| Resides in South (dummy) | −4.8 | −6.1 | 0.07 | 0.06 |
| SELF hours worked outside home | −0.8*** | −0.8*** | −0.005*** | −0.005*** |
| SPOUSE hours worked outside home | 0.2 | 0.2 | 0.002** | 0.002* |
| Benevolent sexism | 2.8 | 11.7* | 0.03* | 0.09** |
| Hostile sexism | 2.1 | 11.5* | 0.00 | 0.06 |
| Benevolent X hostile | - | −3.2 | - | −0.02† |
| R2 | .45*** | .47*** | .37*** | .39*** |
†p = .053. *p ≤ .05. **p < .01. ***p < .001 (two-tailed).
Several variables predicted wives’ proportion of housework hours. The greater wives’ outside work, the smaller was their share of housework. Also, the greater outside work wives attributed to their husbands, the more wives contributed to housework. White women (relative to minorities) reported doing a higher share of housework. Lastly, among main effects, wives’ benevolent sexism was positively associated with their housework share.
There was also a negatively signed Benevolent Sexism X Hostile Sexism interaction (p = .053), which is plotted in Figure 1 (software from Dawson, n.d.). Within Figure 1, “high” refers to one standard deviation above the mean of a given variable, whereas “low” refers to one standard deviation below. Wives contributed the largest housework shares when high in benevolent sexism, but low in hostile sexism. Being low in both forms of sexism was associated with women’s lowest housework share (supporting Hypothesis 1a), although still around 50%. Hypothesis 1b was not supported, as women high on both forms of sexism were not highest in housework proportion. Hypothesis 1c was partly supported, as the combination of women’s low benevolent, but high hostile, sexism was associated with an intermediate housework proportion.
Figure 1
Interaction of Women’s Benevolent and Hostile Sexism in Predicting Their Proportion of Housework
Relatively few significant effects emerged for husbands. The more hours wives worked outside (reported by husbands), the more housework husbands reported performing (B = 0.34, p = .03). The more hours husbands reported working outside the home, the smaller their housework share (B = −.002, p = .052). In contrast, the more hours husbands reported their wives working outside the house, the larger the share of husbands’ housework (B = .002, p = .01). Older husbands’ share of self-reported housework was smaller than that of their younger counterparts (B = −.12, p = .004). However, men’s later age at marriage was associated with their doing a larger share of housework (B = .01, p = .002). Finally, the longer husbands were married, the larger their share of housework (B = .01, p = .02).
Supplemental Analyses
Because the household tasks were heterogeneous, we also examined them individually. Results predicting men’s and women’s proportions appear in Table 4. Several tasks showed significant associations with hostile and/or benevolent sexism, although no single task drove the previous results for the overall housework indices. No Benevolent X Hostile interactions were significant, so we report only from main-effect models. The greater men’s benevolent sexism, the lower proportion of cooking men did (B = −.07, p = .04), whereas women’s internalization of benevolent sexism was associated with women performing a higher share of cooking (B = .08, p = .001). Women’s benevolent sexism was also positively associated with their repair work (B = .08, p = .01). Surprisingly, the greater men’s hostile sexism, the greater share of childcare they performed (B = .07, p = .02). A similar result emerged for men’s hostility and their proportion of family-management work (B = .07, p = .04).
Table 4
Regression Results (Unstandardized) Predicting Self-Proportion of Work on Each Task From Main-Effect-Only Model
| Variable | Cooking
|
Cleaning
|
Laundry
|
Childcare
|
Family Mgmt
|
Repairs
|
Pay bills
|
|||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M | W | M | W | M | W | M | W | M | W | M | W | M | W | |
| Benevolent sexism | −.07* | .08*** | −.05 | .03 | −.04 | .01 | −.00 | .05 | .06 | .00 | .03 | .08** | .07 | −.04 |
| Hostile sexism | .04 | −.02 | .04 | .01 | .03 | −.01 | .07* | .01 | .07* | −.04 | .05 | −.02 | .05 | .02 |
Note. M = men; W = women; Mgmt = management. All control variables in Table 3 were also used in the task-specific analyses but are omitted for simplicity.
*p ≤ .05. **p ≤ .01. ***p ≤ .001 (two-tailed).
Discussion
Wives who exhibited high benevolent sexism but low hostile sexism performed the largest proportion of housework. This finding did not conform with Hypothesis 1b, which predicted women’s greatest housework when they (or their husbands) were high in both benevolent and hostile sexism. Results supported Hypothesis 1a, however, as wives low on both kinds of sexism performed the lowest housework share. Examining not just hostile sexism but also benevolent sexism thus enhanced our ability to predict women’s housework proportion. Illuminating these results for women’s overall housework share, task-specific analyses showed women’s greater benevolent sexism to predict their higher proportions of cooking and home repairs.
A key insight of ambivalent sexism theory is that, although benevolent and hostile sexism serve opposite functions as “the carrot aimed at enticing women to enact traditional roles and… the stick used to punish them when they resisted,” they “both work toward a common aim: maintaining a gender-traditional status quo” (Glick & Fiske, 2011, p. 533). Further, as noted, Hammond et al. (2016) contended men’s direct expression of hostile sexism likely evokes resentment in women, thus requiring the “carrot” of benevolent sexism to incline women toward traditional gendered behavior (e.g., housework). Indeed, women’s benevolent sexism showed a clear positive association with their housework share, especially when they exhibited low hostile sexism (Figure 1). Interestingly, women’s hostile sexism was generally unassociated with their amount of housework. Hammond and colleagues’ suggestion that hostile sexism may do little other than spur women’s resentment could explain our findings. In contrast, Gaunt and Pinho (2018) found women’s internalization of hostile sexism to positively predict their own childcare. Women’s internalized hostility in Gaunt and Pinho’s study also predicted maternal gatekeeping (including disparaging attitudes toward male partners), which would explain women’s greater childcare involvement.
That some women (or members of any group) would internalize patronizing ideologies toward themselves (as some of our female participants appeared to do) may surprise readers. However, research indicates that women high in benevolent sexism view some of its characteristics as appealing. In fact, some women report benevolent sexists as being more attractive than hostile sexists and non-sexists (Bohner, Ahlborn, & Steiner, 2010; Hammond & Sibley, 2011). Seemingly positive/caring qualities of benevolent sexism can have negative consequences, in that benevolent sexism, like hostile sexism, is associated with traditional gender expectations and roles for women. Further, benevolent sexism may deter some women from challenging careers (e.g., lawyer, law enforcement) when men’s discouragement is framed as protecting women’s safety (Moya et al., 2007). Among the reviewed studies, del Prado Silván-Ferrero and López (2007) found high school girls’ benevolent sexism to correlate positively with their performance of highly gender-typed housework, a finding that aligns with ours. del Prado and López speculate that, in Spain, parents may be transmitting traditional—and inequality-perpetuating—gender-role norms to their daughters, which override educational curricula designed to combat sexism. Indeed, research from multiple countries has shown various parental attitudes and socialization behaviors to be positively associated with teenage girls’ benevolent sexism (Dueñas, Santiago-Larrieua, Ferre-Reya, & Cosi, 2020; Mastari, Spruyt, & Siongers, 2019; Montañés et al., 2012). Finally, the present study showed partially overlapping findings with an earlier study (Chen et al., 2009), in which U.S. women’s traditional attitudes toward women’s domestic life were associated with benevolent sexism. Unlike our study, however, Chen et al. found U.S. women’s hostile sexism to covary with traditional attitudes.
Although neither men’s benevolent nor hostile sexism predicted their aggregate housework (hours or proportion), the sexism measures predicted proportional performance of certain tasks. Men’s greater benevolent sexism predicted a lower proportion of cooking they did. More surprisingly, men’s greater hostile sexism predicted their greater proportional involvement in childcare and family management. As reviewed above, Gaunt and Pinho (2018) found women’s hostile sexism toward their own gender to predict their own greater childcare work, indirectly via maternal gatekeeping. Possibly, parallel to Gaunt and Pinho, our obtained positive link between men’s hostile sexism and proportions of childcare and family management could involve gatekeeping on men’s part. Conceivably, men’s gatekeeping would reflect their desire to participate more in child- and family-related activities as a way to limit their wives’ time with children. This conclusion assumes that men high in hostile sexism toward women in general direct some of this hostility toward their wives, a proposition for future research.
Among control variables, those involving one’s own and one’s spouse’s outside work hours showed robust associations with housework performance. In husbands and wives, the more (self) hours one worked outside home, the fewer the (self) hours worked at home. Further, in husbands and wives, the more outside-of-home work hours one attributed to the spouse, the larger share of (self) housework one reported doing. Using a complex multi-national, multi-level design, Sani (2014) found that, in countries characterized by men’s higher average paid work hours, women’s share of families’ housework was also high. Sani (2014) also found that, in countries with high percentages of women in the labor force, women perform smaller shares of housework. Though the percentage of women in the labor force is not the same as their numbers of hours worked, both our findings and Sani’s suggest at a general level that both husbands and wives take on greater shares of housework when their spouses work outside the home. Whereas we found White women to perform a greater share of housework than did women of color, Wight, Bianchi, and Hunt (2012) found the opposite using national data (i.e., female-to-male housework ratios of 1.6 in White, 1.8 in Black, 2.6 in Asian, and 2.7 in Hispanic couples). Because of our relatively low proportion of minority participants, our results on race-ethnicity should be taken with caution.
As with any study, ours had strengths and limitations. Strengths included novel application of the ambivalent-sexism construct to predict household work; participants’ socioeconomic and educational variation; and the ability to compare MTurk and university-based samples (Table 1). Limitations included reliance on self-reports, in which our male participants may have overreported their housework relative to previous surveys (cf. Miller, 2015); availability only of individual rather than dyadic data (precluding cross-spouse corroboration of reported housework; Reifman & Niehuis, 2018); and restriction of the sample to individuals in mixed-gender partnerships.
How might ambivalent sexism theory extend to same-gender couples’ housework allocations? Geist and Ruppanner (2018) sketch out how existing housework theories can encompass diverse families, including same-gender couples. These authors note that, “Those who have investigated same-sex couples’ housework allocations document greater equality than in heterosexual couples… Explanations center on the decreased importance of gender display among partners of the same gender but also on the greater importance of ideas of equality…” (p. 250). Brewster (2017), reviewing relevant literature from 2000 to 2015, documented that lesbian partners share housework more equally than their heterosexual counterparts. Smart, Brown, and Taylor (2017) found via national time-use surveys that “gay men and lesbians spend more time on household serving labor than straight men, and less than straight women” (p. 81), which they interpret as greater egalitarianism in same- rather than mixed-gender couples’ housework. Examining variation in same-gender spouses’ hostile and benevolent sexism will further aid efforts to understand couples’ allocation of household labor.
Our findings have practical applications in at least two areas. The first involves governmental social policies and egalitarian behavior within households. Fuwa and Cohen (2007), examining 33 countries (similar to those in Davis et al., 2007), merged information on national policies with survey responses to study such linkages. Countries with generous family-leave policies exhibited greater egalitarianism in housework allocation. Further, tendencies for wives’ full-time employment and higher income to equalize their housework with husbands were stronger in countries with pro-egalitarian policies (e.g., affirmative action, absence of discriminatory policies) than in those without. Though causation between country-level policies and citizens’ attitudes and behavior is not easily determined, reduced sexism (hostile and benevolent) and greater equality in housework likely can help reinforce a more egalitarian culture. Second, couples’ housework allocation may affect their marital/relationship quality. Niehuis, Skogrand, and Huston (2006) argued that, beyond romance and passion, couples require a realistic impression of the mundane, day-to-day activities of their lives together (possibly including housework) to function harmoniously (see also Huston, Niehuis, & Smith, 2000; Joel et al., 2020). Research also shows that relationship satisfaction/quality is higher when there is less inequality (Lavee & Katz, 2002; Saginak & Saginak, 2005; Schieman, Ruppanner, & Milkie, 2018).
Many substantive and methodological issues discussed above suggest ideas for future research. Most intriguing is the idea of paternal gatekeeping, in which husbands’ hostile sexism leads them to minimize their children’s time spent with their mothers. Future studies, directly assessing gatekeeping, as well as hostile sexism and hostility specifically to one’s spouse, could address this topic. Studies linking ambivalent sexism, housework allocation, and support for egalitarian government policies would also advance our understanding. Expanding the diversity of populations studied, including on race-ethnicity and sexual orientation, would further enrich the literature. Finally, studies that were longitudinal, included both spouses/partners in a couple, and used additional housework measures (e.g., diaries) would also advance the field.
In conclusion, our application of ambivalent sexism theory to mixed-gender couples’ housework allocations—among the first of its kind—has yielded interesting findings via statistical interaction of benevolent and hostile sexism. Though one might expect men’s sexism toward women to play a robust role in housework allocations, it was actually women’s internalized benevolent sexism that did so. Replication and more refined understanding of these findings, as well as extensions to diverse families, await future research.