Marriage and housework

1 Marriage and Housework Cristina Borra ([email protected]) Universidad de Sevilla Facultad de Ciencias Económicas, Universidad de Sevilla Ramón y Cajal 1, 41018 Sevilla (Spain) e-mail: [email protected] Telephone: +34 954551699 Martin Browning ([email protected]) CEBI, Department of Economics, University of Copenhagen, Øster Farimagsgade 5, bygning 26 1353 København K (Denmark) e-mail: [email protected] Almudena Sevilla ([email protected]) Department of Social Sciences University College London (UCL) 55-59 Gordon Square, Office 210 London WC1H 0NT Tel.: +44 (0) 2076126473 e-mail: [email protected] Acknowledgements We thank the comments from two anonymous referees as well as from seminar participants at University of Chicago, Economics of the Family Conference, University of Notre Dame (2016), and the Economic Demography Workshop, Lund University (2018), as well as conference attendants at the Royal Economic Society (2017) and the 39th International Association of Time Use Research Conference (2017). The activities of CEBI are financed by the Danish National Research Foundation. Electronic copy available at: https://ssrn.com/abstract=3524799 2 Abstract This paper provides insights into the gains of forming a couple by estimating how much of the difference in housework between single and married individuals is causal and how much is due to selection. Permanent unobserved heterogeneity explains about half of the observed differences in housework documented in the cross-sectional data. Further ancillary evidence suggests that individuals with a higher preference for marriage also have more traditional views on the division of household labour. There remains a genuine half-an-hour increase per week in housework time for each partner, with women specializing in routine and men in non-routine housework tasks. Key Words: Marriage, Time use, Home production JEL codes: D13, J12, J22 Electronic copy available at: https://ssrn.com/abstract=3524799 3 1. Introduction Spouses' time constitutes an important input into the production of household goods. Across industrialized countries, men spend about 20 hours per week on home production, and women about 30 hours per week. This exceptional amount of time (which excludes the time devoted to child care) is mostly used for the production of household public goods, such as having a clean house, or cooking a home-made meal. The sharing of these household public goods and services, the so-called commodities (see Becker (1965)), forms the basis for arguably one of the most important efficiency gains associated to marriage. In fact, spouses’ similar demands for the consumption of household public goods, rather than increased specialization, can explain the assortative matching along the wage distribution commonly observed in the marriage market (Lam, 1988). In this paper we document and explore the mechanisms behind the differences in housework between single and married individuals (‘marrieds’). Specifically, we estimate how much of the substantial difference is causal and how much is due to selection into marriage. This paper starts from our observations of a robust finding on housework and marriage. 1 We use cross-sectional time-diary data for eleven industrialized countries to document that, for all countries considered, married individuals do much more housework than comparable single individuals. Married women devote almost nine more hours per week to housework than single women from an average of 25 hours per week, although there is a great dispersion in the additional housework time across countries. This increase in housework upon marriage is concentrated in routine housework tasks such as cleaning, cooking, and ironing, which need to be performed on a regular basis. Married men do about an hour less of routine housework tasks per week than single men, but they do about one hour and a half more of non-routine housework (such as home repairs and managing finances) than singles. Taken together, a couple spends 15 per cent longer in housework-related tasks than two singles. These findings are not a result of specialization in the labour market, and are robust to controlling for hours of paid work and a wide set of covariates such as the number of children. The differences in housework between singles and marrieds can be broadly attributed to either state dependence or selection. State dependence refers to effects that arise directly from being together compared to living separately, and captures any indirect effect through characteristics which change from before to after a marriage, such as number of children. Selection involves differences arising from non-random matching in the marriage market due to either permanent observed 1 Throughout the paper we use the term marriage to refer to individuals who are either legally married or in a cohabiting relationship. We also present separate analyses when appropriate. Electronic copy available at: https://ssrn.com/abstract=3524799 4 individual characteristics such as differences in education between singles and marrieds, or as a result from permanent unobserved heterogeneity in the propensity to marry that is correlated with preferences for home produced goods and services. Current available data are not well suited for distinguishing the different explanations for state dependence. Instead we focus on how much of the effect observed in the cross-section data can be attributed to selection into being married due to unobserved time invariant heterogeneity. Thus, we use the term selection throughout the paper to refer to selection effects net of other permanent observed individual characteristics such as education. To that end, we exploit the longitudinal nature of three panel data sets to explore the selection hypothesis: the Panel Study of Income Dynamics (PSID), the harmonized British Household Panel Survey-United Kingdom Household Longitudinal Study (BHPS-UKHLS), and the Household, Income and Labour Dynamics in Australia (HILDA) Survey. For the latter we have information on both routine housework and non-routine housework, whereas for the PSID and BHPS-UKHLS we only have information on routine housework. We show that unobserved time-invariant heterogeneity can go a long way in explaining the differences in routine housework upon marriage. Between about 50 and 70 per cent of the increase of routine housework upon marriage can be accounted for by permanent unobserved heterogeneity for women. Similarly, for men, the decrease in routine housework upon marriage observed in the cross-section virtually vanishes once unobserved time-invariant heterogeneity is taken into account. Further ancillary evidence suggests that individuals with a higher preference for marriage also have a more traditional view on the division of household labour, rather than individuals with a certain preference for housework being perceived as more desirable partners. We fail to find any role for unobserved time-invariant heterogeneity effects in the case of non-routine housework. After permanent unobserved heterogeneity is taken into account, there remains what seems to be a genuine increase of one hour and a half in routine housework upon marriage for women and in non-routine housework for men. Our paper contributes to the existing literature on the welfare gains of forming a couple, which is crucial for the analysis of individual decisions regarding union formation and dissolution (Chiappori et al 2002, 2017, Bruze et al. 2015). The focus of this literature is on the division of labour within marriage, rather than on the changes in the time devoted to home labour upon marriage (Becker 1965, Gronau 1977, Grossbard-Shechtman 1984, Couprie 2007). Whereas selection into marriage has been previously suggested as a potential explanation for differences in housework time between marrieds and singles (Auspurg et al. 2014, Stratton 2015), this hypothesis has never been tested. Here we look at how and why the time devoted to home labour changes upon marriage. We are the first Electronic copy available at: https://ssrn.com/abstract=3524799 5 paper providing robust evidence on the causal impact of marriage on time allocated to home production across different developed countries using longitudinal data spanning over two decades. The paper is organized as follows. Section 2 documents the increase in unpaid labour upon marriage by providing harmonised cross-national evidence drawn from time-use surveys. Section 3 very briefly (and informally) uses economic theory to review the various effects of state dependence that may be operative and justifies our focus on identifying selection effects. Section 4 uses longitudinal data to look at the role of unobserved heterogeneity in explaining the cross-sectional results. Section 5 concludes. 2. Cross-sectional relationship between marriage and housework 2.1 24-Hour Diary Cross-Sectional Data We use 24-hour time diary surveys from the harmonized Multinational Time Use Study Data set (MTUS) in 9 industrialized countries to document the increase in home labour upon marriage. Diary surveys collect information on a respondent’s activities during a 24-hour period, and the diary is completed on a selected day, either on a weekday or on a weekend day or in both days (see Table A.1 in Appendix A for a description of these surveys). MTUS data have been harmonized to minimize differences in survey methodology, and the use of 24-hour diary surveys minimizes comparability issues across surveys in time use categories (see for example Guryan et al., 2008 for a discussion about the conceptualization and comparability of child care time using these surveys). The reliability and validity of MTUS diary data is well established in the literature, just as money expenditure diaries have become the gold standard for describing consumption behaviour (Aguiar and Hurst 2007, Gimenez-Nadal and Sevilla 2012, Ramey and Ramey, 2010). We use a cross-sectional sample of working-age individuals from Austria (1992), Canada (1998), France (1998), Germany (2001), Italy (2003), Norway (2000), Spain (2001/02), the United Kingdom (2000/2001) and the USA (2003-11). We restrict the sample to respondents who are not living with their parents between 24 and 65 years, and perform several robustness checks with subsamples of individuals of different ages. Throughout the paper we compare the time spent in housework of singles (i.e., single individuals not living in the parental home) and marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting). The variable of interest is hours per week spent in housework time. We follow the literature and construct our housework variable from the diary, adding up the time spent in the following categories: cooking, household upkeep, shopping, domestic travel, maintenance, and gardening as in Electronic copy available at: https://ssrn.com/abstract=3524799 6 Stratton and Stancanelli (2014) and Aguiar and Hurst (2007), see Table A.2 in Appendix A for a full description of these housework categories. We also distinguish between routine and non-routine housework. Routine housework activities such as cooking and doing the laundry are those that need to be done on a regular basis in a given week. Non-routine housework such as odd jobs around the house and car repairs have a lower frequency. As a result, compared to routine housework, nonroutine housework can often be postponed and is easier to outsource (see Hersch 1991, Hersch and Stratton, 2002). This distinction is important for our purposes because whereas women tend to specialize in routine housework activities, men tend to specialize in non-routine housework. 2.2 Marriage and Housework: Cross-Sectional Results The raw data in Table 1 compares the hours per week spent in housework for married men and women versus single men and women from the following OLS model: 𝐻𝑖= 𝛼 + 𝛽1 𝑀𝑖+ 𝜀𝑖 (1) where 𝐻𝑖 denotes hours per week devoted to housework by individual i, and 𝑀𝑖 is a dummy for living with a partner (cohabiting or being married). We use MTUS-proposed weights to ensure population, day of the week, and seasonal representativeness (Fisher et al., 2018). There are very few individuals reporting zero time in the diary because of the high degree of aggregation of housework activities into routine and non-routine housework (see Table A.3 in Appendix A with summary statistics for the dependent variables of housework and routine and non-routine housework). The lack of zeros justifies the use of OLS methods rather than other methods, such as Tobit (see Foster and Kalenkoski, 2013). There are marked gender differences in terms of housework time, which are already visible in the single state. Single women tend to spend about 25 hours per week on housework (Column 1), whereas single men spend about 16 hours per week (Column 3). Gender differences are accentuated for married individuals. Married women do on average 8 hours and a half more of housework per week than single women (Column 2), whereas married men do about half an hour less of housework per week than single men (Column 4). Figure 1 shows that in countries where the marriage housework penalty for women is higher, the marriage housework bonus for men is lowest. The increase of housework upon marriage for women is higher than the decrease of housework upon marriage for men, resulting in housework time being higher upon marriage for a couple than for two comparable singles. In particular, Column 6 shows that a couple spends about 3 more hours (15 per cent longer) in housework-related tasks than two singles who spend about 21 hours each (Column 5). The increase in housework upon marriage for women, the decrease in housework upon marriage for men, and the Electronic copy available at: https://ssrn.com/abstract=3524799 7 overall increase in a couple’s housework relative to two singles occur in most countries considered here, although we reject that the size of the marriage coefficient is the same across countries (see 𝜒2- statistics in the last row of Columns 2, 4, and 6 in Table 1). The results in Table 1 may well be driven by other factors unrelated to the marital status of the individual. For example, married individuals may devote more time to housework because there are more persons in the households (such as children or other adults), or because they work less in the labour market. Table 2 presents the results from an OLS regression of housework activities 𝐻𝑖 for individual i on an indicator variable for being married or cohabiting 𝑀𝑖 as in Equation (1), which also controls for individual and household-level characteristics as in Equation (2). These estimates can be interpreted in a descriptive way, as simple means of housework for each country for married and cohabiting individuals versus single individuals, keeping observable individual characteristics constant. 𝐻𝑖= 𝛼 + 𝛽1𝑀𝑖+ 𝛽2 ′ 𝑋𝑖+ 𝜀𝑖 (2) where 𝐻𝑖 denotes hours per week devoted to housework by individual i as before, 𝑀𝑖 is a dummy for living with a partner (cohabiting or being married) as in Equation (1), and 𝑋𝑖 is a vector of covariates. We control for the opportunity cost of time by including age, age squared, and education level dummies as controls. We also include the number of children below the age of 18 living in the household, dummies for the age group of the youngest child in the household, the day of the week the diary refereed to, and hours of paid work in the diary day. Diary surveys do not have information on wages, attitudes, and dwelling size or number of rooms in the house. We use MTUS-proposed weights to ensure population, day of the week, and seasonal representativeness (Fisher et al., 2018). Table A.4 in Appendix A describes how the variables are constructed from the original MTUS codes, and A.5 provides summary statistics for these variables separately by gender. Table 2 shows results for the coefficient on a dummy for cohabiting or being married from estimating Equation (2) for total, routine, and non-routine housework. Table A.6 reports full estimation results. The first row of Table 2 shows that on average, once we control for socio-economic characteristics, marriage continues to have a bigger effect on housework for women than for men albeit smaller than in the unconditional results from Equation (1) presented in Table 1. Women increase the amount of housework upon marriage by about 5 hours (Column 1 in Table 2), instead of 8 hours (Column 2 in Table 1). The increase in housework is entirely driven by increases in routine housework. The first row in Table 2 shows that married women do more routine housework (Column Electronic copy available at: https://ssrn.com/abstract=3524799 8 2), but less non-routine housework (Column 3), than single women. In contrast, married men do more non-routine housework (Column 6), but less routine housework (Column 5), than single men. When taken routine and non-routine housework together however, married women are relatively worse off than married men when compared to their single counterparts. The difference in total housework time between married and single women is almost 5 and a half hours per week (Column 1), whereas the difference in total housework between married men and single men is not statistically significant and amount to less than a quarter of an hour a week (Column 4). As in Table 1, this regularity is qualitatively true in all countries considered, although the 𝜒2-statistics in the last row of Table 2 shows that we cannot reject that the marriage effect is quantitatively different across countries. Robustness Checks The results from Table 3 show that the association between marriage and a couple’s housework described in Tables 1 and 2 remains positive for alternative samples, selecting first, childless individuals, second, cohabiting vs. singles, and third, under 45 years-old individuals. The housework gap between marrieds and singles is higher when we restrict the sample to individuals without children (Columns 1 and 4 in Table 3). In particular, whereas routine housework increased by 5.86 hours a week when all the sample of women was considered (first row of Column 2 of Table 2), it increased by 6.56 hours a week when the sample of women without children is considered (first row of Column 1 in Table 3). Similarly, whereas non-routine housework increased for men by 1.49 hours per week (first row of Column 6 of Table 2), it increased by almost 2 hours per week when only the sample of men without children is considered (first row of Column 4 of Table 3). These results seem to suggest that the relationship between housework and marriage is not driven by the arrival of children. Columns 2 and 5 of Table 3 look at how housework changes between cohabiting and single individuals. An individual who is cohabiting is defined as an individual in a heterosexual relationship who is not legally married. Results show that whereas housework is higher for cohabiting individuals than single individuals in all of the countries, the difference in housework between cohabiting and single individuals is lower than the difference in housework between married and single individuals documented in Table 2. Columns 3 and 6 of Table 3 further show that results remain virtually unchanged when considering a younger sample of individuals between 25 and 44 years of age. Electronic copy available at: https://ssrn.com/abstract=3524799 9 3. Interpreting the data in light of a Household Economic Model of Time Allocation Very broadly there are two classes of explanation for the differences we see between singles and comparable marrieds: state dependence and selection. State dependence refers to effects that arise directly from being together compared to living separately. Selection involves differences arising because of non-random matching in the marriage market due to either permanent observed individual characteristics such as differences in education between singles and marrieds, or as a result from permanent unobserved heterogeneity in the propensity to marry that is correlated with preferences for home produced goods and services. In particular, suppose that there is time invariant unobserved heterogeneity in the propensity to do housework and in the propensity to marry. If these two propensities are positively correlated then the OLS results will over-estimate the causal effect of marriage on housework. Under a simple household model where the spouses have separate utility functions over both public and private goods, and public goods are produced in the home using market goods and spouses’ time, the sources of state dependence are myriad and the sign of the impact of marriage on housework is ambiguous (marriage could lead to an increase or a decrease in housework) (Becker 1965, 1981, 1985; Lam 1988, Browning et al 2013). First, there are economies of scale in time use, arising mainly from production complementarities within the household (Lundberg 2012, Crossley and Lu 2004). For instance, cooking a meal for two does not take twice as much time as cooking a meal for one (Vernon 2010, Stratton 2015). Another example is cleaning; because of setup costs it does not take twice as much time to clean a dwelling for two as one single dwelling (Vernon 2010, Stratton 2015). This effect would tend to reduce total housework for marrieds if other state dependence effects were not operative. A second effect of state dependence arises because some commodities that were necessarily private when single now become public. Examples include a ‘clean house’ or managing finances. This changes the mix of goods that are optimal for co-habiting couples as compared to when they were single which in turn has an impact on time use within the household; see Lam (1988), Deaton and Paxson (1998), Crossley and Lee (2004), and Browning et al. (2014). A third effect of state dependence arises if preferences change on living together (Michaud and Vermeulen 2011, Browning et al. 2013, and Cherchye et al. 2016). An example would be that eating at home is now relatively more attractive than eating in restaurants when single. If the change in preferences is towards commodities that are produced using housework, then this will increase time spent on housework (Stratton 2012, 2015). Finally, if we allow for net affect as suggested by Kahneman and Krueger (2006), doing housework together may be more or less onerous than doing Electronic copy available at: https://ssrn.com/abstract=3524799 16 decrease (increase) it slightly in the BHPS-UKHLS (HILDA). Consistent with the literature (Corrigall and Konrad 2007, Vespa 2009), columns 4 and 5 in Table 6 show that gender role attitudes become less egalitarian for women upon marriage. Women marrying during the sample period in the BHPS-UKHLS (HILDA) go from 0.609 (0.481) to 0.173 (0.149) in a 0/1 scale. a are always single during the sample period. Also, as previously suggested, men’s role attitudes become more egalitarian upon marriage, from 0.269 (2.242) to 0.377 (0.289) for those marrying in the BHPSUKHLS (HILDA) - see columns 9 and 10 in Table 6. Results in column 1 of Table 7 showing the marriage coefficient from estimating Equation 3 in a model additionally controlling for an individual’s egalitarian index (shown as the regression analysis with Group 2 Controls), suggest that changes in gender role attitudes are not responsible for changes in housework time upon marriage. The fact that women’s gender role attitudes become less egalitarian upon marriage (shown in Table 6) has virtually no impact on the marriage coefficients of Table 7; for men, controlling for individual gender role attitudes only increases the size of the reduction in routine housework time upon marriage. In consequence, the marriage coefficient continues to be significant in the OLS specification after controlling for gender role attitudes, and diminishes further once we control for the individual FE as before. Robustness Checks As in the cross-sectional results in Section 2, we can rule out that our results are driven by transitions into having children rather than into marriage. Our main results continue to hold when individuals without children are considered. The FE coefficients in Columns 1 and 4 in Table 8 for the sample of individuals without children are very similar to the FE coefficients for the whole sample of individuals presented in Table 5. We still observe selection on permanent unobservable characteristics for this subsample of non-parents. In particular, selection explains between 55% and 63% of all the increase in housework for women, and as for the results using the whole sample of individuals, controlling for fixed effects can explain over 100 percent of the decrease in housework time upon marriage for men in the PSID and HILDA and about 50 per cent in the BHPS-UKHLS. Similarly, when we select the sample of cohabiting individuals, results continue to hold (Columns 2 and 5 of Table 8). In particular, we consider always-single individuals (i.e., single individuals not living in the parental home), always-cohabiting individuals (i.e., individuals who are in a partnership but not legally married), and individuals entering into cohabitation during the sample period. We exclude transitions from cohabitation into legal marriage. 6 As in the cross-sectional 6 Results are robust to using a different sample selection where those ever legally married are included. In the PSID cohabiting is not a special marital status category but can be computed using available information. First, when Electronic copy available at: https://ssrn.com/abstract=3524799 17 results, the OLS coefficient in the cohabitating sample is smaller than the OLS coefficient when the whole sample of women is considered, and larger for men. However, the FE coefficients are remarkably similar to those obtained for the married (legally married and cohabiting) sample reported in Table 5. Interestingly, the selection effect is generally lower for the sample of cohabiting individuals than for the whole sample. It even turns negligible for the PSID, drops to 36% from the previous 53% in UKHLS-BHPS, and remains roughly similar around 70% in HILDA. 7 4.3 Understanding the Mechanisms behind the Fixed Effects A possible mechanism to explain the above findings is that women are more likely to marry when they have an ex-ante higher preference for routine housework, whereas men with a lower taste for routine housework are more likely to tie the knot. In other words, a necessary condition is that individuals who have a more traditional view on the division of household labour are more likely to marry. If this is the case, we should expect the selection effect to be higher for individuals holding more traditional gender roles about what being a “husband” and a “wife” means (Akerlof and Kranton, 2000). Several pieces of evidence from Tables 5 and 8 seem to corroborate this gender roles hypothesis. First, the smaller explanatory power of selection into marriage for the UK, compared to Australia and the US shown in Table 5, is consistent with more traditional gender roles in Australia and the US vs in the UK. Standard measures of gender equality used for example in Guiso et al., (2008) and Bertrand (2011) show that in 2009 in the UK a majority of 77% of individuals disagreed or totally disagreed with the statement “When jobs are scarce, men should have more right to a job than women”, compared to 73% in Australia and 69% in the US. Similarly, the ratio of women to men in ministerial-level and parliamentary positions is around 30% in the UK, compared to less than 20% in Australia and about 16% in the US (See Table C.8). Second, consistent with the gender roles explanation the selection effect is lower for younger individuals (columns 3 and 6 of Table 8), as well as for cohabiting individuals (columns 2 and 5 in Table 8). Younger individuals are more likely to hold less traditional gender roles, based on existing cohabitation lasts for longer than 12 months the study starts recording information on both members of the couple, just as if a legal marriage had taken place. If not part of the PSID sample, the male cohabitator turns head and the female cohabitator turns “wife”. Second, to detect first-year cohabitations we can use information on other household members. The PSID records whether a boyfriend or girlfriend of the household head lives in the residence, and, therefore, we can know whether the household head is cohabiting. The PSID does not record housework or labour market information for these first-year cohabitors, what may partly explain the lower sample size of cohabiting individuals in the PSID compared to the BHPS and HILDA. In the BHPS and HILDA there are differentiated categories for those legally married and living as couple (categories 1 and 2 when asked about marital status). 7 Regression results using a PSID sample from 1985, the first year the housework variable was asked in the PSID, confirm that our results are robust to the inclusion of earlier years (see Appendix Table C.6). Results are also robust to controlling for the type of dwelling and the type of ownership of the household home (see Appendix Table C.7). Electronic copy available at: https://ssrn.com/abstract=3524799 18 evidence showing that holding less egalitarian attitudes has declined across cohorts over the last years (see Fortin 2005 and Fernandez 2013). Similarly, cohabiting individuals are less likely to hold traditional gender roles than legally married individuals, as shown by their lower degree of couple specialization (Barg and Beblo, 2012). Lastly, the gender roles explanation implies that women with a higher overall preference for housework have a higher preference for marriage, which results in a negative correlation between an individual’s egalitarian gender role attitudes and her predicted fixed effect from running Equation (3) in Column 1 of Table 5. Conversely, the gender roles hypothesis implies that men with a lower overall preference for housework have a higher propensity to marriage, resulting in a positive correlation between an individual’s egalitarian gender role attitudes and his predicted fixed effect from running Equation (3) in Column 2 of Table 5. To test this hypothesis, we run the following OLS regression: 𝐼𝐹𝐸𝑖= 𝛼 + 𝛽2 ′ 𝑋𝑖+ 𝛽3𝐸𝐺𝐴𝐿𝑖+ 𝜀𝑖 (4) where 𝐼𝐹𝐸𝑖 denotes the predicted individual fixed effect term of individual i from Columns 1 and 2 in Table 5. 𝑋𝑖 is a vector of covariates as in Equation (3) that includes average values for a quadratic on age, education at age 25 in years, an interaction term of age and years of education, number of children, hours of paid work, predicted wages, and the number of rooms in the house to control for observed variables that can affect both the housework preference and the egalitarian attitudes. The individual-specific egalitarian index 𝐸𝐺𝐴𝐿𝑖, is a dummy variable that takes value 1 if the average (across waves) of the individual’s egalitarian index in Section 4.2 is greater than the sample mean. As in the main analysis, we restrict the sample to individuals between 24 and 65 years of age not living in the parental home for which we have information for all the variables, including the egalitarian index, and who answered gender role attitudinal questions at least once 8 Columns 1 and 2 in Table 9 show that in both BHPS-UKHLS and HILDA samples, more egalitarian women have a lower predicted fixed effect. The negative coefficient on the egalitarian dummy for women suggests that women more likely to marry have higher preference for housework. Conversely, and also consistent with the gender roles explanation, Columns 3 and 4 in Table 9 show that in both BHPS-UKHLS and HILDA samples, more egalitarian men have a lower predicted fixed effect. The positive coefficient on the egalitarian dummy for men suggests that men marrying are the ones with a lower preference for housework (i.e., less egalitarian). The absolute values of the 8 For the BHPS-UKHLS, we are unable to compute the index for between 4 and 5 per cent of the sample, who never responded to the values questions; in HILDA, between 5 and 6 per cent of the sample could not be used for this analysis for the same reasons. The results in Table 5 hold. Electronic copy available at: https://ssrn.com/abstract=3524799 19 coefficients on the egalitarian dummy are larger for women than for men, indicating that the selection with respect to gender attitudes may be stronger for women. These findings are consistent with the gender roles explanation, and suggestive that the fixed effect in Equation (3) is not capturing a marriage wage premium (penalty) for men (women). A necessary condition for this alternative explanation to be a valid mechanism driving selection is that women with a higher preference for housework tasks before marriage are perceived as more desirable partners in the marriage market, and similarly that men with a lower preference for housework tasks before marriage are perceived as more desirable partners. This last condition is however unlikely, as there is now mounting evidence that men who contribute more to home production are more likely to enter a partnership (see Sevilla-Sanz, 2010; Gimenez-Nadal, J.I., Molina, J. and Sevilla, A., 2012; Burda et al., 2013; and Bertrand et al., 2016). All in all, results from fixed effects estimates using longitudinal data for the US, the UK, and Australia in Tables 5-8 suggest that, to the extent that there are no other omitted variables that could be captured by the individual fixed effect, selection into marriage can account for a substantial portion of the cross-section variation in housework upon marriage, particularly with respect to routine housework activities such as cooking and cleaning. Additionally, the results seem to suggest that the selection runs in opposite directions for women and men. 4.4 Transitions into Marriage and Non-Routine Housework Results based on routine housework seem to suggest a gender imbalance with respect to the effects of marriage on housework time. In contrast to women, where there is a genuine increase in routine housework time upon marriage, routine housework upon marriage may stay the same (in Australia) or decrease (in the UK) for men. However Table 3 documented that married men did more non-routine housework than single men, so routine housework may not give a full picture of the variation of housework time. We observe a measure of non-routine housework in HILDA, which relates to outdoor housework activities such as gardening and home repairs. In particular, HILDA asks “How much time would you spend on Outdoor tasks, including home maintenance (repairs, improvements, painting etc.), car maintenance or repairs and gardening in a typical week?” Results in Table 10 provide a fuller picture of the effect of marriage on housework time by looking at what happens to non-routine housework upon marriage. Column (1) in Table 10 shows that the hour per week increase of routine housework for married women relative to single women observed in Table 5 is compensated by a similar increase in non-routine housework for married men Electronic copy available at: https://ssrn.com/abstract=3524799 20 relative to single men, which increases by 1.32 hours per week (Column 2). Thus, the net effect on housework time from marriage seems to be similar for men and women once a more comprehensive measure of housework is considered. In particular, total housework increases by 1.08 hours per week for women, and by 1.49 hours per week for men (Column 3). Compared to Column (1), 𝜒2-tests comparing OLS and FE coefficients in Column (2) in Table 10 reveal no selection effects in nonroutine housework arising from unobserved permanent heterogeneity, for either men or women, suggesting that the increases in non-routine housework for men are genuine and happen as a result of marrying. 5. Conclusions This paper documents changes in home labour upon forming a household and explores in depth one possible explanation: selection effects arising from unobserved heterogeneity. Using crosssectional data for a wide range of high-income countries we document that married women do about five more hours per week of housework than single women. Compared to married women, men’s housework time is half an hour per week higher when married than when single. Differences in housework across marital states persist after conditioning on a wide set of time-varying and timeinvariant observable characteristics such as age, education, and the number and age of children. These findings follow even after controlling for time spent in the labour market, and for different sub-groups of individuals. Results from fixed effects estimates using longitudinal data for the US, the UK, and Australia suggest that permanent unobserved heterogeneity can account for about 50 and 70 per cent of the increase of routine housework upon marriage for women, and all the decrease in routine housework upon marriage observed in the cross-section for men. There is no evidence that other unobserved heterogeneity, such as gender roles, may be biasing our marriage coefficient even after controlling for permanent unobserved heterogeneity as in Equation 3. All in all our results suggest that the changes in housework observed in the cross-section may be partly driven by individuals with a higher preference for marriage also having a more traditional view on the division of household labour, rather than being more highly valued in the marriage market. Upon marriage there remains a genuine increase of non-routine housework for men and of routine housework for women of between one hour and a half and two hours per week. Women increase routine housework upon marriage, such as meal preparation and cleaning. In contrast men increase the time spent in non-routine housework (such as outdoor and maintenance activities). Compared to non-routine housework, routine housework needs to be performed on job days and is difficult to postpone or contract out and is more likely to impose a penalty upon labour market Electronic copy available at: https://ssrn.com/abstract=3524799 21 activities and wages (see Hersch (1991) and Hersch and Stratton (2002)). The gender specialization across housework tasks upon marriage cannot thus be overlooked, and should inform economic models and the design of policies interested in the distributional effects within households. The causal positive effect of marriage on housework identified here supports the view that gains from marriage go beyond the efficiency gains derived from specialization in paid work. As previously suggested in the literature, the production and consumption of household public goods may be an important source of marital gains. Economists have devoted time and effort to identify the economies of scale associated to changes in the size of the household, which rest on the idea that two can live more cheaply than one (for instance Deaton and Paxson, 1997). The focus in the literature is usually on monetary expenditures. Our findings point towards a substitution of market goods towards home produced goods upon marriage, and highlight the importance of incorporating into household models economic activity that escapes the market economy (see Krueger et al. (2009)). Electronic copy available at: https://ssrn.com/abstract=3524799 22 References Aguiar, M. and Hurst, E. 2007. “Measuring Trends in Leisure: The Allocation of Time Over Five Decades” Quarterly Journal of Economics, 122: 969-1006. Auspurg, K., Iacovou, M., and Nicoletti, C. 2014. "Housework Share between Partners: Experimental Evidence on Gender Identity," IZA Discussion Papers 8569. Becker, G. 1965. “A Theory of the Allocation of Time” Economic Journal, 75: 493–517. Becker, G. 1981. A treatise on the Family. Cambridge, MA: Harvard University Press. Becker, G. 1985. “Human Capital, Effort, and the Sexual Division of Labor”, Journal of Labor Economics, 3: no. 1, part 2: 33-58. Bertrand, M., Cortes, P., Olivetti, C., and Pan, J. 2016. "Social Norms, Labor Market Opportunities, and the Marriage Gap for Skilled Women," NBER Working Papers 22015. Browning, M., Chiappori, P.A., and Lewbel, A. 2013. "Estimating Consumption Economies of Scale, Adult Equivalence Scales, and Household Bargaining Power," Review of Economic Studies, 80(4), 1267-1303. Browning, M., Chiappori, P.A., and Weiss, Y. 2014. Economics of the Family, Cambridge, MA: Cambridge University Press. Bruze, G., Svarer, M., and Weiss, Y. 2015. “The Dynamics of Marriage and Divorce.” Journal of Labor Economics 33:1, 123-170 Burda, M., Hamermesh, D. and Weil, P. 2013. "Total work and gender: facts and possible explanations," Journal of Population Economics, 26(1), 239-261. Cherchye, L., De Rock, B., Surana, K., and Vermeulen, F. 2016. "Marital Matching, Economies of Scale and Intrahousehold Allocations," IZA Discussion Papers 10242, Institute for the Study of Labor (IZA). Chiappori, P.A., Fortin, B. and Lacroix, G. 2002. ‘Marriage market, divorce legislation, and householdlabor supply’, Journal of Political Economy, vol. 110(1), pp. 37–72. Chiappori, P.A., Iyigun, M., Lafortune, J. and Weiss, Y. 2017. “Changing the Rules Midway: The Impact of Granting Alimony Rights on Existing and Newly Formed Partnerships.” The Economic Journal. doi:10.1111/ecoj.12385 Corrigall, E.A. & Konrad, A.M. Sex Roles (2007) 56: 847 Couprie, H. 2007. “Time allocation within the Family: Welfare implications of life in a couple.” The Economic Journal, 117: 287–305. Electronic copy available at: https://ssrn.com/abstract=3524799 23 Corrigall, E. A., & Konrad, A. M. (2007). Gen-der role attitudes and careers: A longitudinal study.Sex Roles, 56, 847–855. Crossley, T. and Lu, Y. 2005. "Exploring the returns to scale in food preparation (baking penny buns at home)," IFS Working Papers W05/03, Institute for Fiscal Studies. Deaton, A., and Paxson, C. 1998. “Economies of Scale, Household Size, and the Demand for Food.” Journal of Political Economy, 106, 897–930. Department of Social Services; Melbourne Institute of Applied Economic and Social Research, 2017, "The Household, Income and Labour Dynamics in Australia (HILDA) Survey, GENERAL RELEASE 13. (Waves 1-13)", doi:10.4225/87/VHRTR5, ADA Dataverse, V4 Eckstein, Z. , Keane, M. and Lifshitz, O. 2019. Career and Family Decisions: Cohorts Born 1935– 1975. Econometrica, 87: 217-253. Fisher, K., Jonathan Gershuny, Sarah M. Flood, Joan Garcia Roman, and Sandra L. Hofferth. Multinational Time Use Study Extract System: Version 1.2 [dataset]. Minneapolis, MN: IPUMS, 2018. https://doi.org/10.18128/D062.V1.2 Foster, G. and Kalenkoski, C. (2013). “Tobit or OLS? An empirical evaluation under different diary window lengths.” Applied Economics, 45, 2994–3010. Giménez-Nadal, J.I. and Sevilla, A. 2012. “Trends in Time Allocation: A Cross-Country Analysis”, European Economic Review, 56(6), 1338-1359. Gronau, R. 1977. “Leisure, Home Production, and Work-The Theory of the Allocation of Time Revisited” Journal of Political Economy, 85(6): 1099-1123. Grossbard-Shechtman, A. S. 1984. A theory of allocation of time in markets for labor and marriage. Economic Journal, 94, 863–882. Gupta, S. 1999. “The Effects of Transitions in Marital Status on Men's Performance of Housework.” Journal of Marriage and the Family 61:700-711. Guryan, J., Hurst, E. and Kearney, M. (2008). “Parental education and parental time with children,” Journal of Economic Perspectives, Summer 2008: 23-46. Hamermesh, D., 2002. “Timing, togetherness and time windfalls.” Journal of Population Economics 15, 601–623. Hersch, J. 1991. “Male-Female Differences in Hourly Wages: the Role of Human Capital, Working Conditions, and Housework.” Industrial and Labor Relations Review 44(4): 746-759. Electronic copy available at: https://ssrn.com/abstract=3524799 24 Hersch, J. and Stratton, L. 2002. “Housework and Wages.” Journal of Human Resources 37(1): 217- 229. Jenkins, S.P. and Osberg, L. 2004. “Nobody to play with? The implications of leisure coordination.” In: Hamermesh , Daniel and Pfann, Gerard, (eds.) The Economics of Time Use: Contributions to Economic Analysis, 271. Elsevier, Amsterdam, The Netherlands, pp. 113-145. Kahneman, D. and Krueger, A.B. 2006."Developments in the Measurement of Subjective Well- Being," Journal of Economic Perspectives, 20(1), 3-24. Kahneman, D., Krueger, A.B., Schkade, D., Schwarz, N., and Stone, A. 2004. "A Survey Method for Characterizing Daily Life Experience: The Day Reconstruction Method (DRM)", Science, 306(5702), 1776-80. Kantar Public, NatCen Social Research, University of Essex. Institute for Social and Economic Research. (2018). Understanding Society: Waves 1-7, 2009-2016 and Harmonised BHPS: Waves 1-18, 1991-2009. [data collection]. 10th Edition. UK Data Service. SN: 6614, http://doi.org/10.5255/UKDA-SN-6614-11 Knowles, J. 2013. “Why are Married Men Working So Much? An Aggregate Analysis of Intra- Household Bargaining and Labour Supply”, The Review of Economic Studies, 80(3): 1055– 1085 Krueger, A.B., Kahneman, D., Schkade, D., Schwarz, N., and Stone, A., 2009. "National Time Accounting: The Currency of Life," NBER Chapters, in: Measuring the Subjective Well-Being of Nations: National Accounts of Time Use and Well-Being, pages 9-86, National Bureau of Economic Research, Inc. Lam, D. 1988. "Marriage Markets and Assortative Mating with Household Public Goods: Theoretical Results and Empirical Implications," Journal of Human Resources, 23(4), 462-487. Lundberg, S. 2012. "Personality and marital surplus," IZA Journal of Labor Economics,. 1(1), 1-21. Michaud, P.C. and Vermeulen, F. 2011. "A collective labor supply model with complementarities in leisure: Identification and estimation by means of panel data," Labour Economics, 18(2), 159- 167. Pollak, R.A. 2012. "Allocating Time: Individuals' Technologies, Household Technology, Perfect Substitutes, and Specialization," Annals of Economics and Statistics, 105-106, 75-97 Pollak, R.A. 2013. "Allocating Household Time: When Does Efficiency Imply Specialization?," NBER Working Papers 19178, National Bureau of Economic Research, Inc. Electronic copy available at: https://ssrn.com/abstract=3524799 25 Ramey, G. and Ramey, V. A. 2010. “The Rug Rat Race”. Brookings Papers on Economic Activity. Spring 129-176 Sevilla, A., 2010. “Household division of labor and cross-country differences in household formation rates.” Journal of Population Economics 23, 225-249. Stancanelli, E. G. F. and Stratton, L. 2014. “Maids, Appliances and Couples' Housework: The Demand for Inputs to Domestic Production.” Economica, 81: 445–467. Stratton, L. 2012. “The role of preferences and opportunity costs in determining the time allocated to housework.” American Economic Review, Papers and Proceedings, 102(3), 606–11. Stratton, L. 2015. “The determinants of housework time” IZA World of Labour 133:1-10. Sullivan, O. 1996. “Time co-ordination, the domestic division of labour and affect relations: time-use and the enjoyment of activities within couples”, Sociology, 30 (), pp. 79.100. University of Michigan. Survey Research Center. Panel Study of Income Dynamics (PSID): Main Interview, 1968-2015. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2018-10-04. https://doi.org/10.3886/ICPSR37142.v1 Vespa, J. (2009). Gender Ideology Construction: A Life Course and Intersectional Approach. Gender & Society, 23(3), 363–387. Vernon, V. 2010. "Marriage: for love, for money…and for time?," Review of Economics of the Household, Springer, vol. 8(4), pages 433-457, December. Electronic copy available at: https://ssrn.com/abstract=3524799 32 Table 4. Hours per week in Routine Housework by Marital Status: Evidence from longitudinal data Women Men (1) (2) (3) (4) (5) (6) Observations Single Married Observations Single Married Panel A. PSID Always single (not in parental home) 286 8.91 327 6.86 Marrying (not from parental home) 153 7.99 12.09 205 6.25 8.02 Always married 3106 18.45 3663 7.07 Panel. B. UKHLS Always single (not in parental home) 260 10.40 327 7.11 Marrying (not from parental home) 146 9.84 12.62 177 6.35 5.74 Always married 3073 17.66 3050 5.39 Panel. C. HILDA Always single (not in parental home) 410 10.26 470 7.11 Marrying (not from parental home) 225 10.03 13.05 264 5.89 6.57 Always married 3959 18.12 3863 6.18 Notes: Data comes from the 1992-2015 PSID, the 1991-2015 Harmonized BHPS-UKHLS, and the 2002-2013 HILDA. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting), and marrying (i.e. individuals who transition from single to married). Housework is reported in hours per week as the answer to the question “About how much time do you spend on housework in an average week? I mean time spent cooking, cleaning, and doing other work around the house?", in PSID, “About how many hours do you spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry?”, in UKHLS, and “About how many hours do you spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry?” (variable lshw), in HILDA. Electronic copy available at: https://ssrn.com/abstract=3524799 33 Table 5. Changes of housework upon marriage: Evidence from longitudinal data (1) (2) Women Men Coefficien t Standard Error % Explained Coefficien t Standard Error % Explained Panel A: PSID OLS without controls 9.56 (0.39)*** 0.41 (0.25) OLS with controls 4.66 (0.43)*** 51% -0.12 (0.28) > 100% FE 1.55 (0.60)*** 67% 0.94 (0.40)** > 100% Chi2 test of equality of coefficients FE-OLS with controls 29.39 (0.000) 5.47 (0.019) N obs. 3545 4195 Panel B: UKHLS OLS without controls 7.24 (0.47)*** -1.56 (0.22)*** OLS with all controls 3.87 (0.42)*** 47% -1.44 (0.24)*** 8% FE 1.83 (0.40)*** 53% -0.73 (0.34)** 49% Chi2 test of equality of coefficients FE-OLS with controls 15.44 (0.000) 3.57 (0.058) N obs. 3479 3554 Panel C: HILDA OLS without controls 7.73 (0.38)*** -0.64 (0.20)*** OLS with all controls 3.49 (0.35)*** 55% -0.83 (0.21)*** -30% FE 1.05 (0.43)** 70% 0.11 (0.27) > 100% Chi2 test of equality of coefficients FE-OLS with controls 20.10 (0.000) 8.56 (0.003) N obs. 4594 4597 Notes: Data comes from the 1992-2015 PSID, the 1991-2015 Harmonized BHPS-UKHLS and the 2002-2016 HILDA. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting) and marrying (i.e. individuals who transition from single to married). Housework is reported in hours per week as the answer to the question “About how much time do you spend on housework in an average week? I mean time spent cooking, cleaning, and doing other work around the house?", in PSID, “About how many hours do you spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry?”, in UKHLS, and “How much time would you spend on housework, such as preparing meals, washing dishes, cleaning house, washing clothes, ironing and sewing, in a typical week?” (variable lshw), in HILDA. The regressions of housework on a dummy for living with a partner (cohabiting or being married). Controls include a quadratic on age, education at age 25 in years, an interaction term of age and years of education, predicted wages, a logarithm in the number of children, hours of paid work, and number of household rooms. Percentage explained by controls obtained as the difference between the OLS coefficient without controls and the OLS coefficient with controls, divided by the OLS coefficient without controls; percentage explained by FE Electronic copy available at: https://ssrn.com/abstract=3524799 34 obtained as the difference between the OLS coefficient with controls and the FE coefficient, divided by the OLS coefficient with controls. In both cases we denote >100% when the formula yields a number greater than 100%. Standard errors in brackets. *** Significant at the 1 percent level; ** significant at the 5 percent level;* significant at the 10 percent level. Electronic copy available at: https://ssrn.com/abstract=3524799 35 Table 6. Routine Housework and Egalitarian Attitudes by Marital Status: Evidence from longitudinal data Women Men (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Observations Housework Egalitarian Index Observations Housework Egalitarian Index Single Married Single Married Single Married Single Married Panel. A. BHPS-UKHLS Always single (not in parental home) 164 10.86 0.009 178 6.93 -0.034 Marrying (not from parental home) 68 8.79 12.05 0.609 0.173 96 6.29 5.58 0.269 0.377 Always married 2336 18.26 -0.041 2300 5.38 0.003 Panel. B. HILDA Always single (not in parental home) 189 10.58 0.226 188 7.06 0.089 Marrying (not from parental home) 63 10.92 13.60 0.481 0.149 52 6.25 6.35 0.242 0.289 Always married 2142 18.46 0.060 2014 6.28 0.114 Notes: Data comes from the 1993, 1995, 1997, 1999, 2001, 2003, and 2005 Harmonized BHPS-UKHLS and the 2005, 2008, 2011, and 2015 HILDA. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting) and marrying (i.e. individuals who transition from single to married). Housework is reported in hours per week as the answer to the question “About how much time do you spend on housework in an average week? I mean time spent cooking, cleaning, and doing other work around the house?", in PSID, “About how many hours do you spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry?”, in UKHLS, and “How much time would you spend on housework, such as preparing meals, washing dishes, cleaning house, washing clothes, ironing and sewing, in a typical week?” (variable lshw), in HILDA. The egalitarian index is the first principal component of the eight (seventeen) questions on the individual’s attitudes towards the division of household labour separately by gender for each of the 3, 5, 7, 9, 11, 13, and 15 (5, 8, 11, and 15) waves in the BHPS-UKHLS (HILDA). Electronic copy available at: https://ssrn.com/abstract=3524799 36 Table 7. Housework and Marriage: Evidence from Longitudinal Data. Controlling for Changes in Family Attitudes upon Marriage (1) (2) Women Men Coefficient Standard Error % Explained Coefficient Standard Error % Explained Panel B: BHPS-UKHLS OLS without controls 7.67 (0.60)*** -1.43 (0.26)*** OLS control group 1 3.96 (0.55)*** 48% -1.41 (0.27)*** 1% OLS control group 2 4.00 (0.56)*** -1% -1.54 (0.27)*** -9% Chi2 test of equality of coeff. OLS group1-OLS group 2 0.75 (0.388) 10.21 (0.001) FE 1.73 (0.62)*** 57% -0.50 (0.60) 68% Chi2 test of equality of coeff. FE-OLS control group 2 8.79 (0.003) 2.86 (0.091) N obs. 2568 2569 Panel C: HILDA OLS without controls 7.76 (0.53)*** -0.69 (0.35)** OLS control group 1 3.04 (0.51)*** 61% -0.77 (0.37)** -12% OLS control group 2 3.02 (0.50)*** 1% -0.85 (0.37)** -10% Chi2 test of equality of coeff. OLS group1-OLS group 2 0.19 (0.661) 3.40 (0.065) FE -0.11 (1.19) > 100% -0.23 (0.69) 73% Chi2 test of equality of coeff. FE-OLS control group 2 6.70 (0.009) 0.70 (0.403) N obs. 2394 2254 Notes: Data comes from the 1993, 1995, 1997, 1999, 2001, 2003, and 2005 Harmonized BHPS-UKHLS and the 2005, 2008, 2011, and 2015 HILDA. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting) and marrying (i.e. individuals who transition from single to married). Housework is reported in hours per week as the answer to the question “About how much time do you spend on housework in an average week? I mean time spent cooking, cleaning, and doing other work around the house?", in PSID, “About how many hours do you spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry?”, in UKHLS, and “How much time would you spend on housework, such as preparing meals, washing dishes, cleaning house, washing clothes, ironing and sewing, in a typical week?” (variable lshw), in HILDA. The regressions of housework on a dummy for living with a partner (cohabiting or being married). Controls in group 1 include a quadratic on age, education at age 25 in years, an interaction term of age and years of education, predicted wages, a logarithm in the number of children, hours of paid work, and number of household rooms. Controls in group 2 additionally include the first principal components of the eight (seventeen) questions on the individual’s attitudes towards the division of household labour separately by gender for each of the 3, 5, 7, 9, 11, 13, and 15 (5, 8, 11, and 15 ) waves in the UKHLS (HILDA). Percentage explained by controls obtained as the difference between the OLS coefficient without controls and the OLS coefficient with controls, divided by the OLS coefficient without controls; percentage explained by FE obtained as the difference Electronic copy available at: https://ssrn.com/abstract=3524799 37 between the OLS coefficient with controls and the FE coefficient, divided by the OLS coefficient with controls. In all cases we denote >100% when the formula yields a number greater than 100%. Standard errors in brackets. *** Significant at the 1 percent level; ** significant at the 5 percent level;* significant at the 10 percent level. Electronic copy available at: https://ssrn.com/abstract=3524799 38 Table 8. Robustness Checks. Changes of housework upon marriage: Evidence from longitudinal data. Different samples. (1) (2) (3) (4) (5) (6) Women Men Without children Cohabitation 25-44 year-olds Without children Cohabitation 25-44 year-olds Coeff. St. Error Coeff. St. Error Coeff. St. Error Coeff. St. Error Coeff. St. Error Coeff. St. Error Panel A: PSID OLS no controls 8.76 (0.39)*** 3.86 (0.69)*** 9.76 (0.40)*** 0.09 (0.28) 0.58 (0.38) 0.60 (0.25)** OLS with controls 4.21 (0.44)*** 2.65 (0.68)*** 3.81 (0.46)*** -0.21 (0.30) 0.07 (0.42) -0.23 (0.27) % Explained by controls 52% 31% 61% > 100% 88% > 100% FE 1.9 (0.51)*** 2.69 (0.81)*** 1.7 (0.65)*** 0.97 (0.45)** 1.49 (0.87)* 0.75 (0.43)* % Explained by fixed effects 55% -2% 55% > 100% <-100% > 100% Chi2 test of equality of coeff. 14.68 (0.000) 0.00(0.969) 7.62 (0.006) 5.37 (0.020) 2.97 (0.085) 4.50 (0.034) N obs. 2303 557 2863 2443 962 2912 Panel B: UKHLS OLS no controls 7.19 (0.41)*** 2.65 (0.52)*** 7.4 (0.52)*** -1.64 (0.23)*** -1.20 (0.25)*** -0.95 (0.21)*** OLS with controls 4.26 (0.42)*** 1.88 (0.44)*** 2.62 (0.42)*** -1.45 (0.24)*** -1.10 (0.27)*** -0.71 (0.22)*** % Explained by controls 41% 29% 65% 12% 8% 25% FE 1.85 (0.38)*** 1.20 (0.50)** 1.72 (0.43)*** -0.72 (0.39)* -0.54 (0.46) -0.49 (0.34) % Explained by fixed effects 57% 36% 34% 50% 51% 31% Chi2 test of equality of coeff. 19.56 (0.000) 1.35 (0.247) 2.81 (0.094) 3.23 (0.072) 1.46 (0.227) 0.37 (0.542) Nobs. 2686 1149 2469 2702 1296 2399 Panel C: HILDA OLS no controls 6.57 (0.36)*** 3.38 (0.45)*** 7.91 (0.46)*** -0.26 (0.18) -0.36 (0.24) 0.14 (0.19) OLS with controls 3.16 (0.35)*** 2.54 (0.40)*** 3.03 (0.42)*** -0.24 (0.19) -0.30 (0.24) -0.11 (0.20) % Explained by controls 52% 25% 62% 8% 17% > 100% FE 1.16 (0.43)*** 0.61 (0.47) 1.26 (0.47)*** 0.45 (0.30) 0.19 (0.32) 0.07 (0.28) % Explained by fixed effects 63% 76% 58% > 100% > 100% > 100% Chi2 test of equality of coeff. 15.19 (0.001) 10.14 (0.001) 8.32 (0.004) 4.37 (0.037) 1.73 (0.188) 0.32 (0.569) Nobs. 3387 1831 3057 1707 2016 2869 Notes: Data comes from the 1992-2015 PSID, the 1991-2015 Harmonized BHPS-UKHLS, and the 2002-2016 HILDA. In Columns 1 and 4 the sample consists of individuals between 24 and 65 years old, not living in the parental home, who have not been divorced, who are observed in at least two waves and who are not living with any child under 18 years of age. In Columns 2 and 5 the sample consists of individuals between 24 and 45 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. In columns 3 and 6 the sample consists of individuals between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are legally married) and marrying (i.e. individuals who transition from single to legally married). Housework is reported in hours per week as the answer to the question “About how much time do you spend on housework in an average week? I mean time spent cooking, cleaning, and doing other work around the house?", in PSID, “About how many hours do you spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry?”, in UKHLS, and “How much time would you spend on housework, such as preparing meals, washing dishes, cleaning house, washing clothes, ironing and sewing, in a typical week?” (variable lshw), in HILDA. The regressions of housework on a dummy for living with a partner (cohabiting or being married). Controls include a quadratic on age, education at age 25 in years, an interaction term of age and years of education, predicted wages, a logarithm in the number of children, hours of paid work, and number of household rooms. Percentage explained by controls obtained as the difference between the OLS coefficient without controls and the OLS coefficient with controls, divided by the OLS coefficient without controls; percentage explained by FE obtained as the difference between the OLS coefficient with controls and the FE coefficient, divided by the OLS coefficient with controls. In both cases we denote >100% when the formula yields a number greater than 100%. Standard errors in brackets. *** Significant at the 1 percent level; ** significant at the 5 percent level;* significant at the 10 percent level. Electronic copy available at: https://ssrn.com/abstract=3524799 39 Table 9. Housework, Marriage and Egalitarian Attitudes Women Men (1) (2) (3) (4) Specif. 1 Specif. 2 Specif. 1 Specif. 2 Panel A: BHPS-UKHLS Egalitarian attitudes dummy -8.25*** -1.13*** 0.37*** 0.86*** (0.52) (0.20) (0.11) (0.11) Other controls No Yes No Yes R-squared 0.090 0.889 0.005 0.120 Observations 2568 2568 2548 2548 Panel B: HILDA Egalitarian attitudes dummy -3.99*** -1.12*** 0.87*** 0.66*** (0.19) (0.15) (0.08) (0.08) Other controls No Yes No Yes R-squared 0.103 0.500 0.031 0.132 Observations 3950 3950 3845 3845 Notes: Data comes from the 1991-2015 Harmonized BHPS-UKHLS and the 2002-2016 HILDA. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, who responded to family attitudes questions in at least one of the waves, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting) and marrying (i.e. individuals who transition from single to married). The dependent variables are predicted individual fixed effects from a fixed effect regression of housework on a dummy for living with a partner (cohabiting or being married), a quadratic on age, education at age 25 in years, an interaction term of age and years of education, a logarithm in the number of children, number of household rooms, weekly work hours and predicted wages. Specification 2 additionally controls for average values of age, education at age 25 in years, an interaction term of age and years of education, a logarithm in the number of children, number of household rooms, weekly work hours, and predicted wages. Standard errors in brackets. *** Significant at the 1 percent level; ** significant at the 5 percent level;* significant at the 10 percent level. Electronic copy available at: https://ssrn.com/abstract=3524799 40 Table 10: Changes of housework upon marriage by type of Housework (1) (2) (3) Routine Housework Non-routine Housework Total Housework Coeffic ient Standard Error % Explaine d Coeffic ient Standard Error % Explaine d Coeffic ient Standard Error % Explaine d Panel A. Women OLS without controls 7.73 (0.38)*** 0.48 (0.19)** 8.21 (0.50)*** OLS with controls 3.49 (0.35)*** 55% 0.06 (0.19) 88% 3.54 (0.45)*** 57% FE 1.05 (0.43)** 70% 0.03 (0.30) 50% 1.08 (0.60)* 69% Chi2 test of equality of coeff. FEOLS with controls 20.10 (0.000) 0.01 (0.937) 11.49 (0.000) Nobs. 4594 4594 4594 Panel B. Men OLS without controls -0.64 (0.20)*** 2.36 (0.18)*** 1.72 (0.32)*** OLS with controls -0.83 (0.21)*** -30% 1.91 (0.22)*** 19% 1.08 (0.35)*** 37% FE 0.11 (0.27) > 100% 1.39 (0.34)*** 27% 1.49 (0.47)*** -38% Chi2 test of equality of coeff. FEOLS with controls 8.56 (0.003) 1.86 (0.172) 0.55 (0.457) Nobs. 4597 4597 4597 Notes: Data comes from the 2002-2016 HILDA. Sample is women between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting) and marrying (i.e. individuals who transition from single to married).All housework variables are reported in hours per week and constitute the answer to the question “How much time would you spend on each of the following activities in a typical week?”. The activity for Column 1 is “Housework, such as preparing meals, washing dishes, cleaning house, washing clothes, ironing and sewing”. In Column 2 it is “Outdoor tasks, including home maintenance (repairs, improvements, painting etc.), car maintenance or repairs and gardening”. The regressions of housework on a dummy for living with a partner (cohabiting or being married). Controls include a quadratic on age, education at age 25 in years, an interaction term of age and years of education, predicted wages, a logarithm in the number of children, hours of paid work, and number of household rooms. Percentage explained by controls obtained as the difference between the OLS coefficient without controls and the OLS coefficient with controls, divided by the OLS coefficient without controls; percentage explained by FE obtained as the difference between the OLS coefficient with controls and the FE coefficient, divided by the OLS coefficient with controls. In both cases we denote >100% when the formula yields a number greater than 100%. Standard errors in brackets. *** Significant at the 1 percent level; ** significant at the 5 percent level;* significant at the 10 percent level. Electronic copy available at: https://ssrn.com/abstract=3524799 41 Appendix A: MTUS Table A1 Survey design Country Year Survey coverage Diary days Time interval Mode of data collection Number of activities Original sample size Austria 1992 Main collection in March and September 1992, some diaries from February, April through August, and October 1992 1-day 15 min. Selfcompletion 202 25,233 diaries Canadá 1998 January - December 1998 1-day Free Recall by telephone 178 10,726 diaries France 1998 January - December 1998 1-day 10 min. Selfcompletion 139 15,441 diaries Germany 2001 April 2001-March 2002 3-day 10 min. Selfcompletion 271 35,813 diaries Italy 2002 April 2002-March 2003 1-day 10 min. Selfcompletion 96 55,773 diaries Norway 2000-01 February 2000’s - February 2001 2-day 10 min. Selfcompletion 122 6,628 diaries Spain 2002-03 October 2002-October 2003 1-day 10 min. Selfcompletion 198 46,774 diaries The United Kingdom 2000-01 June 2000’s - August 2001 2-day 10 min. Selfcompletion 265 20,980 diaries The United States 2003-08 Whole years of 2003, 2004, 2005, 2006, 2007, and 2008 1-day Free Recall by telephone 564 85,177 diaries We restrict the sample to individuals who had time diaries that added up to a complete day (i.e., 1440 min). All surveys include sample weights to ensure each day of the week and each survey are uniformly represented. Source: Multinational Time Use Study (MTUS, www.timeuse.org) version 5.8 release 5 for Austria, France, Spain, UK, and US, and version 5.53 for Canada, Germany, Italy, and Norway.. Electronic copy available at: https://ssrn.com/abstract=3524799 48 the time diary surveys from MTUS. Whereas married men spend 0.58 hours less of housework than single men in the US according to estimates using the MTUS, they do 1.14 hours a week less of housework according to the estimates using the PSID. Similarly, whereas married men spend 1.50 hours less of housework than single men in the UK according to estimates using the MTUS, they do 2.15 hours a week less of housework according to the estimates using the Harmonized BHPSUKHLS. Given that there is no panel diary survey that would allow us to undertake the analysis aimed in this section, these panel data sets seem the best suited data for the task at hand. References to Appendix B Hill, M. 1985. “Patterns of Time Use," In: Juster, T., Stafford, J., (Eds.) Time, goods, and well-being. Ann Arbor, MI: Institute for Social Research, The University of Michigan. Robinson, J.P., 1985. “The validity and reliability of diaries versus alternative time use measures.” In: Juster, T., Stafford, J., (Eds.) Time, goods, and well-being. Ann Arbor, MI: Institute for Social Research, The University of Michigan. Electronic copy available at: https://ssrn.com/abstract=3524799 49 Table B1. Validating Stylized Survey Time Use Measures with Time Use Diary MTUS-PSID MTUS-BHPS (1) (2) (3) (4) MTUS PSID 2003-2007 MTUS UKHLS year 2001 Panel A. Women Married 3.14*** 4.01*** 2.81*** 2.85*** (0.13) (0.25) (0.35) (0.45) Age 0.53*** -0.02 0.86*** 0.34*** (0.05) (0.09) (0.12) (0.08) Age^2 -0.01*** 0.00* -0.01*** -0.00*** (0.00) (0.00) (0.00) (0.00) Number of Children 1.53*** 2.61*** 1.84*** 2.79*** (0.06) (0.14) (0.15) (0.25) Hours of paid work -0.17*** -0.16*** -0.20*** -0.15*** (0.00) (0.01) (0.01) (0.01) Constant 2.34** 12.20*** -1.90 6.00*** (1.03) (1.83) (2.51) (1.74) R-squared 46,480 9,935 5,974 2,824 Observations 0.197 0.173 0.241 0.214 Panel B. Men Married -0.58*** -1.14*** -1.50*** -2.15*** (0.12) (0.24) (0.32) (0.40) Age 0.28*** 0.13** 0.27*** -0.04 (0.04) (0.06) (0.09) (0.04) Age^2 -0.00*** -0.00* -0.00** 0.00 (0.00) (0.00) (0.00) (0.00) Number of Children 0.46*** 0.54*** 0.34*** 0.38** (0.05) (0.08) (0.11) (0.16) Hours of paid work -0.08*** -0.06*** -0.09*** -0.07*** (0.00) (0.01) (0.00) (0.01) Constant 2.55*** 7.16*** 3.92** 10.60*** (0.85) (1.26) (1.93) (0.95) R-squared 36,505 9,192 4,962 2,425 Observations 0.0910 0.0259 0.151 0.0875 Notes: Each column presents regression coefficients as in Table 1 and 2 in the text. As in the analysis presented in Tables 1 and 2 we restrict the sample to respondents between 24 and 65 years old, and exclude retired individuals and students in order to net out life-cycle effects that are closely related to time-use patterns. We also restrict the sample to individuals who are no longer living with their parents. The dependent variable is routine housework and is measured in hours per week. Routine housework is defined as the time spent in cooking and household upkeep in columns 1, 2, 5, and 6 (codes Av6 and av7 in MTUS). In columns 3, 4, 7 and 8 routine housework is the response to the question "About how many hours do you spend on housework in an average week, such as time spent cooking, cleaning and doing the laundry?". Specifications in columns 2, 4, 6, and 8 also include education dummies (secondary education, and post-secondary education or more, less than secondary education being the reference category). Source: MTUS (2000-2001, 2003-2007), Harmonized BHPS-UKHLS (2001), and PSID (2003-2007). Electronic copy available at: https://ssrn.com/abstract=3524799 50 Appendix C: PSID, UKHLS, and HILDA Table C.1 Variable Definitions Variable Variable definition Derived from variable(s) Panel A. PSID Housework Average weekly hours of housework. HHOURS_HEAD, HHOURS_WIFE Married Dummy variable equal to 1 if currently married or cohabiting , and 0 single, divorced or widowed MARST Cohabiting Dummy variable equal to 1 if currently married or cohabiting , and 0 single, divorced or widowed MARST, RELATIONSHIP TO HEAD Always single Dummy variable equal to 1 if single, divorced, or widowed during all waves , and 0 otherwise MARST Always married Dummy variable equal to 1 if married during all waves , and 0 otherwise MARST Transition into marriagesingle Dummy variable equal to 1 for first transition into marriage from single, and 0 otherwise MARST Age Age at date of interview (years) AGE_HEAD, AGE_WIFE Years of education Years of schooling EDUCATION Number of children Number of under 18 in the household NCHILD Number of rooms Number of rooms in the respondent's house ROOMS Paid work Average weekly hours of market work WHOURS_HEAD, WHOURS_WIFE Wages Predicted average hourly earnings WAGE_HEAD, WAGE_WIFE Dwelling type Categories for detached, semidetached, flat, and mobile home TYPE DU Ownership status Categories for own, rent, or neither OWN/RENT OR WHAT Panel B. UKHLS Housework Average weekly hours of housework. HOWLNG Married Dummy variable equal to 1 if currently married or cohabiting , and 0 single, divorced or widowed MASTAT Not in parental home Dummy variable equal to 1 if mother and father not in respondents household, and 0 otherwise HGFNO, HGMNO Cohabiting Dummy variable equal to 1 if currently married or cohabiting , and 0 single, divorced or widowed MASTAT Always single Dummy variable equal to 1 if single, divorced, or widowed during all waves , and 0 otherwise MASTAT Always married Dummy variable equal to 1 if married during all waves , and 0 otherwise MASTAT Transition into marriagesingle Dummy variable equal to 1 for first transition into marriage from single, and 0 otherwise MASTAT Age Age at date of interview (years) AGE Years of education Years of schooling with the following translation: 18 for Higher Degree, 16 for First Degree, 15 for Further Education, 13 for A-levels, 11 for O-levels and other secondary education, 10 for other or no qualification. QFEDHI Number of children Number of under 18 in the household NKIDS Number of rooms Number of rooms in the respondent's house HSROOM Paid work Average weekly hours of market work JBHRS, JBOT Wages Predicted average hourly earnings PAYGU_DV, JBHRS, JBOTPD Dwelling type Categories for detached, semidetached, flat, and other HSTYPE Ownership status Categories for own, rent, or neither TENURE Panel C. HILDA Housework Average weekly hours of housework. LSHW Married Dummy variable equal to 1 if currently married or cohabiting , and 0 single, divorced or widowed MRCURR Not in parental home Dummy variable equal to 1 if mother and father not in respondents household, and 0 otherwise HHMID HHFID Electronic copy available at: https://ssrn.com/abstract=3524799 51 Cohabiting Dummy variable equal to 1 if currently married or cohabiting , and 0 single, divorced or widowed MRCURR Always single Dummy variable equal to 1 if single, divorced, or widowed during all waves , and 0 otherwise MRCURR Always married Dummy variable equal to 1 if married during all waves , and 0 otherwise MRCURR Transition into marriagesingle Dummy variable equal to 1 for first transition into marriage from single, and 0 otherwise MRCURR Age Age at date of interview (years) HGAGE Years of education Years of schooling with the following translation: 18 for Postgrad - masters or doctorate, 17 for Grad diploma, grad certificate, 16 for Bachelor or honours, 14 for Adv diploma, diploma, 13 for Cert III or IV, 12 for Year 12, 11 for Year 11 or equivalent, 10 for Year 10 or equivalent / Junior Seco, 9 for Year 9 or equivalent, 8 for Year 8 or equivalent, 7 for Year 7 or equivalent (NSW, VIC, TAS), O-levels and other secondary education, 6 for Primary school. EDHIGH EDHISTS Number of children Number of under 18 in the household HGAGE Number of rooms Number of rooms in the respondent's house HSBEDRM Paid work Average weekly hours of market work LSEMP Wages Predicted average hourly earnings WSCEI, JBHRUC, ESBRD Dwelling type Categories for detached, semidetached, flat, and other DODTYPE Ownership status Categories for own, rent, or neither HSTENR Source: PSID (1992-2015), Harmonized BHPS-UKHLS (1991-2015), and HILDA (2002-2016). Electronic copy available at: https://ssrn.com/abstract=3524799 52 Table C.2. Summary Statistics Source: PSID (1992-2015), BHPS-UKHLS (1991-2015), and HILDA (2002-2016). Housework Married Cohabiting Age Years of education If children under 5 If children Number of children Number of house rooms Paid work Wages Obser. Panel A: Women PSID 17.53 0.92 0.03 41.86 13.75 0.24 0.52 0.54 25.44 18.51 6.74 27290 (0.08) (0.00) (0.00) (0.06) (0.01) (0.00) (0.00) (0.00) (0.11) (0.04) (0.01) UKHLS 17.03 0.93 0.12 43.39 12.55 0.16 0.43 0.42 22.50 8.58 4.69 36177 (0.06) (0.00) (0.00) (0.06) (0.01) (0.00) (0.00) (0.00) (0.10) (0.02) (0.01) HILDA 17.26 0.90 0.15 43.87 13.14 0.22 0.48 0.50 22.79 27.84 3.38 39220 (0.06) (0.00) (0.00) (0.05) (0.01) (0.00) (0.00) (0.00) (0.09) (0.03) (0.00) Panel B: Men PSID 7.07 0.93 0.03 42.56 13.78 0.25 0.53 0.55 39.34 26.27 6.68 28243 (0.04) (0.00) (0.00) (0.06) (0.01) (0.00) (0.00) (0.00) (0.10) (0.06) (0.01) UKHLS 5.53 0.92 0.14 43.90 12.79 0.16 0.43 0.43 38.76 6.86 4.64 34181 (0.03) (0.00) (0.00) (0.06) (0.01) (0.00) (0.00) (0.00) (0.10) (0.01) (0.01) HILDA 6.28 0.89 0.16 44.73 13.32 0.22 0.48 0.50 38.22 32.12 3.37 37489 (0.03) (0.00) (0.00) (0.06) (0.01) (0.00) (0.00) (0.00) (0.10) (0.04) (0.01) Source: PSID (1992-2015), Harmonized BHPS-UKHLS (1991-2015), and HILDA (2002-2016). Electronic copy available at: https://ssrn.com/abstract=3524799 53 Table C.3. Wage Equations (1) (2) (3) PSID UKHLS HILDA Panel A. Women Age 0.59*** 0.16*** 0.19** (0.10) (0.02) (0.09) Age squared -0.00*** -0.00*** -0.00*** (0.00) (0.00) (0.00) Years of education 2.21*** 0.28*** 1.00*** (0.19) (0.02) (0.11) Age*Years of education -0.00 0.02*** 0.02*** (0.00) (0.00) (0.00) Constant -33.56*** -6.76*** -4.29* (3.13) (0.52) (2.46) Unrestricted obs. 41120 57982 50945 Selection equation Mother employed when respondent was 14 -0.23*** 0.24*** (0.01) (0.01) Constant 0.66*** 0.39*** (0.01) (0.01) Total obs. 57411 87641 76214 Panel B. Men Age 0.55*** 0.35*** 0.32*** (0.13) (0.02) (0.07) Age squared -0.01*** -0.01*** -0.01*** (0.00) (0.00) (0.00) Years of education 0.58** 0.09*** 1.03*** (0.24) (0.02) (0.09) Age*Years of education 0.06*** 0.01*** 0.03*** (0.01) (0.00) (0.00) _cons -27.14*** -8.01*** -11.30*** (4.04) (0.49) (1.88) Unrestricted obs. 45207 51549 53494 Selection equation Father self-employed when respondent was 14 -0.08*** 0.22*** (0.01) (0.01) _cons 0.43*** 0.75*** (0.01) (0.01) Total obs. 52596 64770 65238 Notes: Data comes from the 1985-2015 PSID. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting) and marrying (i.e. individuals who transition from single to married). Housework is reported in hours per week as the answer to the question “About how much time do you spend on housework in an average week? I mean time spent cooking, cleaning, and doing other work around the house?". The regressions of housework on a dummy for living with a partner (cohabiting or being married) also control for a quadratic on age, education at age 25 in years, an interaction term of age and years of education, predicted wages, a logarithm in the number of children, hours of paid work, and number of household rooms. Percentage explained by controls obtained as the difference between the OLS coefficient without controls and the OLS coefficient with controls, divided by the OLS coefficient without controls; percentage explained by FE obtained as the difference between the OLS coefficient with controls and the FE coefficient, divided by the OLS coefficient with controls. In both cases we denote >100% when the formula yields a number greater than 100%. Standard errors in brackets. *** Significant at the 1 percent level; ** significant at the 5 percent level;* significant at the 10 percent level. Electronic copy available at: https://ssrn.com/abstract=3524799 54 Table C.4. Attitudinal Variable Definitions and Summary Statistics Attitude Statement Female Male Min Max Variable Name Obs. Mean Std.Dev Obs. Mean Std.Dev Panel A. UKHLS Pre-school child suffers if mother works 13,047 3.11 1.06 12,461 2.77 1.03 1 5 OPFAMA Family suffers if woman works full time 13,047 3.08 1.08 12,461 2.99 1.03 1 5 OPFAMB Woman and family happier if she works 13,047 3.15 0.73 12,461 3.20 0.75 1 5 OPFAMC Husband and wife should both contribute 13,047 2.65 0.88 12,461 2.68 0.89 1 5 OPFAMD Full time job makes women independent 13,047 2.99 0.97 12,461 2.89 0.91 1 5 OPFAME Husband should earn, wife stay at home 13,047 3.75 0.96 12,461 3.52 0.98 1 5 OPFAMF Children need father as much as mother 13,047 1.92 0.78 12,461 1.84 0.71 1 5 OPFAMG Employers should help with childcare 13,047 2.08 0.86 12,461 2.21 0.88 1 5 OPFAMH Panel B. HILDA Many working mothers seem to care more about being successful at work 7,440 3.12 1.61 7,102 3.45 1.48 1 7 ATWKWMS Many working fathers seem to care more about being successful at work 7,440 3.57 1.66 7,102 3.82 1.51 1 7 ATWKWFS If both partners in a couple work, they should share equally 7,440 6.17 1.12 7,102 5.70 1.26 1 7 ATWKSEH Whatever career a woman may have, her most important role in life is caring 7,440 5.41 1.72 7,102 5.07 1.66 1 7 ATWKWRL Whatever career a man may have, his most important role in life is caring 7,440 5.32 1.71 7,102 5.14 1.62 1 7 ATWKMRL Mothers who dont really need the money shouldnt work 7,440 3.21 1.87 7,102 3.50 1.79 1 7 ATWKMSW Children do just as well if the mother earns the money 7,440 5.48 1.41 7,102 5.17 1.43 1 7 ATWKCDW It is better for everyone involved if the man earns the money 7,440 3.01 1.82 7,102 3.35 1.74 1 7 ATWKBMW As long as the care is good, it is fine for children under 3 years 7,440 3.15 1.81 7,102 3.14 1.69 1 7 ATWKADC A working mother can establish just as good a relationship 7,440 4.71 1.76 7,102 4.18 1.73 1 7 ATWKWMR A working father can establish just as good a relationship 7,440 4.90 1.63 7,102 4.42 1.65 1 7 ATWKWFR A father should be as heavily involved in the care of his children 7,440 5.74 1.30 7,102 5.60 1.28 1 7 ATWKFHI It is not good for a relationship if the woman earns more than the man 7,440 2.37 1.57 7,102 2.33 1.45 1 7 ATWKMMF On the whole, men make better political leaders than women do 7,440 2.33 1.65 7,102 3.14 1.82 1 7 ATWKMPL A pre-school child is likely to suffer if his/her mother works full-time 7,440 3.66 1.87 7,102 3.88 1.72 1 7 ATWKPSC Children often suffer because their fathers concentrate too much 7,440 4.17 1.63 7,102 4.45 1.51 1 7 ATWKCS If parents divorce it is usually better for the child to stay with the mother 7,440 3.69 1.63 7,102 3.42 1.60 1 7 ATWKDCM Source: BHPS-UKHLS (1991-2015), and HILDA (2002-2016). Electronic copy available at: https://ssrn.com/abstract=3524799 55 Table C.5. First Principal Component of Attitudinal Variables (1) (2) Attitudes Females Males Panel A. UKHLS Pre-school child suffers if mother works -0.754 -0.732 Family suffers if woman works full time -0.823 -0.817 Woman and family happier if she works 0.588 0.628 Husband and wife should both contribute 0.419 0.415 Full time job makes women independent 0.415 0.399 Husband should earn, wife stay at home -0.696 -0.707 Children need father as much as mother -0.167 -0.088 Employers should help with childcare 0.408 0.419 Proportion of explained variance 32.7 32.6 Panel B. HILDA Many working mothers seem to care more about being successful at work -0.623 -0.601 Many working fathers seem to care more about being successful at work -0.514 -0.485 If both partners in a couple work, they should share equally 0.060 0.018 Whatever career a woman may have, her most important role in life is caring -0.454 -0.529 Whatever career a man may have, his most important role in life is caring -0.412 -0.485 Mothers who don’t really need the money shouldn’t work -0.635 -0.618 Children do just as well if the mother earns the money 0.462 0.414 It is better for everyone involved if the man earns the money -0.712 -0.708 As long as the care is good, it is fine for children under 3 years 0.369 0.262 A working mother can establish just as good a relationship 0.618 0.534 A working father can establish just as good a relationship 0.519 0.391 A father should be as heavily involved in the care of his children 0.229 0.140 It is not good for a relationship if the woman earns more than the man -0.553 -0.510 On the whole, men make better political leaders than women do -0.512 -0.512 A pre-school child is likely to suffer if his/her mother works full-time -0.676 -0.684 Children often suffer because their fathers concentrate too much -0.509 -0.482 If parents divorce it is usually better for the child to stay with the mother -0.506 -0.442 Proportion of explained variance 27.6 25.4 Notes: This table shows the average weights assigned to each variable for the first principal component in constructing women’s and men’s indices of gender role attitudes. The three highest factor loadings are shaded in grey. The last row also shows the proportion of variance attributable to the first principal component. Source: BHPS-UKHLS (1991-2015), and HILDA (2002-2016). Electronic copy available at: https://ssrn.com/abstract=3524799 56 Table C.6. Housework and marriage: Evidence from longitudinal data. PSID 1985-2015 (1) (2) Women Men Coefficient Standard Error % Explained Coefficient Standard Error % Explained OLS without controls 10.15 (0.43)*** 0.20 (0.25) OLS with controls 4.77 (0.41)*** 53% -0.28 (0.27) > 100% FE 2.16 (0.51)*** 55% 0.69 (0.35)* > 100% Chi2 test of equality of coeff. 18.03 (0.000) 5.87 (0.015) N obs. 3920 4285 Notes: Data comes from the 1985-2015 PSID. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are in a partnership, either legally married or cohabiting) and marrying (i.e. individuals who transition from single to married). Housework is reported in hours per week as the answer to the question “About how much time do you spend on housework in an average week? I mean time spent cooking, cleaning, and doing other work around the house?". The regressions of housework on a dummy for living with a partner (cohabiting or being married) also control for a quadratic on age, education at age 25 in years, an interaction term of age and years of education, predicted wages, a logarithm in the number of children, hours of paid work, and number of household rooms. Percentage explained by controls obtained as the difference between the OLS coefficient without controls and the OLS coefficient with controls, divided by the OLS coefficient without controls; percentage explained by FE obtained as the difference between the OLS coefficient with controls and the FE coefficient, divided by the OLS coefficient with controls. In both cases we denote >100% when the formula yields a number greater than 100%. Standard errors in brackets. *** Significant at the 1 percent level; ** significant at the 5 percent level;* significant at the 10 percent level. Electronic copy available at: https://ssrn.com/abstract=3524799 57 Table C.7. Housework and marriage: Evidence from longitudinal data. Controlling for dwelling type and ownership. (1) (2) Women Men Coefficie nt Standard Error % Explained Coefficie nt Standard Error % Explained Panel A: PSID OLS without controls 9.55 (0.39)*** 0.40 (0.25) OLS control group 1 4.66 (0.43)*** 51% -0.09 (0.28) > 100% OLS control group 2 4.37 (0.44)*** 6% -0.20 (0.28) < -100% FE 1.46 (0.61)** 67% 0.87 (0.40)** > 100% Chi2 test of equality of coefficients FE-OLS control group 3 20.30 (0.000) 5.30 (0.021) N obs. 3545 4188 Panel B: UKHLS OLS without controls 7.28 (0.46)*** -1.56 (0.22)*** OLS control group 1 3.86 (0.42)*** 47% -1.46 (0.24)*** 6% OLS control group 2 3.58 (0.43)*** 7% -1.29 (0.25)*** 12% FE 1.84 (0.42)*** 49% -0.76 (0.35)** 41% Chi2 test of equality of coefficients FE-OLS control group 2 10.63 (0.001) 1.79 (0.181) N obs. 3476 3509 Panel C: HILDA OLS without controls 7.71 (0.39)*** -0.65 (0.21)*** OLS control group 1 3.49 (0.35)*** 55% -0.83 (0.21)*** -28% OLS control group 2 3.36 (0.36)*** 4% -0.84 (0.22)*** -1% FE 1.04 (0.44)** 69% 0.11 (0.27) > 100% Chi2 test of equality of coefficients FE-OLS control group 2 18.28 (0.000) 8.62 (0.003) N obs. 4611 4609 Notes: Data comes from the 1992-2015 PSID, the 1991-2015 Harmonized BHPS-UKHLS, and the 2002-2016 HILDA. Sample is women (men) between 24 and 65 years old, not living in the parental home, who have not been divorced, and who are observed in at least two waves. We consider singles (i.e., single individuals not living in the parental home), marrieds (i.e., individuals who are legally married) and marrying (i.e. individuals who transition from single to legally married). Housework is reported in hours per week as the answer to the Electronic copy available at: https://ssrn.com/abstract=3524799