More chores at home: a price immigrants pay when marrying a native?

Transcription

1 More chores at home: a price immigrants pay when marrying a native? Shoshana Amyra Grossbard San Diego State University, IZA and CES-ifo and Victoria Vernon Empire State College, New York December 2, 2015 Abstract: Using American Time Use Survey we investigate whether marrying a US-born spouse carries a premium or a penalty for immigrants in terms of the amount of household chores they perform in marriage. We predict that intermarriage involves an exchange of assimilation services provided by the native for a higher price that the immigrant spouse is expected to pay in marriage, and therefore that immigrants married to natives work more at chores than their counterparts in all-immigrant marriages. We account for selection into intermarriage and find a sizable chores penalty for immigrants and a premium for natives. The effect is larger for women. Non-employed immigrant women in single earner couples pay the largest chores penalty, and the non-employed native women receive the largest premium. Most American-born men and women spend significantly less time in chores when married to foreigners than when married to other natives, while most foreigners spend at least as much or more time in chores in such marriages relative to what they do in all-immigrant marriages. Our findings are robust to changes in specification and alternative definition of chores. The way that citizenship status affects time spent in chores by intermarried immigrants and natives also suggests that there is a premium to being a legal citizen. JEL CODES: D13, J12, J22 KEYWORDS: Time Use, Chores, Intermarriage between Immigrants and Natives, Marriage Market 1

2 1. Introduction Immigrants tend to benefit from intermarriage with natives in terms of faster wage growth and better job market opportunities in the U.S.A. (Kantarevic 2004, Furtado and Theodoropoulos 2009) and Australia (Meng and Gregory 2005). 1 Comparisons between all-immigrant marriages and intermarriages between natives and immigrants suggest that their native spouses assist immigrants with access to social capital, finding employment, obtaining legal status, and acquisition of language skills (Furtado and Theodoropoulos 2010). Here we investigate whether marrying a native carries a premium or a penalty for immigrants in terms of the amount of household chores they perform in marriage. We also investigate how intermarriage to an immigrant affects the amount of chores that natives perform by comparing time spent doing chores in intermarriages and marriages between two natives. Furthermore we research whether household chores performed by an individual vary with own and spouse s citizenship status. Given that around 13% of the US population is foreign born (Pew Research 2011) these questions are of policy relevance. 2 The American Time Use Survey for the years allows us to compare the time use of immigrants married to natives, endogamously married immigrants, and endogamously married natives. Immigrants and natives who choose to intermarry may possess unobservable characteristics that distinguish them from their endogamous counterparts. To take account of selection into intermarriage we estimate IV (Instrumental Variable) models that include two stages: selection into intermarriage and amount of time spent on chores. In choosing our instruments we learned from prior studies 1 Natives are US citizens who have at least one US-born parent. 2 According to the US Census, of a total million households in the U.S., 15.7 million are headed by immigrants. The majority of these immigrant households--8.2 million--are headed by immigrants who are not U.S. citizens. 2

3 that have linked the probability of intermarriage to being younger at migration, longer time between migration and marriage, better education and availability of mates belonging to the same ethnicity (see e.g. Angrist 2002, Meng and Gregory 2005, Chiswick and Miller 2014). The conceptual framework we use to analyze intermarriage is based on the marriage market analysis pioneered by Gary Becker (1973) and its first application to intermarriage, Grossbard-Shechtman (1983). The latter included the prediction that, given discrimination against Jews in the US in the first part of the 20 th century, Jewish men marrying Christian women would have to compensate for their lower status in the form of more years of schooling or fewer previous marriages relative to their wives characteristics. Similarly, sociologists have hypothesized that preference for lighter and more educated spouses may lead to marriages where blacks trade their high educational status in exchange for the racial advantages of a white partner and to marriages where white partners trade their racial advantage in exchange for higher levels of spouse s education (Merton 1941, Fu 2001, Gullickson and Torche 2014). Underlying these analyses is the assumption that an individual s value in marriage is linked to own traits and those of the spouse. Grossbard-Shechtman (1984) and Lafortune et al. (2012) have linked this value in marriage to willingness to work in the labor force. Intermarriage, which is linked to value in marriage, can be linked to labor force participation as argued in Grossbard-Shechtman and Neuman (1988). The latter predicted that if a woman of a higher status group intermarries with a man from a lower status group she will have a lower participation rate in the labor force than if she marries endogamously. Tests for such intermarriage premium that did not control for the endogeneity of intermarriage provided tentative evidence of lower labor force participation rates for Ashkenazi Jewish women who married Sepharadic men in Israel 3

4 (Grossbard-Shechtman and Neuman 1988) and for Caucasian women who married Hawaiians in Hawaii (Grossbard-Shechtman and Fu 2002). In each case, the intermarried women were compared to their endogamous counterparts. More recently value in marriage was linked to time spent performing chores in the household. Applying a two-stage estimation of household chores and intermarriage to ATUS data, Grossbard, Gimenez and Molina (2014; hence GGM) showed that when intermarried with whites black women in the U.S. perform more hours of chores than comparable black women married to other blacks. In contrast, when intermarried with black men, white women perform fewer hours of chores than their counterparts in white endogamous marriages. In this paper we also use ATUS data and a two-stage estimation method. Our focus here is to investigate the relationship between intermarriage of immigrants and natives and the time that individual spouses spend on chores relative to their non-intermarried counterparts. Our main findings are that when comparing immigrants who intermarry with a native to immigrants in all-immigrant marriages, the intermarried immigrants perform more hours of chores in their household. In contrast, U.S. natives married to an immigrant spend less time on chores than their counterparts who marry another native. We also find that both immigrants and natives benefit from being married to a spouse who is not a US citizen to the extent that they perform fewer hours of chores. 2. Conceptual framework We predict that intermarriage is associated with hours spent doing chores based on the following marriage market analysis. Consider four groups of interrelated heterosexual marriage markets: endogamous markets for natives, endogamous markets for immigrants, markets for marriages between immigrant men and native women, and markets for 4

5 marriages between native men and immigrant women. Within each group there may be submarkets stratified by additional characteristics such as education or age. Following Becker (1973) and Grossbard-Shechtman (1984) it is assumed that in each market an equilibrium has been established associated with a particular price for a man or woman of given characteristics. 3 For instance, there is a price for immigrant men with a high school education and age 25 if they marry immigrant women of the same education and age 27. Supplies express willingness of men with such characteristics to marry a woman with these characteristics at a given price. Demands express willingness of a particular type of woman to pay for marriage to a particular type of man. There appears to be a widespread preference for endogamy in view of the advantages of marrying similar people, implying that many marriages will be all-native or all-immigrant. For that reason, immigrants may prefer to marry other immigrants, especially if they have a similar background. However, in view of the above-mentioned cultural and social benefits that immigrants may derive from marrying natives such as a faster way to assimilate we assume that ceteris paribus there are more immigrants wanting to marry natives than natives wanting to marry immigrants. Consequently, if we compare markets for natives to markets for immigrants, the aggregate demands by immigrants wanting to marry natives is expected to exceed the aggregate demands by natives wanting to marry immigrants. Discrimination against immigrants may also lead to relatively lower demand for immigrant spouses. Assuming similar supplies of immigrants and natives, and that competition leads to clearing prices in all marriage markets, the higher relative demands for native spouses imply that the price of native mates will be high relative to the price of immigrant mates. 3 The prices may be related to the share of the gain from marriage obtained by an individual (see Becker 1973, 1981 and Grossbard 2015). 5

6 Time use surveys offer data on a consequence of variation in price in marriage markets: the time that people work at relatively unpleasant aspects of household production typically called chores is expected to vary inversely with value in the marriage market. The higher people s value in marriage the less we expect them to work at unpleasant tasks such as chores. 4 Accordingly, in the current context of possible intermarriage between immigrants and natives we predict: Immigrants in couple with natives will spend more time on chores than their counterparts in all-immigrant (endogamous) marriages (Prediction 1). Natives in couple with immigrants will perform fewer hours of chores than their counterparts in all-native (endogamous) marriages (prediction 2). These predictions based on marriage market analysis differ from predictions based on theories of intra-household bargaining, such as McElroy and Horney (1981) or Friedberg and Webb (2006) which require detailed information on the resources and time allocation of both partners. Here our focus is on intermarriage: we compare intermarried immigrants to endogamous immigrants, and intermarried natives to endogamous natives. Both natives and immigrants may also prefer to marry individuals who have US citizenship as opposed to those who are either foreign nationals or permanent residents because citizenship offers some entitlements not available otherwise. Consequently to the extent that there are separate marriage markets for citizens and non-citizens, we expect the relative demand for citizen spouses to be higher than that for non-citizen spouses. With comparable supplies in all markets this implies higher prices for citizen spouses than for non-citizens, and therefore we also predict that citizens in couple with 4 Leisure may be more enjoyable than home production to the extent that the former activities provide a higher experienced utility to the individuals (Kahneman et al. 2004). These are also activities generating more happiness according to Connelly and Kimmel (2015). 6

7 non-citizens will spend less time on chores than their counterparts in all-citizen (endogamous) marriages and that non-citizens in couple with citizens will spend more time on chores than comparable citizens married to citizens. When testing these predictions it is important to control for many characteristics of both spouses other than immigration status that could possibly influence individual price in marriage and how that price translates into time performing chores. These additional factors may include spouses relative bargaining power measure by wages as suggested by Pollak (2005). There is evidence that a spouse with higher bargaining power spends less time doing chores and more time enjoying leisure compared to the spouse with lower power (Friedberd and Webb 2005). Another factor may be household income: richer couples can afford to outsource chores. Employment status also influences how much time is available for chores. What if immigrants face inferior options on the job market and therefore choose to spend less time in paid work and more time in household production? In general, immigrants indeed work in inferior jobs and earn less than the native born (Orrenius and Zavodny 2009). However, intermarriage selects for more productive men and women who earn higher wages (Kantarevic 2004, Basu 2015). Even though theoretically families composed of immigrant women and US-born men must have the largest incentive to specialize because of their highest wage differential, in practice these couples are less likely to specialize relative to couples with two immigrants (Nottmeyer 2014). In our sample intermarried men and women are more likely to be employed and to earn high wages than other immigrants. Thus we can rule out preference for home production vs paid work as a competing explanation for more chores among intermarried immigrants. While we do not address the issue of potential endogeneity of paid work hours, which is 7

8 likely to be an issue for women, we examine the impact of intermarriage on chores separately for families with working and non-working wives. We expect the predictions above to receive more support from the data in samples of individuals performing more chores on average. Women typically perform more chores than men, and this is true in the U.S. (see Figure 1). Therefore, we expect the predictions above to be more applicable to women than to men, and more to respondents who are not employed or work few hours in the labor force than to fully employed respondents. Most likely the spouses of these respondents are then employed full-time. If we limit the sample to dual-earner couples we have information on wages of both respondent and spouse, and we can control for another factor expected to influence the time individual men and women spend on chores through household bargaining: their relative wages. If we continue to find intermarriage penalties or premia after controlling for wages, this indicates that our findings are not due to differences in bargaining behavior of intermarrying immigrants or natives. Source country characteristics, including culture and labor force participation rate, may influence the time allocation between work, household production and leisure, especially for women (Fernandez 2007). The influence of source country characteristics diminishes with more time in the US as do rates of ethnic self-identification (Duncan and Trejo 2011). In our regressions we control for region of origin and time since arrival to the U.S. (and possible degree of assimilation). We expect less of a penalty if the individual is a second generation immigrant. 3. Econometric Strategy We first estimate an OLS model of unpaid home production. Then we switch to a twostage model that considers intermarriage as endogeneous. 8

9 The OLS model for immigrants. In this model the impact of intermarriage on unpaid home production work is HPi= α1imit + α2xit + α3yit + uit (1) where HP is the number of minutes per day spent in household production, IM is intermarried status (a binary variable and our variable of interest), and X and Y are two vectors of control variables. X is a vector of human capital, demographic and assimilation controls including four indicators for education, age, age-squared, ethnicity dummies for Hispanic, black and Asian, an indicator for individuals who were born in the US to immigrant parents, immigrants who arrived before age 12, those who arrived at age 12-20, and those who arrived at age The latter set of dummies are meant to account for the fact that length of time in the US affects probability of intermarriage and degree of assimilation (e.g. via language ability for which we do not have a better control in this data set). Assimilation, in turn, influences the allocation of time (Hamermesh and Trejo 2013, Hwang 2015). We also include age difference between spouses (spouse s age minus respondent s age) to account for the possibility that younger immigrants may prefer to marry older more established immigrants or natives, or that natives might be more willing to marry younger immigrants. The difference between spouse s years of schooling and respondent s years of schooling is also included given the importance of education in matching between spouses and the likely importance of education in explaining intermarriage between natives and immigrants (Celisksoy et al. 2006, Furtado 2012). Ideally, we would include indicators for separate countries to account for culture and linguistic distance, which in turn affect assimilation (Chiswick and Houseworth, 2011). However given the modest sample size we group countries into 9

10 nine geographic areas (with English-speaking Western countries as a reference category). We use information on cohorts to cluster error terms: errors in regressions for immigrants are clustered by state, region of origin and cohort of arrival. 5 Vector Y includes controls that affect individual s daily time allocation but not individual s probability of intermarriage (the two events are years apart). These are the presence of children, number of children, presence of a child under age 5, 3 dummies for the region of residence to account for climate differences, dummies for summer and winter, indicators for days of the week (Friday, Saturday, Sunday and holiday). We include dummies for the survey year to adjust for the state of the economy such as the general unemployment rate, as it has been shown to affect the allocation of time (Burda and Hamermesh 2010). We also include an indicator for non-us citizen status of the respondent and the spouse, as we expect these to be related to price in marriage markets, as discussed in the previous section. Being a permanent resident or a foreign citizen, or to a larger extent an illegal alien, may entail fewer options in terms of employment since some jobs require citizenship. In some regressions we include the deviation of total household income from state median household income in thousands of dollars to control for an income effect. This variable does not significantly affect other estimates and is dropped from the first set of results because of its potential endogeneity. In several models we include log wages of both spouses. IV model for immigrants. The decision to intermarry may be endogeneous to the decision on how much time to devote to household production: it may depend on the 5 In some runs we also include 5 cohort of arrival dummies for immigrants who arrived before 1970, in the 70s, in the 80s, in the 90s and after 2000 (with US-born being the reference group) in order to account for possible differences in type of immigrants. We choose not to include these indicators in our final set of results because of concern that the cohort of arrival may be correlated with our instruments. The results are not significantly affected by cohort dummies. 10

11 expected time the individual intends to spend in housework and leisure, and what he or she expects their spouse to produce at home. For instance, selection may be present to the extent that immigrants who are more productive in the labor force prefer to spend less time in household production and choose to marry a partner in or out of their group accordingly. In addition, unobservable traits of both the respondents and their spouses may be associated with the likelihood of intermarriage. In order to account for such possible selection bias we instrument the probability of intermarriage among immigrants and estimate the following first stage equation: IMit = β1zit + β2xit + eit (2) where Z is a measure of the availability of immigrant mates of the opposite sex. Vector X is described above, the error term eit captures other non-observable factors that may affect the probability to intermarry (IM), such as religious beliefs or gender role expectations. We use two different measures of Z: First, we follow Meng and Gregory (2005) in measuring the immigrant mates availability index as the number of unmarried opposite sex potential partners from the region of origin r to all unmarried potential partners observed for age group a in state s in year t: Z1= narst/narst For example for men age from Eastern Europe who married around year 1990 in California, Z1 is the ratio of East European women age in California in 1990 to all women in this age group in California. We also use the following alternative measure of Z suggested by Kantarevic (2004): Z2= (narst/nart)/ (mast/mat), where narst is the number of unmarried opposite sex potential partners from ethnic group r, age group a in state s, 11

12 Nart is the number of potential partners from ethnic group r, age group a in all states, mast is the number of unmarried native individuals in state s, and Mat is the total number of unmarried natives in age group a. Availability ratios are constructed using personal weights data from pooled 1% Census data for 1960 and 1970, and separate 5% Census data for 1980, 1990, 2000, and 2010 by state for 3 age groups: (youngest), (middle), (older) for men, and two years younger, 18-32, 33-47, for women. 6 The two-year difference in the age groups reflects the average age difference between husbands and wives in the US. We use only unmarried men and women as potential partners. The ratios are calculated by state and age group and matched to Census year including the respondent s potential marriage decade: 1960 and 1970 for those who turned 22 before 1975; 1980 Census for those who turned 23 before 1985; 1990 for those who turned 24 before 1995; 2000 for those who turned 25 before 2005, and the 2010 Census for those who turned 25 after We progressively increase marriage age to adjust for the fact that it increased over this period. The three age groups are used only for computing availability of mates measures; in all regressions of household production we combine all those ages We also experimented with adding sex ratio within immigrant group constructed as the ratio of unmarried males to unmarried females within age group a in state s in year t: SRast= narst(men)/narst(women) However, this additional instrument was not significant in most regressions, and therefore was dropped. 6 One concern is the small number of immigrants from Africa and the Middle East in all three age groups in the Census. We reran the results using only instruments Z constructed from Census years , assuming immigrant networks are stable. The results are very similar to those reported in the paper. 12

13 In calculating availability ratios for immigrants who arrived after having reached their presumed marriage age we use the Census year closest to the year of arrival. This assumes that people select a partner at age if born in the US or at the time of arrival to the US. In calculating the sex ratios we implicitly assume that (1) respondents did not divorce or remarry since age 22-25, (2) the state in which they resided around that age is their current state (a strong assumption given high mobility rates in the US), and (3) immigrants re-evaluated their marriage options upon arrival even if they arrived already married. These are strong assumptions because in reality many immigrants who arrive married do not participate in US marriage markets. As one of the robustness checks we perform, we relax this last assumption by also estimating the model for immigrants who arrived in the US before age 25. In order to be good instruments, availability ratios should only affect the probability of endogamous marriage and have no direct effect on time allocation. We assume that the way we construct the availability ratios from past Census data fulfills this condition: our measures of same-region mate availability are decades apart from the measures of chores for many respondents. In all regressions the instruments are highly significant determinants of intermarriage for men and women with F-statistics ranging between 24 and 338 (p<1% in all cases). Yet we are aware of the possibility that the instruments may not be entirely exogenous to chores. For example, immigrants who live in areas with low availability of potential partners of the same ethnicity may be more assimilated and more likely to be intermarried but less willing to trade chores for assimilation services than those who live in immigrant enclaves. In this case the coefficient on intermarriage may underestimate the true effect. On the other hand, natives may be more likely to marry immigrants who are willing to do more chores, and their odds of finding a spouse like this are higher in areas with higher immigrant 13

14 availability ratios. Then the coefficient on intermarriage would overstate its true impact. Nonetheless availability ratios have long become commonly used instruments for intermarriage in the literature. In our two-stage model the first stage is intermarriage probability estimated with probit analysis and the second stage is time spent on household production. The predicted values of intermarriage probability are obtained from the first stage regression (2) and plugged into equation (1). The instruments Z are excluded from the second stage regression, and the control variables in vector Y are excluded from the first stage regression. Among those is citizenship status: since we do not have information on respondent s citizenship status at the time of marriage we do not include citizenship in the probability of intermarriage equations. Models for Natives. We estimate similar OLS and two-stage models for the US natives with a modified set of controls. For the native sample, the instrument Z3 is the availability of foreigners of opposite sex, or the share of all unmarried immigrants of opposite sex from all countries belonging to age group a residing in state s at time t (dast) to all potential unmarried partners of opposite sex including natives (Dat) Z3 = dast/dat It is computed for the same three age groups used for the immigrant regressions, also by state and marriage year. Potential marriage year is defined in the same way, rising from age 22 in the 1970s to age 25 in 2000s. Vectors X and Y are the same, except for the exclusion of variables that apply only to immigrants (such as length of stay in the US and country of origin). Models taking account of whether respondents or their spouses participate in the labor force. Since employment status is a major determinant of the amount of chores that individuals perform we also estimate our two equation models for two subsamples: 14

15 dual earner couples and traditional couples defined as couples in which the wife is not employed, but the husband is. Intermarried probabilities are predicted from the full sample of immigrants using instrument Z2, and from the full sample of natives using instrument Z3. The samples are then restricted in the second stage to dual-earner or traditional couples. 4. Data and sample means ATUS : This dataset is a time use supplement to the CPS. It includes a wide range of demographic, labor market and family status variables, such as countries of birth for respondents and their parents. Unfortunately, only one household member is selected to participate, and thus we can t examine the allocation of time between spouses in intermarried households. Instead, we compare the allocation of time for intermarried and non-intermarried immigrants, as well as natives by marriage type. We examine the following categories of household production: Chores (restricted definition): includes cooking, cleaning, home repairs and maintenance, grocery shopping, garden and vehicle care, and paying bills. It is defined as in GGM (2014). Chores (broad): includes the restricted definition of chores plus nongrocery shopping, buying services, using household and professional services (other than personal and medical services), and making phone calls (excluding those to family and friends). Examples of the extended category thus include purchases in person and online, taking a car for service, hiring a handyman for repairs, and arranging these services on the phone

16 Household production (all): includes the broad definition of chores plus care of children, adults and pets. All categories of time use include related travel. Whether a time use is categorized as leisure, work or household production is to some extent arbitrary. For example, childcare has been shown to behave differently from household production (Kimmel and Connelly 2007), yet mothers multitask more than fathers: they are more likely to combine household production with childcare (Offer and Schneider 2011). Many elements of household production have consumption value (gardening, pet care) and thus can be part of leisure. Whether non-grocery shopping is household production or leisure can be debated, it is likely to be leisure at the margin, but is commonly included in household production. Immigrants are defined as respondents whose parents were both born abroad whereas the respondent could be born abroad or in the US. This definition includes both first and second generation immigrants. In the case of second generation respondents whose two parents came from two different regions of origin (fewer than 5% observations) they are assigned mother s country of origin. Those born in Puerto Rico, Canada, UK, Australia and New Zealand are considered immigrants. Those who have at least one US-born parent are considered natives regardless of birthplace. 8 We removed over 800 records with more than 3 hours of missing activities, missing country of birth, more than 25 years of age difference between spouses, fewer than 15 min of sleep, more than 23 hours of chores, leisure, paid work or personal care each. The rationale behind these exclusions is that we aim to study a typical day of a typical 8 The definition of immigrants varies by study. Meng and Gregory (2005) define immigrant as first generation migrants from non-english speaking countries, and treat those from English-speaking countries as natives. Kantarevic (2004) and Furtado and Theodoropoulos (2009) restrict their samples to first generation migrants from non-english speaking countries. Basu (2015) studies all first generation immigrants. Nottmeyer (2014) uses first and second generation immigrants in her sample. 16

17 couple, and that extremely values may bias our results. The immigrant sample includes 5155 men (18% intermarried) and 6173 women (21% intermarried). Among individuals who arrived in the US before age 25 and those who are born in the US (3407 men and 4086 women), 23% of men and 24% of women are intermarried. These numbers reflect lower rates of assimilation than those reported for Australia by Meng and Gregory (2005). The summary statistics presented in Table 1 suggest that relative to immigrants who are married to other immigrants those who are married to US natives are more educated, more likely to have been born in the US or to have arrived as children, more likely to be US citizens, less likely to have children, less likely to be Hispanic or Asian and more likely to be white. Intermarried spouses are closer in age than immigrant couples. Intermarried immigrant women are significantly more likely to be employed than women married to other immigrants, whereas men s odds of employment are on average the same across the two marriage types. Intermarried men and women receive significantly higher wages than their non-intermarried counterparts. The absolute husband-wife wage differential is the lowest in households composed of an immigrant husband and a native wife, under $4 per hour, followed by endogamous immigrant households, $4-5. The largest spousal wage differential is in couples with an immigrant wife and a native husband, about $7.5 per hour. Both intermarried immigrants and intermarried natives live in richer households than their non-intermarried counterparts, with an average annual household income of over $77,000 compared to $49,000 for allimmigrant couples. We use the information on income with caution since it is collected in intervals and is known to be underreported in surveys. The bottom part of Table 1 reports size of immigrant samples by region of origin. Over half of the immigrants are Hispanic, originating from Mexico and South and 17

18 Central America. Unfortunately, we have fewer than 100 records on intermarried immigrant men and women from India, Middle East, Africa, the Caribbean Islands and men from Asia, which is another reason why we need to exercise caution when interpreting results related to country of origin. Immigrants from Western Europe are more likely to be intermarried, followed by immigrants from English speaking countries: more than half of men and women from developed Western countries are married to US natives. About every fourth immigrant from Eastern Europe is married to a US native. Among immigrants from the Caribbean Islands, one out of 5 is married to a US native. Among immigrants from Mexico and South America, about one in 7 to one in 10 is married to a US native. The lowest rates of intermarriage are for immigrants from India, Bangladesh and Pakistan: about one in 15 to one in 20. Interestingly, women from Southeast Asia and the Middle East are considerably more likely to be intermarried than men of the same origin, while the opposite is true for immigrants from Africa. The sample of natives consists of 20,279 men and 23,885 women, with only 5.2% of men and 4.5% of women intermarried. Respondents in this sample are 3-4 years older than the immigrant respondents. Compared to their endogamously married counterparts, intermarried natives are more educated and more likely to be Hispanic or Asian and to have children. Comparing fertility across the four couple types, on average all-immigrant couples have more children than any other type of couple, allnative couples have fewer children, and couples composed of a native wife and an immigrant husband tend to have more children than couples composed of a native husband and an immigrant wife. Some selection on fertility may be going on: native women who want to have more children may choose to marry immigrant men, and 18

19 immigrant women who want fewer children than is common in their country of origin may prefer to marry native men. Endogamous native women have a labor force participation rate that is 4 percentage points higher than that of native women married to immigrants. Interestingly, both native partners in intermarried unions earn higher wages than men and women in endogamous native unions. As a result, intermarried natives live in higher income households, with the highest incomes (close to $80,000 a year) found for native husbands married to foreign wives. Average daily minutes spent in household production are plotted in Figure 1, using the three measures of household production mentioned above. We see that while native men spend almost 20 min more in chores than immigrant men, native and immigrant men who are intermarried spend about the same amount of time on chores. The contrast between native and immigrant women is also much larger if they are both endogamous than if they are intermarried: intermarried immigrant women work considerably less at chores than their endogamous counterparts, while on average intermarried native women work slightly more at chores than their endogamous counterparts. It can also be seen that endogamous immigrant women spend more than double the time in chores than endogamous immigrant men. The gender gap in chores of endogeneous natives goes in the same direction but is not as large. These comparisons of means confound the impact of many factors such as number of children and income. The regression analysis presented below accounts for many of these factors. 5. Results The cost of being an immigrant. Table 2 summarizes estimates of the coefficient of Intermarried in 3 regressions of time spent in Chores by immigrants: an OLS 19

20 regression and two Two-Stage regressions. The OLS coefficients in Panel A imply that for immigrants intermarriage has no relationship with chores since the coefficients for both men and women are not significantly different from zero. However, this conclusion does not hold once we recognize the presence of selection into intermarriage and use the Two-Stage technique described above that estimates the probability of intermarriage in a first stage. Panels B and C report intermarried effects for immigrants, using two measures of availability of immigrants in region of residence as instruments: Z1 (in panel B) and Z2 (in panel C). The Two-Stage regressions reveal that intermarried immigrants spend significantly more time performing chores than their endogamous counterparts. In the case of female immigrants the coefficients of intermarried are large and significant at the 5% level, suggesting that intermarriage to a native increases immigrant women s daily involvement in chores by around two hours (using instrument Z1) and 1h 30 minutes (using instrument Z2). The coefficients in the regressions for men are also positive but significant only at the 10% level suggesting that many intermarried immigrant men also spend about 1h more in chores per day than men married to other immigrants, yet the pattern is not as consistent for men as it is for women. Since when using instrument Z2 we obtain larger first stage F-statistics and the coefficients obtained with this instrument are more conservative, we only use this instrument in further regressions. The premium obtained by US-born natives. OLS regression coefficients in Panel D do not reveal any effect of intermarriage on chores. As in case of immigrants, this conclusion changes after accounting for endogeneity of intermarriage and chores. Panel E in Table 2 presents the coefficients of intermarried for native men and women using Two-Stage regressions for the full sample of natives. We find that native women who 20

21 are predicted to be married to an immigrant because of favorable immigrant availability ratio at the time of marriage (stage 1 regression) are doing significantly fewer chores (about 1h 20 minutes less) than their counterparts predicted to be married to a native. The same does not hold for men: there is no significant impact of intermarriage on the amount of time the whole sample of native men spend doing chores. Interpretations. We predicted that immigrants intermarried with natives would work more at chores than their counterparts in all-immigrant marriages, whereas natives married to immigrants would work less at chores than their counterparts in all-native marriages. Evidence consistent with these predictions was found for both immigrant men and women when using a two-stage estimation method. Our finding that native women married to immigrants work less at chores than native women in all-native marriages is also consistent with our predictions. Our results echo those of Grossbard, Gimenez and Molina (2014; GGM). In both studies intermarried respondents were compared to their counterparts who did not intermarry, Instrumental Variables were used to obtain predicted probability of intermarriage, and estimation was performed in two stages. According to GGM black women seem to pay a penalty when they intermarry with whites to the extent that they work more at chores than if they were in all-black marriage, and white women intermarried to black men perform fewer chores than their counterparts in all-white marriages. Together the two studies suggest that in U.S. marriage markets the demand for blacks and immigrants is low relative to the demand for whites and natives. It is noteworthy that our predictions were only confirmed when we used a two-stage estimation, not when we used OLS. It is possible that with OLS the true positive effect of IM on chores performed by immigrants is not apparent, for it is neutralized by negative effects due to unobservable traits leading to intermarriage and less need for 21

22 chores in the household. The opposite occurs when we use OLS to estimate chores of natives as a function of intermarriage: the native men who marry immigrant women also want more household production for unobservable reasons. This implies more chores supplied in their households. This positive preference neutralizes the true negative effect of a native/immigrant intermarriage on natives chores. Could the penalties and premia of intermarriage that we uncovered possibly be the result of other factors? For example, could they be the result of the inferior labor market opportunities that immigrants face? Immigrants who marry a native may improve their labor market opportunities, so the fact that we find them working more at chores than their endogamous counterparts can t be explained by differences in labor market opportunities between the intermarried and those who marry other immigrants. The first two columns of Table 3 present the coefficients of interest obtained from two-stage models using instrument Z2 when controlling for the number of hours the respondent and his/her spouse usually works for pay per week. Panel A is for immigrants and Panel B for native respondents. The coefficients of intermarried in the regression for immigrant men and women in the first two columns of Table 3 (Panel A) are very similar to the coefficients we obtained when not controlling for usual hours of work and relative income in Table 2 (panel C). Likewise, the coefficients of intermarried in the two-stage estimations for native men and women (Table 2, panel E) are similar to those coefficients in Table 3. It can be seen that when the new controls are included it is still the case that intermarried immigrants work more at chores and intermarried native women work less at chores than their endogamous counterparts. The penalty for immigrants and the premium for natives are unlikely to be due to differences in hours of work between natives and immigrants. 22

23 The intermarried immigrants also have considerably higher income than their endogamous counterparts. Higher income is expected to be associated with less time devoted to chores. Again, we can rule income differences as a possible explanation for the chores penalty that intermarried immigrants are paying. Coefficients for other variables. Appendix A reports full regression results for immigrants that correspond to the results reported in panels A and C of Table 2. It can be seen from the last two columns in Appendix A that female immigrants work 19 minutes longer at chores when they are not US citizens, possibly a reflection of the value of being a US citizen in marriage or labor markets. However, the size of the coefficient of not US citizen is much smaller than that of intermarried. Immigrant men married to non-citizens work 14 minutes less at chores. This suggests that noncitizens pay a price for marrying a citizen: they do relatively more of the chores. Initially we had reservations about including Not US citizen indicators among controls because potentially it could be an indicator for recent immigrant. However our sample includes many non-citizens among immigrants who have been in the US for 20+ years. Thus the coefficients can reliably be interpreted as a citizenship effect. It can also be seen from Appendix A that, as expected, the higher the ratio of available partners from the same region the lower the probability of intermarriage to a US native (estimated in stage one). A higher probability of intermarriage is also related to more years spent in the US, being more educated and white, a larger difference between spouse s and own education, and a larger difference between spouse s and own age. Everything else held constant, immigrants from the English-speaking countries, Western Europe or Mexico are more likely to intermarry than immigrants from other regions. Overall, the coefficients are consistent with findings in the literature (see Adsera and Ferrer 2014 and Furtado and Trejo 2013). 23

24 Time spent in chores is not affected by spouses differences in age and schooling, yet larger differences in both of these variables increase the probability of intermarriage. Married women s time in chores is lower for second generation and child immigrants compared to immigrants who arrived as adults. Women with a higher education spend substantially less time in household production; Hispanic women spend more time in chores than other women. Men s time in chores is not significantly affected by race, education, years since migration, region of origin and the presence of children. Most region of origin dummies are not significant for women either, possibly due to our modest sample sizes. Yet the few impacts that can be identified suggest that women from East Asia, India, Bangladesh, Pakistan and Mexico spend more time in chores than other immigrant women. Results including all coefficients used in the two equations estimated for natives are presented in Appendix B. It can be seen that native men whose spouses are not citizens work 14 minutes less at chores than those whose spouses are citizens, but spouse not US citizen does not have a significant effect on native women s time in chores. The immigrant availability measure Z3 is strongly associated with intermarriage to an immigrant, and so is education. Separate results for dual-earner and single-earner traditional couples. We deal with the endogeneity of labor force participation by estimating our two equation models for two subsamples: dual earner couples and traditional couples defined as couples in which the wife is not employed (or working less than 10 hours a week), but the husband is. The coefficients of interest for these regressions are reported in the last four columns of Table 3. It can be seen that there is no intermarriage penalty for immigrants in dual earner couples, male or female: when both spouses are employed immigrant men and women 24

25 spend as much time in chores when intermarried as they do when married to another immigrant. In contrast to the results for dual-earner couples with at least one immigrant, regressions for dual-earner couples with at least one native show large and negative coefficients of intermarriage for men and women, suggesting that working natives who are married to working immigrants spend significantly less time in chores than otherwise similar natives with US-born spouses. Intermarried native men work 1h 40 minutes less at chores, and intermarried native women reduce their chores time by about 2 hours per day. The last two columns of Table 3 present results of chore time for traditional couples where the husband is employed but the wife is not (or employed less than 10 hours). Relative to their counterparts married to immigrants, traditional intermarried immigrant women perform 3 hours of extra chores per day. Wives who are not US citizens work an additional 30 minutes at chores. The results in Panel B show that native women married to immigrants and following a traditional division of labor devote 2 hours of chores less than they would have if they had been married to native men. It thus appears that among natives those who gain the most from being married to immigrants are men and women in dual-earner couples and women in traditional couples. Among immigrants (panel A) the chores penalty mostly originates for intermarried traditional women. An advantage of analyzing dual earner couples is that we have actual wage information for both members of such couples. To the extent that chores are agreed upon as part of bargaining that occurs within the couple, and that higher relative wages give more bargaining power, we expect amount of time spent on chores to vary directly with the spouse s wage and inversely with own wage. We don t find that among dual earners with immigrant respondents. However, among natives we find that women with 25

26 higher own wage work less at chores and that both men and women with spouses who earn higher wages spend more time at chores. This indicates more of an effect of the spouse s wage on chores performed by natives than on chores performed by immigrants, possibly due to the considerably larger size of the native sample. In all time use regressions in Table 3 we control for normalized income defined as the difference between household income and state s median income. Even though the time spent in chores is lower for richer dual earner couples, the intermarriage premia and penalty are not explained away by income. Household income does not have an impact on chores in most regressions. We find that the citizenship effect is stronger for traditional women than for women in dual earner couples, which suggests that lack of citizenship status does not only limit labor market options: it may also have a negative effect on women s value in the marriage market. 6. Robustness checks We tested the robustness of our results by expanding the definition of household production, using broader categories that include tasks that are less likely to be chores. Results are found in Table 4 for the entire sample of immigrants, and for immigrants in dual earner and traditional couples, as defined above. Panels A and B report the coefficients of Intermarried in regressions for broader categories of household production, those that include shopping and those that include not only shopping but also adult- and childcare. It can be seen that expanding the definition of household production beyond chores, and thus including more activities that are likely to be enjoyable, leads to the disappearance of significant effects of intermarriage on household production performed by all immigrant women in our 26