Culture and gender allocation of tasks: source country characteristics and the division of non-market work among US immigrants

Abstract

There is a well-known gender difference in time allocation within the household, which has important implications for gender differences in labor market outcomes. We ask how malleable this gender difference in time allocation is to culture. In particular, we ask if US immigrants allocate tasks differently depending upon the characteristics of the source countries from which they emigrated. Using data from the 2003–2017 waves of the American Time Use Survey (ATUS), we find that first-generation immigrants, both women and men, from source countries with more gender equality (as measured by the World Economic Forum’s Global Gender Gap Index) allocate tasks more equally, while those from less gender equal source countries allocate tasks more traditionally. These results are robust to controls for immigration cohort, years since migration, and other own and spouse characteristics. There is also some indication of an effect of parent source country gender equality for second-generation immigrants, particularly for second-generation men with children. Our findings suggest that broader cultural factors do influence the gender division of labor in the household.

Appendices

Theory appendix and additional results

Table 10 Effect of Years Since Migration (YSM) on non-market work for immigrant couples migrating together

Table 11 Effect of source country characteristics on non-market work for immigrants, GGI subindexes

Table 12 Effect of wife earning more than husband on total non-market work for women, dropping relative earnings from 0.489 to 0.50

Table 13 Effect of wife earning more than husband on total non-market work for men, dropping relative earnings from 0.489 to 0.50

Consider a stylized framework where a husband and wife each have $t_g \in [0,1]$ units of time to allocate toward market production and $1 - t_g$ units to allocate toward non-market production, for $g \in \{ m,f\}$. The household has a joint utility function that is increasing and concave in total market good consumption c and household good consumption b. Utility also depends additively on a convex gender norm function $\pi (t_f,t_m)$ that penalizes the wife’s market work relative to the husband’s market work. Market goods can be purchased with wages earned from market work, while household goods are produced via non-market production. The husband and wife can differ in their exogenous wage rate w_g and productivity in home production.

Couples maximize utility subject to a market good consumption (income) constraint and their household non-market goods production function. Before putting any more structure on the model, note two main general results. First, couples will optimize their time at the point when the marginal utility gain from market work is equal to the marginal utility gain of non-market work. Second, the wife’s market work will decrease in the magnitude of the gender norm (and hence her non-market work will increase), while the husband’s market work will increase in the magnitude of the gender norm (and hence his non-market work will decrease). In the absence of income effects, market work will increase in the wage rate and non-market work will increase in home productivity for both men and women.

To see this in a concrete example, we assume that households solve the following maximization problem, where utility is additively separable with a linear gender norm penalty and a Cobb-Douglas household production function:

$$\mathop {{\max }}\limits_{t_m,t_f} c + \varphi \,\log \,b - \pi \left( {t_f - t_m} \right)$$

$${\rm{s}}.{\rm{t}}.\,{c} = w_mt_m + w_ft_f$$

$$b = A\left( {1 - t_m} \right)^\alpha \left( {1 - t_f} \right)^{1 - \alpha }$$

where $\alpha \in (0,1)$ is the relative productivity of the husband in household good production and π is a positive scalar. First order conditions yield the following solution for optimal non-market work time:

$$1 - t_f^ \ast = \frac{{\varphi \left( {1 - \alpha } \right)}}{{w_f - \pi }}$$

$$1 - t_m^ \ast = \frac{{\varphi \alpha }}{{w_m + \pi }}$$

The general conclusions from above may be seen explicitly in these first order conditions. That is, non-market work is increasing in π for women and decreasing in π for men. Further, since this model does not have income effects, non-market work is increasing in relative productivity and decreasing in the wage rate for both men and women. For market work, the converse holds. Note that specific parameterizations must be checked for corner solutions (namely, if the penalty is too large, the wife will spend all her time in home production).

This framework can easily accommodate modeling the wage rate as an endogenous function of the gender norm penalty. For example, we may imagine the gender norm representing a general attitude about women’s accumulation of human capital and thus a stronger penalty decreases her market productivity and hence her wage rate. In this case, the first order conditions for optimization remain the same, though we need to explicitly note the dependence of the wife’s wage on π for the comparative statics with respect to the penalty. If we assume that the wife’s wage decreases in the gender norm, the direction of the effect on non-market allocation is the exact same but with a stronger magnitude.

While the theoretical model considers wages, our empirical work is reduced form with respect to wages in order to avoid well-known problems of estimating (imputing) a wage rate for labor force nonparticipants. This makes it important to control for variables related to human capital to isolate the effect of the penalty. However, the model discussed in this preceding paragraph suggests that those variables are themselves endogenous, motivating our empirical strategy of testing the effect of source country gender norms with and without the inclusion of human capital controls such as age and education.

Data appendix

2.1 Variable definitions

2.1.1 Demographic variables from the ATUS and CPS

Race and ethnicity

• We control for race and ethnicity using a set of indicator variables for five mutually-exclusive categories: White non-Hispanic, Black non-Hispanic, Asian non-Hispanic, other non-Hispanic, and Hispanic.

• Respondent is classified as Hispanic if the respondent reports being Hispanic or reports his/her ethnicity as Spanish, Portuguese, Mexican, Puerto Rican, Latin American Indian, South American Indian, or Mexican American Indian.

• Respondent is classified as black non-Hispanic if the respondent reports being any detailed race that includes black and is not classified as Hispanic.

• Respondent is classified as Asian non-Hispanic if the respondent is not classified as Hispanic or black non-Hispanic and reports race as Asian or any mixed race including Asian.

• Respondent is classified as white non-Hispanic if the respondent is not classified as Hispanic, black non-Hispanic, or Asian non-Hispanic and reports race as white.

• Respondent is classified as other non-Hispanic if none of the above classifications apply.

First- and second-generation immigration variables

• Respondents are classified as first generation if they report their birthplace as outside the 50 states or the District of Columbia. Note that we count respondents born in US territories as immigrants, though they are not included in most immigrant analyses due to not having independent source country characteristics.

• Respondents are classified as second generation if they were born in the fifty states or the District of Columbia and they report that either of their parents was born outside the United States.

  • Second-generation immigrant respondents are assigned the source country characteristics of their mother unless their mother’s source country characteristics are missing or their mother is US born. In that case, they are assigned their father’s source country characteristics. We follow this procedure because the high correlation between father’s and mother’s birthplace when both are foreign born makes estimating separate effects of father’s and mother’s source country difficult in a sample of this size. (Of second-generation immigrants where both parents are foreign born, both parents are from the same source country for 87.2 percent of our sample.) We compare our results to those that prioritize father source country characteristics and find them to be similar.

• We use the term native to refer to those who were born in the United States, with both parents born in the United States. That is, natives may be considered third-plus-generation immigrants.

• We compute years since migration as the difference between the survey year and the mid-point of the binned response to year of immigration. The mid-point of the binned response to year of immigration is also used to assign immigration cohorts.

Earnings variables in Section 8

• We set earnings to 0 for those reporting to be unemployed or out of the labor force. We set earnings to missing if reported earnings are outside the ATUS defined range. We inflate top-coded earnings by 1.5. According to IPUMS: “[The individual earnings variable] was collected at the time of the ATUS interview. However, earnings information was only collected at that time for respondents who changed jobs or employers since the final CPS interview, or whose earnings were allocated in the final CPS interview. For other respondents, earnings information was carried forward from the final CPS interview” (https://www.atusdata.org/atus-action/variables/EARNWEEK#description_section). Further, spousal earnings are not updated at the time of the ATUS interview, but are always carried forward from the final CPS interview. The final CPS interview takes place two to five months before the ATUS interview.

• Our measure of total household income comes from the CPS family income variable, which includes earned and unearned income of all household members. This variable is not updated at the time of the ATUS interview. Family income is reported in ranges, and we use the mid-point of reporting bins.

• We set log earnings to zero for those with 0 earnings.

• Relative earnings are defined as the wife’s earnings divided by the sum of the husband’s and wife’s earnings. If only one spouse is employed, relative earnings are set to 1 or 0 accordingly.

2.1.2 Country characteristics variables

Total fertility

• Total fertility data come from the World Bank, available at http://data.worldbank.org/indicator/SP.DYN.TFRT.IN. In the regressions with country characteristics, we include 2000–2007 country averages of total fertility.

GDP per capita

• Most GDP per capita data come from the World Bank, available at http://data.worldbank.org/indicator/NY.GDP.PCAP.PP.KD. GDP for Argentina, Burma (Myanmar), and Syria is constructed from UN Stats data on GDP by Type of Expenditure at current prices and at constant 2005 prices in national currency units, available at http://data.un.org/Data.aspx?d=SNAAMA&f=grID%3A101%3BcurrID%3ANCU%3BpcFlag%3A0 and http://data.un.org/Data.aspx?q=gdp&d=SNAAMA&f=grID%3a102%3bcurrID%3aNCU%3bpcFlag%3a0, respectively. PPP conversion rates come from http://icp.worldbank.org/icp/QueryResults.aspx?r=-1&ds=0&y=3&ws=1. We use the World Bank methodology to convert to GDP per capita, PPP.

Global gender gap index

• The index of gender equality comes from the World Economic Forum’s “The Global Gender Gap Report”. In the regressions with country characteristics, we include 2006–2007 country averages of the index, unless a 2006 value is not available, in which case we use the earliest value available up until 2012. The 2006 and 2007 reports are available at http://www3.weforum.org/docs/WEF_GenderGap_Report_2006.pdf and http://www3.weforum.org/docs/WEF_GenderGap_Report_2007.pdf.

2.1.3 ATUS variables

We define housework as all activities that fall under the broad ATUS “Household Activities” category. These include housework, food and drink preparation and cleanup, home maintenance, lawn and garden care, pet care, appliance care, and household administrative tasks. We define primary childcare as care for children living in the household, including the “second-tier” ATUS categories of “Caring for & Helping HH Children”, “Activities Related to HH Children’s Education”, and “Activities Related to HH Children’s Health”. We define total non-market work as the sum of these housework and childcare variables, as well as time spent grocery shopping and all the activities included in the “second-tier” ATUS categories of “Caring for Household Adults”, “Helping Household Adults”, and “Caring for & Helping NonHH Members”.

2.2 Sample selection

We use data from the 2003–2017 waves of the American Time Use Survey (ATUS). For the main analysis, we focus on first- and second-generation immigrants as defined above. When natives are included, they are defined as those who are born in the United States with both parents born in the United States. Regressions are weighted by ATUS sampling weights that are normalized to provide equal weighting for each sample year.

We restrict our sample to married respondents in heterosexual partnerships where both the respondent and their spouse are between the ages of 18 and 64. To do this, we keep only respondents who report being married with a spouse present in the household. To collect partner characteristics, we match these respondents to the member of their respective households who lists the respondent as a spouse. If no household member lists the respondent as a spouse, we match the respondent to the member of the household who lists the respondent as an unmarried partner. We then drop any remaining respondents who do not match. We also exclude observations recorded on holidays, natives born abroad, and immigrants whose year of immigration is missing. While we are not able to observe the time allocation of the respondent and their partner, by enforcing these restrictions, we can estimate for the population how married men and women divide household labor. All analyses were repeated including those in heterosexual partnerships but not married. To do this, we do not impose the initial restriction of keeping only respondents claiming to be married with a spouse present. We follow the same partner match procedure outlined above. Our results are similar when partnered respondents are included (results available on request).

In some regressions, the sample is implicitly restricted due to missing data on control variables. Regressions that control for a range of spousal characteristics resulted in a few dropped observations that had missing values for any of these characteristics. Further, some immigrants did not match to source country characteristics and are therefore dropped from regressions including these characteristics. Namely, some source countries do not have GGI scores available, a list that includes any US territory, Taiwan, Haiti, Iraq, Hong Kong, and Laos.

We also recoded the source country of some immigrants to make matching possible. This included assigning individuals from England, Scotland, Wales, and Northern Ireland to the source country characteristics of the UK; assigning individuals from Azores to Portugal; assigning individuals from Kosovo to Albania; assigning individuals from Palestine to Israel; and assigning individuals from “USSR, n.s.” to Russia.