Missing Time with Parents: Son Preference among Asians in the United States

Abstract

We study prevalence of son preference in families of East and South Asian origin living in the U.S. by investigating parental time investments in children using American Time Use Surveys. Estimates show that East and South Asian mothers spend an additional hour of quality time per day with their young (aged 0–2 years) sons than with young daughters; son-preference in mothers’ time allocation declines as children get older. East and South Asian fathers’ time with young children is gender neutral. We find gender specialization in time with children aged 6–17 with fathers spending more time with sons and mothers spending more time with daughters.

Introduction

An extensive body of research documents the existence of son preference in many East and South Asian societies. These studies find that daughters are less likely to be born, and if born, less likely to live past childhood, go to school, receive medical treatment when sick, and live above subsistence compared to sons.¹ In this paper, we investigate if son preference or discrimination against daughters persists in families of East and South Asian origin in the U.S., a fast growing ethnic group, by studying the quantity and quality of parental time investment in children.

Previous research shows that East and South Asian immigrants in the U.S. and Canada have boy-birth percentages at higher parity (second or higher births) that exceed what is biologically normal especially if previous children were girls suggesting that these immigrant parents exercised sex selection.² However, there is no research on whether parents of East and South Asian origin in the U.S. or in other non-Asian countries show preference for sons in allocating family resources including parental time on childcare and other activities with children, a critical, yet least studied, developmental input that can impact abilities and outcomes later in life (Heckman, 2006). Examining East and South Asian immigrants in the U.S. can provide insights into whether the root cause of son preference in East and South Asia is economic or cultural.

Bias against daughters is often linked to cultural norms that relegate daughters to a lower status than sons. For instance, in India and China certain religious and funerary rituals can only be performed by sons (Chung & Gupta, 2007; Das Gupta et al., 2003). Family lineage in these and other patriarchal societies is traced through male offspring. Social institutions and norms in East and South Asian societies also limit the economic and educational opportunities of daughters and create a discriminatory environment against them.³ Additionally, institutions that strengthen and perpetuate these cultural norms make investments in daughters bad economics. High cost of dowry, for instance, implies that daughters are a financial burden on families whereas sons draw dowry into the family. Daughters depart to join their husband’s family after marriage and thus returns on any investments in daughters are unlikely to be reaped by their parents (Das Gupta et al., 2003; Dyson & Moore, 1983; Miller, 1985; Oldenburg, 1992; Rahman & Rao, 2004). Further, because of lack of institutions for elderly care in these countries, sons are considered the primary support in old age and therefore investments in sons have economic payoffs in old age (Chung & Gupta, 2007).

To the extent that economic factors are its primary cause, we expect to find little or no gender bias in parental investments in families of East and South Asian origin in the U.S. where labor market prospects of women are significantly better, where nearly universal Social Security benefits weaken dependence on sons for old age economic support, and where East and South Asian immigrants live in much improved economic conditions. On the other hand, if gender bias is rooted in culture, we expect parental investment in East and South Asian households to reflect son preference or greater son preference compared to other households.

Our research design is based on a comparison group approach. We use U.S. natives or immigrants from non-Asian countries as comparison groups to investigate if there is a pattern in parental investments that is similar across parents from various regions of origin. If the pattern of investment in sons and daughters is similar across families from different regions that would be an indicator that there may be differences across genders that make it optimal for parents to invest more time with children of a certain sex and that differential investment by child’s gender is not specific to East and South Asian parents.

A common assumption in the studies on the prevalence of son preference in allocation of family resources is that boys and girls live in families with similar characteristics. This assumption is untenable given previous research that has found prevalence of sex selection in East and South Asian families in Canada, South Africa and the U.S. If fertility is driven by the desire to have a certain number of boys, as has been documented in East and South Asian countries, girls will end up in families with more children and therefore fewer resources per child. The simple difference in allocation of resources could be due to heterogeneity between families with sons versus those with daughters, and may not necessarily be an indicator of gender discrimination.

We conduct several tests to investigate if our estimates are afflicted by this bias. First, we run models with family fixed effects that help address the issue of sex selection and other differences between families e.g., family size, that may cause differences in time investments on children and may be correlated with the gender of the child. A limitation of fixed effects models is that gender will be correlated with other differences between boys and girls within families (Behrman, 1997; Datar, Kilburn, and Loughran, 2010). In our empirical analysis we control for a rich set of child characteristics such as age, birth order and previous birth spacing to address this issue.

Second, following Barcellos, Carvalho, and Lleras-Muney (2014), we examine gender discrimination in parental investments in first-born children aged 0 to 2 years. This specification relies on the assumption that the parents of children aged 0–2 will not yet have had more children in response to the gender of their youngest child and that parents do not exercise sex-selective abortion for the first born. In supplementary analyses, we investigate if having a son influences the division of household work between parents. Specifically, we study whether presence of a son aged 0–2 changes the time parents (mother or father) spend on household chores and childcare. This analysis is also restricted to families with at least one child aged less than two.

Finally, we conduct a number of tests to assess the prevalence of sex-selection and male-biased fertility stopping among children in East- and South Asian families and estimate the direction and magnitude of bias on account of these practices.

We use data from the American Time Use Survey (ATUS) from 2003–2012. A unique feature of these data is that they provide detailed information on how much time in a given day a parent spent with each child, how the time was spent, and who else was present during each activity. A challenge to studying gender discrimination in allocation of family resources is that researchers often have to rely on household-level data to estimate individual-level allocation for which data are often not available (Kingdon, 2005). The advantage of using ATUS data is that we can study parental time investments made to each child in the family separately.

Our results show that mothers of East and South Asian origin spend 30 more minutes of quality time per day with their young sons (aged 0–5) compared to their young daughters. Further, son-preference in mother’s time is prevalent primarily in the first two years of a child’s life with mothers spending an hour more of quality time with sons compared to daughters. In contrast, mothers of other ethnic origin allocate time in a gender-neutral manner. We find that fathers of East and South Asian, European and Latin American origin are gender neutral in their time investments in young children while U.S. native fathers spend five more minutes of quality time per day with sons compared to daughters. With school-age children, parents across ethnic origins specialize along gender lines: mothers spend more time with daughters while fathers spend more time with sons.

Empirical Evidence on Gender Bias in Parental Investments in Children

Research on gender bias in parental investment has centered on developing countries particularly in East and South Asia where girls have higher mortality rates than boys while the mortality gap is non-existent or reversed in other countries with comparable or even lower economic prosperity and higher poverty (El-Badry, 1969; Guilmoto, 2009; Sen, 1990; UN, 2011). Compared to boys, girls in East and South Asia receive fewer health inputs including less prenatal care (Bharadwaj & Lakdawala, 2013), less medical treatment when ill (Chen, Huq, & D’Souza, 1981; Khanna, Kumar, Vaghela, Sreenivas, & Puliyel, 2003), and poorer nutrition (including shorter duration of breastfeeding) (Barcellos, Carvalho, & Lleras-Muney, 2014; Deaton, 2008; Haddad, Peña, Nishida, Quisumbing, & Slack, 1996; Marcoux, 2002), especially in families with several daughters (Das Gupta, 1987; Pande, 2003) which may, at least in part, explain the gender mortality gap.⁴

Research on education also points to a pro-male bias in East Asia, South Asia, the Middle East, North Africa, Sub-Saharan Africa but not in Latin America or Southeast Asia (Bauer, Wang, Riley, & Zhao, 1992; Dancer & Rammohan, 2007; Dayioğlu, Kirdar, & Tansel, 2009; Grant & Behrman, 2010; Kingdon, 2005; Lancaster, Maitra, & Ray, 2008; Li & Tsang, 2003; Ota & Moffatt, 2007). Mishra, Roy and Retherford (2004) argue that presence and extent of gender discrimination largely depends on the birth order of the index child and the sex composition of older living siblings. They find that discrimination against girls is most visible in families with no living sons, particularly at birth orders 3 and 4+. The lack of evidence of discrimination against girls in other families could be on account of gender selection or heterogeneity in families with boys versus girls.

Two papers have investigated presence of gender bias in parental time with children in developing countries. Barcellos, Carvalho, and Lleras-Muney (2014) examine gender bias using the Indian and South African Time Use Surveys and find that boys receive more childcare than girls in India, but find no gender differences in South Africa. They explicitly assume absence of sex-selective abortion or infanticide, which is implausible given the extensive prevalence of sex selective abortion and reports of infanticide across India during the period covered by their study. Brown (2006) examines if parents spend more time helping boys versus girls on homework in rural China and finds no gender differences. However, Brown’s (2006) analysis does not account for differences between boys’ and girls’ families.

In recent years, researchers have turned attention to gender bias in parental investments in developed countries. Studies based on U.S. data have two primary findings. First, fathers invest more time in sons than daughters and mothers invest more time in daughters than sons (Lundberg, Pabilonia, & Ward-Batts, 2007; Mammen, 2011; Yeung, Sandberg, Davis-Kean, & Hofferth, 2001). Second, time investment in children varies by birth order: parents spend more time on first-born children than second-born children (Price, 2008).

In the U.S., researchers have also investigated how son preference affects parental behaviors, including fertility, marital status and work. Empirical evidence shows that first-born daughters have more siblings than first-born sons (Dahl & Moretti, 2008; Lundberg, 2005) and fathers work more hours and earn more after the birth of a son, which likely influences resources available for investments in children (Lundberg & Rose, 2002).⁵ The last finding has also been replicated in German data (Choi, Joesch, & Lundberg, 2008). Further, women in the U.S. with first-born daughters are less likely to be married and if married more likely to get divorced compared to those with first-born sons (Dahl & Moretti, 2008; Lundberg, 2005). These studies show that boys and girls grow up under different family conditions. Thus, studies of parental investment in children that disregard family heterogeneity are likely to arrive at biased conclusions.

A second category of research has examined differences in children’s own time use by gender. A majority of these studies have also focused on developing countries. Larson and Verma (1999) review this large literature and conclude that in most developing country settings, boys have more free time than girls. More recent studies find corroborating evidence in India (Motiram & Osberg, 2010), Malawi (Nankhuni, 2004), and in Tanzania, Uganda, South Africa, and Kenya (Kes & Swaminathan, 2006). Larson and Verma’s review also finds that in almost all regions of the world, and in both developed and developing countries, girls spend more time in household labor than boys except in the United States where they find no gender differences. These studies, too, assume that boys and girls live in families with similar characteristics, an assumption that is rejected by previous research. When families prefer sons and follow male-biased stopping rules in childbearing, girls will end up in larger households than boys and receive less parental investments even when parents themselves do not discriminate within the household (Yamaguchi, 1989). We control for such fertility preferences and other observed and unobserved family characteristics using a number of strategies including models with a comparison group approach comparing families of East and South Asian origin with those of other ethnic groups; models that restrict samples to families with first-born children aged 0–2; and models with family fixed effects.

Data

Our study uses American Time Use Survey (ATUS) data for 2003–2012. ATUS, conducted annually, is a nationally representative survey of how people spend their time. ATUS surveyed about 136,000 households from 2003 to 2012. From each eligible household one person aged 15 years or more is randomly selected to complete the survey.⁶ Respondents are asked to recall all their activities in the 24-hour period starting at 4 am the previous day, the location of each activity and who else was present during the activity.

ATUS collects demographic information of the respondent and each household member. We refer to the respondents’ co-resident children and grandchildren under the age of 18 as children.⁷ We exclude from our sample respondents who do not have children. Because prevalence of single parent households may differ across ethnic groups, we further restrict the sample to two parent families.⁸ We use information on the respondent’s, respondent’s mother’s and respondent’s father’s country of birth to determine country of origin. The focus of our study is respondents who were born, or have a parent born, in East and South Asia. For comparison we study three other groups: U.S.-born respondents who have U.S.-born parents (henceforth referred to as U.S. natives),⁹ first- and second-generation respondents from Latin America; and first- and second-generation respondents from Europe, Canada, Australia, and Pacific. For convenience, throughout this paper, we use the term Europeans to describe first and second generation immigrants from Europe, Canada, Australia, and Pacific. Appendix Table 1 presents the composition of our East and South Asian sample by country of origin. Eighty percent of the respondents originating from East and South Asia are first- and second-generation immigrants from five countries: China, India, Japan, Pakistan, and South Korea.

Because ATUS collects data from only one person in the household, we observe children’s time use as they interact with the respondent. We therefore have complete information on the time that a respondent parent and his or her children spend together. Following Price (2008), we define quality time as time the child spent with the parent on activities where the child is either the focus of the activity or is interacting considerably with the parent.¹⁰

Table 1 presents descriptive demographic data on East and South Asian families with male and female children and shows that these families are similar on many important characteristics such as child’s age, number of children, whether the respondent (i.e., parent) is female and parent’s age.¹¹ In fact, family characteristics in our East and South Asian sample of male and female children are similar on all characteristics except one: the probability that there is a subsequent birth within an interval of two years is statistically higher for daughters than sons. That, there are not many observed differences in our data between male and female children, and most importantly, between the families of male and female children, suggests low likelihood that differences in parental time use by gender are confounded by unobserved factors. We assess the sensitivity of estimates of time allocation by gender by estimating models with and without these controls.

Table 1.

Summary characteristics of East and South Asian children aged 0–17 years and their families

	Boy Sample		Girl Sample
	Mean	SD	Mean	SD
Age, years	7.05	4.91	7.04	4.86
Birth order	1.58	0.75	1.57	0.72
Previous birth interval: 1 year	0.05	0.22	0.05	0.22
Previous birth interval: 2 years	0.11	0.32	0.12	0.33
Previous birth interval: 3 years	0.09	0.29	0.08	0.27
Previous birth interval: 4 years	0.07	0.26	0.06	0.24
Previous birth interval: 5+ years	0.12	0.33	0.14	0.35
No previous birth	0.55	0.50	0.54	0.50
Subsequent birth interval: 1 year	0.05	0.22	0.05	0.22
Subsequent birth interval: 2 years	0.10	0.30	0.12	0.33†
Subsequent birth interval: 3 years	0.08	0.27	0.09	0.28
Subsequent birth interval: 4 years	0.06	0.24	0.06	0.24
Subsequent birth interval: 5+ years	0.11	0.31	0.09	0.29
No subsequent birth	0.60	0.49	0.59	0.49
Household size	4.24	1.03	4.29	1.05
Number of children	2.02	0.82	2.03	0.80
Number of other boys	0.52	0.67	0.54	0.66
Number of other girls	0.50	0.63	0.49	0.64
Sons-only family	0.57	0.50	-	-
Daughters-only family	-	-	0.54	0.50
Mixed-sons-&-daughters family	0.43	0.50	0.46	0.50
Respondent (parent) is female	0.52	0.50	0.53	0.50
Respondent’s (parent’s) age	40.00	7.50	39.89	7.69
Mother is unemployed	0.46	0.50	0.46	0.50
N	1,216		1,138

Note:

^†indicates that mean for boys and means for girls are different at the 10% significance level. The test accounts for correlation between siblings within families using a sandwich estimator clustered on family.

We also examine the influence of children’s gender composition on parents’ time allocation towards childcare and household chores. Time parents spend on childcare is the sum of time spent caring for and helping household children and on activities related to household children’s education and health. Time on household chores is the sum of time spent on activities coded by ATUS as “household activities”, “household services”, and “grocery shopping”.

Empirical Strategy

We first study differences in parental time that sons and daughters receive in families of various ethnic origins living in the U.S. This analysis uses child-level data that we create using time diaries from parents that contain information on how much time parents spend with each child. Our data provide the time diary of one parent (father or mother) in the family. Because time allocation on children is gender specific, we study the difference in fathers’ time between sons and daughters and in mother’s time between sons and daughters separately. Equation (1) describes the model specification estimated on a combined sample of East and South Asian families and the three comparison groups drawn from the American Time Use Survey for mothers:

$$T_{ij}={\alpha }_0+\sum _e{\alpha }_{e}Et{h}_{ej}+\sum _e{\beta }_{e}Et{h}_{ej}\times ChildMal{e}_{ij}+X_{ij}B+u_{ij}$$ (1)

where $T_{ij}$ is the time that a mother from family j spends with child i and is a function of the mother’s ethnic origin ($Et{h}_j$), child’s gender ($ChildMale$), child characteristics (denoted by X) namely child age (a dummy variable for each year of age), birth order (a set of dummy variables indicating first, second, or higher birth order), previous birth spacing (a set of dummy variables indicating child was born 1, 2, 3, 4, or 5 or more years after previous child, or no previous child), and $u_{ij}$ is the error term.¹² $Et{h}_j$ has four ethnic categories, denoted by subscript e: East and South Asian origin; Latin American origin, European origin and U.S.-born natives (comparison category). All four ethnic categories are interacted with the dummy variable ChildMale (equals to 1 if the child is male, otherwise 0). Because ChildMale is interacted with all four ethnic categories, we do not need to include a control for ChildMale separately. Equation (1) restricts the effect of child characteristics (X) other than gender to be the same across ethnic groups. But in our empirical analysis, we drop this restriction.¹³

In this equation, vector ${\beta }_e$ measures the difference between the time mothers of ethnicity e spend with their sons and the time they spent with their daughters. A similar equation is used to estimate if fathers practice son-preference in the time they spend with their children.

Equation (1) assumes that the gender of the child is randomly determined and, that there is no difference between families with more sons and those with more daughters. Prevalence of sex selection and male-biased fertility stopping rules in East and South Asian families make it likely that these assumptions are not valid. One way to address these issues is to compare parental time investments in sons versus daughters within families. The model specification for this analysis is given by equation (2):

$$T_{ij}={\tilde{\alpha }}_0+\sum _e{\tilde{\beta }}_{e}Et{h}_{ej}\times ChildMal{e}_{ij}+X_{ij}\tilde{B}+{\pi }_j+e_{ij}$$ (2)

which has one additional term (compared to equation (1)): π_j denotes a complete set of family fixed effects that capture family heterogeneity. We have differentiated the parameters in this equation from those in equation (1) using the symbol ~. The main effect of the mother’s ethnicity variable drops out from the model because of the family fixed effects. We run our analyses separately by child age: children aged 0–5 and children aged 6–17 years.

The fixed effects approach described in equation (2) has a number of potential shortcomings. First, the approach imposes restrictions on the sample such that all-boys and all-girls families are excluded from the estimation of son preference. Second, the fixed effects approach may lead to a double counting of son preference if parents have a fixed amount of parental time that they reallocate from daughters to sons. This could occur if parents reinforce or compensate for differences in endowments related to children’s gender as described by Behrman (1997). Third, in fixed effects models gender is likely to pick up other differences across boys and girls that will induce parents to invest differently. Our models include controls for age, birth order, and previous birth spacing to address some of these differences.¹⁴

Following Barcellos, Carvalho, and Lleras-Muney (2014), we estimate equation (1) on (i) children aged 0–2 years and (ii) first born children aged 0–2 years. The intuition underlying this choice is that sex at conception is randomly assigned. If parents do not perform sex-selective abortion, then boys and girls will be born into families with similar characteristics. Parents then have an opportunity to respond to the sex of the newborn by having more children or concluding their child bearing. However, parents need time after the birth of a child before having additional children and thus families with very young (0–2 years) boys will be similar to those with very young girls, especially first born children.¹⁵

Next, we investigate if presence of a young son affects the time parents allocate on childcare and household chores. Ideally, we would like to study if presence of a son affected the time that mothers allocated on household chores and childcare relative to fathers. However, in our data we only observe the time diaries of a single parent. Therefore, we study if presence of a young son aged 0–2 years affects (i) the average time that mothers spend on childcare and household chores, (ii) the average time that fathers spend on childcare and household chores. We restrict our analysis to families with children aged 0–2 to address concerns that differences in parental time allocation between girls’ families and boys’ families might be due to differences in family characteristics. This analysis is conducted with parent level data and controls for parent and household characteristics, namely: the respondent’s gender (mother or father), education (dummy variables representing less than high school, high school, some college or associate degree, and bachelor’s degree or higher), age (dummy variables representing ages 16–20, 21–25, 26–30, 31–35, 36–40, 41–45, 46–50, 51–55, 56–60, 61–65, and 66+), and age at birth of oldest child (dummy variables representing less than 21, 21–25, 26–30, 31–35, 36–40, 41–45, 46–50, 51–55, 56–60, 61–65, and 66+), spouse’s education and spouse’s age at birth of oldest child, and number of adults.

Results

Figures 1 and 2 present the locally weighted scatterplot smoothing plots of quality time mothers and fathers spent with children by child age. Figure 1 shows a distinct son preference in mother’s quality time in East and South Asian families when the children are young, which erodes over time as children age. Point estimates indicate that East and South Asian mothers spend more time with their younger sons (<2 years) than do other ethnic groups, and less time with younger daughters than other ethnic groups. In contrast, the quality time that mothers from other ethnic groups spend with their children appears to be gender neutral.

Fig. 1.

Locally weighted scatterplot smoothing (LOWESS) plots of quality time (in minutes) spent with mother by age of child.Note: Dotted lines show ± standard error at each year of age. LOWESS bandwidth = 0.6.

Fig. 2.

Locally weighted scatterplot smoothing (LOWESS) plots of quality time (in minutes) spent with father by age of child.Note: Dotted lines show ± standard error at each year of age. LOWESS bandwidth = 0.6.

Figure 2 exhibits son preference among East and South Asian fathers of very young children, which quickly erodes as children age, and daughter preference among European fathers, which too erodes as children age. Latin American and American fathers appear to be gender neutral over most of their children’s childhoods.

Table 2 presents regression results from the analyses outlined in equations (1) and (2). Because the quantity and quality of investments in children vary by child age, we do the analysis separately for children aged 0–5 and children aged 6–17. Model 1 controls for child’s age (a dummy variable for each year of age) and ethnic origin of the respondent. Models 2–4 add controls for birth order (dummy variables for first, second, or higher birth order), previous birth spacing (dummy variables for 1, 2, 3, 4, 5 or more years, or no previous child), and Model 4 adds family fixed effects. In all models, we allow the effects of control variables to differ across ethnic groups by including interactions of each of these variables with the dummy variables for the four ethnic groups. The samples of analyses for Models 1–2 are all children within the age category. The samples of analyses for Models 3–4 are children in families with at least a son and a daughter within the age category, i.e., at least one son and one daughter aged 0–5, or at least one son and one daughter aged 6–17. These are the samples of our family fixed effects analysis (Models 3–4) and we present estimates from OLS (model 3) and family fixed effects (model 4) models. We report the coefficients on the interaction term between male child and ethnic group dummy variables, which measure the average additional time that a mother (panel 1)/father(panel 2) spends with sons compared to daughters. The symbol + indicates that the coefficient measuring son-preference in an ethnic group differs from the son-preference coefficient for East & South Asians at the 10% significance level.

Table 2.

Estimates of Son Preference in Parental Quality Time (in minutes)

Sample	Children Aged 0–5 Years				Children Aged 6–17 Years
	(1)	(2)	(3)	(4)	(1)	(2)	(3)	(4)
Panel 1: Mother’s time
East & South Asian Origin X Male	28.2**	30.0**	46.5*	29.2***	8.4	6.5	−6.4	−3.8
Child	(12.0)	(12.2)	(24.5)	(9.7)	(6.9)	(6.8)	(4.5)	(3.2)
US Native X Male child	−1.0+	−1.2+	1.4+	0.5+	−4.6***+	−4.6***	−2.9**	−2.6***
	(2.5)	(2.4)	(2.3)	(1.8)	(1.2)	(1.2)	(1.1)	(0.7)
Latin American X Male child	−3.7+	−5.0+	1.2+	1.1+	−6.6**+	−6.8**+	−7.1***	−4.8***
	(4.9)	(4.8)	(4.3)	(3.4)	(2.8)	(2.8)	(2.4)	(1.6)
European X Male child	2.3+	2.8+	1.7+	3.0+	2.6	3.8	−0.9	−4.1
	(10.2)	(10.1)	(9.8)	(4.5)	(4.9)	(4.8)	(3.9)	(2.5)
Mean of dependent variable	180.7	180.7	199.1	199.1	89.33	89.33	89.01	89.01
N	14,902	14,902	4,219	4,219	26,840	26,840	11,326	11,326
Panel 2: Father’s time
East & South Asian Origin X Male	−0.7	−2.6	−0.3	−2.9	8.8	8.9	2.8	3.6
Child	(10.4)	(10.3)	(11.1)	(7.9)	(6.4)	(6.3)	(5.1)	(2.9)
US Native X Male child	9.4***	9.5***	6.0***	5.4***	5.2***	5.3***	4.3***	3.5***
	(2.5)	(2.5)	(2.3)	(1.5)	(1.1)	(1.1)	(1.0)	(0.6)
Latin American X Male child	3.9	3.0	−3.8	−7.5*	1.0	0.5	2.0	3.9***
	(4.7)	(4.7)	(4.5)	(4.0)	(2.7)	(2.7)	(2.3)	(1.5)
European X Male child	−16.1	−13.3	−5.0	−0.7	4.5	5.4	7.9**	4.8*
	(10.0)	(10.0)	(7.2)	(4.1)	(4.9)	(4.9)	(3.8)	(2.6)
Mean of dependent variable	121.8	121.8	135.0	135.0	67.22	67.22	67.16	67.16
N	12,898	12,898	3,678	3,678	23,427	23,427	9,940	9,940
Controls:
Region of origin	Yes	Yes	Yes	No	Yes	Yes	Yes	No
Age	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Birth order	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Previous birth interval	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Family Fixed Effects	No	No	No	Yes	No	No	No	Yes

Note: Data on parent’s time with each child in the family are obtained from the time diary of one parent (father or mother) per family. Figures in each column in a Panel are based on a separate OLS regression with minutes of quality time with the child per day as the dependent variable. Samples are restricted to families with at least one son and one daughter aged 0–5 in columns 3–4 and with at least one son and one daughter aged 6–17 in columns 7–8. Robust standard errors clustered on family are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1.

⁺indicates that coefficient differs from coefficient for East & South Asian*Male Child at the 10% significance level.

Estimates in Panel 1 indicate son preference in quality time that East and South Asian mothers spend with their children aged 0–5. Models 1–4 suggest that East and South Asian mothers spend 28 to 47 more minutes of quality time with their young sons than with their young daughters. There is no evidence of a corresponding son preference in the time use by mothers from other ethnic groups. In all models, estimated effect of son preference among East- and South Asian mothers is statistically different from those of other ethnic groups.¹⁶

Estimates in Panel 2 suggest that East and South Asian fathers are gender neutral in their time with young children. We find the same pattern of gender neutrality among parents of other ethnic groups except among fathers in American families, however, who spend 5 to 10 additional minutes of quality time with sons than with daughters.¹⁷

Table 2 also has estimates of models for school-age children (aged 6–17). We find some evidence of gender-specialization in quality time that parents spend with school-age children in families of Latin American, European, and U.S. native origin, but the point estimates are small with mothers spending three to seven more minutes of quality time with daughters than sons and fathers spending four to eight more minutes of quality time with sons than daughters. In East and South Asian families, on the other hand, parental quality time with sons and daughters is statistically the same. The sign and size of the point estimates in fixed effects models are similar to those observed for other ethnic groups. It is possible that the sample size for East and south Asian families is too small to detect small effect sizes.

In Table 3, we report results from a sample of children aged 0–2 years (panel 1) and first-born children aged 0–2 (panel 2). We use regression models similar to models 1–2 (models without family fixed effects) in Table 2 for quality time. We also study four different categories of quality time: physical care, playing, eating/drinking, and other activities using model 2.¹⁸ Estimates show that East and South Asian mothers spend 57 more minutes of quality time per day with their very young sons compared to their daughters. East and South Asian fathers are gender neutral in terms of how they spend time with their very young children. The estimated effects for the other ethnic groups are modest and statistically insignificant. For brevity, we do not present those results here.

Table 3.

Robustness Check of Estimates of Son Preference in Parental Quality Time with Children Aged 0 – 2 Years

	Quality time	Quality time	Time on Physical care	Time Playing	Time Eating and drinking	Time in Other activities
Children Aged 0 – 2 Years
Panel 1: Mother’s time
East & South Asian Origin X Male child	56.7***	57.2***	26.0**	32.4***	4.4	−3.7
	(17.8)	(16.8)	(12.8)	(11.0)	(7.1)	(3.6)
Mean of dependent variable	205.2	205.2	90.75	51.33	53.06	9.298
N	7,235	7,235	7,235	7,235	7,235	7,235
Panel 2: Father’s time
East & South Asian Origin X Male child	8.2	10.8	1.7	4.4	4.6	0.2
	(16.4)	(16.1)	(5.1)	(10.6)	(6.6)	(3.2)
Mean of dependent variable	130.4	130.4	39.98	38.54	46.26	5.581
N	6,247	6,247	6,247	6,247	6,247	6,247
First-Born Children Aged 0 – 2 Years
Panel 3: Mother’s time
East & South Asian Origin X Male child	56.1**	61.0**	31.2	34.2**	0.1	−4.4
	(25.9)	(24.9)	(19.4)	(16.3)	(10.1)	(4.8)
Mean of dependent variable	218.8	218.8	90.34	68.35	52.15	7.898
N	2,581	2,581	2,581	2,581	2,581	2,581
Panel 4: Father’s time
East & South Asian Origin X Male child	27.2	15.1	−0.5	17.4	−1.8	0.0
	(21.5)	(21.8)	(7.8)	(13.3)	(8.8)	(5.8)
Mean of dependent variable	142.1	142.1	43.52	46.34	46.50	5.748
N	2,258	2,258	2,258	2,258	2,258	2,258
Controls:
Region of origin	Yes	Yes	Yes	Yes	Yes	Yes
Age	No	Yes	Yes	Yes	Yes	Yes
Birth order	No	Yes	Yes	Yes	Yes	Yes
Previous birth interval	No	Yes	Yes	Yes	Yes	Yes
Household characteristics	No	Yes	Yes	Yes	Yes	Yes
Family Fixed Effects	No	No	No	No	No	No

Note: Data on parent’s time with each child in the family are obtained from the time diary of one parent (father or mother) per family. Figures in the top two rows of each column in a Panel are based on a separate OLS regression with minutes of time spent on activity in header row as the dependent variable. Other activities are: reading, talking, listening, homework, museums, and religious activities. Household characteristics are: respondent’s education and age, whether mother is unemployed, and number of adults, children, and children aged 0–5. Robust standard errors clustered on family are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1.

We believe that parents are less likely to exercise sex-selective abortion with their first born children. Therefore, in Panels 3 & 4, Table 3 we further restrict the sample to children aged 0–2, who are first born. Because of the sample restriction the estimates are less precise. However, this analysis also suggests that East and South Asian mothers spend an additional 56 to 61 minutes of quality time with their first-born sons than with their first-born daughters.

In the last four columns of Table 3, we examine whether there are differences in parental time investments on specific quality time activities with very young children. East and South Asian mothers spend about 26 more minutes of physical care and 32 more minutes of play-time per day with their very young sons compared to very young daughters. On the other hand, East and South Asian fathers spend statistically the same amount of time on quality activities with very young sons and daughters.

We conduct analysis similar to Table 2 where the dependent variable is one-on-one quality time that parents spent with only one of their children. The estimates observed for one-on-one quality time are presented in Appendix Table 3 and are similar to those for quality time presented in Table 2, but smaller sized and sometimes statistically insignificant, indicating that our results are robust to an alternative specification of parental time investment.

We also investigate if son preference in East and South Asian mother’s time that we observed in Table 2 differed by whether the mother is a first- or second-generation immigrant in the US.¹⁹ The results of this analysis, presented in Table 4, suggest that second generation mothers spend more additional time with sons than daughters than first generation mothers. The difference is statistically insignificant in most models and the difference in point estimates disappears in family fixed effects models. In a separate analysis, conducted on a sample of children of first-generation East and South Asian parents, we find that years since immigration in the US has no effect on son preference among East and South Asian families (Appendix Table 4).

Table 4.

Comparing First and Second Generation East and South Asian Immigrants’ Son-Preference in Parental Quality Time with Children Aged 0 – 5 Years

	Quality time (in minutes)
	(1)	(2)	(3)	(4)
Mother	52.8	42.4	25.1
	(34.2)	(34.8)	(62.8)
1st generation parent	−3.9	−1.9	−79.6
	(23.9)	(23.8)	(49.3)
Mother X 1st generation parent	32.3	38.2	70.9
	(36.6)	(36.2)	(74.4)
Mother X 1st generation parent X Male child	25.3*	29.0**	38.3*	30.6***
	(12.9)	(13.3)	(20.6)	(11.0)
Mother X 2nd generation parent X Male child	78.5**	94.6**	70.0*	33.1
	(34.6)	(36.9)	(38.3)	(28.7)
Father X 1st generation parent X Male child	−3.6	−6.6	−4.2	−4.2
	(11.4)	(11.4)	(12.5)	(10.1)
Father X 2nd generation parent X Male child	26.7	31.3	10.8	−4.2
	(26.0)	(25.0)	(21.7)	(13.2)
Constant	150.6***	75.9	217.7	190.6***
	(23.5)	(57.1)	(138.2)	(27.3)
P-value For Test
Coefficient of Mother X 1st generation parent X Male child = Mother X 2nd generation parent X Male child	0.151	0.094	0.428	0.934
Coefficient of Father X 1st generation parent X Male child = Father X 2nd generation parent X Male child	0.285	0.165	0.540	0.998
Mean of dependent variable	186.3	186.3	204.5	204.5
N	1,030	1,030	237	237
Controls:
Age	Yes	Yes	Yes	Yes
Birth order	No	Yes	Yes	Yes
Previous birth interval	No	Yes	Yes	Yes
Father’s education attainment	No	Yes	Yes	No
Mother’s education attainment	No	Yes	Yes	No
Father’s age at birth of oldest child	No	Yes	Yes	No
Mother’s age at birth of oldest child	No	Yes	Yes	No
Family Fixed Effects	No	No	No	Yes

Note: Data on parent’s time with each child in the family are obtained from the time diary of one parent (father or mother) per family. Figures in each column are based on a separate OLS regression with minutes of quality time with the child per day as the dependent variable. Samples of Models 3 and 4 are restricted to families with at least one son and one daughter aged 0–5. Robust standard errors clustered on family are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1

Next, we investigate if presence of a young son affects the time parents allocate to childcare and household chores (Table 5). We restrict our analysis to families with children aged 0–2 to address concerns that differences in parental time allocation between girls’ families and boys’ families might be due to differences in family characteristics. Each column presents the results of a unique OLS regression using parent-level data. Model 1 provides unadjusted differences in time spent by respondents from different ethnic origins and Model 2 adds controls for respondent’s age, education, and age at birth of oldest child, spouse’s education and age at birth of oldest child, and number of adults in the family. Model 3 adds a control for whether mother is unemployed. The results from the three models are similar.

Table 5.

Gender Differences in Parents’ Time Spent on Child-Care and Household Chores (Families with a child 0–2 years)

	Time on Caring for Children			Time on Household Chores
	(1)	(2)	(3)	(1)	(2)	(3)
Panel 1: Mother’s time
East & South Asian Origin X At least 1 son 0–2yrs	70.2***	74.4***	76.6***	−12.8	−21.9	−20.7
	(19.4)	(19.3)	(18.4)	(17.9)	(17.9)	(17.5)
US Native X At least 1 son 0–2yrs	12.0***+	9.9**+	11.0***+	−4.3	−3.4	−2.5
	(4.1)	(4.1)	(4.0)	(3.9)	(3.8)	(3.8)
Latin American X At least 1 son 0–2yrs	6.6+	6.5+	7.5+	−3.7	−4.3	−3.4
	(8.2)	(8.1)	(7.9)	(9.6)	(9.8)	(9.6)
European X At least 1 son 0–2yrs	34.7**	29.1*+	27.4*+	3.1	7.3	5.8
	(17.7)	(17.0)	(16.6)	(14.3)	(14.9)	(14.7)
Mean of dependent variable (in minutes)	176.7	176.7	176.7	165.9	165.9	165.9
N	6,387	6,387	6,387	6,387	6,387	6,387
Panel 2: Father’s time
East & South Asian Origin X At least 1 son 0–2yrs	6.0	4.8	4.7	9.0	14.7	14.9
	(14.0)	(15.0)	(15.1)	(10.3)	(10.7)	(10.8)
US Native X At least 1 son 0–2yrs	9.1**	9.2**	9.5**	3.5	2.9	3.1
	(3.8)	(3.8)	(3.8)	(4.0)	(4.0)	(4.0)
Latin American X At least 1 son 0–2yrs	3.6	3.9	3.8	−1.8	−1.1	−1.5
	(6.6)	(6.5)	(6.5)	(7.9)	(7.9)	(7.8)
European X At least 1 son 0–2yrs	−2.9	0.4	0.4	−14.6	−16.5+	−17.1+
	(14.0)	(15.0)	(14.9)	(13.5)	(14.2)	(14.0)
Mean of dependent variable (in minutes)	94.77	94.77	94.77	91.96	91.96	91.96
N	5,534	5,534	5,534	5,534	5,534	5,534
Controls:
Parent’s gender	Yes	Yes	Yes	Yes	Yes	Yes
Parent and household characteristics	No	Yes	Yes	No	Yes	Yes
Mother is employed	No	No	Yes	No	No	Yes

Note: Figures in each column of a panel are based on a separate regression with time spent on household chores or time spent on caring for children as the dependent variable. Time is reported in minutes per day. Parent and household characteristics are respondent’s education (dummy variables representing less than high school, high school, some college or associate degree, and bachelor’s degree or higher), age (dummy variables representing ages 16–20, 21–25, 26–30, 31–35, 36–40, 41–45, 46–50, 51–55, 56–60, 61–65, and 66+), age at birth of oldest child (dummy variables representing less than 21, 21–25, 26–30, 31–35, 36–40, 41–45, 46–50, 51–55, 56–60, 61–65, and 66+), and spouse’s education and spouse’s age at birth of oldest child, and number of household adults. Robust standard errors are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1.

⁺indicates that coefficient differs from coefficient for ‘East & South Asian Origin X At least 1 son 0–2yrs’ at the 10% significance level.

In East and South Asian families, mothers with a young son aged 0–2 allocate 70 to 77 additional minutes on childcare per day than mothers with only young daughters; they also spend 13–21 fewer minutes in household chores, but the latter is statistically insignificant.²⁰ We find similar evidence of increased time of caring for children among native US and European mothers with a son <2 years, but the effect size is smaller and statistically different from the effect size for East and South Asian mothers. There is no statistical evidence that East and South Asian fathers’ participation in household chores or childcare is influenced by the gender of their young child. Interestingly, in native U.S. families, fathers allocate between 9 to 10 additional minutes in childcare if they have a young son than a young daughter, but there is no evidence of such gender preference among fathers in other ethnic groups.

Assessments of Bias

Son preference in East and South Asian immigrant families is likely to manifest in several ways: sex selection at higher parity, son-preferring fertility stopping rules, and son preference in parental investments including time allocation. Prevalence of sex selection and son-preferring fertility stopping rules, while consistent with our results of son preference, are likely to bias our estimates. Arguably, parents adopting sex selection and son-preferring rules would be more preferential in their time-allocation for sons compared to parents who do not adopt such rules, which would mean that our estimates contain a bias.

We conduct a number of tests to assess if our estimates are biased on account of sex-selective abortion and fertility stopping rules. In the first test, we examine if the differences in time allocation across sons and daughters that we estimate are the result of son preference or due to differences in characteristics between boy- and girl-families. We run models with the following specification for each ethnic group:

$$Child\_is\_mal{e}_{if}=X_f\beta +{\mu }_{if}$$ (3)

where X is a set of child and family characteristics, namely age of the child, birth order, number of the child’s brothers aged 0–17, number of the child’s sisters aged 0–17, sex, age and educational attainment of the respondent parent. We ran the above specification for children aged less than 18, children less than 6 and children less than 2. Results presented in Appendix Table 5 show that for the East and South Asian sample, the p-value for a test that all co-variates are jointly zero is 0.30 for all children; 0.41 for children <6, and 0.47 for children <2 (bottom row). Thus, in all three cases we reject the null hypotheses that the co-variates are jointly different from zero.²¹ Our analysis suggests that observed family characteristics fail in determining the sex of the child, providing some evidence the observed family characteristics do not explain son preference in parental time investments. Appendix Table 5 also presents results (p-values) of the hypotheses that the female and male child samples have the same child and family characteristics. In all cases, we fail to reject the null that these characteristics are statistically the same for the two sets of samples.

Bias on account of Fertility-stopping Rules

In the second test, presented in Table 6, we investigate if East and South Asian families are more likely to adopt son-biased fertility stopping rules using the Annual Social and Economic Supplement of the Current Population Survey (CPS) data for 2003–2012. We use the CPS data because ATUS respondents are randomly selected from CPS sample households and because of its large sample size. For this analysis, we study three outcomes: (i) a dichotomous variable indicating that the family has a second child, (ii) the number of children, and (iii) the gap between the first and second child (in years). We regress each outcome on parent’s ethnicity, a variable indicating that the first child is a boy and an interaction between these two variables (model 1). Additionally, we estimate each of these models by including a rich set of demographic variables (model 2). A limitation of the CPS is that it does not provide data on children who are not living with their parents. One way to address this limitation is to study families with relatively young parents. We did all analyses on a sample of families where mother’s age <40 and oldest child is <16. Our findings are: one, if the first child is male, the probability of having a second child is four percentage points lower in East and South Asian families (or 4% of all East and South Asian families) and one percentage point higher in families of US origin. Two, if the first child is male, East and South Asian families have 0.06 fewer children and families of US origin are have 0.03 more children. And three, having a first male child has no statistically significant effect on the age gap between first and second child in East and South Asian families, but it lowers the gap in US origin families by 0.05 years. For all three outcomes, we do not find any statistically significant effects for European and Latin American families.

Table 6.

Test for Son-Biased Fertility Stopping Using Current Population Survey (CPS - Annual Social and Economic Supplement) Data for 2003–2012

Dependent variable	Has Second Child = 1		Number of Children		Age Gap Between 1st and 2nd Child
	(1)	(2)	(1)	(2)	(1)	(2)
Omitted category: European	−	−	−	−	−	−
US Native	0.03***+	0.05***+	0.08***+	0.09***+	0.03+	−0.02+
	(0.01)	(0.01)	(0.02)	(0.02)	(0.06)	(0.06)
East/South Asian	−0.06***	−0.05***	−0.19***	−0.17***	0.15*	0.31***
	(0.02)	(0.02)	(0.03)	(0.03)	(0.09)	(0.09)
Latin American	0.07***+	0.05***+	0.20***+	0.05**+	0.33***+	0.29***+
	(0.01)	(0.01)	(0.02)	(0.02)	(0.07)	(0.07)
Oldest is male X US Native	0.01***+	0.01***+	0.03***+	0.03***+	−0.05***+	−0.05***+
	(0.00)	(0.00)	(0.01)	(0.01)	(0.02)	(0.02)
Oldest is male X East/South Asian	−0.03**	−0.04**	−0.06**	−0.06**	0.14	0.14
	(0.02)	(0.02)	(0.03)	(0.02)	(0.10)	(0.10)
Oldest is male X Latin American	0.00+	0.00+	0.02+	0.02+	0.04	0.03
	(0.01)	(0.01)	(0.01)	(0.01)	(0.04)	(0.04)
Oldest is male X European	−0.01	−0.00+	−0.01	−0.00	0.02	0.02
	(0.01)	(0.01)	(0.03)	(0.03)	(0.09)	(0.08)
Controls:
Number of household adults	No	Yes	No	Yes	No	Yes
Mother’s age	No	Yes	No	Yes	No	Yes
Father’s age	No	Yes	No	Yes	No	Yes
Mother’s education attainment	No	Yes	No	Yes	No	Yes
Father’s education attainment	No	Yes	No	Yes	No	Yes
Mean	0.670	0.670	2.002	2.002	3.341	3.341
Observations	113,276	113,276	113,276	113,276	73,576	73,576

Note: Sample restricted to families where the mother is <40 years old and the oldest resident child is <16 years old. Figures in each column are based on a separate regression with variable in column header as the dependent variable. Robust standard errors are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1.

⁺indicates that coefficient differs from corresponding coefficient for East & South Asian at the 10% significance level.

The second sensitivity test suggests that we cannot rule out the possibility of fertility stopping rules in East and South Asian families. The issue is whether the decision to adopt fertility-stopping rules by these four percent of parents can cause a large bias in our estimates and what would be the direction of the bias. We find that in East and South Asian families if the first child is male, the probability of having a second child is four percentage points lower (or 4% of all East and South Asian families adopt son-preferring fertility stopping). We also find that having a male first child reduced the number of children by 0.06 in East and South Asian families. This implies that the four percent of families that adopt fertility-stopping rules have 1.5 fewer children (=0.06/0.04). This will clearly put sons in families adopting fertility-stopping rules in an advantage on account of family size. The advantage would be greater in one-on-one time that parents spend with their children than in parental quality time where all children are present. We find that on average East and South Asian mothers spend 100 minutes of one-on-one time with a young child. Assuming that mothers who exercised fertility-stopping rules spend the entire 150 minutes of time they “saved” on account of a smaller family on their sons yields an upward bias of 6 minutes (=150*0.04) or 25 percent of our OLS estimate of the gender difference in one-on-one time (25%=6/24). Note that this is an upper-bound of the size of bias as mothers will most likely divide their time across multiple activities, and one-on-one time with sons would be one such activity. The level of bias would be more modest on total quality time with children. Also note that these estimates are based on the CPS data. In the ATUS, the data used in our analyses, we find no difference in family size by the gender of the child (Appendix Table 5).

Family fixed effects models ensure that there will be no confounding on account of family size. However, our estimates are based on a selective sample. The fixed effects sample, for instance, excludes a small proportion (four percent) of parents who exercise son-preferring fertility-stopping rules. Arguably, these parents are more likely to invest in sons preferentially. Exclusion of these parents will therefore result in a small downward bias in our estimates of son-preference in parental time investments. Our fixed effects sample also excludes parents with only one child or same gender children within a specific age category. If parents with children of the same gender have lower son preference in parental time investments, exclusion of these parents would result in an upward bias in our estimates. On the other hand, if they have greater son bias or daughter discrimination, our fixed effects estimates will be downward biased. Given these various sources of biases, we exercise caution in interpreting these results and argue that our fixed effects models estimate son preference within families.

Arguably, if these parents had not adopted son-targeting fertility behavior they would have exercised at least as much son preference in time-investments on their additional children as parents who did not adopt fertility stopping rules. Thus, adoption of fertility stopping rules by these four percent of one-child families with a son will cause a downward bias in our estimates. To estimate the size of the bias, we compute the upper bound of the difference between sons and daughters in parental time under the assumption that these four percent of parents would have been among the bottom 20^th percentile in terms of their time spent with daughters of the corresponding age. In the OLS model (model 2 in Table 2), in families with children less than age 5, our estimate of son preference in mothers’ time investment has a downward bias of one minute or 3%; in the fixed effects model (model 4 in Table 2), the estimated downward bias on account of fertility stopping rules is 4.7 minutes or 16 percent.

In the final specification, where we restricted the sample of analysis to families with only one child under 2, our sample is unaffected by fertility stopping rules. But parents who expect to have more children might invest in their children differently from parents who do not expect more children. If there is son preference, more parents who have a firstborn daughter will expect (or desire) a larger family than parents who have a firstborn son and if the expectation of more children results in parents altering their investments in existing children, our estimated coefficient will reflect the combined effect of son preference from the desire to have another child as well as higher time allocation on sons than daughters.

Our analysis shows that an additional child increased the time East and South Asian mothers spend on their children aged 0–2 years by a statistically insignificant 12.4 minutes and lowered the time East and South Asian fathers spend in children aged 0–2 years by a statistically insignificant 7.3 minutes. We also find that if the first child is male, East and South Asian families have 0.06 fewer children (Table 6). To estimate the effect of child gender on parental time investment that is on account of anticipated family size, we multiply the two coefficients (effect of family size on investments and effect of male firstborn on family size). This analysis shows that our estimates have a downward bias of 0.74 minutes (=12.4*0.06) or 1.3% (=0.74/56.1) in mothers’ quality time with young children (0–2 years old) and an upward bias of 0.44 minutes (=7.3*0.06) or 1.6% (=0.44/27) in fathers’ quality time with young children.

Bias on account of Sex Selection

In the third test, we estimate the sex ratios of children in families across ethnic groups, overall and by birth order using the Annual Social and Economic Supplement of the Current Population Survey (CPS) data for 2003–2012. Here too we estimate sex ratios for families where mother’s age is less than 40 and oldest child is less than 16. Our data do not suggest any statistically significant difference in sex ratio in South and East Asian families.²²

Arguably, the CPS is not the appropriate dataset to measure sex selection. Previous studies based on U.S. Natality data for 1991–2014 have estimated the sex ratios at birth to be between 1.08 and 1.19 at third or higher parity among immigrants from East and South Asian countries (Abrevaya 2009; Almond and Sun, 2017). Assuming the sex ratio at birth of U.S. White children, which is 1.058, to be normal, these studies suggest that between 2% to 13% of East and South Asian families at third or fourth parity engage in sex selection (Abrevaya 2009). In our data 10.7% of the families have three or more children, which implies that our estimates may be biased on account of the absence of 0.2% (=0.02*0.107) to 1% (=0.13*0.107) of East and South Asian families that may have exercised sex selection. These are crude calculations but they indicate that bias on account of sex selection is likely to be much less in our analyses than bias on account fertility stopping rules, even though the direction of the bias would be similar. Because sex selection is documented mostly at higher parity, bias on account of sex selection is less likely in the empirical specifications where the sample is restricted to families with first-born children aged 0–2 years or families with children aged 0–2 only.

Conclusion and Discussion

In this paper, we investigate if son preference or discrimination against daughters persists in families of East and South Asian origin that have migrated to the U.S. by studying the quantity and quality of parental time investment in children, a critical, yet least studied, developmental input that can impact abilities and outcomes in later life. Our analysis has five main findings. First, East and South Asian mothers spend 30 more minutes of quality time with their young (aged 0–5 years) sons than with their young daughters. There is no corresponding evidence of gender discrimination in time that mothers of other ethnic groups spend with their young children. We find that East and South Asian fathers are gender neutral in their allocation of total and quality time with young children. We are, however, unable to rule out prevalence of son-favoring fertility stopping rules in a small number, about four percent, of families in our data. Arguably, parents adopting sex-selection and son-preferring rules would be more preferential in their time-allocation for sons compared to parents who do not adopt such rules, which would result in a small downward bias in our estimates.

Second, in analysis restricted to families with a first-born child aged 0–2, we find that East and South Asian mothers spend about 56–61 minutes of additional quality time with sons aged 0–2 than with similarly aged daughters.

Third, our analysis suggests some evidence of gender specialization in quality time that parents spend with their children, but the point estimates are small for all groups and statistically insignificant for East and South Asian families.

Fourth, activity specific analyses suggest that East and South Asian mothers with children aged 0–2 spend 31 additional minutes on the physical care and 34 minutes in playing with their very young sons (aged 0–2) than with similarly aged daughters. Our result differs from previous research (Barcellos et al., 2014) that found mothers in India spending half hour more in breast feeding their very young sons than daughters. We find mother’s time allocation towards feeding their young children (all kinds of feeding including breast feeding) to be gender neutral.

In the final analysis, we investigate if presence of a son affects the time parents allocate in childcare and household chores in families with children aged 0–2 and find that mothers in East and South Asian families spend 70–76 additional minutes in childcare and 13–20 fewer minutes (statistically insignificant) in household chores if they have a young son than a young daughter.

One possible explanation could be that due to cultural factors East and South Asian mothers feel “blessed” with sons and think that sons need more of their time, but as their children get older they adapt more gender neutral parenting following other ethnic groups in the US. Alternatively, it may be easier for parents to exercise preferential treatment when children are very young. As children age, they are more likely to have siblings, which makes preferential time investment more challenging since parents would need to sequester one or more of their children. Besides, as they get older, sons may not want to spend more time with their parents.

Our findings are consistent with differential parental investments favoring young sons because of cultural factors, such as social pressure, or expectations about future payoffs (Puri, Adams, Ivey, & Nachtigall, 2011; Das Gupta et al. 2003). An alternative explanation is that East and South Asian mothers compensate for differences between boys’ and girls’ emotional and physical needs whereas mothers from other ethnicities do not. It is argued that East- and South Asian boys in particular “need” more attention at home because they “look different” from majority of boys and therefore are more likely to be called out for disruptive behavior. If so, we would expect more parental time with school-age boys. Our primary finding that son preference in time allocation in East and South Asian families is most prevalent for children less than 2, and disappears with child’s age, suggests that it is not in response to external environment – e.g., East and South Asian boys being bullied or called out in school.

Our research highlights that son preference in parental investments in East and South Asian families persists after immigration to the U.S. We find no evidence that the gender bias in mothers’ investments in young children declines across generations. These findings thus suggest that norms of son preference in East and South Asian cultures continue to affect parenting behavior even after migration. Our findings also imply that East and South-Asian countries should not treat daughter discrimination as an outcome of low economic development and poverty, but as a gender issue that needs to be addressed with increased awareness about gender bias and the harm that it would do to children of both sexes. Son preference and daughter discrimination is likely to influence family well-being adversely and result in sub-optimal outcomes for East and South Asian families and the society as a whole. Increasing awareness about gender based discrimination through education or outreach within communities can go a long way in affecting family well-being.

Appendix Table 1.

South and East Asian Children by Respondent’s (i.e., Parent’s) Country of Origin

Country	n	%
Bangladesh	63	2.68
China	507	21.54
India	926	39.34
Japan	168	7.14
Korea	269	11.43
Nepal	9	0.38
Pakistan	110	4.67
Sri Lanka	10	0.42
Taiwan	139	5.9
Multiple South & East Asia Countries	153	6.5
N	2,354	100

Appendix Table 2.

Estimates of Son Preference in Parental Quality Time (in minutes)

	Children Aged 0–5 Years				Children Aged 6–17 Years
	(1)	(2)	(3)	(4)	(1)	(2)	(3)	(4)
Panel 1: Mother’s time
East & South Asian Origin X Male	28.2**	30.0**	46.5*	29.2***	8.4	6.5	−6.4	−3.8
	(12.0)	(12.2)	(24.5)	(9.7)	(6.9)	(6.8)	(4.5)	(3.2)
US Native X Male child	−1.0+	−1.2+	1.4+	0.5+	−4.6***+	−4.6***	−2.9**	−2.6***
	(2.5)	(2.4)	(2.3)	(1.8)	(1.2)	(1.2)	(1.1)	(0.7)
Latin American X Male child	−3.7+	−5.0+	1.2+	1.1+	−6.6**+	−6.8**+	−7.1***	−4.8***
	(4.9)	(4.8)	(4.3)	(3.4)	(2.8)	(2.8)	(2.4)	(1.6)
European X Male child	2.3+	2.8+	1.7+	3.0+	2.6	3.8	−0.9	−4.1
	(10.2)	(10.1)	(9.8)	(4.5)	(4.9)	(4.8)	(3.9)	(2.5)
West Asian/Middle East X Male child	−43.6*	−51.1*	89.4**	−24.6**	−5.9	−2.8	20.2	−7.9
	(24.9)	(27.3)	(39.4)	(10.6)	(17.3)	(16.9)	(21.5)	(7.8)
Mean of dependent variable	180.6	180.6	199.1	199.1	89.57	89.57	89.26	89.26
N	14,980	14,980	4,235	4,235	26,994	26,994	11,395	11,395
Panel 2: Father’s time
East & South Asian Origin X Male	−0.7	−2.6	−0.3	−2.9	8.8	8.9	2.8	3.6
	(10.4)	(10.3)	(11.1)	(7.9)	(6.4)	(6.3)	(5.1)	(2.9)
US Native X Male child	9.4***	9.5***	6.0***	5.4***	5.2***	5.3***	4.3***	3.5***
	(2.5)	(2.5)	(2.3)	(1.6)	(1.1)	(1.1)	(1.0)	(0.6)
Latin American X Male child	3.9	3.0	−3.8	−7.5*	1.0	0.5	2.0	3.9***
	(4.7)	(4.7)	(4.5)	(4.0)	(2.7)	(2.7)	(2.3)	(1.5)
European X Male child	−16.1	−13.3	−5.0	−0.7	4.5	5.4	7.9**	4.8*
	(10.0)	(10.0)	(7.2)	(4.1)	(4.9)	(4.9)	(3.8)	(2.6)
West Asian/Middle East X Male child	23.8	11.7	−20.4	−17.1**	3.7	8.0	29.3	5.1
	(34.2)	(36.4)	(24.0)	(7.4)	(14.5)	(17.2)	(20.6)	(4.2)
Mean of dependent variable	121.8	121.8	135	135	67.29	67.29	67.27	67.27
N	12,987	12,987	3,699	3,699	23,593	23,593	10,020	10,020
Controls:
Region of origin	Yes	Yes	Yes	No	Yes	Yes	Yes	No
Age	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Birth order	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Previous birth interval	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Family Fixed Effects	No	No	No	Yes	No	No	No	Yes

Note: Data on parent’s time with each child in the family are obtained from the time diary of one parent (father or mother) per family. Figures in each column in a Panel are based on a separate OLS regression with minutes of quality time with the child per day as the dependent variable. Samples are restricted to families with at least one son and one daughter aged 0–5 in columns 3–4 and with at least one son and one daughter aged 6–17 in columns 7–8. Robust standard errors clustered on family are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1.

⁺indicates that coefficient differs from coefficient for East & South Asian*Male Child at the 10% significance level.

Appendix Table 3.

Estimates of Son Preference in One-on-one Quality Time with Children

	Children aged 0 – 5				Children aged 6 – 17
	(1)	(2)	(3)	(4)	(1)	(2)	(3)	(4)
Panel 1: Mother’s time
East & South Asian Origin X Male	15.8	23.8**	16.2*	25.6**	5.1	4.0	−2.6	−3.6
	(11.3)	(10.2)	(9.4)	(9.9)	(4.5)	(4.5)	(3.2)	(3.2)
US Native X Male child	2.3	1.8+	0.7	0.7+	−2.2***	−2.3***	−1.5***	−1.6***
	(1.9)	(1.7)	(1.6)	(1.7)	(0.6)	(0.6)	(0.5)	(0.5)
Latin American X Male child	1.8	−0.8+	2.4	2.8+	−0.3	−0.4	−0.9	−1.2
	(3.4)	(3.1)	(3.0)	(3.2)	(1.2)	(1.2)	(1.1)	(1.0)
European X Male child	4.5	1.9+	5.9	5.4+	−1.9	−1.5	−1.6	−2.0
	(7.5)	(6.7)	(5.1)	(4.7)	(2.6)	(2.6)	(2.2)	(1.9)
Mean of dependent variable	58.69	58.69	21.03	21.03	16.69	16.69	8.046	8.046
N	14,902	14,902	4,219	4,219	26,840	26,840	11,326	11,326
Panel 2: Father’s time
East & South Asian Origin X Male	10.6	9.4	1.7	0.2	6.8*	7.0*	2.2	1.7
	(7.9)	(7.2)	(7.6)	(7.7)	(3.7)	(3.6)	(2.3)	(2.6)
US Native X Male child	3.6**	4.0***	3.5***	3.5***	2.4***	2.4***	1.8***	2.0***
	(1.6)	(1.5)	(1.3)	(1.3)	(0.5)	(0.5)	(0.4)	(0.5)
Latin American X Male child	2.8	2.6	−5.7	−6.8*	2.4**	2.4**	2.8***	2.7***
	(2.9)	(2.7)	(3.6)	(3.6)	(1.0)	(1.0)	(1.0)	(1.0)
European X Male child	1.9	5.6	−2.5	−2.5	−0.5	0.1	3.4*	3.5*
	(6.6)	(6.2)	(3.6)	(3.9)	(2.7)	(2.6)	(2.0)	(2.0)
Mean of dependent variable	34.36	34.36	9.058	9.058	11.25	11.25	4.543	4.543
N	12,898	12,898	3,678	3,678	23,427	23,427	9,940	9,940
Controls:
Region of origin	Yes	Yes	Yes	No	Yes	Yes	Yes	No
Age	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Birth order	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Previous birth interval	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Family Fixed Effects	No	No	No	Yes	No	No	No	Yes

Note: Data on parent’s time with each child in the family are obtained from the time diary of one parent (father or mother) per family. Figures in each column in a Panel are based on a separate OLS regression with minutes of quality time per day as the dependent variable. Samples are restricted to families with at least one son and one daughter aged 0–5 in columns 3–4 and with at least one son and one daughter aged 6–17 in columns 7–8. Robust standard errors clustered on family are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1.

⁺indicates that coefficient differs from coefficient for East & South Asian*Male child at the 10% significance level.

Appendix Table 4.

Estimates of the Association between Parental Quality Time with Children Aged 0 – 5 Years and Years in the US, among East and South Asian First Generation Immigrant Families

	Quality time (in minutes)
	(1)	(2)	(3)	(4)
Mother	71.6***	56.6**	50.6
	(22.3)	(26.3)	(81.0)
Male child X Father	−3.7	−9.0	3.3	13.7
	(18.2)	(19.2)	(27.3)	(20.2)
Male child X Mother	18.7	29.0	9.9	23.5
	(21.4)	(21.6)	(25.8)	(21.1)
Years in the US	−0.1	0.2	−1.2
	(0.7)	(0.9)	(3.6)
Years in the US X Mother	1.0	1.6	2.2
	(1.3)	(1.5)	(3.7)
Years in the US X Male Child	−0.0	0.1	−0.7	−1.3
	(1.0)	(1.0)	(2.0)	(1.6)
Years in the US X Mother X Male Child	0.6	−0.1	3.2	1.9
	(1.8)	(1.8)	(2.6)	(2.0)
Constant	145.1***	63.4	207.2	155.0***
	(17.7)	(60.1)	(133.3)	(30.8)
Mean dependent variable	185.0	185.0	197.2	197.2
N	891	891	200	200
Controls:
Age	Yes	Yes	Yes	Yes
Birth order	No	Yes	Yes	Yes
Previous birth interval	No	Yes	Yes	Yes
Father’s education attainment	No	Yes	Yes	No
Mother’s education attainment	No	Yes	Yes	No
Father’s age at birth of oldest child	No	Yes	Yes	No
Mother’s age at birth of oldest child	No	Yes	Yes	No
Family Fixed Effects	No	No	No	Yes

Note: Data on parent’s time with each child in the family are obtained from the time diary of one parent (father or mother) per family. Figures in each column are based on a separate OLS regression with minutes of quality time with the child per day as the dependent variable. Robust standard errors clustered on family are in parentheses.

^***p<0.01

^**p<0.05

^*p<0.1

Appendix Table 5.

Mean characteristics of male and female East and South Asian children by age of sample

	Age 0–2 years			Age 0–5 years			Age 0–18 years
	Females	Males	P-value female=males	Females	Males	P-value females=males	Females	Males	P-value females=males
Child characteristics
Age (years)	1.07	1.03	0.544	2.48	2.42	0.566	7.08	7.10	0.897
Birth order (reference: first)
Second	0.36	0.39	0.513	0.38	0.41	0.300	0.36	0.35	0.820
Third or higher	0.14	0.12	0.405	0.12	0.11	0.435	0.092	0.089	0.861
Responding parent characteristics
Female	0.53	0.57	0.304	0.53	0.53	0.958	0.52	0.53	0.936
Age (years)	34.8	34.3	0.336	35.9	35.7	0.638	39.3	39.5	0.453
Educational attainment (reference: less than high school)
High school	0.073	0.081	0.743	0.080	0.091	0.509	0.10	0.12	0.185
Some college or associate degree	0.11	0.088	0.334	0.11	0.080	0.147	0.11	0.10	0.700
Bachelor’s degree or higher	0.81	0.82	0.669	0.80	0.82	0.643	0.77	0.76	0.746
Family characteristics
Number of sisters aged 0–17 years	0.39	0.35	0.526	0.46	0.42	0.334	0.49	0.51	0.556
Number of brothers aged 0–17 years	0.38	0.43	0.350	0.46	0.48	0.616	0.54	0.52	0.312
R2 regressing child sex on all characteristics			0.036			0.024			0.017
P-value characteristics are jointly zero			0.47			0.41			0.30

Note: P-value comparing females and males obtained from separate OLS regressions where each characteristic is regressed on child gender. P-values in the bottom row are obtained from regressing child’s gender on all the characteristics listed in the table and performing a Wald test. In this regression, child and parent’s age are entered as categorical variables.

^***p<0.01

^**p<0.05

^*p<0.1.