https://orcid.org/0000-0002-6343-4071
Abstract
Recent cohorts of U.S. children increasingly consist of immigrants or the immediate descendants of immigrants, a demographic shift that has been implicated in high rates of child poverty. Analyzing data from the 2014–2018 Current Population Survey and using the U.S. Census Bureau's Supplemental Poverty Measure, we describe differences in child poverty rates across immigrant generations and assess how these disparities are rooted in generational differences in the prevalence and impact of key poverty risk factors. Our estimates show that poverty rates among Hispanic children are very high, particularly among first-generation children and second-generation children with two foreign-born parents. Low family employment is the most significant risk factor for poverty, but the prevalence of this risk varies little across immigrant generations. Differences in parental education account for the greatest share of observed intergenerational disparities in child poverty. Supplemental comparisons with third+-generation non-Hispanic White children underscore the disadvantages faced by all Hispanic children, highlighting the continued salience of race and ethnicity within the U.S. stratification system. Understanding the role of immigrant generation vis-à-vis other dimensions of inequality has significant policy implications given that America's population continues to grow more diverse along multiple social axes.
Introduction and Background
Racial and ethnic diversity is steadily increasing in the United States, driven largely by the changing composition of the youngest age cohorts (Lichter 2013). As of 2018, the race and ethnicity of more than half of all school-age children was something other than non-Hispanic White, and Hispanics alone made up one quarter of these youngest age-groups (Frey 2019; U.S. Census Bureau 2019b).1 Importantly, these changes in ethnoracial composition reflect shifting patterns of immigration and fertility among immigrants (Johnson and Lichter 2008). An implication of such shifts is that recent cohorts of children will reflect increasing diversity in terms of both race and immigrant generation, each of which operates as an important axis of socioeconomic inequality (Lichter 2013; Rumbaut and Portes 2001). These changes raise the prospect that the youngest cohorts of U.S. children will be characterized by large and complex differences in well-being and opportunity.
Racial and ethnic disparities in poverty and other socioeconomic outcomes among children have been well documented (Eggebeen and Lichter 1991; Lichter et al. 2015; Timberlake 2007), but considerably less attention has been placed on variation in children's economic status across immigrant generations (National Academies of Sciences, Engineering, and Medicine 2016).2 Much of the evidence on this issue is based on data from the early 2000s and from studies using a poverty measure that may systematically mischaracterize economic conditions across immigrant generations. An updated analysis of intergenerational disparities in child poverty simply described rather than explained these differences (Thiede and Brooks 2018). Further study of these patterns can provide insight into processes of economic integration or exclusion across immigrant generations in at least two respects. First, children's early-life circumstances shape their likelihood of experiencing upward mobility over their life course (Duncan et al. 1998; Duncan et al. 2010). Early exposure to poverty “at the starting line” can leave children permanently disadvantaged (Lichter et al. 2015). Second, cross-sectional patterns of intergenerational inequalities provide an analog to the trajectory that new immigrants and their descendants may experience as they form emergent first- and second-generation cohorts (Alba and Nee 2009).
With these motivations in mind, we document the levels and correlates of child poverty differences across immigrant generations. We focus on the Hispanic population because it is the modal ethnoracial group of first- and second-generation children. Drawing on the logic of regression decomposition and the related prevalences and penalties framework for comparative poverty research proposed by Brady et al. (2017), we begin by answering three empirical questions about Hispanic children. First, how does the prevalence of poverty risk factors vary across immigrant generations? Second, how does the penalty (i.e., the magnitude of poverty risk) associated with each risk factor vary across immigrant generations? Third, given the patterns we document in the first two steps, what share of observed poverty rates, and intergenerational differences therein, can be explained (statistically) by the uneven distribution of risk factors? We then expand our focus beyond Hispanics and ask how the prevalence and penalties of poverty risk factors differ between Hispanics and the third+ generation of non-Hispanic Whites. This final, comparative analysis sheds preliminary light on the intersection of race and immigrant generation.
Immigration and Inequalities in Child Poverty
Growing ethnoracial diversity among recent cohorts of U.S. children shapes and contextualizes policy debates about child poverty and the social safety net. These demographic trends—driven by immigration, marriage and fertility patterns, and related social changes—are broadly correlated with large and growing shares of children born into disadvantaged circumstances (Lichter et al. 2015). Over the 2008–2014 period, for example, 43.2% of non-Hispanic Black infants were born into poor families (as defined by the U.S. official poverty measure), as were 36.5% of Hispanic infants and 17.7% of non-Hispanic White newborns (Thiede et al. 2018). These differences partially represent the most recent manifestation of long-standing ethnoracial inequalities in child poverty (Call and Voss 2016; Eggebeen and Lichter 1991; Lichter et al. 2005). However, new forms of diversity—particularly along the lines of nativity—are being driven by immigration and high fertility among first- and second-generation immigrants relative to other groups (Woods and Hanson 2016). Social changes have heightened discrimination and conditionality by nativity and immigrant generation simultaneously, making them increasingly salient dimensions of socioeconomic inequality (Dohan 2003; Heinrich 2018; Laird et al. 2019). Understanding current child poverty dynamics and developing effective social policy therefore require attention to how the levels and correlates of poverty vary across immigrant generations.
However, relevant evidence available to policymakers is limited. Social scientists have documented substantial variation in socioeconomic outcomes across immigrant generations (Jiménez et al. 2018; National Academies of Sciences, Engineering, and Medicine 2016; Park and Myers 2010; Parrado and Morgan 2008), but few studies have analyzed such differences in poverty among children (we discuss exceptions later in this section).3 This knowledge gap is particularly important because childhood exposure to poverty has large and sometimes irreversible effects on developmental outcomes that shape attainment over the life course (Bradley and Corwyn 2002; Duncan et al. 1998).4 Over the long run, an unequal distribution of child poverty across immigrant generations may therefore contribute to intergenerational disparities in education, health, and other commonly studied later-life outcomes (Abraído-Lanza et al. 2016; Qian and Qian 2019; Tran and Valdez 2017).
The few prior analyses of child poverty differences across immigrant generations provide clear motivation for additional research. For example, Oropesa and Landale (1997) showed that more than four in 10 (40.7%) first-generation Hispanic children were poor in 1990, more than 10 percentage points higher than their third+-generation peers and more than 30 percentage points higher than the non-Hispanic White third+ generation. Likewise, Lichter et al.'s (2005) analysis of the 2000 census showed that poverty among first-generation Mexican-origin children (36.1%) was nearly 14 percentage points higher than among third+-generation Mexican Americans (22.8%).5 Intergenerational disparities are also dynamic. Comparing all children of immigrants with children of native-born adults, Van Hook et al. (2004) documented growing inequalities in poverty from 1969 through 1999. Overall, they found that the gap in child poverty between children of immigrants and children of natives increased from −2.5 percentage points (11.6% vs. 14.1%) in 1969 to 6.9 percentage points (21.6% vs. 14.7%), representing a swing of 9.4 percentage points. This rapid increase in poverty among children of immigrants partly reflected compositional changes—namely, shifts in race and ethnicity, parental employment and marital status, and time in the United States—and disproportionate increases in education and work hours among native-born parents. Large intergenerational disparities in child poverty have also been observed in more recent data. Thiede and Brooks (2018) found that, overall, first-generation noncitizen children and second-generation children with two foreign-born parents experienced much higher rates of poverty (as officially defined by the U.S. government) in 2015–2016 (30.2% and 25.7%, respectively) than second-generation children with one foreign-born parent and third+-generation children (approximately 17% each). Beyond these studies, surprisingly little analysis has explored differences in child poverty across immigrant generations (see also Borjas 2011).
Knowledge on this topic is also limited by the use of the U.S. government's official poverty measure (OPM). The OPM has been widely critiqued (Brady and Parolin 2020), and many of its limitations are amplified in comparisons of child poverty across immigrant generations. For example, the OPM does not account for near-cash transfers, such as the Supplemental Nutrition Assistance Program (SNAP), which have important implications for children's welfare and to which access varies according to immigrant generation (and citizenship).6 A recent comparison of OPM-based estimates of child poverty by immigrant generation with rates based on historical estimates of the U.S. Census Bureau's Supplemental Poverty Measure (SPM) (Fox et al. 2015) revealed differences that are consistent with expectations about disparities in program eligibility and use (Thiede and Brooks 2018).7 Thus, prior evidence based on the OPM may mischaracterize levels and differences in poverty across immigrant generations. In the context of projected future growth in the demographic diversity of youth cohorts, these evidence gaps motivate our use of the SPM and focus on poverty among immigrant children.
Conceptual Framework
We draw on Brady et al.'s (2017) prevalences and penalties framework for comparing the poverty risk across populations to conceptualize and analyze intergenerational economic disparities among Hispanic children.8 Drawing on the logic of demographic standardization and decomposition, this framework disaggregates disparities in child poverty rates into (a) differences in the prevalence of characteristics associated with poverty (i.e., a “prevalence”) and (b) differences in the degree of risk associated with a given characteristic (i.e., a “penalty”). Importantly, this framework emphasizes risk factors that are modifiable—and thus amenable to policy interventions—rather than purely ascriptive (Brady et al. 2017:742). We focus on five such dimensions of risk: parental age, family structure, parental education, family employment, and place of residence.9
Parental age is expected to be inversely associated with the risk of poverty. Poverty rates are particularly high for young parents, especially teenagers and those in their early 20s (Kearney and Levine 2012). These are the ages at which earnings tend to be lowest and thus when parents are least able to meet the additional income needs of a new child. Prior research suggests that age-at-childbearing patterns vary by immigrant generation. For example, Rumbaut and Komaie (2010) found that first-generation immigrants are much more likely to have children in their early adult years (56%), compared with both the second (31%) and third generations (38%).
Family structure is correlated with child poverty risk, in part because single headship constrains family labor supply. Family structure may also capture poverty risks associated with gender discrimination, such as the compounded disadvantage that single mothers face vis-à-vis single fathers (Kramer et al. 2016). Marriage and cohabitation patterns as well as family structure at birth have been shown to vary across immigrant generations: higher marriage rates are typically observed among the earliest generations (Brown et al. 2008; Glick 2010; Lichter et al. 2005). For example, whereas first-generation Hispanic immigrants have high marriage rates, those in the second and third+ generations have experienced the same retreat from marriage as the U.S. population at large (Oropesa and Landale 2004).
Parental education may affect poverty risk through employment, job quality, and wages. Given increased matching on education among partners (Smith et al. 2014), education may also influence poverty by affecting the probability that one's spouse or cohabitating partner is employed and earning above-poverty wages. Prior studies demonstrated significant intergenerational disparities within the Hispanic population (e.g., Rumbaut and Komaie 2010). Likewise, our analysis of the 2014–2018 Current Population Survey (CPS) reveals that 12.3% of all first-generation Hispanic adults completed a college degree, compared with 20% or greater of Hispanic adults from later generations.
Employment (henceforth “family employment” in our typology) is the main determinant of family income and is therefore an important correlate of child poverty (Baker 2015). Although even full-time employment is sometimes insufficient to avoid poverty (Brady et al. 2010), families with unemployed adults face exceptionally high poverty risks. Differences in employment between native- and foreign-born adults have been well documented. Gonzalez-Barrera and Lopez (2013) showed that the unemployment rate in the early 2010s was much higher among native-born Mexican-origin adults (14.1%) than their foreign-born peers (10.3%). Other evidence also suggests that foreign-born adults may be more likely to work full-time than comparable native-born individuals (National Academy of Sciences, Engineering, and Medicine 2017; Rumbaut and Komaie 2010).
Finally, place of residence may affect children's poverty risk by shaping the economic opportunity structure for their parents. Local social ties and community institutions constitute important dimensions of the assimilation context that immigrants and their descendants face (Rendón 2019). Additionally, economic well-being is influenced by structural and institutional constraints that vary spatially (e.g., by region, as Baker (2020) demonstrated recently).10 Finally, the cost of living, safety net eligibility, and benefit generosity vary across the country, which influences household well-being and estimates of poverty based on measures that capture such costs and resources (as in our study) (Pacas and Rothwell 2020; Renwick 2011).
Spatial assimilation models predict intergenerational changes in the types of places where families reside, with neighborhood quality and socioeconomic attainment positively associated with immigrant generation (Alba and Logan 1991; Alba and Nee 2009). However, such models do not typically account for differences in the cost of living and have been complicated by the emergence of new, disproportionately rural immigrant destinations (Crowley et al. 2006; Kandel and Parrado 2005; Lichter and Johnson 2009; Lichter et al. 2012; Ludwig-Dehm and Iceland 2017; Marrow 2020). These new destinations not only represent different assimilation contexts than traditional gateways, but they are also characterized by unique political and institutional structures, costs of living, and other such factors that influence well-being (Carr et al. 2012; Hall 2013; Light 2006). Although each type of place may provide economic opportunities (and barriers), recent work by Lichter et al. (2015) revealed exceptionally high poverty risks among newborn children in new, rural destinations. We consider this finding to be a reasonable basis for provisionally treating residence in such places as one of our primary risk factors.
In addition to intergenerational differences in these prevalences, the penalties associated with each factor may vary systematically across groups (Brady et al. 2017). For instance, the penalty for low parental education may vary across immigrant generations because of real or perceived differences in the quality of education received internationally versus domestically (Kaushal 2011). Likewise, the low family employment penalty may vary because of intergenerational differences in wages (Massey and Gelatt 2010), and the effects of place of residence may vary given geographic differences in safety net eligibility and use among foreign-born children and native-born children of foreign-born parents (Bitler and Hoynes 2011; Heinrich 2018). In addition to these and other examples supported by prior research, differences in such penalties may reflect intergenerational disparities in contextual factors and individual characteristics that cannot be easily observed in our data. Although there is a strong rationale to expect systematic differences in penalties, such patterns make it difficult to develop directional hypotheses about the exact nature of these patterns. We therefore consider this issue an empirical question.
Research Objectives
The overall goals of this paper are to estimate differences in poverty rates across immigrant generations of Hispanic children and to evaluate the correlates, or sources, of these differences. We have four specific objectives. First, we produce estimates of poverty rates for Hispanic children in each of four immigrant generations (as defined later) using the SPM. Second, we compare the prevalence of five risk factors—parental age, family structure, parental education, family employment, and place of residence—and the penalty associated with each of them across generations. Third, we assess the substantive importance of these patterns by simulating group-specific poverty rates for a series of counterfactual scenarios in which intergenerational differences in select risk factors are eliminated. Fourth, we draw comparisons with the non-Hispanic White population, which is sometimes used as the reference group in studies of immigrant integration across generations (e.g., Mattingly and Pedroza 2018; Orrenius and Zavodny 2019).
Analytic Strategy
Data
We draw on microdata from the Annual Social and Economic Supplement (ASEC) of the Current Population Survey, which is based on a nationally representative sample of approximately 60,000 households and is a primary source of U.S. labor force statistics. The ASEC includes detailed information on prior-year income and employment and has a number of advantages for our purposes. First, these data include information on the parental birthplace of all individuals in respondent households, allowing us to measure children's immigrant generation regardless of whether the parent(s) resided with the child at the time of the survey. Second, these data include all information needed to construct the SPM, our use of which advances research on poverty among immigrant children. Third, the ASEC is designed to increase the precision of estimates among the Hispanic population and households with children aged 18 or younger (U.S. Census Bureau 2019a). Nonetheless, these data have two important limitations. First, despite recent improvements in data quality, households with foreign-born adults—especially undocumented and recently arrived individuals—remain undercounted in the CPS (Passel and Cohn 2018; Van Hook et al. 2014). Because this undercounted population is likely to be disproportionately disadvantaged, our sample may produce conservative estimates of poverty among the first and second generations. Second, the ethnic and racial classification used in the CPS may introduce measurement error that varies generationally and socioeconomically (see below) in a manner that upwardly biases poverty estimates among later immigrant generations. These two sources of bias operate in opposite directions, with the net effect unknown and contingent on which influence dominates.
We compile files from the 2014–2018 ASEC using the IPUMS database (Flood et al. 2018).11 This period includes the five most recent waves of the ASEC that were available at the time of analysis. The analytic sample is restricted to individuals aged 17 years or younger at the time of the survey. Because some children will be observed twice (in consecutive samples) given the sampling structure of the CPS, cases should be interpreted as person-period observations. We account for these repeated observations of individuals by clustering on person identifiers. The main analytic sample of Hispanic children includes a total of 50,875 unweighted person-year observations.12
Measures
The outcome of interest is a given child's poverty status, which we measure using the U.S. Census Bureau's SPM. Unlike the official measure, the SPM accounts for near-cash transfers (e.g., SNAP, housing vouchers), taxes and tax credits, and geography-specific costs of living (Fox et al. 2015).13 It also includes cohabitating partners, foster children, and nonattached children under age 15 in family size calculations. Further, the SPM uses a different basket of items and cost calculations to determine poverty thresholds than the decades-old formula that underlies the OPM (Meyer and Sullivan 2012; Wimer et al. 2016). As we argued earlier, the ability of the SPM to capture the welfare effects of many factors that vary by nativity is particularly advantageous for our purposes.
Our primary stratifying variable is immigrant generation, which we measure using a four-category typology. The first group includes foreign-born children14 and is classified as the first generation. The second generation includes native-born children with at least one foreign-born parent, and we stratify this population into two groups according to whether (a) one or (b) both parents are foreign-born.15 Finally, we define the third+ generation as native-born children with native-born parents; this category includes children who were born abroad to parents who were U.S. citizens. We cannot consistently distinguish between the third and fourth+ generations. Our estimates will therefore mask a fourth+-generation disadvantage that has been documented elsewhere (Orrenius and Zavodny 2019).16
We further stratify children on the basis of race and ethnicity, restricting our main analyses to the sample of Hispanic children (of all races) and conducting one additional comparative analysis using the non-Hispanic White third+ generation as our reference group. We define children's race and ethnicity according to how they were characterized by the respondent in the CPS interview. This approach allows us to classify children consistently regardless of the number of coresident parents, but it has important limitations (Alba et al. 2018). First, children of multiple ethnoracial backgrounds are classified inconsistently and in a manner that may be correlated with the ethnoracial identities of their parents and the balance of power within the household (Lichter and Qian 2018). Second, attachments to ethnic identities may decline across immigrant generations (particularly after the second generation), and this ethnic attrition may be positively selected on socioeconomic status (Duncan and Trejo 2011, 2018; Fernández et al. 2018; Orrenius and Zavodny 2019).17
We focus on the prevalence of five poverty risk factors. The first is low parental age, which we define as children residing in families headed by adults aged 24 years or younger.18 We estimate the penalty of this risk factor relative to children in families headed by adults aged 35–44 years. We include additional controls to account for children with family heads aged 25–34 years, 45–54 years, and 55+ years, respectively.
The second factor is family structure, with residence in a family with a single, never-married head being the primary risk factor of interest. However, because we also expect children in families with other unmarried heads to face high poverty risks, we distinguish between family heads that are cohabitating; divorced, separated, or widowed; and married.
The third risk factor is low parental education, defined as no high school diploma for the child's family head. We measure the penalty associated with this risk relative to children of family heads with a bachelor's degree or higher. We also include controls to account for family heads with a high school diploma or an associate's degree, respectively.19
The fourth risk factor is low employment among adult family members. Family employment is defined as the average number of full-time equivalents (FTEs; 1 FTE =1,750 hours) worked by all working-age (ages 24–64 years) adult members of the SPM family unit during the calendar year before the CPS.20 We standardize by the number of working-age adults because the upper bound of the absolute FTEs worked is a function of household labor supply. We then construct an indicator that distinguishes between families with less than 0.5 FTE per adult and those with 0.5 or more FTE per adult.21
The fifth risk factor is residence in a new nonmetropolitan destination (reference =metropolitan areas of established destinations). Our residence type variable distinguishes between individuals living in (a) a new, established, or other immigrant destination state, and (b) a metropolitan or nonmetropolitan county. Following Massey and Capoferro (2008), we define immigrant destination status at the state level (see Figure A1 in the online appendix) because we cannot consistently identify place of residence at a finer scale.22 Metropolitan status is provided in the CPS and is defined using the Office of Management and Budget's metropolitan classification system.23
Methods
The analysis proceeds as follows. We begin by estimating the poverty rate among Hispanic children by immigrant generation, estimating the prevalence of all five risk factors for each generation, and quantifying the penalty associated with each factor. Penalties represent the percentage-point increase or decrease in poverty associated with each risk. They are estimated using a series of linear probability models that predict children's SPM poverty status as a function of all five risk factors and a set of control variables, stratified by immigrant generation. Controls include child's age (in years), relationship to the family head (distinguishing between child, grandchild, and other relation), family head's sex, family size, and region of residence.24
We then quantify the contribution of observed intergenerational differences in the prevalence of risk factors to disparities in child poverty. Specifically, following Brady et al. (2017), we produce counterfactual predictions of poverty rates for each immigrant generation under the assumption that group i had the same distribution of a given poverty risk factor as observed among the third+ generation. Although not a formal decomposition, this exercise follows the logic of Kitagawa–Blinder–Oaxaca decomposition (Fairlie 2005; Kitagawa 1955) and allows us to produce easily interpretable estimates of whether and how poverty among first- and second-generation immigrant children would change in the absence of compositional differences with the third+ generation.
Our main analysis focuses on Hispanic children, who represent the largest group of immigrant children. Third+-generation Hispanic children serve as our primary reference group, which allows us to understand how nativity, the presence of foreign-born parents, and correlated characteristics influence children's economic circumstances. These cross-sectional comparisons are valuable because they compare nativity groups within the same social and policy context, where discrimination and conditionality vary by generation. We then compare the poverty rates and the prevalence and penalties of poverty risk factors observed among Hispanic children with those of the third+-generation non-Hispanic White population. These analyses build on our first set of comparisons to evaluate whether and how racial disparities might moderate the influence of immigrant generation. Given evidence of discrimination against the Hispanic population in general (Pager et al. 2009; Pager and Shepherd 2008), we argue that it is useful to evaluate changes in poverty that would occur if characteristics converged to the average levels of the most advantaged group within the U.S. stratification system.25 In doing so, we account for the possibility that convergence among Hispanic immigrant generations might still leave some children disadvantaged relative to other groups.
Results
Poverty Among Hispanic Children
We begin by describing Hispanic child poverty rates by immigrant generation (Table 1). Point estimates of poverty are highest among first-generation children (32.2%) and native-born children with two foreign-born parents (32.1%). At more than 30%, these rates are exceptionally high by most standards. By contrast, poverty rates are more than 10 percentage points lower among both second-generation children with only one foreign-born parent (20.8%) and the third+ immigrant generation (19.1%). However, in both cases, poverty remains above the national average for children during this period (16.2%). Overall, the results suggest a bifurcation in poverty between immigrant generation groups. Poverty rates among first-generation children and second-generation children with two-foreign-born parents converge around 32%, whereas approximately 20% of both second-generation children with one foreign-born parent and the third+ generation reside in poor families.
Prevalence of Poverty Risk Factors Among Hispanic Children
We next describe the prevalence of five major poverty risk factors across immigrant generations (see Figure 1 and Table 1). We begin by examining the share of children residing in families headed by a young adult, aged 24 or younger. Third+-generation children (6.9%) and second-generation children with one foreign-born parent (6.9%) are most likely to reside in such families, followed closely by first-generation children (6.6%). Second-generation children with two foreign-born parents are least likely to reside in a family unit headed by a young adult (5.3%). The range of prevalence rates across generations is 1.6 percentage points, which is relatively small in absolute terms but proportionately represents approximately 23%–30% of the observed rates.
We next consider parental marital status, as proxied by the marital status of the family head. Hispanic children in the first two immigrant generations are least likely to reside in a family headed by a single, never-married adult: 11.8% of second-generation children with two foreign-born parents live in such families, as do 12.3% of first-generation children and second-generation children with one foreign-born parent. In contrast, 14.7% of third+-generation children reside in a family with a single, never-married head. A similar pattern is observed for the share of each generation in families headed by an unmarried adult, irrespective of whether that family head had ever been married or is cohabitating. We find that 32.3% of first-generation children and between 30.6% and 34.9% of second-generation children reside in such families, compared with 41.4% of the third+ generation.
The third risk factor is low parental education, defined as when a family head lacks a high school diploma. More than 40% of first-generation children (41.5%) and second-generation children with two foreign-born parents (49.6%) have a family head with low education. These prevalence rates for low parental education are more than 20 percentage points higher than observed among second-generation children with one foreign-born parent (23.0%) and members of the third+ generation (16.6%). Although this pattern suggests a clear disadvantage among the former two groups, the share of first-generation children in families with a college-educated head (19.2%) is comparable to that among the latter two generations (20.5% for second-generation children with one foreign-born parent and 20.0% for children of the third+ generation). That is, the parental education distribution of the first generation is bifurcated. The prevalence of children with a college-educated family head is considerably lower among second-generation children with two foreign-born parents (7.8%), making these children appear distinctively disadvantaged with respect to parental education.
Fourth, the prevalence of low family employment is distributed evenly across immigrant generations. According to our point estimates, the share of children in such families ranges from 15.6% among second-generation children with two foreign-born parents to 16.8% among first-generation children. Importantly, this is the one of our five risk factors for which differences in the distribution among the first three generational groups are not statistically different relative to the third+ generation.
Finally, we consider the distribution of children by place of residence. Our estimates reveal modest intergenerational variation in the share of children residing in new nonmetropolitan destinations (our residence type of interest) (see Figure A1 in the online appendix). Almost 3% of third+-generation children (2.9%) and second-generation children with one foreign-born parent (2.7%) reside in such places, which is approximately 35%–45% more than the share of first-generation children (2.0%) and second-generation children with two foreign-born parents (1.9%). Although these figures are small in an absolute sense, Hispanic children are underrepresented in nonmetropolitan areas of the United States. Looking beyond our focal residence type, we also find that, across immigrant generations, members of the third+ generation are most likely to reside outside of traditional destination states. Just more than one third (33.9%) of this group resides outside of such states, compared with 28.5% to 29.8% among the first and second generations.
Penalties to Poverty Risk Factors Among Hispanic Children
We next describe levels and intergenerational differences in the penalties associated with each factor of interest (see Figure 2 and Table 2). First, with respect to parental age, the penalty associated with residing in a family headed by a young adult is substantively small and statistically nonsignificant among first-generation children and second-generation children with two foreign-born parents. Among second-generation children with one foreign-born parent and third+-generation children, however, the penalties are statistically significant and substantively meaningful at 6.8 and 6.7 percentage points, respectively. That is, relative to children residing in families headed by adults aged 35–44 years within their own immigrant generations, such children face a probability of poverty that is (respectively) 6.8 and 6.7 percentage points higher net of controls. This pattern of only the latter two generations experiencing benefits to delayed childbearing could reflect a flatter age–earnings profile among foreign-born workers than among their native-born peers.
Second, we consider the penalty for residence in families headed by a single, never-married adult. Such children in the second and third+ generations face significantly higher poverty risks than their same-generation peers living in families with a married head. The point estimate of the penalty is smallest—and not statistically significant—among the first generation and largest among second-generation children with one foreign-born parent (18.1 points). The estimated penalty among second-generation children with two foreign-born parents is 15.3 points, and that among the third+ generation is 12.7 points. We also observe statistically significant penalties for residing in families with a cohabitating head and, with the exception of the first generation, a separated, divorced, or widowed head. For example, the penalty associated with parental cohabitation is 9.9 to 12.3 points for first- and second-generation children, compared with 8.6 points for the third+ generation. These intergenerational differences are a matter of degree rather than kind. However, it is notable that the estimated penalty for living in a family with an unmarried head is lower among the first and third+ generations than among the other two groups. This pattern may reflect different processes, including the relatively low prevalence of dual-earner couples among the first generation (and thus low returns to parental marriage) and higher earnings among the third+ generation (and thus a single parent's greater ability to earn above-poverty wages).26
Third, the penalties associated with low parental education (no high school diploma) are statistically and substantively significant for all groups. With referenceto same-generation peers with a college-educated family head, point estimates of penalties are largest among second-generation children with two foreign-born parents (21.0 points), followed by first- and third+-generation children (17.2 points each). The penalty is lowest among second-generation children with one foreign-born parent (14.0 points). We also note that children of all generations with high school–educated parents face statistically significant penalties of 5.8 points (among the third+ generation) to 11.2 points (among second-generation children with two foreign-born parents). The pattern of point estimates reflects uniformly high penalties for low parental education and does not reveal a clear intergenerational gradient. Of course, this does not mean that the penalties are generated through common processes across immigrant generations.
Fourth, low family employment comes with very high penalties that, at more than 42 points across all immigrant generations, are by far the highest of the five risk factors of interest. The implication is that compared with children in families with adequate family employment (as defined by ≥0.5 FTE per adult), children in low-employment families face exceptionally high poverty risk. Point estimates of the penalties observed across the four immigrant generations range from 42.2 to 43.8 points among the second and third generations to 49.1 points among first-generation Hispanic children. This particularly high penalty for low family employment for first-generation children may reflect their parents' limited safety net eligibility and use, which amplifies the centrality of earnings for family income and thus the importance of work for escaping poverty.
Finally, we consider the penalties associated with residence in new nonmetropolitan destinations, using established metropolitan destinations (the modal residence type for all generations) as the reference group. Contrary to expectations, first-generation children and second-generation children with two foreign-born parents face statistically lower poverty risks than their same-generation peers in metropolitan areas of traditional destination states.27 Point estimates of the returns (i.e., benefits) to residence in new rural destinations are 15.1 points and 18.8 points, respectively. Second-generation children with one foreign-born parent and third+-generation children face statistically similar poverty risks in new nonmetropolitan destinations and established metropolitan destinations. This result suggests that new arrivals (in the generational sense) may benefit from bypassing the social and economic structures of traditional gateways or, relatedly, are subject to particularly strong selection processes into new destinations. These findings are consistent with prior research on new destinations, which has emphasized the declining economic conditions (e.g., low income-to-rent ratios) in traditional gateways, the draw of economic opportunities in new destinations, and positive selection on education and related factors associated with economic success (Jensen 2006; Kandel and Parrado 2005; Lichter and Johnson 2009; Light 2006; Light and Johnson 2009).
Notably, our findings run contrary to results from Lichter et al.'s (2015) study of Hispanic newborns (irrespective of immigrant generation), which revealed significantly higher rates of newborn poverty in new nonmetropolitan destinations. We hypothesized that these differences could be explained by our use of the SPM (vs. Lichter et al.'s use of the OPM) because this measure accounts for the lower costs of living in nonmetropolitan areas and certain states (Laird et al. 2018; Pacas and Rothwell 2020). We confirmed this interpretation by conducting supplementary analyses using the OPM, which revealed substantively similar findings as Lichter et al. (2015) regarding the difference in child poverty between new nonmetropolitan destinations and established metropolitan gateways (see Table A6 in the online appendix).28 This finding underlines the implications of poverty measurement decisions, which we argue are particularly salient (and understudied) for research on poverty among immigrant children. It is also consistent with the aforementioned suggestion that relatively low costs of living have partially driven the emergence of new destinations.
Decomposing Intergenerational Poverty Differences Among Hispanic Children
Our next analysis quantifies the contribution of differences in the distribution of risk factors to intergenerational disparities in child poverty (see Figure 3 and Table 3). Given the schedule of penalties observed for each immigrant generation, we produce counterfactual simulations of the poverty rate for a given group had they been characterized by the same prevalence of the five risk factors of interest as the third+ generation. We consider the contribution of each factor individually and then estimate the net effect of all five factors simultaneously. Throughout these analyses, we hold all control and nonfocal variables at their observed means for each generation.
We begin by discussing the contribution of each risk factor on its own terms. First, differences in the age distribution of family heads account for very little of the poverty differences between the third+ generation and all three immigration generation groups of interest. In each case, the counterfactual poverty rate is within 0.2 percentage points of the group's predicted probability of poverty.
Second, we find that the family structure of first- and second-generation children has a protective effect on poverty relative to that of the third+ generation. All three of these groups would have experienced higher rates of poverty than observed if they were characterized by the same distribution of parental marital status as the third+ generation, among whom unmarried family heads were most prevalent. The absolute differences between the simulated and observed values range from 0.8 percentage points among the first generation to 1.1 percentage points among second-generation children with two foreign-born parents. In proportional terms, simulated changes in family structure produce 2.5%–4.3% increases in poverty over the observed rates.
Third, parental education represents a substantively important source of disadvantage among first-generation children and second-generation children with two foreign-born parents. If first-generation children had the same levels of parental education as the third+ generation, they would have experienced a 2.4-percentage-point (7.5%) lower risk of poverty than was observed. Likewise, second-generation children with two foreign-born parents would have had a poverty risk that was 5.0 percentagepoints (15.6%) lower than observed. In contrast, education accounts for little of the difference in poverty risk between the third+ generation and second-generation children with one foreign-born parent. Our simulation produces a counterfactual poverty rate of 20.3% for the latter, less than 0.1 percentage point lower than observed.
Fourth, intergenerational differences in family employment play a minor role in poverty differences. The difference between observed and simulated poverty probabilities ranges between −0.2 and just 0.4 points, a trivial share of the baseline poverty rate. On the one hand, the limited explanatory role of work patterns is expected given the nonsignificant intergenerational differences in the prevalence of low family employment. However, this weak effect is still notable given the large penalty associated with low levels of employment, highlighting the importance of distinguishing between the respective contributions of risk prevalences and penalties to poverty differences.
Fifth, we simulate child poverty rates for the first two generations, assuming that they share the same geographic distribution across our immigrant–destination typology as the third+ generation. In all three cases, the simulated poverty rate is lower than observed, suggesting that place of residence is a source of disadvantage for these groups relative to the third+ generation. Differences between the observed and simulated values range from 0.8 percentage points (2.5%) among second-generation children with two foreign-born parents to 0.3 percentage points (1.4%) among members of the second generation with one foreign-born parent.
Finally, we consider the net contribution of all five risk factors to intergenerational child poverty differences. If first-generation Hispanic children and second-generation children with two foreign-born parents had the same prevalence of risk factors as the third+ generation (as observed here), they would have experienced probabilities of poverty that were 2.3 (7.1%) and 4.5 (14.0%) percentage points lower than observed. Under this scenario, the respective gaps between these two groups and the third+ generation would have been reduced by 17.6% and 34.6%, respectively. Overall, then, such compositional differences represent a substantively important, but not deterministic, source of disadvantage among these two groups relative to the third+ generation. In contrast, if second-generation children with one foreign-born parent had the same characteristics as the third+ generation, their expected probability of poverty would have remained essentially unchanged (increasing from 20.8% to 20.9%). Hence, this exercise suggests that differences in population characteristics contribute modestly, at most, to poverty differences between these latter two groups.
Comparisons With the Non-Hispanic White Third+ Generation
Our final objective is to provide preliminary evidence of how immigrant generation and ethnoracial identity intersect to advantage or disadvantage Hispanic children relative to their non-Hispanic White peers who are not subject to ethnoracial discrimination. We begin with a simple descriptive comparison of Hispanic children and the third+ generation of non-Hispanic Whites. We then produce an alternative set of counterfactual estimates that use this non-Hispanic White subpopulation as the reference group. These analyses yield three main sets of findings.
First, we find that Hispanic children of all generations have much higher rates of poverty than third+ generation non-Hispanic White children, of whom 9.2% fall below the poverty line. This rate is more than 20 percentage points lower than observed among both first-generation Hispanic children and second-generation Hispanic children with two foreign-born parents. It is also approximately half the rate of the latter two generations of Hispanic children. These ethnoracial differences in child poverty are substantively large by most standards.
Second, prevalence rates are markedly lower among the non-Hispanic White third+ generation than among any of the Hispanic subpopulations for four of the five focal risk factors: young parental age (2.3%); single, never-married parent (3.9%); low parental education (5.3%); and low family employment (9.5%). In comparison with the most advantaged Hispanic generation for each risk factor, we observe nontrivial disparities of 3.0, 7.9, 11.3, and 6.1 points, respectively. Additionally, a larger share of the non-Hispanic White third+ generation (9.4%) resided in new nonmetropolitan destinations, compared with less than 3% of Hispanic children across all generations. Although we had initially hypothesized that such locations were associated with elevated child poverty risks, our empirical results demonstrate that there are returns rather than penalties to residence in new nonmetropolitan destinations for both Hispanic and non-Hispanic children.
Finally, we evaluate how poverty rates among the first through third+ generations of Hispanic children would change if the distribution of all five poverty risk factors converged with that of the non-Hispanic White third+ generation (see Figure 4 and Table 4). For first-generation Hispanic children, such reductions in risk factor prevalences would result in a 13.7-percentage-point decline in poverty. This hypothetical reduction represents a 42.5% decrease in the observed poverty rate. The population of second-generation Hispanic children with two foreign-born parents would experience an even larger reduction in poverty (by 17 percentage points) under such a scenario, lowering poverty to less than half (a 52.9% reduction) of the observed rate. Although such changes represent a significant convergence of poverty rates with third+-generation non-Hispanic White children, they leave substantively important gaps of 5.9 to 9.3 percentage points. Second-generation Hispanic children with one foreign-born parent and third+-generationHispanic children experience reductions of 10.3 (49.5%) and 9.2 (48.2%) percentage points, respectively, under our simulation. The predicted probabilities of poverty fall to within 0.7 to 1.3 percentage points of third+-generation non-Hispanic White children.
Discussion and Conclusions
In this article, we have examined patterns of poverty across immigrant generations of Hispanic children, engaging with prior work on immigrant attainment and the poverty literature. Our analyses point to six overall conclusions with implications for our understanding of immigrant integration, social policy, and future research on child poverty. First, poverty rates are exceptionally high among the first generation of Hispanic children and second-generation children with two foreign-born parents. More than 30% of children in each group live in poor families, which is significantly higher than the poverty rates of approximately 20% observed among the latter two generations of Hispanic children. Of course, a poverty rate of 20% is still above the national average and is far above the poverty rates observed for third+-generation non-Hispanic White children (9.2%).
Second and relatedly, our study highlights the importance of poverty measurement decisions, which to our knowledge have received relatively little attention in the literature on immigrant child poverty. We demonstrate that our use of the SPM leads to substantively different conclusions about the penalty to residence in new nonmetropolitan destinations (e.g., vis-à-vis results based on the OPM from Lichter et al. 2015), in large part because the SPM detects geographic variation in the cost of living (Laird et al. 2018; Pacas and Rothwell 2020). More broadly, we also argue that the use of the SPM is particularly important for research on immigrant poverty given differences in eligibility and use of safety net programs, family size and structure, and local costs of living that this measure captures. The results of this study reinforce findings from our recent comparison of OPM- and SPM-based estimates of intergenerational poverty differences by immigrant generation (Thiede and Brooks 2018).
Third, the penalties associated with the five poverty risk factors of interest vary modestly across immigrant generations. Notably, we find an elevated penalty for low family employment among the first generation, which may reflect the limited access to and use of safety net programs among these children and their parents. However, the lack of a pronounced, consistent gradient across generations—akin to some of the cross-national patterns in penalties that Brady et al. (2017) and Rothwell and McEwan (2017) documented—is also notable. The absence of a strong gradient in penalties may reflect the multiplicity of factors that determine such penalties, which may operate inconsistently across generations. For example, a low penalty for an unmarried parent may reflect the low prevalence of dual-earner couples or the high wages among the parents of a given immigrant generation. Alternatively, other factors, such as racial and ethnic discrimination, may shape penalties more than immigrant generation per se. For example, our supplementary analysis of the third+ generation of non-Hispanic White children reveals a lower penalty for low family employment than any of the four Hispanic subpopulations considered in the main analysis. It may also simply be that penalties are less salient in explaining intranational than international poverty differences. Indeed, in their study of between-state differences in poverty, Laird et al. (2018) concluded that geographic variation in prevalences is central to why poverty rates vary within the United States.
Fourth, intergenerational differences in the prevalence of risk factors explain nontrivial shares of the poverty differences between the third+ generation and both first-generation children (17.6%) and second-generation children with two foreign-born parents (34.6%). Differences in family structure are a source of advantage for the latter two groups relative to the third+ generation, but this effect is offset and reversed by disadvantages in terms of parental education. That is, these groups are not uniformly disadvantaged relative to the third+ generation. This analysis also demonstrates that the factors with the largest penalties are not necessarily those that explain the largest share of intergenerational child poverty gaps. For example, low employment is strongly correlated with poverty, but it explains relatively little of the between-group differences in poverty given the relatively uniform distribution of this risk across immigrant generations. Analysts and policymakers should therefore be careful to differentiate between the marginal effects of changing a risk factor and the aggregate impact such a change would have given the baseline distribution of that factor.
Fifth, intergenerational differences in poverty risk factors—and their apparent contribution to observed poverty rates—among Hispanic children pale in comparison with inequalities with third+-generation non-Hispanic White children. These large inequalities underline the salience of race and ethnicity over and above nativity. Here, two specific points are worth emphasizing. First, intergenerational convergence in poverty risks among Hispanic children is not sufficient to close the child poverty gap between Hispanic and non-Hispanic White children. Second, convergence in the prevalence of the observed poverty risk factors between Hispanic children and third+-generation non-Hispanic White children would reduce the former's poverty rates dramatically, but nontrivial inequalities would remain. This finding demonstrates that advantages that non-Hispanic White children experience extend well beyond the five compositional variables that we focus on here, likely including factors such as English language proficiency (see Table A1 in the online appendix) and many systemic conditions that are not easily measured in our framework. Further attention to these disparities is clearly merited, including further study of ethnoracial differences in the penalties to poverty risk factors.
Finally, our results speak to general theories of the causes of poverty, variously providing support for behavioral, structural, and political explanations (Brady 2019). For example, the substantial penalties associated with single headship and, for some groups, young parental age are consistent with theories of poverty rooted in individual behaviors and decisions. The persistent ethnoracial disparities that we document across our results are consistent with structural explanations that emphasize the role of systemic racism in the U.S. stratification system. Finally, our use of the SPM—which more fully accounts for the impact of the safety net (e.g., near-cash supports, such as SNAP) than the OPM used in most prior research on this topic—underscores the role of the safety net and thus political determinants of poverty in America today (see also Curran 2021; Thiede and Brooks 2018).29
In a context of exceptionally high child poverty rates (Shaefer et al. 2018; Smeeding and Thèvenot 2016) and rapid demographic change, identifying the most salient axes of inequality among children and determining which factors explain such differences are increasingly needed to develop appropriate anti-poverty interventions. Our results point to at least four target areas for interventions aiming to diminish intergenerational inequalities among Hispanic children and reduce child poverty overall. First, improving the educational attainment of foreign-born parents would reduce poverty, as well as intergenerational disparities therein, among Hispanic children. With parents generally beyond their schooling years, policies are needed to enhance and promote adult learning and other training opportunities that lead to upward mobility for parents.30
Second, although low family employment is not an important source of intergenerational differences in poverty among Hispanic children, the penalties are very high (and higher than among non-Hispanic White children). Reducing barriers to parental employment is therefore likely to be an effective means of reducing Hispanic child poverty overall, and supplemental analysis suggests that there is considerable room for improvement: the average FTE worked per adult in Hispanic children's families ranges from 0.764 to 0.862, well below the figure of 0.965 FTE per adult among the third+ generation of non-Hispanic White children's families.31 Of course, beyond some point, increasing parents' work hours outside the home could be detrimental for child outcomes (not to mention parents themselves) (Waldfogel 2006). Accordingly, efforts to increase family labor supply should be paired with wage and tax-credit legislation (e.g., minimum wage increases, childcare tax credits, child benefits) to ensure that, at minimum, one full-time worker is sufficient to keep families out of poverty. At the same time, strengthening anti-discrimination efforts in U.S. workplaces is needed to enhance the opportunities and returns to work among Hispanic parents (Pager et al. 2009; Pager and Shepherd 2008).
Third, the high penalty for low family employment among the first generation highlights the vulnerability of early-immigrant generations to labor market dislocations. Such populations often are ineligible for safety net benefits or avoid them (Laird et al. 2019), despite the importance of these benefits for protecting children from the adverse effects of poverty. Thus, expanded eligibility, utilization promotion, and job protection (to reduce shocks in the first place) are all necessary to protect workers and their children. Normative considerations aside, doing so is necessary to avoid the long-run social costs of child poverty.
Fourth, the economic benefits to residence in new nonmetropolitan destinations suggest the need to encourage and promote both settlement and incorporation into such places. Efforts are needed to reduce the hostility that foreign-born and non-White populations face in some new destinations and to document and popularize the economic and demographic vitality that new arrivals can bring to localities (Carr et al. 2012; Hall 2013; Heinrich 2018; Jensen 2006).
In addition to these policy implications, our results and the limitations of our study should also serve as a basis for more research on this important topic (National Academies of Sciences, Engineering, and Medicine 2016). For one, future studies should look beyond the Hispanic population to more fully consider how nativity and immigrant generation intersect with race—and racism within the U.S. stratification system—to influence children's economic circumstances (Thomas 2011). High and growing levels of diversity among recent cohorts of children raise the possibility of increasingly complex patterns of inequality that attention to race or nativity alone may mask.
Additional policy-focused analysis is also needed. For example, more evidence is needed to unpack the policies—or factors amenable to policy interventions—that explain the processes that generate penalties. There also is a need for updated analyses of whether and how the role of safety net programs varies across immigrant generations and how such disparities may be driven by patterns of eligibility or utilization (Bean et al. 1997; Heinrich 2018; Jensen 1988). Laird et al.'s (2019) work on the potential influence of a change in the “public charge” policy provides one recent example that should be built upon. Future research could also take a historical perspective to assess how changes in immigration policy and the safety net have shaped immigrant children's circumstances, disparities across immigrant generations, and the overall trajectories of integration and social mobility among such children.
Finally, as other scholars have noted (e.g., Alba et al. 2018; Prewitt 2018), new data collection and measurement efforts are needed to better understand the social and economic circumstances of immigrants and their descendants in the United States. Two specific priorities are to better correct—or at least more precisely quantify—the biases associated with the undercount of recent arrivals and the selective ethnic attrition among later generations. In the absence of such measures, comparative analyses of poverty and other outcomes will continue to be limited by ethnic attrition and the resulting biases introduced to each generation's estimates. Attention to these and related questions is needed to understand the economic underpinnings of new cohorts of children, who are increasingly diverse along multiple axes. Such evidence can inform policies to ensure that growing diversity among youth does not result in increasing fragmentation with respect to socioeconomic conditions during childhood and overall life chances (Lichter 2013).
Acknowledgments
Previous versions of this paper were presented at the University of Wisconsin–Madison's Institute for Research on Poverty (IRP) and the 2018 annual meeting of the Population Association of America. Thiede gratefully acknowledges support from the IRP extramural grants program. The authors thank Lonnie Berger for constructive feedback on an early version of this paper. We also acknowledge assistance provided by the Population Research Institute at Penn State University, which is supported by an infrastructure grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD041025). The work of Thiede and Jensen was also supported by the USDA National Institute of Food and Agriculture and Multistate Research Project #PEN04623 (Accession #1013257). The authors take responsibility for any and all remaining errors.
Notes
We use the term Hispanic throughout for two reasons. First, according to a recent Pew Research Center analysis, a majority (51%) of the population identifying as Hispanic or Latino has no preference regarding the term, and most of the remaining 49% prefer Hispanic (32%) over Latino (15%) (Lopez et al. 2020; statistics influenced by rounding error). Second, the use of the term Hispanic is consistent with most of the literature that we engage in this study. However, we acknowledge that Latino and Latinx are sometimes used as alternatives or preferred labels for this population.
As one anonymous reviewer and other scholars (e.g., Baker 2020) have noted, this gap in research also reflects the disproportionate emphasis among U.S. poverty scholars on cities in the Northeast and Midwest. This focus overemphasizes poverty among Black populations relative to Hispanic populations and populations in the South, Southwest, and West regions more broadly.
All of the studies discussed in this section measured poverty using the U.S. government’s official poverty measure unless otherwise noted.
These findings for children exposed to poverty are compared with the counterfactual of being nonpoor in the United States.
Lichter and colleagues (2005) also found that the magnitude and direction of these intergenerational disparities varied by race. For example, poverty rates among non-Hispanic Black children in the first immigrant generation (26.2%) were lower than among the third generation (34.0%), a finding that Thomas (2011) replicated and demonstrated is moderated by familial context. Such findings demonstrate that “downward assimilation” is possible and that the choice of the reference group has conceptual implications (e.g., within- or between-race comparisons).
The OPM also does not account for geographic variation in living costs, which are likely to moderate economic hardship unevenly across immigrant generations given nonrandom settlement patterns (Lichter and Johnson 2006; Pacas and Rothwell 2020).
Thiede and Brooks (2018) described trends in poverty rates by immigrant generation and ethnoracial group from 1993 to 2016 using both the OPM and SPM. Using that paper as a point of departure, here we address more analytic questions about the social and demographic factors that explain contemporary intergenerational differences in child poverty as defined by the SPM.
See Rothwell and McEwen (2017) for another excellent application of Brady et al.’s framework to the analysis of child poverty. Likewise, Laird et al. (2018) used this framework to study interstate differences in poverty in the United States using the SPM.
Citizenship, authorization status, and English proficiency are also important and modifiable correlates of economic outcomes among immigrants (Mattingly and Pedroza 2018; Sullivan and Ziegert 2008). However, language proficiency data were not collected in the data we use from the Current Population Survey (CPS), and children’s citizenship status varies only within the first generation, so defined. Authorization status has been estimated using the CPS (Van Hook et al. 2015), but it is not included as a predictor because it does not vary within all generations. In one effort to address these limitations, we used data from the 2014–2018 American Community Survey to compare the English language proficiency of native- and foreign-born adults with children (under age 18) in their household—the closest proxy to the cohort of parents in our data (see Table A1 in the online appendix). We found that foreign-born adults are much more likely to speak poor or no English (44.2%) than native-born adults (3.1%). Low English language proficiency is therefore most likely to be a risk factor for first-generation children and second-generation children with two foreign-born parents in our sample.
For example, Baker (2020) showed that disproportionately high poverty in the U.S. South can be partially explained by the weakness of collective power resources that influence labor market structure and resource distribution.
For 2014, we include only the three eighths of the sample that received an experimental redesign. This approach is consistent with guidance for using the 2014 ASEC with subsequent samples (i.e., 2015 and beyond) (Flood et al. 2018).
All analyses are weighted using the ASEC-specific person weights provided by IPUMS.
The SPM does not account for noncash transfers, such as public health care and education programs (Fox et al. 2015; Garfinkel et al. 2006).
This group includes children who are both foreign-born noncitizens and foreign-born naturalized citizens. Although compositional differences exist between these two subgroups, these subpopulations are combined in this analysis because of the relatively small sample size of foreign-born naturalized citizens. A supplementary analysis with the first generation stratified by citizenship status is provided in the online appendix (see Table A2).
As Masferrer et al. (2019) documented, hundreds of thousands of U.S.-born children with Mexican-born parents—members of the second generation—have returned to Mexico over recent years. To the extent that this process is selective, it will influence our poverty estimates for both second-generation groups, so defined.
Among the Mexican-origin population in our sample, many later-generation individuals are from families that were not immigrants but were rather incorporated into the United States through conquest (Montejano 1987). These populations faced high levels of discrimination in the United States, which—combined with ethnic attrition—has been hypothesized to explain declining socioeconomic attainment from the third to the fourth+ generation of Hispanic individuals (Orrenius and Zavodny 2019).
The results of a supplemental analysis exploring this issue are reported in Table A3 of the online appendix. Among first- and second-generation children with at least one Hispanic coresident parent, between 94.0% and 98.9% are identified as Hispanic in the CPS. Nearly all others are identified as non-Hispanic White. Among the third+ generation, 89.9% of children with at least one Hispanic coresident parent are identified as Hispanic. This percentage is lower than that for the other groups, a finding consistent with the ethnic attrition documented in other research.
We measure age, marital status, and education for family heads rather than mothers (or fathers) because not all children have coresident parents. This decision allows for consistent measurement and is in line with our conceptualization of the family as an economic resource–sharing unit. However, for some outcomes (e.g., child development), it may be important to link children to mothers or fathers directly (Crosnoe et al. 2016). We provide descriptive statistics using an alternative measurement approach in the online appendix (see Table A4). For this alternative analysis, we use the characteristics of the coresident mother (if present), father (if only the father is present), and family head (if the child has no coresident parents).
The high school diploma category includes those with CPS education categories of (a) high school diploma or equivalent and (b) some college but no degree.
We set 1.0 FTE to 1,750 hours because this is the total number of hours worked per year given a 35-hour workweek and 50 workweeks per year. We include the work of all working-age adults under the assumption that the family, as defined by the SPM, is a resource-sharing unit. We exclude adults at ages commonly associated with college attendance and adults who are at or beyond retirement age.
Our choice of 0.5 FTE per adult is reasonable but ultimately arbitrary. We assume that a household with two working-age adults that has an equivalent of one full-time worker (0.5 FTE per person) would be deemed adequately employed. A supplemental analysis using a threshold of 1.0 FTE per adult revealed substantively similar conclusions (see Table A5 of the online appendix).
The state-level measures that we necessarily use mask substate heterogeneity in destination type that could be detected with county- or metropolitan area–level data (Johnson and Lichter 2008; Ludwig-Dehm and Iceland 2017).
The final categories used in the analysis are established metropolitan, nonmetropolitan, new metropolitan, new nonmetropolitan, other metropolitan, other nonmetropolitan, and all counties without an identifiable metropolitan status.
Our estimation uses robust standard errors to account for the heteroskedasticity inherent in these models.
Our approach differs from that of other important studies. For example, Jiménez et al. (2018) emphasized the changing circumstances of recent cohorts of children vis-à-vis their parents’ cohort (i.e., comparing the 2010 third generation with the second generation in 1980).
Among children in families with a married head, the average FTE worked per adult was 0.748 among the first generation compared with 0.883 among the third+ generation.
In addition, children in all immigrant generations face statistically lower poverty risks in metropolitan areas of new destination states, suggesting that our main finding is not driven purely by the lower costs of living in nonmetropolitan areas. That is, residence in new destination states itself is associated with reduced child poverty risk.
The conclusions regarding the penalty associated with residence in new nonmetropolitan destinations are identical when we followed Lichter et al.’s (2015) approach and pooled all Hispanic children. The results of models stratified by immigrant generation (Table A6 in the online appendix) were consistent with the pooled results but also revealed intergenerational heterogeneity. These findings therefore highlight the implications of choices regarding poverty measurement and stratification by generation.
Of course, we also speak to political explanations by merely redocumenting the high rates of child poverty in the United States, which prior research has demonstrated are largely a function of social policy decisions (Brady et al. 2017).
Perversely, full-time employment (often in multiple jobs) may represent an important barrier to such training programs, given the related time constraints.
Hispanic families are also characterized by substantial gender disparities in employment. In our sample of families with children, the average FTE worked by female adults is 0.636, compared with 0.938 among male adults.
