Introduction

Just over 40 years ago, Hispanics comprised less than 5% of the US population, according to the 1970 Decennial Census. By 2010, they represented approximately one-sixth of the US population, registering in excess of 50 million people. Additionally, the Hispanic population is young and will continue to grow rapidly (Johnson et al. 2014). The growing share of the population identifying as Hispanic demands a deeper understanding of how Hispanics have fared over time. Examining key indicators can inform what we know about the social position of Hispanics relative to other racial/ethnic groups in the country (Bean et al. 2009; Massey and Eggers 1990; Tienda and Lii 1987; Waldinger 1999). A long-term view can help inform research, debate, and policy geared toward addressing challenges and opportunities facing growing Hispanic populations. This is particularly critical in the current policy landscape, where discussions about immigration—especially undocumented immigration—are at the fore. Understanding the characteristics of Hispanics can illuminate the potential impact of proposed legislation.

This paper carefully considers the poverty experiences of Hispanics over the past 40 years, including the impact of the Great Recession’s disproportionate impact on Hispanic employment (Mattingly and Pedroza 2015; Taylor et al. 2011; Young 2012). Using Decennial Census and American Community Survey data, we ask: have Hispanic poverty rates deteriorated relative to those of whites? And, just as importantly, what factors explain both changes in Hispanic poverty and the gaps between Hispanics and whites since 1980? In order to answer these questions, we analyze how well key social and economic indicators explain changes in Hispanic poverty.

Literature Review

The relative position of Hispanics to whites, as measured by poverty rates, has changed little over the past four decades. However, the underlying characteristics of both groups did not stay the same. Our study synthesizes the main drivers of Hispanic poverty discussed in the literature. Below, we discuss previous research on how the determinants of poverty are similar to or different from other racial/ethnic groups. We then consider the extant literature on the drivers of poverty gaps between relatively advantaged and disadvantaged populations and how poverty trends have changed over time. We build on past research by analyzing the extent to which changes in the white–Hispanic poverty gap between 1980 and 2010 reveal either a convergence of advantages between the two populations or rising racial/ethnic disadvantages.

Determinants of Hispanic Poverty

Poverty researchers have demonstrated clear linkages between a host of demographic characteristics and poverty. Analyzing poverty rates among Hispanics entails accounting for factors affecting the general population as well as capturing the influence of the foreign-born population and English language proficiency on Hispanic poverty rates. Many studies find Hispanic poverty is driven, in large part, by usual suspects, including age, gender, educational attainment, family composition, and fertility (Aponte 1991; Garcia 2011; Lopez 2013; Siordia and Leyser-Whalen 2014). However, some evidence suggests labor market attachment and labor market participation may be particularly associated with Hispanic poverty (Garcia 2011; Lopez 2013).

Since most undocumented immigrants are Hispanic, past research has indirectly examined the role of legal status and English language proficiency as predictors of Hispanic poverty. Passel and Cohn (2009) estimated one-fifth of undocumented Hispanic adults lived below the poverty line, and one-third of children with undocumented parents lived in poverty. Orrenius and Zavodny (2013) find immigrant status explains part (about half as much as educational attainment) of the poverty gap between Hispanics and non-Hispanic, whites. Among Mexican immigrants, Garcia (2011) finds likely undocumented status (a proxy indicator based on English proficiency among other variables) strongly predicted poverty as much as did being unemployed, followed by working in an immigrant job, defined as low-status and low-wage jobs typically occupied by male and female immigrants as operationalized by Douglas and Saenz (2008). Notably, education reduced the odds of poverty among non-Mexican immigrants but offered only limited protection from poverty among Mexican immigrants. Moreover, Sullivan and Ziegert (2008) find Hispanic poverty rates remain high partly due to limited English language proficiency.

Decomposition of Gaps in Poverty

Scholars have raised additional questions regarding Hispanic poverty, including: What accounts for the gap between Hispanic poverty and poverty among other groups? What changes in the underlying composition of Hispanic populations would make the most impact in reducing poverty? Knowing the relative importance of the contributing factors related to poverty has clear implications for addressing disparities in poverty.

Two known studies (Orrenius and Zavodny 2013; Sullivan and Ziegert 2008) have employed Blinder-Oaxaca decompositions to estimate how well different correlates of poverty explain the white–Hispanic gap in poverty rates, while another study discussed below decomposes child poverty by parents’ nativity (Van Hook et al. 2004). Orrenius and Zavodny (2013) compare US-born and foreign-born Hispanic poverty to poverty among non-Hispanic whites. Sullivan and Ziegert (2008) compare Hispanic immigrants (across different countries of origin) to the non-Hispanic, foreign-born population. Both studies find Hispanic heads of households’ English proficiency and educational attainment best explain the gap in poverty between Hispanic populations and other groups. Both studies suggest that if Hispanic educational attainment and English language proficiency more closely resembled those of non-Hispanic groups (whites and immigrants), then Hispanic poverty rates would be lower. The 2000 poverty gap between Hispanic immigrants and other immigrants would drop by half in this scenario (Sullivan and Ziegert 2008). The 2009 white–Hispanic poverty gap would drop by more than half (from roughly 12.1 to 2.3%) if Hispanics had the same characteristics as whites, driven in large part by the education and English proficiency differences between whites and Hispanics (Orrenius and Zavodny 2013). Finally, the observed influence of language and education on Hispanic poverty is partially offset by “a relatively high level of family work effort” (Sullivan and Ziegert 2008, p 683).

Hispanic Poverty Trends Over Time

Key studies have examined poverty trends over time. While overall Hispanic poverty has remained fairly constant over the past four decades, several within-Hispanic variations in poverty trends are evident. Lopez (2013) notes, among heads of household, Hispanic female poverty fell between 1990 (38%) and 2000 (32%)—and then again by 2006–08 (to 27%)—but stood twice as high as Hispanic male poverty at each time period. In addition, Hispanic poverty rates differ by nativity. Hispanic immigrant male poverty outpaced poverty among US-born men at each time period. Among Hispanic women, immigrant poverty was slightly lower than among the US-born in 1990 before climbing in 2000 to surpass US-born Hispanic female poverty and rising further still in 2006–08 (Lopez 2013, pp 55–57).

Van Hook et al. (2004) document differences in child poverty by parents’ nativity between 1970 and 2000. They find poverty remained stable among children with US-born parents. Poverty among Mexican Hispanic children with US-born parents was higher than among other Hispanic children (30 and 23%, respectively), but their poverty rates converged by 2010 (25 and 26%, respectively). Among children of immigrants, Mexican child poverty remained very high (above 30%) between 1970 and 2000, while poverty among other Hispanic children of immigrants rose from 15 to 25% during this time. Moreover, differences in racial/ethnic composition, parental education, and employment help account for some of the poverty gap between children of immigrants and other children.

Iceland (2003) also examines poverty trends over the long term. He finds the largest declines in poverty—due chiefly to income growth—took place before 1970. He writes, “As economic growth slowed in the 1970s and 1980s, inequality and demographic changes together served to keep it that way. Yet the strong economy, coupled with the waning effect of income inequality and demographic change, once again decreased absolute poverty in the 1990s” (Iceland 2003, p516). In addition, he attributes poverty among Hispanics to rising income inequality in the 1970s. Whereas Hispanic concentration in low-wage work appears to have prevented their poverty rate from falling after 1970, family structure had a prominent effect in increasing poverty among blacks (but not Hispanics) through 1990.

Taken together, these studies highlight the importance of examining Hispanic trends broadly. In this paper, we examine how demographic correlates of poverty influence Hispanic poverty rates. We also analyze Hispanic poverty across time, through the recent recession.

Research Questions

Our approach directly informs debates regarding the future of Hispanics’ socioeconomic position compared to whites. The extant literature finds support for two competing scenarios: (a) the increasing entrenchment of a new underclass of Hispanics (Garcia 2011 on Mexican immigrants in the southwest; Massey and Pren 2012a, b) and (b) the potential for resilient Hispanic households whose likelihood of poverty may drop (Orrenius and Zavodny 2013) and become less tied to ethnic background than in the past (Van Hook et al. 2004). We pose two questions to address these debates.

First, we ask: How have poverty rates among Hispanics changed relative to whites between 1980 and 2010? In answering this, we also examine how individual and household characteristics predict Hispanic poverty status over time in comparison with non-Hispanic whites.

Second, we ask: How much do key explanatory factors—including US nativity, citizenship status, and English language proficiency—account for changes in Hispanic–white poverty gaps? That is, are the apparent drivers of differences in poverty by race/ethnicity due to rising disadvantages or a convergence of advantages across these two groups? Specifically, if Hispanic disadvantages (i.e., poverty-enhancing characteristics relative to whites) have grown or become more closely related to poverty over time, then we expect the poverty gaps across groups to stem from rising disadvantages. We call this the rising racial/ethnic disadvantage hypothesis. By contrast, the observed racial/ethnic gaps in poverty may be increasingly driven by poverty-reducing gains made across the groups. We call this the convergence of advantages hypothesis. We measure changes in each group’s characteristics over time not whether the returns to these characteristics differ along racial/ethnic lines. Past work has established wage inequality, for example, can be partially explained by a skill premium enjoyed by whites as well as discrimination affecting blacks (Kim 2010).

Data and Methods

This paper addresses changes in poverty trends among Hispanics between 1980 and 2010 using Census data. We begin in 1980 as that is the first year for which we can capture language proficiency, an important correlate of Hispanic poverty. Using the Integrated Public Use Microdata Series (IPUMS) from the Minnesota Population Center, we combine data from the Decennial Census (1980, 1990, 2000) and the American Community Survey (ACS, 2006–2010, hereafter 2010) (Ruggles et al. 2015). The data are from the 1980 Decennial Census (5% state sample); 1990 Decennial Census (1% sample); 2000 Decennial Census (5% sample); and the 2006–2010 ACS (5-year sample). We chose the 1990 metropolitan (1-in-100) sample because it allows us account for differences by metropolitan area status. Only the Decennial Census (prior to 2010) and the ACS include measures of poverty, English language proficiency, and US nativity.

We use IPUMS data to analyze poverty trends across racial/ethnic groups over time. In addition to presenting descriptive analyses and regression models predicting poverty for each decennial year, we also use decomposition techniques to better understand what factors are driving change across time. Data limitations preclude similar analysis before 1980 because English language proficiency, an important correlate of Hispanic poverty, is not included in earlier censuses. Consistent with earlier research (Garcia 2011; Lopez 2013; Orrenius and Zavodny 2013), our analyses are limited to heads of household. We limit our analysis to household heads age 25 and over to observe respondents who have had an opportunity to complete formal education. All analyses use household-level weights. Note that in preliminary analyses, we observed similar results when all adults over age 25 were included.

Variables

Following is a discussion of variables in our analyses. Table 1 presents descriptive statistics for all independent variables by year.

Table 1 Descriptive statistics for all variables used in multivariate analyses
Full size table

Dependent Variable

Our dependent variable is poverty status. We predict whether or not the head of household’s total family income was below the official poverty threshold (versus at or above the threshold) for their family size and composition in the previous year (Census Bureau 2017). The official poverty measure can be calculated across time, which allows for analyses of poverty trends over several decades. Across the years, between 9.7 and 10.2% of the full analysis sample were poor (Table 1).

Independent Variables

Race/ethnicity We begin with descriptive comparisons and consider differences between Hispanics and non-Hispanic whites. Although whites remain the majority across years, the sample became more diverse by race/ethnicity across the decades (Table 1). To help answer our main question, we discuss the relationship between Hispanic ethnicity and poverty over time relative to whites (Table 2). Finally, the decomposition models focus on two groups (Hispanics and whites) because decomposition methods are meaningful where poverty gaps remain large and persistent across all time periods. By contrast, other poverty gaps nearly converged by 2010 (e.g., Asian and white; Hispanic and black), and narrow disparities limit the applicability of decomposition methods. Since multiracial identity is not recorded until 2000, we omit people who report two or more races.

Table 2 Odds ratios predicting poverty status (household heads)
Full size table

English language proficiency We also include an indicator of English language proficiency. We examine the effect of knowing English “not well” or “not at all” compared to speaking English “well” or fluently (“only English”) as the reference category. While in 1980, only 1.9% of the sample (Hispanics and non-Hispanics combined) did not have strong English proficiency, by 2010 this had grown to 4.1%.

Nativity and citizenship We include a three-category variable to capture nativity and citizenship status: US-born (including those born in territories and those born abroad to US citizen parents), our reference category; naturalized citizen; and noncitizen. In 2010, the proportion of noncitizens was more than twice as large as in 1980, and naturalized citizens also represented a larger share of the population.

Other explanatory variables In addition to accounting for diversity by race, Hispanic origin, language, nativity, and citizenship status, we also examine poverty by: employment (working in a low-wage job typically associated with immigrant workers, hereafter “immigrant job”; working in a professional occupation; number of adults working in the family; whether the household head worked 50 or more weeks the previous year); gender (coded 0 for male, 1 for female); age (10-year intervals from 25 to 34 to include only people who have had a chance to complete schooling; householders age 55–64 are the reference category); educational attainment (less than high school; 12 years of education; some college, the reference category; and four or more years of college); marital status (married with both spouses present, the reference category; married with one spouse present; divorced; separated; never married; widowed); the presence of young children (coded 1 for presence of any children under five); number of children (categorical variable for number of own children under age 18—from 0 to 3 or more with 1 child as the reference category); and metropolitan status (suburban, the reference category; rural; urban; other metro; and not identifiable).

Analyses

We begin by presenting Hispanic and white differences in poverty rates in 1980, 1990, 2000, and 2010 (Fig. 1 and Table 1). Next, we use logistic regression models to predict poverty status among heads of household age 25 and older (Table 2). Logistic regressions for each decade allow us to address whether the relationship between poverty and educational attainment, for example, remains significant net of other characteristics.

Fig. 1
figure 1

White and Hispanic Poverty, 1980–2010. Authors’ calculations of IPUMS data from Ruggles et al. (2015). Analysis sample of heads of household age 25 years and older (excludes group quarters). White figures reflect non-Hispanic populations. Hispanic figures include Hispanics of any race

Full size image

Finally, we examine characteristics associated with poverty gaps over time between whites and Hispanics. We employ Fairlie decompositions to estimate the portion of the poverty gap due to differences in characteristics between two groups (for a detailed discussion, see Van Hook et al. 2004). Fairlie decompositions are expressed as a nonlinear equation:

$$P_{a} - P_{b} = \left[ {\mathop \sum \limits_{i = 1}^{{N_{a} }} \frac{{F\left( {X_{i}^{a} \beta_{b} } \right)}}{{N_{a} }} - \mathop \sum \limits_{i = 1}^{{N_{b} }} \frac{{F\left( {X_{i}^{b} \beta_{b} } \right)}}{{N_{b} }}} \right] + \left[ {\mathop \sum \limits_{i = 1}^{{N_{a} }} \frac{{F\left( {X_{i}^{a} \beta_{a} } \right)}}{{N_{a} }} - \mathop \sum \limits_{i = 1}^{{N_{a} }} \frac{{F\left( {X_{i}^{a} \beta_{b} } \right)}}{{N_{a} }}} \right] ,$$

where N is the sample size for each group (N a and N b ) (Fairlie 2006). Similar to other decomposition methods, the Fairlie approach estimates what would happen to poverty if Hispanics as a group had higher mean levels of, for example, educational attainment and English language proficiency without also assuming Hispanics had the same returns to those characteristics as whites. We thus present the explained difference in poverty gaps attributable to underlying differences in key characteristics between each set of groups. Unlike Blinder-Oaxaca decompositions—a common alternative approach—the Fairlie method was developed to analyze nonlinear outcomes. The equation above allows the mean poverty rate to differ across two logistic functions (F) predicting the relationship between poverty and characteristics and coefficients for two groups. Fairlie decompositions feature other advantages. The calculation of the Fairlie estimates adjusts for differences in (a) the underlying size of the two groups being compared and (b) the distribution of characteristics between the two groups. The Fairlie method draws random subsamples of two groups to ensure both are equal in size. Then, the random subsamples are ranked according to each person’s predicted probability of being in poverty to account for the contribution of group differences (in characteristics) to the poverty gap. Finally, the decomposition technique repeats the process, randomizes the ordering of the explanatory variables, and reports mean results across 100 replications. Differences across time were statistically tested by comparing 1980 and 2010 using t tests, calculated as follows:

$$\frac{{C_{1} - C_{2} }}{{\sqrt {(\sigma x_{1} )^{2} + (\sigma x_{2} )^{2} } }},$$

where C 1 is the coefficient for the first model, C 2 is the coefficient for the second model, σx 1 is the standard error corresponding to C 1, and σx 2 is the standard deviation corresponding to C 2. We also ran separate models for each comparison using the Blinder-Oaxaca decomposition method (Jann 2008) as well as a decomposition method (KHB) developed to address the problem of rescaling in nonlinear models (Kohler et al. 2011). The results (available upon request) closely mirror the Fairlie results. Black–white and black–Hispanic decomposition results are also available from the authors upon request. Black–white poverty inequality are well known and documented in the literature on segregation (Rugh and Massey 2014), urban poverty (Iceland 1997), decomposition of black–white wage inequality (Couch and Daly 2002; Kim 2010), and decomposition of black–white unemployment (Couch and Fairlie 2010); therefore, they are not presented in this paper.

Results

Descriptive Statistics

Since 1980, poverty rates have remained relatively stable among Hispanics even as the number and proportion of Hispanics has grown. In 1980, 21.4% of Hispanics heads of household (age 25 an older), or over 753 thousand households, lived in poverty. Thirty years later, the percent of Hispanics in poverty remained similar (22.1%, Fig. 1), yet the number of households in poverty had increased to over 2.7 million (analyses not shown), more than a threefold increase. The stable trend in poverty rates persisted even as immigration and natural increase contributed to Hispanic population growth and as Hispanic populations became increasingly diverse. Mexicans remain the largest group of both US-born and foreign-born Hispanics and comprise the vast majority of undocumented Hispanic immigrants (Massey and Pren 2012a). However, other Hispanic origin groups have grown in number, especially Central Americans (Massey and Pren 2012b). By 2010, most (60%) immigrants from Latin America had arrived in the USA after 1990 (analyses not shown). Yet, despite the presence of new waves of arrivals, poverty among Hispanics as a group did not increase. The relative stability in the Hispanic poverty rate appears to be driven by US-born Hispanics.

Compared to the rest of the population, Hispanics have occupied a middle position between non-Hispanic whites and non-Hispanic blacks for decades, although black and Hispanic poverty rates have converged over time (Firebaugh and Farrell 2016; Mattingly and Pedroza 2015). In our sample, among household heads age 25 and older in 1980, the poverty rate among blacks (27.6%) exceeded the white poverty rate (9.0%) nearly three to one. At the time, the Hispanic poverty rate (21.4%) represented a relatively middle position between black and white poverty. Importantly, poverty declined for all groups between 1980 and 2000, but increased slightly by 2010, concurrent with the Great Recession. Despite the gains represented by a falling poverty rate, black poverty—which fell 10% points between 1970 and 2007, before the Great Recession (analyses not shown)—remained nearly 2.5 times higher than white poverty by 2010.

Multivariate Analyses

Each of our multivariate models presents findings for each decade: 1980, 1990, 2000 and 2010. Echoing our descriptive findings, a very strong relationship between race/ethnicity and poverty is apparent (Table 2). However, consistent with research on the declining significance of race (Wilson 2015) and Hispanic child poverty (Van Hook et al. 2004), we see that the magnitude of the relationship appears to decline over time. The models also show the consistent importance of English language proficiency, and the reversal of US nativity. In 1980, naturalized citizens were less often poor than their US-born counterparts but by 2010, they were 20% more likely to be poor. Noncitizens have consistently been at higher risk of poverty than the US-born, and their risk of being poor has increased over time. Notably, working in an immigrant job (low-wage occupations typically associated with immigrant workers) was associated with 1.3 times the odds of living in poverty in 2010 but roughly the same as working other jobs before 2010. The other explanatory variables in the models acted as we might anticipate: working outside of a professional occupation, fewer workers, fewer weeks of employment, being female, being younger, having completed less education, having more children, having a young child, and living outside of the suburbs are all associated living above the poverty threshold.

Next, we consider what explains the gaps in poverty rates between Hispanics and whites (Table 3). For example, English language proficiency accounts for a substantial portion of the white–Hispanic poverty gap. English proficiency matters at each time period and is tied to 22% of the explained difference in 2010 (1.8% out of the 8.1% total) in the Fairlie decomposition results. Notably, we can account for a much larger portion of the Hispanic–white poverty gap in recent years than in 1980. By 2010, the gap is primarily a function of differences in the characteristics of Hispanic and white householders rather than race/ethnicity alone. In other words, knowing a householder’s racial/ethnic identification (e.g., Hispanic or non-Hispanic white or black) in 1980 was more informative about the household’s likelihood of living in poverty than it was in 2010. These results echo the multivariate findings: racial/ethnic classification seems to decline as a correlate of poverty, net of compositional differences.

Table 3 Fairlie decomposition models
Full size table

The Hispanic–white gap persisted largely unchanged between 1980 and 2010 (Table 3). Most of the persistent white–Hispanic gap (ranging from 11.6 to 13.5% age points across the decades) can be traced to differences in the two groups’ characteristics. Taken together, three main differences in the two groups account for approximately half of the full white–Hispanic gap (regardless of the time period): having less than a high school education, having three or more children, and English language proficiency. Education is the key driver of differences in white–Hispanic poverty rates. Notably, the white–Hispanic educational attainment gap grew between 1980 and 2010, even as attainment rose among household heads in both groups. In retrospect, although a substantial education gap remains, rising education attainment among Hispanics may explain why Hispanic poverty did not increase over time. In addition, English language proficiency among householders helps explain the white–Hispanic gap. English may be more important among some Hispanic groups than others, however. Indeed, English proficiency improved among Puerto Rican and Cuban Hispanics in our sample but declined among Mexican and other Hispanics observed at each decade.

By 2010, three competing factors mattered more than in earlier decades and held the white–Hispanic poverty gap from widening or shrinking. On the one hand, single and noncitizen householders became more common among Hispanics than in earlier time periods. Given the growing share of Hispanic householders who were noncitizens (24% in 1980 and 33% in 2010) or single (10% in 1980 and 20% in 2010), we would have expected Hispanic poverty to increase over time, absent other compositional shifts. On the other hand, the number of adult workers in Hispanic households (1.25 in 1980 and 1.31 in 2010) outpaced white households (1.14 in 1980 and 1.09 in 2010). It appears adult workers in Hispanic households roughly canceled out the poverty-inducing effects of single and noncitizen householders. Looking ahead, two divergent futures await the white–Hispanic poverty gap: (a) rising educational attainment and smaller family sizes narrowing the gap as Hispanic householders rely on high numbers of adult workers or (b) rising poverty as Hispanic households are increasingly headed by single and noncitizen adults.

Based on the decomposition results, we can also describe how many Hispanic households enjoy poverty-reducing advantages. As discussed above, Hispanic poverty would have fallen if Hispanic characteristics more closely resembled those of white householders. We pinpoint six poverty-related factors along which the two groups differ: (1) educational attainment, (2) English language proficiency, (3) number of children, (4) lack of US citizenship, (5) number of adult workers, and (6) single status. We also know how many Hispanic households report these risk factors, which helps further answer whether the study period is a time of rising disadvantages or a convergence of advantages between Hispanics and whites. Table 4 splits the sample at each time period into Hispanic households reporting (a) two or fewer of the above factors (advantaged Hispanics) versus (b) three or more of the same factors (disadvantaged Hispanics). First, the Hispanic poverty rate is now on par with whites among advantaged Hispanic households. Among this group, poverty fell from 12 to 13% in 1980 and 1990 to 10–11% in 2000 and 2010. These households kept poverty rates from increasing, and they represent a sizeable portion (approximately 60%) of Hispanic households in our study. In stark contrast, the remaining Hispanic households reported a consistently high (between 32 and 38%) poverty rate.

Table 4 Hispanic poverty and characteristics by factors associated with white–Hispanic poverty gap (1980–2010)
Full size table

Hispanics as a group have made strides in keeping poverty from rising over the past 30 years, and a majority of Hispanic households enjoy a convergence of advantages with a typical white household. On balance, advantaged Hispanics report lower poverty largely because their characteristics are less associated with the determinants of poverty identified above. Although the weight of the evidence recommends optimism regarding the potential for a falling poverty rate, we also find cause for concern for cleavages within the Hispanic population. Indeed, a segment of Hispanic households face multiple risks associated with especially high poverty rates. As displayed in Table 4 above, a majority of these households were headed by noncitizens since 1990 and a majority also did not speak English very well. Moreover, very few (less than 7% in 2010) of the multiply disadvantaged householders completed any college, and a third are headed by a single parent.

Discussion

We set out to better understand changes in Hispanic poverty over time relative to non-Hispanic whites. We find Hispanics have poverty risks that are closer to whites in 2010 than in 1980. We also find the persistent white–Hispanic poverty gap is, at least in part, attributable to demographic differences. Based on our results, the likelihood of living in poverty is less tied to being Hispanic in 1980 than in 2010, at which point poverty status operates primarily through Hispanic–white disparities in education and other correlates of poverty. Nonetheless, this emphasizes the importance of understanding heterogeneity within the Hispanic population and paying particular attention to those who are most disadvantaged (Tran and Valdez 2017), the same people currently facing extreme hostility and challenges to integrating into US society given the current sociopolitical context.

We set out to understand the following question: are the apparent drivers of differences in poverty by race/ethnicity due to rising disadvantages or a convergence of advantages across groups? We find support for our convergence of advantages hypothesis because of poverty-reducing gains made by both Hispanic- and white-headed households. White poverty remained lower than Hispanic poverty at each point in time due, in part, to white householders’ relative advantage in educational attainment and family size. Hispanic poverty, however, would have been higher if not for gains on both of these poverty-reducing characteristics. Looking ahead, the potential for poverty reduction is greater among Hispanics households because so many such householders have lower education and larger families compared to whites. In 2010, Hispanics reported a higher share of householders with less than a high school education (33% compared to 8% among whites) and three or more children (19% compared to 6% among whites) (analyses not shown). Holding other factors at current levels, if both groups make equal gains in educational attainment (say, a one-third reduction in householders with less than a high school degree) in years to come, poverty rates among Hispanic householders would fall relative to whites.

We find only partial support for the rising disadvantages hypothesis because Hispanic householders’ poverty-enhancing disadvantages (relative to whites) are either in slow decline or offset by other factors. First, relative to whites, Hispanic households remain disadvantaged by limited English language proficiency and citizenship status. Even these relative disadvantages, however, are not currently large enough to offset Hispanic households’ major advantage compared to whites (i.e., more adult workers per household), which kept Hispanic poverty rates from rising through 2010. Whether citizenship status will continue to disadvantage Hispanics depends on changes in immigration policy affecting Hispanic immigrants.

The decomposition results presented rely on counterfactuals; namely, determining how much poverty rates would change if one group enjoyed the compositional advantages (e.g., more and better employment opportunities) of a second group. Of course, although the results are instructive and provide valuable insights into poverty trends along race/ethnicity lines, we cannot determine whether the trends observed through 2010 will hold in the future. Gains in educational attainment could, we suppose, translate into a drop in poverty rates as observed thus far, but we cannot be certain. It is possible the gaps in poverty between whites and Hispanics could remain high if gains in educational attainment become stratified along racial/ethnic lines. For example, even if Hispanics enroll in post-secondary schools in greater numbers, the options available to these groups could be, on average, qualitatively different (and with lower expected earnings) than the options available to whites. In addition to these limitations, the analyses use cross-sectional data. Longitudinal and nationally representative data with large Hispanic samples dating back to 1980, however, do not exist. Moreover, we do not assign legal status among noncitizens, a particularly disadvantaged group. This precludes us from directly analyzing the importance of being undocumented above and beyond the factors we know can disproportionately affect undocumented immigrants (e.g., education, citizenship, language proficiency, and working in an immigrant job). Despite these limitations, this paper presents a range of trends over a long period of time. The approach helps identify how trends in poverty have evolved and suggests explanations for such changes.

The Hispanic population will continue to grow and already comprises a rising proportion of the youth population. Understanding how demographic characteristics are related to poverty, both among Hispanics and in comparison with other groups, can help determine how similar (or different) Hispanic poverty is compared to other groups. For example, it appears Hispanic poverty rates can fall with investments in human capital (educational attainment and English language learning) as well as legalization routes for noncitizens. Whether or not those without legal authorization are offered a path to legal status in the coming years has the potential to profoundly influence Hispanic poverty through reduced work and educational opportunities available, as well as associated stigma and prejudices. Indeed, Hispanic immigrants who have been in the country a long time tend to report lower poverty rates. Orrenius and Zavodny (2013) observe a downward trend in poverty rates among Hispanic immigrants who arrived in earlier time periods, which suggests a road for socioeconomic integration evolving today. Not surprisingly, among immigrants who resided in the country long enough to benefit from amnesty under federal immigration reform enacted in 1986, those who adjusted their legal status reported wage benefits compared to immigrants who remained unauthorized (Kossoudji and Cobb-Clark 2002).

At the same time, research on white–Hispanic segregation trends suggests the potential for persistent and growing inequality between whites and disadvantaged Hispanics (Firebaugh and Farrell 2016; Iceland and Wilkes 2006; Lichter et al. 2012). Based on our results, if Hispanics with low educational attainment, limited English proficiency, and/or lack of US citizenship remain excluded from socioeconomic opportunities, then the white–Hispanic poverty gap may not drop in the foreseeable future. As discussed earlier, Hispanic households reporting multiple poverty-related disadvantages are much more likely to headed by noncitizens (Table 4). Poverty could fall among noncitizen-headed households who can adjust their legal status or otherwise attain the right to work legally. Notably, temporary employment authorization is associated with lower poverty rates among noncitizen beneficiaries of the Obama Administration’s Deferred Action for Childhood Arrivals (DACA) (Amuedo-Dorantes and Antman 2016). Conversely, closing routes to legalization and work permits coupled with ramping up deportations could make poverty even more likely among households headed by Hispanic noncitizens. During the period leading up to the US presidential election of 2016, exposure to immigration enforcement in local communities raised the likelihood of living in poverty among households most likely affected by deportations and police-based enforcement activity (Amuedo-Dorantes et al. 2016). As mass deportations escalate unabated, the prospects of living in poverty among disadvantaged Hispanic households could rise even higher than observed in this study.