The March issue of Race and Social Problems published a paper analyzing poverty trends: "Convergence and Disadvantage in Poverty Trends (1980–2010): What is Driving the Relative Socioeconomic Position of Hispanics and Whites?"
I co-authored the paper with Beth Mattingly (Ph.D., Sociology, University of Maryland, College Park). Beth is Director of Research on Vulnerable Families (Carsey School of Public Policy, University of New Hampshire) and research consultant at Stanford University's Center on Poverty and Inequality. We have been collaborating since I began graduate school on Hispanic poverty trends, differences in poverty by race/ethnicity and place, and gender disparities in safety net services.
We start with the fact that the gap between White and Hispanic poverty has remained stable for decades even as these groups' compositions changed. Using Census and American Community Survey (ACS) data, we examine a cross-section of trends (1980, 1990, 2000, 2010) and ask:
How much do key 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?
Decomposition of Poverty Rates
After reviewing sociological, demographic, and economics research on racial/ethnic poverty trends, we examine household poverty as a function of the usual suspects (i.e., individual, household, and contextual). We conducted analyses separately for Hispanics and for whites.
First, we look at differences in the composition of both groups (i.e., the characteristics, or Xs).
Second, we apply the characteristics of the group with generally lower rates of poverty (whites) to the second group (Hispanics) to determine what Hispanic poverty would be if the Hispanic population shared the same characteristics as the non-Hispanic, white population.
We report what Hispanic would have been at each time period if the distribution of Xs were different but the returns to those Xs (i.e., the coefficients) remained the same. That is, we assume the Hispanic population had the same returns as the data suggest and simply varied the characteristics. This decomposition approach is commonly used to measure disparities between groups. Note that we could have also focused on what Hispanic poverty would have been if Hispanic returns to, say, education resembled those of white householders’ investment in formal education. In the paper, we discuss the implications of a more realistic counterfactual:
What would hypothetically happen to Hispanic poverty rates if Hispanics enjoyed the relative advantages (e.g., higher educational attainment) and disadvantages (i.e., fewer adult workers per household) of whites?
We conduct all analyses using parallel decomposition techniques and focus on results shared by all three approaches below:
Blinder-Oaxaca decomposition method, which is common in poverty research (see Orrenius & Zavodny (2011) for an earlier analysis of white-Hispanic poverty)
An extension of the Blinder-Oaxaca technique by Robert Fairlie for probit and logit models
a new KHB decomposition method for non-linear outcomes
We report Fairlie decomposition results because the method has a number of unique features:
"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" [Mattingly & Pedroza, 2018, page 58].
We find a convergence of advantages between Hispanics and whites; largely because both groups report, for example, higher levels of educational attainment and smaller family size over time.
We also find that Hispanic households are more often headed by people who are not U.S. citizens and who report limited English language proficiency, contributing to a persistent concentration of disadvantage.
Let’s dive a bit deeper into the decomposition results. For simplicity, the figure below is based on decompositions that account for the top 6 categories related to the Hispanic-white gap. The paper relies on the full range of correlates of poverty, and the results are mostly the same. In other words, if you can only measure a handful of differences between these two groups, you can capture nearly all of the relevant variation by focusing on education, language proficiency, citizenship status, family size, presence of young children, and (most recently) single status.
Here is what we found:
Source: Authors' analyses using the Fairlie decomposition approach applied to Hispanic and white heads of household (age 25 and over, not in group quarters) in IPUMS data: Ruggles, S., Genadek, K., Goeken, R., Grover, J., & Sobek, M. (2015). Integrated public use microdata series: Version 6.0 [Machinereadable database]. Minneapolis, MN: University of Minnesota.
the Hispanic-white poverty gap was similar at each time period (11.6-13.5%)
We can explain more of the poverty gap in 2010 (8.3 of the 11.6% gap, which is 71% of the gap) than 1980 (6.7 of the 12.4% gap, or 54% of the gap)
Extending U.S. citizenship to Hispanics on par with whites would close the poverty gap: we could expect as much as 11% of the gap to narrow in 2010, assuming such new citizens experienced the same benefits to citizenship as current citizens.
To put these results in perspective, Hispanic households reporting 2 or fewer of the 6 factors above report relatively low poverty rates (between 10% and 13%) at each time period. By contrast, approximately one-third of Hispanic households with 3 or more of these factors reported living in poverty. For detailed results, see Table 4, page 62 in Mattingly & Pedroza (2018).
Decomposition techniques assume we can observe what would hypothetically happen to an outcome if one group's characteristics were swapped with those of a second group. But is it realistic to assume current trends in higher education enrollment will inevitably translate into poverty-reducing gains? Maybe not: “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” [Mattingly & Pedroza, 2018, page 63].
Here, we are revisiting an old dilemma. In 2005, Doug Masssey quotes what Stanley Lieberson observed 20 years earlier about the lengths an advantaged group might go to maintain high status (in this case, during the hunt for housing):
"...a complicated causal analysis of factors contributing to the racial gaps in income has not the causal value one might have assumed. It describes the given set of events at a given time; it describes what a black person might well follow as a get-ahead strategy if he or she can assume that not many other blacks will follow the same strategy and hence the basic [social] matrix will remain unaltered. But there is no assurance that this matrix will continue to operate—indeed, there is virtual certainty that the matrix will not continue to operate if some superficial factor that appears to cause the income gap is no longer relevant (for example, if the groups end up with the same educational distribution). In which case, new rules and regulations will operate; the other regression coefficients will change in value in order to maintain the existing system. (pp. 191–92)" [quoted in Massey, 2005, page 148]
Massey then restates: "Lieberson’s pessimistic analysis suggests that the problem of racism is not likely to be 'solved' easily or quickly by passing a few reforms and calling it a day. Racial discrimination is a moving target. One cannot simply ban prevailing discriminatory practices and declare the struggle for racial equality won" [pages 148-49].
It is too early to tell whether a moving target might contribute to persistently high Hispanic poverty.
In my research of existing decomposition methods, I read journal articles and technical documentation authored by a range of scholars. It wasn't until May 2018 (via an eye-opening exchange on twitter) that I learned about Evelyn Kitigawa's groundbreaking work on decomposition. We owe Evelyn credit for her pioneering work as a sociologist and demographer. In her own words, here is where we stood 62 years ago:
"As yet, very little attention has been directed to the problem of formalizing the analysis of standardized rates, and of systematically explaining which factors account for the differences between standardized rates in comparison with corresponding differences between their unstandardized rates. If standardization alters a difference between two total rates, it should be possible to measure the amount of change, and to break it up into components attributable to the various factors for which the data were standardized" [Kitigawa, 1955, page 1169].
And what about the poverty gap between Hispanics and whites then? In 1960, the gap was 15% for householders age 25 and older, only slightly higher than it's been for decades.
Acknowledgements: This research was partially supported by a Poverty Center Grant awarded to the Center on Poverty and Inequality at Stanford University (Grant Number AE00101) from the U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation, and awarded by Substance Abuse Mental Health Service Administration and a sub-award (Grant Number H79 AE000101-02S1) titled "Poverty, Inequality, and Mobility among Hispanic Populations: An Innovative Subgrant Research Program at the Stanford Center on Poverty and Inequality” funded by ACF’s Office of Planning, Research and Evaluation. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation, or Office of Planning, Research and Evaluation. We acknowledge valuable feedback from David Grusky, Doug Massey, Van Tran, Jayanti Owens, panel session attendees at the 2014 annual meeting of the Population Association of America, and workshop participants at Stanford University (at the Center on Poverty and Inequality; and at the Migration, Ethnicity, Race, and Nation graduate research workshop) . We appreciate their suggestions and feedback, but any errors or omissions remain our own.