Using nationally representative survey data, this research note examines the association between immigrant legal status and poverty in the United States. Our objective is to test whether estimates of this association vary depending on the method used to infer legal status in survey data, focusing on two approaches in particular: (1) inferring legal status using a logical imputation method that ignores the existence of legal-status survey questions (logical approach); and (2) defining legal status based on survey questions about legal status (survey approach). We show that the two methods yield contrasting conclusions. In models using the logical approach, among noncitizens, being a legal permanent resident (LPR) is counterintuitively associated with a significantly greater net probability of being below the poverty line compared with their noncitizen peers without LPR status. Conversely, using the survey approach to measure legal status, LPR status is associated with a lower net probability of living in poverty, which is in line with a growing body of qualitative and small-sample evidence. Consistent with simulation experiments carried out by Van Hook et al. (2015), the findings call for a more cautious approach to interpreting research results based on legal status imputations and for greater attention to potential biases introduced by various methodological approaches to inferring individuals’ legal status in survey data. Consequently, the approach used for measuring legal status has important implications for future research on immigration and legal status.
The noncitizen population in the United States comprises immigrants with diverse legal statuses. Although the unauthorized immigrant population receives substantial attention in political discourse and the media, lawful permanent residents (LPRs) outnumbered unauthorized immigrants by 1.5 million as of 2016 (Passel and Cohn 2018). The legal residency status of green card1 holders confers rights not guaranteed to other noncitizen immigrants, including work authorization, property ownership rights, eligibility for public assistance, financial assistance for tertiary education, and a pathway to citizenship (Department of Homeland Security 2019). Legal residency status facilitates access to invaluable public resources and opportunities for social mobility frequently denied to other noncitizen—primarily unauthorized—immigrants (Yoshikawa 2011), which subsequently affects intergenerational socioeconomic outcomes (Bean et al. 2011; Gonzales 2016; Prentice et al. 2005; Yoshikawa 2011). Presently, quantifying the relationship between legal status and socioeconomic outcomes is complicated by limited availability of nationally representative data, including measures of both citizenship and (especially) legal status. Surveys that measure citizenship status, but not legal status of noncitizens, prohibit the disaggregation of the noncitizen population into LPR and non-LPR components.
To address the limitation in federal survey data, Jeffrey Passel, a leading demographer on legal-status population estimation, developed an innovative logical imputation method (hereafter, logical approach) to assign legal status at the individual level (Passel and Clark 1998; Passel and Cohn 2014; Passel et al. 2004). Although specific details of Passel and Cohn’s (2014) algorithm are unclear, the logical approach has facilitated a recent surge of empirical research with direct relevance to national immigration policy discussions. Researchers have increasingly employed variations of the logical approach to overcome survey data limitations and assign legal status to immigrants in microdata (Borjas and Slusky 2018; Bustamante et al. 2014; Cohen and Schpero 2018; Gunadi 2019; Pourat et al. 2014; Zuckerman et al. 2017).
However, to our knowledge, the logical approach has not been evaluated against survey-based values of legal status (hereafter, survey approach). Little is therefore known about the extent to which estimated effects of legal status on a given outcome vary across micro-imputation approaches. We address this gap by estimating whether the association between legal status and poverty depends on whether legal status is imputed using a logical approach versus assignment inferred from survey items.
Although methodological decisions related to legal status assignment may affect any number of substantive outcomes, we examine poverty for several reasons. First, it is a widely used summary measure of immigrant integration (Altman et al. in press; National Academies of Sciences, Engineering, and Medicine 2015). It is an important indicator of integration that also serves as a barometer for the likely socioeconomic position of the children of immigrants (Lichter et al. 2015; Van Hook et al. 2004). Second, previous research suggests that legal status impacts the chances of being in poverty among immigrant families in the United States. Specifically, we expect that compared with immigrants without official permanent residency status, lawfully present immigrants will have lower odds of being in poverty. If conclusions about the relationship between legal status and poverty differ across imputation approaches, our findings will have strong implications for past and future research that assigns legal status to immigrants in survey microdata.
Data and Methods
We use data from the 2008 Survey of Income and Program Participation (SIPP), a nationally representative household-based survey conducted by the U.S. Census Bureau. Our analyses include foreign-born respondents ages 18 and older who reported not being naturalized citizens; that is, U.S.-born and naturalized immigrants are excluded from our analyses. Unlike other census surveys, the SIPP asks nonnaturalized immigrants whether they arrived in the United States as an LPR and, if not, whether they have since adjusted to LPR status. This series of questions allows us to use self-reported immigration status to distinguish LPRs from a group likely comprising mostly unauthorized immigrants and a small proportion of legal nonimmigrants on temporary visas.2 Some studies have used the term “unauthorized” to refer to all nonnaturalized, non-LPRs, but we find this categorization imprecise and henceforth refer to this group as nonLPR.
To create an indicator of LPR status derived from a logical approach, we use an approach recently employed by Borjas (2017) and Borjas and Cassidy (2019). Borjas (2017) developed his method by reverse engineering the assignment method used by Pew Research Center (Passel and Cohn 2014). We choose the method published by Borjas (2017) because its clear articulation allows us to generate the same indicators and replicate the procedure in the SIPP. Following Borjas (2017), we assign LPR status to any sample individual who meets one or more of the following criteria3:
Arrived before 1980
Is a citizen
Receives Social Security benefits, Supplemental Security Income (SSI), Medicaid, Medicare, or military insurance
Is a veteran or is currently in the Armed Forces
Works in the government sector
Resides in public housing or receives rental subsides, or is a spouse of someone who resides in public housing or receives rental subsidies
Was born in Cuba4
Is in an occupation requiring some form of licensing (such as physician, registered nurse, air traffic controller, and lawyer)
Is married to a legal immigrant or citizen5
We estimate two logistic regression models predicting poverty status, measured dichotomously indicating whether a respondent’s total family income is less than the federal poverty line. The first model includes an LPR status indicator derived from the logical approach ignoring SIPP’s reports of immigration status. The second model uses survey approach indicators of LPR status. Both models include controls for sex, age, race (Black/non-Black), Hispanic ethnicity, world region of birth, duration of U.S. residence, limited English proficiency, education, employment status, insurance coverage, homeownership, marital status, region of residence, and metropolitan status. We then generate predicted probabilities of being in poverty by legal status for each model separately and graph the results.
Table 1 lists the criteria used in the logical imputation method, showing the relative influence of each criterion in the identification of LPRs. Receiving public assistance and being married to a legal immigrant or citizen are the most influential in identifying LPRs: 42% of LPRs identified receive public assistance (LPRs Identified columns), and 21% receive public assistance and meet no other criteria (Unique Contribution columns). More than 60% have a legal or citizen spouse, and 36% are identified as LPR only because they have a legal/citizen spouse.
The Matches Survey Values column in Table 1 shows for each logical criterion the percentage of logical LPRs identified whose survey value also identifies them as an LPR. If we take the survey responses for legal status as the “true” values, then the percentages in this column can be interpreted as the accuracy of a given criterion and any deviation from 100% should be considered error. The percentages range from 71% to 88%, with the Cuban identifier as the most accurate and the public housing criterion having the greatest error. Although meeting any given criterion suggests a relatively high likelihood of being identified as an LPR using the survey approach, the logical approach adopted here and used elsewhere implicitly assumes zero measurement error with respect to indicators. The implication of this assumption is that any given indicator of LPR status (e.g., public health coverage) is not misreported or allocated incorrectly and thus that all noncitizens coded to have a given characteristic must be an LPR. Table 1 demonstrates that this strict assumption does not hold for any of the criteria used by the logical approach.
Table 2 shows estimated characteristics of the LPR population by legal status assignment method. Compared with the survey approach, the logical approach produces a profile of the LPR population that is slightly older, a higher proportion female, a higher percentage Hispanic, and more likely to be married relative to the survey approach. Most notably, LPRs identified by the logical approach are significantly more likely to have incomes below the federal poverty level and to be unemployed or not in the labor force relative to survey-based LPRs.
In Table 3, we show results from two logistic regression models estimating the log odds of being in poverty. A separate adjusted logistic regression model was specified for each LPR identification method confirming the bivariate relationships between LPR status and poverty. Using the logical approach to assign legal status, the model suggests that LPRs have higher odds of being in poverty relative to non-LPRs (log(θ) = 0.43, p < .001) when all other variables in the model are held constant. The coefficient for LPR status derived from the survey approach is negative (log(θ) = –0.16, p < .10), indicating that LPRs have lower odds of poverty relative to non-LPRs, although the coefficient is only marginally significant. The associations between poverty and other demographic and socioeconomic variables appear to be less sensitive to legal status imputation approach: nearly all other coefficients of predictors differ only marginally in magnitude, but not direction, between the two models.
To more clearly illustrate how the relationship between LPR status and poverty differs across legal status imputation approach, we present predicted probabilities of being in poverty for LPRs and non-LPRs in Fig. 1, generated from the models represented in Table 3. All covariates are held at mean values. The predicted probability of being in poverty for LPRs identified by the logical approach is .27, compared with a probability of .19 for those identified as non-LPRs using the same method. In contrast, survey LPRs are less likely than non-LPRs to be in poverty, with a probability of .22 compared with .24, respectively.
Legal status methodological decisions have significant impacts on conclusions about the association between legal status and socioeconomic outcomes. Using a survey identification approach, we found that among noncitizens, LPRs were less likely to be in poverty than non-LPRs. In contrast, when we assigned LPR status to noncitizens using a logical approach, LPR status was positively associated with living below the poverty line. These findings appear to be robust across survey year. To test the sensitivity of our results, we estimated the multivariate analysis using the 2004 SIPP and observed relationships in the same direction and of comparable magnitude for each legal status assignment approach.
Overall, the finding from the logical approach runs counter to evidence provided by field studies suggesting that legal residency standing affords immigrants access to public benefits and other mobility-generating opportunities that place them at a much lower risk of economic hardship relative to unauthorized immigrants (Gleeson and Gonzales 2012; Gonzales 2016; Yoshikawa 2011).
Although some criteria used to identify LPRs in the logical approach were accurate in identifying certain survey LPRs—such as those who are married or who receive public assistance—they also appear to systematically overlook other characteristics of LPRs. In this case, many high-income legal immigrants were missed. Additionally, the strong influence of the public assistance criterion in the logical imputation method’s algorithm produced an LPR population that was disproportionately low-income and impoverished.
These findings raise significant concerns over the validity of knowledge produced using logical approaches to estimate legal status in federal surveys. Although we acknowledge the need for methods to estimate legal status in microlevel data, any such method should be validated and checked against survey reports or administrative data whenever possible. We also urge caution against reliance on deterministic assumptions and instead suggest that future research examine the viability of probabilistic approaches that incorporate a random element into status assignments.
The authors thank the fellows of the Public Policy Lab of the College of Liberal Arts at Temple University for helpful comments and suggestions. Spence also acknowledges funding through a fellowship with the Public Policy Lab.
All authors contributed to the study conception and design. Cody Spence performed all data management and statistical analysis tasks and wrote all portions of the manuscript. James Bachmeier wrote some portions of the manuscript, reviewed and edited all portions of the manuscript, and contributed to the development of methods and interpretation of data. Claire Altman and Christal Hamilton reviewed and edited all portions of the manuscript.
The data set used is publicly available at https://www.census.gov/programs-surveys/sipp/data/datasets.2008.html. Analysis code is available upon reasonable request from corresponding author.
Compliance With Ethical Standards
Conflict of Interest
The authors declare that they have no conflict of interest.
Ethics and Consent
This article does not contain any studies with human participants or animals performed by any of the authors.
The lawful permanent residence (LPR) visa is often referred to as a “green card.” To minimize redundancy, we use “LPR” and “green card” interchangeably.
Individuals who failed to respond to certain questions were allocated values by the Census Bureau, as is customary in most publicly released census microdata. We treat these values the same as self-reports.
We omit criterion b—“Is a citizen”—from our logical imputation because our sample is already limited to noncitizen immigrants. Our focus is on lawful permanent residents and noncitizen, non-LPRs.
Because of the absence of national origin information in the 2008 public-use SIPP, we are unable to identify Cuban nationals. We instead use a proxy that identifies individuals born in the Caribbean who either speak Spanish at home or identify as Hispanic.
If someone’s spouse is identified as legal using any of the criteria in this list, then that individual is also coded as legal.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.