House Poor in Los Angeles: Examining Patterns of Housing-Induced Poverty by Race, Nativity, and Legal Status

Abstract

Housing affordability in the United States is generally operationalized using the ratio approach, with those allocating more than thirty percent of income to shelter costs considered to have housing affordability challenges. Alternative standards have been developed that focus on residual income, whether income remaining after housing expenditures is sufficient to meet non-housing needs.

This study employs Los Angeles Family and Neighborhood Survey data to consider racial/ethnic, nativity and legal status differences in one residual income standard. Logistic regression analyses of housing-induced poverty focus on whether there are differences among five distinct groups: U.S.born Latinos, Non-Hispanic Whites, and African Americans, authorized Latino immigrants, and unauthorized Latino immigrants. Results suggest that: 1) Latino natives are significantly more likely to be in housing-induced poverty than African Americans and Latino immigrants, and 2) unauthorized Latino immigrants are not more likely to experience the outcome than other groups.

The present work extends previous research. First, the results provide additional evidence of the value of operationalizing housing affordability using a residual income standard. Alternatives to the ratio approach deserve more empirical attention from a wider range of scholars and policymakers interested in housing affordability. Second, housing scholarship to date generally differentiates among Latinos by ethnicity, nativity, and citizenship. The present study contributes to emerging research investigating heterogeneity among Latinos by nativity and legal status.

Housing costs are the single largest expenditure for U.S. households (Bureau of Labor Statistics 2009). The generally accepted indicator of housing affordability relies on a ratio of housing costs to income; in recent decades, those spending more than thirty percent of income on shelter are considered to have housing affordability problems (e.g., Combs, Combs et al. 1994; Stone 2006a; Jewkes and Delgadillo 2010). For many households, allocating a higher proportion of income to housing costs constrains other necessary expenditures (e.g., Brennan and Lipman 2008; Lipman 2005). A substantial proportion of U.S. households experience housing affordability challenges according to this standard: 29 percent of all U.S. households in 2001 and 36 percent of U.S. households by 2008 (Joint Center for Housing Studies 2010).

The ratio approach to measuring housing affordability has achieved widespread acceptance among policymakers, housing professionals, and scholars in the United States (e.g., Elmelech 2004; Joint Center for Housing Studies, 2010; Brennan and Lipman 2008).¹ However, there are powerful critiques of using a thirty percent rule of thumb as a housing affordability standard. For example, Stone (1993; 2006a; 2006b) notes that the thirty-percent cutoff is arbitrary and, more significantly, does not accurately represent whether households can afford other goods and services after paying for housing. Critiques of the ratio approach have led researchers in the United States and elsewhere to explore other ways to operationalize housing affordability (e.g., Stone 1993, 2006a; Kutty 2005; Haffner and Heylen 2011; Thalmann 2003).

The residual income approach, the most commonly employed alternative to the ratio approach, focuses on the income remaining after housing expenditures. Residual income standards provide a sliding scale of housing affordability that identifies whether households have sufficient income to meet non-housing needs (e.g., Kutty 2005; Stone 1993; 2006a). Using residual income, then, addresses a key problem with the ratio approach: some households earn such low incomes that they cannot afford to allocate even ten percent of their income to shelter costs, much less thirty percent (Stone 2006a). Although the residual income approach to housing affordability offers important advantages over the ratio approach, the former is “neither well known, particularly in this country, nor widely understood, let alone accepted” (Stone 2006a 179).

The present study focuses on housing-induced poverty, a residual income standard developed by Kutty (2005) that identifies whether households are below federally-defined income thresholds after housing expenditures. Housing-induced poverty offers insight about the challenges faced by many lower-income households. Indeed, experiencing a poverty standard of living after shelter costs suggests an involuntary response to low incomes, high housing costs, or both. With the ratio approach, it is less clear whether the allocation of a higher proportion of income to housing is involuntary or a deliberate decision to access higher-quality housing, a better neighborhood, or other reasons. Further, although there are significant critiques regarding how the federal government defines poverty and calculates income thresholds, described later in the paper, official poverty measures are influential and widely-used metrics. For example, federal definitions of poverty are used to generate official statistics about poverty in the U.S. (Fischer 1997) and to determine eligibility for assistance from federal programs serving millions of people, such as the Supplemental Nutrition Assistance Program (U.S. Department of Agriculture 2011).²

Drawing from data collected in Los Angeles County, this is the first residual income study that investigates racial/ethnic differences among native-born Latinos, African Americans and Non-Hispanic Whites and nativity and nativity and legal status differences among Latinos.³ Although Latinos are heterogeneous along many dimensions, a focus on nativity is relevant, given that Latin American immigrants comprise more than half of all immigrants in the U.S. in recent decades (Dockterman 2011). Contrasting Latino immigrants by legal authorization to reside in the United States is also appropriate. In the contemporary social, economic, and political context, possessing legal status appears to be an increasingly significant factor in American life for immigrants. For instance, local and state-level legislation increasingly focuses on unauthorized immigrants (Chavez and Provine 2009), with implications for the incomes that they are able to earn in the United States and their access to housing (e.g., Oliveri 2009, Bender 2010). Indeed, a growing body of research emphasizes how legal status shapes the experiences of contemporary Mexican and other Latin American immigrants in the United States (e.g., Abrego 2006; Menjívar 2006; Massey and Sánchez 2010; Gonzalez 2001). Taken together, the study provides important new information about housing affordability challenges for the three largest native-born groups in the United States, investigates differences among Latinos by nativity and legal status, and offers more insight about the factors linked with a residual income outcome like housing-induced poverty.

Literature Review

The Residual income approach to housing affordability

Scholars have developed diverse measures based on the residual income approach to housing affordability.⁴ Several employ federal poverty figures (Combs, Combs et al. 1994; Combs and Park 1994; Kutty 2005; McConnell 2006). For example, households experience housing-induced poverty, the focus of the present study, when income remaining after housing costs is less than two-thirds of the appropriate federal poverty threshold (Kutty 2005). A similar outcome, “housing poverty,” refers to those with residual income less than three-quarters of the federal poverty guidelines (Combs and Park 1994).⁵ The advantages of these operationalizations include the accessibility of official poverty thresholds or guidelines that are released each year by the federal government, the ease of calculation, and the straightforward interpretation. However, there are numerous critiques of federal poverty definitions, including the methods by which the income levels are calculated and the lack of geographic adjustment for differences in cost of living (e.g., Boushey et al 2001; Citro and Michael 1995; Pearce 2009; Reed 2006; Ruggles 1990; Stone 1993). For instance, when the thresholds were first developed in 1963, they were set at three times a minimum food budget; these calculations do not capture child care expenditures or other significant costs (e.g., Ruggles 1990; Citro and Michael 1995).⁶ A specific critique of Kutty’s (2005) concept is the use of two-thirds of the federal poverty threshold as the cut-off point for housing-induced poverty (Stone 2006a).

Other residual income measures are based on budgets data. One prominent residual income standard is “shelter poverty,” developed by Stone (e.g., 1993, 2006a, 2006b). Shelter poverty is defined as paying more for housing than the maximum shelter amount that a household can afford, using 1981 Bureau of Labor Statistics (BLS) Lower budgets that have been adjusted for household characteristics, are updated using the Consumer Price Index, and account for personal taxes and tax credits (Stone 1993, Appendix A). Issues noted for BLS budgets data, even when updated, are that they rely on the direct pricing of budget items in 1969 (Kutty 2005), and the best method to adjust BLS figures for different household compositions and sizes (Johnson, Rogers et al 2001).

Alternative budgets have been developed, in response to critiques of federal poverty definitions and, perhaps, because of “the perceived obsolescence” of BLS budgets (Pelletiere 2008 11). Although such budgets were not explicitly developed as residual income measures of housing affordability, subtracting housing costs from these budgets provides additional estimates of non-housing income needed to meet basic needs. The Economic Policy Institute’s (EPI) basic family budgets identify the income necessary for a “safe and decent standard of living” for diverse family types in many U.S. cities using geographically-specific costs of housing, food, childcare, taxes, and other items (Boushey et al 2001 15). Pearce (2009) developed the Self Sufficiency Standard (SSS) in the mid-1990s and has provided regular updates; the SSS estimates the after-tax income required to support working-age, non-elderly, non-disabled households without public or private assistance. Like EPI budgets, the SSS uses information about the number and age of minor children to explicitly account for child-care costs.⁷

Table 1 provides illustrative examples of non-housing income amounts calculated when using the diverse concepts described in this section, for a four-person household with two minor children in Los Angeles.⁸ The household experiences housing-induced poverty (Kutty 2005) if residual income is less than $11,630 and housing poverty (Combs and Park 1994) if residual income is less than $12,786 (first and second columns, Table 1). As expected, using other data sources provide substantially higher estimates of non-housing income needed to support a household. For instance, the sample household experiences shelter poverty (Stone 1993) if it spends more than $6,033 on housing, leaving non-housing income of $20,540 or less. EPI and SSS figures estimate higher non-housing income thresholds to meet basic needs, particularly when child care costs are explicitly included in the budget (fourth and sixth columns, Table 1). As these examples suggest, there are diverse ways to operationalize residual income standards of housing affordability. Such alternatives to the ratio standard of housing affordability deserve more empirical attention from a wider range of scholars and policymakers interested in housing affordability and housing outcomes generally. The present investigation of the relationship between housing-induced poverty and race, nativity and legal status contributes to this effort.

Table 1.

Residual Income for a four-person household in Los Angeles: Two adults and two children under 18

Housing-Induced Poverty, United States, 2000a	Housing Poverty, United States, 2000b	Shelter Poverty, Los Angeles-Long Beach MSA, 2001c	Economic Policy Institute Basic Family Budgets, Los Angeles-Long Beach MSA,1999d	Self-Sufficiency Standard, Los Angeles County, 2000e
$11,630	$12,786	$20,540	$29,792	$23,200 (teens)	$43,408 (infants)

^aThe 2000 federal poverty threshold for a family unit of this size and composition is $17,463 (Dalaker 2001), housing-induced poverty (Kutty 2005) is non-housing income that is less than two/thirds of the federal poverty threshold.

^bThe 2000 federal poverty guideline for a family unit of this size and composition is $17,050 (Health and Human Services 2010), housing poverty (Combs and Park 1994) is non-housing income that is less than three-quarters of the federal poverty guidelines.

^cResidual income for a family of four is derived from BLS Lower Budget Estimates data excluding personal taxes for a four-person household in the Los Angeles-Long Beach metropolitan statistical area in Autumn 2001. The estimated maximum allowable cost for shelter is $6,033. Provided by Professor Michael E. Stone (6/17/2010, personal communication).

^dEPI family budgets data for the Los Angeles Los Angeles-Long Beach MSA in 1999 were compiled by the staff of the Economic Policy Institute and provided by Dr. Sylvia Allegretto (6/16/2010, personal communication). Residual income is annual income of $38,780 minus estimated housing costs.

^eSelf-Sufficiency Standard figures for Los Angeles County in 2000 were developed by Dr. Diana Pearce, Center for Women’s Welfare, University of Washington School of Social Work and provided by Lisa Manzer (6/10/2010, personal communication). Residual income is annual income minus housing costs, estimated at $32,392 for a family/household with two adults and two teens (requiring no childcare expenditures) and $52,600 for a household with two adults and two infants requiring childcare.

Multivariate Analyses of Residual Income

To date, only a few U.S.-based studies have conducted multivariate analyses of outcomes derived using the residual income approach (e.g., Combs, Combs et al. 1994; Combs and Park 1994; Kutty 2005; McConnell 2006). Of these, only one includes specific race indicators in a multivariate framework (McConnell 2006).⁹ That study finds that near-poor households headed by Whites are more likely to experience housing-induced poverty than Latinos or African Americans, controlling for tenure status, the age of the household head, and other factors (McConnell 2006). In contrast, studies that use less detailed race indicators report no differences between Whites and non-Whites (Combs and Park 1994; Combs, Combs et al. 1994). For instance, Combs and Park (1994) find that elderly female household heads who are members of a “minority” group are as likely as their White rural female counterparts to experience “housing poverty,” controlling for income, homeownership, and other factors.

The present study focuses on whether there are initial racial disparities and residual racial differences between groups after accounting for a comprehensive set of individual, household, and neighborhood-level variables linked with housing affordability. The results are expected to show that though there may be baseline differences in the incidence of housing-induced poverty between native Whites, Latinos, and Blacks; controlling variation in individual, household, and contextual characteristics will fully explain initial racial/ethnic disparities. This hypothesis arises because the study explicitly compares U.S. born members of all three groups in one geographic area, which may reduce disparities that might otherwise be found between groups that include both immigrants and natives or in a national sample. Moreover, the analyses control for an extensive set of factors identified in previous housing affordability scholarship, described at the end of this section. As many of these characteristics vary between native Whites, Blacks, and Latinos, controlling for these differences may eliminate potential residual differences in the likelihood of housing-induced poverty.

Studies of residual income outcomes in the U.S. have not explicitly considered differences by nativity. However, previous work using the ratio approach of housing costs to income generally report that immigrants tend to have higher housing costs or spend a larger proportion of their income on housing than natives (e.g., Capps, Ku et al. 2002; Elmelech 2004; Krivo 1995; Lipman 2003; McArdle and Mikelson 1994; Schill, Friedman et al. 1998). Latino immigrants in the present study are hypothesized to be more likely to experience housing-induced poverty than U.S. born Latinos, as well. However, as the multivariate analyses control for differences in a range of factors of relevance to immigrants, residual disparities among Latinos by nativity are expected to be smaller than when such differences are not controlled.

To date, published work has not examined disparities in housing affordability based on immigrants’ authorization to live and work in the United States.¹⁰ Housing research generally shows that naturalized citizen immigrants in the United States tend to have more favorable housing outcomes than non-citizen immigrants (e.g., Coulson 1999; Krivo and Kaufman 2004; Toussaint-Comeau and Rhine 2004). In the present study, unauthorized Latino immigrants are expected to have a higher incidence of housing-induced poverty than their authorized Latino immigrant counterparts. This hypothesis stems from the more limited economic resources and constrained housing options of undocumented immigrants compared to immigrants with legal status. However, in line with expected patterns by race and nativity, controlling for other differences in characteristics will likely reduce baseline gaps in housing-induced poverty between authorized and unauthorized Latino immigrants.

The present analyses incorporate a range of established in individual, household and neighborhood-level predictors of housing-induced poverty and other housing affordability standards. For example, Kutty’s (2005) regression analyses of 1999 American Housing Survey data finds that among near poor households, being older, and owning a home are linked with a lower likelihood of housing-induced poverty than younger individuals or renters. In contrast, those with no children have a higher probability of the outcome than those with one to three children; those with four or more children had a lower probability (Kutty 2005). The current study includes indicators tapping into age of respondent, number of children, and homeownership. Although these variables are not explicitly tested in previous residual income analyses cited earlier, research focusing on the ratio approach documents that persons who are married or have more education allocate a lower proportion of their income on housing than their unmarried or lesser-educated counterparts (e.g., Devaney, Chiremba et. al 2004; Elmelech 2004). Marital status and years of education are included in the present study to consider whether these factors help predict housing-induced poverty. Turning to established predictors of immigrant outcomes, prior work shows that as immigrants gain more U.S. experience and become more integrated in the United States, their demographic and housing profiles approach those of the native born (e.g., Alba and Logan 1992; Alba and Nee 2003; Myers, Painter et al. 2005; Painter, Gabriel et al. 2001; Rosenbaum and Friedman 2007). For this reason, immigrants who have more U.S. experience are hypothesized to be less likely to experience housing-induced poverty than their less-experienced immigrant peers.

Contextual features related to housing, such as urban/rural location, region of the country, or immigration context, also are linked with housing affordability and other housing outcomes (e.g., Combs, Combs et al., 1994; Combs and Park 1994; Krivo and Kaufman 2004; Kutty 2005; Myers, Painter et al. 2005; Painter, Gabriel et al. 2001; Rosenbaum and Friedman 2007). As the present work focuses on Los Angeles County, contextual characteristics tap into neighborhood-level economic and immigrant characteristics. More specifically, those in neighborhoods with higher median home prices may be less likely to be in housing-induced poverty, as they need higher incomes to afford more expensive housing. In contrast, those living in areas with higher concentrations of recent immigrants may be more likely to experience the outcome, as persons may choose such neighborhoods explicitly because of economic constraints.

Other factors may be linked with a residual income standard like housing-induced poverty, but have not yet been examined with a sample of natives and immigrants. Participation in mainstream financial institutions is one such factor. Whites are more likely to have a checking or savings account than Latinos and African Americans (e.g., Federal Deposit Insurance Corporation (FDIC) 2009; Hogarth, Anguelov, et al. 2005) and natives are more likely to “banked” than immigrants (Osili and Paulson 2004). Research with new legal immigrant homeowners suggests that those without bank accounts are more likely to spend more than thirty percent of income on housing than those who are banked (McConnell and Akresh 2010). Respondents who are banked are expected to have a lower incidence of the outcome than those without such financial access. This is because the former may have access to more credit, more favorable loan terms or rental contracts, and more information that can lower the cost of housing and stretch available economic resources than those without bank accounts.

Whether the respondent lives only with immediate family versus more extended living arrangements is expected to be connected with the outcome. Previous work shows that living arrangements vary by race, ethnicity, nativity, and duration in the U.S. (e.g., Glick 2000; Lara-Cinisomo and Griffin 2007). Extended living arrangements are due, in part, to economic constraints such as low incomes and high housing costs. Indeed, other work suggests that immigrant families who have difficulty paying for housing move in with others as a coping strategy (Capps, Ku et al 2002). Thus, those living with immediate relatives only (spouse/partner and/or children) are expected to be less likely to be in housing-induced poverty than those with additional household members. The multivariate analyses undertaken here offer the first test of whether type of living arrangements is linked with a residual income approach to housing affordability.

Finally, three other lesser-studied factors are expected to predict housing-induced poverty. Housing unit size is one, as some respondents may willingly reduce available non-housing income to access larger housing. Those residing in larger units are expected to have higher probabilities of the outcome than those in smaller units. The number of earners in the family is also likely connected with the outcome. Respondents with more family members earning income via employment likely have lesser probabilities of housing affordability problems; as they have more economic resources to pay for housing and other expenses. Such families are expected to be less likely to be in housing-induced poverty than those with fewer earners. Finally, whether the respondent has moved within the previous year, may address otherwise unmeasured differences related to economic instability associated with the outcome.

Data and Methods

The present study conducts multivariate analyses of the incidence of housing-induced poverty, using data from the Los Angeles Family and Neighborhood Survey (L.A.FANS), cross-sectional data collected in Los Angeles County between April 2000 and January 2002 from about 3,000 households in Los Angeles County (Sastry and Pebley 2003). Poor and very poor census tracts, households with children, and Latinos were oversampled. Approximately 40 randomly selected households completed the survey in each of 65 census tracts, used to represent neighborhoods. Personal interviews were conducted with respondents in English and Spanish, depending on the language preferred by the respondent. L.A.FANS respondents were assured of the confidentiality of their responses and that their names and addresses would not be made available to non-L.A.FANS staff (Pebley and Sastry 2004); these safeguards protect human subjects and encourage respondents to answer sensitive questions more honestly. These data are particularly valuable because the Latino immigrant sample can be stratified by legal status, an advantage over other data sources that cannot be used to consider the role of legal authorization in shaping immigrant outcomes. L.A.FANS data are generally representative of Los Angeles (Goldman, Smith et al. 2005; Clark and Ledwith 2006) and have been employed to investigate housing-related issues such as residential mobility and neighborhood choice (Clark and Ledwith 2006; Clark and Ledwith 2007; Cort 2010).

The analyses use information from the public and restricted versions of several L.A.FANS modules. One randomly selected adult (RSA) was selected from the roster of full-time adult household residents provided information about their education, nativity, residential history, and other data. A member of the RSA’s immediate family who was the most informed about finances reported information about income, assets, housing costs, and housing characteristics. These data are linked with a restricted-version of L.A.FANS that identifies respondents’ census tract of residence and the L.A. Neighborhood Services and Characteristics database (L.A.NSC), a publicly available database of census-tract level information created by L.A.FANS staff (Peterson, Pebley et al. 2007). The files are merged so that each record includes information about the respondent and immediate family, household, and census tract.

The final analytic sample is limited to native-born Whites, Blacks/African Americans, Latinos and immigrant Latinos.¹¹ Although some households in L.A.FANS had multiple adult respondents, the analytic sample includes only one adult respondent per housing unit.¹² Consistent with prior analyses (Kutty 2005; McConnell 2006), the sample is limited to lower-income individuals to isolate the likelihood of the outcome from the incidence of poverty generally. More specifically, only respondents with total family incomes at or below 200 percent of the federal poverty threshold, depending on family size or composition, are included in the sample. The final sample size is 716; comparable to the numbers of immigrants in other specialized surveys in Los Angeles County (e.g., Capps, Ku et al., 2002; McConnell and Marcelli 2007; Zhou et al 2008). The complex sampling design of L.A. FANS is addressed in the descriptive and multivariate analyses with the appropriate strata and cluster option in Stata 11.

Analytical approach

Two logistic regressions are modeled to examine the relationships between the independent variables and housing-induced poverty. The first model estimates only main effects for the indicators pertaining to race, nativity, and legal status. The second model incorporates these main effects and the full set of covariates listed in Table 2. The goal of this approach is to identify whether initial differences among groups exist in the baseline model and disparities across groups persist in the fully-specified model. These analyses also extend the housing affordability literature by identifying a more complete set of factors linked with the outcome. Three identical sets of analyses are carried out. In the first analyses, the omitted group is U.S. born non-Hispanic Whites, which allows for formal contrasts among natives by race; the reference group is U.S. born Latinos in a second set of analyses to focus on nativity differences among Latinos, and in a third, unauthorized Latino immigrants are omitted to explore whether there are legal status differences from their authorized Latino immigrant counterparts.

Table 2.

Description of Variables Used in the Analyses

Variable label	Operationalization
Dependent Variable
Housing-Induced Poverty	1 if annual income after housing costs is less than or equal two third of poverty threshold, 0 otherwise
Independent Variables
U.S. born Black	1 if respondent was born in U.S. and Non-Hispanic Black, 0 otherwise
U.S. born White	1 if respondent was born in U.S. and Non-Hispanic White, 0 otherwise
U.S. born Latino	1 if respondent was born in U.S. and Latino, 0 otherwise
Authorized Latino immigrant	1 if respondent not U.S. born and authorized to be in the country, 0 otherwise
Unauthorized Latino immigrant	1 if respondent not U.S. born and not authorized to be in the country, 0 otherwise
Age between 18 and 29	1 if respondent is between 18 and 29 years of age, 0 otherwise
Age between 30 and 44	1 if respondent is between 30 and 44 years of age, 0 otherwise
Age between 45 and 59	1 if respondent is between 45 and 59 years of age, 0 otherwise
Age 60 years or older	1 if respondent is 60 years of age or older, 0 otherwise
Years of education	Respondent’s number of years of education
Married	1 if respondent is married or living with a partner, 0 otherwise
0 children	1 if 0 children in family, 0 otherwise
1 to 3 children	1 if 1-3 children in family, 0 otherwise
4 or more children	1 if 4 or more children in family, 0 otherwise
Immediate family	1 if only respondent, spouse/co-resident partner and children in household, 0 if others live in household
Zero earners	1 if no family member (respondent or spouse/partner) earns income from a job, 0 otherwise
One earner	1 if one family member earns income from a job, 0 otherwise
Two earners	1 if two family member earn income from a job, 0 otherwise
Bank account	1 if family has checking, savings, or money market account, 0 otherwise
Own home	1 if home is owned, 0 otherwise
Number of rooms	Number of rooms in house/apartment excluding bathrooms
Recently moved	1 if moved to current residence in previous year, 0 otherwise
More than half of life in U.S.	1 if has lived in U.S. for 50 percent or more of life, 0 otherwise
LQ recent immigrant	LQ: percent of census tract are immigrants arriving after 1995
LQ median price	LQ: median price of homes in tract, year before surveyed

Housing-Induced Poverty

Following Kutty (2005), housing-induced poverty is operationalized as a binary variable with a value of one signifying having family income that is less than or equal to two-thirds of the federal poverty threshold after paying for housing costs. This variable is created using information about housing costs, income, household characteristics, and the federal poverty thresholds. L.A.FANS asked renters and owners with mortgages to provided information about the cost of rent or mortgage payments. For renters, housing costs comprise the annual total of rent payments provided in the survey. The survey asked homeowners with mortgages about whether their mortgage payments include property taxes and property insurance.¹³ L.A.FANS did not ask renters or homeowners about utility or other housing-related expenditures.¹⁴

L.A.FANS collected information about family income, that is, income earned by the RSA and RSA’s immediate family (co-resident spouse/partner and/or minor children) (Peterson, Sastry et al. 2004). Descriptive data presented later shows that most respondents live in households where the only half-time or greater residents are the RSA, their spouse/parent and children. Following the U.S. Census Bureau (n.d.) procedure for defining income in the poverty calculation, income includes salary and wages earned from employment, public assistance, and assets such as rental property, stocks and bonds before taxes and excludes capital gains or losses, non-cash benefits and excludes income of non-relatives living in the unit.¹⁵ The reference year for income, the number of half-time or greater resident children in the RSA’s family, and total family size are used to identify whether respondents and their family experience housing-induced poverty. Age of the householder is also needed for family units of one or two persons, as the poverty thresholds differ for those over 65 years old. All of this information is used in conjunction with the appropriate federal poverty thresholds (Dalaker 2001; Bishaw and Iceland 2003) to identify whether a family’s remaining income after housing costs is two-thirds or less of the federal poverty threshold.¹⁶

Race/Ethnicity, Nativity, and Legal Status

U.S. born Latino respondents were born in the United States and identify as Latino/Hispanic/Latin American (of any race), U.S. born Whites and Blacks are native-born, non-Hispanic respondents who identify as “White” or “Black/African-American.”¹⁷ Foreign born respondents who identify as Latino/Hispanic/Latin American are differentiated by legal status. L.A.FANS did not directly ask immigrants whether they lacked legal permission to reside in the United States. This study categorizes Latino immigrants who report being a naturalized citizen, permanent resident, asylee, possessing refugee status, temporary protected status, or having a valid visa as authorized. All other Latino immigrants are categorized as unauthorized. This is an accepted approach for identifying unauthorized immigrants in survey research (Capps, Ku et al. 2002; Goldman, Smith et. al 2005) and shares similarities with the “residual” methodology used in federal reporting to identify unauthorized immigrants (e.g., Hoefer, Rytina et al. 2010).

Other Covariates

Table 2 describes the additional variables included in the fully-specified logistic regression of housing-induced poverty. These include whether the respondent is married/cohabitating, respondent’s years of education and age, number of children in family, number of family members earning income from employment, number of rooms in the unit to represent housing size, recently moved to the unit, and whether residence is owned.¹⁸ Financial access/participation is operationalized as having a bank account. Living arrangements are categorized as lives only with immediate family versus more extended situations. Following another L.A.FANS-based study of natives and immigrants (Greif 2009), a binary variable representing having spent fifty percent of his/her life in the United States is incorporated in the final model to examine whether U.S. experience helps predict the outcome.¹⁹

Finally, the full specification includes two neighborhood-level variables: the concentration of recent immigrants and median home prices. Following previous analyses of L.A.FANS data (e.g., Frank and Bjornstrom 2011), these characteristics are entered as measures of relative concentration, specifically location quotients (LQs) that range from a value of 0 to more than 1. These LQs compare the respondent’s tract to the average for all census tracts in Los Angeles County, as reported in Census 2000 data. For example, a value of less than 1 for the LQ of recently arrived immigrants indicates that a respondent lives in a census tract with a lower concentration of immigrants arriving after 1995 than the average tract for L.A. County in 2000; and an LQ of more than 1 means that the respondent lives in an area with higher concentration of post-1995 immigrants than the county average. The second LQ variable corresponds to the median price of residences in the respondent’s census tract relative to the county average.

Results

Table 3 provides weighted descriptives pertaining to the pooled sample and to each of the five groups. Approximately 44 percent of the pooled sample experience housing-induced poverty. Family income, housing costs, and residual income are used in the calculation of housing-induced poverty and are included in the table for illustrative purposes. The analytic sample has mean income of less than $16,500, housing costs that average more than $6,200, and residual income of about $10,100. These figures reveal that lower-income respondents not allocate a very large proportion of their incomes to shelter costs but also have very limited resources to spend on other necessary expenses. The sample selection of those with incomes less than 200 percent of the federal poverty threshold helps explain many of the significant economic disadvantages noted in Table 3. For instance, consistent with these constraints, less than 30 percent of the analytic sample are homeowners, less than half have a bank account, and reside in neighborhoods with higher concentrations of recent Latino immigrants and lower average median prices than the average neighborhood in Los Angeles County. The majority of the analytic sample is Latino, of varying nativity/legal statuses, live with their immediate family only, are married, and have one family member with earned income from a job.

Table 3.

Weighted Descriptives of Analytic Sample

	Pooled Sample	U.S. born Whites	U.S. born Blacks	U.S. born Latinos	Authorized Latino immigrants	Unauthorized Latino immigrants
Dependent Variable
Housing-Induced Poverty (%)	44.4	43.2	46.5	60.9	43.2	41.1
Mean annual family income ($)	16,409	14,500	16,005	15,044	17,978	16,903
Mean annual housing costs ($)	6,270	5,238	6,105	6,174	6,793	6,658
Mean residual income ($)	10,139	9,262	9,900	8,870	11,184	10,245
Independent Variables
Race/nativity/legal status (%)	100.0	23.4	9.7	8.6	29.3	29.0
Age (%)
18-29 years	30.0	20.6	16.5	52.9	19.5	45.9
30-44 years	36.6	25.9	30.7	21.5	43.0	45.4
45-59 years	14.1	14.3	15.9	5.7	21.9	8.0
60 years or older	19.3	39.3	36.9	19.8	15.5	0.1
Mean years of education	10.0	13.1	12.5	11.7	7.7	8.6
Married (%)	53.4	44.1	19.1	48.5	68.0	59.3
Number of Children (%)
0 children	49.7	70.9	43.8	55.9	33.7	49.0
1-3 children	44.0	27.5	40.6	39.7	58.9	44.5
4 children or more	6.3	1.7	15.6	4.4	7.5	6.1
Immediate family	88.4	92.9	92.4	87.5	91.3	80.7
Number of earners (%)
Zero	25.7	58.0	47.6	24.5	17.0	1.5
One	58.2	32.3	49.9	66.5	61.4	76.2
Two	16.1	9.7	2.5	8.7	21.6	22.2
Bank account (%)	42.5	63.8	47.0	59.5	45.7	15.5
Own home (%)	29.1	57.8	33.5	36.8	27.0	4.1
Mean number of rooms	3.4	4.6	3.8	3.6	3.3	2.45
Recently moved (%)	32.2	23.1	22.7	27.5	22.2	54.4
More than half of life in U.S. (%)	86.7	100.0	100.0	100.0	60.9	93.5
Mean LQ of recent immigrants	1.2	0.7	1.2	1.1	1.4	1.6
Mean LQ of median home price	0.8	0.8	0.7	0.8	0.8	0.8
Total N	716	69	73	68	280	226

Source: Author’s calculations of Los Angeles Family and Neighborhood Survey, Wave 1. Due to rounding, percents may not equal 100.0 and sums may not add up correctly.

Tests of group differences, not shown, indicate statistically significant differences across groups. For instance, compared to U.S. born Latinos, U.S. born Whites are more likely to be older and to have no family members earning income from employment, and less likely to live in neighborhoods with high concentrations of recently-arrived immigrants. Unauthorized Latino immigrants are significantly different from all other groups in many domains, such as being less likely to live only with immediate family, less likely to reside with no earners with income from employment, to have a bank account, to own their home, or be over 60 years old. Latino authorized and unauthorized immigrants are similar in characteristics, such years of education and neighborhood immigrant and economic context, but differ in other attributes, like percent of life in the United States. The relatively-high proportion of unauthorized Latino immigrants spending more than fifty percent of their life in the U.S. is partially due to their youth relative to their authorized immigrant counterparts coupled with many years of U.S. experience. Other studies have documented the long period of U.S. residence for many unauthorized immigrants (e.g., Taylor, Lopez, et al 2011).

Table 4 provides the odds ratios and standard errors for the baseline and fully-specified models of housing-induced poverty.²⁰ The first two columns present the results with U.S. born Whites as the omitted group. These results indicate no statistically significant differences in the odds of housing-induced poverty between U.S. born Blacks and Whites in either the baseline or fully specified models (columns 1 and 2, Table 4). U.S. born Latinos have about twice the odds of housing-induced poverty than their native White counterparts in the baseline model (odds ratio of 2.0493), but do not have significantly higher odds when an extensive set of covariates have been included (first and second columns, Table 4). The contrasts between Latino immigrants and the reference group are notable, though not strictly testing differences by race, nativity, or legal status. The baseline model indicates that U.S. born Whites and both Latino immigrant groups have similar odds of experiencing the outcome. However, in the fully-specified model, authorized Latino immigrants have about 78.2 percent lower odds (1- odds ratio of 0.2180) of housing-induced poverty than U.S. born Whites. Unauthorized Latino immigrants have 77.8 percent lower odds (1-0.2220) than White natives (column 2, Table 4).

Table 4.

Odds Ratios from Logistic Regression Analyses of Housing-Induced Poverty: Main Race, Nativity, and Legal Status Effects

	US born Whites (reference)		US born Latinos (reference)		Unauthorized Latino immigrants (reference)
	Baseline	Full	Baseline	Full	Baseline	Full
U.S. born White	----	----	0.4880† (0.1949)	0.5729 (0.2621)	1.0903 (0.3290)	4.5052** (2.1231)
U.S. born Black	1.1431 (0.4851)	0.6600 (0.2674)	0.5578 (0.2493)	0.3781† (0.1868)	1.2463 (0.4617)	2.9734* (1.323)
U.S. born Latino	2.0493† (0.8187)	1.7456 (0.7985)	----	----	2.2343† (0.9387)	7.8644*** (3.9467)
Authorized Latino immigrant	1.0012 (0.3326)	0.2180** (0.1005)	0.4886† (0.1940)	0.1249*** (0.0550)	1.0916 (0.2855)	0.9819 (0.2983)
Unauthorized Latino immigrant	0.9172 (0.2768)	0.2220** (0.1046)	0.4476† (0.1880)	0.1272*** (0.0638)	----	----

Source: Los Angeles Family and Neighborhood Survey, Wave 1.

Notes: Standard Errors in parentheses. Baseline specification: main effects of race and nativity/legal status; Full Model includes: age, education, married, number of children, immediate family, number of earners, bank account, own home, number of rooms, own home^* number of rooms, recently moved, more than half of life in the U.S., LQs of recent immigrant and median home price.

^†p< .10,

^*p<.05,

^**p < .01,

^***p<.001

The third and fourth columns of Table 4 present the logistic regression results when U.S. born Latinos are the omitted group. These results show that both authorized and unauthorized Latino immigrants have substantially lower odds of the outcome than their native Latino counterparts in the baseline model (odds ratios of 0.4886 and 0.4476, respectively, third column). Their lower odds of housing-induced poverty relative to Latino natives decline even further when controlling for the full set of covariates (0.1249 and 0.1272, respectively, fourth column). African Americans are equally likely to be in housing-induced poverty as U.S. born Latinos in the baseline model, but less likely to experience the outcome in the full model; the opposite pattern previously described for Latinos and Whites. The fifth and sixth columns of Table 4 present odds ratios when unauthorized Latino immigrants are the reference group. The results reveal no significant differences in the incidence of housing-induced poverty between unauthorized Latino immigrants and authorized Latino immigrants, even in the full model. Finally, in addition to the aforementioned higher odds for White and Latino natives than unauthorized Latino immigrants, this specification shows that African Americans also are more likely to experience the outcome than unauthorized Latino immigrants (sixth column).²¹ The discussion elaborates on these results.

Table 5 presents the complete set of results for the full model and identify which established predictors and previously untested variables are linked with housing-induced poverty. The odds ratios for all variables other those identifying the race/nativity/legal status groups are the same irrespective of which group is omitted from the specification. U.S. born Latinos are the reference group in these specifications. The results support many of the hypotheses regarding variables typically included in multivariate analyses of housing affordability. For example, consistent with prior work (e.g., Combs, Combs et al. 1994; Kutty 2005; McConnell 2006), respondents who are sixty or older are less likely to be in housing-induced poverty than those who are between thirty and forty-four years old. More educated respondents are substantially less likely to be in housing-induced poverty than their lesser-educated counterparts, odds that decline by 6.8 percent (1- 0.9322) with each additional year of education. Previous work documents the housing affordability challenges of renters relative to homeowners (e.g., Schwartz 2010; Brennan and Lipman 2008); the present study confirms that homeowners have lower odds of housing-induced poverty than renters, controlling for the full set of variables. The logistic regression results also show that respondents in larger housing units are more likely to experience the outcome; odds that increase by a factor of about 1.3 for each additional room in the unit.

Table 5.

Odds Ratios from Logistic Regression Analysis of Housing-Induced Poverty

	Full Modela
U.S. born Latino (omitted)	----
U.S. born White	0.5729 (0.2621)
U.S. born Black	0.3781† (0.1868)
Authorized Latino immigrant	0.1249*** (0.0550)
Unauthorized Latino immigrant	0.1272*** (0.0638)
18 and 29 years of age	0.7563 (0.2397)
30 and 44 years of age (omitted)	--
45 and 59 years old	0.7864 (0.2537)
60 years or older	0.3521† (0.2136)
Years of education	0.9322* (0.0287)
Married	1.2659 (0.3450)
0 children (omitted)	----
1 to 3 children	1.9400* (0.5039)
4 children	1.1942 (0.4754)
Immediate family	0.3132* (0.1477)
Zero earners	2.2942* (0.8732)
One earner (omitted)	----
Two earners	0.2489** (0.1134)
Bank account	0.3756*** (0.0879)
Own home	0.2508** (0.1290)
Number of rooms	1.2979* (0.1413)
Recently moved	1.0978 (0.2348)
More than half of life in US	0.2225*** (0.0691)
LQ recent immigrants	1.3617 (0.2546)
LQ median home price	1.1833 (0.4389)

Notes: Standard Errors in parentheses.

^†p< .10,

^*p<.05,

^**p < .01,

^***p<.001

^aThese are complete results for the “Full Model” in Table 4 (fourth column). Analyses when the reference group is U.S. born White or unauthorized Latino immigrant (second and sixth columns of Table 4) rely on the same specification and produce the same estimates beginning with the indicator for “18 and 29 years of age.”

Other traditional predictors of housing affordability do not operate as expected. For instance, contrary to expectations, neither being married nor median home price in the neighborhood predicts housing-induced poverty, controlling for other variables. With respect to number of children, respondents residing with one to three children have nearly twice the odds of housing-induced poverty than those with no children (odds ratio of 1.9400). Kutty (2005) and McConnell (2005) find the opposite pattern; the present results may be related to unmeasured measured differences in stage of the life cycle or other factors that vary across groups. Recently moving to the unit is not associated with housing-induced poverty, accounting for the full set of variables.

Table 5 confirms that, beyond the importance of disaggregating U.S. Latinos, some variables absent from multivariate analyses of residual income standards are linked with the outcome. Living arrangements are relevant: those living only with immediate family have significantly lower odds of housing-induced poverty than those in extended living arrangements (odds ratio of 0.3132). This result is consistent with past work showing that extended living arrangements sometimes reflect difficulties in affording housing expenses (Capps, Ku et al 2002). Number of earners also matter. Indeed, relative to families with one earner, those with no earned income from employment have higher odds of housing-induced poverty and those with two earners have far lower odds (2.2942 and 0.2489, respectively). As expected, those who have a bank account are less likely to be in housing-induced poverty than those lacking mainstream financial access. Although neighborhood immigrant context is not linked with housing-induced poverty, respondents who have spent half of their life or more in the U.S. have 77.8 percent lower odds (1- 0.2225) of housing-induced poverty than immigrants spending less than half of their life in the country. Taken together, the regression results presented in Table 5 align with previous research about the factors linked with housing affordability and suggest the role of lesser-studied characteristics, such as financial access, in shaping housing affordability.

Discussion

More than forty percent of the analytic sample experiences a poverty standard of living after housing costs. Although this level of housing affordability problems is high, it is not unexpected, given the sample selection of those with relatively low incomes and the higher median rents and home prices in Los Angeles County relative to other U.S. counties (Brennan and Lipman 2008). The logistic regression results support some hypotheses. For example, as expected, U.S. born Whites and Blacks are equally likely to experience housing-induced poverty in the fully specified models. However, the results for Latinos diverge significantly from the hypotheses: 1) the better-than-expected results for Latino immigrants, 2) authorized and unauthorized Latino immigrants have similar odds of housing-induced poverty, and 3) the significant disadvantage of U.S. born Latinos relative to other groups.

Why might Latino immigrants have lower odds of housing-induced poverty than other groups, such as U.S. born Whites (column 2, Table 4)? Clearly, the variables included in the full specification explain some of these results. As shown in Table 5, respondents with more education, who have a bank account and own their residence have lower odds of housing-induced poverty. Latino immigrants, both authorized and unauthorized, are less likely to possess these characteristics than White natives. Thus, controlling for variation in these factors may help account for the results in the fully-specified models (e.g., second and sixth columns, Table 4). Yet, that does not explain why there were no significant differences in housing-induced poverty in the baseline models for Latino immigrants relative to U.S. born Whites or Blacks (first and fifth columns, Table 4). One possibility is that the sample limitation to similarly low-income respondents may have “evened” the playing field with respect to unmeasured differences that might have otherwise disadvantaged Latino immigrants versus White natives and others in the study.

However, the explanation may be related to another central finding of this study. Contrary to expectations, the analyses revealed neither baseline nor residual disadvantages for unauthorized Latino immigrants in the incidence of housing-induced poverty compared to authorized Latino immigrants. In fact, immigrants lacking legal status either were equally likely or less likely to experience the outcome relative to all other groups, even in the baseline model. A recent study of Mexican immigrants also using data for Los Angeles County in 2001 found no statistically significant relationship between legal status and housing tenure (McConnell and Marcelli 2007). In explaining the results, the authors point to the specificity of the location and the time period under study. In 2000, Los Angeles County had the largest Latino population (Guzmán 2001), the largest foreign born population (Suchan, Perry et al. 2007), and the largest unauthorized immigrant population (Fortuny, Capps et al. 2007) of any county in the United States. McConnell and Marcelli (2007) contend that these demographics, particularly the sizable unauthorized Latino immigrant population, the increasing national interest in Latino immigrants as consumers, and changes in mortgage lender practices during this period likely facilitated Mexican immigrant homeownership. The same demographic context and time period also may explain why Latino immigrants in this study, both authorized and unauthorized, could have roughly comparable incomes and housing costs to natives. Together with the sample limitation to relatively low-income respondents, this could account for why neither nativity nor legal status seems to disadvantage Latino immigrants vis-à-vis housing-induced poverty.

What remains unanswered, however, are why Latino natives are more likely to experience housing-induced poverty than Latino immigrants, in the baseline specification or when controlling for predictors such as education, financial access, number of earners, and other variables. Additional analyses, not shown, indicate that removing predictors, such as having a bank account, nor adding new variables, such as household crowding (e.g., Solari and Mare 2012), substantively change the pattern of results for U.S. born Latinos and Latino immigrants. Although it cannot be tested in this study, the segmented assimilation framework offers one way to interpret the results. This perspective argues that immigrants and their children may experience differentiated assimilation trajectories, based on variability in immigration policies related to the social/political context of reception, features of the co-ethnic community, and other factors (e.g. Portes and Zhou 1993; Portes and Rumbaut 2005; Portes, Fernández-Kelly, and Haller 2005; Zhou et al 2008). As a result, some groups may experience upward mobility while others may experience lessened or even downward mobility.

A recent study of Mexican Americans in Los Angeles and San Antonio finds that third and fourth-generation Mexican Americans experience stagnation when it comes to increases in socioeconomic status and decreases in poverty rates relative to first and second generations (Telles and Ortiz 2009). Perhaps the uneven economic integration and downward mobility for the children and grandchildren of immigrants compared with the first generation observed in the Telles and Ortiz (2009) study could explain the present patterns, especially since the focus is on relatively low-income Latino natives and immigrants. Results presented in Table 4 offer tentative support for this possibility. For example, Latino immigrants are less likely to experience housing-induced poverty than U.S. born Latinos from the outset (third column), even though Latino immigrants possess many characteristics shown to be linked with higher odds of the outcome. These results are generally consistent with the segmented assimilation perspective that intergenerational upward mobility is not inevitable for all groups. Clearly; however, more empirical testing is needed.

Also puzzling is why U.S. born Latinos fare worse than similarly low-income African Americans, controlling for the full set of covariates. Tests of group differences, not shown, indicate no significant differences between Latino and Black natives in annual incomes, housing costs, or residual incomes after housing expenditures. In fact, the only statistically significant differences between the two groups are that U.S. born Latinos are more likely to be married, to be young (between the ages of 18 and 29), and less likely to have no family members earning income from a job than native Blacks. As number of earners is linked with housing induced-poverty (Table 5), it would seem that controlling for differences in this variable might help explain the results. However, ancillary analyses, not shown, suggest that removing indicators of earners from the specification does not change the lower likelihood of housing-induced poverty for African Americans relative to native Latinos. Future analyses that can tap into unmeasured differences between these groups, perhaps variations in mortgage terms for homeowners or other factors, may account for the observed disparity between them.

Conclusion

This study focuses on the residual income approach to housing affordability, an alternative perspective to the ubiquitous housing affordability standard based on the proportion of income spent on housing costs. Although some scholars in the U.S. and elsewhere have advocated for the residual income approach, such alternatives have not received much attention from policymakers or widespread acceptance among housing researchers. The present work hopes to draw more attention to this growing body of literature, highlight the diversity of operationalizations using the residual income approach, and expand empirical analyses using such measures.

The multivariate analyses offer the first explicit contrasts of U.S. born Whites, Blacks, and Latinos and among Latinos by nativity and legal status, using any residual income measure. The results offer insight about which of the five groups analyzed are more likely to experience housing-induced poverty in the largest county in the U.S. in the years preceding the recession. They also confirm the importance of previously-documented factors and links additional variables with the outcome. Yet, the limitations of this study, such as the analyses of cross-sectional data for one location and small sample size, require that the results be interpreted with caution. Clearly, additional scholarship is needed. For example, research in contexts other than Los Angeles will clarify whether observed results for Latino natives and immigrants hold in other parts of the country. Studies using recent data collected in Los Angeles emphasize how legal status shapes immigrant mobility (e.g., Abrego 2011; Zhou, Lee, et al. 2008); analyses with updated data could identify whether the results for unauthorized Latino immigrants revealed in this work still holds. On a related point, housing affordability problems have increased since the recession (Joint Center for Housing Studies 2010; 2011) and the economic downturn has had differential impacts of Latinos and African Americans relative to Whites (Taylor, Kochhar et al, 2011; Joint Center for Housing Studies 2011). Thus, analyses using data collected after the economic recession may suggest that patterns of housing-induced poverty among the U.S. by race and ethnicity also may have changed.

Despite these caveats, the analyses confirm Kutty’s (2005) original work showing that housing-induced poverty offers a useful vantage point for documenting housing affordability challenges using a conservative residual income measure. Similarly, like Stone’s research (e.g., 1993, 2006a, 2006), the present study hopes to encourage more housing scholars to examine, develop, refine, and use alternatives to the ratio approach of housing affordability. Housing-induced poverty is one of several possible residual income alternatives, and all of them deserve more examination. Also needed are thorough analyses of residual income standards that focus on inter-group and intra-group differences with a wider array of racial groups, a careful disaggregation of pan-ethnic populations such as Latinos and Asians by specific group (e.g., Mexican, Salvadoran, Chinese, Vietnamese), nativity, generation in the United States, and legal status of immigrants.

In terms of implications for housing policy, this study offers additional evidence about the housing affordability difficulties experienced by low-income households. Nearly half of the respondents in this study are in housing-induced poverty, some of whom earn up to twice as much income as outlined by federal poverty thresholds. Housing affordability challenges would be even more prevalent if more generous estimates of the income required to sustain households, such as the Self Sufficiency Standard, were used in the calculation of housing-induced poverty. The majority of federal housing subsidies are not directed to addressing the affordability needs of low-income households. The largest federal housing subsidy by far is the mortgage interest tax deduction for homeowners, which totaled nearly $101 billion in 2009 (Schwartz 2010 Table 4.1).²² This is more than twice the total direct housing assistance provided by the federal government to lower-income households, which amounted to about $41 billion in 2008 (Schwartz 2010 89). Further, high-earning homeowners are much more likely to be able to take mortgage interest deductions than low and moderate-income homeowners (Schwartz 2010). These and other reasons help explain why, despite the important role that the U.S. government plays in subsidizing lower-income households, substantial proportions of low and moderate-income households continue to have substantial housing affordability challenges. Solving the country’s need for affordable housing requires policy discussions and empirical research that acknowledges the necessity and validity of diverse housing affordability standards.