Introduction

Latinos/Hispanics in the United States make up almost one-fifth of the total population. They are the second largest group after the non-Hispanic white population, the largest ethnic minority group, and the fastest growing population in the United States. According to 2023 U.S. Census estimates, the Latino population was 65.2 million, representing 19.5% of the total population. In Michigan, the Latino population in 2023 was estimated at 600,102, representing 6% of the total.

The Latino population in the U.S. significantly increased over the last 5 decades, from 14.6 million in 1980 to 62.2 million in 2023, or an increase of about 50.6 million. The number of Latinos increased by 3 million between 2020 and 2023, including an increase of 1.16 million Latinos between 2022 and 2023 alone. Michigan has seen a similar trend. The Latino population grew from 157,626 in 1980 to 600,112 in 2023, an increase of 442,476. The number of Latinos in Michigan grew by about 34,226 between 2020 and 2023, with an increase of 11,799 Latinos between 2022 and 2023 alone.

The Latino population growth is primarily due to natural increase (births-deaths), but also to international immigration. According to the 2019-2023 American Community Survey (ACS) 5-year estimates, about 45.3 million people in the United States were foreign-born, representing 13.7 percent of the U.S. population. Among the foreign-born population, 50% were born in Latin America, 31.2% in Asia, 10.6% in Europe, 5.7% in Africa, and 2.4% in other regions.

Historically Latinos in the United States have concentrated in the largest metropolitan areas in the Southwest and the Northeast. Latinos have traditionally resided in five states: California, Texas, New York, Florida, and Illinois. Within those states, they tended to concentrate in gateway and ethnic hub cities such as Los Angeles, Houston, Dallas, New York City, Miami, and Chicago. More recently, Latinos have settled in new destinations outside of traditional gateway cities, particularly in the Midwest and Southeast.

Population changes in both the U.S. and in Michigan, on one hand, are attributable to the increase of the Latino and Asian populations, but on the other hand, to the decline of the non-Hispanic white population. These demographic shifts are linked to economic changes in many places and have implications for the communities experiencing them. The Latino population has been attracted to new destinations by, or recruited for, employment opportunities in agriculture, meat packing, other food-processing jobs, timber harvesting and processing, and other industries. In Michigan, the Latino population is attracted by local labor market opportunities, especially in nonmetropolitan areas in agriculture-related occupations. The out-migration of non-Hispanic white and other native populations from many places is also linked to changes in local economies of those places. Local economic changes are directly linked to the availability or lack of job opportunities, which have implications for family income, income security, racial/ethnic inequalities, including increased poverty, housing shortages, and overall well-being of the population in those places. This article focuses on the well-being of racial/ethnic and immigrant population in Michigan, focusing on household poverty. This study addresses the following research questions:

1. How do racial/ethnic minority households in Michigan compare with non-Hispanic white households in terms of poverty?

2. How do immigrant households in Michigan compare with native households in terms of poverty?

3. Do the associations between race/ethnic minority and immigrant status and poverty persist after controlling for other individual and household confounders?

4. Do local labor market opportunity structures influence household poverty, net of the effects of nonmetropolitan/ metropolitan location, concentrated disadvantage, and individual and household confounders?

Background

The causes of poverty are multifaceted and include both micro- and macro-level theories. In this article, I will focus only on three main explanations of the causes of poverty: (1) Human capital explanations that emphasize characteristics such as human capital, attitudes, and behavior of the individual (Becker, 2009; Lichter et al., 1993); (2) Structural explanations such as economic restructuring and global processes that highlight macro forces that affect the distribution and changes in opportunity structures (Wilson, 2012); and (3) Social stratification that highlights inequalities across social groups, including race/ethnicity, immigrant status, gender, and geopolitical space.

Human Capital Explanations

Individual explanations of poverty suggest that individuals with lower levels of education and job experience are employed in low-wage jobs and are likely to have lower earnings and therefore be in poverty. Alternatively, individuals with higher education and better job experience should earn higher wages and hence are less likely to be poor. These views are used to explain why racial/ethnic minorities, who tend to have lower levels of education, are in low-wage jobs, and likely to be in poverty.

This view has dominated research and policy on poverty, however, emphasis on individual attributes and actions often overlooks the enormous impact of social, economic, and political systems on poverty. Structural factors, such as not having enough good jobs, rather than not having enough skills or motivation (Iceland, 2013), may be the main cause of poverty. While individual attributes such as human capital may partially explain poverty differentials and income gaps, existing opportunity structures beyond individuals’ attributes and abilities may better explain the level of poverty and why it persists.

Economic Restructuring Explanations

The ongoing restructuring of economies has not only created new structures of work but has also constrained choices available to workers in different labor markets and at home. Deindustrialization, the increase in new technologies, and globalization are key macro-level factors that may be related to poverty. Economic restructuring led to the loss of many good jobs, especially low-skilled, blue-collar jobs with greater incomes as well as health insurance and retirement benefits. The newly created jobs in the service sector of the economy were of two kinds: (1) those requiring higher education and technical skills, and (2) those requiring lower job skills. With deindustrialization, a growing number of new jobs are part-time, contingency, subcontracted, or temporary, with irregular work schedules and high layoff and turnover rates (Seccombe, 2000).

Technological changes in the economy play a role in increasing inequality by raising the demand for high-skilled workers, such as engineers and programmers, while reducing the demand for lower-skilled, low-paid workers (Iceland, 2013). A major effect of deindustrialization has been the loss of employment. Many families, especially those with lower skills and educational levels, were unable to find jobs, especially good jobs that pay well enough to lift them out of poverty and economic uncertainty and offer fringe benefits, such as health insurance and pensions.

The process of globalization may have triggered the incidence of poverty. The importation of manufactured goods from less developed countries places workers in advanced countries in direct competition with those in less developed countries. The lower labor costs and lack of environmental regulation costs make the import of manufactured goods economically beneficial to companies but costly to workers because they reduce wages and increase unemployment, especially for the least-skilled workers.

Another aspect of globalization has been the continuous decline of labor unions. According to Desmond (2023), “only around one in ten American workers today belong to a union, and most of them are firefighters, nurses, cops, and other public sector workers. Almost all private sector employees (94 percent) are without a union” (Desmond, 2023:49). Greater unionization is normally associated with reduced income inequality and greater well-being. However, employers have an arsenal of tactics that prevent collective bargaining (Desmond, 2023). Nonunionized workers typically are paid lower wages and have less job security (Iceland, 2013), which may contribute to poverty.

Social Stratification Explanations

Social stratification explanations of poverty underline the hierarchical and uneven access to opportunities across race/ethnicity, social class, gender, and immigrant status. Racial/ethnic minorities are on average more likely than non-Hispanic whites to have lower levels of education, lower levels of employment, lower wages, and chronic health conditions—all characteristics associated with higher poverty rates (Iceland, 2013). Women, compared to men, especially minority, female-headed households, continue to occupy lower economic positions and are more likely to be in poverty.

Immigrant families are at greater risk of poverty and have lower incomes than nonimmigrant families. Limited language proficiency and unfamiliarity with U.S. customs and the labor market considerably hinder immigrant economic mobility in the short run. But over time and in subsequent generations, labor market barriers become less important (Iceland, 2013). The influx of low-skilled migrants is often viewed as increasing poverty, in part because they displace native workers and threaten their wages, although this relationship has been contested (Portes & Zhou, 1992; Waldinger, 1996). Other studies argue that overall levels of racial and class polarization have increased, with immigrants concentrated in “casual” jobs and native whites concentrated in professional jobs (Frey & Liaw, 1998; McCall, 2001).

Economic well-being is not only unevenly distributed across race/ ethnicity, social class, gender, and other social strata, but it is also unevenly distributed across geopolitical spaces. The impact of economic restructuring has been uneven across spaces, affecting individuals, families, and communities in different locations (Lobao, 1990; Lyson & Falk, 1993; Tickamyer et al., 1993). In urban areas, poverty persists because of the combined and interacting effects of joblessness, deteriorating neighborhoods, and the “oppositional” culture that these forces generate (Duncan, 1999; Wilson, 2012). In rural areas, economic restructuring has intensified existing disadvantages of rural communities (Lyson, Falk, Henry, Hickey, & Warner, 1993; Tickamyer & Duncan, 1990). Nonmetropolitan families are more likely to be in economic distress and poverty than their metropolitan counterparts (Castle, 1993). Nonmetropolitan areas have relatively limited employment and earnings opportunities and less diversified labor markets (Tickamyer & Duncan, 1990).

Data and Measures

Data

Data for this article are drawn from two sources. Individual-level data on 121,651 households are extracted from the 5-year (2018–2022) American Community Survey (ACS) Public Use Microdata Sample (PUMS). Local labor market area (LMA) data are from the 2018–2022 ACS Summary file (ACS-SF). The ACS PUMS data is a national sample of population and housing unit records. This study uses a sample of population and housing units in Michigan. The primary level (level - 1) is the household. Only householders of working age (i.e., 16–64 years) are selected. Excluded from the analysis are subfamilies within households, military households, households with zero household income, and individuals living in group quarters. The ACS-SF contains sample data about the characteristics of different geographic units. The level - 2 unit of analysis is the local labor market area (LMA). We used the PUMA level as a proxy of the LMA. For the U.S. Census Bureau confidentiality requirements, PUMA contains at least 100,000 people.

Measures

The primary outcome variable is household poverty, a dichotomous variable that takes the value of 1 if the household falls below the official poverty line (i.e., below 100% of the poverty threshold) and 0 otherwise. Descriptive statistics of household and LMA levels (means, standard deviations, and percentages) are displayed in Appendix Table A1.

Individual- and Household-Level Variables

The basic household characteristics include householder’s age (years), sex/gender (male or female), race/ethnicity, immigrant status, educational attainment of householder and/or spouse/partner if present (the highest), marital status ((i.e., married, formerly married (divorced, separated, and widowed), and never married)), number of children, employment status, home ownership (i.e., housing owner or renters), and residential stability (i.e., 10 years or more in the same house or otherwise). The main predictors at the household level are race/ ethnicity, immigrant status, and employment status of householder and/or spouse/partner. The other variables are controlled for in the level - 1 model. Race/ethnicity is constructed using race and Hispanic origin variables. First, Latino householders are distinguished from nonLatino householders. Second, non-Latino householders are categorized by race to include whites, African Americans, Asians, including Pacific Islanders, and other races, including Native Americans and multiple races. The latter category of other races was excluded from the analysis. Immigrant status is constructed as a dummy variable indicating if the householder is foreign-born or not. Employment status is a composite variable indicating whether either the householder or spouse/partner (if present) are employed or not. This variable was aggregated at the household level to indicate the number of earners in the household and was recoded as a dummy variable to indicate whether at least one earner in the household was employed or otherwise.

Local Labor Market Area Variables

The analysis includes several demographic, socioeconomic, and employment characteristics at the PUMA level. These include nonmetropolitan status, percentages of the population 65 years and older and under 18 years of age, high minority concentration (i.e., 40% or more of African Americans and Latinos), immigrant concentration (i.e., percentage of the population that is foreign-born), concentrated disadvantage (i.e., a standardized principal factor scale) combining the following variables: percentage of residents in poverty, percentage of residents unemployed, percentage of households receiving public assistance, percentage of female-headed families with children under the age of 18, percentage of residents 25 years or older with less than high school education, and in negative coded: percentage of residents 25 years or older with college education, percentage of residents in managerial, professional, and technical occupations, percentage of affluent households (i.e., with household income ≥ $150,000), and median household income. At the PUMA level, industrial structure is measured by the following industries: percentage of manufacturing industry, percentage of consumer service industry, and percentage of good jobs (i.e., professional, managerial, and technical occupations, information, and finance, insurance, and real estate occupations).

Statistical Methods

A multilevel logistic regression model for binary outcomes (Raudenbush & Bryk, 2002) is used to model the odds that a household in each LMA is in poverty. The odds that a household is in poverty are modeled as a function of individual, household, and LMA characteristics. The analysis proceeds from examining first the household and then the LMA effects on odds of poverty. The first set of models examines the effects of race/ ethnicity and immigrant status on the odds of being poor, controlling for age and sex (Model 2, Table 1). The second stage of models adds household background characteristics, including educational attainment (Model 3) and employment status and examines their effects on the odds of being poor, controlling for marital status, homeownership, and residential stability (Model 4). The last set of models examines the effects of nonmetropolitan status and LMA concentrated disadvantage on the odds of being poor, net of the effects of household characteristics. They also examine the effect of having at least one earner in the household on the odds of being poor and whether LMA employment opportunities reduce poverty via employment (Model 5, Table 2).

Results

Model 1 of Table 1 displays the results of an unconditional model (no predictor) of household poverty. This shows that the average odds of poverty are 0.094, or about 1 to almost 11. This corresponds to a probability of 1/ (1 + (1/0.094)) = 0.086. Assuming that the odds of poverty are approximately normally distributed with mean 0.094 and variance of 0.374, it is expected to be a 95% confidence interval of (0.028, 0.311) or, in terms of probabilities, a 95% confidence interval of (0.027, 0.237). It appears that some labor market areas (LMAs) have poverty rates near zero, while in others, 24% of households are in poverty, which shows significant variations in poverty across LMAs in Michigan.

Model 2 consists of race/ethnicity and immigrant status combinations, adjusting for householder’s age, gender, marital status, homeownership, and residential location. The results show that the odds of poverty are significantly higher among non-Hispanic white immigrant, Latino native, Latino immigrant, Black native, and Black immigrant households than non-Hispanic white native households. More specifically, the odds of poverty are 3.1 times higher for non-Hispanic white immigrant, 1.5 times higher for Latino native, 1.6 times higher for Latino immigrant, 1.8 times higher for Black native, and almost 2 times higher for Black immigrant households than non-Hispanic white native households, respectively (Model 2). These results adjust for significant control variables, which show that the odds of poverty increase by householder’s age and are significantly higher among female than male householders, among formerly married (i.e., divorced, separated, or widowed) and never married than married householders, and higher in nonmetropolitan than in metropolitan areas. The results also show, as might be expected, that the odds of poverty are significantly lower among homeowners than renters. These demographic variables explain about 56% of variance in poverty across LMAs (Model 2).

Model 3 shows the effects of educational attainment on poverty. The results show, as expected, that the odds of poverty are significantly lower for households with high school, some college, and college or higher education than those with less than high school education. Specifically, the odds of poverty are 56% lower in households with a high school education, 71% lower in households with some college education, and 89% lower in households with a college education compared with those in households with less than high school education, respectively. Notice that the odds of poverty are reduced for every racial/ethnic and immigrant group and rendered for Latino immigrant households not significantly different from those of non-Hispanic white native households, once education is taken into consideration. Educational attainment explains an additional 21% variance in poverty across LMAs.

Model 4 in Table 1 adds the number of earners in the household. Having at least one earner in the household is associated with significantly lower odds of poverty. Notice that once the number of earners in the household is accounted for in Model 4, the odds of poverty for racial/ ethnic and immigrant groups remain significantly higher than those of non-Hispanic white native households and slightly increase for immigrant households. Adding the number of earners in the household explains additionally 29% variance in poverty across LMAs.

Models 5 – 7 in Table 2 combine household characteristics and LMA characteristics in predicting the likelihood of poverty. The results in Model 5 show that the odds of poverty remain significantly higher in nonmetropolitan than in metropolitan LMAs, adjusting for household characteristics. Specifically, the odds of poverty are 10% higher for households in nonmetropolitan than in metropolitan LMAs. The results also show that the odds of poverty are significantly higher in concentrated disadvantaged than in nonconcentrated disadvantaged LMAs. Specifically, each standard deviation increase in the concentrated disadvantaged scale increases the odds of poverty by 18%, net of the effects of individual and household confounders in the model. Adding nonmetropolitan residence and concentrated disadvantaged scale in Model 5 accounts for an additional 28% variance in poverty between LMAs.

Model 6 in Table 2 tests whether characteristics of the local labor market affect poverty. The results show that the greater the proportion of manufacturing industries the lower the odds of poverty. Specifically, each standard deviation increase in the proportion of manufacturing industries reduces the odds of poverty by 12%, net of the positive effects on poverty of living in nonmetropolitan/metropolitan and concentrated disadvantaged local labor market areas. Adding the percentage of manufacturing industries in LMAs in Model 6 explains an additional 26% variance in poverty between LMAs.

The final model (Model 7) tests the degree to which the risk of poverty is reduced by the local labor market opportunity structure and the number of earners in the household. The results in Model 7 show that the odds of poverty are significantly lower in local LMAs with more manufacturing industries, but higher in local LMAs with high concentration of disadvantage and in nonmetropolitan LMAs. The results also show that the odds of poverty are significantly lower in households with at least one earner and that those odds are further reduced if the household is located in local LMAs with high rates of manufacturing industries and good jobs (i.e., professional, information, and finance, insurance, and real estate services).

Discussion and Conclusion

The results highlight racial/ethnic and immigrant status’ differences in household poverty in Michigan. The results show significant differences in household poverty by race/ethnicity and immigrant status. Immigrant and minority (Latino and Black) households are more likely to be overrepresented among the poor than non-Hispanic white native households. Drawing on the results in the final model in Table 2, the odds of poverty are 3 times higher for non-Hispanic white immigrant, 1.3 times higher for Latino native, 1.4 times higher for Latino immigrant, 1.4 times higher for Black native, and 2.7 times higher for Black immigrant households than those on non-Hispanic white native households, respectively. One of the key findings of this analysis is that racial/ethnic and immigrant differences in poverty remain significant, even after accounting for confounders such as educational attainment, number of earners in the household, householder’s age, gender, marital status, homeownership, residential location, and LMA characteristics.

The results also show that the odds of poverty are significantly higher in nonmetropolitan than in metropolitan LMAs, adjusting for individual and household and LMA characteristics. Although the overall adjusted effect of residential location on poverty is diminished, it remains statistically significant. The odds of poverty are in the end 5% higher in nonmetropolitan than in metropolitan LMAs, adjusting for individual/household and LMA characteristics. This implies that the uneven development may explain poverty differences between LMAs, with households in nonmetropolitan LMAs experiencing greater burden in terms of poverty than those in metropolitan LMAs.

The results also show, as expected, that odds of poverty are significantly higher in LMAs with higher concentration of disadvantage than in concentrated advantaged LMAs, net of the influence of individual/ household characteristics and nonmetropolitan/metropolitan residence. Overall, each standard deviation increase in the concentrated disadvantage scale increases the odds of poverty by 15%, net of the effects of individual/household characteristics and nonmetropolitan/ metropolitan residence. These areas are characterized by high poverty, unemployment, proportions of less skilled workers, proportions of households receiving public assistance, and high proportions of female-headed families with children. They are also characterized by a low proportion of workers employed in managerial, professional, and technical occupations and a low number of affluent households.

Another important finding is that the greater the proportion of manufacturing industries in a labor market area, the lower the odds of poverty, net of the effects living in nonmetropolitan/ metropolitan and concentrated disadvantaged LMAs, and individual and household characteristics. The adjusted odds of poverty are reduced by 12% for each standard deviation increase in the proportion of manufacturing industries.

Another important finding is that the odds of poverty are significantly lower in households with at least one earner, and those odds are further reduced if the household is in a local LMA with high rates of manufacturing industries and good jobs. More importantly, the odds of poverty for those living in a nonmetropolitan or concentrated disadvantaged LMA are significantly reduced when at least the head and/ or the spouse in the household works and they live in LMAs with high rates of manufacturing and availability of great jobs (i.e., in professional, information, and finance, insurance, and real estate services).

The results reported above highlight persistent racial/ethnic and immigrant inequalities in household poverty. They also suggest that poverty is likely to remain unchanged unless one pays attention to both individual and structural characteristics that create poverty and poor places. More employment opportunities with quality jobs are needed in nonmetropolitan areas to reduce the differences in poverty rates between nonmetropolitan and metropolitan LMAs. The lack of economic opportunities in those LMAs may be associated with high poverty. High minority and immigrant populations tend to be concentrated in those areas. Economic restructuring hit these LMAs the hardest with the loss of manufacturing jobs that were paying relatively well and the flight of middle-class families (Wilson 2012). Most residents in manufacturing industries, especially those with low levels of education, lost their jobs and were not able to secure comparable new jobs in the service sector and other newly created jobs. With economic restructuring, the flight of middle-class families produced social and economic environments with limited tax bases and reduced social resources. Investing in better job opportunities in those areas may reduce poverty and other socioeconomic disadvantages.

The findings reported also suggest that more employment opportunities in manufacturing industries, which traditionally pay relatively higher incomes, are crucial to reducing poverty, especially in nonmetropolitan and concentrated disadvantaged LMAs. More importantly, having at least one earner in the household, especially in LMAs with high rates of manufacturing and availability of great jobs, has a higher propensity to reduce poverty.

Finally, the findings above reveal that even after accounting for the effects of LMA opportunity and composition structure, residential location, and individual and household confounders, racial/ethnic and immigrant status’ differences in poverty remain significant. This implies that policymakers at all levels—federal, state, and local government— can intervene to reduce poverty overall and differences in poverty between race/ethnicity and immigrant groups. This can be done in ways that increase investment in human capital through spending, for example, more on schools so they can hire qualified teachers, but also through investment that improves job training and skills. This can also be done by investing in economic opportunities that create more and sustainable jobs in different labor markets in both nonmetropolitan and metropolitan areas and forgotten places with high concentration of economic disadvantage, specifically by creating job opportunities that pay a decent and livable incomes above poverty, such as those in manufacturing and professional services. This can also be done by supplementing incomes of those in or near poverty through local, state, or federal programs such as income-tax credits and housing programs.

Table 1. Effects of Household and LMA Characteristics on Household Income (Odds)

Variable	Model 1	Model 2	Model 3	Model 4
Household characteristics
Race/Ethnicity and Immigrant Status
White nativea		-----	-----	-----
White immigrant		3.084***	2.804***	2.961***
Latino native		1.502***	1.290**	1.283**
Latino immigrant		1.612***	0.942	1.387*
Black native		1.796***	1.531***	1.460***
Black immigrant		1.954**	1.801*	2.633***
Educational attainment
Less than high schoola			-----	-----
High school			0.422***	0.566***
Some college			0.289***	0.397***
College or higher			0.109***	0.172***
Number of earners
Nonea				-----
At least one earner				0.126***
Control variables
Age (years)		1.007**	1.001	0.977***
Malea		-----	-----	-----
Female		1.401***	1.511***	1.537***
Marital status
Marrieda		-----	-----	-----
Formerly married		2.643***	2.197***	1.640***
Never married		2.830***	2.380***	1.61***
Homeownership
Rentera		-----	-----	-----
Owner		0.351***	0.423***	0.490***
LMA characteristics
Intercept	0.094***	0.072***	0.249***	0.838†
Metropolitan		-----	-----	-----
Nonmetropolitan		1.245***	1.155**	1.125**
Variance components
LMA	0.374***	0.165***	0.131***	0.093***
		55.88	20.61	29.01

*** p < 0.001; ** p < 0.01; * p < 0.05; † p < 0.10 (two-tailed tests)

Notes: N = 121,295 (Level - 1); N = 68 (Level - 2). ^aReference category.

 

Table 2. Effects of LMA Characteristics on Household Income and Number of Earners (Odds)

Variable	Model 5	Model 6	Model 7
Household characteristics
Race/Ethnicity and Immigrant Status
White nativea	-----	-----	-----
White immigrant	2.972***	2.976***	2.952***
Latino native	1.275*	1.275	1.269*
Latino immigrant	1.377*	0.379*	1.403*
Black native	1.434***	1.428***	1.430***
Black immigrant	2.625***	2.605***	2.703***
Educational attainment
Less than high schoola	-----	-----	-----
High school	0.567***	0.567***	0.568***
Some college	0.399***	0.398***	0.399***
College or higher	0.174***	0.173***	0.175***
Number of earners
Nonea	-----	-----	-----
At least one earner	0.126***	0.126***	0.122***
Control variables
Age (years)	0.977***	0.977***	0.976***
Malea	-----	-----	-----
Female	1.537***	1.538***	1.536***
Marital status
Marrieda	-----	-----	-----
Formerly married	1.639***	1.639***	1.643***
Never married	1.679***	1.674***	1.674***
Homeownership
Rentera	-----	-----	-----
Owner	0.489***	0.490***	0.491***
LMA characteristics	Intercept	Intercept	Intercept	Number of Earners (Slope)
Intercept	0.825†	0.818*	0.811**
Nonmetropolitan	1.104**	1.086**	1.054*
Concentrated disadvantage	1.184***	1.175***	1.146***
Percent manufacturing		0.876***	0.924†	0.882**
Percent of good jobs				0.900*
LMA variance components
Intercept	0.066***	0.049***	0.052***
Number of earners, slope			0.099***
% of variance explained	29.03	25.76	-22.45

*** p < 0.001; ** p < 0.01; * p < 0.05 (two-tailed tests)

Notes: N = 121,295 (Level - 1); N = 68 (Level - 2). ^aReference category.

Appendix A. Descriptive Statistics of Selected Variables

Variable	Mean (%)	Standard Deviation	Minimum	Maximum
Household Characteristics (N = 121,295)
Household poverty	9.41		0.00	1.00
Householder's age (years)	44.82	12.52	16.00	64.00
Female (%) (ref. = male)	50.01		0.00	1.00
Race/Ethnicity (%) (ref. = non-Hispanic White)
Latino	5.12		0.00	1.00
Black	14.39		0.00	1.00
Foreign-born (%) (ref. = native)	5.64		0.00	1.00
Race/Ethnicity and immigrant (ref. = non-Hispanic White native)
Non-Hispanic white immigrant	3.5
Black native	13.9
Black foreign born	0.5
Latino native	3.5
Latino foreign born	1.6
Educational attainment (%) (ref. = Less than high school)
High school	19.95		0.00	1.00
Some college	35.93		0.00	1.00
College or higher	39.78		0.00	1.00
Marital status (%) (ref. = Married)
Formerly married	21.27		0.00	1.00
Never married	29.16		0.00	1.00
Number of children (5 = 5 or more)	0.65	1.06	0	5
Employment status (%) (ref. = At least one earner)
Number of earners	85.31		0.00	1.00
Homeownership (%) (ref. = renters)	69.36
Residential stability (%) (ref. = Less than 5 years)	13.39
LMA Characteristics (N = 68)
Nonmetropolitan (%) (ref. = metropolitan)
Nonmetropolitan (%)	17.65		0.00	1.00
Percent 65 and over
Percent under 18	21.43	2.65	12.19	27.29
High minority concentration (%)  (≥ 40%) (ref. = < 40%)
High minority concentration			0.00	1.00
Immigrant concentration	9.12	8.73	1.51	38.73
Concentrated disadvantage	0.00	6.56	-14.51	17.59
Percent of good jobs	0.00	2.28	-4.20	6.82
Percent in manufacturing	18.50	4.42	8.92	33.55
Percent in services	24.43	3.01	18.29	31.71

 

---