Mitigating Poverty through the Formation of Extended Family Households: Race and Ethnic Differences

Abstract

In times of hardship, moving in with family is one strategy for alleviating economic deprivation and uncertainty. The ability of the family to buffer against poverty may vary by the resources available to and the economic needs of individuals. I assess how the formation of extended-family households is associated with a move into or out of poverty and how this association varies by race and ethnicity, since economic resources and norms around extended-family households differ. Using longitudinal data that span four years, I estimate linear fixed effects regression models to assess how changes in living arrangements are related to changes in poverty. I find that moving into an extended-family household reduces poverty, especially for the joining family unit. Most of this poverty reduction occurs through a family safety net, with a non-poor family taking in poor family units.

Extended-family households have become increasingly common in the United States; the proportion of the population in such households has increased from 12 percent in the 1980s to 18 percent in 2012 (Fry and Passel 2014). Recent increases in the prevalence of extended-family households have occurred across all racial and ethnic groups, though racial and ethnic minorities have historically been more likely to live in extended-family living arrangements and are still more likely to share households (Fry and Passel 2014).

Despite the documented increase in the prevalence of extended-family living arrangements, little is known about whether and how moving into an extended-family household might change the amount of resources (e.g., money, food, and shelter) available to individuals in these households. Moving into an extended-family household can increase the resources available to an individual because he or she can now pull from the collective resources of a larger household (assuming resource sharing in related households). However, moving into an extended family household can also strain the limited resources of some individuals — namely, those who take in family — as each individual is an additional consumer of resources (Pfeiffer, Anacker, and Louton 2016). At the same time, when households combine, they gain economies of scale; that is, the cost per person of meeting basic needs declines as household size increases. Thus, while a host family (the householders of newly formed extended-family households) taking in a guest family (those moving in with others) with a lower per person income, would experience a net reduction in income per person, both families might still reduce their total living costs by sharing the costs of housing such as rent, utilities, and other shared household goods. Additionally, members of such households may gain other benefits, such as childcare or other household labor; these benefits might also translate into cost savings for some.

The theory of family adaptive strategies implies that families as a collective of individuals function to make rational calculations in the best interest of the family within the social and economic constraints of their lives (Moen and Wethington 1992). Building on this theory suggests that family members are unlikely to move in with others unless doing so can be expected to provide some perceived or anticipated improvement in their situations, whether purely economic or otherwise. However, it is important to understand more concretely how changes in living arrangements relate to changes in poverty. While the advantages may go beyond household income gains for both movers and hosts, this paper will focus specifically on the nature of the financial changes experienced by forming an extended-family household. I leverage longitudinal data from 2008–2013 to assess how the formation of an extended-family household is associated with changes in poverty for hosts and guests using fixed effects regression.

Additionally, I will examine how forming an extended-family household may have differing effects on poverty by race and ethnicity. Normative household structures vary by race and ethnic group, with blacks, Hispanics, and Asians all reporting higher rates of extended-family living arrangements than non-Hispanic whites (Keene and Batson 2010; Van Hook and Glick 2007). For racial and ethnic minorities, extended-family households may reflect long-term strategies that have been used to cope with the disadvantage and adversity these groups have historically encountered (Roschelle 1997). However, while some groups may employ extended-family households as a resiliency strategy in the hopes of overcoming economic hardship thereby, the actual effectiveness of these living arrangements across groups is still unclear from an empirical standpoint. Depending on what the data reveal, then, about how extended-family formation affects the poverty status of various racial and ethnic minorities, extended-family household formation might reflect higher levels of informal family support for these groups, or it might merely represent yet another disadvantage among the options that minorities find for coping with economic hardship.

ECONOMIC MOTIVATIONS FOR EXTENDED-FAMILY LIVING ARRANGEMENTS

Housing is expensive and typically accounts for 30–50 percent of total income among low-income families in the United States (Schwartz and Wilson 2008). Indeed, when living on their own becomes too onerous, low-income families and individuals may turn to other relatives for housing support as part of an adaptive family strategy (Moen and Wethington 1992). Adaptive family strategies are the social and economic decisions that families make with the intention of honoring all family members’ best interests (McCubbin and Patterson 1983). Families’ response to the social realities of an economic recession, for instance, may require a redistribution of resources within the family to cope with limited economic resources (Moen and Wethington 1992). Combining households can be a cost-effective strategy for improving the circumstances of the individuals involved; co-residence is a mechanism of resource transfer to young, old, unemployed, and sick family members (Bianchi et al. 2008), and the resources transferred are not limited to the monetary (e.g., time and energy spent on caregiving). For low-income families, reducing this rent burden by doubling up is a common strategy (Pilkauskas, Garfinkel, and McLanahan 2014).

The decision to live in an extended-family household is often determined largely by necessity rather than choice, and structural constraints often limit the potential strategies that are available to individuals and families in the first place (Moen and Wethington 1992). For instance, economic limitations (e.g., limited income to make in vivo transfers) and institutional barriers (e.g., policy restrictions on public housing eligibility) can both constrain an individual’s housing choices. Forming an extended-family household, requires both the host family and the guest family who is moving into the host’s household to decide this is in the best interest of the collective group. Financial motivations to combine households likely occur in one of two ways: the pooling of collective resources of two family units to benefit both units or to provide family safety, whereby one family unit benefits.

As the service economy expands and the middle class erodes in the United States, an increasing number of workers have wages so low that, despite working full time, they still live in poverty (Burkhauser and Sabia 2007). One strategy that potentially allows low-income earners to avoid poverty is pooling wages in the household across multiple earners. Forming extended-family households with multiple earners who can pool resources may be increasingly necessary for low-wage earners to meet increasing costs of living (Burkhauser and Sabia 2007; Joint Center for Housing Studies 2017). This is especially true, since many low-wage individuals have more income volatility, making it difficult to smooth out expenses. One strategy for coping with volatile income may be reliance on family through shared housing or borrowing, despite the often limited resources available within their kin network (Heflin and Pattillo 2006; Morduch and Schneider 2017). Further, given the high cost of childcare, having an additional adult in the household may be helpful to offset this cost as well (Sarkisian and Gerstel 2004).

Although resource pooling can benefit both host and guest families, research on within household income and rental contributions finds that a majority of income is earned by the host family and nearly all rental expenses are paid by the host family (Glick and Van Hook 2011; Reyes 2018). Additionally, individuals are more likely to move in with family in response to economic shocks such as a job loss (Wiemers 2014). Taken together, these findings suggest that the formation of extended family households provides a safety-net for family members in need of economic support rather than a mutual benefit for most families.

RACIAL AND ETHNIC DIFFERENCES IN HOUSEHOLD COMPOSITION AND FAMILY SUPPORT

In 2012, Asians were most likely to live in extended-family households, followed closely by blacks and Hispanics; non-Hispanic whites were least likely (Fry and Passel 2014). Historical and contemporary racial differences in household composition have been attributed to both structural and cultural explanations (Angel and Tienda 1982). For example, racial and ethnic differences in household composition have been linked to cultural differences in family cohesion and in the importance placed on extended family ties (Burr and Mutchler 1999; Roschelle 1997; Vega 1995). Meanwhile, other scholars have argued that strong kin networks are an adaptive response, in general, through which ethnic minorities are forced by their circumstances to compensate for limited economic resources (Baca Zinn and Pok 2002; Tienda and Angel 1982).

Different cultural norms and expectations across racial and ethnic groups may increase the chances that minority families can rely on relatives to move out of poverty, but these expectations may also mean that families feel pressured to help even when their own resources are limited. In studies of norms of filial responsibility, for instance, blacks and Hispanics are more likely than non-Hispanic whites to believe that each generation should provide co-residential assistance when needed (Burr and Mutchler 1999). Moreover, racial and ethnic differences in familial expectations have been observed for both young and older adults (Burr and Mutchler 1999; Fuligni 2007), suggesting that such norms span generations.

Racial differences observed in the types of support offered also suggest that whites are able to provide more support through economic transfers rather than through the formation of extended-family households, perhaps because white families have more financial resources (Sarkisian and Gerstel 2004). The fact that type of support varies by racial or ethnic background might reflect not just that cultural norms vary, but also that the economic benefits of extended-family household formation also varies by race and ethnicity.

The primary structural explanation for racial and ethnic differences in living arrangements is that blacks and Hispanics have lower socioeconomic status, which is reflected in their higher rates of poverty and lower levels of income, education, homeownership, and wealth (Keene and Batson 2010; Schwartz 2014; Shapiro 2017). Lower socioeconomic status among blacks and Hispanics could affect the economic benefits of forming extended-family households in three key ways: (1) fewer resources increase the need and prevalence of these moves; (2) fewer resources among kin limit the ability to rely on family; or (3) fewer resources among kin increase the prevalence of mutual benefit through income pooling in extended-family households.

Indeed, many studies of racial and ethnic differences in living arrangements have found that structural factors such as socioeconomic status explain a large portion of these differences — but do not fully account for them (Burr and Mutchler 1999; Keene and Batson 2010). Current racial differences in wealth are large, larger even than income differences; the median wealth of white households is about 10 times that of black households and eight times that of Hispanic households (Killewald, Pfeffer, and Schachner 2017). These large racial and ethnic gaps in wealth stem largely from the legacy of slavery and discriminatory policies that provided opportunities for land acquisition, higher education, and homeownership to white Americans while specifically excluding blacks and other minority groups (Shapiro 2017). Further, many of the tax provisions in the United States are aimed at helping those with more wealth, which helps perpetuate current wealth inequalities, making it difficult for minority groups to build wealth (Shapiro 2017). Thus, with lower levels of savings and especially of homeownership, minority families, especially black and Hispanic families, may have less stable housing and a greater need to form extended family households as an adaptive family strategy.

Racial and ethnic disparities with respect to financial well-being extend to kin networks as well (i.e., beyond individual or even household economic circumstances), a fact which may also influence living arrangements and the potential benefits an individual can reap from an extended-household structure. The middle class for blacks and Hispanics, for instance, is much smaller than it is for non-Hispanic whites and Asians (Goodman, Pendall, and Zhu 2015; Wheary 2006). Furthermore, even those minority individuals who achieve middle-class status are more likely than whites both to have and to live closer to poor kin (Heflin and Pattillo 2006; Pattillo 2013). Impoverished networks have been suggested as an explanation for racial and ethnic wealth gaps also; instead of saving and investing, it has been found that these minority groups are supporting their kin (O’Brien 2012). In the minority “culture of mobility” paradigm, it is suggested that those minority-group individuals who become middle-class must negotiate relations with poorer kin, limiting their further upward mobility (Chiteji and Hamilton 2002; Neckerman, Carter, and Lee 1999). Indeed, blacks and Hispanics have been found to take more responsibility for the welfare of their kin, expressed in the idea of “linked fate” (Dawson 1994). Upwardly mobile Mexicans in the U.S. middle class often feel obligated to take on the responsibility of supporting family members who are less well-off (Vallejo 2012).

All of these findings suggest that the structural barriers for minority groups involve not only an individual’s own socioeconomic status, but also the socioeconomic circumstances of his or her larger kin network. Put simply, if those around one have limited resources, one’s ability to rely on family in times of economic need is tenuous. Given their heightened sense of obligation, racial and ethnic minorities may be more likely to take family in, but, given blacks’ and Hispanics’ statistical likelihood of having poorer kin networks, the prospects for a minority individual to move her/himself or her/his kin out of poverty through sharing (already limited) resources may be less likely. Alternatively, fewer resources across kin networks may increase the economic need for both host and guest to form an extended-family household by gaining economies of scale. In such households, the formation of the extended-family household may represent an economic benefit to both family units, instead of serving as a safety-net for one disadvantaged family within the household.

I use longitudinal data to examine changes in families’ poverty status due to changes in their living arrangements. This is a significant contribution, because most studies simply compare poverty measures among those in different living arrangements through a cross-sectional lens, whereas using longitudinal data will provide a deeper understanding of the role that extended-family households play as part of a familial safety net. Specifically, the current study’s longitudinal data come from the nationally representative Survey of Income and Program Participation (SIPP) from the years 2008–2013, and the analysis examines how changes in extended-family living arrangements are associated with changes in poverty over time. Much of the SIPP data were collected during and immediately after the Great Recession, a time when many Americans were experiencing job loss and poverty (Nichols and Simms 2012). Therefore, this study captures the experience of relying on family during a time with limited economic opportunities.

Previous research has demonstrated the economic benefits and the prevalence of extended-family households among low-income groups (Pilkauskas et al. 2014); however, the extent to which individuals are able to alleviate material deprivation by moving in with family remains unclear. The current study will address that question more directly. Further, it will contribute to the literature on extended-family households by examining changes in poverty status for both guests and hosts by asking: Does forming an extended-family household change the poverty status for hosts and/or guests? Drawing on the family adaptive strategy theory (Moen and Wethington 1992), I hypothesize:

H1: Forming an extended-family household will reduce household poverty among guests.

However, drawing on previous research that suggests extended family households function primarily as a safety-net for family members in need:

H2: Forming an extended-family household will not change the poverty status of hosts.

This study expands upon previous research on racial and ethnic differences in poverty and living arrangements by addressing a second question: does change in poverty status vary according to race or ethnicity for individuals forming extended-family households, be they guests or hosts? Moving into an extended-family household can raise individuals out of poverty only if the collective income is sufficient. Moreover, the benefits experienced thereby may not be the same for all racial and ethnic groups. Drawing on theory and previous literature, I formulate competing hypotheses about the role of racial and ethnic differences in predicting the outcomes for those who move into an extended-family household as guests (H3) and as hosts (H4). Drawing on the minority culture of mobility paradigm (Chiteji and Hamilton 2002; Neckerman et al. 1999):

H3a: I hypothesize that black and Hispanic guests will experience a smaller economic benefit than non-Hispanic whites and Asians from moving into an extended-family household.

This is because, despite close family ties that pull together in times of need, the kin networks of blacks and Hispanics are much more limited than those of non-Hispanic whites and Asians. Given racial and ethnic differences in living arrangements and economic resources, I would expect whites and Asians who move in with family to experience larger poverty reductions than blacks and Hispanics, because the former groups’ family networks possess more resources.

H3b: On the other hand, given racial and ethnic differences in types of intergenerational assistance — with whites being more likely to receive monetary support as opposed to shared residences (Sarkisian and Gerstel 2004) — we may observe smaller reductions in poverty for whites compared to all other racial and ethnic groups upon the formation of extended-family households, because some poverty reduction for whites is already and/or simultaneously occurring from direct economic transfers.

H4a: Again drawing on the minority culture of mobility paradigm, as well as racial and ethnic differences in norms of familial support, I hypothesize that black and Hispanic hosts will potentially overextend their resources and thus experience an increase in poverty. More specifically, strong family norms that prize giving assistance, coupled with limited economic resources, may lead to a drain on the resources of black and Hispanic host families. Conversely, non-Hispanic white and Asian host families have more resources and may more easily absorb additional household members. Therefore, I hypothesize concurrently that white and Asian hosts will see no change in poverty when extended-family households are formed.

H4b: Alternatively, minority host families may see greater benefits because an increase in household size provides economies of scale, which offer a mutual advantage when resources are pooled. I hypothesize that the adaptive family strategies of blacks and Hispanics will include more mutual assistance than those of non-Hispanic whites, and therefore that black and Hispanic host families will experience larger reductions in poverty than non-Hispanic white host families.

METHODS

Data and Measures

I use the 2008 panel of the Survey of Income and Program Participation (http://www.census.gov/sipp/), a nationally representative longitudinal survey of the civilian noninstitutionalized population of the United States. Respondents are interviewed over a period of 56 months at four-month intervals. However, the information they provide at each interview reflects their circumstances month by month. SIPP collects information on all individuals in a household and attempts to follow all original sample members for the length of the survey. This unique feature allows for the study of changes in living arrangements over time.

At baseline, interviewers collected information for 105,303 respondents. The sample is restricted to individuals present at baseline, since only original sample members are followed for the duration of the survey. Respondents must be present for at least two waves to be included in the current analyses, thus reducing the sample to 95,577. Attrition rates between waves ranged from 6 percent to 12 percent, with a cumulative sample loss of 53 percent.¹ Because individuals do not need to be present for all waves to be included in the final sample, attrition bias is minimized.

I restrict my sample to adults aged 25 and older, since respondents aged 18–24 may have transitory living arrangements for a variety of reasons, such as moving into a college dorm and back; this limits the sample further to 63,608 individuals.² Respondents classified as “other” race are excluded as well, because they represent a small proportion of respondents (three percent), and attempting to treat and interpret a heterogeneous “other” race category as a cohesive group would not add to our understanding of racial and ethnic differences. The final analytic sample, then, is 60,208 respondents aged 25 and older, contributing, on average, 44 months or 11 waves.

Poverty

Poverty is a measure of economic deprivation. I use a modified version of the official poverty thresholds that includes cohabiting partners; previous research has referred to this modification of the official poverty measure as “social poverty” (Manning and Brown 2006). To be more specific, official measures of poverty are made at the family level and include all members in the household who are related by blood, marriage, or adoption. By contrast, the modified poverty variable defines families more broadly, such that it also includes cohabiting partners. Households in my analysis match closely with official family definitions for poverty, with two exceptions: (1) my analysis considers cohabiting partners also to be family, and (2) individuals who live in households comprised of non-related adults are each treated as their own unit. Many argue that poverty measures should include cohabitating partners’ income (National Research Council 1995), and research has found that many cohabitating partners do pool resources (Kenney 2008). Including cohabiting partners’ income also provides a more conservative measure of poverty.

I calculate poverty at the household level at each wave using poverty thresholds from the U.S. Census Bureau, which account for household size and age composition. Household income represents the sum of income for each member of the household and includes all household members’ earned income, social security income, pension income, asset income, and other income from government programs. Individuals in households where household income lies below the poverty threshold (again, considering household size and age composition) are coded = 1 (in poverty). I use this binary measure of poverty as a blunt way of measuring changes in economic deprivation, though results are similar using a continuous measure of the poverty ratio.³

Extended-Family Household Structure

In the United States, there is a growing proportion of households that expands beyond the traditional nuclear-family household structure. Here, I define household structure based on the number of minimal household units (MHU). The term “MHU” refers to the smallest economic decision-making unit in the household (Ermisch and Overton 1984), of which a household can have more than one. Using information on relationship to household reference person, I classify married or cohabitating couples and their unmarried, childless children younger than 25 as belonging to the same MHU. Co-residential children under the age of 25 who are currently married or have their own child represent their own family unit (or MHU) as well. I am not able to identify cohabiting unions among members who are not the householder, because only the relationship to the householder is recorded; however, only about six percent of all cohabiting unions do not include the householder (Kreider 2008). Among householders, about six percent report an unmarried partner living in their household. All other family members, such as parents, siblings, and aunts, are considered to be part of a separate family unit with their spouse and/or children (under age 25), if any. Therefore, two cohabitating parents and their children younger than 25 would be considered a single-family unit, whereas the addition of one of their parents would be considered a second family unit. This definition differs from that of the U.S. Census Bureau by including cohabitating couples and their children (under age 25) in one family unit, as well as by distinguishing among immediate and extended familial relationships. In so doing, the current study’s approach is similar to those used by other researchers (Glick and Van Hook 2011; Van Hook and Glick 2007).

Hereafter in this paper, the terms “MHU” and “family” will be treated as interchangeable; an individual is coded as living in an extended-family household if the household contains more than one related family unit or MHU. This definition expands the traditional definition of family because, by this definition, a “family” can consist of one adult (given that a single adult without children can be an MHU).

In the formation of an extended-family household, there are two simultaneous components: those allowing others to move into their home, and those agreeing to move into a home of which they are not the owner/renter of record. The formation of such a household could benefit the host, the guest family, or both. Therefore, I differentiate individuals within extended-family households as members of the host family or members of the guest family in the analysis. Families are considered host families when their family unit contains the householder (i.e., the owner/householder of record).

This analysis focuses on changes in extended-family household structure over time for individuals. Changes occur through both the formation and dissolution of extended-family households. Sensitivity analysis finds that results are similar for formation and dissolution. Therefore, the interpretation of the results focuses on forming extended-family households, though they could be interpreted from the perspective of dissolution as well.⁴

Because the definition of a family unit for this study might (or might not) consider an adult child to be an independent family unit, depending on his/her age, parenthood status, and marital status, some changes in household structure occur over the course of the survey even where no one has moved. That is, a small number of households in the sample experience a change in household members without anyone changing residence. A single family could become an extended-family household, for instance, if a child under age 25 becomes a parent; the householder’s child (who has not moved) and grandchild (who has not lived elsewhere previously) would be considered their own family in the now-extended-family household (N = 29). Sensitivity analysis excluding these respondents indicates that results are robust to the exclusion of these non-household composition changes in extended-family household formation.

Control Variables

In the multivariate analysis, I include controls for time-varying factors that may also be associated with changes in poverty. At the individual level, these controls include employment and marital status. Employment is a measure of “employed” versus “not working.” “Marital status” is a categorical variable: married, single, divorced, widowed, or cohabitating. At the household level, the number of children under age 18 is included as a continuous measure;⁵ a change in the household’s number of non-adult children may change the household’s poverty status because it changes the poverty income threshold (without changing living arrangements).

Race and ethnicity are included as interaction terms in the models, with the following categories: non-Hispanic white, non-Hispanic black, non-Hispanic Asian, and Hispanic. Race and ethnicity capture the individuals’ race and ethnicity, which may differ from other members in the household, including that of the host or guest family in extended-family households. Although, in about 90 percent of extended-family households the race and ethnicity of other residents is the same as that of the householder.

Analysis Strategy

I use linear fixed effects regression models to estimate how a change in living arrangements is associated with a change in poverty for individuals. Although the outcome is a binary variable, I use linear fixed effects to include in the estimates those who do not change in the dependent variable, to avoid potential bias (Miller, Shenhav, and Grosz 2019).

Individuals are nested within families, which are then nested within households; however, membership in both a family unit and a household can change during the survey. Because family and household composition change over time, analysis is conducted at the individual level to capture how the formation and dissolution of extended-family households are related to individual-level changes in poverty status. All analyses are conducted at the individual level, though some of the measures draw on household-level characteristics; in such cases, these characteristics are attributed to all members of the household.

Those who move into extended-family households likely differ in both measured and unmeasured ways from those who remain in single-family households. Fixed effects models attempt to remove selection bias by using respondents as their own controls. In other words, the fixed effects method focuses on change within individuals rather than between them, effectively controlling for all observed and unobserved time-invariant characteristics, such as gender, ideology, and personality traits (Allison 2009). This method is useful for this research question, as one’s household organization is often attributed to an underlying cultural preference, a theoretical variable for which no measures are available in the data. Assuming that one’s cultural inclination towards a given living arrangement is a stable characteristic, the fixed effects method controls for it. The general model specification is as follows:

$$y_{\mathit{it}}={\mu }_t+{\beta x}_{\mathit{it}}+\mathrm{\gamma }{z}_i+{\alpha }_i+{\mathrm{\epsilon }}_{\mathit{it}}$$

µ_t represents purely random variation at each point in time, and X_it is a vector of time-varying covariates. The terms z_i and α_irepresent time-invariant predictors (that is, both observed and unobserved individual traits), while ${\epsilon }_{\mathit{it}}$ represents random variation at each point in time for each individual. These models use change in the independent variables to predict change in the dependent variable. All time-invariant variables are controlled implicitly in the fixed effects models; however, these models do not remove the biasing effects of unmeasured variables that do change over time (Allison 2009). Health of family members is one example of a variable that may change over time, and previous research has found this to be related to co-residence (Choi 2003). Moreover, another limitation of fixed effects is the possibility of reverse causality. Some potential implications of these limitations are expanded in the discussion.

Models are estimated to assess the role that the formation of extended-family households plays on poverty. The first of the models estimates a change in poverty following the formation of an extended-family household for both host family and guest family, accounting only for survey time to assess the baseline relationship. The second model introduces the time-varying covariates, employment and marital status. I am interested in whether a relationship between a time-varying covariate (household composition) and poverty status differs over time. In the third model, I introduce interaction terms between race and ethnicity and extended-family household formation to assess how race and ethnicity moderate the relationship.

All models are weighted using person weights, and standard errors are adjusted to account for the complex survey design, the clustering of individual observations over time, and the clustering of individuals within households. Because fixed effects models focus on changes, the models work best when a significant amount of change occurs. An examination of the data shows that such variation does exist; during the sample period, about 14 percent of the sample will experience a change in household structure from a single-family household to an extended-family household.

Some evidence of seam bias has been found in the SIPP data. Seam bias occurs because individuals are more likely to report changes that occur in the months when their interviews take place, so transitions are lumped at four-month intervals. However, in the more recent panels of SIPP, although seam bias still exists, it is lower than in earlier panels, due to probes about the specific month in which events occurred over the preceding four (Moore 2008). To account for this bias, I include a dummy variable for interview month (Ham, Li, and Shore-Sheppard 2007).

RESULTS

Table 1 presents descriptive statistics for key variables in the first wave of the survey for the sample as a whole, as well as by racial and ethnic groups. For all groups, single-family households are the most common household structure, though the prevalence varies significantly. Non-Hispanic whites are the least likely to be in extended-family households (13.8 percent), with much higher proportions of blacks, Asians, and Hispanics living in extended-family households (27.9 percent, 29.0 percent, and 31.7 percent respectively). Poverty rates across racial and ethnic groups differ as expected, with non-Hispanic whites reporting the lowest levels of poverty at 9.1 percent, followed closely by Asians at 12.7 percent, and blacks and Hispanics having poverty rates over 20 percent.

Table 1.

Descriptive Statistics by Race and Ethnicity at Wave 1 

	NH White	NH Black	% or mean NH Asian	Hispanic	Total
Extended Family Household	13.8	27.9	29.0	31.7	17.8
Poverty	9.1	21.7	12.7	23.7	12.1
Income-Poverty Ratio	449.0	293.1	481.1	252.8	413.4
Employed+	76.1	66.9	74.1	68.2	74.0
Homeownership	80.2	56.7	65.6	54.8	74.4
Proportion Ever Change
Enter Poverty	25.1	37.3	30.4	47.1	28.9
Exit Poverty	23.3	36.9	30.7	44.3	25.2
Move into Extended Family HH	13.1	16.3	13.0	19.3	14.1
N Persons	44,903	6,719	2,565	6,021	60,208

Notes: Descriptive statistics were weighted and accounted for complex survey design. NH=Non-Hispanic +Proportion Employed is restricted to respondents aged 25-64.

During the nearly five years of the survey, transitions into and out of poverty are common, with close to 29 percent of respondents entering and 25 percent exiting poverty. Non-Hispanic whites are least likely of all racial and ethnic groups to enter (25 percent) or exit poverty (23 percent), while almost half of Hispanics experience a change in poverty status.

As noted above, 14 percent of the sample is observed moving into extended-family households during the survey. Racial and ethnic differences in the transition into extended-family households mirrors racial and ethnic differences in the proportion of individuals in extended-family households, with one exception: despite the high proportion of Asians living in extended-family households in the first wave, their transition rate into extended-family households during the survey is relatively low (13 percent). This finding implies that single-family households are perhaps more stable among Asians. Meanwhile, Hispanics experience the highest transition rate into extended-family households (19 percent).

To test the hypothesis that forming an extended-family household reduces poverty for guests (H1) and not for host families (H2), I estimate linear fixed effects regression models predicting changes in poverty status by forming an extended-family household structure (Table 2). Model 1 shows that the formation of an extended-family household subsequently reduces the probability of living in poverty, both for those who are moving into the household and, to a lesser extent, for the host family. More specifically, transitioning from a single-family household to an extended-family household reduces the probability of poverty by 13 percent for movers and 3 percent for host families. The significant reduction in the likelihood of poverty among host families suggests that some poverty reduction is derived either through a mutual benefit and joining of incomes, which pushes the household above the poverty threshold, and/or through (in some cases) a higher-income family joining a poor household. That said, the reduction in the probability of poverty is much larger among guests who move into extended-family households.

Table 2.

Fixed Effects Regression of Poverty by Extended Household Formation

	Model 1		Model 2		Model 3
Survey Month	0.00***	(0.00)	0.00	(0.00)	0.00	(0.00)
Seam Indicator	0.00***	(0.00)	0.00***	(0.00)	0.00***	(0.00)
EF Host	−0.03***	(0.00)	−0.03***	(0.00)	−0.02***	(0.00)
EF Guest	−0.13***	(0.01)	−0.15***	(0.01)	−0.14***	(0.01)
Employed			−0.16***	(0.00)	−0.16***	(0.00)
Number of Children			0.02***	(0.00)	0.02***	(0.00)
Marital Status (Ref=Never Married)
Married			−0.08***	(0.01)	−0.08***	(0.01)
Divorced			0.01*	(0.01)	0.01*	(0.01)
Widowed			0.00	(0.01)	0.00	(0.01)
Cohabitating			−0.10***	(0.01)	−0.10***	(0.01)
Race by Family Extension Interactions
Black EF Host					−0.04***	(0.01)
Asian EF Host					−0.04**	(0.01)
Hispanic EF Host					−0.02**	(0.01)
Black EF Guest					−0.04*	(0.01)
Asian EF Guest					−0.01	(0.03)
Hispanic EF Guest					−0.00	(0.02)
Constant	0.12***	(0.00)	0.26***	(0.01)	0.26***	(0.01)
N Persons	60,208

Notes: EF=Extended Family; Standard errors in parentheses. All models use person weights and adjust standard errors for clustering; * p<0.05 ** p<0.01 *** p<0.001 (two-tailed tests).

Model 2 indicates that this effect is robust to changes in family composition, marital status, and employment. Changing from not working to being employed is associated with a 16 percent reduction in the probability of poverty. Meanwhile, an increase in the number of children increases the probability of poverty. For families already on the margin, having a child is unlikely to be associated with an increase in income; therefore, since the poverty thresholds are based on adjusted per-person income, having a child may increase poverty for some families. A change in marital status to married or cohabitating reduces the probability of poverty, while becoming divorced increases that probability slightly.

Having established that moving into an extended-family household is associated with a reduction in the likelihood of poverty, I next examine racial and ethnic differences (H3a-H4b). Model 3 adds an interaction term between extended-family household composition (for hosts and guests) and race and ethnicity to test whether the reduction in the probability of poverty resulting from a change in extended-family household status is consistent across racial and ethnic groups. The results indicate some racial and ethnic differences, but mostly among extended-family hosts. Specifically, black, Asian, and Hispanic host families have larger reductions in the probability of poverty from taking in relatives than do non-Hispanic white host families. This finding suggests that these racial and ethnic minority families are more likely to benefit from the pooling of incomes and resources brought into the household from guest families than are non-Hispanic white host families.

Among those moving into an extended-family household, black guests have a larger reduction in the probability of poverty than do non-Hispanic white guests. Asian and Hispanic guests, meanwhile, have similar poverty reductions as non-Hispanic white guests. All in all, all racial and ethnic groups experience a significant reduction in the probability of poverty to varying degrees, with larger reductions for the families that move in rather than for host families.

Previous research demonstrates that changes in employment status have the strongest associations with changes in poverty status (McKernan and Ratcliffe 2005), and, indeed, employment status is one of the largest coefficients in all the models thus far. In order to isolate the effect of changes in extended household formation from changes in employment, as a further robustness check, supplemental analysis was conducted on a sub-sample of respondents who were continuously employed and finds similar results (available upon request).

The robust association of a lower likelihood of poverty with changes into extended-family households begs the question of what strategy underlies this change in poverty. A family that becomes part of an extended-family household can change its poverty status in two primary ways: first, it could be that both families are poor prior to the move but gain in the economies of scale in the way poverty is measured, such that, once combined into a single household, they are simply re-categorized as non-poor. Second, a poor family can join with a non-poor family, such that now, once combined into one household, the previously poor family is no longer in a poor household. These are two very different pathways out of poverty but are not distinguishable in the analysis thus far. Hence, among those observed exiting poverty and transitioning into an extended-family household, I examine the poverty status of each family within the household and how such poverty transitions vary by race and ethnicity (Figure 1). If all families in the household would be poor if living on their own, then a decrease in the probability of poverty for them occurs through economies of scale and income pooling; whereas if only one family in the extended-family household is poor and the other family is not poor, then this suggests that the extended family household is operating as a family safety net. The last possibility is that neither family unit in the extended-family household would be poor on its own and that one family simultaneously exited poverty and moved into an extended family household.

Figure 1.

Family Poverty in Extended Family Households That Exit Poverty

The most common family poverty structure observed across all racial and ethnic groups is a safety net organization, in which one family within the household would be poor on their own but is no longer in poverty through the added resources of the non-poor household (see the solid gray bar in Figure 1). In most households, it is the guest family that would be poor if on their own, while less than 10 percent of host families would be in poverty if not for non-poor guest family (results not shown). These results echo fixed effects results that demonstrate consistently larger reductions in the changes of poverty among guest families moving into the extended-family household, with the guest family likely being in poverty if on their own, and the host family providing shelter and economic resources sufficient to boost the former household out of poverty. This pathway out of poverty is especially common among Asians and Hispanics, with about two-thirds of Asian and Hispanic households operating as a safety net. This may suggest these groups have a different cultural norm that includes helping family, especially recently arrived immigrants that commonly live in extended family households. Overall, the high proportion of safety-net households suggests that extended-family living arrangements occur primarily to the economic benefit of one family unit within the larger family structure.

A less common pathway out of poverty is for families to gain economies of scale, since all the family units within their household would be classified as poor if they lived on their own. Gaining economies of scale (solid black bar on Figure 1) is most common among blacks (16.7 percent) and least common among Asians (7.0 percent). Households where economies of scale are the primary pathway out of poverty may form under different expectations than safety-net households, given that both families may have expected to benefit from the arrangement. Not only are blacks more likely to move out of poverty through forming extended-family households (compared to non-Hispanic whites); among those black families that do move out of poverty, they are more likely to do so through economies of scale. This suggests that structural forces may make extended-family households more attractive to blacks, given the need to pool incomes to live above official poverty thresholds. That said, for all groups, some amount of changes in poverty and household composition may be occurring simultaneously.

DISCUSSION

Overall, forming extended-family households does appear to be a strategy that helps individuals move out of poverty. For all racial and ethnic groups, a move into such a household reduces poverty significantly, supporting the first hypothesis (H1) that moving into an extended-family household is an economic strategy and that it can help alleviate poverty. This finding supports a family adaptive theory that the increase in extended-family households that has been occurring since the 1980s reflects how families are coping with increased economic uncertainty. In an era of reduced welfare support, familial support networks may become more important than ever to the survival of low-income families (Scott et al. 2004).

However, the benefits of forming extended-family households are not the same for hosts as for guests. The formation of extended-family households is a two-way decision process that requires both the guests and the hosts to agree to the arrangement. No support is found for the second hypothesis (H2) that speculates that host families will not benefit from forming extended-family households. Host families experience a small reduction in poverty from the formation of extended-family households, which suggests that most host families are also benefiting from this arrangement, either through pooling incomes or freeing up workers in the household from the burden of extra childcare or eldercare costs. The results do demonstrate that reductions in poverty are significantly smaller for the host family than for those moving in.

Results do not support the hypothesis that black and Hispanic guests will experience a smaller reduction in poverty than non-Hispanic whites and Hispanics (H3a). Rather, the results provide mixed support for the competing hypothesis (H3b): black guest families experience a larger reduction in poverty than whites. However, Hispanic and Asian guest families experience a similar reduction in the likelihood of poverty as whites. Strong norms of family assistance may account for the larger poverty reduction among blacks. Although blacks and Hispanics have more impoverished networks, this does not appear to limit their ability to move out of poverty through forming extended-family households. Moreover, we may simply see lower reductions in the probability of poverty from the formation of extended-family households among Hispanics and Asians because of alternative motivations for forming extended-family households; for these groups, stronger filial norms for assistance across generations may motivate some moves, rather than poverty (Burr and Mutchler 1999). Further, if whites are reducing the probability of poverty through economic transfers rather than through forming extended families, we may also see smaller effect sizes for whites compared to other groups (Sarkisian and Gerstel 2004). Also, whites may move in with family when income is reduced but still above poverty, though sensitivity analysis suggests that racial and ethnic differences are similar when income–poverty ratio is examined. Finally, results of those who do move out of poverty show that Asian and Hispanics are more likely to operate as a safety net with one family unit in poverty, suggesting a strong sense of obligation to family may motivate this household dynamic.

Black, Asian, and Hispanic host families experience a larger reduction in the probability of poverty than white host families, supporting the competing hypothesis (H4b). There is no evidence to suggest that racial and ethnic minority groups would be more likely to overextend their resources upon the formation of an extended-family household than non-Hispanic whites (H4a). One possible reason that racial and ethnic minority host families experience greater reductions in the probability of poverty is that these groups may be more likely to draw mutual benefit from extended-family households, rather than in the familial safety-net arrangement. In particular, pooling income across families may help all individuals involved to gain economies of scale, thus moving them above the poverty threshold. While reliance on income pooling to move out of poverty may provide a net positive, it also signifies that these households are still in a precarious economic situation. This disadvantage among minorities exacerbates educational and labor force inequities that dampen mobility and may even reproduce educational deficits among young adults, who are unable to rely on family to help extend their education because of their need to contribute financially to the household. Moreover, blacks’ larger reduction in the likelihood of poverty from forming extended-family households may reflect, in part, lower rates of marriage for this demographic and increased reliance on parents/grandparents for support, rather than on spouses (Roschelle 1997; Taylor, Chatters, and Celious 2003). Additionally, there may be assistance that flows from guests to hosts more often in racial and ethnic minority extended-family households, as opposed to from hosts to guests in non-Hispanic white extended-family households.

Although black, Asian, and Hispanic host families achieve a significantly larger reduction in the probability of poverty compared to non-Hispanic whites, even among these groups, the guests moving into the extended-family household still have a 3–4 times larger probability of moving out of poverty from the formation of an extended-family household than the hosts do. This finding, coupled with the analysis on extended-family households’ pathway out of poverty, supports the idea that a majority of these extended-family households are forming as part of a family safety-net system that provides relief to relatives in poverty. While informal family support was evident for all racial and ethnic groups, the level of benefit may vary; moving into a household barely above poverty is likely not the same as moving into a household living at four times the poverty threshold. Ultimately, it appears that all groups use extended-family households to some degree to help provide at least temporary relief from living in poverty and that families are resilient and do, indeed, pull together in times of need by providing housing to poor relatives.

The current results support previous research, especially qualitative work that has described moves into extended-family households as an economic strategy (Newman 2012). Individuals, especially young adults, are moving in with family as an economic strategy to lift themselves out of poverty, either because the host family is providing a safety net or because the household is pooling the collective resources of all members to help everyone move out of poverty.

Most extended-family households that exit poverty still have at least one family unit that would be in poverty if living on their own. One might argue that this result is an artifact of the poverty threshold scales or represents specialization within the household, the significance depends on whether individuals have equal access to the income of others in the household. Indeed, among entirely related individuals, it is reasonable to assume that some degree of resource sharing is occurring, even if not entirely equal. Pooled resources will go further, and extended-family households, on average, likely provide real economic benefits to family members; however, the degree of sharing and benefits likely varies across families.

It is important to emphasize that the benefits that this study identifies from moving into an extended-family household are represented only as measurable changes in total household income; in reality, there may be other attendant complications to extended-family living that erode or enhance quality of life. Moving into an extended-family household may provide economic benefits in the aggregate — theoretically reducing household poverty for many — this does not mean that such a move is a happy experience for all. Extended-family households may be physically and psychologically taxing. Families’ ability to buffer individuals from the harshest elements of economic deprivation by sharing households does not extend to an ability to shelter them from all hardships, either outside or inside the home.

Reductions in household-level poverty, furthermore, may mask the economic hardships of individual household members. Considering the role of family in the reproduction of inequality, while extended family households may provide a net benefit, they may also limit the accumulation of wealth and upward mobility for some family members. Additionally, the ability to rely on family to weather economic rough patches is contingent both on having family and on having family that is willing to share a household. Therefore, this option may not be available to all, and, in fact, may not be available to the most vulnerable segments of the population. For instance, blacks are less likely to have living parents to move in with, compared to non-Hispanic whites (Daw, Verdery, and Margolis 2016).

Furthermore, because the official poverty threshold may mask economic hardships within extended-family households, the need for poverty-reducing social programs may be underestimated. When poverty guidelines are used as eligibility criteria and all household members’ income is factored into these calculations, impoverished individuals’ eligibility for some social programs, such as SNAP, may be jeopardized. In these cases, moving into an extended-family household may demand a tradeoff between family assistance and government assistance. Some scholars argue that an alternative measure of poverty is needed to expose hidden poverty and to redirect the policy conversation away from marriage promotion and towards employment and wage-improvement policies instead (Meulders and Dorchai 2013).

One of this study’s limitations is the omission of time-varying covariates; for instance, changes in the health status of family members. Poor health is related to both the increased risk of poverty and to higher rates of co-residence. I am unable to estimate the role of health as the underlying mechanism, or as the reason, for poverty and subsequent moves into extended-family households. Aside from assuming a static health status, the second assumption implicit in these models is that a move into an extended-family household has preceded the change in poverty; that is, that there is no reverse causation. For instance, individuals could change in income (and associated poverty level) and then move into an extended-family household. The attempts of supplemental analysis to gain traction on this issue by estimating models with lagged dependent variables and limiting the sample to only those continually employed still find that changing from a single-family household to an extended-family household resulted in a reduction in poverty. Still, this does not entirely rule out the possibility of reverse causation.

Other potentially pertinent information on housing tenure is also not accounted for in the analysis. During and immediately following the recession, there was an increase in foreclosures, but information on foreclosures is not available in the data. Although, these data overlap with part of the Great Recession, a time when foreclosures were more common and poverty rates increased, sensitivity analyses did not find a difference in the timing of the effects during/post-recession (available upon request).

Another limitation is that these results do not account for how long individuals remain in extended-family households. Previous research has found extended-family households to be less stable, but I am unable to assess whether the formation of these households represents a short-term reprieve from poverty or a long-term strategy for poverty reduction. Finally, the poverty calculations at the household level assume resource sharing among all related adults in the household; in reality, actual resource sharing in the household is unknown.

Despite these limitations, this study advances our understanding of the relationship between poverty and living arrangements, and it demonstrates one adaptive family strategy for allowing individuals to move out of poverty: co-residence. At the same time, this paper raises several questions about what the true poverty rate would be, without government thresholds assuming an equitable redistribution of resources within each household. The current study’s results also demonstrate that differences exist across race and ethnicity in the ability to reduce poverty through the formation of extended-family households. Black respondents experience some of the largest benefits from moving in with family, likely due to the strained economic opportunities and familial norms regarding support for the demographic (Roschelle 1997).

Future research should investigate how this process unfolds over a longer period of time for individuals and families. Does a temporary moving-in with family allow individuals to get on their feet, such that, even upon venturing out on their own, they are out of poverty — or does it only add to the many poverty transitions that individuals experience throughout their lifetimes? On the other hand, for some groups, perhaps these moves become long-term solutions to avoid poverty. Given gender differences in the effects of the recession as well as racial and ethnic differences in single-parenthood, future research should examine the economic benefits of extended-family formation at the intersection of gender, parenthood, and race and ethnicity.