Neighborhoods, Cities, and Economic Mobility

Abstract

Most of the research literature explaining the level of economic mobility in the United States focuses on characteristics of individuals or families. This article expands the focus beyond the individual and the family to consider features of communities and cities. Although evidence is strong that features of neighborhoods and cities have causal effects on individual economic mobility, there is much less evidence on the most relevant mechanisms. The article reviews the available evidence at both levels of analysis before concluding with a discussion of the implications for social policy.

Although the literature on intergenerational economic mobility in the United States has advanced considerably over time, less progress has been made in explaining the mechanisms leading to the persistence of economic status across generations. Empirical research designed to explain why economic advantage and disadvantage tend to be transmitted from parents to children has focused on characteristics of individuals or families. This article expands the focus to consider features of communities and cities.

This focus on the spatial foundations of economic mobility in the United States is based on two basic claims (Sharkey and Faber 2014). The first is that systems of stratification are organized, in part, along spatial lines. This claim is uncontroversial. The spatial organization of American social, economic, and political life is reflected in patterns of discrimination and segregation, zoning decisions, the establishment of boundaries for political districts, school catchment areas, and police precincts, the siting of environmental hazards, and the location decisions of public institutions and private firms (Dreier, Mollenkopf, and Swanstrom 2001). The second claim is that the spatial organization of America’s stratification system affects the life chances and the economic trajectories of different segments of the population in ways that maintain, and reinforce, inequality. This claim has been the subject of more vigorous debate, and the empirical evidence that has been generated to support or refute this claim is the focus of this article.

The article is guided by three questions: How do residential contexts affect prospects for mobility? How do cities and metropolitan areas affect economic mobility? What are the implications for social policy?

HOW DO RESIDENTIAL CONTEXTS AFFECT PROSPECTS FOR MOBILITY?

Only a few studies have analyzed the relationship between childhood neighborhood conditions and adult economic outcomes, primarily because there are few datasets that follow sample members across multiple generations. Most of the observational studies that allow for cross-generational analysis draw on data from the Panel Study of Income Dynamics (PSID) and find an association between measures of neighborhood economic status during childhood and adult economic status after adjusting for observed individual and family characteristics, although the strength of the association varies widely depending on the methods used, the specific neighborhood measures considered in the analysis, the outcome under study, and the subpopulations examined. Studies conducted by Linda Datcher (1982), Mary Corcoran and Terry Adams (1999), Corcoran and her colleagues (1992), Steven Holloway and Stephen Mulherin (2004), and Thomas Vartanian (1999) report conditional associations between neighborhood economic status and adult outcomes related to employment and income.

Taking a different approach, Daniel Aaron-son (1998) and Vartanian and Page Buck (2005) exploit variation in childhood neighborhood conditions experienced by siblings within families and find significant effects of neighborhood economic status on adult educational and economic outcomes. However, Robert Plotnick and Saul Hoffman (1999) conduct a similar analysis with a sample of sisters in the PSID to study outcomes related to welfare receipt and fertility, and find null effects of childhood neighborhood conditions when using family fixed-effects specifications.

A much larger literature examines how neighborhoods affect some of the key mechanisms influencing later economic outcomes such as academic success, cognitive skills, and educational attainment. Several of these studies have found a strong association between different compositional characteristics of children’s neighborhoods, such as the level of neighborhood poverty, the presence of affluent neighbors, and rates of residential mobility, and individual outcomes like dropping out of high school or scores on assessments of cognitive skills (Leventhal and Brooks-Gunn 2000; Sastry 2012). David Harding’s (2003) study of the effect of neighborhood poverty on high school dropout is one example of a carefully designed observational study that analyzed matched pairs of children who look extremely similar in every aspect of their lives but their neighborhood. Harding estimated that living in a high-poverty neighborhood during adolescence doubles the likelihood that a child will drop out of high school relative to living in a low-poverty neighborhood among both blacks and whites. His findings were found to be robust to a conservative sensitivity analysis.

Harding’s study also is unique because he measured neighborhood conditions of children over an extended duration of childhood. Several descriptive studies have documented the persistence of neighborhood advantage and disadvantage over long periods and across generations, suggesting the need for a greater focus on the temporal dimensions of exposure to neighborhood poverty (Briggs and Keys 2009; Quillian 2003; Sharkey 2008; Timberlake 2007).

A set of recent studies has used methods that adjust for time-varying confounders to estimate the cumulative consequences of exposure to neighborhood disadvantage on academic and cognitive outcomes. Robert Sampson, Patrick Sharkey, and Stephen Raudenbush (2008) find that exposure to concentrated disadvantage altered the development of cognitive skills of African American children in Chicago, with consequences that persist years after exposure to neighborhood disadvantage. Geoffrey Wodtke, David Harding, and Felix Elwert (2011) find that exposure to concentrated disadvantage over the course of childhood reduces the probability of high school graduation by 20 percentage points for black youth and by 10 percentage points for all other youth. Sharkey and Elwert (2011) use a similar approach but look further back into families’ histories, and find that exposure to neighborhood poverty over consecutive generations reduces children’s performance on tests of cognitive skills by between 8 and 9 points, more than half of a standard deviation. A formal sensitivity analysis demonstrated that the effect of multigenerational neighborhood poverty is robust to high levels of potential bias arising from unobserved selection processes. The common conclusion reached by these studies is that the effect of neighborhood disadvantage on cognitive and academic outcomes is more severe if disadvantage is persistent, experienced over long periods of a family’s history.

Evidence from Housing Mobility Programs

A second strand of evidence comes from studies that exploit quasi-experimental or experimental changes in families’ neighborhoods and schools arising from low-income housing assistance programs (Briggs 1997; DeLuca and Dayton 2009). Among the many residential mobility programs that have been studied in the literature, the two most prominent examples are the Gautreaux Assisted Housing Program in Chicago and the Moving to Opportunity program, which was conducted in five U.S. cities.

Gautreaux was the result of a desegregation settlement that required the Chicago Housing Authority to provide housing to eligible families in neighborhoods across the Chicago metropolitan area. Specific units were identified across a range of neighborhoods that included the affluent and predominantly white suburbs surrounding Chicago, and families were offered specific units at least partly on the basis of their position on the waitlist for housing. Early studies based on samples of families that moved in the Gautreaux program found that families moving outside Chicago’s city limits experienced substantial changes in adults’ economic outcomes and children’s educational attainment (Kaufman and Rosenbaum 1992; Rosenbaum and Popkin 1991; Rubinowitz and Rosenbaum 2000). For instance, 54 percent of children in families who moved to the suburbs attended any college, against 21 percent of children in families who remained in the suburbs (Rubinowitz and Rosenbaum 2000).

However, subsequent research has questioned whether the changes in neighborhood conditions induced by the program should be thought of as exogenous. Mark Votruba and Jeffrey Kling (2009) document a correlation between the characteristics of Gautreaux families’ origin neighborhoods and their destination neighborhoods, suggesting that families’ preferences played at least some role in determining the neighborhoods to which they were assigned. Subsequent research on families in the Gautreaux program adjusts for observed differences between families and finds that caregivers were more likely to remain in the labor force and and earn higher wages when they left the deeply segregated, high-poverty neighborhoods of Chicago and moved to more integrated, less-poor communities across the metropolitan area (Mendenhall, DeLuca, and Duncan 2006).

Motivated in part by the strong findings from Gautreaux, the Moving to Opportunity Program (MTO) was a social experiment conducted in five cities—Baltimore, Boston, Chicago, Los Angeles, and New York—to test whether moving into low-poverty neighborhoods affected the social and economic outcomes of families living in areas of concentrated poverty (Briggs, Popkin, and Goering 2010; Goering and Feins 2003; Sanbonmatsu et al. 2011). In each of the cities, families in designated public housing developments who volunteered for the program were randomized into one of three groups: an experimental group that received housing vouchers that could only be used to rent in low-poverty neighborhoods; a Section 8 group that received standard Section 8 vouchers without requirements on where the voucher could be used; and a control group that received no voucher at all.

The results from MTO are complex and difficult to summarize. The most recent reports have found that ten to fifteen years after the initial random assignment, adults in the experimental group experienced substantial improvements in mental and physical health and overall subjective well-being, but no improvements in economic outcomes related to labor force participation or income (Ludwig et al. 2012). The impact of the program on children appeared to vary by gender. Girls in the experimental group experienced improved mental health and were less likely to participate in some risky behaviors, but boys showed few changes in their lives as a result of the program but increases in some risky behaviors (Clampet-Lundquist et al. 2011; Sanbonmatsu et al. 2011).

This crude summary of results from MTO obscures an even more complex set of findings that has emerged in the five sites at different times following the implementation of the program. A review of studies that report outcomes related to cognitive and academic skills reveals an erratic set of findings that vary across the five cities and across subgroups within the cities. To begin with, analyses that pooled all children across the five cities found no effects on cognitive skills either four to seven years after the program began or ten to fifteen years after (Sanbonmatsu et al. 2006; Sanbonmatsu et al. 2011). However, African American children in the experimental group scored 0.08 standard deviations higher than African American children in the control group on an assessment of broad reading skills conducted four to seven years after random assignment (Sanbonmatsu et al. 2006). Further, among families who remained in low-poverty neighborhoods for a longer time, Margery Turner and her colleagues (2012) find positive effects of moving to low-poverty neighborhoods on both reading and math scores for boys and girls. Raj Chetty, Nathan Hendren, and Larry Katz (2015) find that younger children in the experimental group may have fared much better than children whose families moved at an older age. Children who were younger than thirteen when their families received vouchers were 2.5 percentage points were more likely to attend college than children in the control group and made roughly $1,600 more per year in income in their mid-twenties.

Studies focusing attention on samples of families from specific cities have generated even more divergent findings. In the New York site, Tama Leventhal and Jeanne Brooks-Gunn (2004) find no overall effects of the program after three years following implementation, but positive effects on assessments of cognitive skills for boys. Subsequent research on the New York City sample finds negative effects of moving to low-poverty neighborhoods ten to fifteen years after the program implementation (Sanbonmatsu et al. 2011). Research on the Baltimore sample finds strong effects on children’s test scores four to seven years after random assignment that were no longer present ten to fifteen years after the program was implemented (Burdick-Will et al. 2011; Ludwig, Ladd, and Duncan 2001; Sanbonmatsu et al. 2011). Research on the Chicago site, on the other hand, documents similarly strong effects on children’s test scores four to seven years after random assignment, and smaller effects that persisted through the latter follow-up ten to fifteen years after the program was implemented (Burdick-Will et al. 2011; Sanbonmatsu et al. 2011).

Making sense of these conflicting findings is challenging because of the nature of the MTO experiment and the variation in its implementation and impact across the five cities (Briggs, Popkin, and Goering 2010). Research examining where families moved demonstrates that in some cities families in the experimental group moved into areas of the city in proximity to communities of families who received no vouchers and were in the control groups (Sampson 2008). In other cities, families assigned to the experimental group experienced much more substantial changes in their residential environments, but these changes were short lived (Clark 2008). In the latest follow-up, conducted ten to fifteen years after random assignment, families in the experimental group lived in neighborhoods with poverty rates just 3 percentage points lower, on average, than families in the control group (Ludwig et al. 2012).

Making general conclusions from MTO becomes even more difficult when one considers the changes taking place in major U.S. cities at the same time as the experiment. The intervention was implemented at a time when employment opportunities were expanding rapidly in high-poverty communities, welfare reform was being implemented, public housing was being demolished in many cities around the country, and violent crime was just beginning to decline after several decades of rising violence in central cities (Sharkey 2013). None of these observations make MTO any less useful for understanding the effects of a policy designed to move families into lower-poverty communities. However, it is important to consider MTO in the context of its time, and to consider results from MTO alongside those from Gautreaux and the many other housing mobility programs that have been implemented and studied over time.

Many of these other housing mobility programs have generated evidence suggesting that moving out of highly disadvantaged communities can lead to positive effects on children’s academic trajectories and economic outcomes. George Galster, Anna Santiago, and Jessica Lucero (2015a, 2015b) analyze data from Denver County’s “Dispersed Housing Program,” in which low-income families were offered specific housing units based on the family type and what was available when the family reached the top of the list. The authors argue that the unique nature of this housing assignment process created exogenous variation in the locations of units offered to families. They find that children from families moving to neighborhoods with higher crime, greater social problems in the community, and lower average socioeconomic status have worse developmental, health, education, and early labor market outcomes.

Jens Ludwig and his colleagues (2010) analyze data from housing assistance recipients in Chicago who were randomly assigned a position on a wait list when the local housing authority opened this wait list for the first time in years. Exploiting variation in the timing of when families were offered housing in lower-poverty neighborhoods, the researchers found that children offered housing vouchers scored 0.05 standard deviations higher on reading scores and 0.08 standard deviations higher on math scores than children in the control group. Because most families offered a voucher did not actually move, the estimated effects of moving into lower-poverty neighborhoods among those families that moved are substantially larger.

Douglas Massey and a team of researchers analyze the outcomes of families who were able to move into a new housing development in the Philadelphia suburb of Mt. Laurel, New Jersey (2013). Matching families that moved into the new housing development with families on the waiting list who were not offered housing, Massey and his colleagues find that those who moved had higher earnings and employment rates than those who did not, but that moving had no effect on welfare receipt. Children in families that moved to the new development attended higher quality schools and their parents were more involved with their schools, although children’s grades did not change as a result of their moves.

Whereas these studies focus on variation in neighborhood conditions, other research has exploited exogenous variation in school quality arising from natural experiments in order to identify how the school setting affects academic success. Heather Schwartz (2010) analyzes data on test performance among low-income students in Montgomery County, Maryland, one of the wealthiest urban school districts in the nation. Montgomery County is unique not only because of the quality of its public schools, but also because it has the nation’s oldest and most extensive inclusionary zoning program. As part of this zoning policy, the county’s housing authority is able to purchase up to one-third of the units set aside by developers to be rented or sold at below market rates. The housing authority randomly assigns families selected for housing assistance to these units, which are scattered across all neighborhoods and school attendance zones throughout the county.

Exploiting the random assignment of low-income families to housing units, Schwartz estimated the effect of attending elementary schools with relatively low levels of student poverty versus moderate or high levels of poverty. The study tracked academic performance among 850 low-income students over five to seven years, and found that students in low-poverty elementary schools performed 0.4 standard deviations higher in math and 0.2 standard deviations higher in reading than similar students assigned to schools with 20 percent or higher poverty rates. By the end of elementary school, the gap between low-income students assigned to low-poverty schools and their peers in the larger student body had been cut by half in math and by a third in English.

Will Dobbie and Roland Fryer (2011) analyze the effect on academic performance of attending a charter school run by the Harlem Children’s Zone (HCZ). The HCZ is a well-known community organization targeting a roughly hundred-block area of Harlem with high-quality social services, schools, and programs for youth and families. To identify the effect of attending a HCZ school, Dobbie and Fryer exploit the fact that attendance at the HCZ Promise Academy Schools was based on a lottery among all applicants. As a second identification strategy, the researchers used variation in the probability of attending HCZ schools derived from the interaction of the student’s address and birth cohort. The study found that both older and younger students who were able to attend a Promise Academy experienced substantial improvements in English and math performance, and were less likely to be absent from school. Effect sizes ranged from one-quarter to four-fifths of a standard deviation improvement in standardized test performance, with larger gains in the math assessments.

In considering this evidence, it is important to be clear about what it reveals and what it does not. The studies by Schwartz (2010) and Dobbie and Fryer (2011) both focus on school composition and school quality, but do not indicate that schools are the sole mechanism underlying neighborhood effects, nor that simply offering alternatives to poor-performing public schools will sever the link between neighborhood disadvantage and academic inequality. Analyzing data on student performance derived from thirty-six charter schools that used attendance lotteries, Philip Gleason and his colleagues (2010) find that, overall, attending a charter school had no detectable effects on on academic or behavioral outcomes. Some schools showed strong positive impacts on student performance, particularly those that served more disadvantaged student populations. Other schools produced null or negative impacts. These results suggest that the findings from a specific program or school, such as the Promise Academies within the Harlem Children’s Zone or other programs that offer their own unique and effective approaches,¹ do not necessarily generalize to other schools or programs that may differ in quality, approach, or in the skill of teachers and administrators.

The evidence from the studies of school quality does reveal that when low-income students living in highly disadvantaged residential settings are able to attend high-quality schools, their academic performance improves substantially.² This research provides tangible evidence that the explanations for persistence at the bottom of the academic distribution do not lie fully within low-income individuals or families. Instead, aspects of the residential environment surrounding such families, such as schools, can play an important role in influencing their academic trajectories and, in turn, their prospects for economic mobility.

Evidence on the Effects of Community Change

A third strand of evidence analyzes how change in the neighborhood or local labor market that occurs around individuals affects individual economic trajectories. Sharkey (2012a) compares matched pairs of African American children who lived in neighborhoods that had similar economic and demographic composition but began to change in different ways as they aged into early adulthood. Conditional on initial neighborhood conditions and the trajectory of change in the past, Sharkey argues that it is plausible to think of subsequent neighborhood change as exogenous and to assume any impacts of neighborhood change are causal. The study found that African American children in neighborhoods where the level of concentrated disadvantage declined by one standard deviation had roughly $4,000 higher annual earnings and $6,000 higher annual family income in adulthood. This finding appears to have been driven by economic opportunities because the effect was null on other outcomes, such as educational attainment and marital status.

Other studies have exploited local economic shocks to identify the effect of changes in economic opportunities around individuals and families. For example, Elizabeth Ananat, Anna Gassman-Pines, and Christina Gibson-Davis (2011) use factory plant closings in North Carolina counties to identify the effect of local job losses on aggregate measures of children’s academic performance. Changes in local economic conditions arising from plant closings were found to have large effects on children’s reading and math scores in the state, and impacts were larger for eighth graders than fourth graders. Statewide job losses of 1 percent of the working-age population were estimated to reduce eighth grade math scores by 0.076 standard deviations.³

Another example comes from the experience of American Indian tribes in the aftermath of the 1988 law that allowed for the development of large-scale gaming facilities on reservation land (Wolfe et al. 2012). Several researchers have used variation in the timing at which gaming facilities have opened on reservation land to identify the effect of an influx of income and economic opportunities into highly disadvantaged areas. These studies have found substantial effects on educational attainment, on median income and employment, and on physical and mental health, including mortality (Copeland and Costello 2010; Costello et al. 2003; Wolfe et al. 2012). Although the establishment of casino gaming is a unique form of change in the local economic environment and one that may come with serious social consequences, the evidence from American Indian reservations does indicate that large-scale transformations of local economic opportunities can generate substantial economic benefits for the residents of the area (on the impact of the Tennessee Valley Authority, see Kline and Moretti 2013).

Summary and Next Steps

Several conclusions can be made from the range of empirical work summarized, some of which stand on firmer ground than others. First, evidence from observational studies typically documents an association between child neighborhood conditions and adult economic outcomes, although this relationship is not found in all studies that have been conducted. A much larger literature has examined the relationship between neighborhood conditions and local school quality and outcomes related to educational attainment and academic performance. This strand of research has generated consistent evidence that growing up in disadvantaged residential environments and attending low-quality schools impedes children’s academic trajectories and development of cognitive skills. Recent contributions to this literature have made improvements in modeling selection into high-poverty neighborhoods over time, and have shown that the consequences of long-term exposure to disadvantaged environments appear to be cumulative, more harmful effects arising from sustained or multigenerational exposure to neighborhood disadvantage.

Evidence derived from housing mobility programs is more difficult to interpret. Research from several quasi-experimental and experimental studies has yielded strong evidence that moving out of concentrated disadvantage can have substantial benefits for the developmental trajectories of youth and for parents’ well-being, but these findings depend on the nature of the housing mobility program and the types of moves that families make. This evidence is more useful for evaluating the impact of specific policies implemented in unique locations and times, than for making general conclusions about the relationship between neighborhoods and social and economic mobility.

Last, several studies focusing on shocks in local labor market opportunities have documented strong impacts on adults’ labor market outcomes and children’s academic outcomes. These studies confirm the intuitive idea that the presence or absence of opportunities in the residential environment can play important roles in affecting prospects for economic success, with impacts that extend across generations.

The research reviewed here focuses primarily on economic opportunities and schools as key pathways linking neighborhoods with economic mobility. However, the literature on neighborhood effects has developed much more extensive theoretical models focusing on how the residential environments surrounding children may affect their developmental trajectories and ultimately influence their prospects for economic mobility (Harding et al. 2011; Galster 2012). Patrick Sharkey and Jacob Faber (2014) argue that greater attention should be devoted to the full range of mechanisms linking residential settings with children’s outcomes, including environmental exposures such as air pollution, lead, and violence; the quality and quantity of local institutions such as day-care centers, nonprofits, and churches; and peer groups, networks, and role models. Although a large literature documents the effect of these various dimensions of children’s lives on health, education, and cognitive skill development during childhood, very few studies have attempted to make the connection to economic mobility. Moving beyond schools and jobs to provide evidence on the full range of mechanisms linking neighborhoods with economic mobility is a clear next step for this literature.

HOW DO CITIES AND METROPOLITAN ARE AS AFFECT ECONOMIC MOBILITY?

Virtually all of the research on intergenerational economic mobility in the United States describes the level of mobility in the nation as a whole. Whereas early estimates suggested only a weak relationship between the economic status of parents and their children, more recent research that corrects for methodological problems in the original studies finds much lower levels of economic mobility (Mazumder 2005a, 2005b; Solon 1992). Recent estimates of the intergenerational elasticity of family income in the United States, measured as the strength of the relationship between the natural logarithm of total family income measured at the same age across successive generations, range from around 0.40 up to as high as 0.60. The former estimate, which represents a rough lower bound of the estimated intergenerational elasticity in the United States, can be interpreted to mean that if a parent’s income is roughly twice as high as the national average, then the child’s adult income would be expected to be about 40 percent higher.

This estimate indicates much greater persistence of economic status across generations in the United States than in European nations or Canada (Corak 2006; Smeeding, Erikson, and Jantii 2011; Solon 2002). Recent research has looked to national policy as a way of explaining the relatively low level of economic mobility in the United States, but this approach overlooks the potential to explore variation within the nation to begin to understand what drives intergenerational economic mobility. Considering the tremendous diversity in population characteristics, regional and local economies, politics, and culture within the United States, there are good reasons to expect substantial variation across U.S. states and cities in levels of economic mobility. Several recent studies confirm this expectation.

An analysis by Bhashkar Mazumder and published by the Pew Charitable Trusts Economic Mobility Project (Economic Mobility Project 2012) offers estimates of intergenerational economic mobility across the states, identifying a pocket of northeastern states with high levels of economic mobility, and another pocket of southern states with much lower levels. Bryan Graham and Sharkey (2013) use data from three national surveys and document substantial variation in levels of economic mobility across urban areas. Chetty and his colleagues (2014) use data from the Internal Revenue Service to create several measures of relative and absolute income mobility across the nation’s commuting zones, which are sets of contiguous counties that surround central cities and cover the entire nation. These authors find substantial variation in levels of economic mobility, and that some commuting zones have mobility levels equal to those of the most mobile nations in Western Europe and others with lower levels of mobility than any of the developed nations.

All of these studies indicate that national measures of intergenerational economic mobility obscure substantial geographic variation in levels of mobility across urban areas, commuting zones, states, and regions of the country. Recent evidence suggests that this variation is a function of places themselves, rather than the people within them. Chetty and Hendren (2015) use data from the Internal Revenue Service to analyze the relationship between time spent in low-and high-mobility areas and economic outcomes later in life. The authors exploit variation across siblings and other “exogenous displacement shocks” and find that “spending more of one’s childhood in an area with higher rates of upward mobility . . . leads to higher earnings in adulthood” (Chetty 2015, 25). In other words, something about high-mobility areas seems to affect the chances for economic mobility among residents. Although this evidence suggests a causal effect of places on economic mobility, less progress has been made in explaining what it is about those places that increases or reduces the chances for residents to move upward in the income distribution.

Mechanisms Explaining Variation

Chetty and his colleagues (2014) analyze the correlations between a range of social, demographic, and economic characteristics of commuting zones and the level of economic mobility, focusing primary attention on the probability of upward mobility from the bottom of the income distribution. The level of upward mobility in a commuting zone was found to be most strongly associated with the degree of racial and economic segregation in the commuting zone, with the rate of high school dropouts and single parents, with the level of violent crime and measures of social capital, and with the level of economic inequality in the commuting zone. This descriptive analysis provides several suggestive conclusions about the types of commuting zones with high and low levels of economic mobility, but the analysis is exploratory and not designed to generate convincing causal evidence.

Graham and Sharkey (2013) focus on the connection between economic segregation, defined as the proportion of overall variance in income within a metropolitan area that lies between neighborhoods, and levels of economic mobility. This focus is motivated by a theoretical model in which transmission of parents’ economic status is driven by both family-level mechanisms and place-based mechanisms such as the quality of local schools and other institutions, property values, crime, and other aspects of the residential environment. This model predicts that in urban areas where the rich live in separate neighborhoods from the poor, the benefits of economic resources and the costs of poverty are exacerbated because of the tight connection between family economic status and neighborhood economic status. As a result, family economic status is transmitted more easily to the next generation (for similar propositions, see Loury 1977; Durlauf 1996). An association between economic segregation and economic mobility was found in three different datasets, and also was found in analyses that examine change in economic segregation and change in mobility within urban areas. Similar to that of Chetty and his colleagues (2014), the analysis by Graham and Sharkey (2013) provides suggestive evidence linking economic segregation with variation in levels of mobility, but not evidence that allows for strong causal claims.

Summary and Next Steps

Two summary conclusions are possible on the basis of these studies. First, geographic variation in levels of economic mobility within the United States is substantial. This observation means that national estimates of income mobility, though informative, pool data from places that have widely divergent patterns of both absolute and relative mobility. Perhaps the most notable geographic pattern is found in maps Chetty and his colleagues (2014) present, which document a large swath of the southeastern part of the country featuring extremely low levels of mobility. This striking pattern reveals a large section of the country where upward mobility is rare, and suggests the need for more empirical research to explain regional variation in economic mobility.

Second, new evidence from Chetty and Hendren (2015) suggests that geographic variation in levels of economic mobility reflects the causal effects of places, rather than the selection of more or less mobile people into areas. This evidence is based on the empirical finding that children who spend more time in areas with greater levels of mobility have a higher probability of upward economic mobility relative to other children who spend less time in the high-mobility area.

Third, some evidence has been generated on the characteristics of places that are associated with economic mobility, but minimal progress has been made in providing evidence that allows for causal claims. Graham and Sharkey (2013) provide a strong theoretical motivation for the focus on economic segregation, and Chetty and his colleagues (2014) document conditional associations between economic mobility and a range of characteristics of commuting zones. As a whole, however, the research explaining geographic variation in economic mobility remains at a very early stage. Generating causal evidence on the mechanisms explaining variation in economic mobility is the most pressing challenge for researchers in this area.

WHAT ARE THE IMPLICATIONS FOR SOCIAL POLICY?

Two broad approaches are commonly proposed to reduce neighborhood inequality and its consequences. The first confronts neighborhood inequality with investments in communities, or families within them, designed to weaken the link between growing up in a disadvantaged neighborhood and its consequences for children’s economic trajectories. The second confronts neighborhood inequality more directly by attempting to alter the distribution of neighborhoods occupied by different segments of the population. The most common policy tool used to implement the second approach is residential mobility or housing assistance programs for low-income populations. In addition to such mobility programs, however, a set of more basic changes in housing and urban policy represent alternative approaches to compressing the distribution of neighborhood advantage and disadvantage.

Place-Based Investment

Several examples of investments and initiatives have been designed to target disadvantaged places or the individuals and families within them. The New Hope program, implemented in Milwaukee in the mid-1990s, offered extensive work supports, wage supplements, and temporary guaranteed jobs for individuals willing to work at least thirty hours per week (Duncan, Huston, and Weisner 2009). Certain features of the program distinguished it from many other welfare-to-work programs implemented in the same period, one of which was to target low-income individuals living in low-income neighborhoods. New Hope is thus an example of a “place-conscious” program that directed resources and supports toward individuals and families within disadvantaged areas (Pastor and Turner 2010).

Applicants to the program were randomly assigned to a treatment and control group, and several studies have tracked the outcomes of participants over an extended period. Results show that the treatment group had higher rates of employment and earnings while the program was in operation, a finding that was driven, in part, by the guarantee of community service employment for participants unable to find a job in the private market (Huston et al. 2003). The program reduced family poverty from 60 percent among control group members to 52 percent among program group members; and multiple studies have documented improvements in academic performance and behavior among the children of families in the program’s treatment group (Duncan, Huston, and Weisner 2009; Huston et al. 2001; Huston et al. 2003). Children in the program group scored 0.12 standard deviations higher than their counterparts in the control group on an assessment of reading and language skills several years after the program began (Huston et al. 2003). The program costs totaled roughly $16,000 per individual over three years, or $5,300 per person per year (Huston et al. 2003).

Another place-conscious intervention focusing on individuals’ prospects in the labor market was the Jobs-Plus program implemented in the 1990s by the federal Department of Housing and Urban Development in five very different cities: Baltimore, Chattanooga, Dayton, Los Angeles, and St. Paul. Jobs-Plus saturated public housing developments both with services designed to enhance individuals’ capacity to obtain and retain employment over time, and with rent incentives designed to encourage work. Over the course of the program, resident employment and income rose steadily in three of the sites that implemented the full package of services and incentives offered through the program (Bloom et al. 2005). In these three sites, the program was estimated to increase employment by roughly 10 percent and to increase annual earnings of participants by between 8 and 19 percent (Bloom et al. 2005). The full costs of the rent incentives and on-site services were roughly $1,800 per individual per year (in 2003 dollars), although if public housing agencies are already administering some of the services offered through Jobs-Plus then the incremental costs of this program would be lower (Bloom et al. 2005).

Whereas New Hope and Jobs-Plus targeted individuals within high-poverty areas, an alternative set of interventions attempt to create greater demand for labor through incentives designed to encourage firms to invest and to hire local residents. The most notable example is the Empowerment Zones/Enterprise Communities (EZ/EC) program that began in the mid-1990s.⁴ This federal program provided firms, in dozens of communities, tax incentives designed to encourage expansion, investment, and employment of local residents. The most persuasive evidence demonstrating positive effects of the designation as Empowerment Zones comes from Matias Busso, Jesse Gregory, and Patrick Kline (2013), who estimate the effect of the program by comparing the six selected sites with others that applied and were rejected or that applied and were later accepted. The researchers estimate that the program increased the number of jobs for residents within the zone boundaries by 15 percent, and increased wages of workers within the zone by 8 percent (in addition to benefits for workers not living in the zone). The authors’ best estimates of program costs indicate that the federal government spent roughly $850 per resident on grants and wage credits, though the various sources of funding that went into the Empowerment Zones were difficult to measure and all of the point estimates from the analysis are imprecise (Busso et al. 2013).

This study’s findings stand in contrast to other empirical findings of minimal effects of the program in different sites (see Elvery 2009; Oakley and Tsao 2006). Further, several commentators have raised doubts, from both theoretical and empirical perspectives, that any spatially targeted economic development program can generate cost-effective, positive effects on residents in the absence of supplemental investments (Glaeser and Gottlieb 2008; Ladd 1994).

The interventions discussed to this point focus on reducing poverty, improving prospects in the labor market, and creating job opportunities for individuals in disadvantaged neighborhoods. However, the presence or absence of economic opportunities is only one of several mechanisms through which neighborhood inequality may be linked with economic mobility. A range of interventions targeted toward disadvantaged areas have been conducted over time to improve the quality of schooling, reduce violent crime, or improve community health (see, for instance, Braga 2005; Cook et al. 2015; Dobbie and Fryer 2011; Heller 2014; Heller et al. 2013; Papachristos et al. 2007; Schwartz 2010). These types of place-based or place-conscious interventions may indirectly affect levels of economic mobility through exposure to violence, school quality, health, or the many additional mechanisms by which neighborhood advantage and disadvantage are linked with economic mobility.

The recognition that economic disadvantage tends to be concentrated in areas that face a range of associated challenges is the motivation for a set of interventions that have come to be identified as Community Change Initiatives (CCIs). CCIs focus on comprehensive neighborhood revitalization designed to flood an area with resources focused on economic development, institutional support, physical infrastructure, and social services (Kubisch et al. 2010). As Anne Kubisch notes in a report describing how CCIs have been implemented over time, most Community Change Initiatives have in practice not received the level of sustained resources necessary to generate transformative community change. Reviews of the field have identified other challenges faced by any effort to implement comprehensive community change, including the coordination of services and supports, building institutional capacity, and engaging residents and other important local actors in the effort (Chaskin, Joseph, and Chipenda-Dansokho 1997; Kubisch et al. 2010). Like many other efforts to revitalize communities through place-based investment, the impact of CCIs is difficult to assess because programs rarely have been designed in ways that allow for a clear assessment of program impact, and because these types of programs typically are not implemented at a scale that could generate tangible change that is sustained over time (O’Connor 1995, 1999; Sharkey 2013).

Although interventions designed to generate neighborhood change have not demonstrated a clear track record of success, empirical work on large-scale changes in local economic opportunities provide proof-of-concept evidence suggesting that major investments that alter the local economic environment in a fundamental way can have substantial, long-term effects on residents. Research on federal investments in the Tennessee Valley Authority and the introduction of casino gaming to American Indian reservations provides suggestive evidence that large-scale transformations of local economic opportunities can generate substantial economic benefits for the residents of the area (Kline and Moretti 2013; Copeland and Costello 2010; Wolfe et al. 2012). For example, Barbara Wolfe and her colleagues (2012) estimate that the introduction of casino gaming increased average household income by 5.3 percent, leading to indirect declines in smoking (9.6 percent), anxiety (7.3 percent decline in days of anxiety), heavy drinking (5.2 percent), and health outcomes related to obesity (2 to 4 percent). These estimates do not consider any larger costs associated with casino gaming, of course, and are included here only as an illustration of the potential gains that arise from large-scale economic transformation of a clearly defined area or zone. How to turn these examples into effective policies or programs targeting disadvantaged communities is a much more challenging question.

Expanding Mobility and Reducing Inequality

The alternative approach to confronting neighborhood inequality is to implement new policies or revise existing policies specifically to alter directly, or compress, the distribution of neighborhood advantage and disadvantage. One method of moving toward this goal is residential mobility programs for recipients of housing assistance.

I have already discussed results from various housing mobility programs that have been implemented and evaluated over time. A more basic consideration is whether such programs are effective mechanisms to generate meaningful changes in families’ residential environments. Research from the most well-known residential mobility program, the Moving to Opportunity experiment, has demonstrated that families in the experimental groups from several sites did experience changes in exposure to neighborhood poverty but were highly likely to move into communities that were both near and similar in racial-ethnic composition and school quality to their origin neighborhoods (Clark 2008; Sampson 2008). Over time, the change in neighborhood poverty induced by the program faded away (Ludwig et al. 2012).

This pattern of findings reflects the challenges families face in navigating a highly stratified urban landscape, but it also reflects the “psychological constraints” (Shroder 2002) that condition the choices made by different groups of families in urban housing markets. For example, one of the primary predictors of whether families in the MTO experiment were able to “lease up” in a new apartment was their uncertainty about whether they would like their new neighborhood if they were to move (Shroder 2002). This type of uncertainty arises from unfamiliarity with communities around an urban area (Krysan and Bader 2009), from concerns about how families would be treated in new communities (Thompson 2001), and from a very real historical legacy of discrimination and violence. As a result, residential moves made by families receiving housing assistance and navigating the private rental market often are found to reinforce, rather than disrupt, patterns of urban inequality (Sharkey 2012b).

A few housing assistance programs have been more successful in generating transformative changes in families’ neighborhood contexts. One common feature of these programs is that they take a more active role in expanding the choice set of families deciding where to relocate. For example, many housing experts call for more intensive counseling for recipients of housing assistance; and some argue for efforts to alter the choice architecture of families as they begin their search for housing. Xavier de Souza Briggs, Susan Popkin, and John Goering (2010) suggest providing families with a “default” set of two or three units that are available in different communities within the city. An extreme version of this approach is the Gautreaux Assisted Housing Program in Chicago, where participating families were offered specific units located throughout the Chicago metropolitan area based on their position on a waitlist (Rubinowitz and Rosenbaum 2000). Unlike in most residential mobility programs, the Gautreaux moves took families across the entire Chicago metropolitan area and brought about a change in families’ neighborhood environments that persisted over time (Keels et al. 2005).

A few current housing assistance programs follow a similar approach by providing extensive information, support, and resources necessary to allow families to make the kind of residential moves that bring them into entirely new sections of their metropolitan areas. The Baltimore Housing Mobility Program also arose from a settlement with the federal Department of Housing and Urban Development, and features intensive counseling designed to bring families into new, racially diverse communities with low rates of poverty and abundant economic opportunities (Darrah and DeLuca 2014). This active approach is necessary to allow families to make the kinds of moves that disrupt the structure of residential stratification within the metropolitan area, and that are rare among low-income families navigating the rental market on their own.

However, providing vouchers that allow low-income families to move is not the sole mechanism to expand residential options for low-income families or members of racial and ethnic minority groups. An alternative approach involves taking active steps to break down barriers that limit housing choice. Manuel Pastor and Margery Turner (2010) review an extensive list of options that include expanding the supply of affordable housing, confronting exclusionary zoning policies, promoting and enforcing fair-share housing plans, taking active steps to reduce residential discrimination by race and ethnicity, and developing coordinated metropolitan-wide plans for transportation, housing, education, and economic development (see also Goering 2006; Katz and Turner 2001; Katz 1999, 2000; Rusk 1999; Quigley 2011; Quigley and Raphael 2004; Turner and Ross 2005).

These proposed policy shifts reinforce the point that to weaken the connection between neighborhood inequality and economic mobility it may not be necessary to implement new interventions, programs, or initiatives with substantial costs attached to them. Altering or ending several existing housing and land use policies that exacerbate inequality, and instead implementing programs that confront inequality, would be an initial step in an urban policy agenda designed to reduce neighborhood inequality.

One example is the home mortgage interest deduction. In 2012, the estimated costs of the mortgage interest deduction were $70 billion, most of which went to homeowners making $100,000 or more in annual income (Fischer and Huang 2013; Turner et al. 2013). Various proposals to reform this deduction have been put forth in an effort to make it less regressive, more efficient, and less expensive while limiting any potential negative impacts of the reforms on the housing market. To confront neighborhood inequality, one could argue for shifting federal tax expenditures forfeited to the mortgage interest deduction toward programs designed to provide affordable housing or to expand housing options in high-opportunity neighborhoods. For example, Barbara Sard and Will Fischer (2012) argue for reforming the mortgage interest deduction and instituting a renter’s tax credit as part of an effort to create a more balanced approach toward housing policy. In advocating for this type of approach, it is important to acknowledge the dearth of research on the impact of the mortgage interest deduction. With the exception of research focusing on the mortgage interest deduction and home ownership (Hilber and Turner 2014), little empirical work has been conducted that would inform our understanding of the likely consequences of these proposed reforms on economic or racial-ethnic segregation or on neighborhood inequality.

It is certainly true that promising new programs and reforms of existing housing assistance programs can be effective in reducing the consequences of neighborhood disadvantage. However, it is also true that neighborhood inequality is in part the result of active intervention into the housing market through law and public policy. A basic approach to confront neighborhood inequality is to change the way that the federal and state governments invest in places. This can be done with new programs or investments, but it can also be done by altering existing policies in basic ways designed to compress the distribution of neighborhood economic status and to reduce the consequences of neighborhood disadvantage.

Download Robert Putnam Discussant Comments on Sharkey Conference Presentation.