Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

American housing policymakers have been confronted with the geographic implications of their strategies for almost half a century.Footnote 1 Following the end of legal segregation of public housing with the Executive Order of 1964 and the publication of the Kerner Commission report in 1968 in the aftermath of the prior four summers’ urban civil disruptions, the federal government began to grapple with the possibility that where they were supplying housing assistance was perhaps contributing to the poverty problem more than its solution (Goering 1986; Galster 2008). Both public housing and assistance provided to privately owned developments for low-income tenants began to come under criticism by scholars (e.g., Rainwater 1970) and federal courts (e.g., Gautreaux case; see Polikoff 2006) for their role in creating and maintaining ghettos. This geographic analytical focus gained academic if not policy salience with the publication of Wilson’s The Truly Disadvantaged (1987) and my introduction of and formal conceptualization of the term “geography of opportunity” (Galster and Killen 1995). Over two decades of ever-intensifying inter-disciplinary research and policy discussion on “neighbourhood effects” followed (see van Ham et al. 2012, 2013).

Rising concerns by the courts, scholars, and activists over the personal and social costs arising from concentrating low-income (typically minority) households in urban neighbourhoods with high proportions of similarly disadvantaged households prompted several types of programmatic responses by federal government housing policymakers (Popkin et al. 2000; Goetz 2003; McClure 2006, 2008). Arguably, the earliest was an attempt by the U.S. Department of Housing and Urban Development (HUD) to redirect the vestiges of new housing construction and acquisition under the public housing program toward small-scale sites outside of neighbourhoods of concentrated disadvantage beginning in the late 1960s and early 1970s (Hogan 1996). The second was HUD’s increasing emphasis on attaching housing assistance to the needy tenant instead of to a dwelling unit, beginning formally with the creation of tenant-based housing assistance certificates in Section 8 of the 1974 Housing and Community Development Act. Since the inception of the “Section 8” (re-titled Housing Choice Voucher, HCV, in the 2000s) program, there have been a few changes in program administrative rulesFootnote 2 and experiments with providing pre-move assistance and counseling to subsidized tenants in an effort to encourage them to use their voucher to move to lower-poverty neighbourhoods offering superior quality of life and opportunities. By 1994 the conditions of some public housing estates had grown so dire that HUD initiated a third response: the HOPE VI (Housing Opportunities for People Everywhere) program. The notion was to demolish or rehabilitate the worst public housing estates, ultimately replacing them with mixed-income (often mixed-tenure) developments. Original low-income residents would either inhabit the affordable units on the redeveloped sites or would be helped to move elsewhere with tenant-based housing assistance or conventional or scattered-site public housing.Footnote 3

At the outset I should make it clear that the deconcentration of poverty has never been a major, consistently pursued goal of federal housing policy, nor have HUD programs or administrative rules been comprehensively and systematically oriented toward achieving this goal. Indeed, the federal effort at poverty deconcentration could be described as token, fragmented, and reluctant. Scattered-site public housing was rarely adopted by the local housing authorities that manage public housing and HCV programs, and often only under the impetus of a court order. This initiative never represented more than a tiny share of public housing units nationwide (Hogan 1996). Though there have been several small poverty deconcentration demonstration programs involving HCVs (Schwartz 2010), they have involved only a few dozen local housing authorities representing a small share of all HCVs. In addition, HCVs with stipulations for deconcentration have frequently been required as elements of court-ordered public housing desegregation decrees, not because of HUD initiatives (Popkin et al. 2003). Arguably, the HOPE VI program was more motivated by an urgent political need to defuse Congressional Republican efforts to abolish HUD than by an overarching commitment to deconcentrate poverty. Moreover, federal housing efforts are bureaucratically fragmented. By far, the major current program for the construction of affordable housing in the U.S. is the Low Income Housing Tax Credit (LIHTC) program, yet this is administered by the Treasury Department, not HUD. As explained below, this program’s rules are schizophrenic regarding poverty deconcentration.

In this paper I do not challenge the notion that deconcentrating poverty and reducing future concentrations of poverty is a worthy goal of federal housing policy (cf. Cisneros 1996; Galster 2002; Goetz 2003; Arthurson, Chap. 12, this volume). Nor in this paper do I raise the thorny questions of whether deconcentration ultimately benefits the low-income households who may be involved (cf. Goering and Feins 2003; DeLuca and Dayton 2009; Galster 2013), the host communities that may become more diverse as a consequence of these programs (cf. Galster et al. 2003), or the communities from which the poor move (cf. Galster 2003). Rather, here I take a distinctly geographic perspective and consider the degree to which these aforementioned federal housing programs succeeded in opening up a wider variety of spatial opportunities for low-income households to live in lower-poverty, less minority concentrated neighbourhoods.Footnote 4 I then address what individual, structural, and administrative forces may have influenced the success of these programs in this geographic regard. Finally, I consider what spatial lessons the U.S. experience with deconcentration strategies offers to an international audience.

I rely upon secondary analyses of studies of the aforementioned four types of federal deconcentration programs. As such, I am constrained in my operationalization of neighbourhood indicators and bases for comparison.Footnote 5 I thus am unable (with rare exceptions) to explore the degree to which these programs have fostered deconcentration along lines of improving access of low-income households to appropriate employment, superior school districts, or other dimensions of opportunity that are often weakly measured by neighbourhood poverty rates and minority population percentages that have been traditionally employed in research. Similarly, I am often unable to compare geographic outcomes against alternative benchmarks: locations of recipients pre- vs. post-program participation, locations of recipients across programs, and locations of comparable households who are not recipients.

I also note as introduction that virtually all extant research is descriptive or quasi-experimental in its design; the exception is the Moving To Opportunity (MTO) demonstration, which used a random assignment design. Because there is a great deal of administratively-based and recipient-based selection into programs, claims about the independent causal impact of a program on recipient location cannot be made, with the possible exception of MTO. Moreover, because studies typically report locations only of those who succeed in participating in the given program (i.e., either passed the screening for site-based projects or successfully leased an apartment through the HCV program), the full program effects can be overstated (Clark 2005).Footnote 6 To complicate matters still further, there is a great deal of functional interdependence among the programs. HCV holders often reside in LIHTC developments. HOPE VI projects rely on HCVs to relocate most of their original tenants. Comparisons among randomly assigned groups in the Moving To Opportunity (MTO) demonstration were confounded by the fact that many control group households were later required to leave their public housing unit due to HOPE VI initiatives. Court-ordered public housing desegregation mandates (such as in Baltimore) also offered them HCVs for relocation that placed geographic limitations on their use. Thus, the independent geographic impact of a particular program on the locations of subsidy recipients may be camouflaged behind the forthcoming statistics showing their geographic patterns.

Despite these caveats, some clear and important findings can be discerned. I proceed by analyzing the geographic patterns of participants in each of the aforementioned strands of federal policy—scattered-site public housing, HCVs, LIHTC, and HOPE VI—and compare these patterns to other low-income renters not receiving subsidies and across programs to the extent feasible. I turn next to characteristics of the low-income participants, the structural aspects of local housing markets, and housing program administrative rules that may influence the geographic outcomes of the programs. Finally, I draw lessons from this analysis for policymakers in other nations who may wish to pursue their own deconcentration strategies.

Scattered-Site Public Housing

Local Public Housing Authorities (PHAs) that develop and manage public housing in the American system were encouraged by U.S. Department of Housing and Urban Development (HUD) beginning in the 1970s to develop more housing on a “scattered-site” basis. This typically was operationalized as the construction and/or acquisition of low-density buildings with fewer than 15 units per site in locations that where not disproportionately minority-occupied (Hogan 1996). This strategy was not widely adopted across the nation, and cross-PHA variations in the density and locations of “scattered sites” were huge.

Beginning in the early 1980s and continuing for two decades, the impetus for scattered-site public housing was primarily provided by the federal courts. In dozens of locales across the country, PHAs and HUD were sued by minority public housing tenants claiming a variety of discriminatory and segregationist actions (Julian and Daniel 1990). All these cases were resolved with court-ordered deconcentration efforts, most involving a combination of scattered-site public housing and HCVs issued to plaintiffs (Popkin et al. 2003).

With very little research attention focused on it, the most recent data on scattered-site public housing come from a 1994 survey of selected large- and medium-sized PHAs. It found that in the large PHAs such housing represented 8 % of all PHA units, and were scattered with 6.2 units per site, on average. The comparable figures for the medium-sized PHAs were: 9.5 % share with 4.7 units per site (Hogan 1996). Though the case studies revealed the popularity of scattered-site compared to conventional public housing on the part of program administrators, tenants, and the general public, a near-elimination of funding for new public housing development of any sort (especially after the advent of HOPE VI) relegated this strategy to a marginal or “boutique” status.

Nevertheless, the 1994 survey provided some suggestive evidence from nine case study sites about where scattered-site units were located. A unit-weighted average of reported data (Hogan 1996: Tables 3–6, 3–7) reveals the following percentages of scattered-site units in census tracts with 1990 higher-than-citywide: median income (28.5 %); poverty rates (58.8 %); minority occupancy rates (61.4 %); and high school graduation rates (43.5 %). These figures suggest that the sampled scattered-site public housing developments were located in better neighbourhoods than conventional public housing, but nevertheless were in neighbourhoods with higher rates of poverty, minority occupancy, and high-school leaving than the average for their cities. However, there is so much variation in neighbourhood characteristics of scatted-site developments both within and across cities, so these averages should be interpreted with considerable caution.

Hogan (1996) also investigated two special cases of Chicago (IL) and Yonkers (NY), which were required to build scattered-sites in response to public housing desegregation court decisions; see Table 11.1. In both cases, the scattered-sites were located in census tracts that had substantially higher incomes and lower poverty rates, unemployment rates, and rates of black occupancy, compared to the conventional public housing developments that previously were their only public housing options. Other generalizations are more difficult. In Yonkers the scattered sites were in places with more white and fewer Hispanic residents and more with college degrees; not so for Chicago. Moreover, it is clear that the scattered sites in Yonkers were considerably more advantaged locales in multiple dimensions than scattered sites in Chicago. Thus, while it is clear that in general scattered site public housing programs have offered superior neighbourhood environments for low-income tenants compared to conventional, large-scale, concentrated public housing developments, the gain achieved is contextualized by local market structures, including frequent neighbourhood opposition.

Table 11.1 Comparison of neighbourhood characteristics in Chicago and Yonkers, by scattered-site and conventional public housing developments (participants are black and Hispanic)
Full size table

Tenant-Based Rental Assistance: Housing Choice Vouchers

Since its inception in 1974, tenant-based rental assistance had followed the general formula that the qualifying household must contribute a share of its income (currently 30 %) toward rent of an apartment that meets certain quality standards and whose landlord agrees to sign a minimum 1-year lease with the tenant and the PHA administering the voucher. The value of the subsidy is computed as the difference between metropolitan area’s “fair market rent” (currently the 40th percentile of rents for the given apartment size the family qualified for, based on a recent survey of that entire metro area) and the tenant contribution. In its early “certificate” incarnation, the program required the tenant to find apartments at or below fair market rent. This was modified in the current “voucher” version, though the value of the subsidy was not increased if the tenant chose to occupy a more expensive apartment; very few can afford this extra expense. Once issued a voucher, the tenant has usually 3 or more months depending on the PHA to locate an apartment with a landlord willing to participate in the HCV program, complete requisite lease paperwork and have the apartment inspected.Footnote 7 As of 2009, almost a third of all households receiving federally financed housing assistance received their benefit through a HCV (Schwartz 2010: Table 1.1)

In principle, the augmented purchasing power provided by the HCV should reduce the financial constraints on low-income households’ ability to occupy apartments in lower-poverty, lower minority concentration neighbourhoods. Those who hoped the HCV program would produce different spatial outcomes for low-income households were sobered, however, by the findings of the Experimental Housing Allowance of the late 1970s (Cronin and Rasmussen 1981) and the first wave of city-specific case studies after program rollout (e.g., Hartung and Henig 1997; Newman and Schnare 1997; Turner 1998b). They showed that most HCV holders went to neighbourhoods that indeed were slightly less poor and minority-occupied, but still were relatively poor, segregated places compared to the generic neighbourhood. Moreover, many HCV users did not move at all, preferring instead to reduce their rent contributions to their current landlord.Footnote 8

Subsequent national studies of HCV geographic outcomes provided a more nuanced portrait but did not alter its fundamental contours. Pendall (2000) compared the neighbourhood circumstances of a 1998 nationwide sample of HCV holders and low-income renters who received no assistance. He found that those with HCVs were only 75 % as likely to live in distressed neighbourhoods, on average. There were substantial variations, however, across metropolitan areas and race of HCV holder (with blacks being much more likely to use their vouchers in distressed neighbourhoods). McClure (2006) found in fiscal year 2002 that HCV holders experienced modestly lower average neighbourhood poverty rates than all very low-income renters (those earning less than 50 % of the metro area median income) in central cities (23.2 % vs. 24.4 %). Yet, the opposite relationship proved true in the suburbs (13.5 % vs. 12.1 %), producing only a small difference nationwide (18.9 % vs. 19.8 %). In a companion study (2008), McClure found that only 26 % of HCV holders resided in census tracts with less than 10 % poverty rates. This figure was one only percentage point higher than the average poverty rate in the locations of unsubsidized renters in the same income bracket as most HCV tenants (i.e., under 30 % of metro area median income). The performance of minority HCV holders was even worse in this regard: only 17 % of black and 19 % of Hispanic HCV holders resided in neighbourhoods with under 10 % poverty rates. In the only study to examine cross-housing program comparative safety characteristics of neighbourhoods, Lens and colleagues (2011) recently found that HCV holders in ten large cities, resided in neighbourhoods with lower crime rates than those in place-based assisted housing, on average.

It should be noted that comparing HCV holders to other low-income renters obscures some important unobservable differences between the groups, so the aforementioned differences (or lack thereof) might be due purely to selection bias of who takes up HCV and/or succeeds in leasing up. This possibility was tested explicitly in a random assignment experiment (Patterson et al. 2004), but this did nothing to shake the central conclusion reached above: use of a voucher resulted in only small improvements of neighbourhood conditions on many dimensions.

A different basis of comparison—longitudinal changes in households’ locations before and after receipt of a HCV—paints a similar portrait.Footnote 9 Feins and Patterson (2005) conducted the most comprehensive longitudinal analysis using a national sample of those entering the HCV program during 1995–2002. They discovered that the trajectory of moves was not into significantly better neighbourhoods (measured on many characteristics) over time; see Table 11.2. Several things are significant from Table 11.2. First, by most indicators the post-HCV neighbourhoods were inferior to the pre-HCV ones. Second, the second neighbourhood occupied by HCV holders was generally inferior to the first neighbourhood they occupied with their HCV. Third, the average differences in either direction were small.Footnote 10

Table 11.2 Comparison of neighbourhood characteristics of HCV participants’ pre-program, initial program, and second program locations (national sample 1995–2002; all ethnic groups)
Full size table

Feins and Patterson’s (2005) multivariate models showed an interesting geographic aspect, however. Moving greater distances with HCVs was associated with lower neighbourhood poverty rates and greater neighbourhood owner-occupancy rates. For example, those moving 1–5 miles saw an average 1 % point decrease in neighbourhood poverty rate, whereas those (few recipients) moving over 10 miles saw at least double that reduction.Footnote 11

These results all suggest that merely increasing the effective affordability of decent-quality vacant apartments via a HCV is insufficient to get much average improvement in the geographic outcomes for program participants compared to comparable renters who are not subsidized. But how much of this is related to distance of initial move and longevity since departure from original neighbourhood? What happens to outcomes if stronger or weaker constraints on geography are imposed? What happens if more mobility assistance and counseling is provided to HCV recipients? Three special programs involving HCVs provide some answers to these questions: Gautreaux phases I and II and the Moving To Opportunity (MTO) demonstration; for a good overview of these three programs and their results, see Goering and Feins (2003) and Duncan and Zuberi (2006).

The Gautreaux Phase I and II Programs

In 1966 Dorothy Gautreaux, representing the class of black residents of Chicago Housing Authority (CHA) public housing projects, filed suit against CHA and HUD, alleging a variety of discriminatory practices. After extended court battles, the Supreme Court found in her favour (Rubinowitz and Rosenbaum 2000). The first court-mandated remedy provided 7,000 HCVs issued from 1976 to the late 1990s to black CHA tenants and waiting list candidates, and mandated extensive pre-move counselling and moving assistance for participants provided by a local non-profit fair housing organization. Initially the court required that all such HVCs be used in suburbs with less than 30 % black populations. Roughly four-fifths of the participants were ultimately placed in such suburbs. However, as the Chicago rental market tightened in particular years, some families were permitted to be placed in Chicago City neighbourhoods that were considerably less advantaged and had higher minority shares than 30 %, but were deemed “revitalizing” (Rubinowitz and Rosenbaum 2000).

Recent follow-up investigations of “Gautreaux I” revealed that movers to suburbs indeed succeeded in getting into and staying in much safer, whiter neighbourhoods with better schools than the neighbourhoods they left (Keels et al. 2005; DeLuca and Rosenbaum 2003).Footnote 12 Origin communities were on average 42 % poor and 83 % black, whereas most suburban relocatees at the time of survey 15–20 years after initial move lived in neighbourhoods that were 16 % poor and 48 % black, on average. However, over time they tended to move (or have their neighbourhoods change) in such a way that some of the initial drop in the percentage of black residents in the neighbourhood was erased. Even more impressive was the durability of these gains for the second generation (Keels et al. 2005). Both original heads of households (mostly mothers) who moved and their children who moved with them but were adults by the time of follow-up research showed impressive persistence of residential environments; see top panel of Table 11.3.

Table 11.3 Comparison of origin and subsequent neighbourhoods for suburban participants in Gautreaux I and II programs (all participants are black)
Full size table

There was a supplementary phase of the Gautreaux litigation (commonly called Gautreaux II) that commenced in 2002, which provided another few hundred more HCVs (Pashup et al. 2005). Unlike the first phase, these HCVs had to be used in neighbourhoods that were less than 30 % black-occupied and less than 23.49 % (the city-wide average) poor. Compared to Gautreaux I suburban movers, the second-phase movers evinced larger initial reductions in neighbourhood percentages of poverty and black residents, but a much stronger erosion of these contextual gains over a shorter period; see the bottom panel of Table 11.3.Footnote 13

The Moving To Opportunity Demonstration

The Congressionally authorized MTO demonstration operating from 1994 to the late 1990s also employed HCVs but differed from the Gautreaux programs in many ways (Goering and Feins 2003; de Souza Briggs et al. 2010a, b). It is an experimental research effort undertaken in five metropolitan areas, not an effort to redress past discrimination in one city. It was established as a classic random assignment experiment, wherein families with children living in public housing complexes in highly disadvantaged neighbourhoods who volunteered to participate were randomly assigned to one of three groups and then tracked over a now nearly 20-year span. One was the control; one was given a HCV with neither restrictions nor mobility assistance (like the generic HCV program); the third (“experimental” group) was given a HCV that only could be used in a neighbourhood with less than a 10 % poverty rate but also provided mobility assistance by a local non-profit agency. Importantly, the experimental group (like all Section 8 tenants) was only required to remain in their initial, low-poverty neighbourhood for 1 year.

Feins (2003) analyzed the initial moves of participants. She found a 35 %-point reduction in average neighbourhood percentage poor for experimental movers and a 23 %-point reduction for HCV holders, in contrast to the comparison group. Much more modest reductions (8–10 % points) were observed for average neighbourhood percentages of black population. But these early gaps significantly narrowed by the time of the interim study 5 years after initial assignment (Orr et al. 2003; Clark 2005); see Table 11.4. This narrowing occurred because: (1) many experimental households moved to higher-poverty, higher-percentage black neighbourhoods after the first and second year of assignment to a low-poverty neighbourhood; (2) many neighbourhoods initially occupied witnessed rising trends in poverty; and (3) many control households moved out of public housing, often due to their fears of crime and gangs or the demolition of their projects due to HOPE VI or court-ordered desegregation plans (Clark 2005).

Table 11.4 Comparison of neighbourhood characteristics, by MTO control group, experimental movers, and HCV movers (5 years after random assignment; primarily black with some Hispanic participants)
Full size table

Nevertheless, as Table 11.4 shows, on every quantitative measure of neighbourhood employed, both the HCV group and the experimental group attained a superior neighbourhood environment than the control group 5 years after initial assignment. The same can be said when the battery of participant-assessed characteristics of neighbourhoods is considered; see Table 11.5. However, the gaps in either objective or subjective measures of neighbourhood were considerably narrowed when HCV and experimental groups were contrasted, though the latter resided in superior environments in every aspect except household member victimization.

Table 11.5 Comparison of MTO participants’ assessments of neighbourhood, by control group, experimental movers, and HCV movers (5 years after random assignment; primarily black with some Hispanic participants)
Full size table

Subsequent qualitative investigations of movers in MTO turned up some additional insights about altered neighbourhood conditions. de Souza Briggs and colleagues (2010a, b) concluded that, in addition to notable gains in mental and physical health, the major environmental gains experienced by the experimental group were gender-specific. Girls in experimental families gained substantially from the reduced stress associated with enhanced personal security in their new locations. In particular, they were removed from rampant predatory behaviours, including gangs who pressured them into early sex. A less salutary outcome was also revealed: 70 % of experimental household children were attending same school district as originally (Orr et al. 2003). This was produced by a combination of short-distance moves, open (non-geographically based) enrolment policies of local schools, and parents who did not wish to disrupt children’s social and school networks.

Other Efforts at Deconcentrating HCV Holders

As suggested in the introduction, there were additional efforts initiated by the federal courts or by HUD to use HCVs to significantly alter the geographic outcomes of HCV recipients. Besides the famous Gautreaux case noted above, there were over a dozen other PHA racial desegregation case settlements begun in the 1980s and 1990s that used HCVs (Goetz 2003; Popkin et al. 2003). Two other HUD-initiated programs of the 1990s tried to change the geography of HCV use, though their efforts have never been evaluated systematically (Schwartz 2010). The Regional Opportunity Counselling Program was established in 1997 in 16 metropolitan areas. It tried to build collaborations between central city and suburban housing authorities designed to increase residential options for HCV holders by overcoming bureaucratic barriers and offering counselling assistance. The Vacancy Consolidation Program, targeted at public housing developments slated for demolition in 15 housing authorities, also provided encouragement and counselling for relocates using HCVs. Neither imposed any requirements on the types of neighbourhood that recipients must select.

Housing Development Through the Low-Income Housing Tax Credit Program

The LIHTC program was created as part of the Tax Reform Act of 1986, which, among other things, removed substantial tax incentives for investments in rental housing development. The LIHTC is administered by the Department of Treasury (not HUD), which grants a per capita value of tax credit allocations to each state’s designated housing agency. Subject to broad guidelines, each state develops criteria for awarding these tax credits and holds annual competitions among prospective developers for projects designed with a minimum share of “affordable” units.Footnote 14 Developers awarded the credits sell them (at prices that reflect a variety of market and institutional conditions) to large companies seeking tax relief via a secondary market specifically designed for this purpose, thereby providing equity to the development. Designated LIHTC units must remain affordable for 15 years (Schwartz 2010). The subsides involved are not as deep in prior site-based federal assisted housing programs and rents are not set at 30 % of income, and thus the clientele targeted by the LIHTC program (typically earning 50–60 % of metro area median incomes) is not as low-income as typically served by the public housing or HCV programs (Khadduri and Wilkins 2008). As of 2009, the LIHTC program was assisting 21.5 % of all households receiving federally financed housing aid (Schwartz 2010: Table 1.1) and represents the largest contemporary producer of affordable housing.

Several early studies of the program revealed that there was a clear tendency for most LIHTC units to be built in areas of higher-than-average poverty and minority concentrations (Newman and Schnare 1997; Roisman 1998; Freeman 2004; Rohe and Freeman 2001), especially if they were located in central cities. McClure (2006) found, however, that as the LIHTC program has evolved it has increasingly developed units in the suburbs; in the most recent year analyzed (2002) almost equal shares went to central cities and suburbs. After examining construction produced over all the first 15 years of the program, McClure (2006) showed that 29 % of all LIHTC units were built in neighbourhoods with less than 10 % poverty rates, and only 8.5 % were built in those with higher than 40 % poverty rates. Nevertheless, these figures were virtually identical to those for all renters earning less than 50 % of the metro area median income (27 and 9 %, respectively), suggesting that little deconcentration was achieved by LIHTC developments.. The same conclusion was reached by McClure (2008) in a follow-up investigation of the locations of the 80,000 LIHTC units put into service nationally during 2002. He found that only 32 % were located in neighbourhoods with less than 10 % poverty rates. Moreover, this figure was 2 % points lower than the average poverty rate in the locations of unsubsidized renters in the same income bracket as most LIHTC tenants. Interpretation of these statistics must be done with caution, however, as we know nothing about LIHTC occupants’ origins. We therefore have no idea if occupants in the suburban, low-poverty area LIHTC projects were primarily relocatees from poor, central city neighbourhoods or suburbanites.

Moreover, the independent geographic impact of the LIHTC program is particularly difficult to assess due to the functional overlap between this program and the HCV program. States have often encouraged LIHTC developers to lease to HCV holders because otherwise very low-income unsubsidized renters would often be unable to afford rents in these projects. Williamson and colleagues (2009) discovered that 63 % of Florida LIHTC developments housed HCV holders, an average of 9 % of the tenantry per development, constituting a whopping 16 % of all the HCV holders in the state. Thus, it is likely that the siting of LIHTC units affects opportunities for a non-trivial number of HCV holders and, conversely, the impact of the LIHTC program is strongly influenced by the presence of the HCV option. It is also significant that while only 12 % of LIHTC units were located in Florida concentrated poverty neighbourhoods, 30 % of all voucher-holders living in LIHTC units resided in such neighbourhoods. The authors conclude for Florida that LIHTC contributed to concentration of disadvantage directly via their siting and indirectly by drawing disproportionate numbers of voucher holders to these distressed neighbourhoods. We do not know the extent to which this conclusion can be generalized.

Mixed-Income Redevelopment of Former Public Housing Estates Through the HOPE VI Program

Initiated in 1994 (in the same statute as MTO), the sixth program within the Housing Opportunities for People Everywhere (HOPE) programs was saddled with a host of oft-conflicting goals which partly frustrated the objective of poverty deconcentration (Turbov 2006). The program called for the revitalization of “severely distressed” public housing sites (characterized by physical decay, high vacancies, drugs, gangs and violence) through locally developed PHA-private developer/financier partnerships. These partnerships competed for HUD grant allocations, which helped finance the demolition and rehabilitation of public housing units, the construction of new units on site, the temporary relocation of displaced tenants, and the provision of HCVs to displaced tenants who were unwilling or unable to return to the redeveloped sites. Though there are vast differences in the features of the redeveloped sites (Popkin et al. 2004), a universal feature is a mixture of public housing tenants with those of higher incomes and, often, some owner-occupants. In all cases there has been a net reduction in the number of units for public housing tenants on site. HOPE VI was discontinued in 2006, though new variants are now being proposed under HUD’s “Neighbourhoods of Choice” rubric. All totaled, HOPE VI demolished about 150,000 dilapidated public housing units in 224 different projects nationwide (Landis and McClure 2010).

The net geographic impact of the HOPE VI program is a complex amalgam of both who ends up residing on the redeveloped sites and what happens to those who were displaced. Among the last group, some moved to other public housing projects, some were able to use HCVs (and go to other communities comparable to those of other HCV holders), and some were not and had to fend in the private rental market. The national HOPE VI tracking study found that after the first eight years of the program only 19 % of original residents were living on the redeveloped sites, 29 % were in other public housing, 33 % were using HCVs, and 18 % had left housing assistance (Popkin et al. 2004), proportions that roughly matched those obtained in a slightly earlier study of a different sample of 73 projects (Kingsley et al. 2003).

The most detailed information about the geographic outcomes associated with residents of PHA sites that were redeveloped under HOPE VI has been provided in Buron’s (2004) study of eight longitudinal case study sites. He found that comparing initial conditions on-site to those at first HCV relocation residence, the average neighbourhood poverty rate dropped from 40 to 28 %, but the average share of minority residents only dropped from 92 to 87 %. The main gain was in residents’ perceptions of safety: reports of “big problems with…:” “shootings and violence” fell from 67 to 20 %, “people selling drugs” fell from 77 to 30 %, “gangs” fell from 49 to 17 %, and “people being attacked or robbed” fell from 25 to 9 %. However, these effects were distinctive according to whether the relocatees moved to another public housing unit on the site being redeveloped, or moved off-site without assistance, with a HCV, or into another public housing development elsewhere; see Table 11.6. Nevertheless, relocatees on average experienced neighbourhoods that they perceived as much safer than the original HOPE VI sites before redevelopment (cf. Popkin and Cove 2007).

Table 11.6 Comparison of HOPE VI participants’ assessments of neighbourhood, by type of assistance and location (primarily black with some Hispanic participants) All figures as percentages
Full size table

Kingsley and colleagues (2003) studied geographic outcomes for movers from all 73 HOPE VI sites as of 2000. They found that 31 % used HCV, 49 % went to other public housing and 20 % moved elsewhere without assistance. On average, relocatees moved 3.9 miles from their original HOPE VI site, saw a decline in their neighbourhood poverty rates from 61 to 27 % and a decline in their neighbourhood percent minority from 88 to 68 %. Relocatees using HCVs followed roughly comparable patterns of clustering as those in the generic HCV program, though relocates were slightly more likely (13 % vs. 10 %) to cluster in tracts that already had 10 % HCV households or more.

Buron and colleagues (2007) also found that most HOPE VI relocatees using HCVs saw a large improvement in their neighbourhood quality in terms of poverty rates and safety, compared to their former public housing estates. However, these relocatees not only faced the normal challenges as general HCV holders but also the extra adjustments associated with moving out of public housing (e.g., being responsible for timely utilities and rent payments). The fact that there were no additional program funds allocated within the HOPE VI program to counsel and assist such HCV-using relocates move to substantially lower poverty, lower minority neighbourhoods implicitly suggests their lack of serious commitment to the deconcentration goal. Thus, there should be some concern over the sustainability of these initial gains by HOPE VI relocatees using HCVs.

Over the past decade, the Chicago Housing Authority (CHA) has been the national leader in transforming most of its public housing estates in HOPE VI-like manner, though using additional sources of funds beside this federal program. As such it has been the object of intense study that, among other things,Footnote 15 has reinforced some of the worrisome prospects concerning geographic sustainability noted above. Popkin and colleagues (2000) found that former CHA public housing residents had different and more severe needs that inhibited their successful leasing up of apartments with HCVs than generic HCV holders. For example, because of rampant gang activity and drug markets, it was common for CHA relocatees to have at least one member of their family with a criminal history, rendering them both disqualified to return to CHA units on the redeveloped site and easy to deny by private landlords in the rental market. Many CHA relocatees also had severe personal challenges (e.g., mental and physical disabilities; responsibilities for many children) that made it difficult to find appropriate private rental dwellings and successfully lease up. Most had no experience in searching for housing, interfacing effectively with landlords or, once housed, behaving appropriately as a private tenant (such as paying utility bills or allowing informal occupancy arrangements in violation of leases). But the problems were not only confined to relocatees. Popkin and colleagues (2000) found in that CHA residents who lived on the original sites but now occupied the redeveloped sites were having trouble complying with the new, tough lease requirements because they were not getting the supportive services they needed. This evidence speaks to the minimal successes that HOPE VI has had in substantially increasing housing opportunities for former public housing residents in non-poor environments.

Comparing Geographic Outcomes Across U.S. Assisted Housing Programs

The prior analysis has relied upon studies that have essentially made within-program comparisons of geographic outcomes for participants. Here I turn to the handful of studies that used common bases to compare outcomes across programs. An introductory note of caution in interpreting the following results is in order because of the functional overlaps between many programs. HCV holders may have been moved under the auspices of the generic program (while many stayed and leased up in place), the HOPE VI program, or the “vouchering out” of tenants in privately owned subsidized apartments developed under the Section 8 or Section 236 programs noted below. Moreover, some number of HCV holders reside in LIHTC units.

In the first cross-program comparative study of geographic outcomes of an older set of programs, Newman and Schnare (1997) found during the early 1990s that only 15 % of HCV holders resided in neighbourhoods with 30 % or higher poverty rates, which compared favourably to 54 % of residents in public housing and 23 % of residents in privately owned, HUD-subsidized projects.Footnote 16 Only 5 % of HCV holders resided in neighbourhoods of over 40 % poverty, compared to 36 % of residents in public housing and 13 % of residents in privately owned, HUD-subsidized projects. These data form a pattern that has often been observed subsequently: among those receiving U.S. federal housing subsidies, public housing residents generally live in the most-disadvantaged neighbourhoods, followed by residents in other-site-based assistance programs, followed by HCV holders residing in the private rental market.

Pendall (2000) provided a more comprehensive, cross-program comparative study of geographic outcomes; see Table 11.7. His figures showed that, on average, residents of public housing (25 % of all assisted tenants as of 1998) experienced the highest rates of neighbourhood poverty (36 %), minority occupancy (59 %), and renter occupancy (74 %). Mean neighbourhood features of units produced by the Section 8 Moderate Rehabilitation and miscellaneous site-based programs (10 % of all assisted tenants) ranked closely behind. Next in rank, with virtually identical conditions for residents, came locations associated with the LIHTC program (7 % of the total), Section 236 new construction subsidy program (9 % of the total), Section 8 New or Substantial Rehabilitation program (19 % of total), and HCV program (30 % of total). This cluster of both site-based and tenant-based subsidy programs had their average tenants occupying neighbourhoods that were: 20–21 % poor, 34–41 % minority, and 60–67 % renter-occupied. Compared to public housing, these differentials were greatest in the aspect of neighbourhood poverty rates and least in the aspect of neighbourhood renter-occupancy rates.

Table 11.7 Comparison of participants’ neighbourhood characteristics, by Federal Housing Assistance Program
Full size table

Important geographic nuance to this portrait has been provided by McClure (2006), DeFillippis and Wyly (2008), and Talen and Koschinsky (2011). In his nation-wide study, McClure (2006) discovered that neighbourhood poverty rate differentials between the HCV and LIHTC programs and compared to unsubsidized, lower-income renters depended upon whether a central city or suburban location was considered; see Table 11.8. In all cases the average neighbourhood poverty rates experienced by program participants and generic low-income renters were roughly twice as high in the central cities. However, whereas in the central cities the HCV holders’ mean neighbourhood poverty rate was 3 % points less than residents in LIHTC developments, the reverse was true (by 1 % point) in the suburbs. Nevertheless, in both geographic contexts the HCV holders only slightly “out-performed” what McClure used as their unsubsidized comparison group (renter households in poverty) and LIHTC residents slightly “under-performed” what McClure used as their unsubsidized comparison group (renter households earning less than half the metro area median income). From a different perspective, the performance of the LIHTC program in expanding options in the suburbs appears more favourable. McClure (2006) showed that if only suburban destinations are considered, a substantially higher share of all units provided by the LIHTC program go to low-poverty (0–10 %) neighbourhoods compared to shares of HCV holders (50 % vs. 43 %). Nevertheless, Table 11.8 makes it clear that, on average, neither the LIHTC nor the HCV program operating in either central cities or suburbs produces a substantially different distribution of low-income households by neighbourhood poverty rates than what is produced by comparable unsubsidized renters in the market place.

Table 11.8 Comparison of participants’ neighbourhood poverty rates, by federal housing assistance program and area where assistance used
Full size table

DeFillippis and Wyly’s (2008) study of New York City revealed that HCV holders were not more likely than residents in subsidized housing (supported by either the federal government and/or the city itself) to live in lower-poverty or less minority-concentrated neighbourhoods. They concluded that, especially in tight housing markets (partially made so by rent control in the case of New York) and markets undergoing much gentrification (such as New York), preserving the site-based subsidized housing stock was more important for poverty deconcentration efforts than more vouchers. But this is debatable as the sole goal.

Talen and Koschinsky (2011) found in Chicago that although on average HCV holders experienced less poverty and minority concentrations in their neighbourhoods than residents of site-based assisted units, there was an important distinction related to degree of concentration of assistance in the neighbourhood. In areas with high concentrations of either HCV or assisted sites, the above relationship was reversed, and the households in site-based assisted developments lived in more advantaged neighbourhoods. This suggests that when forces lead to high concentrations of HCV holders in particular neighbourhoods it often erases their potential locational advantages. I explore this potential endogeneity further below.

Explanations for Findings: Individual, Structural, and Administrative Rules

Taken at face value, the aforementioned studies lead to the following conclusions:

  • 1. Residents of U.S. public housing on average reside in significantly more disadvantaged neighbourhoods compared to participants in any other assisted housing program and most other low income renters.

  • 2. Residents of other types of site-based assisted housing programs (particularly LIHTC) do not, however, reside in significantly different residential environments than tenant-based HCV holders

  • 3. HCV holders fare somewhat better in neighbourhood poverty rates than equivalent households who do not receive subsidies, but the comparative differences are even smaller when the LIHTC program is compared to its equivalent private renter standard.

  • 4. HCV holders in general do not substantially improve their neighbourhood circumstances with subsequent moves; indeed if their initial move was (perhaps with counselling assistance) to a low-income, predominantly white neighbourhood, their subsequent moves were to higher-poverty, higher-minority share neighbourhoods.

Why do these patterns emerge? There is little debate regarding conclusion 1. Due to their construction at large scales and high densities, their explicit history of racial segregation, their historical evolution to house only the neediest, and the concomitant negative spill over effects on their environs, traditional American public housing has almost tautologically resulted in concentrations of disadvantage (Hirsch 1983; Goering 1986; Julian and Daniel 1990; Massey and Kanaiaupuni 1993; Schill and Wachter 1995; Coulibaly et al. 1998).

There is more contention over the sources of conclusions 2., 3., and 4. There are three not mutually exclusive but distinct sets of arguments here: the “individualist,” “structuralist,” and “program rules” explanations. The individualist view focuses on characteristics of the program participants that influence how they use their HCV geographically, such as preferences, housing search patterns, social networks, personal psychological and intellectual resources, family responsibilities, criminal histories, and the like (e.g., Clark 2005, 2008). The structuralist view emphasizes geographic constraints imposed on program participants by the operation of metropolitan housing markets (such as low vacancy rates, racial discrimination, and selective participation of landlords in the HCV program) and public transportation systems (e.g., de Souza Briggs et al. 2010b; DeLuca and Rosenblatt 2011). The program rules view emphasizes the constraining impacts of the regulations associated with the major housing programs in question and their administration in local housing markets.

Unfortunately, some research findings do not help us sort out the individualist and structuralist explanations for the geographic performance of HCV holders because they are consistent with both. For example, a common consensual finding emerges from a variety of investigations of the geographic impact of the HCV program: ethnic-racial differences. Inferior outcomes were uniformly associated with minority ethnic status, even after controlling for other characteristics. Being black was especially associated with reaping small geographic gains from use of HCV (Hartung and Henig 1997; Newman and Schnare 1997; Turner 1998a, b; Pendall 2000; Basolo and Nguyen 2005; McClure 2008). This, of course, could be the result of all the (unmeasured) individualist factors above that are correlated with race, and/or racial discrimination in rental markets. As another example, Turner (1998b) found that in most metropolitan areas she studied there were greater shares of below-FMR units in low-poverty areas than shares of HCV holders residing there, suggesting something either about the search patterns and/or preferences of the HCV holders and/or the willingness of landlords in such areas to participate in the voucher program.Footnote 17 However, other research clearly offers support to elements of all three positions; I discuss these next.

Evidence Supporting the Individualist Position

There is convincing evidence that low-income households in general and HCV holders in particular are deeply embedded in highly localized social networks. This “bonding social capital” can provide invaluable sources of support (money, child-care, in-kind assistance) and information, though this also sometimes comes with a burdensome set of responsibilities (de Souza Briggs et al. 2010a, b). These networks can provide a geographic centre of gravity for residents who are granted HCVs in two ways: spatially biased information and a need for proximity for assistance family can offer. HCV holders without counselling support typically activate local networks to help them locate a place to use their voucher (Deluca et al. 2011). Unfortunately, given the constrained geographic scope of these networks and other search strategies (such as personally looking for “For Rent” signs), few vacant apartments in advantaged neighbourhoods are uncovered. Even if they have information about rental opportunities in advantaged neighbourhoods, HCV holders may not wish to move there if the distance to their kin and friendship networks and institutional ties seems prohibitive. A particularly poignant if unique example was provided by Goetz (2003), who observed Asian immigrant residents of a Minneapolis public housing project vigorously opposed moving in compliance with a court-ordered desegregation decree and, when issued a HCV and forced to leave, stayed very close. Moreover, with little first-hand or second-hand information about alternative neighbourhoods, new HCV holders often have limited bases upon which to compare them and make more globally informed choices. de Souza Briggs and colleagues (2010a, b) determined that many MTO movers focused mainly on avoiding danger, not moving to “places of opportunity.” But some also moved to escape predatory relatives and neighbours. They did not choose what might be perceived as much better options because they had never experienced them and thus did not know what they were missing; “information poverty” they called it. In their extensive, open-ended interviews with low-income black households in Baltimore and Mobile AL, Deluca and colleagues (2011) found little salience of “neighbourhood” in the residential choices of their interviewees, other than a desire for relative safety at the small geographic scale around the dwelling; dwelling characteristics dominated the selection process.

Of course, lack of information with geographic breadth is not a pure individualist trait but likely is reflective of housing market structure, as amplified below. Furthermore, evidence on moving destinations of HCV holders should not be taken as proof of “revealed preference.” Mobility reflects a variety of structural constraints, including limited information and limited housing options, not just preferences. This point gains powerful nuance with recent discoveries by Deluca and colleagues (2011). Remarkably, they found that 70 % cited reasons for last move that were beyond their control, what the authors called “reactive moves.”Footnote 18 Dwelling unit failure was the most frequently cited cause of mobility (25 %). Reactive moves must occur in a matter of weeks to avoid homelessness, so expediency is salient. Not surprisingly under these circumstances, search processes rely upon “leads” from family and friend networks and seeking nearby “for rent” signs being paramount, with highly localized moves aimed at securing decent dwellings (not necessarily decent neighbourhoods) being the result. Though family ties were activated by necessity during these reactive housing searches, many did not express a desire to do so or to retain close contacts. Similarly, though low-income black households typically moved to places with high proportions of black neighbours, it was not because they in any way “preferred” that racial composition, contrary to the conclusions of Clark (2008).

Evidence Supporting the Structuralist Position

Many studies clearly indicate various types of housing market structural barriers that influence where HVC holders reside. One set relates to the availability of vacant, appropriately priced rental units located in low-poverty, low minority concentration neighbourhoods whose landlords are willing to participate in the HCV program. Pendall’s (2000) regressions based on a nationwide sample of HCV holders showed that structure of local housing market, specifically the supply of rental housing in non-distressed and distressed tracts, was a key determinant of the share of HCV holders living in distressed tracts, controlling for metro area poverty and racial composition. Turner’s (1998b) study of HCV holders in six metro areas found substantial differences in outcomes depending on local housing market conditions. In a few metropolitan areas the combination of good locations for public and other site-based assisted housing and tight private rental markets led HCV holders to underperform their site-based counterparts in terms of poverty and minority neighbourhood indicators. These conclusions were echoed in the subsequent study by Finkel and Buron (2001) involving more metropolitan areas. The MTO research documented substantial differences in the ability of experimental and generic HCV groups to lease an apartment due to the relative paucity of units available in low-poverty neighbourhoods (Orr et al. 2003). Even with their counselling, the experimental group’s lease up rate was 14 % points lower than the HCV participants,’ though much higher than the success rate in Gautreaux (Shroder 2003). Subsequent qualitative research has further emphasized how difficult it was for MTO experimental households to find housing in low-poverty neighbourhoods, even with the assistance of counselors (de Souza Briggs et al. 2010a, b). Finally, tight rental markets in the Chicago suburbs forced the Gautreaux I program to modify its desegregation criteria (Rubinowitz and Rosenbaum 2000), as noted above.

A closely related structural bias induces HCV use in disadvantaged neighbourhoods because landlords often eagerly recruit HCV holders there. In these neighbourhoods private landlords are more likely faced with high vacancies and respond by aggressively marketing their units to voucher holders (Galster et al. 1999), especially at local housing authorities (Deluca et al. 2011).

Taken holistically, the studies discussed in the preceding two paragraphs clearly indicate synergistic biases in the way housing markets operate to limit geographic opportunities for lower-income households, whether they have HCVs or not. Areas of opportunity are often inaccessible to HCV holders because they are too expensive, have few vacant rental units, and/or have few landlords willing to participate in the program, all precisely because they are areas in high demand by more affluent segments of the housing market. Simultaneously, areas of disadvantage have all the opposite factors at work: lower rents, many vacant rental units, and landlords so desperate for tenants that they actively recruit HCV holders. Thus, the same market forces that produce income group segregation remain powerful determinants of the geography of HCV holder residence.

What’s more, racial differences in structural barriers are also apparent. Basolo and Nguyen’s (2005) study of HCV users in a large Southern California PHA found that HCV users’ perceptions of barriers to mobility were primarily “too few homes to rent” and “landlord will not rent to Section 8.” Huge racial differences in responses were revealed, as blacks and to a lesser degree Hispanics were more likely to cite these obstacles than whites, with the incidence of black responses at least 22 % points higher than whites’. Finally, the most recent national investigation of racial discrimination that employed carefully matched “testers” revealed substantial rates of differential access afforded black and Hispanic apartment seekers (Turner et al. 2002). Although the study did not have the testers use HCVs, we can presume that such discriminatory barriers based on race would be of relevance to the outcomes of black and Hispanic HCV users. If a landlord wished to discriminate illegally against a minority HCV holder, it is an easy and virtually undetectable subterfuge to merely decline on the legal basis of unwillingness to participate in the HCV program.

A final structural barrier is that many holders of a new HCV cannot lease up because they lack the requisite savings to cover the costs of moving, security deposits, and other fees associated with acquiring a new apartment (Popkin and Cunningham 1999). Though there may be means of covering these costs through special emergency grant funds or charitable contributions, such often do not arrive in time before the voucher lease up period expires (Marr 2005).

Though all these structuralist scenarios provide plausible explanations for the modest geographic performance of HCV holders in any given cross-section, they are less persuasive explanations for the erosion of geographic advantages over subsequent HCV-supported moves, as observed by Feins and Patterson (2005) and especially in the MTO demonstration (Orr et al. 2003; Clark 2005; Turney et al. 2006). Here, different sorts of structural barriers must be brought to bear. First, once in an advantaged neighbourhood (perhaps through the assistance of mobility counsellors), HCV holders are often forced to move because the landlord is unwilling to continue participating in the program. After examining the geographic patterns of HCV holders who received their vouchers as part of a public housing desegregation suit in Baltimore, Deluca and Rosenblatt (2011) found that nearly half of those who had moved after 4 years from their original site (in a low-poverty, low-minority, low-assisted household neighbourhood) were forced to do so because landlords refused to continue participation. This echoed results from MTO (Orr et al. 2003), where the two most frequently cited reasons by experimental group households for leaving their first, low-poverty neighbourhood were “leasing problems” (22 %) and “problems with landlords” (20 %). Second, HCV holders in advantaged neighbourhoods may face harassment or more subtle forms of discrimination and ostracism (based on their class and/or racial status) that makes them uncomfortable and desirous of more diverse environs. Even more neutrally, they are unlikely to form close social bonds with their new neighbours or get deeply involved with new institutions, thus a subtle sense of alienation may remain (de Souza Briggs et al. 2010b). Third, these locations may raise insurmountable challenges to negotiate the spatial mismatch of home, work, socialization, and childcare, especially if the HCV holder lacks an automobile (de Souza Briggs et al. 2010a, b). MTO families that initially moved to low-poverty neighbourhoods often excessively distanced themselves from pre-existing job and social networks and eventually felt compelled to move closer to the urban core (Turney et al. 2006; de Souza Briggs et al. 2010a, b). Once any or all of these reasons trigger a move from the initial location, all the individualist and structuralist factors noted above return to play and produce the observed pattern of mobility into somewhat less-advantaged places, closer to the original, disadvantaged neighbourhood.

Evidence on Assisted Housing Program Rules and Administration

There are several fundamental elements in the design and administration of HCVs that limit their efficacy as a vehicle for deconcentrating poverty among recipients. First, the aforementioned asymmetry in landlords’ willingness to participate in the program (less in more desirable neighbourhoods and vice versa) would be rendered moot if all landlords were required to participate in the program, but this compulsion is contrary to current rules. Second, HCV Fair Market Rents (FMRs) have been consistently lowered since the inception of the program, from 50th to 45th to the current 40th percentiles of the metropolitan-wide distributions of rental units for the particular category of bedroom in question. Thus, the purchasing power of the HCV has been eroded and thus the regions over which recipients can afford to use it have shrunk. Third, the fundamental nature of the FMR creates an economic incentive for recipients toward HCV use in more disadvantaged neighbourhoods. Since the value of the HCV is based on metropolitan-wide rent distributions, a HCV holder can find cheaper accommodations in more disadvantaged submarkets within the region. Some PHAs permit reductions in tenant out-of-pocket contributions to rent if they can lease such below-FMR apartments, thus unwittingly providing an incentive for choosing disadvantaged neighbourhoods. Fourth, the time limitation of how quickly recipients must lease-up encourages them to settle on neighbourhoods with which they are already most familiar. Fifth, qualitative evidence suggests that some PHAs discourage those to whom they issue vouchers from using them outside of their jurisdiction (Marr 2005). Finally, the operation of the LIHTC program that works to recruit HCV holders to developments in more distressed neighbourhoods has been alluded to earlier (Williamson et al. 2009).

Why the LIHTC program does not generally out-perform the unsubsidized rental market in providing neighbourhood options for low-income households can be explained by program rules that encourage development of exclusively low-income projects in disadvantaged neighbourhoods. From the outset, the LIHTC program structure has not favoured poverty deconcentration, though some had hoped that this program could more easily overcome political opposition from suburbs than other subsidized housing vehicles. First, Treasury Department rules require that states favour applications for developments in “qualified census tracts” (QCTs) that are part of “comprehensive redevelopment initiatives.” QCTs are neighbourhoods wherein half or more of the residents have incomes below 60 % of the metro area median family income and the poverty rate is 25 % or more. Once granted to QCTs, the associated developers get 30 % more tax credits than would otherwise have been allocated. The QCTs ideally are areas that will be redeveloped for middle- or upper-income private housing, whereupon the location of a LIHTC development might create a more diverse community in the long run than otherwise would have been the case. Unfortunately, this “creation of an island of affordability in a sea of gentrification” has happened rarely; the gentrification has typically not materialized. Second, Treasury Department rules stipulate that at least one-tenth of all credits be allocated to non-profit developers; currently over a fifth are so allocated nationwide, on average. Most non-profit housing developers in the U.S. are community development corporations that are based in disadvantaged neighbourhoods, have substantial governance representation from these neighbourhoods, and focus on providing affordable housing to these places. Not surprisingly, these first two rules have resulted in a disproportionate number of LIHTC projects being built in areas that have remained primarily low-income and minority-occupied. Third, because program rules typically favour developments that provide larger proportions of affordable units, there is little incentive for developers to design mixed-income projects that would enhance economic diversity at the small geographic scale. Finally, within the federal guidelines there is latitude for states to value various aspects of applications for tax credits, including geographic criteria, and there have been few state schemes that that heavily favoured LIHTC developments that provide opportunities for deconcentration.

Policy Implications for U.S. and European Contexts

Proposals to Better Deconcentrate Poverty in the U.S.

For almost a quarter century there have been discussions of wide-ranging policy reform proposals aimed at (among other things) improving the geographic distribution of U.S. federal housing assistance, (see: Goering 1986; Goering et al. 1995; Turner 1998a; Turner and Williams 1998; Popkin and Cunningham 1999; Katz and Turner 2001, 2008; Pendall 2000; Achtenberg 2002; Galster et al. 2003; Grigsby and Bourassa 2004; Khadduri 2005; Popkin et al. 2004, 2005; McClure 2008; Khadduri and Wilkins 2008; Pendall 2008; de Souza Briggs et al. 2010a, b; Landis and McClure 2010). There seems to be an emerging consensus that what is required is a broad palette of reforms involving both supply-side (dwelling-based) and demand-side (tenant-based) housing strategies (tailored to the particulars of the local metropolitan market), plus complementary non-housing strategies. The suggested reform proposals have included:

Demand-Side Housing Assistance (HCV) proposals:

  • Direct leasing and brokerage for connecting HVC holders to market-rate rental housing and LIHTC developments in good neighbourhoods

  • Financial incentives to HCV holders and potential HCV landlords in desirable areas, such as raising Fair Market Rent levels there

  • Intensified pre-move mobility counselling and aid, coupled with post-move follow-up, support, and assistance when necessary

  • PHA performance incentives rewarding those who help HCV holders move outside disadvantaged neighbourhoods and promote a more effective use of inter-PHA portability of HCVs

  • End PHA administration of vouchers and contract to non-profit organizations with metro-wide coverage

  • Prohibitions on the use of HCVs in certain neighbourhoods/requirements that they can only be used in more “opportunity rich” neighbourhoods

  • Requirements for all landlords to participate in HCV program upon request

  • Making comparative school performance data more available to HCV parents

  • Beefed-up fair housing enforcement aimed at users of HCVs who are minority and families with children

Supply-Side Housing Assistance (scattered-site public housing, LIHTC, HOPE VI) proposalsFootnote 19:

  • Changing rules of LITC allocations to discourage development in poor neighbourhoods and to create more income mixing within developments

  • Changing the basis for state allocations of tax credits from per capita to favor those state with tighter housing markets

  • Limitations on where developments can be sited (“neighbourhood impaction standards”) to avoid concentrations of low-income or assisted households

  • Preserving affordable housing in gentrifying areas, perhaps by offering tax abatements or freezing assessed values for property tax purposes

  • Empowering metropolitan planning organizations to tie receipt of federal grants to suburban jurisdictions with their creation of “fair share” assisted housing development

Non-housing proposals:

  • Car vouchers to help navigate the tricky transportation requirements for home-work-childcare-church transitions

  • Attaching child-care vouchers and training assistance to housing assistance

  • Holistic matching of housing and other supportive welfare and educational services across agencies

Potential Lessons for Western Europe and Further Afield

What can housing policymakers in Western Europe learn from the past U.S. experience and this panoply of proposals? Before addressing this question I must emphasize that: (1) fundamental differences in U.S. and Western European housing systems make the origins of the poverty deconcentration problem and its resolution distinctive; (2) Western European housing systems cannot be generalized without perils of oversimplification; and thus (3) detailed policy recommendations for Western Europe based on American experience should not be made.Footnote 20

The first point is essential and bears amplification. On the demand-side of the equation, the Western European challenges related to poverty deconcentration are considerably less severe than in the U.S. Because tenant-based housing allowances are an entitlement and after-tax income distributions are much more compressed in Western Europe, there will be less severe neighbourhood sorting according to income transpiring through market processes. On the supply-side, there are several factors that also make the issue considerably different in Western Europe. These include a large social rental sector encompassing a wide range of incomes, centralized or regionalized planning systems that can exert direct control over where this social housing is located and how it is interspersed with market-rate dwellings, universal participation of private market landlords in housing allowance schemesFootnote 21 and, frequently, some form of rent restrictions and/or a relatively small and underdeveloped private rental sector. Because of these differences, the fundamental distinction in the origins of the problem of poverty concentration is that it is primarily market-driven in the case of the U.S. and state-driven in the case of Western Europe.

The fundamental difference in the nature of the potential solutions to the problem is that the U.S. has indirect and relative weak policy levers while Western Europe has the opposite. At its core, federal housing programs designed to assist lower-income households represent only 5.4 % of the entire U.S. housing stock and assists only a third of eligible households as of 2009 (Schwartz 2010). Thus, even if these programs were systematically designed to deconcentrate poverty (which, as I argued above, they are not) it is arguable that they would have only a modest impact on the geography of disadvantage. Moreover, U.S. federal policy is delivered against a backdrop of fragmented local governments that typically lack regional coordination for deconcentration efforts. Finally, the politics surrounding the deconcentration of poverty in the U.S. is indelibly stained with racism that arguably has constrained the aggressiveness with which any such initiatives could be pursued. Though such racial-ethnic issues certainly are present in Western Europe, I do not believe that they have attained the degree of longstanding cultural and political salience as they have in the U.S.

Despite these fundamental trans-Atlantic differences, I think some broad lessons from the American experience do have international traction. The efficacy of tenant-based (demand-side) efforts to deconcentrate poverty will be inversely related to the:

  • Tightness of the local metropolitan housing market in neighbourhoods that represent desirable destinations for assisted tenants

  • Extent that concentrated low-income households constitute racial-ethnic-immigrant minorities and the private housing market is balkanized by discriminatory barriers

  • Strength of local social ties among concentrated low-income households and the density of location-specific institutions purveying cultural capital to these communities

  • The degree of safety and other aspects about quality of residential life and public services available in neighbourhoods occupied disproportionately by the poor that would make them less likely to seek alternative locations

The efficacy of dwelling-based (supply-side) efforts to deconcentrate poverty will be directly related to the:

  • Regulatory powers granted to local public planning and housing development authorities to plan regionally in a dispersed manner to avoid concentrations de novo

  • Geographic area over which these powers may be exercised

  • Tightness of the local metropolitan housing market overall that will limit the ability of higher-income households to avoid living in mixed-income neighbourhoods

  • The nature of neighbourhood-based facilities and services, including schooling quality and jobs programs

Two Final Policy Caveats

In closing this policy discussion two caveats are in order. First, while there may be a general consensus on both sides of the Atlantic that concentrations of poverty are bad things that should be avoided to the extent feasible, the counterfactual is rarely specified in any detail. Neighbourhood diversity is hard to both define and make operational in practice. Five thorny practical issues arise in particular (Tunstall and Fenton 2006: 25–26; Kleinhans 2004; Galster 2013):

  • Composition: On what basis(es) are we mixing people in the deconcentrated neighbourhood alternative: ethnicity, race, religion, immigrant status, income, housing tenure…all, or some of the above?

  • Concentration: What is the desired amount of mixing in question? Which amounts of which groups comprise the ideal mix, or are minimally required to produce the desired outcomes?

  • Scale: Over what level(s) of geography should the relevant mix be measured? Does mixing at different spatial scales yield different outcomes?

  • Distance: How far away from the prior area of concentrated deprivation should low-income households be moved to achieve a more socially mixed neighbourhood?Footnote 22 And how is distance managed for low income residents who need services only offered in central places?

  • Speed: How rapidly do such programs need to be brought up to scale? Are demonstrations helpful in the transition?

Many different combinations of the above elements characterize different aspects of poverty deconcentration policies in different national contexts, though often not explicitly. Indeed, the counterfactual to concentrated poverty – “social mix” – is an intrinsically vague, slippery term; it is typically used to mean different things by different people. Planners and policymakers must be more precise and explicit in specifying the parameters of these five aspects of social mix before they can recommend specific policies and practices to deconcentrate poverty.

The second caveat relates to the efficacy of assisted housing policy in general to radically change the socioeconomic opportunities of low-income households and their families by changing their geographic contexts. I think the evidence is clear that for certain families in certain contexts the differences supplied by geography alone can be substantial. However, for many low-income residents of concentrated poverty neighbourhoods it will take more than changing their location given their durable and potentially constraining connections to social networks based in disadvantaged neighbourhoods and a variety of personal attributes that will continue to limit their upward mobility unless addressed directly (de Souza Briggs et al. 2010a, b; Goetz 2010). This is clearly recognized by the current administration at HUD, as embodied in the principles of their Choice Neighbourhoods program. Several local housing authorities at this writing are experimenting with new collaborations between local educators and other service providers to comprehensively and holistically help subsidized households improve their economic and social prospects (U.S. Department of Housing and Urban Development 2011).

Conclusions

At a descriptive level, three conclusions can be drawn regarding the impact of U.S. federal assisted housing policy on deconcentrating poverty. First, residents of U.S. public housing on average reside in significantly more disadvantaged neighbourhoods compared to participants in any other assisted housing program. Second, residents of other types of site-based assisted housing programs (particularly LIHTC) do not reside in significantly different residential environments than tenant-based HCV holders. HCV holders fare somewhat better in neighbourhood poverty rates than equivalent households who do not receive subsidies, but the comparative differences are even smaller when the LIHTC program is compared to equivalent private renters. Third, HCV holders do not substantially improve their neighbourhood circumstances with subsequent moves; indeed if their initial move is (perhaps with counselling assistance) to a low-income, predominantly white neighbourhood, their subsequent moves are generally to higher-poverty, higher-minority share neighbourhoods. In other words, the public housing program historically intensified the concentration of poverty and subsequent demand-side and supply-side policies have had relatively little impact in improving the neighbourhood conditions of recipients.

Why these subsequent policy efforts over several decades have produced so little deconcentration of poverty is subject to considerable debate. Arguments involving the characteristics of residents of concentrated poverty neighbourhoods (such a binding local social ties), arguments citing structural barriers of many sorts in the housing market, and arguments involving the rules and administration of HCV and LIHTC programs all have merit. It is extremely difficult to quantify precisely the relative contributions of these three strands of argument. Thus, I believe that all “pure” explanations should be rejected in favour of some middle-ground position.

What should be done in the U.S. to enhance the efficacy of assisted housing programs to deconcentrate poverty has been the longstanding object of vigorous debate. Some amalgam of supply-side and demand side reforms, coupled with non-housing strategies hold most promise. The U.S. experience in this regard offers several broad lessons to housing policymakers in Western Europe, even though there are vast differences in the origins and policy options available for addressing concentrated poverty.

Scholars should recognize how challenging it is to measure precisely the independent causal impacts on the residential geography of recipients emanating from specific programs providing federal housing assistance in the United States. There is selection bias in terms of program participation, with distinctly different clienteles participating in the various programs. There may be substantial functional overlaps and interrelationships among the programs. Though experimentally designed research holds promise in sorting out some of these confounding biases, it is costly. Perhaps most importantly, the major studies describing the geography of housing assistance assume that the locations and mobility behaviours of other households that define aggregate neighbourhood characteristics are unaffected by the geographic decisions of assisted households or developers of site-based assisted housing. Clearly this is untrue in some circumstances, as we know that over-concentrations of assisted housing can lead to endogenous neighbourhood reactions (Galster et al. 1999, 2003, 2008; Varady and Walker 2003a, b). Some interesting efforts to model holistically these dynamic neighbourhood mobility interrelationships among assisted and unassisted households are being undertaken at this writing by Owens (2011). Not only are the effects of current programs hard to discern, but the tools to incrementally build poverty deconcentration and neighbourhood mixing programs are in its infancy. All these realms offer fertile areas for future scholarship that are likely to yield important insights for policymakers interested in altering the geography of opportunity.