«Moving to opportunity voluMe 14, nuMber 2 • 2012 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
MTO is not the first study to take advantage of a lottery that randomly assigns some families to a housing voucher treatment group and others to a control group. Whereas other voucher experiments have compared voucher-based subsidies with no subsidies, however, MTO is unique in contrasting voucher receipt with living in subsidized public housing. All past voucher lottery studies have emphasized outcomes in the domains of labor force participation, welfare use, criminal behavior, and child and adult education. None of these studies observed consistent effects in these domains, and MTO is no exception to this disappointing picture. As MTO progressed, however, it became more health-focused than its predecessor studies thanks to its interim survey findings (Orr et al., 2003), which suggested that MTO reduced depression and anxiety among female heads of family and female youth (Kling, Liebman, and Katz, 2007). These interim findings led MTO researchers to increase the number of health assessments they made in the long-term followup survey, when the results (Ludwig et al., 2011; Sanbonmatsu et al., 2011) showed that, among adult women, upgrading neighborhood quality (1) maintained the superior mental health status previously noted, (2) reduced extreme obesity and diabetes, and (3) improved glycosylated hemoglobin levels (HbA1c).
This last is a biomarker of likely future cardiovascular complications also associated with diabetes and obesity. These conceptually consistent MTO health results suggest that all past voucher studies may have looked for effects under the light of the wrong lamppost. It is now clear that, because of how MTO evolved and what it discovered, health outcomes deserve a higher profile in research on housing in general and on housing vouchers in particular.
Of course, no single study can do everything. MTO has several features that make it look like an evaluation of the Section 8 Housing Choice Voucher Program (Section 8 Program), the current budget of which is about $16 billion per year. These features include the use of housing vouchers in both treatment groups, one of which was called the Section 8 group because group members could use their voucher just like any family exiting public housing with a voucher. Nonetheless, we argue that MTO has limited relevance as an evaluation of the Section 8 Program writ large because of its restriction to families who were living in public housing at baseline. Of course, MTO did not 170 Moving to Opportunity Making MTO Health Results More Relevant to Current Housing Policy: Next Steps set out to be an evaluation of the Section 8 Program. Rather, it sought to describe how an enriched neighborhood alternative affects many different adult and youth outcomes, including health, relative to living in public housing. It is not, therefore, the fault of the MTO research design that so few new voucher holders come from public rather than private housing or that some important program requirements that might affect outcomes differ between these two groups. We use the MTO demonstration not to cavil about how the study was framed, designed, or analyzed, but rather to describe some possible next research steps in the study of housing voucher effects on health.
The argument we make is along four main lines. First, MTO’s population does not represent the Section 8 Program population. MTO sought to maximize differences in neighborhood poverty concentration by studying public housing residents whose neighborhoods had some of the highest poverty rates in the United States and by requiring the principal treatment group of residents to use vouchers to move to neighborhoods with very low poverty rates. This dual strategy created the theoretically desired large neighborhood poverty contrast but, in so doing, led to a side effect that reduced MTO’s relevance to the Section 8 Program writ large. In the Section 8 Program, most families applying for a voucher are already in the private housing market and so are not receiving a public housing subsidy. They also tend to be better off, less frequently members of racial or ethnic minorities, less female-headed, and almost certainly healthier—given the usual gradient linking health to socioeconomic status (Adler and Stewart, 2010)—than public housing families. These differences mean that the Section 8 Program involves a less vulnerable population than the public housing families in MTO, leading us to ask: Would MTO’s health effects on public housing residents be replicated for the larger, more heterogeneous, and less vulnerable population of Section 8 Program voucher holders?
Second, the treatment contrasts achieved in MTO are greater than the mobility changes most Section 8 Program voucher holders spontaneously experience. Before their move, the average family in the Section 8 Program tends to live in less densely poor settings than public housing residents; when the family moves, it is probably to neighborhoods less affluent than those into which the MTO low-poverty housing voucher families were constrained to move, thus entailing a larger mobility difference in MTO than in the Section 8 Program. As we describe in the following sections, the MTO experimental group families moved from neighborhoods with about 50-percent poverty rates to those with about 10-percent poverty rates. Few families in the Section 8 Program make such dramatic neighborhood mobility changes. Of course, MTO also included a Section 8 group with no constraints on the poverty levels of the new neighborhoods. As we again describe in the following sections, however, the Section 8 group’s initial 50-percent neighborhood poverty rate exceeds that of the average participants in the Section 8 Program, who are already in the more affluent private housing market when they get a voucher. Therefore, the mobility treatment contrast is probably even greater in the MTO Section 8 group than in the Section 8 Program, in which former public housing residents are rare.
Third, MTO probably involved causal mechanisms different from those found in the Section 8 Program. Public housing residents can use their housing vouchers only to change neighborhood and residence. By contrast, most Section 8 Program families already in private housing can use some of their voucher’s monetary value to increase disposable income. More specifically, families already spending more than 30 percent of their adjusted income on rent in the private market—the vast
Cityscape 171Cook and Wing
majority of voucher holders—can reduce out-of-pocket spending on rent by using their new voucher to pay a portion of the rent they used to pay and pocketing the difference. Section 8 Program rules place limits on how much substitution is possible, with the total amount depending on a family’s income, their new and old rent payments, and local Fair Market Rent (FMR) values.
The more a family wants to pay in rent after receiving a voucher compared with their prevoucher spending, the greater the gain in housing quality. By contrast, when families opt for a lesser difference between premove and postmove rent, the implicit income supplement is greater. The difference between MTO and the Section 8 Program is that MTO families could use their vouchers only to move to better housing, and most new voucher holders in the Section 8 Program are free to choose how they trade off between increasing their housing quality and supplementing their disposable income. Section 8 Program rules—and Chicago data we present in the following section— indicate that neighborhood upgrades are therefore greater but income supplements are therefore less for MTO families. MTO activates one mechanism to a greater degree than the Section 8 Program, but the Section 8 Program can activate two mechanisms—better housing and more disposable income.
Fourth, how MTO’s health results would scale up to the national level is unclear because the Section 8 Program disproportionately comprises families leaving private-sector housing. These much greater numbers give rise to concerns about the limited supply of affordable rental units in neighborhoods with a poverty rate of 10 percent or less. Also, many poorer families are doubtlessly reluctant to relocate spontaneously in neighborhoods that are socially very different from those they know. MTO results we present in the following sections suggest this reluctance exists. By contrast, families already in private housing do not have to make such dramatic changes to move into neighborhoods that are 10 percent poor, and they are less likely to be racial or ethnic minority families reluctant to move into predominantly White settings. Therefore, scaling up MTO’s findings would probably be problematic in the larger Section 8 Program, in which public housing families are quite rare in housing lotteries.
For all four reasons, we argue that the MTO demonstration’s exciting health consequences cannot
yet be responsibly extrapolated to the Section 8 Program. We call for a new voucher lottery study:
a study in which (1) the population is all new Section 8 Program-eligible households, not just those currently living in subsidized public housing; (2) the variation in neighborhood poverty rate is one that spontaneously occurs rather than one that is experimentally imposed; (3) study families are free to use their vouchers not just for better housing but also to increase their disposable income;
and (4) the major outcomes are a wide array of health and biological statuses assessed, not just on adult females and youth as in MTO, but on young children as well.
The MTO Population Is Different From the Current Voucher Population The MTO participants were families living in public housing units in census tracts where at least 40 percent of the household incomes fell below the federal poverty line. In fact, the average tract poverty of the initial sample was 53 percent, emphasizing that concentrated poverty is especially prevalent in the public housing population. These facts make it plausible to assume that the MTO study population lived in worse housing and neighborhood conditions than current eligible 172 Moving to Opportunity Making MTO Health Results More Relevant to Current Housing Policy: Next Steps voucher holders in the Section 8 Program. Indeed, Jacob and Ludwig (2012) examined an expansion of the Section 8 Program in Chicago in the late 1990s. More than 80,000 people applied for a new voucher. About 90 percent were living in unsubsidized private housing when they applied, suggesting that any voucher evaluation results limited to public housing residents will not necessarily apply to the average voucher holder nationally.
In the same Chicago study, the average voucher applicant lived in a neighborhood with a poverty rate of about 29 percent. By contrast, the Chicago MTO sample’s average baseline neighborhood poverty rate was about 50 percent across MTO’s three groups. It seems likely, therefore, that the MTO study population is poorer than the overall voucher-eligible population and lives in poorer quality housing and worse neighborhoods. If so, these poverty rate differences are also likely to be associated with worse initial health status (Adler and Stewart, 2010), including the extreme obesity, diabetes, and HbA1c obtained in MTO. Were the families receiving vouchers through MTO initially less healthy in the aggregate than Section 8 Program voucher holders? If so, would MTO’s health findings be replicated with the relatively more healthy (and more economically advantaged) national population of Section 8 Program voucher holders?
The Average Size of MTO’s Treatment Contrast in Neighborhood Poverty Exceeds What We Would Expect in the Section 8 Program HUD designed MTO to maximize differences in neighborhood poverty concentration, so it chose a public housing population whose pretreatment poverty rate averaged 53 percent. Some families were then assigned to the MTO experimental group. After 1 year, those so assigned who actually moved were living in neighborhood tracts averaging 11 percent poor. This 42-percent difference in neighborhood poverty is very large and totally commendable from MTO’s theory-testing perspective.
The size of this contrast decreased over time. Ten to 15 years later, the control group had moved, on average, from 53 to 31 percent poor tracts, whereas the experimental group movers had gone from 53 to 21 percent poor tracts. Movers among the Section 8 group had gone from 54 to 24 percent poor tracts. Thus, by the end of the study, a contrast of about 10 percentage points characterized how the control group differed from both the experimental and Section 8 groups. In the following sections, we examine some reasons for this temporal decrease in contrast size. For now, however, we point out that the health differences between the control and experimental groups was always statistically significant and large enough over time to obtain health effects in intention-to-treat analyses with only modest compliance rates (Ludwig et al., 2011)—a considerable achievement.
From the perspective of evaluating the Section 8 Program, however, the pertinent question is, “How big of a neighborhood poverty contrast would we expect when members of the broader Section 8 Program population move?” It is impossible to know exactly, but consider the following. Members of MTO’s “traditional voucher group” were randomly assigned a Section 8 voucher and were free to move wherever they wanted. They initially moved to tracts with 29 percent poor, on average, appreciably better than the neighborhoods they left but not as affluent as the tracts, with an average 11 percent poor, to which the experimental group moved. The traditional housing voucher group, however, moved into neighborhoods with as many poor families (29 percent) as characterized the
Cityscape 173Cook and Wing
premove Chicago voucher applicants in Jacob and Ludwig (2012), a stark contrast to the neighborhoods with 53 percent poor when MTO began. It is therefore impossible for the Chicago group, beginning at 29 percent, to experience the MTO low-poverty voucher group’s poverty reduction of 42 percentage points. It would also be next to impossible to achieve the 24-percent reduction obtained in MTO’s Section 8 group.