«Moving to opportunity voluMe 14, nuMber 2 • 2012 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
The articles in this symposium by Lisa Sanbonmatsu and co-authors and by Lisa A. Gennetian and co-authors summarize the mixed pattern of impacts that MTO had on the outcomes of adults and youth over the long term. In the fifth section of my introduction, I offer my own thoughts about what these results mean for social science hypotheses about neighborhood effects on adults and children. The MTO findings reject the hypothesis that “neighborhoods always matter,” because we did not find detectable effects on schooling or labor market outcomes across the five demonstration sites in either the interim (4- to 7-year) or long-term (10- to 15-year) followup studies. Because MTO enrolled a very disadvantaged set of families living in severely distressed areas, these findings may not generalize to less disadvantaged samples and settings. These are exactly the sorts of disadvantaged families who have commanded (for good reason) a disproportionate share of media and policy attention, however, and there is little in the existing research literature that would have predicted that the most disadvantaged families should be less affected by their neighborhood environments than are other types of families. MTO teaches us that neighborhood effects are more contingent than we had thought.
The fact that MTO moves had impacts on several important outcome domains—physical and mental health—that are to my way of thinking quite large in size also lets us reject the overly sweeping conclusion that neighborhoods don’t matter. What remains something of a puzzle is why neighborhood environments seem to matter much more for health than for other outcomes. I speculate about some answers to this question, which are motivated by some suggestive evidence that changes in neighborhood safety could be one of the key reasons behind MTO’s impacts on health outcomes.
In the final section, I consider the implications of MTO for public policy. Many people have concluded that mobility programs that are more intensive than MTO in terms of achieving changes in neighborhood or school environments of families may be necessary to change those outcome domains like schooling and employment that were not affected by MTO. This is, for example, the spirit of the articles in this symposium by Philip Oreopoulos, by Margery Austin Turner, and by Kathryn Edin, Stefanie DeLuca, and Ann Owens. My own reading of MTO and other research suggests this need not be the case. I also consider what we might learn from MTO about the design of community-level interventions, with a focus on safety, given the role this might play in driving the MTO impacts on health and the importance of safety to the MTO families themselves.
The MTO Experiment The MTO story begins in 1966 on the South Side of Chicago, actually not very far at all from my office at the University of Chicago. The first quasi-experimental evidence to support the idea that neighborhoods may exert large effects on poor families arose from a discrimination lawsuit against the Chicago Housing Authority (CHA) filed on behalf of an African-American public housing resident named Dorothy Gautreaux (Rubinowitz and Rosenbaum, 2000). As a result, starting in the 1970s, a total of 7,100 families were moved either into different parts of Chicago that were poor and segregated, but improving, or else into low-poverty, racially integrated suburbs (Keels et al., 2005).
A 1988 followup survey by Northwestern University sociologist James Rosenbaum found that moving to the suburbs instead of other parts of Chicago was associated with better job outcomes for mothers and schooling outcomes for children (Rosenbaum, 1995; Rubinowitz and Rosenbaum, 2000).
Rosenbaum’s findings were interesting and provocative, but left open the question of whether at least part of the difference in outcomes between Gautreaux suburban versus city movers might be due to other differences in the characteristics of the families themselves. Followup research has provided some support for this concern and has also tended to find smaller impacts on family outcomes (Deluca et al., 2010; Mendenhall, Duncan, and Deluca, 2006; Votruba and Kling, 2009).
The initial Gautreaux findings were nonetheless important enough to motivate HUD to sponsor the first true randomized experimental test of what happens to families when they move into very different neighborhood environments—the MTO demonstration. Eligibility for MTO was limited to low-income families with children living in selected distressed public housing or project-based housing in five cities: Baltimore, Boston, Chicago, Los Angeles, and New York. The housing projects from which MTO families came were among the most distressed in the country, with an average tract poverty rate of fully 53 percent. These projects were also extremely racially segregated.
Almost all MTO participants from the Baltimore and Chicago sites are African American, whereas the other three sites are split about evenly between African-Americans and Hispanics. There were very few White families in these housing projects, and as a result there are very few Whites in the MTO study sample.
Between 1994 and 1998, MTO enrolled 4,604 families. Surveys collected at baseline (exhibit 1) show just how disadvantaged those families were when they signed up for the MTO program. The average annual household income was $12,827 (in 2009 dollars). Fewer than two of five MTO household heads had a high school diploma, whereas three-quarters were on welfare.
Perhaps the most striking result from exhibit 1 is that over 40 percent of MTO applicants had someone in the household victimized by a crime during the 6 months before the baseline survey.
It is perhaps not surprising, then, that far and away the most important reason families signed up for MTO was safety. Three-quarters of MTO applicants said getting away from gangs and drugs was the first or second most important reason they signed up for MTO.
The families who volunteered for MTO were then randomly assigned them to one of three conditions.
The experimental group was offered the chance to use a housing rent-subsidy voucher2 to move into private-market housing in lower poverty areas. As part of the MTO design, the vouchers offered to families in this group could only be redeemed in census tracts with a 1990 poverty rate under 10 percent. Families had to stay in these neighborhoods for 1 year. If they moved before the year was up, they would lose their voucher. After their initial 1-year lease was up, they could use their housing voucher to move again, including moves into a higher poverty area. Families in this group also received housing search assistance and relocation counseling from local nonprofit organizations.3 The Section 8 group was offered a traditional housing voucher to move into private-market housing, with no special MTO-imposed constraints on where they move. Families in this group also did not receive any special housing mobility counseling beyond what is normally provided to voucher holders.
The control group did not receive access to any new services through MTO, but did not lose access to any housing or other social services to which they would otherwise have been entitled.
Random assignment in MTO helps overcome the self-selection concerns with previous observational (nonexperimental) studies by creating groups of families who are comparable in all respects but differ in the housing and neighborhood conditions that they experience. As a result, any differences across groups in their average outcomes can be attributed to the MTO mobility intervention itself.
Not all of the families who were offered an MTO housing voucher used them. Around 47 percent of those families offered an experimental group voucher and 63 percent of those offered a Section 8 group voucher relocated through MTO. Although many people outside the housing policy research community have been surprised that these takeup rates are not higher, the voucher utilization rates observed in MTO are generally similar to what has been found in previous studies of other housing voucher programs (Olsen, 2003; Rubinowitz and Rosenbaum, 2000). One reason some families do not move is because they cannot find a unit that is affordable under the voucher program rules, within the time limit that the voucher program allows families to search for housing. The difficulty of finding affordable housing in the allowable time may have been particularly challenging for Housing vouchers provide families with a subsidy for their private-market rent, equal to the difference between the local area Fair Market Rent (set to equal between the 40th and 50th percentile of the local metropolitan area’s rent distribution, depending on the city and year in question) and 30 percent of the family’s adjusted income (see Jacob and Ludwig, 2012, and Olsen, 2003, for details). The family’s required rent contribution is the same for public housing and housing vouchers and so receipt of a voucher does not free up any extra disposable income to families by enabling them to change their own out-of-pocket spending on rent.
The interim (Orr et al., 2003) and long-term (Sanbonmatsu et al., 2011) HUD technical reports summarizing the MTO results describe the three groups using the same terminology I use here: experimental, Section 8, and control groups. In some of our research team’s other writings (for example, Ludwig et al., 2011), we used instead the more descriptive terms low-poverty voucher group, traditional voucher group, and control group.
6 Moving to Opportunity
Guest Editor’s Introductionfamilies in the experimental group, who were restricted to looking in low-poverty census tracts.
Some families in the experimental group did not relocate because they did not attend all of the lifeskills counseling sessions that the local nonprofit organizations assisting with the housing search required them to complete before looking for housing. It is worth keeping in mind that many of the proposals to increase voucher takeup rates that have been suggested may create some difficult tradeoffs for policymakers.4 The fact that only some of the families who are offered MTO housing vouchers actually use them does not introduce any selection bias into our analyses (for additional discussion, see Ludwig et al., 2008). Families who are assigned to a voucher group who use a voucher are surely different from
those who do not. The analyses presented in this Cityscape issue show two types of estimates:
(1) the effect of being offered a housing voucher through MTO, known as the “intention to treat” (or ITT) effect and calculated as the difference in average outcomes of all families assigned to one of the treatment groups with all families assigned to control; or (2) the effect of actually moving with a housing voucher in MTO, known as the “effect of treatment on the treated” (or TOT), which is calculated using a method that preserves the strength of the MTO experimental design.5 It is also important to keep in mind when reading the MTO findings that the control condition in the MTO demonstration does not correspond to a situation of “no mobility.” Families in the MTO control group were allowed to move on their own, even if they did not receive any assistance through MTO to move. In addition, many of the public housing projects in which MTO families were living at baseline were demolished through HUD’s HOPE VI and other programs (see, for example, Katz, 2009), which further contributed to control group mobility.
Finally, we should be clear about what policy questions MTO can and cannot answer. MTO compares the effects of being offered a housing voucher with the chance to stay in public housing, which leads to sizable changes in neighborhood conditions (as I describe in the following section) but no change in out-of-pocket household spending on rent. This comparison helps answer the policy question of what would happen if we changed the mix of means-tested housing programs to include a larger share of housing vouchers and a smaller share of project-based units.
For example, one potential way to improve voucher takeup rates is to provide families with a longer window of time to search for units. This, however, creates the risk of reducing the share of vouchers that are being used by low-income families at any given point in time and increasing the share of voucher subsidies that are idle while families continue to search for housing. As an alternative, we could spend more money on housing-mobility counseling assistance for voucher recipients or efforts to encourage landlords to accept housing vouchers. Even if these efforts were successful in increasing voucher lease-up rates, spending more on these types of activities necessarily comes at the cost of diverting money that could have gone to providing actual housing subsidies to the two-thirds of income-eligible households in America who are not enrolled in means-tested housing programs (Olsen, 2003).
We do not try to estimate the effects of moving with an MTO voucher by doing something nonexperimental, such as comparing just the experimental group movers with the control group, because the families in the experimental group who move with a voucher are a self-selected subset of families assigned to that group—and so this self-selected subset cannot be compared with all the families assigned to the control group, which would be an apples-to-oranges comparison. Instead, we estimate the TOT in a way that exploits the experimental design of MTO, as follows. If we are willing to assume that being assigned to the experimental (or Section 8) group does not have much effect on families who do not use an MTO voucher to move, then the TOT effect will equal the ITT effect divided by the share of families assigned to the experimental (or Section 8) group who use an MTO voucher to relocate (H. Bloom, 1984). Because no control group families can use an MTO voucher by construction, the TOT estimate for some outcome of interest is basically the ratio of two ITT effects that are fully experimental—the ITT effect on the outcome divided by the ITT effect on use of an MTO voucher.