«Moving to opportunity voluMe 14, nuMber 2 • 2012 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
With that qualifier in mind, one hypothesis is that the MTO health impacts may be due in part— perhaps even in large part—to changes in neighborhood safety. It is easy to see why MTO’s impacts on mental health outcomes like the K6 index of psychological distress might be due to improvements in neighborhood safety. An important role for safety in explaining MTO impacts on mental health would also be consistent with the fact that three-quarters of MTO adults said safety was one of the most important reasons they signed up to move as part of the program.
Safety might also be an important contributing factor to the large impacts of MTO on physical health measures like extreme obesity and diabetes, through the effects of safety on stress that previous research has in turn linked to sleep and metabolism. One reason to suspect this safety-stress-health link in MTO is by process of elimination: we do not see large, consistent MTO impacts on other candidate mechanisms around diet, exercise, and access to medical care, though it should be said that our measures of these mechanisms are not as detailed as one might ideally wish. It is worth keeping in mind, however, that the effects of moving through MTO on diabetes and extreme obesity are extremely large. It would be surprising if diet, exercise, or access to medical care could change by enough to explain entirely such significant MTO effects on diabetes and obesity yet give such little indication of change in the mediating measures included on our surveys.13 If safety is an important mechanism behind MTO’s health impacts, then why do we not also see MTO impacts on other outcomes like schooling? After all, Sharkey (2010) finds some evidence in the PHDCN data for very large (.50 to.66 standard deviation) short-term effects of neighborhood homicide rates on children’s achievement test scores. Perhaps the contrast between the PHDCN An alternative approach to understanding more about the mediating pathways through which MTO affects these health outcomes is that of Kling, Liebman, and Katz (2007): to use interactions of indicators for treatment group and demonstration sites as instrumental variables for different neighborhood characteristics as endogenous explanatory variables. This method basically estimates a dose-response relationship and asks whether, in those demonstration sites where randomization to a given treatment group generates a relatively larger change in a candidate mediating measure, randomization to that treatment arm also generates relatively larger changes in the outcome variable of interest. The method assumes that the only reason why different randomized groups in different sites respond differently to treatment group assignment is because they experience a relatively larger change in the mediating measure, not because of some other systematic variation across sites and groups in how people would respond to a given unit change in some neighborhood characteristic or other mediator. The method also assumes that the only pathway through which treatment assignment affects the outcome of interest is through the endogenous explanatory variables (the candidate mediators) included in the model. Given the large number of candidate mediators through which MTO might affect outcomes and the limited number of instrumental variables available with this design, this assumption will be challenging to meet. We can, however, interpret candidate mediators used in this way as markers or proxies for the collection of neighborhood attributes that covary (for example, Kling, Liebman and Katz, 2007, interpreted census tract poverty rates as a marker for a collection of features of neighborhood economic disadvantage that are correlated). Using this approach to explore the mechanisms behind MTO’s health impacts should be a priority for future work.
and MTO data could reflect in part the difference between the short-term and long-term effects of exposure to neighborhood crime and violence. Over the longer term, parents may engage in a variety of protective behaviors that try to shield their children from the harmful effects of dangerous neighborhoods, although, in principle, adaptations like this could wind up generating costs in other ways. When we examine the data in ways that extend beyond MTO’s pure experimental design, we see some hints that schooling outcomes for female youth could actually be better in more unsafe neighborhoods. One imagines children being kept inside more often in dangerous areas and so having more time to do homework, but that is just speculation. Understanding more about the safety-schooling link should be an important priority for future research.
Implications of MTO for Public Policy One way to read the MTO demonstration is as an evaluation of a program (voucher-assisted residential mobility) that policymakers might consider carrying out at scale. One thing we have learned from MTO is that this sort of mobility program can have surprisingly large, beneficial impacts on important mental and physical health outcomes. Whether these benefits from MTO are large enough to justify the costs of the mobility program is difficult to determine with the available data. As Olsen notes in his article in this symposium, the costs to government housing agencies of an MTO-like switch from public housing to housing vouchers is likely to be negative—that is, to save money. Some of the most important potential costs of MTO are unlikely to show up on any government budget spreadsheet, however. The whole logic behind MTO—that being surrounded by relatively more affluent neighbors could be good for the life outcomes of low-income families— raises the possibility that MTO moves could have adverse effects on other families outside of the MTO demonstration who are living in destination areas or the origin neighborhoods that MTO families left.
In principle, it could be that mobility programs like MTO are just a zero-sum game, with whatever benefits arise to MTO families from living in a lower poverty area being exactly offset by adverse impacts on other families in destination areas from experiencing an increase in the poverty rate of their neighborhood. If every family responds the same way to living in a neighborhood of a given type, and if the relationship between people’s outcomes and neighborhood poverty or other characteristics are linear (so that a 1-percentage-point change in tract poverty or some other neighborhood attribute always has the same effect on people’s outcomes, regardless of whether we are moving from 50 to 49 percent poor or from 16 to 15 percent) then mobility programs like MTO will change the geographic distribution of social problems, but not their overall rates in society.
MTO is great for studying the effects of MTO moves on the movers, but it is not well suited to learning anything about these larger societywide effects.
Even if the health benefits from MTO were sufficient to justify the program’s costs, there is still the question of what else we need to do in order to improve those outcome domains that were not affected in MTO, particularly schooling and labor market outcomes. A common reaction to MTO is to conclude that because MTO-like moves did not generate detectably large gains in schooling and labor market outcomes, then more intensive mobility interventions are needed. It is not obvious, however, that such mobility programs will necessarily have the effects on schooling and labor market outcomes that proponents hope for, or that such policies are even feasible at large scale.
20 Moving to Opportunity
Guest Editor’s IntroductionOne reason I am not sure that more intensive mobility programs will necessarily generate big schooling or labor market gains comes from previous quasi-experimental analyses that have tried to learn more about mechanisms. These results suggest that MTO participants who experience relatively larger changes in neighborhood poverty or related characteristics have larger improvements in physical or mental health outcomes (Ludwig et al., 2011). In the interim MTO data, however, Kling, Liebman, and Katz (2007) did not see the same “dose-response” relationship for schooling or labor market outcomes, which means that a larger neighborhood “dose” need not lead to larger changes in education or work outcomes. One qualification here is that there is one particular type of move—namely, to affluent, mostly White suburbs—are not very well represented in the MTO data. Although MTO itself does not have much to say about those sorts of moves, followup Gautreaux research using longitudinal administrative records has not found large beneficial effects from moving to the suburbs (DeLuca et al., 2010).
A different sort of question is whether mobility programs that achieve even more socioeconomic or racial integration than did MTO are feasible at large scale. The standard concern has to do with political feasibility, given some of the political opposition that arose to MTO itself (Goering, 2003).
I do not claim to have any special insight on this question of political feasibility, although it is perhaps worth noting that the few programs that I know of to have moved poor urban families to affluent suburbs (Gautreaux in Chicago, Thompson in Baltimore) were enacted by judges rather than elected politicians.
There is another important constraint on our ability to achieve even greater levels of economic integration than what we saw in MTO, which is the sheer amount of poverty itself that we have in the United States. A common measure of residential segregation is the “dissimilarity index,” which is defined as the share of people who would need to be moved across census tracts within a given area in order to have the share of poor people in each tract equal the share of the larger area that is poor. The five MTO demonstration cities have poverty rates right now in the ballpark of 20 percent.14 The average tract poverty rate of MTO experimental group movers (about 21 percent) corresponds basically to the benchmark of perfect poverty integration in these MTO cities. Even if we implemented a residential mobility program that would move inner-city families all over the country, the poverty rate in the United States as a whole right now is 15 percent.15 There is just not that much room to achieve more economic integration at large scale when the overall poverty rate is on the order of 15 to 20 percent.16 Another way to read the MTO demonstration is as a way to help inform community-level interventions (not just mobility programs), by trying to shed light on the specific neighborhood attributes that might matter most for people’s life outcomes. If we had all the money in the world, the first, Data from the Census Bureau’s American Community Survey for 2006 through 2010 show the poverty rates for the five MTO cities are 21.3 percent for Baltimore, 21.2 percent for Boston, 20.9 percent for Chicago, 19.5 percent for Los Angeles, and 19.1 percent for New York. See http://www.census.gov.
It is always possible to have some poor families live in tracts with poverty rates below 15 percent. Because 15 percent of the population is poor, however, that would require some other poor families to then live in tracts with poverty rates above 15 percent. The key point is that if 15 percent of all Americans are poor, it is simply not possible to have each and every poor family live in a tract in which less than 15 percent of all tract residents are poor.
best way to learn about community-level interventions is to carry out randomized experiments that test community-level interventions. Implementing most community-level programs in enough communities to provide adequate statistical power to detect effects quickly becomes cost prohibitive, however. A second best approach for learning about community-level interventions might be to study the effects of moving families into different types of communities, as in MTO and in the spirit of “mechanism experiments” suggested by Ludwig, Kling, and Mullainathan (2011).
Although one potential concern is that MTO might have less beneficial impacts on people’s lives than would community-level interventions, given the potentially disruptive effects of moving itself, this concern strikes me as less serious than it initially appears once we recognize the high rates of residential mobility that we see in general in the United States. Typically around 18 to 22 percent of Americans change addresses each year, about twice the rate we see in other developed countries like Japan or Britain (Long, 1992). Mobility rates are higher still among American renters, around
32.5 percent per year (Crowley, 2003). If we implemented a community-level program in a subset of neighborhoods, after a 10- to 15-year followup period, a large share of the original residents would have turned over. A large share of the people who currently lived in the new-and-improved neighborhood would have moved in from somewhere else; that is, the net effect of the community improvement effort would be to enable a subset of low-income people to move into a new, less disadvantaged neighborhood. Over the long term, therefore, MTO and a community-level intervention might wind up looking not all that different.
Given my discussion of the MTO results so far, it is probably not surprising that I think safety seems like a particularly important target for community-level interventions. The MTO families themselves reported on the baseline surveys that safety was far and away the most common reason they signed up to participate in MTO. The beneficial effects of MTO on neighborhood safety may be one of the key drivers for MTO’s impacts on mental health outcomes, and potentially on physical health outcomes like extreme obesity and diabetes. Improving safety would also have important direct effects on public health of low-income populations by reducing the toll of violence. Homicide is the leading cause of death to African Americans ages 15 to 24, by far. Homicides, because they are so heavily concentrated among young people, are responsible for nearly as many years of potential life lost before age 65 among African Americans as is the nation’s leading overall killer, heart disease. Devoting more attention to the crime problem that plagues our inner cities might be one of the most helpful things we could do for the low-income families living there.