Where Do We Go from Here?
Research presented in Chapter 3 not only highlights ways that spatial mismatch and neighborhood effects can significantly impact residents in disparate neighborhoods but also begins to suggest opportunities for future research and policy that might ameliorate some of these negative influences. Identifying promising directions for research, ways to improve datasets and methodologies focused on measuring the effects of place, and policy options to improve the conditions of inner-city neighborhoods were the final goals of the workshop, and these issues were explored in the final session. George Galster focused on methodological challenges to conducting research on neighborhood effects, and Tim Smeeding discussed policy options that could be appropriate responses based on the current status of research. This chapter summarizes their remarks as well as those of discussants Paul Jargowsky, John Goering, and Harold Wolman. In addition, comments made during the day that were particularly salient to the discussion of general research and policy issues are presented.
Discussion of the direction of future research on neighborhood effects covered two major areas: (1) the desirability of developing an overarching conceptual framework that could account for all the interactions among variables as contrasted with breaking up the question into at least two parts—the effect of neighborhoods on outcomes and an understanding of
how neighborhoods came to be the way they are and (2) the strengths of various methodological approaches for understanding the first question— the effect of neighborhoods on outcomes.
George Galster contended that the way to understand the challenges facing researchers investigating the role of place or neighborhood in shaping opportunities for children and youth is to articulate an overarching conceptual framework. Such a framework is based on the premise that, in order to suggest the most promising directions for future research, one needs to comprehend in a holistic fashion all the factors that contribute to the outcome in question and the causal interrelationship among those factors. Specifically, Galster suggested using a model in which housing tenure, neighborhood, mobility expectations, and housing wealth are determined simultaneously.
Five methodological issues presented challenges to researchers interested in measuring the mechanisms of neighborhood impacts on human outcomes. First, neighborhood characteristics and social processes have to be operationalized in clever ways. A better understanding of social processes, such as social capital and group norms, can yield conclusions about how these processes link to the variables that are available for geographic areas such as census tracts.
Second, some of the links between neighborhood characteristics and human outcomes may be nonlinear and indirect. As a result, the relationship between neighborhood and outcomes may be quite complicated. For example, the poverty level in a neighborhood may have to decline below a particular level in order to have a positive effect on children’s outcomes. Gordon Berlin and Sue Popkin both discussed this point of nonlinearity in the context of the Moving to Opportunity (MTO) experiment. Both noted that while it is impractical to adopt a national policy of moving all poor people into neighborhoods that are less than 10 percent poor, it is possible to stimulate opportunities for some families to move to neighborhoods with lower rates of poverty. Important research questions then become: At what threshold do people begin experiencing benefits? Would people who live in extremely poor neighborhoods (i.e., with poverty rates of 40 percent or more) be markedly better of in areas with poverty rates of 20 percent or even 30 percent? Is there some threshold of neighborhood poverty below which people experience benefits but above which residents would experience few, if any, changes from their current neighborhoods?
Galster’s third point on challenges for future research was that a complete understanding of neighborhood characteristics and outcomes has to
take account of omitted variables. In order to solve this challenge it is necessary to work into statistical models such factors as housing tenure status, mobility expectations, and housing wealth that have not been included consistently or extensively in other studies. For example, homeowners tend to take better care of their dwellings than renters, and better-quality housing environments might have positive impacts on children. Alternatively, because homeowners have a greater financial stake in their neighborhood, they are more likely to get involved in community efforts such as those discussed earlier. As a result, homeowners may be more likely to monitor the collective behaviors of young people in their neighborhood, which could result in a reduction in crimes by youth and more positive social environments for young people.
Selection bias represented yet another, if relatively well-known, methodological challenge. In this case the unobserved variables, such as parental characteristics, would affect not only neighborhood characteristics through the choice of neighborhood but also residents’ savings behavior, their choice of home ownership, and their mobility expectations and behavior. Galster suggested that databases with a much richer set of characteristics relating to parenting style and the whole panoply of attitudes and behaviors that factor into the outcome in question could help to solve this challenge. An alternative to generating new databases would be to develop better methods for measuring neighborhood choice and characteristics, as well as the other intervening variables in a model. He also suggested that researchers probe more deeply into the use of data from natural policy experiments that mimic random experiments such as the Denver Housing Authority’s Dispersed Housing Program.
The final research challenge was the problem of simultaneity. For example, if an individual has a predisposition to become a homeowner, that individual is likely to choose a neighborhood carefully because one would not want such an asset to depreciate in value. Conversely, if for information or economic reasons a potential homeowner has access to only a small set of neighborhoods in which to purchase a home, that individual may opt to continue renting. This is one kind of endogeneity between the key variable of interest, neighborhood characteristics, and a key intervening variable, tenure status.
Paul Jargowsky summarized Galster’s proposal about the future direction of neighborhood effects as follows: Galster argued that researchers need to understand the complex set of factors that produce the neighborhood characteristics that in turn produce the neighborhood effects; the best way
to understand the relationships is to use a statistical model. Jargowsky had two major comments on Galster’s ideas. First, he suggested that there are still more things that need to be included in Galster’s model, things such as the role of politics, policies, zoning, fragmentation, and public service provisions. He remarked that Galster’s work usually includes these dimensions and that they ought be part of any consideration of future research ideas. However, the limitation of making the problem so comprehensive is that it becomes entirely intractable. Jargowsky argued that understanding the effect of neighborhood on a given child or children requires only knowledge of the current characteristics of a neighborhood if all parental characteristics were controlled. How the neighborhood came to be the way it is represents a separate and interesting question, but linking these questions may make it impossible to solve either one. The workshop discussion did not resolve the conflict as to whether a complete specification of factors affecting neighborhoods is needed in order to understand their effects on particular outcomes.
Jargowsky and several other discussants addressed the issue of the advantages of various research methods for enhancing our understanding of the importance of neighborhoods and the mechanisms through which neighborhoods contribute to particular outcomes. Clearly, Galster made the case that instrumental variables set within a series of simultaneous equations was the preferred method for moving researchers’ understanding forward. Jargowsky suggested that setting up these statistical models, with appropriate variables, is at least as challenging as pursuing the role of neighborhoods through experimental models.
Gordon Berlin made a strong case that we can advance researchers’ understanding of neighborhood effects primarily with evaluations and social experiments. He discussed the merits of two specific experimental projects. The first social experiment, MTO, a long-term randomized experiment conducted by the U.S. Department of Housing and Urban Development, relocates certain families into neighborhoods with poverty rates that are dramatically lower than those from which the families are moving and in which the control groups remain. This methodology offers an important opportunity to understand whether there is some independent effect on outcomes related to the move.
The second experiment Berlin described was the Jobs Plus study in which pairs or triplets of housing projects were matched. In seven cities one of those three matched housing developments was saturated with “best practices” employment and training incentives that help to make work more
valuable than government subsidies. Results for both studies will be available in the next few years, and these data may help answer questions about the importance of neighborhoods and the mechanisms by which neighborhoods affect outcomes.
Berlin also discussed why some experimental studies, such as MTO and Jobs Plus, might show little effect of neighborhoods. Specifically, in a robust economy such as that which existed through much of the 1990s and with the tremendous amount of resources that have been expended for welfare reform, there may be few low-income adults or welfare recipients who are able to hold a job but do not now have one. In effect, these forces have accomplished most of what the experiments on neighborhoods were designed to do.
John Goering’s major comment on Galster’s presentation related to the importance of social experimentation in understanding neighborhood effects. Goering suggested that Galster’s comprehensive model was not really necessary to answer questions related to the effects on neighborhoods of people who were at the bottom of the housing market and who would be unlikely to become homeowners. Goering also took issue with the importance of natural experiments in understanding the effects of neighborhoods. He suggested that natural experiments may become too entangled with local politics and public opinion to be particularly helpful in understanding neighborhood effects. Despite this, it may prove useful to compare the outcomes that result from two different types of policies: demolition and relocation.
One aspect of the discussion of the research methodology that was raised by several workshop participants was the difference between theory and practice in randomized experiments. In particular, there were several comments on the fact that not all the families in the experimental group in the MTO study actually moved. This suggested that the families that did move may not be a random selection of all families in their neighborhoods. In other words, the “movers” may be different in some important respects from the control group.
Jeffrey Morenoff, in the course of his presentation on the importance of place for public health outcomes, also discussed the types of questions that are best answered with experimental data and those that are better answered with multivariate statistical controls. He talked about two different types of questions: Does place matter and why? He argued that the first question is best answered with experimental data. However, the second question—why place matters—is much harder to get at with experimental
data because it would require a very complex type of experiment to be able to control for a vast number of things that might be changing. Morenoff believes that this second question is better handled through multivariate statistical controls. Over the past five years or so there have been significant advances in research design and measurement that enable researchers to measure factors and mechanisms at the neighborhood level. Finally, Morenoff argued that when researchers are trying to measure neighborhood processes, it is important to have multiple people from the same neighborhood responding to the same survey.
Jargowsky suggested that research can be furthered by undertaking a large new data collection effort comprised of 500 neighborhoods and 100 young adults in each. Neighborhoods would differ by resources, such as quality of school districts, income status, and public services. By selecting neighborhoods with various combinations of resources, researchers would be able to understand which resources affected outcomes and by how much. He pointed out that in the MTO study everything changed for the treatment group—schools, neighborhoods, and public services. This makes it virtually impossible to isolate the contribution of each of the various influences.
Tim Smeeding also suggested new datasets that might enhance our understanding of the dynamics of cities. His suggestions differed from Jargowsky’s because Smeeding focused on datasets that already exist or that are in the planning stages. Smeeding stated that adding geographic coding to the Panel Study of Income Dynamics will provide a new source of city-level data. Also, the American Community Survey, which is currently being planned by the Census Bureau, will provide more current information on cities and where people live than does the decennial census. While these datasets may be important for policy makers in understanding what is happening in their cities, using such datasets in research will depend on the level of detail that is available. Many of the research questions raised in this discussion require data at the census tract level or even the neighborhood level.
Susan Popkin discussed the role that large tracking studies could play in identifying mechanisms through which neighborhoods affect outcomes. She also described the use of qualitative research to help with decisions about the quantitative research that will help people decide what issues to explore when attempting to identify the causes of various effects. Her point was that there is a range of complementary research going on that can contribute significantly to our general understanding.
Hal Wolman commented in support of Galster’s first point that there is a great need to understand and do a better job of defining social processes and neighborhood effects. Wolman also stated that, while participants spent much time during the workshop discussing how to obtain data that are difficult to obtain, some of what researchers must accomplish is to do a better job of conceptualizing the underlying processes or mechanisms that social experiments are trying to measure.
The discussion of future directions for policy focused primarily on the following points: many policies are linked to places, but the emphasis should be on policies for people; direction for policy makers given that it is not completely understood how neighborhood affects work; and knowledge of the problem is not the same as knowledge of how to create a solution.
Tim Smeeding began the discussion of policy options with the difficult linkage between policies related to place and those that focus on people. He pointed out that there are some policies that focus in a general way on specific places without adequate attention to the needs of the people in the area. In these cases, place policy has become political and is used to “spread the bacon.” But this sort of untargeted community development only benefits those most able to develop and benefit in a particular city. Smeeding acknowledged that this sort of policy may be necessary in the world of politics, but it is not the way to help the people who most need to be helped. The more important starting place is with people policy because it can be better targeted to individual economic needs and circumstances.
This was not to suggest that geographic and locational realities are unimportant: people live in particular places, and targeted populations are often reached through place-based policies. Gordon Berlin echoed this point, noting that sometimes a concentration of people with similar needs in the same geographic areas provides opportunities to deliver services that otherwise would not exist. Clearly, some policies are almost necessarily place based—public health policies and school policies—because these services need to be delivered from fixed locations. Smeeding concluded that, even though place and people policies are intertwined, the way to begin thinking about policy that can capitalize on place-based opportunities is to consider what people need, not what places need.
Smeeding then moved to a discussion of particular policy options he
believes can be recommended to policy makers even though we do not completely understand the ways in which neighborhoods affect the lives of their residents. Overall, the best policy to improve people’s well-being is a strong economy with strong demand for lower-skilled labor. The boom of the late 1990s decreased poverty rates and raised the employment rates and incomes of African Americans and Hispanics like no social policy or intervention program has been able to do.
Smeeding’s second policy point was that increasing mobility is a powerful way to allow individuals living in areas of concentrated poverty to deal with their economic and social needs. As Claudia Coulton noted, more than any other factor, access to a car most positively affects job access. It is possible to promote mobility through improvements in public transportation, but it still may not be possible to address the problems of mothers who work unusual hours or have to go in one direction to get to child care and then in another to get to work.
Rather than directing every city to rebuild its public transportation system, Smeeding suggested dealing with mobility issues through programs that give people access to cars. “Poor people,” he stated, “need what everybody else in America has: a car.” Smeeding acknowledged the difficulties of such an idea but was clear that he believed at least one part of a policy agenda was to enhance people’s mobility through whatever means possible. He argued that focusing policy attention on mobility was likely to be more productive than either trying to redo central cities to bring in all the needed employment and services or permanently moving people into other neighborhoods where employment and services already exist. Smeeding added a caveat to these policy ideas by pointing out that all cities are different and that actions having a substantial payoff in one place may create little benefit someplace else. Each city has its own barriers and its own economic structure. As a result, no single set of policies can be put in place to accomplish the same outcomes everywhere.
The third major policy topic Smeeding discussed was that “problem knowledge” is not the same as “solution knowledge.” For example, research suggests that collective efficacy is a key mechanism by which neighborhood effects have a positive influence on young people. One powerful way to shift neighborhood dynamics might be to stimulate collective efficacy in a neighborhood. However, the question that remains is how to accomplish this goal. What are the policy levers for promoting the mechanism or the factors identified as important to intervening and overcoming a given neighborhood issue? If research suggests that schools are an important
mechanism to improve neighborhood outcomes, what particular aspects of schools are important and should be targeted for change? Researchers have yet to determine if it is smaller classes, school choice, or better facilities that make some schools good places. As a result, even when researchers have learned more about the mechanisms that improve individual outcomes in neighborhoods, there will still be more to learn before recommending and putting specific policies into place.
Hal Wolman made a comment on defining place that provided a useful amplification of Smeeding’s point on the complexity of the relationship between people and place policy. Wolman highlighted why political jurisdictions may be more practical than neighborhoods as the locus and focus of what we mean by place. He contended that there are other dimensions of space and place, such as local jurisdictions, to which questions of place can be addressed more productively because it may be a lot easier to address policy for local jurisdictions, for example, than neighborhoods. If you discuss local jurisdictions, you can more easily assess an area’s fiscal capacity to tax and provide services, and fiscal capacity is something that can vary significantly among jurisdictions. He continued by saying that if one believes fiscal capacity is related to the quantity and quality of services provided, and if one further believes that the quality of services provided is related to important individual outcomes, then one would want to know what the fiscal capacity is and how it varies among local governments. Knowing about fiscal policy is important because there are ways to provide funds that can equalize spending among jurisdictions and therefore offset the variations in fiscal capacity across local governments. Wolman did remark that, while technically possible, this matter of equalization might be politically difficult.
Paul Jargowsky also commented on the equalization of expenditures among areas. He supported working toward providing public services such as safety and schooling more equally. Even without looking to mechanisms such as establishing metropolitan governments, he contended that there are ways to provide services more equally. Jargowsky also added another element to Smeeding’s list of policies that can be recommended to policy makers even before the research is completed—namely, that it is necessary to change the way urban environments are built. The current approach is to build an exclusionary development pattern with outer-ring suburbs taking the lion’s share of the resources. The residents of the outer ring can avoid responsibility for all the public problems that exist while still exploiting the benefits of being near a major cultural and economic center. This is
a long-term perspective because even if it were possible to immediately implement the type of policy that would solve this problem, the housing stock is still there and will be for another 30 years. It will take time to achieve some sort of consensus about things that can be done to slow the current expansion process and turn it around. Because of this, discussion about how to stop making so many bad neighborhoods should begin soon so that new policies can facilitate changes before we have to contend with the negative effects of new and larger disadvantaged neighborhoods.
Susan Popkin, in her comments on one of the earlier papers, raised some policy questions based on research results suggesting that neighborhood effects are positive but small. She was concerned about the advisability of investing in neighborhood interventions because they are very expensive and may not have more than a modest positive effect. Under the circumstances, it seems appropriate to understand who is most likely to benefit from interventions and in what circumstances. She also commented that even if the effects are small, they may be important and they may affect something we really care about—the well-being of young people.
One issue mentioned by several workshop speakers but not addressed in detail either in terms of research or policy implications was how racial discrimination interacts with other neighborhood mechanisms that affect the outcomes of individuals. At the beginning of this workshop summary, the increased concentration of low-income African Americans in central cities was considered, and Tim Smeeding discussed employment discrimination, housing discrimination, and mortgage discrimination, which are entangled with other problems of low-income neighborhoods. Workshop participants did not discuss how various policy proposals would address this interaction effect.
Another topic that received comment from several workshop participants was the wisdom of thinking about expanding MTO into national policy if it is judged to be successful in improving the outcomes of people who move out of high-poverty neighborhoods. In particular, questions such as the following must be addressed: What are the budgetary and social limits of moving people into new neighborhoods? How should this type of program fit with alternative policy initiatives such as increasing mobility for residents of high-poverty neighborhoods and enhancing low-income neighborhoods by increasing services?
The final discussion of the workshop raised a number of questions for future research and policy directions. In general, participants thought that further research on the importance of place could be productive and had the potential to address a number of important issues concerning building better neighborhoods, reducing spatial barriers and disadvantage, and creating places that might yield positive rather than detrimental effects on residents’ well-being.
The comments of workshop participants consistently converged around three points—each of which are highlighted in the findings of the Governance and Opportunity report that was the original impetus for the workshop and each of which tied back into the workshop theme of how place relates to the well-being and opportunity of individual residents. First, understanding neighborhood dynamics and effects more clearly is important. The implications of ignoring the effect of place may permit the continued growth of neighborhoods of concentrated poverty and disadvantage. In contrast, identifying the mechanisms that create barriers to opportunities in a given city has the potential to yield cost-effective, well-targeted, and successful social policy and intervention programming. The path to gaining this knowledge is not without challenge, but a number of advances in this type of research make this process considerably more viable. In general, workshop participants were encouraged by the new steps taken in current research on neighborhood effects and were eager to learn the final results of ongoing studies such as MTO and the Project on Human Development in Chicago Neighborhoods.
Second, incorporating contextual and governance issues, in addition to neighborhood characteristics, would lead to better problem definitions and improved research and policy approaches. Every neighborhood possesses a contextual set of factors (e.g., socioeconomic status, racial composition, employment opportunities, crime, access to public transportation) that will affect outcomes. In addition, neighborhoods are subject to exterior policies such as zoning, tax structures, and other governance issues that also will affect outcomes. Any attempt at conceptualizing neighborhood and place-based challenges must consider both the contextual and the governance issues. Perspectives that conceive of neighborhood effects as results produced only by the “culture” or “personality” of a neighborhood will be incomplete, as will approaches that focus only on governance or contextual factors. All of these factors together have an effect on residents and on the
social dynamics of a community, and research and policy must bear this in mind.
Finally, it is important to create linkages between neighborhoods in which certain effects are occurring and political jurisdictions where programs to address these effects are located. Furthermore, linkages are needed between people policy and approaches that are strongly tied to specific places. Place-based policy will not necessarily solve all problems and runs the risk of missing the people it intends to influence. However, certain types of policies carried out through specific places will be considerably less effective if a connection between space and residents is overlooked.