Neighborhood Effects Chapter (1995)

[Download PDF version here]

Penultimate Draft

February 13, 1995







Chapter 19:  Neighborhood Effects and Federal Policy


Jeffrey S. Lehman,

Law School

University of Michigan




 Timothy Smeeding,

Maxwell School

Syracuse University and CASBS














This paper was originally prepared for the SSRC Committee for Research on the Urban Underclass, Working Group on communities and Neighborhoods, Family Processes and Individual Development Conference  May 19-20, 1994, Baltimore, Maryland, and has been revised based on conference discussion and on comments received subsequently.  The support of the Russell Sage Foundation and the Center for Advanced Study in the Behavioral Sciences is gratefully acknowledged.  The authors would like to thank Esther Gray,  Karin D'Agostino, and Leslie Lindzey for secretarial assistance and Michael McLeod, Susan Mayer and Greg Duncan for helpful comments.  We, however, assume all responsibility for error of commission and omission.

In this chapter, we reflect on what the social science research included in this volume implies for federal policy. How should Congress react to this new learning about neighborhood effects on children? What direction should policymakers take from this new scholarship?  We approach these questions slowly and carefully.  For, in fact, they subsume some very difficult general questions about the relationship between academic scholarship and the domain of public policy, and about the relationship of federal policy to children in prospering as well as failing neighborhoods.  By making these general questions explicit in the first portion of this chapter, we hope to make it easier to grapple with the specific questions later on.

I.  Introduction

We presume that the overarching goal of federal policy is to deploy public resources toward the efficient and equitable promotion of public goals, particularly providing a fair opportunity for upward social and economic mobility for all Americans.

Stating our premise so baldly helps to signal some of the deeply contested normative issues that shape and reshape debates over federal policy, particularly as it affects local neighborhoods and their residents. What are appropriate public goals? What role do political "leaders" play in shaping those goals? How large a pool of public resources is available to promote those goals? Faced with scarce resources, how should priorities be established among worthy goals?  Which level of government should be responsible for reaching worthy goals?  And, if the federal government is to take the lead, how should state and local governments be involved in shaping and implementing these policies?

One strand of academic policy analysis holds that social science scholarship has little to contribute to the discussion of such normative issues. The issues seem too big, too philosophical, too personal, or too political to be susceptible to scientific evaluation. By this line of thinking, all social scientists can do is study and describe the world they see, and the way that different forms of governmental activity influence that world. The world cannot be subdivided so neatly. In almost any public debate, one quickly discovers that positions of  "principle" or "philosophy" are mixed together with empirical assumptions about how people behave or about how the world at large works, and some of those assumptions can be tested. 

Thus, some people who oppose a guaranteed minimum income may say that they believe, "able-bodied people should not be allowed to depend on state support if work is available." But when pressed for reasons, they may say, "otherwise, so many people will quit their jobs that we will face an intolerable labor shortage and a large tax burden to support nonworkers." While all individuals can weigh in on the first statement, social scientists in particular may well have something useful to say about that second proposition.

But the linkages between normative debates and empirical analysis go in the opposite direction as well. A social scientist's choice of what problem to study is influenced by a sense of what seems "important." That may reflect a personal judgment about what normative goals the public agenda ought to be directed toward, and what social facts might help one to sensibly develop such an agenda. Or it may reflect an assessment of what normative goals seem to be dominating the public debate at a particular moment, and of what social facts might be most salient in shaping that debate.

Even more importantly, the social scientist's methodology may reflect certain conventional normative assumptions about how people behave. The economist may be methodologically committed to an assumption that people are wealth maximizers. The anthropologist may be methodologically committed to an assumption that cultural norms are adaptive and functional, given a particular set of environmental constraints.  The developmental psychologist may be committed to the notion that parents can be trusted to do whatever is good for their children. An average citizen's acceptance of the social scientist's "findings" might therefore properly depend on her or his willingness to accept those methodological and normative commitments as well.

To ask what the research findings presented in this volume imply for the policy arena is therefore to ask how those findings, and their normative and methodological underpinnings, relate to the political environment in which they are being received. We will therefore begin by offering our own assessment of that environment. We will then consider the way that environment has already been shaped by the work of other social scientists. Next, we will consider the difficulty of the task we have addressed and the significance of neighborhood research at a general level, speculating about what kinds of findings might have emerged, and what their implications might have been. Finally, we will look at the particular findings that did emerge, and assess their implications for the future and the ways in which they might shape the policy debate. 

The reader should be forewarned that modesty is a virtue not to be ignored in this paper.  When research results are mixed, when existing policy has only small or uncertain effects on outcomes, when large scale efforts ("social engineering") have a very high absolute dollar cost and a high opportunity cost, and when the ultimate determinants of child well-being  (much less the effects of neighborhoods on well-being) remain elusive, modest responses may be called for.  But this is not to say that zero federal policy response is the appropriate policy direction either--going slow is not the same as doing nothing.

II.  The Current Political Environment

American social welfare policy has long been characterized by conflicts, constraints, and uncertainty. The dominant commitment to a government of limited scope made ours a late arrival on the scene of welfare states, and has left it relatively small in comparison with those of other industrialized nations. Fears for the work disincentive effects of public assistance have made the programs categorical, have seriously limited the generosity of cash programs, and have thus created some set of their own disincentives to escape welfare.  We seem to prefer social programs and benefits tied to work or to noncash benefits that reduce choices for recipients.And our federal structure of governmental authority has made implementation a central concern in the design of any national program.

And yet, it would seem that in today's political environment, the 1994 elections notwithstanding, some rough consensus has emerged concerning some of the appropriate goals of federal action. In particular, we would mention the following:

The  goal of investment in children. Central to America's self-understanding is the ideology that offers every child a "reasonably fair opportunity" (albeit not an "equal" opportunity) to prosper economically. Children are deemed to be morally blameless, and society is thought to have an obligation to mitigate the most extreme inequalities that follow from the accident of birth. A parallel ideology would hold that "children belong to everyone" (or, more parochially, investments in children redound to the whole society). 

At least since 1935, those ideologies have justified federal support for various different forms of assistance to deprived children. And the easier it has been for the federal government to ensure that a given program helps children without "leaking" over and helping their parents, the easier it has been to garner support for the program.  However, U.S. programs and policy thinking has still not progressed to the level of other modern nations where all children receive universal benefits (child allowances) due only to their citizenship, or whereby governments provide insurance for child support payments to the custodial parents of children living with only one natural parent.

The goals of reducing poverty, promoting individual dignity, and enhancing opportunity for social mobility.  These goals are more deeply contested, more heavily qualified, and  more weakly respected than the goal identified in the prior paragraph. Nonetheless, at least since the Great Society, it appears to us that Americans are disturbed by the presence of poverty amid affluence. Moreover, they believe that the federal government has a role to play in fighting poverty, at least as long as other public commitments are not sacrificed for the cause. The sentiment for federal intervention appears especially strong when individuals, particularly families with children, who are willing and able to work find themselves unable to earn enough to live above the poverty line.   However, appropriate policy action to meet these sentiments has been only slowly  forthcoming, even for the  working poor.  And while most Americans agree that "welfare dependency" is a major social problem, there appears to be little or no consensus as how to best attack the problem.  In fact, current policy thinking finds at least 50 ways to proceed -- one for each state.

The goal of minimizing crime and social disorder.  The maintenance of civil order is a defining characteristic of civil society. For the most part, America has left that critical task to the jurisdiction of state and local governments. If and when those governments prove inadequate to the task, however, local crime is easily transformed into a national issue. During the mid‑1990s, concern over urban violence was voiced frequently at the highest levels of government. That concern was highest when it intersected with the goal of investment in children: when crime seemed to touch the schools children attend, or the neighborhoods where they live (Herbinger 1994).  And, with the 1994 Crime bill, federal policy produced "three (convictions) strikes and you're out  (lifetime imprisonment)", 100,000 new police, and hundreds of new jails, and increased gun control laws.  Neighborhood safety and protection from violent crime are two overriding issues which have support from policymakers of all political stripes.

Historically, the federal government has pursued those goals primarily by designing programs that are defined by reference to the traits of individuals and families, as opposed to the traits of neighborhoods or communities. Eligibility for Medicaid, Aid to Families With Dependent Children (AFDC), Food Stamps, the Women Infants and Children (WIC) program, Head Start,  and School Lunches depends on the characteristics of the child and the family, not on  the characteristics of the place where the child lives. Such an approach to program definition is politically attractive.  It does not require Congress to specify ex ante which districts get more resources and which get fewer; inequalities in the distribution of resources flow, after the fact, from the application of seemingly neutral principles. And as long as a child's "need" depends more heavily on individual and family-level traits than on neighborhood-level traits, such programs are a logical and efficient response (Diamond 1994).  While the aggregation of large majorities of poor people in specific locations (underclass areas) may produce situations where economies of scale in service delivery and in program implementation argue for a strategy of community or neighborhood based service centers, these are still individual and family programs, not "neighborhood" programs per se.

Nonetheless, over the years the federal government has at times sought to pursue its social welfare goals through programs that use the neighborhood or the community as their unit of analysis. The first such program to be pursued on a grand scale was Urban Renewal, a program which targeted "blight" because of a belief that visible structural decay was an important contributing factor to the decline of community structures. Thereafter such diverse efforts as Community Action, Model Cities, and (most recently) Empowerment Zones and Enterprise Communities have attempted to bring federal resources to bear on social needs and problems that are defined by reference to specific geographic areas. At still other times, the federal government has sponsored programs that are designed to help specific geographic spaces in a different way. It has taken resources that were collected through national-level taxes and transmitted them to state and local units of government. The programs have varied in the amount of discretion they have left to the decentralized governmental entities concerning how the funds are to be spent: Community Development Block Grants have afforded great discretion to local political leaders, while Head Start funds have come with many more strings attached.

In all these programs, the same design issues have presented themselves. Should the details of program administration be dictated uniformly "from above," or should they vary from jurisdiction to jurisdiction as they are developed in cooperation with state and local governmental and private-sector representatives? Should the workers who implement the programs be employees of the federal government, of local government, or of private nonprofit or for-profit organizations?  How should the initiatives be interrelated to assume a coherent set of programs for youth and families with multiple problems?

Recent Actions

During its first year, the Clinton administration frequently signaled its interest in pursuing the connection between federal policy and the goals of social welfare. President Clinton and prominent members of his government spoke frequently on such matters as "ending welfare as we know it," "putting America to work," "investing in all Americans," "assisting the 'underclass,'" and "empowering distressed communities to join the economic mainstream." The first year's legislative and administrative agenda included the following items:  a substantial expansion of the Earned Income Tax Credit (to $25.0 billion per year in 1996);  increased funding for Head Start; the enactment of  limited direct federal assistance to businesses that invest in "empowerment zones" and "enterprise communities" that are locally nominated but federally chosen; the passage of an anti-crime package that included the so-called Brady Bill; and the  grant of waivers to states interested in experimenting  with radical modifications in federal-state welfare programs, particularly the AFDC program.

The second year's domestic agenda promised an emphasis on health care reform, crime and  perhaps welfare reform, but the only major success was the crime bill.  As Clinton enters his third year in office, welfare reform has risen to the top of the national agenda.  

From his first days in office, President Clinton has made a point of publicly including Professor William Julius Wilson in his inner circle of advisors about the federal role in social welfare policy. That last development, as much as anything else, captures the political significance of research by social scientists about how neighborhoods and neighborhood policies affect child development and the important of these findings to policy formation and implementation.   A brief review of Dr. Wilson's thesis is therefore useful to understanding some of the impetus behind federal policy.

III.  The Wilson Thesis and Federal Policy

The publication in 1987 of Wilson's book, The Truly Disadvantaged, was a significant event on many levels. It had a substantial impact on public debate about race and urban poverty;  it had a clear impact on the thinking of then Arkansas governor Bill Clinton; and it triggered  a resurgence of foundation support for social science research in those domains. Not the least significant of the book's effects was to stimulate renewed and heightened interest in the sociological and ethnographic investigation of neighborhoods. A critical link in Wilson's complex thesis about the emergence of an urban underclass during the 1970's was the proposition that declining neighborhoods and the disintegration of community social buffers led directly to a decline in what might generally be termed "youth outcomes" - the likelihood that a child will finish high school, become regularly employed, and avoid producing children outside marriage.

Because of its importance, it is worthwhile to summarize Wilson's thesis in some detail. Wilson claims that the emergence of an urban underclass resulted, in the first instance, from the development of socially isolated neighborhoods. Those neighborhoods, he claims, were produced through the interaction of several important macro-structural changes in American society after World War II, most notably: (a) the great migration of Southern blacks to segregated northern cities, (b) the age profile of the black community (tilted toward youth), (c) the restructuring of the American economy away from manufacturing and the relocation of remaining American manufacturing out of central cities, with the consequent drop in demand for unskilled urban labor, and (d) the opening-up of the suburbs to middle class blacks, and their consequent out‑migration from inner cities. The neighborhoods were socially isolated in the sense that their residents lacked contact with individuals and institutions that represent mainstream society. According to Wilson, that lack of contact left ghetto residents cut off from job networks and engendered a set of ghetto-specific norms and behaviors that made steady work even less likely.

What followed from this social isolation was a downward spiral of neighborhood dislocation. Employment plummeted, marriage and education lost their attractiveness, and crime rose. The cycle became self-perpetuating as families living in those neighborhoods had to cope with the set of experiences that Wilson called concentration effects. In his more recent work (Tanner lecture 1993), Wilson has argued that during the 1980s those concentration effects interacted with the arrival of crack cocaine, AIDS, and rising homelessness on the one hand, and the drop in countercyclical governmental interventions on the other. The cycle became even more vicious as flight from the cities accelerated and urban racial tensions heightened in an atmosphere of mutual recrimination.  At the same time, Wilson stresses that the vicious cycle is only partly self-perpetuating: the concentration effects are produced by dynamic processes, and those processes have situational bases that are susceptible to change through public and private action.

Wilson's emphasis on social isolation and concentration effects has renewed interest in the neighborhood as a unit of analysis, an interest that had flagged over the course of the 1970s and 1980s.  Wilson's argument resonated with the intuitions of a great many people in both the scholarly and political communities, not to mention the average person thinking about buying a house.  It suggested that some children were growing up to experience substantially less satisfying lives than they would have experienced if their families had lived in a different neighborhood, even if all other elements in their lives, such as family structure, were the same.  Wilson's theory was not about small marginal differences in well-being; it was about the difference between a life of productivity in the economic mainstream and a life of marginalized despair.

And yet, Wilson's thesis also has an optimistic side.  It suggests to us that these neighborhood effects could be mitigated through the right kinds of policy intervention.  To be sure, it was not at all obvious what the "right kinds" of policy intervention might be.   Even if one knows that neighborhood effects both exist and are significant, it is not at all obvious how one should go about designing and implementing effective policy responses. For instance, it is not clear whether policies should be aimed at improving neighborhoods or aimed at helping people move out of poor neighborhoods.  Not all of the neighborhoods dealt with in this book are as bad as the Chicago neighborhoods where Wilson based his studies.  Nonetheless, knowledge about neighborhood effects can help to shape and direct the development of government programs at all levels.


IV.  Our Model of Child Well-Being and Issues it Raises for Urban Policy

Optimists and believers in neighborhood based theories should also be exposed to a dose of modesty.  The effects and channels by which public policy directed toward neighborhoods or toward families affect children's well-being are complex and not always obvious (Smeeding 1995).   In fact, Chapter 4 of this volume lays out a skeletal micro-model of child well-being which helps organize thinking about the way in which neighborhoods affect and are affected by changes in child well-being and other factors.  A brief review of  the multidisciplinary nature of this model will help us understand the complicated nature of child well-being and the forces - neighborhood and other - which might bring about positive child outcomes.

Until recently, many social scientists (economists and sociologists, especially) and most public policy analysts have measured children’s well-being by the well-being of their parents.  They use measurable socio-economic variables which are really inputs into children's well-being:  variables such as household consumption, income, wealth, household capital goods, and neighborhood characteristics.  Social standing or the lack thereof is also measured largely by parent's characteristics until children reach the age of majority, labor force participation, or criminal institutionalization, whichever comes first.  Beyond birthweight and Apgar scores, children have not traditionally been individual social, economic, and statistical entities as far as state record keeping and social scientist household survey practice was concerned.  Most large scale ("macro oriented") household surveys have included little more than their age and sex.  Most of the databases which are exceptions to this rule were used earlier in this volume to help ferret out neighborhood effects on families and children.

Developmental psychologists, educators, anthropologists, and pediatricians approach children's well-being from the other or "micro" end of the spectrum.  That is, they employ direct measures of some aspect of children's well-being: cognitive, social, intellectual, educational, or other developmental outcome measures for psychologists, educators, and anthropologists;  and physical and mental health status for pediatricians.  The measures are usually for a small subset of children drawn from a local sample or institution.  Even in cases where children studied vary widely by background and/or health status (as  do many of the children studied here) the economic, social, environmental, and other contextual measures on which the macro disciplines focus are often poorly measured or ignored. 

It is to their great credit that the scholars represented in this volume are wedded to the idea of marrying these two perspectives to produce a more holistic approach to measuring the impact of various forces, and ultimately in analyzing the impact of public policy, on children.  If we are to truly measure the impact of adult focused programs like welfare reform on children's well-being, or if we ultimately want to find out why some children in poor neighborhoods  succeed while others do not, these two perspectives must be brought together, if not married.  Important public policy questions such as the relative effects on children of family-based subsidies and neighborhood-based initiatives can only be addressed in this manner.

The model developed in chapter 4 is intended to suggest the types of conditions, processes and life events which shape children's lives and well-being.  It is an attempt to represent the "big picture",” and it highlights the role of neighborhoods and communities in affecting child outcomes.  The model is sketched in Figures 1A, 1B and 2 of Chapter 4.  It suggests that various exogenous (to the neighborhood) forces such as economic restructuring, migration, and existing public policies create an environment which presents the constraints and/or opportunities facing children and their families at a point in time.  Public policy may affect children's well-being by affecting any of  the variables that, directly or indirectly affect the child:  family, peers, other adults, or such neighborhood features as stability, ethnic and racial heterogeneity, social organizations, and cultural values.

These clusters of static variables set the stage for the dynamic processes involving the family and the community that also contributed to child development.  Together, they shape various measurable child outcomes including health status and educational attainment.

How should neighborhood effects and policy responses be viewed through the lens of this model?  Immediately, we realize that it is not enough to know, as a general proposition, that neighborhoods matter.  It matters how neighborhoods matter.  That is, one  needs to know the mediators:  social organizations and networks; cultural processes; parental, peer and other processes, which link to neighborhoods and to children's developmental courses.  Consider one example.  The Omnibus Budget Reconciliation Act of 1993 and the Crime Bill of 1994 called for the identification of various "empowerment zones" demonstrating various characteristics (e.g., community plans) to be designated by the Secretary of Housing and Urban Development after having been nominated by the relevant state and local governments.  To be contenders, zones and communities must satisfy specific criteria with regard to size and must have a condition of pervasive poverty, unemployment, and general economic distress.  In addition, the relevant "state and local governments must have committed themselves to a strategic plan" that would include direct public investments in the nominated area and provide indirect support for private for-profit and nonprofit institutions to do likewise.  Once chosen, the communities received various awards (from $20 to $100 million) to implement their plans.

But what, precisely, are the kinds of "investments" that institutions should be making in the empowerment zones? What kinds of activities should be subsidized? It depends on why we are concerned about with these neighborhoods.  Our concern here is the effect of bad neighborhoods on children.  To be sure, there are a host of reasons why a government might be concerned about depressed neighborhoods that have nothing to do with concentration effects on child development. But let us assume (accurately, in this case) that at least part of the motivation for the creation of empowerment zones is a desire to respond to the implications of Wilson's thesis. Then, we would contend, it matters how children are affected by growing up in neighborhoods that suffer from concentrated high unemployment, crime, and social decay.

Suppose, for example, that high levels of neighborhood unemployment are detrimental to children because those children are deprived of a particular kind of socializing experience: that of interacting socially with adult neighbors whose lives are structured around the experience of steady employment (and who are likely to reflect the primacy of employment in their lives through their conversations). If that is the major mediating mechanism, then the logical aim of policy intervention should be to subsidize the employment of adults who reside in the neighborhood, regardless of where the job is located.  But now suppose, conversely, that high levels of neighborhood unemployment are detrimental to children because those children are deprived of a different kind of socializing experience: that of observing a critical mass of adult neighbors at work in steady jobs during the day.  In that case, the logical aim of policy intervention would be to subsidize the employment of adult neighbors in neighborhood jobs that are visible to children.  One policy stresses job creation regardless of place, the other stresses job creation within a specific neighborhood.  Both polices affect family well-being and presumably then, parental/family processes.  But only the second affects neighborhoods as well and hence neighborhood and community characteristics and environments.  Moreover, the exact ways by which each of these policies filter down to children - directly or indirectly (via the mediators of social organizations, cultural processes or family efforts) is not well known.

  In fact, Congress chose subsidies that were closer to the latter description than to the former. To oversimplify somewhat, in the 1993 Employment Zone law, employers were given wage subsidies for wages paid to zone residents if they work in the zone (although admittedly even if they work in places that are not accessible to children), and additional tax incentives were granted to businesses that carry out substantially all their operations within the zone as long as at least 35 percent of their employees are zone residents. In the 1994 Crime Bill, there was considerably greater flexibility in the use of federal funds, but they were still targeted on specific areas and on their plan of action.  Whether these were the right choices is not important for present purposes.  Similarly, we are not concerned here with whether Congress was right to emphasize employment policy; we would make the same point if the centerpiece had been anti-crime policy or school assistance policy.  The point is that, in deciding which type of policy to pursue, it matters how children are affected by their environments.

V.  The Research Findings and Federal Policy:  What Might Have Been Found and What These Findings Might Have Implied

Since the publication of The Truly Disadvantaged, social scientists have been systematically testing various elements of Wilson's thesis.  Variations have been proposed, stressing factors other than those Wilson stressed, or placing different emphasis on different aspects of the overall story.  And one particularly significant piece of Wilson's analysis that has received special attention is his suggestion that some neighborhoods have significant, independent, detrimental effects on the life courses of the children who grow up in them.  That particular aspect is one that gave rise to this book, and one which produced neighborhoods and communities as key elements in our structural model (Figures 1A, 1B, and 2 in Chapter 4).  To give an appropriate sense of the policy significance of the research that is reported in this book, it is useful to consider what is sometimes referred to as (Sherlock) Holmes's "dog that didn't bark." It is useful to think more generally about what the researchers might have found, and what that might have implied for public policy. We will consider three "silent dogs", three possibilities that did not come about from the findings in this book:  (a) a finding of absolutely no neighborhood effects, (b) a finding that children who grow up in "bad" neighborhoods end up doing "better" than children who grow up in "average" neighborhoods (a "survivor's effect"), and (c) a finding that children who grow up in "bad" neighborhoods end up doing "worse" than children who grow up in "average" neighborhoods (the anticipated findings)

. A.  Silent Dog #1: No discernible neighborhood effects at all.

. The research in this volume could have discerned no distinct neighborhood effects whatsoever. The so-called "Core A" analyses in Chapters 6, 7, 9, 11, and 13 [Note:  Check “Core A” chapter numbers] could have concluded that, once one included a full range of individual- and family-level variables, measurable characteristics of the neighborhood added nothing to our ability to explain variation in child outcomes. What would that have implied?

First off, we should note that such a finding would have been surprising. Casual empiricism would suggest that most people believe quite strongly that neighborhoods matter, and in recent years, researchers have tended to find such effects, although the policy significance of their magnitude has been a subject of much debate. (See Chapter 3?) And among the many methodological hurdles that confront researchers in this field, at least one has been noted to generate a potential bias in favor of finding such effects: the fact that most if not all  individuals have some control over which neighborhood they live in means it is easy to attribute certain effects to neighborhoods that in fact reflect unmeasured family characteristics. (See Tienda 1989; and Burton, Price-Spratkin, Spenser Chapter). Of course, studies that overcome any such bias and still find no neighborhood effects could themselves have been methodologically contaminated. 

The research methodology might have been unable to discern neighborhood influences that were really there. The research might have been trapped with Census data that required "neighborhoods" to be defined differently from the sociological neighborhoods that have real-world effects (Spencer, Coles Burton chapter). The thirty-five census tract variables available to the researchers might not have captured some characteristics of "neighborhoods" (such as levels of gang activity, or school characteristics) that are highly relevant to child development. (See generally Burton, Price-Spratkin and Spencer chapter). Traits that were deemed "family" traits might have been shaped by neighborhood characteristics before they could be measured. The samples might have been too small. Nonlinear effects might have been invisible to linear models. Interneighborhood social linkages might be ignored (Jarrett chapter).  And the researchers might have failed to test for an interaction that was in fact present.

But if no neighborhood effects had been discerned, and if a policymaker had been comfortable enough with the methodology to place weight on that finding, there would have been clear policy implications. In a world of limited resources, one would be more inclined to "go slow" on neighborhood-level interventions wherever one could as easily (as efficiently, and with no greater collateral costs) implement macroeconomic, family-level, or individual-level interventions.  Thus:

Ÿ One would be inclined to continue to emphasize programs that respond to the needs of individual children for education, health care, nutrition, and general income support, where "needs" are measured without regard to the child's place of residence.

Ÿ One would be inclined to continue to emphasize programs that direct adult education, training, and wage subsidies to individual adults in need, without regard to their neighbors' employment situations. 

Ÿ And one would be inclined to continue to emphasize the importance of macroeconomic policies that influence the resources available to children's family environments and shape the longer-term opportunities those children will enjoy when they are older.

B.  Silent Dog #2:  "Bad" neighborhoods lead consistently to "better" outcomes than "average" neighborhoods.

  Jencks and Mayer (1990) identify several different varieties of neighborhood effects theory. Two of those varieties are "competition" theories (in which children compete with their peers for access to scarce developmentally significant resources) and "relative deprivation" theories (in which a child's developmentally significant self-concept is shaped by comparison with peers).  Both those varieties of neighborhood effects theory predict that, even after one accounts for a full range of individual- and family-level characteristics, the presence of certain "bad" neighborhood attributes will be associated with positive child outcomes, as children thrive from being the "biggest fish" in their local ponds.  Consistently with such theories, the so-called "Core A" analyses in Chapters 6, 7, 9, 11, and 13 [Note:  check chapter numbers] could have found that children who live in "low SES" neighborhoods do better than comparable children who live in "middle SES" or "high SES" neighborhoods. What would that have implied?  Once again, such findings would have been quite noteworthy. They would have been inconsistent with the spirit of evaluations of the Gautreaux housing voucher program  (Rosenbaum 1991); and they would accordingly have complicated straightforward arguments for "deconcentrating the inner city poor."  (Schill 1992).


C.  Silent Dog #3: "Bad" neighborhoods lead consistently to "worse" child outcomes than "average" neighborhoods.

  Two other varieties of neighborhood effects theory that were identified by Jencks and Mayer (1990) are "contagion" theories (in which child development is shaped by peer behavior) and "collective socialization" theories (in which child development is shaped by adult role modeling and monitoring).  The Jencks/Mayer labels connote subsets of more general categories: theories that stress peer influence (whether or not one would want to brand it with the pejorative "contagion") and theories that stress community resources (including but not limited to role modeling and mentoring). Both peer-influence and community-resource theories predict that, even after one accounts for a full range of individual- and family-level characteristics, the presence of certain "bad" neighborhood attributes will be associated with negative child outcomes.[1]  Consistently with such theories, the so-called "Core A" analyses in Chapters 6, 7, 9, 11, and 13 [Note:  Check chapter numbers] could have found that children who live in "low SES" neighborhoods do worse than comparable children who live in "middle SES" or "high SES" neighborhoods. What would that have implied?

In thinking about government policy more generally, such findings would have implied only that one would have needed to read the "Core C" analyses with great care.  Peer-influence theories have very different policy implications from community-resource theories. Or, to return to the more general theme stressed in Section IV of this chapter, the mediators would matter.

Consider, for example, the possibility that the Core C analyses had found that children in "low SES" neighborhoods did worse because, as Wilson theorized, the individuals living in that neighborhood lacked the wherewithal to produce an adequate stock of critical community resources ("social buffers," "quality schools," "job networks," or "physical security"). Responsive policies might then involve systematic efforts (a) to identify which such resources were critical to child development, and then (b) to make the production of such resources less dependent on the individuals who live in a particular neighborhood. Such efforts might involve simple funds transfers from one governing body to another, in the form of revenue sharing or block grants. Or, more grandly, they might involve more aggressive transformation of metropolitan governance structures along the lines advocated by Rusk (1993).

Or consider the possibility that the Core C analyses had found that children in "low SES" neighborhoods did worse because their peers spoke a different language from the Standard American English that is most efficacious in the job market. That is the "language of segregation" theory advanced by Douglas Massey and Nancy Denton (1993). Responsive policies might take any of several directions. They might attempt to restructure the likely peer environments that children would find themselves in, through heightened attacks on residential segregation, or, as a fall-back strategy, through heightened efforts to promote school integration even in a society where residential segregation persists. Or, more speculatively, they might explore ways to generalize the strategies that have allowed immigrant non-speakers of Standard American English to penetrate the American job market.

Or consider a more complex mediation possibility.  Suppose that the Core C analyses had found that children in "low SES" neighborhoods did worse because of the interaction of three forces:

Ÿ High degree of perceived rewards for investing time in socially unproductive or even antisocial activities.  The perception of such rewards could be acquired from peer influence or from weighing the costs and benefits of criminal activity as opposed to legal costs (Freeman 1991; Anderson 1994)..

Ÿ Low degree of perceived rewards for investing time in skills acquisition.  This absence could be because the rewards are in fact not there, because job opportunities are constrained.  Or it could be because informational resources (either in the form of peer influence or of community resources such as role models or counselors) are not there to show children that the rewards are in fact there.

Ÿ Low degree of effective assistance with skills acquisition, e.g., due to poor quality schools.

Once again, responsive policies might take any of several different courses.  Community-level efforts to raise the perceived cost of antisocial activities might involve counseling, the criminal justice system, or public support for private authoritarian institutions. Strategies for giving children more heterogeneous opportunities for peer interaction were discussed above. Strategies for increasing the availability of returns to investment in skills acquisition might depend on why such investment had been found to be currently ineffective; Lehman (1994) provides a general review of such strategies.  Strategies for increasing the availability of effective assistance with skills acquisition might involve school finance reform or support for extracurricular education-and-training programs.  They might also involve programs to produce safer schools.

To be sure, even if the research in this volume had provided powerful support for one or another peer-influence or community-resource theory of neighborhood effects, and even if one were to conclude on the basis of such research that a policy intervention along one of the foregoing lines was appropriate, it would still not be clear what role the federal government should be playing and how states and localities might implement these interventions. That is an important overarching question that we will return to in Part VII below.

In summary, the silent dog analogies and the model review in the previous section point to the sequential nature of our enterprise: (a) what is the nature and type of neighborhood effect? (b) how large and how consistent and how linear are these effects? (c) and even if we know that these effects are worth the attention of policy makers (i.e., an identified problem), do we know enough about the mediators through which they operate to be able to design effective policies to offset or neutralize the effects?  Knowledge of the problem does not always point to knowledge of  an effective solution.

VI.  The Research Findings and Federal Policy - What Was Found

Having carefully and deliberately set the context for interpreting the findings of this body of research, we turn from dogs that didn't bark to those  which at least made some noise.  We begin by noting that it is somewhat risky to distill the nuanced and still warm findings of many different researchers into a thematic synopsis.  But that must be done if those findings are to be accessible and useful to policymakers.  And several themes do appear to have emerged clearly within the research findings of this volume.  The following eight themes strike us as being of the greatest potential policy interest:

A.  The researchers found measurable neighborhood effects.  Consistently, across data sets, and across ages, the researchers found that models which included neighborhood variables were able to explain more of the variance in child outcomes than models which did not include neighborhood variables.  Even though which neighborhood variable or variables proved significant varied from study to study, there was evidence that the neighborhood variables which mattered in the Core A analyses matched up well with the neighborhood variables derived from the "windshield method" (Spenser, Cole and Burton chapter).  This fits with our (and others’) general reactions to the Wilson hypothesis, with investigations of its veracity in other locations, and with common sense:  different neighborhoods have different problems and hence different types of effects on the children which live in them.

B.  The neighborhood effects that were found were in some respects more consistent with community-resource theories and in some respects more consistent with adult-level contagion theories (peer contagion effects for children per se could only be inferred from these data).  The presence of high-SES neighbors seemed to be associated with good child outcomes, for at least some children, in three of the five age groups (Chapters 6, 7, and 13).  The presence of low-SES neighbors did not seem to be associated with bad child outcomes for any of the age groups and ironically seemed to be associated with better behavior by early adolescent African American girls in the Atlanta sample (Chapter 11, p. 21).  The presence of high rates of neighborhood male joblessness did seem to be associated with bad child outcomes, for at least some children, in  two age groups (Chapter 9, p. 13, African American males Chapter 11, p. 15, Rochester African American males, and p. 21, Atlanta girls) but was, conversely, associated with good child outcomes, for at least some children, in three age groups (Chapter 6, p. 15, girls; Chapter 11, p. 16, Atlanta boys; .Chapter 13, p. 10, PSID African American girls).

The variety of neighborhood effects sugests that it would be a mistake to focus exclusively on male joblessness as the key element in the causal process of neighborhood decline.  While male joblessness, particularly as it manifests itself in the lives of African American youth, is a key problem to be solved (e.g., see Bluestone, Stevenson and Tilley 1994), it is far from the only problem, and it may not even be the primary problem in black neighborhoods (see Duncan, et al., chapters xx, xx).

C.  Some neighborhood effects were found at all ages, even for three-year-old children.

Even after controlling for family-level variables, the researchers found that the presence of high-SES neighbors contributed to better outcomes in the cognitive domain for the three-year-olds they studied.  (Chapter 6, p. 13).  Somewhat surprisingly, living in a worse neighborhood (measured with a composite variable) seemed to be associated with diminished behavioral problems in that age group when family variables were controlled for.  (former Chapter 6, pp. 15, 17).

D.  The neighborhood effects that were found varied by race and gender.   In the three chapters where high-SES neighbors were associated with positive outcomes, the association was stronger for whites than for blacks; sometimes it was stronger for boys, and sometimes for girls.  [Chapter 6, pp. 13, 19; Chapter 7, pp. 4, 5; Chapter 13, p. 9; chapters and pages need recalibration].  In all the chapters, race and gender differences were common.  For now, we only note that this finding may significantly complicate policy design.

E.  The so-called "Core C" analyses of mediators were often indeterminate, but yielded interesting alternative hypotheses to account for complex interactions among variables, such as race and gender.  Several different chapters suggested interactions among a child's race, the child's neighborhood quality, and the child's perceptions of home and school support.  These interactions were sometimes consistent with radically different interpretations.  For example, high-SES neighborhoods were not associated with the same kinds of positive outcomes for black males as for white males.  "Core C" analysis suggested that black males from high-SES neighborhoods were less likely to find schools supportive than black males from low-SES neighborhoods.  Thus, other benefits from living in a high-SES neighborhood may have been offset (or "suppressed" in the vernacular) by harms in the school context.  The precise reason for the school-context effect remains unclear (See Chapter 7, pp. 20‑29).

F.  The neighborhood effects that were found were generally much weaker than the effects of family-level and individual-level factors.  While some of the neighborhood variables were significant predictors, they were always secondary predictors.  In no case did they explain as much of the variance in child outcomes as the family-level and individual-level variables did.

G.  The data did not always permit the researchers to distinguish linear, additive neighborhood effects from nonlinear, "epidemic-style" neighborhood effects.  A finding that the fraction of high-SES neighbors in a neighborhood is positively associated (in a way identifiable through linear regression analysis) with good child outcomes is consistent with either of two possibilities.  The relationship might be linear, so that moving a high-SES neighbor out of one neighborhood into another would harm the outcomes in the original neighborhood in a way that roughly corresponds to the benefits it would provide for the outcomes in the new neighborhood.  But it might also be nonlinear, so that moving a high-SES neighbor from a well-off neighborhood to a neighborhood with very few high-SES residents would provide benefits to the new neighborhood without harming the outcomes in the original neighborhood.  Chapter 15 (Duncan et al.) was the only chapter to begin to get leverage on this important question.

H.  The researchers' own methodological reservations counsel policymakers to exercise caution in relying too heavily on their findings.  Chapters 15 and 16 include a long list of potential sources of bias that permeate the quantitative analyses in this volume; and the qualitative commentary chapters (Janett; Burton, Price-Spratlan and Spenser) indicate the many meanings and nuances of "neighborhood" which quantitative variables can only hope to measure. One of the implications of that list is that, among the many ways that social science research differs from a Sherlock Holmes mystery, a particularly important difference concerns the inferences one should draw from silent dogs.  Some of those methodological limitations imply that, even if the dogs did not bark for these researchers, a different research methodology or measure of neighborhood might have coaxed them into action. The research here has thus not shown that any of the policy responses discussed in this section are necessarily wrong-headed.  The arguments in favor of those approaches were not refuted; rather, they were only held in abeyance.  For that reason, the potential conclusions that the present research did not produce should be referred in our dog metaphor to as "abeyance hounds."

VII.  The Research Findings and Federal Policy - What the Findings Imply

The research in this volume, which has taken advantage of the most sophisticated interdisciplinary theories available,  which has embraced multidisciplinary teams of well-qualified researchers,and which has exploited some rich and well constructed data sets, ultimately has very little definitive to say to federal policymakers concerning neighborhood effects on children.  For a first attempt at such a multi-phased and difficult collaboration, little else should have been expected.  The most extreme potential findings - that could have in theory supported a simpler vision of the world - did not materialize.  Instead, we have received additional support for the proposition that the world is complex.  And so, the most significant debates about the future course of policy on behalf of children growing up in distressed urban neighborhoods should continue unabated.

To us, that implies at least three things:

ŸIf the federal government is already undertaking macroeconomic, family-level, and individual-level programs that help low-income children who live in distressed urban neighborhoods, this research does not give any reason to stop. 

This research joins the existing literature suggesting that family income matters most to child development.  If the federal government can use macroeconomic policy to promote employment opportunities for low-income parents and to supplement earned income, such a policy should continue to be understood as pro-child-development.  If the federal government can use redistributive transfers to boost family incomes, such transfers should continue to be understood as pro-child-development as well.  If anything, this research supports continued efforts to develop new forms of family-level and individual-level intervention that are likely to help low-income children who live in distressed urban neighborhoods.  Some form of universal health insurance guaranteeing access to basic preventive and acute health care for all children, a federal policy initiative not enacted in 1994, is surely an obvious place to begin.  But it is certainly not the place to stop, as we suggest below.

Moreover, the belief that nothing can be done, or worse, that federal policy hurts poor families more than it helps them  is clearly not supported by the research completed in this volume.

ŸIf the federal government is already undertaking neighborhood- (or community-) oriented programs that help low-income children who live in distressed urban neighborhoods, this research does not give any reason to stop.

The evaluation literature suggests that Head Start and the WIC program are effective federal policies (NAS, 1993, National Commission on Children, 1991).  One might point to Chapter 6's finding that neighborhood effects show up even before school begins, to support expanded support for Head Start, but that seems unnecessary.  For Head Start attempts to influence individual-level characteristics through an intervention that is at once neighborhood-based and family-directed.   Robert Sampson has aptly observed that it is a common fallacy to believe that, when one uses a "community" as one's unit of analysis, all one's observations result from "community processes" rather than from the cumulative effects of the actions of the "individuals" who happen to compose the community at that moment  (Sampson 1993).  As is often the case, such an observation about the empirical analysis of social phenomena can be translated into an analogous admonition about the analysis of the effects of social programs.  Here, we would note that it is also a fallacy to assume that a "community-based program" makes sense only if community processes are significant and the program influences those processes.  A community-based program may simply be the most effective way to reach the individuals who live there.  Anti-crime programs and neighborhood safety programs might be couched in the same terms.

ŸWe still have breathtakingly little knowledge about the effects of public interventions on child development, and about the way neighborhood effects on children are mediated.  Accordingly, it would seem premature (or at least a significant leap of faith) to press for some bold, new, nationally mandated, universal effort to address neighborhood effects.  Conversely, there would seem to be great potential value in federal support for  carefully structured programmatic experimentation, coupled with rigorous evaluation, in order to augment that knowledge.

In Chapter 20 of this volume, the authors discuss some of the existing state and local initiatives that are already a potential source of programmatic experimentation.  How should federal efforts dovetail with those initiatives?  We would offer the following tentative suggestions, both substantive and procedural.

Substantive Suggestions: The Loci of Additional Inquiry

The research findings in this volume do not dictate a determinate list of fields for policy experimentation and evaluation.  Nonetheless, we would be surprised if any reader of the quantitative and qualitative analyses did not come away with an appetite whetted for a better understanding of the dynamics of neighborhood effects.  The following list reflects our own personal sense of domains for fruitful experimentation in  public interventions:

1.  Jobs and their Correlates .  In the research in this volume, "high socioeconomic status" was defined in terms of an annual income of $30,000, a level that almost invariably requires one and perhaps even two family member to have a  good job, and at least one job which has good benefits, including health insurance.   Jobs along with "male joblessness" were the variables that most frequently generated independent effects on the development of children in the neighborhood.  What kinds of policy interventions might increase employment levels?  A number of concrete proposals have been on the table for several years, and many have been implemented in various ways.  Various forms of government-run welfare-to-work transitional programs and education-and-training programs have received substantial trial and evaluation over the years.  (Smith 1993; Heckman 1993; Gueron and Pauly 1991).  We would expect the marginal returns from further trial and evaluation in those domains to be diminishing rapidly.  Other proposals, some directly and others indirectly linked to work promotion, have received less experimental testing.  The following strike us as the most promising candidates:

Ÿ Guaranteed public sector employment.  The literature on past efforts is dated and inconclusive.  (Lehman 1994).  Moreover, there are strong theoretical arguments for the proposition that the current dynamics of racial distrust require more radical interventions than before.  The New Hope demonstration project in Milwaukee is evaluating such an experiment.  (Hollister 1993; Aikman 1993).  More such experiments, if targeted to areas of high joblessness and  few job openings; would provide opportunities for careful and up-to-date evaluation.

Ÿ Transportation.  The literature on the so-called "spatial mismatch" is conflicting and inconclusive.  (See Burtless and Mishel 1993). Nonetheless, there is almost irresistible intuitive appeal to the proposition that job opportunities expand greatly if one has ready access to an effective means of transportation.  Senator Bill Bradley has introduced legislation entitled the "Mobility for Work Act" that would expand support for public transit systems in twenty metropolitan areas.  But as attractive a service as public transportation is likely to be on many levels, its impact on the lives of low-income children who live in distressed urban neighborhoods is likely to be slow and unmeasurable.  A better targeted, less expensive, and more easily evaluated alternative experiment would be to provide target groups with private transportation vouchers or "car stamps":  in-kind subsidies for welfare-to-work programs enrollees, for the rental or purchase of transportation services, including cars  (Smeeding 1994).

Ÿ Child Care.  As was true for transportation, there is almost irresistible intuitive appeal to the proposition that job opportunities expand greatly for parents who have ready access to a reliable source of quality day care. The Expanded Child Care Options Project, being run by the Department of Health and Human Services, is intended to provide a comprehensive and systematic long-range test of the impact of different forms of child care assurance on adults and children in low-income families.  (Hollister 1993).  The project would seem to deserve full federal support over a very long term.

Ÿ Non-public employment and training initiatives and placement.  In recent years, private for-profit entrepreneurs such as America Works and Maximus have attracted attention for their efforts to help welfare recipients obtain and keep positions in the formal economy.  (Aikman 1993).  It is at least theoretically conceivable that the interpersonal dynamics of dealing with a for-profit service, coupled with a well-structured profit motive, could make such services more effective than their public sector counterparts.  But absent rigorous evaluation, it remains just as conceivable that such services are merely skimming the most motivated participants and claiming substantial credit for outcomes in which their own contributions were slight.  Additional experimentation, coupled with rigorous evaluation, seems both feasible and desirable.

Note that in conducting evaluations of interventions such as these, the primary concern must be long-term differential impact.  Neither car stamps nor public service employment should be required to catapult beneficiaries into positions of high income.  The question is whether, over the long term, they can be said to have helped beneficiaries move onto a path that might lead to such a position where no such path existed before.

2.  Social Capital.  The quantitative research in this volume did not attempt to capture in its independent variables the kinds of private, community-based institutions that Wilson referred to in his discussion of "social buffers."  Yet much of the theoretical and qualitative discussion suggests that such institutions can make an important difference in the lives of children.  A literature on community based organizations documents (usually although not invariably through case study) the systematic efforts by planners, social workers, and organizers to nurture such organizations. (Sullivan 1993a; Naparstek 1993; Mayer 1981).  In recent years substantial attention has been given to the efforts of community-based financial institutions and housing-oriented community development corporations to redevelop neighborhoods that are the locus of concentrated urban poverty.  (Aikman 1993; Sullivan 1993a; Taub 1988).

Yet that same literature is quite explicit about how difficult it is to do careful evaluation research concerning such institutions (Aikman 1993).  Our knowledge about what kinds of public interventions are most likely to enable such programs to survive remains impressionistic, more a reflection of a priori theory than post hoc evaluation. (Lehman and Lento 1992).  And if they do survive, we know very little about how to measure their effects on neighborhood processes (as opposed to their effects on the individuals who live there).  It may ultimately be that social scientists have little in the way of evaluative rigor to offer to supplement local  policymakers' hunches about which of these programs are effective and which are not.  But at this point it would seem worthwhile to support efforts to develop evaluation protocols that would identify the most rigorous evaluations of such potentially critical programs that we are capable of (see Brown and Richman).

3.  Public Safety.  The quantitative research in this volume was unable to include measures of neighborhood safety in the research designs.  Yet a significant qualitative literature, and growing epidemiological and psychiatric literatures, suggest that some of the most destructive effects of ghetto neighborhoods are associated with the extremely high levels of violence found there.  (Marans and Cohen 1994; Fagan 1993; Sampson 1993; Earls 1992).  The need to develop interventions that bolster mechanisms of social control, especially of teenage boys, is an old one.  But it seems to have taken on pressing significance  and policy importance in recent years.  While we remain unconvinced that more prisons and longer jail sentences offer a long term cost-effective solution to reduce crime and increase public safety, other public safety innovations warrant ongoing study.

4. Drugs.  Too often in the urban landscape the buying and selling and consumption of illicit drugs is highly correlated with violence, imprisonment, and neglected children.  Regulation and criminalization continue to enjoy only mixed success.  But the case for decriminalization remains speculative (Nadelman 1989).  Any local experiments that do take place should certainly be subjected to rigorous evaluation.

5.  Schools.  Some of the most interesting "Core C" analysis in this volume concerned the interaction among race, gender, neighborhood, and the child's perception of the degree of support available in school.  (See Part V.E. above).  We still do not know nearly enough about the role schools play in exaggerating or mitigating the effects of variable neighborhood quality.  We do not know nearly enough about the differential effects schools may have on children of different races and genders.  We do not know nearly enough about how the school environment interacts with and may influence the home learning environment.  But, the Atlanta and Rochester data sets seem to promise exceptionally rich sources of further research on these topics.   Moreover, schools may be much better equipped than the larger neighborhood to provide effective loci for delivery of social and health services, and to provide safe havens from violence, guns, and drugs. While the growing segregation by race, income, and poverty status in American schools is becoming all too apparent (Orfield,, 1993),  schools still provide a focal point for intensive, synergistic interventions that can be effectively targeted on children in bad neighborhoods.

6.  Policy Analogues of Interaction Effects and Nonlinearities.  We mentioned earlier that observations about the empirical analysis of social phenomena can be translated into an analogous admonition about the analysis of the effects of social programs.  The empirical research in this volume made at least two methodological observations whose analogues in the policy domain warrant discussion.

The first such observation concerns so-called "interaction effects":  the fact that when one finds two characteristics (such as race and poverty) in the same observational unit (neighborhood or individual), the effects may be different from the linear sum of the effects of finding the two characteristics in isolation.   In the policy domain, the analogous concept is "synergy."  At least as a theoretical matter, it is quite possible that Program A (e.g., child care) might have no effects on its own, and Program B (e.g., transportation assistance) might have no effects on its own, but that the simultaneous implementation of Program A and Program B might have significant effects. In recent years, several commentators have pressed the theoretical case for interactions among policy interventions on behalf of low-income children who live in distressed urban neighborhoods.  These arguments have sometimes been expressed as calls for "integrated services," and sometimes in terms of "one-stop shopping." 

The abstract possibility of programmatic synergy does not strike us as sufficient to justify experiments that randomly mix and match different combinations of interventions.  But where there are strong theoretical reasons to expect synergistic benefits, experimental interventions can be designed to test for them directly.  This volume itself offers two fruitful approaches to developing theoretical criteria for ascertaining when synergistic benefits are likely to be obtained: the theoretical discussion of individual development processes, and the theoretical discussion of neighborhood processes (Chapter X, and XX).  And the chapter on state and local policy implications indicates that such synergism might be integrated  with various types of  "social capital" at the neighborhood level to produce positive outcomes.

The second such observation concerns so-called "nonlinearities":  the fact that when  a significant level of a phenomenon is present, the effects on behavior may be more than double the effects that exist at half that level.  (The best-known application in the social science context is probably Schelling's (1971) analysis of "tipping" in racially segregated neighborhoods.)  In the policy domain, this phenomenon has a direct analogue:  at least as a theoretical matter, it is quite possible that Program A (e.g., counseling) might have no effects at one level (e.g., 1 hour per week) but might have significant effects at a much higher level (e.g., 1 hour per day).  And that possibility is frequently invoked in political debates, sometimes defensively to distinguish past experimental failures (e.g., "the fact that lightly funded state enterprise zones have failed to produce significant results tells us little about whether heavily funded federal enterprise zones would"), and sometimes offensively as an argument for "saturation" ("if we spend a huge amount in a massive intervention, we are bound to make a difference").

Our own intuition is that nonlinearities pervade the policy domain.  At the low end, virtually any successful program is likely to produce no detectable benefits if it is attempted on a small enough scale.  At the high end, our sense is not that one is likely to see accelerating benefits from a single program, but the opposite:  that one is more likely to see diminishing marginal returns.  (This is different from the possibility of beneficial interactions among multiple programs, discussed above, but as we noted there our intuition is to believe that one should have a priori theoretical reasons to suspect the presence of such interactions before investing in testing them.)  The net result is that experimental interventions are likely to provide the most valuable information for those who want to assess the costs and benefits of a prospective policy if the effects of several, significantly different levels of intervention are tested simultaneously.  That is what is being attempted in the Expanded Child Care Options Project, discussed above.  (Hollister 1993).  And the positive impacts of the MDRC evaluation of the GAIN program in Riverside County, California (as compared to other counties in California) points to the importance of organizational culture in achieving good results for welfare to work programs (Gueron, 1994).

Procedural Suggestions: The Levers of Federalism

One of the most problematic questions under our federal system concerns the design of programs that are supported at the federal level.  Indeed the question of devolvement from federal to state control of a number of social programs is now on the front burner of American national policy.  How much specificity in design should be handled at the federal level, how much left to more decentralized units of activity?  How much administration should be carried out by federal employees and how much by others?  How much funding should come from the federal purse and how much from others?

These questions arise most strongly in the context of full-scale national programs, but they arise in the context of federally supported experiments as well.  In the past, the federal government has sometimes launched and monitored localized experiments and evaluations on its own.  Sometimes it has been able to do so at times through new appropriations of funds (e.g., EHAP, SIME/DIME programs).  Sometimes it has been able to do so in conjunction with programs of cooperative federalism (e.g., MDRC evaluations of JOBS in California).  And sometimes it has been able to do so by placing conditions on program grants to state and local units of government (e.g. JTPA programs).

We do not have strong a priori commitments about how such questions of allocation should be dealt with in the context of experimentation.  We do believe, however, that close attention to the problems associated with the policy analysis of neighborhood-level interventions might provide some useful insights.  In particular we would stress the specific domains of implementation and evaluation.

Two of the most vexing problems in funding and evaluating new experiments may not be at the level of high-level design.  Rather, those problem may be at the level of implementation and evaluation.  At the level of implementation, we would emphasize that whether a program has beneficial effects or not may depend as much on who runs it as on how it is structured (Manski, 1990).  That fact creates something of a theoretical conundrum in the formation of federal policy.  Are we interested in experiments that involve the best imaginable quality of implementation?  Or are we interested in experiments that involve only a representative quality of implementation?  The former approach may leave us overly optimistic about what benefits may flow from generalizing an experimental intervention.  (That is what some have argued about the evaluation of the Perry Preschool Head Start program.)  But the latter approach may leave us overly pessimistic about what is feasible.

We would argue that, at least at the stage of experiment and evaluation, it makes more sense to design experiments that make use of the best available mode of implementation.  That approach allows one to treat any benefits found as a plausible upper bound on the benefits to be obtained from a generalized, replicated program.  Moreover, careful evaluation should permit systematic discussion of how expensive or difficult it might be to generalize and replicate successes.

What forms of implementation are most likely to maximize success?  Once again, one should be wary of overgeneralizations.  But some of the best impressionistic accounts of neighborhood-level interventions share two common themes:  charismatic leadership and participatory development.  Energetic, creative, and charismatic leaders can sustain programs that would otherwise fail.  (Aikman 1993; Smith 1993). Participation of residents in program development can have two benefits.  It can leave them more committed to making sure the program succeeds.  (Aikman 1993; Halpern 1993).  And, perhaps even more significantly, that participation can generate an important "product" of the program:  higher levels of human capital in community residents.  (Brown and Richman 1993).

At the level of evaluation, we would emphasize two related observations. First, as Aber (1993) has persuasively argued, new evaluation research should be shaped by basic theory and research.  To the extent variations in neighborhood context are able to shape individual child development outcomes, one should also expect that variations in neighborhood context will shape the results of programmatic intervention.  Comprehensive evaluations should explore between-neighborhood variations in program effects, as well as within-neighborhood variation between control and intervention subjects, in order to provide the most useful information to policymakers.  Second, evaluators must be sensitive to the point about implementation made in the preceding paragraphs.  Several observers have suggested that idiosyncratic features of program implementation, such as leaders' charisma and community participation in development, may be critical to program success or failure.  To the extent evaluators can track and measure such characteristics, evaluations are likely to be much more meaningful.

And we would conclude with two observations about the problem of funding.  First, federal administrators are unlikely to have access to the sorts of information that might allow them to discern which leaders will be perceived as charismatic by the community, and which leaders will be able to galvanize community participation.  Over the past few decades, however, private foundations have been actively involved in making precisely those kinds of distinctions.  It might therefore make sense for federal funds to be disbursed as matching grants:  directed towards programs that are willing to submit to rigorous evaluation and that have also been able to obtain equivalent levels of funding from major foundation sources.

Second, one of the most important unanswered research questions is also an important unanswered funding question.  As we have noted above, confirmation of the existence of neighborhood effects does not answer the question whether those effects are linear or nonlinear.  If the effects are nonlinear, a case can be made that when major programs are implemented on a universal level, such as the assistance to schools that educate poor students pursuant to Title I, funds should be distributed in a nonlinear manner as well.  However, if the effects are related (for example) in a linear manner to the rate of child poverty in the neighborhood, then any disproportionate distribution of those funds might be less effective than a proportionate distribution would be.  More basic research is thus likely to continue to pay policy dividends well into the future.

VIII.  Conclusion

Perhaps we should conclude by pointing out one traditional type of urban policy which we did not  mention in our experimentation  recommendations.  Specially targeted federal assistance for urban economic development is not on our list of priorities.  To the extent that such activities promote political patronage and other objectives unrelated to antipverty or child development goals, we believe that federal efforts should be designed to do the least harm (Bartik 1994).  The case for massive "neighborhood" economic development policy remains unmade.

On the other hand, we would be remiss if we did not observe that the research in this volume has been conducted in an environment where we believe there is already a compelling case to support policy reform.  It is now apparent that full-year, full-time work will not by itself ensure that parents with little education or experience will be able to lift their children out of poverty (Jencks and Edin 1993). The expected earnings trajectories for undereducated single parents are so low that longer term income support is necessary if many families with children are to escape poverty (Burtless 1994; Danziger and Lehman 1994).  The expansion and extension of the EITC is helpful; it may be sufficient for two-parent families when both can work and  parent as a team.  But for single parents the EITC  is not enough.  Substitutes for the absent parent in the form of child care and child support are also needed.  Child care may come from family members or via government subsidy to formal providers.  Child support may come from absent fathers (formally or informally), government child support assurances, or even refundable income tax credits.  The point is that something beyond work and the EITC -- a comprehensive income support package (Rainwater 1993) -- is necessary if we are going to turn work into a vehicle by which single parents have better choices for their children. 

It is unfortunate that much of today’s welfare reform rhetoric is more concerned with the choices and opportunities of state governments than with the choices and opportunities of poor children.  We remain hopeful that in the end a compassionate concern for the vulnerable will lead to constructive income support reform.  Beyond the domain of income support, the research in this volume has not led us to abandon the sense of priorities we held before we were exposed to it.  We would support carefully designed experimental efforts to give both adults and their children a better set of true life choices:  residential mobility, an education, a job, a sense of hope for the future.  And we would make “neighborhood” policy a secondary priority, except where required to promote a safe environment in which “people” policy and income support policy reform can operate.

It is perhaps predictable that a chapter discussing the federal policy implications of social science research concerned with the effects of poor neighborhoods on children should be cautious.  Calls for experimentation, further evaluation, and still more research are de rigueur.  But our conformity to that norm should not be read as in any way critical of the careful research that is included in this volume.  If the dogs are standing mute, good researchers have to leave policymakers to make their own music.  We can only hope that it will be the type of music that we would like to hear.

References (Partial List)


Aber, J. Lawrence.  1993.  "Effects of poor neighborhoods on children, youth and families:  theory, research and policy implications."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Aikman, Lisa.  1993.  "Fighting Urban Poverty:  Lessons from Local Intervention Programs."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Anderson, David.  1994.  "The Crime Funeral", N. Y. Times Magazine.  Sunday, July 4.


Anderson, Elijah.  1991.  "Neighborhood Effects on Teenage Pregnancy." The Urban Underclass. C. Jencks and P. Peterson, eds.  Washington, D.C.:  Brookings.


Bartik, Tim.  1994.  "What Should the Federal Government Be Doing About Urban Economic Development?"  Upjohn Institute Working Paper #94 - 25:  Kalamazoo:  Upjohn Institute.  March.


Brooks-Gunn, J. and F. F. Furstenberg, Jr.  1987.  "Continuity and Change in the Context of Poverty:  Adolescent Mothers and Their Children."  In J.J. Gallagher & C. T. Ramey (Eds.), The Malleability of Children (Baltimore:  Brookes Publishing Company), pp.171-188.


Brown, Prudence, and Harold A. Richman.  1993.  "Communities and Neighborhoods:  How Can Existing Research Inform and Shape Current Urban Change Initiatives?"  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Burtless, Gary.  1994.  "The Employment Prospects of Welfare Recipients" prepared for the Urban Institute conference on "Self-Sufficiency and the Low Wage Labor Market", Arlington, VA.  April 20.


Burtless, Gary, and Lawrence Mishel.  1993.  "Recent Wage Trends:  The Implications for Low-Wage Workers."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Clark, Rebecca.  1992.  "Neighborhood Effects on Dropping Out of School Among Teenage Boys."  Mimeo.  Washington, D.C.:  Urban Institute.


Crane, Jonathan.  1991a.  "Effects of Neighborhoods on Dropping Out of School and Teenage Childbearing."  The Urban Underclass. C. Jencks and P. Peterson, eds.  Washington, D.C.:  Brookings.


----------.  1991b.  "The Epidemic Theory of Ghettos and Neighborhood Effects on Dropping Out and Teenage Childbearing."  American Journal of Sociology.  96:1226‑59.


Diamond, Paul.  1994.  "The Emerging Realities for Addressing Urban Poverty and Decline in the United States" mimeo.  Washington, D.C.:  White House, July 26.


Earls, Felton.  1992.  "A Commentary on Papers of the Working Group on the Social Ecology of Crime and Drugs."  Paper presented at the Research Conference on the Urban Underclass.  Ann Arbor.  June.


Fagan, Jeffrey.  1993.  "Crime, Drugs and Neighborhood Change: The Effects of Deindustrialization on Social Control in Inner Cities."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Freeman, Richard.  1991.  "Crime and the Economic Position of Disadvantaged Youth", NBER WP # 4027, Cambridge, MA:  NBER.


Gueron, Judy.  1994.  Seminar presented to Russell Sage Foundation, May 6.


Gueron, Judy and Edward Pauly.  1991.  From Welfare to Work (New York:  Russell Sage).


Halpern, Robert.  1993.  "Neighborhood-Based Services / Best Practice in the Inner City: A Memo in Two Parts." Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Hechinger, Fred M.  1994.  "Saving Youth from Violence" Carnegie Quarterly Vol.39, No.1, Winter.


Heckman, James.  1994.  "Is Job Training Oversold?"  Public Interest, Sprint: 91-115.


Hollister, Robinson G.  1993.  "Social Policy Research Through Major Data Gathering Projects:  Four Contemporaneous Examples."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Jencks, Christopher and Kathy Edin.  1993.  "Welfare" in C. Jencks, Rethingking Social Policy.  New York:  Harper-Collins, 1993:  205 - 235


Jencks, Christopher, and Susan Mayer.  1990.  "The Social Consequences of Growing Up in a Poor Neighborhood:  A Review."  Concentrated Urban Poverty in America.  M. McGeary and L. Lynn, eds.  Washington, D.C.:  National Academy Press.


Lehman, Jeffrey.  1994.  "Updating Urban Policy."  Confronting Poverty.  S. Danziger, G. Sandefur, and D. Weinberg, eds.  Cambridge:  Harvard University Press.


Lehman, Jeffrey and Rochelle Lento.  1992.  "Law School Support for Community-Based Economic Development in Low-Income Urban Neighborhoods." Journal of Urban and Contemporary Law 42:65‑84 (May).


Lehman, Jeffrey and Sheldon Danziger.  1995.  "Ending Welfare as We Know it:  Values Economics and Politics." Domestic Affairs.


Marans, Steven R., and Donald J. Cohen.  1994.  "Children and Inner-City Violence:  Strategies for Intervention."  Psychological Effects of War and Violence on Children.  L. Leavitt and N. Fox, eds.


Massey, Douglas, and Nancy Denton.  1993.  American Apartheid.  Cambridge: Harvard University Press.


Mayer, Neil.  1981.  Keys to the Growth of Neighborhood Development Organizations.  Washington, D.C.:  Urban Institute.


Nadelman, Ethan.  1989.  "Drug Prohibition in the U.S.:  Costs, Consequences and Alternatives."  Science Vol.249, Sept.1:  939-946.


Naparstek, Arthur. 1993.  "Rethinking poverty Through a Community-Building Approach:  Policy Memorandum on Community Reinvestment."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


National Academy of Sciences.  1993.  Head Start: Final Report, mimeo.


National Commission on Children.  1991.  Beyond Rhetoric  Washington:  National Commission on Children.


Orfield, Gary.  1993.  "The Growth of Segregation in American Schools:  Changing Patterns of Separation and Poverty since 1968".  1993.  Harvard University, mimeo.


Rainwater, Lee.  1993.  "The Social Wage in the Income Package of Working Parents" Luxembourg Income Study, Working Paper #89.  Luxembourg:  LIS.  August.


Rosenbaum, James. 1991.  "Black Pioneers - Do Their Moves to the Suburbs Increase Economic Opportunity for Mothers and Children?"  Housing Policy Debate.  2:1179‑1213.


Rusk, David.  1993.  Cities Without Suburbs.  Baltimore:  Johns Hopkins University Press.


Sampson, Robert J.  1993.  "Concentrated urban poverty and crime: a synopsis of prior community-level research."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Schelling, Thomas.  1971.  "Dynamic Models of Segregation."  Journal of Mathematical Sociology.  1:143‑86.


Schill, Michael.  1992.  "Deconcentrating the Inner City Poor." Chicago-Kent Law Review.  67:795‑853.


Smeeding, Timothy.  1994.  "Car Stamps:  Economic Mobility for the Geographically Challenged." Mimeo. Syracuse University.


----------.  1995.  "An Interdisciplinary Model for Studying Poor Children" in P.L. Chase-Lansdale and J. Brooks-Gunn, eds., Escape from Poverty:  What Makes A Difference for Poor Children?  Cambridge:  Cambridge U. Press.


Smeeding, Timothy and Barbara B. Torrey.  1988.  "Poor Children in Rich Countries"  Science, No.247.  Nov. 11:1135-1140.


Smith, Thomas J.  1993.  "Improving Practice in the Youth Employment Field."  Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Spenser, Margaret, Janet Blumenthal and Elizabeth Richards, 1994, "Caring for Poor Children:  Child Care and Children of Color"  to appear in P.L. Chase-Lansdale and J. Brooks-Gunn, eds., Escape from Poverty:  What Makes A Difference for Poor Children?  Cambridge:  Cambridge U. Press.


Sullivan, Mercer.  1993.  More Than Housing:  How Community Development Corporations Go About Changing Lives and Neighborhoods.  New York:  New School for Social Research.


----------.  1993a.  "Community Development as an Anti-Poverty Strategy." Memorandum prepared for SSRC Policy Conference on Persistent Urban Poverty, Washington, D.C.  Nov. 9‑10.


Tienda, Marta.  1989. "Poor People and Places:  Deciphering Neighborhood Effects on Policy Outcomes." Paper presented at the Annual meeting of the American Sociological Association, San Francisco, August 9‑13.


Wilson, William Julius.  1987.  The Truly Disadvantaged.  Chicago:  University of Chicago Press.


----------.  1993.  "Public Policy Research and The Truly Disadvantaged." The Urban Underclass. C. Jencks and P. Peterson, eds.  Washington, D.C.:  Brookings Institution. 1993.  "The New Urban poverty and the Problem of Race." The 1993 Obert C. Tanner Lecture, Ann Arbor.  October 22, 1993.

    [1]Jencks and Mayer (1990) note that if peer influence effects are linear, so that "good eggs" are equally beneficial for their neighbors wherever they are, and "bad apples" equally harmful, the argument for the social benefits of "deconcentrating" the bad apples is much more complicated than the argument is if the effects are nonlinear.  Crane (1991a, 1991b) found such nonlinear "epidemic" effects in his analysis of dropping out and teenage childbearing using 1970 Census data.  His findings are consistent with the enthnographic account of teenage pregnancy in Anderson (1991).  But Clark (1992) was unable to find such effects in her analysis of dropping out using 1980 Census data.