INTRODUCTION AND OVERVIEW

The sea change in social welfare policy that occurred with passage of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 has appropriately been accompanied by much discussion and debate. Most of this attention focuses on work requirements and time limits, the core elements of welfare reform. But a second tier conversation has also been taking place about the roles that child care, transportation, and health care will play in enabling welfare recipients to meet the work requirements. This second tier discussion is significant not only for its substantive content but also for its symbolic value. It reflects the acknowledgment that moving recipients from welfare to work requires more than job skills and work readiness. It involves, in addition, solving the logistic problems of everyday life and satisfying very basic needs so that getting and keeping a job is possible.

What is noteworthy about this second tier discussion is that housing is not part of the conversation, even though housing policy poses two obvious questions for welfare reform. The first is whether living in a decent, affordable home in a good neighborhood improves a welfare recipient's chances of achieving self–sufficiency. Is there an argument for housing subsidies that is analogous, for example, to the argument that good and affordable child care and transportation to and from work improve the ability to get and keep a job? The second is how welfare reform will reverberate on the housing assistance system for the poor. This system, broadly defined,¹ represents a $28 billion component of the federal budget, larger than the $20 billion Aid to Families with Dependent Children (AFDC) program that welfare replaces.²

Why is housing being left out, as it has been left out of the last 20 years of angst and analysis of how the welfare system should be fixed? Part of the answer lies in the bureaucratic separation of housing and welfare in congressional committees, executive branch agencies, and delivery systems (Hornburg 1997). But much of the explanation is that the housing assistance system has been preoccupied with its own major transformation, which has been unfolding over the last two decades but, unlike welfare reform, is still very much in process. Two to three decades ago—when only a relatively small proportion of households receiving housing assistance also received welfare—it was reasonable that the housing assistance system consciously, even proudly, distinguished itself from the welfare system. The fact that the housing assistance system has always been only one small part of a very large housing sector and, therefore, only one focus of U.S. housing policy, made this separation seem even more justifiable.³ Indeed, during this period, inclusion of the housing assistance system in the welfare debate would have been not only unexpected but unwelcome by many in the housing sector.

Today, the situation is substantially different. Even though the welfare and housing systems were originally designed to assist different clienteles, there is now substantial overlap (Newman and Schnare 1992). The proportion of households living in assisted housing that also receive assistance from one of the three main income assistance programs (AFDC, Supplemental Security Income [SSI], and General Assistance) has more than doubled since 1966, now reaching more than 50 percent.⁴ Similar increases have occurred in the proportion of income assistance recipients who also receive housing assistance. Between 1981 and 1995, for example, this fraction increased from about one–fifth to one–third.⁵ The overlap is even greater when AFDC recipients alone are considered. In the mid–1990s, for example, nearly 50 percent of households with children living in public housing were receiving AFDC.⁶ Thus, nearly half of all families with children who live in public housing are potentially affected by welfare reform. Taken together, these changes have moved a significant part of housing policy squarely into the safety net.

This transformation has been furthered by the shift in public opinion regarding the principles that ought to govern all assistance programs for the poor. The assistance programs developed in the New Deal era had the basic goal of income replacement. The social values and attitudes undergirding current welfare reform efforts—most prominently, that economic independence and self–sufficiency are the ultimate goals, and that there are reciprocal obligations between recipients and government—logically apply to housing assistance as well.

The cumulative effect of all these changes has been a rethinking of the goals of housing policy for the poor and the tools for achieving those goals. While the debates continue, strong arguments are now being made for a new orientation, in which helping poor people live in decent housing—the traditional goal of housing policy for the poor over its more than 50–year history—is no longer viewed as an end in itself, but as a means to economic independence.

Housing has been left out of the welfare reform conversation, then because it hasn't been ready to participate. One important measure of not being ready is not having much hard information to bring to the table about how welfare policy does—and how it should—interact with housing assistance policy. This book begins to provide that information. It has three specific goals: (1) to present new research on the projected impacts of welfare reform on housing; (2) to distill the most important lessons learned from past research and demonstration programs, in both housing and welfare arenas, aimed at moving recipients to self–sufficiency; and (3) to identify the special challenges welfare reform presents to low–income housing practitioners and researchers, and promising approaches to addressing these challenges. The immediate issue of the ripple effects of welfare reform on housing, the focus of this book, can best be addressed within a broader agenda of questions that must be answered if we are to advance our knowledge of the role of housing in helping move poor people into the economic mainstream.

Does Decent Housing Improve the Changes of Achieving Economic Independence?⁷

The most compelling evidence for providing a role for housing policy in welfare reform would show that decent, safe, and affordable housing in a decent neighborhood actually increases the ability of the poor to achieve sustained economic self–sufficiency. To answer this question definitively would require a controlled experiment in which otherwise similar sample members were assigned to an experimental "decent housing" group, and a control group were not. Since the two groups would be the same except for their housing situation, any difference in subsequent income, employment, welfare receipt, and so on, could confidently be attributed to the housing difference. Unfortunately, no such experiment has been undertaken, forcing us to rely on more indirect evidence.

At the low–quality extreme, of course, are the homeless. Although studies suggest that 20–30 percent of the homeless have jobs, most observers take it as being self–evident that being homeless increases the difficulty of pursuing education or job training, or getting and keeping a job. But, given that one does have housing, does the quality of that housing—measured by such characteristics as affordability, physical adequacy, location (e.g., accessibility to training, work, or school; safety or other conditions of the neighborhood; quality of the environment for children), and stability—make a difference to one's chances of achieving economic independence? The three major housing dimensions that have been studied in this connection are proximity to attainable jobs (the "spatial mismatch" hypothesis), housing quality, and neighborhood quality more generally.

The Spatial Mismatch Hypothesis

The spatial mismatch hypothesis holds that unemployment (and therefore welfare use) is high in inner cities because there are no entry–level jobs in the areas where the poor can afford to live.⁸ At this writing, the accumulated evidence from the most rigorous studies of this topic suggests that spatial mismatch plays some role in employment, though the size of its influence on getting and keeping a job remains in dispute (Cutler and Glaeser 1995; O'Regan and Quigley 1996; Raphael 1998; Ihlanfeldt 1997; Holzer 1991; Kain 1992). Nonetheless, on at least the implicit assumption that these effects are large enough to present real impediments to work, transportation subsidies have been included in many states' welfare reform plans. But we do not know the relative benefits and costs of such subsidies relative to the benefit–cost picture for subsidizing housing in the vicinity of available work.

Some observers question whether "dispersal," such as through housing opportunities outside inner–city poverty areas can, in fact, make much difference to long–term economic independence (e.g., Hughes 1987). They argue that the basic problem facing poverty area residents is lack of job qualifications, which housing does not affect. The work of Case and Katz (1991) supports this view. They find that neighborhood influences (peer criminal activity, substance abuse, teen pregnancies) were important even in the tight labor market in which they did their research. None of these researchers address the relative advantages and disadvantages of developing those qualifications while surrounded by various neighborhood constraints, however. Such constraints characterize a significant fraction of assisted housing developments.

The Housing Quality Hypothesis

One version of the housing quality hypothesis holds that housing quality affects health, and that health affects one's chances of achieving economic independence. The proven deleterious effects of lead–based paint on children is a clear example of such a link. Except for extreme housing dilapidation and infestation, analysts have generally not found a direct connection between housing and health (Kasl 1976; Wilner 1962; Weicher 1980, 1982). But much of this research is nearly two decades old or older and methodologically primitive when judged against current standards for testing complex hypotheses.

Housing quality may also mean a more spacious dwelling (in which each resident can find a quiet place to read or study), a more comfortable dwelling (with adequate heat in winter and air conditioning in summer), the presence of appropriate furnishings and perhaps time–saving amenities (such as a dishwasher, washing machine, and dryer). Here, again, there is little systematic research. Mayer's (1997) recent work suggests that there is some effect, but the measures of physical quality used in this research are difficult to evaluate.⁹

The Neighborhood Quality Hypothesis

The neighborhood quality hypothesis argues that high rates of drug abuse, other criminal activity, unemployment, and teenage pregnancy affect a resident's willingness and ability to work toward economic independence in the mainstream economy. The research literature offers different conceptualizations of how this hypothesized link might operate (Jencks and Mayer 1990; Ginther et al. 1993) which, for simplicity, I collapse into two broad categories. The first holds that children's perceptions of appropriate values, norms, and behavior—both good and bad—are influenced by their peers and adult role models (Wilson 1987; Crane 1991; Clark 1992). For example, children who grow up in "good" neighborhoods are, therefore, more likely to adopt mainstream values and behaviors than those who grow up in problem neighborhoods. The second broad conceptualization is the very different notion that growing up in a "good" neighborhood may actually be detrimental. According to this view, children judge their success by comparing themselves with their peers. Research has shown that, other things equal, children from poor families do worse academically than those from more affluent backgrounds. Depending on many factors (including an individual's psychological makeup), the response to growing up in a neighborhood in which the majority of families are more affluent, or attending a school in which the majority of students are from more affluent families, may be to work harder. But it may also be to give up, which is the response hypothesized by this second view.

Findings about the effects of neighborhood quality are inconsistent. Wilson's research on the ghetto poor established a convincing theoretical case for the negative effects of isolation and concentration on aspirations and achievement. The empirical studies of Datcher (1982), Brooks–Gunn et al. (1993), Case and Katz (1991), Crane (1991), Clark (1992), and Aaronson (1985) all suggest that growing up in a more affluent neighborhood has positive influences on a variety of outcomes for adolescents from poor families—including educational attainment, criminal behavior, drug abuse, and labor force participation. It is now clear, however, that beneficial effects of good neighborhoods have a direct counterpart in deleterious effects of bad neighborhoods. Some of these same researchers, for example, find that the positive effect of middle–income neighbors is more powerful than the detrimental effect of poor neighbors. But male adolescents who are black and poor appear relatively less affected by the presence of middle–class neighbors than other groups (Duncan 1994), although they become responsive when the middle–class neighbors themselves are black. And Corcoran et al. (1991), Evans et al. (1992), Plotnick and Hoffman (1993), and Ginther et al. (1993) find few, if any, neighborhood effects.

It is important to note that even the research that report significant neighborhood effects find the size of these impacts, at least on child outcomes, to be modest. In probably the most rigorous study to date, Aaronson (1995) found that a 10 percent increase in the rate of neighborhood poverty would reduce the likelihood of graduating from high school by 7 percent. The Gautreaux program provides the most direct test of neighborhood effects to date. This program resulted from a successful court case charging the Housing Authority of the City of Chicago with racial discrimination in provision of public housing to its minority population. Under the Gautreaux decision, the Chicago housing authority provided housing certificates and housing counseling to public housing families to help them move out of inner city Chicago to the suburbs. Research on the program suggests overwhelmingly positive effects for those who moved out of the inner city compared to those who remained in the city. Suburban movers were 25 percent more likely than city dwellers to have a job after their move, and school dropout rates were much lower for the suburban group of households (Rosenbaum 1991). The lack of a true experimental design, no follow–up data on dropouts, and other methodological questions, however, indicate caution in interpreting the results.

Both the methodological and substantive questions raised by the Gautreaux program will potentially be addressed by the Moving to Opportunity (MTO) Demonstration Program, which the U.S. Department of Housing and Urban Development (HUD) launched in six metropolitan areas around the country in 1994. MTO used an experimental design, randomly assigning public housing tenants to either an experimental group, who were given Section 8 certificates and housing counseling to help them locate a unit in a low–poverty neighborhood and to refer them to other assistance they might need, or to one of two control groups. One of these groups received a certificate but no housing counseling, the other received neither a certificate nor housing counseling.¹⁰ Participating households are to be tracked for 10 years. MTO represents the best opportunity to date for disentangling the effects of neighborhood characteristics from the effects of parental and family characteristics.¹¹

Is Assisted Housing Linked to Decent Housing and Neighborhoods?¹²

Households living in assisted housing confront both a macroenvironment comprising all the units in the development and surrounding neighborhood, and the microenvironment of their individual housing unit. According to the most recent empirical evidence, the macroenvironment of assisted housing is bad. Public housing is disproportionately located in neighborhoods where incomes are very low, unemployment and poverty rates are high, and the quality of the surrounding housing stock is poor. And neighborhoods surrounding other project–based developments are roughly equivalent to the neighborhoods of welfare households who do not receive housing assistance (Newman and Schnare 1997). Thus, roughly 75 percent of all assisted housing units are located in distressed neighborhoods. Some research suggests that, for a sizable fraction of residents living in large public housing developments, venturing out into the neighborhood is rare, and the orbit of activity is restricted to the development itself (Shlay and Holupka 1991). These are also the developments that have large concentrations of households with very low incomes, high welfare dependency rates, and low educational achievement (Newman and Schnare 1993). If the concentration of disadvantaged neighbors has a negative effect on child outcomes, then both components of the macroenvironment of much assisted housing—the surrounding neighborhood and the development itself—suggest that children growing up there are likely to experience worse outcomes than other children.

How about the microenvironment? One element of the microenvironment is the quality of the dwelling. Because assisted housing regulations require that all dwellings meet housing quality standards, the overall physical adequacy of these units should be higher than for comparable households who do not receive assistance. This expectation is generally borne out by recent empirical evidence (Newman and Schnare 1993), although the General Accounting Office's review of a sample of dwellings rated by housing inspectors as meeting the housing quality standards of HUD found that "some properties clearly violated" these standards (U.S. General Accounting Office 1994). The quality of furnishings and other material possessions in the dwelling should also be greater because assisted housing residents pay only 30 percent of their income for rent (a smaller proportion than that paid on average by nonassisted poor households), but there is no empirical evidence about these expenditures. On balance, then, assisted housing environments may be more physically adequate than other low–income rental properties. Whether this difference matters for a range of outcomes is not yet known.

A second element of the microenvironment is the family. As alluded to earlier, previous research suggests that family background characteristics are closely related to a number of child outcomes such as educational attainment (e.g., Aaronson 1995; Duncan 1994). It is also plausible that such characteristics are related to adult outcomes, such as work activity. In public housing, which constitutes about one–fifth of all assisted housing, somewhat less than half of all household heads have a high school education, nearly 90 percent of households are headed by a single female, and nearly half receive welfare. The profile is slightly less disadvantaged among certificate and voucher users, and substantially so among those living in private developments that receive housing subsidies (the group that accounts for the largest share of all assisted housing). On balance, then, in assisted housing as a whole there may be modest positive effects, at best, on such behavioral outcomes as labor force participation or educational attainment.

Another element of the microenvironment is how the household responds to the housing subsidy. As noted in chapter 2, there are plausible arguments for expecting both more positive, and more negative, effects on behavioral outcomes. If the receipt of a housing subsidy allows the family to increase the consumption of necessities that are forgone when housing cost burdens are very high, the effects may be positive. But to the extent that the housing subsidy acts as a disincentive for the parent(s) to engage in work or other activities to move to self–sufficiency, the effects on achievement by both children and adults may be negative. Such disincentive effects of housing assistance are discussed further below.

Does Assisted Housing Act As a Bonus or Enhancer?¹³

Beyond its direct effects on behavioral outcomes, housing subsidies could be used as a "bonus" to spur welfare recipients to comply with the requirements of welfare reform. Compared to transportation allowances, Medicaid, or even day care, housing assistance provides a substantial, tangible, and immediate reward. Such a bonus effect could be a powerful argument for linking housing assistance to welfare reform. Many states are reporting that sizable fractions of recipients are not showing up for the required orientation sessions and some are not complying with other new rules (Holcomb et al. 1998; Pavetti et al. 1995). While the reasons for the high rate of no–shows and noncompliance are not yet understood, one possibility is lack of motivation among some welfare recipients which, in turn, will reduce the effectiveness of welfare reform. Providing a large bounty like housing assistance might make a difference.

A second way in which housing might combine with services to increase self–sufficiency is if increased housing security made the impact of other services stronger. For example, in the welfare–to–work demonstrations, those with the most persistent dependency experienced smaller gains in employment and earnings than the somewhat less dependent (Burtless 1989; Holcomb et al. 1998; Pavetti et al. 1997; Friedlander and Burtless 1995). This has prompted speculation that more intensive services might be more effective (Burtless 1989; Friedlander and Burtless 1995) in making welfare reform a success.

HUD demonstrations have tried both approaches, although methodological weaknesses in the associated evaluations prevent us from learning much about the behavioral effects of combining housing and services. (Chapter 5 reviews some of these demonstrations, chapter 6 summarizes the major lessons from the welfare–to–work demonstrations, and chapter 7 critiques some of them.) In addition to formal federal demonstration programs, elements within the housing assistance system itself have been moving in the direction of combining housing with services. For example, many public housing authorities¹⁴ have either implemented supportive service programs in some of their public housing projects or are running demonstration programs of this kind. And many housing managers have cobbled together supportive services in their privately owned, federally subsidized developments. Unfortunately, there has not been much support available to systematically evaluate these initiatives. For example, research on a family–oriented social service intervention in Baltimore public housing reported limited effects across a wide range of outcomes including income, education, and social and behavioral characteristics. But since the evaluation was conducted early in the intervention, no information will be available on long–term effects (Shlay and Holupka 1991) There has been one rigorous evaluation of the effects of child care centers located in public housing projects, although not on exit rates from public housing, and the results are positive. The proximity to such centers was associated with increases in the probability of being employed, increased earnings, increased work hours, and decreased welfare rates (Robins 1988). Not surprisingly, these effects were particularly strong for parents with preschool children.

Is Housing Assistance Itself a Disincentive?¹⁵

Although survey data indicate that welfare recipients who also receive housing assistance have lower employment rates than those who do not receive housing assistance, these differences may be the result of administrative decisions, such as preference rules for housing programs that gave priority to the most disadvantaged—not the result of the disincentive effect of housing assistance, per se.

Theoretically, there are plausible arguments for expecting both more positive, and more negative, effects. Because housing assistance provides a stable and affordable residence, housing assistance may be a springboard into the job market, and, therefore, success under welfare reform. On the other hand, by offering housing security, housing assistance may dull the incentives to work or cushion the risk of noncompliance with welfare reform requirements, thereby contributing to negative outcomes. Perhaps a more significant way housing assistance may have the negative outcome of discouraging work is the housing assistance system's 30 percent tax rate on income. The tenant rent contribution is set at 30 percent of adjusted gross income. Thus, every dollar increase in income is effectively taxed at 30 percent. Housing assistance benefits decline as income rises and phase out entirely when income rises beyond a certain point. Under the certificate and voucher programs, for example, when 30 percent of income has been greater than or equal to the rent for six months, the tenant is no longer eligible for a subsidy. Since this rule is equivalent to a tax on earnings, it should be a work disincentive, other things equal.

What are the likely effects of this rule under welfare reform, which requires recipients to work and, presumably, to have earnings? Past research on the effects of employment and training programs for welfare recipients and on state work–welfare programs indicates only a modest short–run net gain in earnings (between $560 and $1500 a year) (Cottingham 1989; Burtless 1989; Ellwood 1989; Gueron and Pauly 1991; Friedlander and Burtless 1995; Nightingale and Holcomb 1997; Strawn 1998). If welfare reform follows this pattern, the earnings impact would not be enough to make most recipients lose eligibility for housing assistance. Even if incomes rose to the poverty line, housing affordability problems would still be pervasive.

There are no data on the longer–term impacts of housing programs on earnings, however, leaving unanswered the question of a recipient's motivation to increase earnings when the housing assistance payment is either reduced or lost as a result. By contrast, earnings effects of cash welfare have received great attention. Research on the Seattle–Denver Income Maintenance Experiment, for example, showed "substantial and significant reductions" in annual earnings for both males and females (SRI International 1983). These effects established the rationale for introducing more rigorous work requirements into welfare programs (Aaron 1984). It is unclear how these findings relate to the case of housing, however. Prior to welfare reform, the AFDC program had a tax rate on additional income that was more than double that for housing assistance.¹⁶ The most that can be said about housing is that, since the income eligibility cutoffs for housing assistance are roughly 33 percent higher than the poverty line, housing benefits continue over a substantial earnings range. Since the total loss of a housing benefit is unlikely for this reason, the disincentive may not be very large. This conclusion would change, of course, if housing policy either neutralized the disincentive effects by reducing or eliminating the tax on earnings for some period of time or, alternatively, eliminated the automatic reduction in out–of–pocket rent payments for those not in compliance with the welfare reform requirements.

One other feature of housing assistance may also play a direct role in employability: neighborhoods. Even though the empirical evidence is not unambiguous, neighborhood conditions are likely to have some effect on access to education and jobs (as demonstrated by chapter 4), and the absence of suitable role models in distressed neighborhoods may discourage individual initiative and integration with the broader community. The sobering findings regarding the neighborhoods of public housing and, to a lesser extent, private developments built and operated with federal subsidies suggest that housing assistance may be hurting rather than helping these recipients achieve economic independence.

The Rest of the Book

Part 1 scopes out the problem and addresses such basic questions as: What kinds of impacts will welfare reform have on housing? How large are these impacts on numbers and characteristics of people affected, and on government budgetary outlays? And what is the geography of these impacts? Will some geographic areas feel the housing impacts more acutely than others, and why? In fact, geography is another way to think about the first part of this volume: We begin with a national level analysis, then narrow in to examine a sample of metropolitan areas, and then look intensively at one city and its surrounding region.

Part 2 reviews the policies and programs that the housing sector has tried in the past to improve the economic independence of housing assistance recipients. A thorough description of these efforts is followed by a critical assessment from each of two perspectives: their structure and implementation, using the lessons learned from welfare–to–work demonstration programs as the standard; and the policy value of research on these policies and programs more generally. The goal of both critiques is to assess whether any useful information can be salvaged from existing programs, and, perhaps more importantly, to suggest improvements in the design, implementation, and analysis of future programs.

Part 3 examines welfare reform from the housing program practitioner's vantage point. One source of concern is the incompatibility or inconsistency of welfare reform and housing policies and programs. The "mixed signals" to housing assistance recipients on AFDC noted earlier in this first chapter are causing problems for both recipients and for program practitioners. Several examples of such incompatibilities are reviewed, as are other tensions between the "new" welfare thrust and the "old" housing rules. Where the base of experience and research evidence is strong, suggestions are offered for addressing these tensions. Where it is not, a practitioner–oriented research agenda emerges that is largely similar to the policy researcher's agenda implicit in the first two sections of this volume.

Part 1
Chapter 2

The implications of the new work requirements, work incentives, and time limits on the work effort, earnings, and housing cost burden of welfare recipients in assisted housing are the subject of chapter 2. Since federal welfare reform only became law in 1996, and states are still at varying stages of implementing their new programs to comply with these federal requirements, Sandra J. Newman and Joseph Harkness used evidence from states that instituted welfare reform waiver programs similar to Temporary Assistance for Needy Families (TANF) (the component of the welfare reform law that eliminates the AFDC system) to stimulate the likely effects of TANF on housing assistance recipients. Four such programs were chosen on the basis of two criteria. First, the programs had to reflect a range of stringency with regard to the key elements of welfare reform—time limits, work requirements, and sanctions. Second, there had to be some impact data available. The states chosen are Florida and Virginia at the stringent end of the spectrum, Michigan in the middle, and Vermont at the liberal end. These waiver–based estimates provide the first quantitative evidence on how welfare reform is likely to affect recipients of housing assistance.

Assisted households turn out to be particularly vulnerable to welfare reform. More of them are not working, they have less outside income on average, and they have been on welfare for longer periods than nonassisted households. The potential impact of the cumulative duration on welfare could be particularly dramatic. If the waiver programs in Virginia and Florida had been in full operation in 1994, for example, the more than 75 percent of assisted households who had received AFDC for 24 months or longer at that time could have been dropped from the welfare rolls.

According to the simulation results, welfare reform is likely to lead to modest net increases in the proportions of households that are not exempt from work requirements who receive housing assistance under each waiver program. This increase could be as high as 14 percent (about 119,00 households nationwide) in Vermont, the most liberal waiver state in the sample. The proportion of assisted households working in unsubsidized jobs would also increase in all four states. But these simulated increases in employment vary widely, from 25 percent under Michigan's policy (which has no time limit on welfare receipt) to 151 percent under Virginia's. The proportion of assisted cases exempt from work requirements would shrink under all waivers, from about half of all cases prewaiver to between one–third and one–sixth post–waiver. The simulated effects on the incomes of assisted households also vary widely. Under the Michigan and Vermont waivers, average income would rise by 2–4 percent, reflecting work incentives that are somewhat more enticing than those under the old AFDC program and no time limits on assistance. Under the Florida and Virginia waivers, in contrast, average incomes for assisted renters would fall (by 8 and 21 percent, respectively), indicating that gains in income for those in subsidized employment while on welfare will be more than offset by losses in income for those who do not find such jobs and have their benefits reduced by sanctions or eliminated due to hitting the time limit.

The estimated impact on the HUD budget for assisted housing indicated by these simulations varies with the overall income impacts. But the simulated savings under the Michigan and Vermont reforms (less than one–half of one percent of HUD's budget) are not enough to offset the 1.2 percent budget increase associated with Florida– and Virginia–type policies. Finally, none of the waiver programs has a large average effect on housing burdens. For those hitting time limits, however, the simulated national impacts are substantial. The worst cases are under the Virginia waiver program. The one–third of assisted renters hitting the time limit and not working will experience more than a 50 percent decline in income, triggering HUD expenditures of over $400 million. For the one–fifth of unassisted recipients hitting the time limit, housing cost burdens will rise by more than 50 percent, almost certainly necessitating a move to lower quality dwellings.

Chapter 3

A three–pronged thesis undergirds the chapter by G. Thomas Kingsley and Peter Tatian: The response of the assisted housing sector to federal welfare reform—if it is to be maximally effective—should be strongly influenced by the degree of overlap in the two recipient populations in a specific area, the importance of HUD presence in an area, and the predominant form that presence takes. If urban areas differ substantially along these dimensions, in other words, no "one size fits all" approach makes much sense.

Their analysis of a major new HUD data set, A Picture of Subsidized Households (APSH), takes a first step in providing the relevant data for assessing such differences. The authors focus on the 100 largest metropolitan areas, which together account for 58 percent of the U.S. population, 60 percent of all assisted households in public housing and publicly assisted housing, and 53 percent of those receiving tenant–based assistance. The authors include only areas for which APSH response rates are at least 60 percent, a limitation that forces them to drop some major metropolises from at least parts of their analysis.

Wide variations are revealed in patterns of housing assistance and in the welfare/housing assistance overlap. A major reason is that some cities have been much more aggressive than others in seeking HUD assistance. In addition, HUD's allocation procedures have enabled them to reallocate funds initially designed for some areas to other areas with more projects ready in the pipeline. Among these largest metropolitan areas, the proportion of low–income renter households receiving housing assistance varies from a low of 8 percent to a high of 50 percent. The share of assisted households ranges from 0 to 60 percent in public housing, 0 to 74 percent in publicly assisted housing, and 0 to 97 percent for tenant–based assistance. The share of HUD–assisted households whose dominant source of income is welfare ranges from 11 percent to 36 percent overall, and reaches a high 44 percent for tenant–based assistance and 50 percent for public housing.

As for how initiatives to help these "overlap" households move toward self–sufficiency should be tailored to different type of geographic concentration, the authors have several general points to make. Obviously, opportunities for community–building approaches and intensive support services are greatest within the project environment. They could also work well in publicly assisted housing if the prior required step of motivating project owners to cooperate is successful. In a metropolis where most housing assistance is tenant–based, more individualized self–sufficiency initiatives will be required. Where a large proportion of local welfare recipients are HUD–assisted, partnerships with other local services agencies should be central to the overall response to welfare reform. But where housing assistance is minor and the economy weak, counteracting the danger of major increases in homelessness because of welfare reform may need strong interventions in the supply side of the local housing market. In areas with high unemployment and few new jobs, public employment strategies may be needed before any linkages with other local agencies can make a difference.

Chapter 4

Major mismatches between the areas where the jobs are and the areas where low–income city residents live have been much discussed as a cause of joblessness in the inner city. Whether the new welfare emphasis on moving recipients out and into the regular job market is successful will depend in good part on availability of jobs within commuting distance of the neighborhoods in which they live. Claudia Coulton, Laura Leete, and Neil Bania examine how welfare recipients' locations match low–skill job openings and availability of affordable rental housing for the City of Cleveland—using county welfare program records for recipient residence information, decennial census data on occupational characteristics, Ohio Bureau of Employment Services data on job openings, published bus schedules and estimated automobile commute times from the Northeast Ohio Area Coordinating Committee, and decennial census data on affordable rental housing availability in the Cleveland/Akron labor market.

Welfare recipients in Cleveland, according to this detailed analysis, will not have an easy time finding jobs they can get to. First, they typically live in the inner city. The inner city is the only major source of low–wage jobs within striking distance of where they live. But a large proportion of welfare recipients can only find jobs there by displacing existing workers in the city. All other major jobs sources are long automobile commutes from the city and not well served by public transportation. Overcoming the current public transit problems looks difficult, since half of the forecast job openings are served by transit authorities other than the one that serves inner–city Cleveland residents. The likelihood of moving to where the jobs are is not high, because housing affordable by low–wage workers is hard to come by outside the center city. Even if welfare recipients do move to the suburbs where the jobs are, transportation problems will remain, because public transit within those areas is much less frequent and pervasive than in the inner city.

Part 2

In the past 20 years, there have been numerous studies of the impact of programs to move low–income individuals into jobs and welfare recipients into work, but far less attention has been paid to the connection between housing assistance and self–sufficiency.

Chapter 5

Amy Bogdon describes the design, implementation, and findings of three major HUD programs that have used housing assistance to encourage recipients to become economically self–sufficient: Project Self–Sufficiency, Operation Bootstrap, and Family Self–Sufficiency.

Project Self–Sufficiency (PSS) was designed to encourage communities to develop mechanisms for integrating public and private support services for the participants. HUD provided participating housing authorities with a special allocation of Section 8 certificates but no additional funding for support services, making the involvement of the community a key design feature. HUD required the programs to follow the same basic model, but communities were given substantial flexibility within that model. A total of 154 communities were selected to participate in the PSS demonstration, with selection based on public housing authority (PHA) diversity, past PHA performance, and program resources and planning. The target group was low–income single–parent households from the Section 8 waiting lists, with programs encouraged to select motivated participants.

Operation Bootstrap was very similar to Project Self–Sufficiency, the major difference being inclusion of two–parent families, a requirement that participants develop individual action plans, and even more local flexibility under Bootstrap. Almost 12,000 Section 8 certificates were awarded to 353 communities. The Family Self–Sufficiency program (FSS) was similar in some respects to the other two initiatives but had different PHA selection criteria, participant eligibility, incentives, and sanctions. In the first two program years, PHAs were selected through an award process. In the third year, participation became mandatory for all PHAs receiving incremental Section 8 or public housing units. Screening for motivation was still permitted, but the participation of individual families was limited to current holders of Section 8 certificates or vouchers or current residents of public housing. For the first time, participants faced sanctions for failure to comply with their contract of participation, and participants deposited rent increases resulting from increased earnings into an escrow savings account for later use in purchasing a house. As Bogdon notes, none of the three followed the type of experimental design necessary to assign causation. Participants were not randomly selected, most sites screened out less motivated participants, and some even used previous educational attainment or job experience as a screen. Design and implementation flexibility across sites also prevented outcomes from being compared across sites.

Even so, Bogdon argues, the outcomes tell us some things that are useful. People did volunteer for self–sufficiency programs and are making progress toward self–sufficiency. Receipt of Section 8 certificates did improve housing outcomes for many participants, and local programs have been reasonably successful in coordinating a wide range of services to assist participants. However, few participants have become completely independent of all types of assistance; there have been difficulties in recruiting participants, particularly for programs with sanctions; and dropout rates are considerable. Bogdon concludes that understanding the potential role of housing assistance in the process of attaining self–sufficiency requires that future programs be designed so that their impacts, and not just their outcomes, can be measured. She also recommends improving program standardization, introducing experimental or quasi–experimental designs, and improving data collection.

The next two chapters are considerably more pessimistic about whether anything can be learned about the impact of housing programs designed to promote self–sufficiency from these HUD initiatives, because there is no way to measure the counterfactual—what would have happened in the absence of those programs.

Chapter 6

Starting from his view that the programs Bogdon describes cannot tell us about program effectiveness, James Riccio reviews what the large–scale social experiments that have evaluated welfare–to–work and job training programs, including Jobs Training Partnership Act (JTPA) programs, can tell us about whether at least the assumptions underlying the housing/self–sufficiency programs are plausible. He then describes an ambitious new employment initiative for public housing residents, the Jobs–Plus Initiative, that will be subject to rigorous evaluation.

Two assumptions are shared by all three of the housing/self–sufficiency programs described in chapter 5: (1) that support services should give priority to education and training activities before encouraging participants to enter the labor market, and (2) that program effectiveness is enhanced by choosing participants selectively. An additional assumption of the third demonstration, FSS, is that stronger financial incentives for residents to work help achieve that goal, and that an escrow savings plan is an effective incentive. The overall message from the rigorous evaluation literature is that some welfare–to–work strategies have been effective (and cost–effective) in increasing employment and earnings and reducing welfare payments. But these gains are modest, none of the programs has had much success in reducing poverty, and the problem of high job turnover has not been resolved.

With respect to the three particular assumptions Riccio highlights, he has the following responses: (1) The housing programs' approach to selecting participants who show particular motivation may not be the most effective strategy. It is at least as plausible that recipients who do not show individual initiative are the ones that self–sufficiency programs must try to reach if they are to be effective in moving large numbers of residents into work. (2) Evidence from the welfare–to–work and JTPA evaluations indicates that putting up–front job search and work–based strategies ahead of education and training in the program participation sequence, and combining them with substantial resources into job development efforts, are more effective in moving participants into the job market than those that stress more general development of human capital. (3) There is some evidence that financial incentives can increase earnings when combined with strong work requirements. But diverting rent increases into escrow accounts may be too distant, and too restrictive, a payoff to be an effective inducement to change behavior.

The Jobs–Plus Initiative Riccio goes on to describe is an ambitious undertaking, designed to build on the successes of past welfare–to–work initiatives within the public housing context and within a rigorous evaluation design. In seven cities, eight public housing developments have been selected to form new partnerships—collaboratives—of residents, welfare departments, JTPA and other employment agencies, and private–sector employers. The evaluators and a team of consultants are assisting the PHAs in building their collaboratives. In addition, the strategies the Jobs–Plus communities must follow include work incentives and requirements built into the PHA rent and welfare rules, best practices in preparing residents for sustained employment and linking them with jobs, and enhanced community support for work (through promoting social networks that support work both inside and outside the development). The target group encompasses all working–age residents in the PHA hosting the program—providing an opportunity to test whether a comprehensive, intensive, place–based employment intervention can transform housing developments into communities of workers and, in turn, improve their quality of life.

Since place–based interventions cannot be studied with a random assignment design, the initiative builds in several other methodological approaches. First, the sites' implementation strategies and experiences will be a major focus of the research so that the interventions will be rigorously delineated. Second, groups of similar housing developments were identified. One of these was randomly selected to be a Jobs–Plus host and the other(s) served as comparison sites. Data will include administrative records and survey information spanning five years before and five years after the start of Jobs–Plus. As Riccio puts it, "Dramatic change requires bolder interventions that can draw only partial guidance from past research. Jobs–Plus will carefully test an employment initiative created by committed local stakeholders (with guidance from national partners). And it will provide a scope and intensity of assistance to public housing residents that is unprecedented. Success is not guaranteed, of course, but the stakes are too high not to make bold efforts to outperform programs of the past."

Chapter 7

Craig Thornton, Robert G. Wood, and Pamela M. Jones discuss how the policy value of demonstration research on housing and self–sufficiency can be increased. On the issue of whether anything can be salvaged from the three demonstrations described by Bogdon, they conclude that "the impact estimation methods and supporting data cannot provide evidence that would persuade skeptics about the value of housing–based self–sufficiency programs." They follow this verdict with a discussion of two housing–related self–sufficiency studies—Jobs–Plus, as discussed in chapter 6, and Moving to Opportunity (MTO).

They agree with Riccio that the Jobs–Plus evaluation design was innovative and a great step forward from previous housing/self–sufficiency studies. Their main concern is that the broad program design flexibility given to each community may produce such different programs that evaluators will not be able to pool the impact results across sites. If so, the power of the evaluation will be substantially reduced.

The MTO seeks to improve economic outcomes for public housing residents by offering them Section 8 certificates on condition that they use them to obtain private housing in low–poverty neighborhoods. The program also provides housing search assistance and counseling. The evaluation involves five cities and will collect 10 years of data. It also includes random assignment of MTO applicants to (1) a primary treatment group receiving the full MTO treatment, (2) a secondary treatment group receiving Section 8 certificates that can be used anywhere but no housing search or counseling services, and (3) a control group not receiving a certificate or services.

This research design is powerful and will provide clear and reliable results on the effects of the housing initiative being tested. But it has two limitations. First, it can only provide information on a voluntary program. Second, to the extent that the programs differ across cities, analysts may have to treat each as a separate program, which will reduce effective sample size. Differences across sites may be useful in helping identify which parts of the treatment caused which impacts, however. This concern can also be addressed by comparing outcomes for the primary treatment group with outcomes for the secondary treatment group.

Part 3
Chapter 8

The housing practitioners who staff the more than 3,000 local PHAs view welfare reform and its impacts on housing assistance from the unique vantage point of implementation. Newman's last chapter looks at their views for the hypotheses they generate about the role of assisted housing in achieving economic independence. She finds practitioners concerned about a wide variety of disincentives in current housing policy with respect to fostering self–sufficiency.

The first of these—discussed in earlier chapters from the analysts' perspective—is the disincentive to work provided by the sliding scale rent calculations for housing assistance recipients. This implicit tax on earnings, particularly since housing assistance is not an entitlement and has long waiting lists, constitutes a powerful work disincentive. It can also be expected to discourage tenants from reporting income, as required by housing program regulations. There is no systematic information on either receipt of underground income or the extent to which it is investigated by PHAs. But administrators quote substantial evidence that tenants often reduce work as they approach the six–month limit in the certificate and voucher programs. And PHAs who investigate fraud allegations report large numbers of cases of unreported earnings. Many PHAs do not pursue fraud, however, because the investigative costs are likely to be higher than the back rent to be collected from the effort.

One strategy for encouraging recipients to work, and to report earnings, is the earned income disregard. Such disregards have been required in housing programs since 1990. But few PHAs have implemented them. PHAs may be unwilling to implement such a policy because they risk the loss of operating subsidies—calculated prospectively as estimated revenues minus estimated costs over the next year. If only earned income about the level used to project rents is disregarded, the PHAs are not at risk. But if all earnings are to be disregarded—which is the case for tenants participating in any HUD–funded or JTPA employment/training program—operating subsidies are affected, at least in the short run. For Section 8 certificate and voucher programs, operating subsidies are not relevant. The risk for them is that the subsidy slots vacated as families earn their way out of the program will not be replaced and the total subsidies available will, therefore, shrink. PHAs may also be deterred from implementing earned disregards because they would no longer be able to recapture any income increases through increased rent.

Because HUD has been almost continuously in the throes of reorganization, reinvention, and downsizing throughout the 1990s, it would not be surprising if the many PHAs who have not implemented the earned income disregard requirement have been betting that HUD would not detect such violations. It is also very likely—since the relevant regulations were not published until four years after the statute was passed, and HUD has yet to train PHAs in how to interpret and implement them—that PHAs chose not to implement because of confusion about the interpretation of the requirement. The historical record of many features of housing assistance is replete with such examples of unclear regulations—including core elements such as how rents are set, how income is calculated, and what percentage of the income is to be paid as rent. Such confusion suggests that there is probably much more variation in the implementation of housing programs across the country than is commonly recognized—with attendant complications for analysts trying to isolate the impact of housing assistance on economic independence.

How would housing practitioners fix housing programs? Their consistent answer is that government should greatly increase the PHAs' flexibility to run their programs as they see fit. The Moving–to–Work demonstration (MTW), launched in 1997, is doing just that by allowing participating PHAs to combine their different streams of HUD subsidies for use in any way they think will best encourage tenants to work. MTW provides no additional monies but allows PHAs to keep any savings they can generate. Unfortunately, research goals played no role in MTW site designs or implementation decisions, drastically limiting what can be learned from the effort about impacts on self–sufficiency generally or about the relative promise of different strategies.

Although this volume takes an important first step in presenting new analytical and critical thinking about the implications of welfare reform for housing, it goes without saying that it is only a first step, and much, much more needs to be done. For example, we need to understand the responses of housing assistance recipients to welfare reform using a dynamic, behavioral model, not a static cross–sectional model. We also need to learn whether assisted housing residents respond differently to welfare reform from those who do not receive housing assistance, and the causes, nature, and magnitude of those differential responses. In addition, because housing policy is also undergoing major changes, we need to carefully analyze the interactions between these changes and welfare reform. And to be of assistance to federal policy development, we need to generate solid national–level impact estimates, albeit reflecting important local variations in both housing and welfare administration and implementation.

But as valuable as such research would be to the particularly timely issue of welfare reform and its effects on housing, it still leaves the most fundamental questions about the impacts of housing, such as those noted earlier, largely unanswered. We have the research tools to study these questions. And as the final chapter in this volume makes clear, practitioners and researcher agree on the questions. But support for housing research of any kind is modest, and for research of the kind suggested here, nearly nonexistent. Unless such support is forthcoming, we will continue to lament our perpetual inability to answer fundamental questions about housing, and our exclusion from the "big debates" about social policy. Our state of knowledge about whether housing is a route to independence will continue to be, as the title of a paper I wrote 10 years ago put it, "A Case That Has Yet to Be Made."

1. Throughout this book, "housing assistance" refers to federal housing subsidy programs for the poor. Included here are public housing, private developments that receive federal subsidies through either the U.S. Department of Housing and Urban Development or the Internal Revenue Service in the form of low–income housing tax credits, and certificates and vouchers that allow recipients to rent modest, affordable apartments in the private rental market and to pay no more than 30 percent of income for rent.

2. All estimates are for FY1996 to capture pre–welfare reform levels. The housing assistance estimate includes subsidies for rental housing ($25 billion) plus the revenue loss associated with the low–income housing tax credit ($2.6 billion) (U.S. House of Representatives 1998; Budget of the U.S. Government, FY1998, 1997; Pedone 1998). AFDC numbers reflect the federal and state share of benefit outlays (administrative costs are excluded). (Numbers expressed in 1996 dollars.)

3. The universal scope of housing policy will always distinguish housing assistance from other safety net programs.

4. Figures range from 53 percent in the Panel Study of Income Dynamics–Assisted Housing Database for 1993 to 61 percent from Burke (1995).

5. Author's tabulations from the 1981 and 1995 national American Housing Surveys.

6. Figures range from 45 percent in the Panel Study of Income Dynamics–Assisted Housing Database for 1993 to 48 percent from Burke (1995).

7. This section draws heavily from Newman and Schnare (1994) and Newman and Harkness (1998).

8. As first stated by Kain (1968), this hypothesis referred to black central city residents. More recent work has extended it to low–skilled white workers (Ihlanfeldt and Sjoquist 1989).

9. For example, several key items on Mayer's physical quality index were collected through interviewer observations, which often have reliability problems.

10. High–poverty tracts are those with poverty rates of at least 40 percent. Low–poverty tracts are those with poverty rates of less than 10 percent.

11. Because children in the MTO experimental groups will have moved not only from high– to low–poverty areas but also from public housing to private rental housing with a Section 8 certificate, it may be difficult to disentangle the effects of the change in neighborhood, the change in housing, and the move itself.

12. This section is based on Newman and Harkness (1998).

13. This section is based on Newman and Schnare (1994).

14. Public housing authorities are quasi–governmental units vested with the responsibility for managing federal housing subsidies for public housing and for Section 8 certificates and vouchers.

15. This section is based on Newman and Schnare (1994).