Reexamining the Moving to Opportunity Study and its Contribution to Changing the Distribution of Poverty and Ethnic Concentration

Abstract

For the past decade and a half, a concerted effort has been undertaken to determine whether policy interventions in residential location can solve the problems of inner-city poverty and racial concentration. Studies based on data from the Gautreaux litigation and the U.S. Department of Housing and Urban Development (HUD)–sponsored Moving to Opportunity (MTO) program have provided an overall optimistic interpretation of the possibilities of improving inner-city lives via mobility vouchers and counseling. A reanalysis of the data from the MTO program, focusing specifically on African American households, suggests greater caution in the interpretation of the findings from either Gautreaux or the MTO program. No statistically significant difference exists between the percentage of poor or the percentage of African Americans in the current neighborhoods between MTO and Section 8 experimental groups. In some cases, there is no statistically significant difference between those who move with a voucher and those who move without any assistance at all. Although there is some evidence that MTO programs have brought specific gains for individual families, claims for the MTO program as a whole need to be treated with a great deal more caution than they have been to date.

The half-century-long U.S. concern with inner-city poverty and the inner-city concentration of minority populations now focuses on help through individual assistance rather than the construction of either project-based housing or scattered site housing, although sizeable amounts of both still exist in most metropolitan areas. The shift to individual assistance, usually through some form of housing voucher, reflects the increasing concern of the U.S. government to refocus attention on individuals rather than on government intervention via housing demolition and construction. This shift reflects a fundamental change in social thinking on how best to combat poverty and help underprivileged populations.

Within the context of concerns over the potential for a growing urban underclass and the associated concerns with growing poverty concentrations (although the percentage of those in poverty actually declined in the 1990s), interest has increased as to whether inner-city minority households can escape impoverished neighborhoods and whether government assistance can increase those probabilities. In the 1990s, a tentative consensus emerged that enabling low-income families to move from high- to low-poverty neighborhoods had the potential to reduce the levels of income segregation and, as a corollary, the degree of racial separation. The specific rationale for voucher-based programs for poor, inner-city households was to increase access to neighborhoods that would enhance employment and educational opportunities and diminish exposure to crime, violence, and drugs. Certainly, some commentators suggested that these programs benefit individual families and have the potential to deconcentrate poverty (Briggs 1998, 2005). Although the program was not specifically designed to integrate minority populations, the implication of moving to lower-poverty neighborhoods would have gains of living in more racially mixed neighborhoods as well.1

A specific voucher policy designed to reduce economic residential segregation might also have the effect of increasing opportunities for labor market success. Some research suggested that relocation to suburban area would increase job opportunities for low-income populations, but other studies found high unemployment rates for suburban movers compared with city movers (Rosenbaum 1995). Still others have questioned the role of housing vouchers altogether. Grigsby and Bourassa (2004) argued that the housing choice voucher program is no longer effective as a mechanism for housing assistance. They noted that housing quality has improved substantially and that now only a very small proportion of all the housing stock in the United States is severely inadequate. They concluded that the housing choice voucher program is little more than an income subsidy and should be merged into other aspects of the U.S. federal social safety net.

These questions about the role of housing choice vouchers are at the heart of this paper. There have been other critiques of the Moving to Opportunity (MTO) demonstration program, but this paper examines whether there are overall program gains from the MTO program versus the Section 8 voucher program, or even from no intervention at all. The focus of the paper is the evaluation of a federal program intervention in the mobility process and the outcomes in poverty levels and levels of separation in the cities in the MTO study. This paper also takes a specific spatial focus, which is lacking in other studies of the MTO demonstration program. I do not argue against the view that some families may have benefited from receiving vouchers and mobility counseling. However, I do argue that the advantages of the policy have been emphasized at the expense of a more balanced analysis of the strengths and weaknesses of voucher-based interventions in the residential mobility process.

Specifically, I ask whether the distributions of African American households who move with one kind of assistance versus another (MTO vouchers with counseling, or Section 8 vouchers) are different in the kinds of poverty neighborhoods they enter and the levels of racial mixing they experience initially and in the longer term. I specifically examine whether the proportion of movers who live in lower-poverty and more racially mixed areas is sustained over time. A central focus of the research is to contrast what is sometimes called the “intent-to-treat” sample (all persons in the study) with the “treated” sample (those who participated). The tests in the paper examine whether the distributions of those with vouchers and counseling (treated) are different from those in the baseline sample (intent-to-treat sample), who did not receive assistance. I use the subsample of African American households in this study as the group that is most often targeted with programs to alleviate poverty and to offer integrative opportunities, as argued in several U.S. federal court cases on housing availability. I use Kolmogorov-Smirov (K-S) tests of the difference of distributions to test for significant differences.

PREVIOUS RESEARCH ON RELOCATING POOR PEOPLE AND THE ROLE OF VOUCHERS

The growing emphasis on geographically dispersing housing-subsidy recipients is based on the assumption that residence in concentrated poverty neighborhoods abets socially dysfunctional behavior—or, more simply, that high-poverty households will fare better outside of poverty neighborhoods (Galster and Zobel 1998). Although there are a number of individual studies of voucher use, the main body of analysis has grown up around data sets that emerged as part of the Gautreaux litigation in Chicago, the Holman litigation in Minneapolis, and from the MTO program. The conclusions are by no means consistent. Although some see positive effects from moving families from poor neighborhoods to less-poor and sometimes suburban neighborhoods (Briggs 2005; Goering 2005; Johnson, Ladd, and Ludwig 2002), others question whether the programs can deliver substantial gains in dispersing poverty (Clark 2005; Varady 2005). In a recent paper, Galster (2007) provided an important distinction between a focus on whether the poor benefit, or whether there is an aggregated societal gain. It is whether there is societal gain in the aggregate that is the major concern of this analysis. Considerable debate is ongoing about “voucher intervention” in the residential fabric, including policy arguments in favor of enlarging the current MTO project, making this analysis especially relevant.

As part of the Gautreaux housing litigation in Chicago, a selection of inner-city households living in Chicago Housing Authority public housing was provided opportunities to move to neighborhoods of a low percentage African American composition and to suburban locations. The Gautreaux research did not specifically disentangle poverty and race effects, but still it did provide some evidence that those who moved to suburban communities were more likely to be employed (although their salaries were not necessarily higher compared with those of movers within the city; Rosenbaum and Popkin 1991), and that suburban youth did better on several educational measures (Rosenbaum 1995). At the same time, the Gautreaux findings were criticized for selection bias (the 7,000 participants were a small proportion of all applicants to the program) and for a general focus on suburban movers and less attention to movers within the city. Movers in the Gautreaux program were given extensive mobility counseling and assistance, but this is not the same as the MTO demonstration program because there was no comparable baseline group in the program. Overall, however, the evidence supports the view that participating tenants do gain from the dispersed moves; these gains may come not from the lower concentration of poverty per se, but rather from the “structural advantages of the suburban areas, such as schools, public services, and job accessibility” (Galster and Zobel 1998:615).

A recent paper (Keels et al. 2005) selected data on 1,506 cases, which is little more than one-half the sample of 3,000 Gautreaux participants who were in the program prior to 1990. However, the analysis focused on 1,171 families who were in Illinois. This paper, which examined the outcomes for these participants, is a recent example of the positive interpretation of the Gautreaux data; it reported that, “helping families relocate into communities that are racially/ethnically integrated . . . appears beneficial in both the short and the long run” (p. 71). The study made a particular point of the stability of the movers and noted that they do not move back to their old neighborhoods. The data include 1,171 cases: 574 cases in the city, and 597 cases in the suburbs. The study reported that 57% of suburban placements remained in the suburbs, although nearly 30% moved back to the city. For city placements, 78% remained in the city, and 12% moved to the suburbs.

The conclusion that the Gautreaux program was successful reiterates the basic motivation of this reanalysis of the MTO data. How does one decide about whether to focus on “program” successes against “individual outcome” successes? A tendency in the research literature on the Gautreaux program has been to emphasize the positive of the program as a whole, when in fact it may better be argued as limited successes for particular participants. One aim of the Gautreaux program (at least as emerges in much of the literature) was to move households to the suburbs (which was not a requirement of the program). In this evaluation, the program was halfway successful. If, as one reviewer argued, the outcome was related to how vouchers were distributed, this is an important caveat on evaluating the program contribution, albeit beyond the scope of this review. With respect to levels of integration, families are still living in concentrated African American areas after moves, as the authors point out (Keels et al. 2005). However, the areas were slightly less populated with African Americans after the move. The origin neighborhoods in the city were 9.8% white for city movers (movers who remained in the city) and 15.0% white for suburban movers. Their neighborhoods at the time of reanalysis in 1998 were 24.1% white for city movers and 40.5% white for suburban movers (Table 1). Clearly, suburban movers are more likely to live in less–African American neighborhoods but still in majority–African American neighborhoods. What the analysis does not address is that, again, city dynamics enter the picture. The initial gains of greater integration were cut in half by the end of the analysis period, from 84% white to 40% white. It is true that the current neighborhoods are a mix of white, African American, and other minorities (41% white, 38% African American, and 21% other). This division does not signal failure for individuals, but it shows that the program outcomes are more subtle and less clear-cut than the sometimes overly optimistic interpretations of those who seem to want to empower the various voucher programs.

Table 1.

Outcomes for Gautreaux Movers, by Origin and Current Locations

	City Movers (N = 574)	Suburban Movers (N = 597)
Origin Location
Percentage African American	83.7	81.8
Percentage minority	90.2	85.0
Percentage white	9.8	15.0
Placement Location
Percentage African American	47.2	6.5
Percentage minority	67.6	15.6
Percentage white	32.4	84.4
Current Location
Percentage African American	57.2	38.4
Percentage minority	76.9	59.5
Percentage white	24.1	40.5

Source: Data are from Keels et al. (2005).

The results of an attempt to redistribute low-income public-housing residents in Minneapolis also provided mixed findings on the ability to successfully relocate households via vouchers. As part of a consent decree in Minneapolis (Holman v. Cisneros 1995), a large public housing complex in the inner city of Minneapolis was demolished, and the residents were provided with relocation assistance. Not all residents of the projects that were demolished were willing participants in the relocation project. Southeast Asian households were much more resistant to forced relocation than were African American households, and most wanted to stay in Minneapolis (Goetz 2003:203). Those who indicated a desire to leave the city wanted, for the most part, to move only to the inner ring of suburbs directly north of the city: that is, close to where they lived before they moved. Preferences for familiar neighborhoods are especially strong for the project residents.

The largest study and a controlled investigation of the outcomes of mobility behavior is the MTO program, a program that was based on the reported gains from the Gautreaux relocation program. In 1992, with a mandate from Congress, the U.S. Department of Housing and Urban Development (HUD) initiated this experimental program as a method of testing whether providing vouchers and special counseling would improve the outcomes for households who moved from inner-city neighborhoods.2 The aim of the program was to find out “what happens when very poor families have the chance to move out of subsidized housing in the poorest neighborhoods of five very large American cities” (Orr et al. 2003:i). The program divided potential voucher holders into three groups. Those in the baseline group did not receive a voucher but could continue to live in public or assisted housing. Participants of the Section 8 group received a voucher and regular housing assistance counseling and could move wherever they could find a suitable unit. Those in the experimental group received a voucher and special mobility counseling, but participants were required to move to a low-poverty neighborhood (less than 10% poverty, according to the 1990 U.S. census).

The most recent comprehensive report from the study concluded that there are greater gains from living in lower-poverty neighborhoods and in more-integrated settings for the special program MTO movers than for regular Section 8 voucher movers or for those who received no assistance (Orr et al. 2003). This finding is supported in global reviews by Johnson et al. (2002), Goering (2005), and Briggs (2005). The Orr et al. (2003) study tested regular Section 8 movers and MTO special program movers against the baseline group and showed that both groups were significantly different from the baseline participants. The tests were run separately for those who were able to “lease up” and move, and for the total sample (the intent-to-treat sample). Tests for the latter were statistically significant at the .05 level. In addition to accessing lower-poverty neighborhoods and integrated neighborhoods, the study examined outcomes for education, health, employment, housing, and changes in criminal/behavioral problems.3

The findings of the Orr et al. report (2003) have been questioned, and others (Kling et al. 2004; Varady 2005) have queried the positive conclusions of the MTO findings. The tests in the Orr report may be insufficient, on at least two grounds, to decide that the MTO special program achieved significantly better outcomes than the program using regular Section 8 vouchers. First, the Orr report did not conduct two important tests: a test of the difference between MTO and Section 8 voucher recipients, and a test of city-specific differences. The Orr study compared MTO special voucher and Section 8 each to the baseline sample, concluding that MTO has a larger impact and is, therefore, “better” than Section 8. There has not been a direct test of the difference between MTO vouchers and Section 8 vouchers. Second, and equally problematic, is that the data for all five cities were aggregated, thereby masking geographic impacts and averaging out differences across cities. One cannot know from the analysis whether MTO gains are city-specific or general and whether the positive gains in one city are weighting the aggregate outcomes positively.

Goering (2005) suggested that a “well designed extension of MTO could offer opportunities to thousands of additional low income and public housing families” (p. 145). Alternatively, Varady (2005) was not convinced that voucher programs are the answer to concentrated poverty or racial segregation because of both the strong desire of households to move nearby (many of the poor do not want to move away from friends and relatives) and the many involuntary moves that are part of the mobility process. However, in a series of papers (Varady and Walker 2000,Varady and Walker 2003), Varady suggested that moving to the suburbs (in one, a case study of Oakland, California) leads to improvements in housing conditions. Given the range of opinions and different outcomes, the current paper reexamines the MTO outcomes on a city-specific basis and examines the issue of program rather than individual effects.

REEVALUATING VOUCHERS AS A TOOL FOR DISPERSING POVERTY AND INTEGRATING NEIGHBORHOODS

The data for the MTO study were drawn from five cities: Baltimore, Boston, Chicago, Los Angeles, and New York.4 The total sample was 4,610 families, divided into 1,440 baseline cases (households who did not receive a voucher), 1,350 families who were offered vouchers, and 1,820 families who were given vouchers and special counseling. The data analyzed in this paper are for the subset of 2,298 African American families in the sample. There were 534 cases for Baltimore, 282 cases for Boston, 782 cases for Chicago, 302 cases for Los Angeles, and 398 cases for New York. (The sample sizes for each subset are given in the relevant tables.) There is a small difference between the total of the cities analyzed in this study and the aggregate reported for all five cities. This small difference is generated by a very small number of cases for which it was not possible to identify the tracts to which households moved.

The analysis used the 2000 census poverty level and percentage African Americans as the context for analyzing locational changes of the sample.5 The “lease up” dates—the terminology for participation in the project—were between 1994 and 1997. The current locations were evaluated as of 2002. The current tracts for the sample were used to evaluate the current outcome in percent poverty and racial composition. The original locations were concentrated in the central tracts of the city (where the public housing projects are located), and the mapping from these core areas gives a good indication of the “relocation” of the samples over time.6 It is important to recall a major focus of this study, a test that has not been undertaken before: determining whether a difference exists for outcomes between MTO and Section 8 voucher holders.

An Empirical Analysis of Vouchers as a Tool for Dispersing Poverty

Experimental movers (MTO movers) in all five cities were more likely to be in lower-poverty areas than Section 8 movers in their first lease up (Table 2). This also shows up visually in Panel b of the maps (Figures 1–3) for New York, Chicago, and especially Los Angeles.7 Of course, this is to be expected because the program required leasing in a low-poverty neighborhood. Although nearly all experimental movers chose neighborhoods that were less than 20% poverty neighborhoods (2000 census measures), only 22% of the regular Section 8 movers did so. The results are far less compelling when testing the results for the total sample of movers and nonmovers (Table 3). Table 3 shows that for some cities, the distributions are different; but for others, there are no differences in the patterns.

Table 2.

Percentage of MTO and Section 8 African American Respondents by Their Original Move Location and in Their Current Location, and by Poverty Composition of the Neighborhood

Poverty Composition	Original Move		Current Location
	MTO Mover	Section 8 Mover	MTO Mover	Section 8 Mover
Baltimore
0%–10%	35.29	4.46	19.64	9.17
10%–20%	63.03	17.86	36.61	19.27
20%–30%	0.84	22.32	14.29	26.61
30%–40%	0.84	39.29	16.07	23.85
40%–50%		14.29	8.04	13.76
50%–60%		1.79	5.36	6.42
60%–70%				0.92
70%–80%
80%–90%
N	119	112	112	109
Boston
0%–10%	65.31	3.23	29.17	10.00
10%–20%	30.61	48.39	37.50	33.33
20%–30%	4.08	32.26	22.92	30.00
30%–40%		16.13	8.33	23.33
40%–50%			2.08	3.33
50%–60%
60%–70%
70%–80%
80%–90%
N	49	31	49	31
Chicago
0%–10%	32.31	1.79	21.60	4.55
10%–20%	56.92	16.96	39.20	20.91
20%–30%	8.46	21.43	18.40	25.45
30%–40%	1.54	23.21	8.80	20.91
40%–50%	0.77	18.75	8.00	11.82
50%–60%		10.71	4.00	8.18
60%–70%		7.14		6.36
70%–80%				0.91
80%–90%				0.91
N	130	112	130	112
Los Angeles
0%–10%	13.54	1.89	4.40	0
10%–20%	57.29	19.98	30.77	14.81
20%–30%	25.00	26.42	26.37	24.07
30%–40%		32.08	18.68	27.78
40%–50%	4.17	22.64	16.48	27.78
50%–60%			1.10	3.70
60%–70%			2.20	1.85
70%–80%
80%–90%
N	96	53	96	53
New York
0%–10%	20.78	2.94	12.50	3.33
10%–20%	68.83	19.12	42.19	18.33
20%–30%	9.09	22.06	17.19	21.67
30%–40%	1.30	22.06	10.94	25.00
40%–50%		29.41	10.94	30.00
50%–60%		4.41	4.69	1.67
60%–70%			1.56
70%–80%
80%–90%
N	77	68	77	68

Note: Some columns may not sum to 100.00 because of rounding.

Figure 1.

Current Locations of Households That Moved in New York: Regular Section 8 Moves, MTO Moves, and Control Sample Moves

Note: Although the initial origins are distributed across several central tracts in the city, for visualization purposes, they are shown as initiating from one central location.

Figure 3.

Current Locations of Households That Moved in Los Angeles: Regular Section 8 Moves, MTO Moves, and Control Sample Moves

Note: Although the initial origins are distributed across several central tracts in the city, for visualization purposes, they are shown as initiating from one central location.

Table 3.

Percentage of Total MTO, Total Section 8, and Baseline African American Respondents in Their Current Location, by Poverty Composition of the Neighborhood

% Poverty	MTO Movers and Nonmovers	Section 8 Movers and Nonmovers	Baseline Sample
Baltimore
0%–10%	12.24	6.49	3.21
10%–20%	25.00	21.43	17.31
20%–30%	13.78	24.03	19.23
30%–40%	17.86	20.13	16.03
40%–50%	13.78	12.99	13.46
50%–60%	15.82	14.29	26.92
60%–70%	1.53	0.65	3.85
70%–80%
80%–90%
90%–100%
N	196	154	156
Boston
0%–10%	15.74	4.62	2.08
10%–20%	23.15	24.62	15.63
20%–30%	23.15	20.00	16.67
30%–40%	33.33	41.54	46.88
40%–50%	4.63	6.15	14.58
50%–60%		3.08	4.17
60%–70%
70%–80%
80%–90%
90%–100%
N	108	65	96
Chicago
0%–10%	8.38	4.62	4.35
10%–20%	20.68	15.03	13.04
20%–30%	16.49	21.39	19.02
30%–40%	13.87	21.39	14.13
40%–50%	8.90	12.72	11.96
50%–60%	5.76	7.51	9.24
60%–70%	17.54	11.56	19.57
70%–80%	7.59	4.05	7.07
80%–90%	0	0.58	0
90%–100%	0.79	1.16	1.63
N	382	173	184
Los Angeles
0%–10%	3.48	0	0
10%–20%	25.22	13.33	4.39
20%–30%	21.74	21.67	8.77
30%–40%	20.00	25.00	15.79
40%–50%	20.87	33.33	28.07
50%–60%	6.09	3.33	0.88
60%–70%	2.61	3.33	23.68
70%–80%			18.42
80%–90%
90%–100%
N	115	60	114
New York
0%–10%	6.92	2.56	0
10%–20%	22.31	10.26	2.04
20%–30%	12.31	12.82	6.12
30%–40%	13.08	20.51	16.33
40%–50%	31.54	33.33	39.80
50%–60%	12.31	19.66	30.61
60%–70%	1.54	0.85	4.08
70%–80%			0
80%–90%			0
90%–100%			1.02
N	130	117	98

Notes: Nonmovers are those in households who were given vouchers and special counseling (MTO) and/or Section 8 vouchers but were not able to convert those vouchers to actual moves. Some columns may not sum to 100.00 because of rounding.

Source: MTO data for combined files for Baltimore, Boston, Chicago, Los Angeles, and New York; prepared by the HUD Office of Policy Development and Research.

I report the analysis in the following series of tests. For movers only, I test (1) original MTO moves with original Section 8 moves; (2) original MTO and current MTO locations (to see whether regression occurs—that is, to see whether the moves maintain their low-poverty gains); and (3) current MTO versus current Section 8 moves. For the total sample (intent-to-treat), movers and nonmovers, I test (1) the current MTO locations against current Section 8 locations; (2) current MTO locations against the current baseline (control group); and (3) Section 8 current locations against the current baseline group. I use Kolmogorov-Smirnov (K-S) two-sample tests at the .01 level and .05 levels, as shown in Table 4.8 All the values for the tests are shown in Appendix Tables A1 and A2.

Table 4.

Kolmogorov-Smirnov Two-Sample Tests of Differences Between Programs: Movers Only (treated) and the Total Sample (intent-to-treat) for Moves by Poverty Locations

City	Movers			Total Sample
	Original MTO Versus Original Section 8	Original MTO Versus Current MTO	Current MTO Versus Current Section 8	Current MTO Versus Current Section 8	Current MTO Versus Baseline	Current Section 8 Versus Baseline
Baltimore	**	**	**			*
Boston	**	**			**
Chicago	**	**	**
Los Angeles	**	**			**	**
New York	**	**	**		**	*

^*Significant at the .05 level.

^**Significant at the .01 level.

For all cities, original MTO locations are significantly more likely to be in low-poverty neighborhoods. For all cities, the changes over time are also significant; they have not been able to maintain the low-poverty locations. Differences between the current MTO and the current Section 8 locations are different for Baltimore, New York, and Chicago but not for Los Angeles and Boston. That is, there is no distinguishable gain for MTO over Section 8 for those who actually moved for two of the five cities. That three cities show greater gains in being in a lower-poverty neighborhood for experimental movers than for Section 8 movers is a finding that can be cited as evidence for the gains of the MTO program in specific cities. The tests here suggest that aggregating the data, as in the Orr study, hides important outcomes by specific geographies. Different cities have different outcomes.

Turning to the major concern of the paper—namely, measuring program effects (that is, when movers and nonmovers are aggregated and subjected to tests of difference)—for no city does a difference exist between current MTO and current Section 8 locations.9 This is direct evidence that MTO as a program does not deliver gains over regular Section 8 vouchers as a program. For three cities (Los Angeles, Boston, and New York), current MTO outcomes are better than the baseline outcomes. In Los Angeles and New York, both MTO and Section 8 participants fared better than the baseline. For the other cities, there is no significant difference. For only one city (Los Angeles) is the current Section 8 pattern an unequivocal gain over the baseline sample. Two other cities (Baltimore and New York) show gains at the .05 level. This outcome reiterates the dynamism of the city and the fact that the many baseline households who were not given vouchers still managed to improve their housing situations. In other words, the sample respondents—without help—have made gains in moving to low-poverty neighborhoods. It is not always true, but it is sufficiently prevalent to raise questions about the nature of the MTO intervention as a program. This is not totally unexpected because all households who participated in the sample, those who were selected to receive a voucher, and those in the control group who did not receive a voucher were all motivated to move. In Los Angeles, the difference between the samples and the baseline sample is almost certainly a product of the city’s demography and the high level of poverty among the Hispanic population, many of whom are undocumented and for whom moving is more difficult if not impossible.

The maps are a critical element of understanding any intervention in the urban fabric and provide a spatial representation of the outcomes (Figures 1–3). The overall similarity and the tendency to move to nearby neighborhoods reiterate basic mobility behavior in cities generally. The current Section 8 and current MTO, and even the control patterns, are remarkably similar visually for Chicago and New York. Even so, the Section 8 patterns in Chicago are more dispersed than the MTO locations, and the baseline/control patterns are not very different. Los Angeles is a significant contrast. The MTO patterns show significant gains. Large numbers of movers accessed housing in the San Fernando Valley to the north of the central city, as did some of the baseline movers. Section 8 mover patterns are much more circumscribed (Figure 3). Overall, the tables and figures emphasize that despite initial gains, those gains decline over time as individuals make additional location choices. It will be a recurrent theme of this analysis that intervention in a dynamic system of residential choices and moves is inherently difficult and that people do not “stay put”: they move, often frequently, to bring their housing needs into adjustment with their housing space as has been established in consistent and substantial research on residential mobility (Clark and Dieleman 1996). Perhaps it is obvious now, but the notion that “one-shot” intervention with a voucher and counseling would change spatial patterns was certainly overly optimistic. The fact that some authors continue to argue for such programs is not supported by this reanalysis.

There is considerable debate about whether nonmovers should be added to movers,10 but a program evaluation of whether vouchers are successful cannot be based on only those who were successful in moving. If, as some suggest, MTO should be expanded, we must have some sense of the overall success of the program as a whole. Reasonable evaluations will differ on the specifics; to reemphasize a recurrent theme of this paper, I am attempting a program evaluation, not testing whether specific households report gains from being in lower-poverty neighborhoods. How to target such individuals is a critical dimension of overcoming poverty.

Finally, an analysis of the mobility rates across the five cities provides additional data on the problems of controlled choice programs (Table 5). Although the (relatively uncontrolled) Section 8 recipients had relatively high mobility rates across all cities, only in Los Angeles was there a rate of mobility that provides confidence in the ability of a controlled program to generate successful mobility. For the other cities, mobility rates for the MTO sample hovered around 50%, with dramatically lower results in Chicago. Thus, the findings with respect to dispersing poverty are not sanguine, or at least with the interventions strategies envisaged in the MTO experimental program.

Table 5.

Geographic Mobility of the MTO and Section 8 African American Respondents

	Baltimore	Boston	Chicago	Los Angeles	New York
MTO Sample
Moved	112	48	125	91	64
No move	84	60	257	24	66
Percentage moved	57.1	44.4	32.7	79.1	49.2
Section 8 Sample
Moved	109	30	110	54	60
No move	45	35	63	6	57
Percentage moved	70.8	75.0	63.6	90.0	51.3

Source: MTO data for combined files for Baltimore, Boston, Chicago, Los Angeles, and New York; prepared by the HUD Office of Policy Development and Research.

An Empirical Analysis of Vouchers as a Tool for Racial Integration

Although the MTO program was never designed as an integrative program, several commentators (Briggs 2005; Goering 2005) have specifically discussed the “gains” in integration from the MTO program, and the Orr et al. (2003) report provides data on levels of integration. It seems relevant to take up this programmatic issue as well as the discussion of poverty.

An earlier study of all voucher households in Baltimore (the sample was predominantly African American) showed the difficulty in using vouchers to increase racial integration (Clark 2005). A replication of that analysis with African American households in all five cities shows that the initial moves of the MTO experimental group do result in greater integration for the initial move (Table 6). When movers and nonmovers are aggregated, the differences between the MTO sample and the Section 8 sample largely disappear (Table 7). For movers in all cities at the .05 level and three of the cities at the more conservative level of .01, the MTO sample movers are in more-integrated settings than the Section 8 movers (Table 8). Interestingly, there is less regression in racial integration than in poverty reduction. In Baltimore and Los Angeles, there was regression to less-integrated settings (significant at the .05 level). In several cities, either at the .05 or .01 levels, there were no differences between current MTO and current section 8 patterns.

Table 6.

Percentage of MTO and Section 8 African American Respondents, by Their Original Move Location and Their Current Location, and by Racial Composition of the Neighborhood

% African American	Original Move		Current Location
	MTO Mover	Section 8 Mover	MTO Mover	Section 8 Mover
Baltimore
0%–20%	8.40	1.79	5.36	2.75
20%–40%	26.89	13.39	12.50	16.51
40%–60%	17.65	10.71	9.82	8.26
60%–80%	14.29	14.29	16.96	10.09
80%–100%	32.77	59.82	55.36	62.39
N	119	112	112	109
Boston
0%–20%	67.35	12.90	47.92	13.33
20–40%	24.49	19.35	16.67	16.67
40–60%	0	19.35	2.08	20.00
60–80%	8.16	22.58	16.67	33.33
80–100%	0	25.81	16.67	16.67
N	49	31	49	31
Chicago
0%–20%	6.92	0	3.20	0
20%–40%	5.38	3.57	6.40	3.64
40%–60%	12.31	1.79	8.80	2.73
60%–80%	4.62	5.36	5.60	6.36
80%–100%	70.77	89.29	76.00	87.27
N	130	112	130	112
Los Angeles
0%–20%	69.79	43.40	46.15	40.74
20%–40%	4.17	24.53	20.88	24.07
40%–60%	12.50	22.64	27.47	18.52
60%–80%	4.17	5.66	4.40	9.26
80%–100%	9.38	3.77	1.10	7.41
N	96	53	96	53
New York
0%–20%	6.49	5.88	6.25	8.33
20%–40%	7.79	36.76	18.75	35.00
40%–60%	11.69	29.41	21.88	31.67
60%–80%	24.68	11.76	23.44	6.67
80%–100%	49.35	16.18	29.69	18.33
N	77	68	77	68

Source: MTO data for combined files for Baltimore, Boston, Chicago, Los Angeles, and New York; prepared by the HUD Office of Policy Development and Research.

Table 7.

Percentage of Total MTO, Total Section 8, and Baseline African American Respondents, by Their Current Location by Racial Composition of the Neighborhood

% African American	MTO Movers and Nonmovers	Section 8 Movers and Nonmovers	Baseline Sample
Baltimore
0%–20%	4.59	2.60	2.56
20%–40%	8.16	12.34	4.49
40%–60%	9.69	5.84	7.05
60%–80%	11.73	12.34	6.41
80%–100%	65.82	66.88	79.49
N	196	154	156
Boston
0%–20%	27.78	16.92	22.92
20%–40%	17.59	12.31	17.71
40%–60%	12.04	16.92	14.58
60%–80%	28.70	41.54	34.38
80%–100%	13.89	12.31	10.42
N	108	65	96
Chicago
0%–20%	2.36	1.16	2.17
20%–40%	3.14	2.89	3.80
40%–60%	4.97	2.89	4.89
60%–80%	3.14	4.62	4.35
80%–100%	86.39	88.44	84.78
N	382	173	184
Los Angeles
0%–20%	41.74	38.33	14.91
20%–40%	18.26	23.33	14.91
40%–60%	31.30	23.33	47.37
60%–80%	7.83	8.33	21.93
80%–100%	0.87	6.67	0.88
N	115	60	114
New York
0%–20%	5.38	4.27	5.10
20%–40%	28.46	31.62	35.71
40%–60%	27.69	41.03	40.82
60%–80%	21.54	10.26	16.33
80%–100%	16.92	12.82	2.04
N	130	117	98

Note: Some columns may not sum to 100.00 because of rounding.

Source: MTO data for combined files for Baltimore, Boston, Chicago, Los Angeles, and New York; prepared by the HUD Office of Policy Development and Research

Table 8.

Kolmogorov-Smirnov Two-Sample Tests of Differences Between Programs: Movers Only (treated) and the Total Sample (intent-to-treat) for Moves, by Race of Neighborhood Locations

City	Movers			Total Sample
	Original MTO Versus Original Section 8	Original MTO Versus Current MTO	Current MTO Versus Current Section 8	Current MTO Versus Current Section 8	Current MTO Versus Baseline	Current Section 8 Versus Baseline
Baltimore	**	**
Boston	**		*
Chicago	*
Los Angeles	*	*	*		**	**
New York	**		*		*

^*Significant at the .05 level.

^**Significant at the .01 level.

Again, the main test (of program effects between MTO and Section 8 effects) shows almost no difference at either the .01 or the .05 level. In fact, the test values are extremely low. Nor are there important differences between total sample current MTO and baseline or between current Section 8 and baseline except in Los Angeles and in New York for current MTO versus baseline at the .05 level (Table 8). Although there were some gains in poverty, there are almost none (Los Angeles excepted) across the five sample cities in terms of increased living in mixed-race settings. In Los Angeles, the MTO movers made initial gains and, to some extent, maintained those gains. There is no statistical regression over time. The differences between the experimental MTO sample, the Section 8 sample, and the baseline sample are almost certainly due to the nature of the composition of the ethnic population in Los Angeles, where many tracts are, in fact, integrated but integrated with combinations of Hispanic and African American populations. Even without vouchers and the special counseling of the MTO program, many households in Los Angeles ended up in more-integrated settings, in tracts that are 40%–60% African American. The sample from Los Angeles and the particular dynamics of that city are certainly affecting the aggregated positive outcomes detected in the Orr et al. (2003) report. This finding reiterates the affect of local demographics on program intervention.

Perhaps to make gains in integration, specific targeting of integrated tracts will be necessary. Overall, the findings of associated outcomes for integration are like those for poverty: less than compelling in the context of a policy. The fact that households in the control group are about as integrated as either the MTO sample or the Section 8 sample (except in Los Angeles) reemphasizes the outcomes for self-selected households who expressed a desire to move and indeed had significant mobility rates.

CONCLUSION

How easy is it, as a policy, to intervene in poverty distributions and to integrate neighborhoods? At least some researchers suggest that well-designed voucher programs will work nationally (Goering 2005:139). The research reported here suggests otherwise and raises important questions about such policy interventions. Although it may be possible to disperse some individual households, whether voucher programs can be used as a policy intervention is far from clear. Indeed, the research shown here suggests proceeding with caution in using such programs to change the concentrations and patterns of poverty. Others have suggested that it would require many individual moves and a lot of money to effect any substantial deconcentration of the poor (Goetz 2003).

The results from comparing aided and ordinary mobility reiterate the difficulty of intervening in the dynamic of household relocation. Consistent with our knowledge of mobility in general, subsequent moves by the MTO group were often to neighborhoods like the ones they came from—and, in some cases, back to their old neighborhoods. The geographic patterns illustrated in the maps emphasize the constraints on mobility and the selection process that favor known neighborhoods with friends, family, and support relationships. Households vote with their feet, and decisions by governments are always embedded in the dynamic demography of the city (Tiebout 1956). Income and assets are important and integral parts of the choice process as are preferences, and these forces play an ongoing role in the way in which households choose places to live. The evidence that the baseline sample also made gains and that their distributions are sometimes not different from the distributions of the combined samples of MTO movers and nonmovers and the Section 8 movers and nonmovers suggests viewing the calls for national voucher programs with caution.

None of this, of course, denies the finding that MTO special programs made initial changes in the distributions and that these distributions for movers had the effect of dispersing poverty. There were even gains over time in the dispersal; but, at the same time, those gains—when the MTO program is evaluated as a program—are not statistically significant. This is troubling to those who wish to emphasize the contributions of the MTO program, but it forces us to refocus our attention on the division between gains for individuals and gains from programs. For individuals, there were gains; as a program, it cannot sustain the claims that have been made for it. Perhaps better counseling would have made a difference, and counseling after the move might also be important. Indeed, there is evidence that on an individual basis, vouchers can work, and recipients of both MTO and Section 8 vouchers fared better than those without any assistance in some cases. However, this study is not about individual outcomes and localized situations, but about overall policy outcomes of an intervention program. Tests of better post-move counseling and other forms of assistance and their impact on the outcomes are tests that can be conducted only in the future.

Behavioral changes are impacting the metropolitan structure. Exit and voice have long been opposites available for urban households in their locational decisions, especially the decision of whether to stay in the central city. Many are leaving. Reich (1991) called it the secession of the successful (not of poor households, it is true, but it certainly raises the issue of the mobility processes related to income and education), and Wolfe (1998) pointed out that the propensity to secede is even higher among African Americans than other groups. These processes have been going on for some time, and they are not likely to change in the near future. Such movements are the context within which governments and agencies intervene in the urban fabric, and those interventions may not have the anticipated outcome. At the very least, we must be cognizant of the strong forces built into choice and selection, processes that daily make and remake our urban fabric. These forces are often more powerful than our limited ability to intervene with specific programs of assistance. It may be reasonable to suggest that redirecting attention to the root problems of education, jobs, and affirmative opportunities in the job market will provide greater gains in solving issues of inequality in the urban fabric.

Figure 2.

Current Locations of Households That Moved in Chicago: Regular Section 8 Moves, MTO Moves, and Control Sample Moves

Note: Although the initial origins are distributed across several central tracts in the city, for visualization purposes, they are shown as initiating from one central location.

Appendix Table A1.

Kolmogorov-Smirnov Tests of Significance for All Pairs of Possibilities: Poverty

City	Movers			Total Sample
	Original MTO Versus Original Section 8	Original MTO Versus Current MTO	Current MTO Versus Current Section 8	Current MTO Versus Current Section 8	Current MTO Versus Baseline	Current Section 8 Versus Baseline
Baltimore	0.7590 (0.2146)** (0.1790)*	0.4207 (0.2140) (0.1790)	0.2781 (0.2193) (0.1830)	0.0930 (0.1755) (0.1464)	0.1311 (0.1749) (0.1459)	0.1631 (0.1852) (0.1545)
	119/112	119/112	112/109	196/154	196/156	154/156
Los Angeles	0.5196 (0.2789) (0.2327)	0.4567 (0.2385) (0.1990)	0.2265 (0.2800) (0.2336)	0.1543 (0.2596) (0.21660	0.3341 (0.2154) (0.1797)	0.3631 (0.2600) (0.2169)
	96/53	96/91	91/54	115/60	115/114	60/114
Boston	0.6208 (0.3741) (0.3121)	0.3614 (0.3310) (0.2762)	9.2334 (0.3794) (0.3165)	0.1281 (0.2559) (0.2135)	0.2766 (0.2286) (0.2135)	0.1485 (0.2618) (0.2185)
	49/31	49/48	48/30	108/65	108/96	65/96
New York	0.6755 (0.2731) (0.2263)	0.3492 (0.2757) (0.2300)	0.3302 (0.2929) (0.2444)	0.1641 (0.2077) (0.1733)	0.3338 (0.2181) (0.1819)	0.216 (0.2232) (0.1862)
	77/68	77/64	64/60	130/117	130/98	117/98
Chicago	0.7048 (0.2101) (0.1753)	0.2743 (0.2042) (0.1704)	0.3535 (0.2131) (0.1778)	0.0941 (0.1494) (0.1246)	0.1167 (0.1463) (0.1220)	0.1264 (0.1726) (0.1440)
	130/112	130/125	125/110	382/173	382/184	173/184

^*p < .05;

^**p < .01

Appendix Table A2.

Kolmogorov-Smirnov Tests of Significance for All Pairs of Possibilities: Race

City	Movers			Total Sample
	Original MTO Versus Original Section 8	Original MTO Versus Current MTO	Current MTO Versus Current Section 8	Current MTO Versus Current Section 8	Current MTO Versus Baseline	Current Section 8 Versus Baseline
Baltimore	0.2705 (0.2146)** (0.1790)*	0.2526 (0.2146) (0.1790)	0.0401 (0.2193) (0.8300)	0.0609 (0.1755) (0.1464)	0.1339 (0.1749) (0.1459)	0.1261 (0.1852) (0.1545)
	119/112	119/112	112/109	196/154	196/156	154/156
Los Angeles	0.2639 (0.2788) (0.2327)	0.2364 (0.2385) (0.1990)	0.1118 (0.2800) (0.2336)	0.0630 (0.2596) (0.2166)	0.3018 (0.2154) (0.1797)	0.3185 (0.2600) (0.2169)
	96/53	96/91	91/54	115/60	115/114	60/114
Boston	0.5958 (0.3741) (0.3121)	0.2726 (0.3310) (0.2762)	0.3458 (0.3794) (0.3165)	0.1614 (0.2559) (0.2135)	0.0474 (0.2286) (0.2185)	0.1140 (0.2618) (0.2185)
	49/31	49/48	48/30	108/65	108/96	65/96
New York	0.4609 (0.2731) (0.2263)	0.2091 (0.2757) (0.2300)	0.2812 (0.2929) (0.2444)	0.1538 (0.2077) (0.1733)	0.2009 (0.2181) (0.1819)	0.1078 (0.2232) (0.1862)
	77/68	77/64	64/60	130/117	130/98	117/98
Chicago	0.1926 (0.2101) (0.1753)	0.0622 (0.2042) (0.1704)	0.1204 (0.2131) (0.1778)	0.0353 (0.1494) (0.1246)	0.0161 (0.1463) (0.1220)	0.0366 (0.1726) (0.1440)
	130/112	130/125	125/110	382/173	382/184	173/184

^*p < .05;

^**p < .01