The Australia
A
an National Univ
versity
Centre fo
or Econ
nomic Policy
P
Researrch
DISCUS
D
USSION PAPER
R
L
Long-T
Term Efffects off Publicc Low-In
ncome H
Housin
ng
V
Voucherrs:
Work
k, Neighborhoood, Fam
mily Com
mpositioon and
Child
dcare Usage
U
Roobert Havem
man
University
U
off Wisconsin-Madison an
nd
Research School of Economics
E
The Australlian Nationaal University
y
B
Barbara Wollfe
University
U
off Wisconsin-Madison an
nd
Research School of Economics
E
The Australlian Nationaal University
y
DISCUSSI
D
ION PAPE
ER NO. 66
67
JJuly, 2012
2
ISSN
N: 1442-8636
ISBN
N: 978-1-9216693-48-9
Long-Term Effects of Public Low-Income Housing Vouchers:
Work, Neighborhood, Family Composition and Childcare Usage*
Robert Havemana
University of Wisconsin–Madison
Research School of Economics
The Australian National University
Barbara Wolfeb
University of Wisconsin–Madison
Research School of Economics
The Australian National University
*The several research studies from which these results are drawn are coauthored with Deven Carlson
and Thomas Kaplan, of the University of Wisconsin-Madison, USA. The research presented in this
paper was generously supported by a grant from the John D. and Catherine T. MacArthur Foundation.
We gratefully acknowledge that support. Finally, we owe a debt of gratitude to Dan Ross for his work
in securing, cleaning, and organizing the data, and to Deborah Johnson for help in editing.
a
Institute for Research on Poverty, University of Wisconsin—Madison, 3412 Sewell Social Science Building, 1180
Observatory Drive, Madison, WI 53706, United States. [email protected]
b
Institute for Research on Poverty, University of Wisconsin—Madison, 3412 Sewell Social Science Building, 1180
Observatory Drive, Madison, WI 53706, United States. [email protected]
Abstract
Using a propensity score matching approach coupled with difference-in-differences regression
analysis, we estimate the effect of receiving a low-income housing voucher on the employment and
earnings, mobility, neighborhood quality, household/family composition and childcare utilization of a
large longitudinal sample of low-income families in the U.S. We observe these effects over six years
following voucher receipt. Our results indicate that voucher receipt has little effect on employment,
but a negative effect on earnings. The negative earnings effect is largest in the years immediately
following initial receipt, and fades out over time. Full-sample results show voucher receipt to have
little effect on neighborhood quality in the short-term, but some positive long-term effects. We also
find that voucher receipt is tied to a higher probability of change in household/family composition in
the year of voucher receipt, but greater stability in subsequent years. The results of our propensity
score matching procedure show voucher receipt to be tied to a greater take-up of public child care
subsidies. Several robustness tests are run to support the reliability of our findings. We discuss the
implications of our findings for research and policy.
Long-Term Effects of Public Low-Income Housing Vouchers
on Work, Neighborhood, Family Composition and Childcare Usage
1.
Introduction
In recent decades, low-income housing policy in the United States has expanded the provision of
vouchers that can be used to secure housing in the private market. Currently, the U.S. Section 8 housing
voucher program serves about 2.2 million families, with these households containing approximately 5.1
million family members (U.S. Department of Housing and Urban Development, 2008). Research has
followed implementation of this policy expansion, and an ever-growing body of studies examines the
behavioral responses of individuals to receipt of voucher-based housing assistance (Jacob and Ludwig,
2008; Mills et al., 2006; Susin, 2005; U.S. Department of Housing and Urban Development, 2003).
In addition to presenting evidence on behavioral responses from a new sample—one that is
larger, more diverse, and more representative of voucher recipients nationwide than samples used in most
previous studies—we compare and reconcile our findings with previous findings by noting the varying
counterfactuals, samples, and time horizons that characterize previous studies, and exploring the
implications of this variation. Our primary estimates of the effects of voucher receipt reflect a
counterfactual world in which comparison cases do not receive any housing assistance, and only seek
housing in the private market. However, our data also allow us to evaluate the effects of voucher receipt
against a counterfactual of public housing assistance. This analysis contributes to an understanding of the
disparate results in the existing literature on the labor market effects of vouchers.
In addition, we have analyzed the relationship between voucher receipt and labor market
outcomes for a wide variety of policy-relevant socioeconomic subgroups, an approach that allows us to
investigate the possibility of heterogeneous effects of voucher receipt. Previous studies have not been able
to conduct such detailed analysis. We summarize these findings here; details are available upon request
from the authors.
We perform our analyses using an original dataset assembled for this study. We constructed the
dataset from administrative records contained in two databases maintained by the State of Wisconsin
2
combined with information from the U.S. Census Bureau. This dataset contains information on up to six
years of post-voucher receipt earnings and employment patterns—and several additional years of prevoucher receipt information—for more than 350,000 households from all parts of a medium-sized,
diverse state.
2.
Low-Income Housing Policy and the Section 8 Voucher Program
2.1.
Low-income housing subsidy programs: tenant- vs. project-based approaches
Housing policy experts in the United States have long disagreed over the extent to which
government-provided low-income housing assistance should be project-based (with funds going to public
agencies or private developers to construct or remodel and to operate housing units for low-income
households) or tenant-based (with subsidies provided directly to low-income households who seek
housing in the private market that is compliant with Department of Housing and Urban Development
(HUD) and Public Housing Agency (PHA) requirements). Each approach has been employed to varying
degrees over the years. Initially, all government low-income housing assistance was project-based in
nature and was provided through the construction and operation of housing units by government agencies
(Orlebeke, 2000). This policy, generally referred to as the “public housing program,” monopolized lowincome housing policy from the mid-1930s through the early 1970s. At that point, dissatisfaction with
several aspects of the program—routine cost overruns, missed production targets, some issues with high
crime rates in major cities and dilapidated structures—resulted in policymakers beginning to take lowincome housing subsidies in a new, tenant-based direction. Specifically, the Housing and Community
Development Act of 1974 authorized the Section 8 voucher program, the details of which are described
below. This program expanded throughout the late 1970s and early 1980s and during this time it appeared
that the vast majority of low-income housing assistance would be provided through this tenant-based
approach for the foreseeable future.
However, a decade later, Section 42 of the Tax Reform Act of 1986 authorized a program that has
come to be known as the Low-Income Housing Tax Credit (LIHTC). This program provides subsidies to
3
private developers who construct housing units that will be subject to rent ceilings and tenant income
limits for at least 15 years following construction (for a more comprehensive description of the LIHTC,
see Eriksen, 2009). Since its inception in 1986, the LIHTC has become the predominant form of projectbased housing assistance, with nearly 1.7 million housing units subsidized by the program in use in 2008
(U.S. Department of Housing and Urban Development, 2010). The 1.7 million LIHTC units exceed the
1.2 million public housing units in use—another major project-based housing assistance program—but
falls short of the 2.2 million tenant-based Section 8 vouchers in existence in 2008.
Although with different emphases, substantial research literatures have studied these three major
low-income housing assistance programs identified above. By examining the labor market, mobility,
neighborhood, household composition and child care usage effects of voucher receipt, we are able to
provide insight into a variety of important aspects of the primary tenant-based housing assistance program
in the United States—the Section 8 housing voucher program. These results inform the continuing debate
over the direction of national housing policy, and the effects of tenant- versus place-based housing
subsidy programs. And in doing so we hope to provide insight for Australia into likely effects of
providing housing subsidies to lower income families, especially to those families with children.
2.2.
Section 8 voucher program background
As described above, the U.S. Department of Housing and Urban Development (HUD) currently provides
housing assistance to low-income households through a variety of programs, including the Section 8
tenant-based voucher program (officially known as the Housing Choice Voucher Program since 1999). 1
This program, which is operated by HUD in conjunction with over 3,000 local public housing authorities
(PHAs), currently serves about 2.2 million families nationally, including more than one million families
1
The “Section 8” designation refers to the program’s statutory authorization under Section 8 of the United
States Housing Act of 1937, as amended by the Housing and Community Development Act of 1974. Although the
official title of Section 8 tenant-based assistance is now called the Housing Choice Voucher Program, most
researchers and administrators still refer to it as the “Section 8 voucher” program. We use the “Section 8”
designation in this paper.
4
with minor children (U.S. Department of Housing and Urban Development, 2010). The primary objective
of the program is to enable “very low-income families, the elderly, and the disabled to afford decent, safe,
and sanitary housing in the private market.” 2 A secondary objective of the program involves facilitating
the relocation of recipients to better neighborhoods. 3
The process of securing a Section 8 voucher begins with the submission of an application to a
PHA at a time when the waiting list is open and the PHA is accepting applications; upon submission,
applicants are assigned a position on the waiting list. 4 Each PHA has the autonomy to establish
preferences for individuals or households with particular characteristics. 5 When the applicant’s name rises
to the top of the waiting list, the household meets with housing authority staff who outline the rules and
requirements of the Section 8 program and provide recipients with instructions for seeking housing
compliant with the program. Under the program, voucher recipients are responsible for locating housing
in the private market that meets a minimum standard of health and safety. This housing can take the form
of a single-family home, a townhouse, or an apartment, but any unit a recipient locates must be owned by
a landlord who is willing to rent under the terms of the program. If a voucher recipient—whose income
must, in general, be below 50 percent of the median income of the county or metropolitan area in which
2
http://www.hud.gov/offices/pih/programs/hcv/about/fact_sheet.cfm#10
3
As the program has expanded over time, a number of constraints have partially interfered with the goal of
geographic mobility for voucher recipients. One constraint has been the limited geographic span of many local
PHAs that serve only parts of metropolitan areas. While some PHAs allow recipients to find housing in other
jurisdictions, administrative burdens and the need to transfer supporting funds constrain this practice. Such transfers
also impose additional costs on recipients in the form of duplicate application, orientation, and program criteria
(Katz and Turner, 2000).
4
At many PHAs, particularly those in urban areas, Section 8 waiting lists are open for limited periods of
time—often only a week or two each year. In an extreme example, after opening for a couple weeks in 2007 the
Section 8 waiting list at the PHA in Madison, Wisconsin was closed throughout all of 2008 and 2009. Applicants
can apply for project-based housing assistance at the same time they apply for Section 8 assistance. Waiting lists for
project-based housing assistance are generally more likely to be open and to be shorter than are the Section 8
waiting lists. Applicants who received a Section 8 voucher in Wisconsin in 2008 spent an average of 27 months on
the waiting list. PHAs differ in their policies toward project housing residents who reach the head of a Section 8
waiting list; some authorities skip past their name and do not inform the residents that they have reached the head of
the Section 8 list, whereas other authorities allow project residents to move out of the project and into a Section 8
apartment if they reach the head of the Section 8 waiting list.
5
For a comprehensive description of waiting list policies—including common examples of preferences
instituted by PHAs—see http://www.hud.gov/offices/adm/hudclips/guidebooks/7420.10G/7420g04GUID.pdf.
5
they live—is able to locate suitable housing, then the recipient household generally contributes 30 percent
of its income toward rent. 6 The Section 8 program then subsidizes the difference between the tenant
contribution and actual rent, up to a locally defined “fair market rent” payment standard.7
3.
Conceptual Issues in Understanding the Behavioral Effects of Voucher Receipt
The complex nature of the Section 8 program generates a diverse set of opportunities and incentives that
may influence the behavior of voucher recipients in many different ways. As a result, it is difficult to
make clear, unambiguous predictions regarding expected effects of the program (Jacob and Ludwig,
2008; Shroder, 2002a). This theoretical ambiguity gives greater significance to the empirical
relationships, which have been studied by researchers in recent years. Prior to reviewing the literature on
this topic, and situating our study within this body of research, we explore the variety of conceptual issues
that arise when considering the effect of voucher receipt on a number of outcomes beginning with
employment and earnings and then turning to neighborhood, family composition and childcare.
3.1
Employment and earnings effects
In theory, voucher receipt enables recipients to make location and other changes that may result
in improved labor market outcomes. Specifically, voucher receipt presents households with an
opportunity to relocate to areas with available jobs, better child care options, lower crime rates, and other
characteristics that may encourage and enable individuals to locate and maintain steady, desirable
employment. Relocation of this sort could result in increased adult earnings and incomes, reduced
reliance on welfare assistance, and better outcomes for children.
At the same time, however, the difficulties and disruptions associated with preparation for and
execution of a move to a different neighborhood, even one with better job opportunities, may lead a new
6
See http://www.hud.gov/offices/pih/programs/hcv/about/fact_sheet.cfm. Each PHA must provide 75
percent of its vouchers to applicants whose incomes do not exceed 30 percent of the area median income.
7
This standard is set by the HUD at the 40th percentile of the local rental market, as calculated by the
monetary value of leases commenced in the previous year.
6
voucher recipient to temporarily work fewer hours in an existing job, or to search for a different job. A
move to a new neighborhood may also disrupt previous social and support arrangements from friends and
neighbors to childcare arrangements (Ross, Reynolds, and Geis, 2000; Swartz and Miller, 2002). These
factors could lead to lower short-run attainments across a variety of economic and social dimensions—
including labor market outcomes.
In addition to the potential short-term decrease in earnings and employment caused by such
disruptions, several features of Section 8 program design may also lead voucher recipients to reduce
employment and earnings relative to their pre-receipt levels. First, because the Section 8 program requires
recipient households to contribute 30 percent of income toward rent and then subsidizes the difference
between the tenant contribution and actual rent, the subsidy value of recipients’ Section 8 voucher falls as
their income rises. In terms of standard economic theory, voucher receipt increases the marginal tax rate
on earnings of all program beneficiaries, hence increasing work disincentives (Van Ryzin, Kaestner, and
Main, 2003).
Second, because the Section 8 program subsidizes the difference between the 30 percent of
income tenant contribution toward rent and the rent that is actually charged, the program provides an
effective income boost to households. Standard economic theory predicts that increases in income will
affect the trade-off between labor and leisure, resulting in an increase in leisure consumption and an
equivalent decrease in hours worked and earnings.
Finally, the Section 8 program specifies an income threshold that households must fall below in
order to be eligible for voucher receipt. If a household has income above the eligibility threshold,
continued voucher receipt worth thousands of dollars annually is jeopardized. Hence, after receipt,
households may choose to reduce their earnings in order to ensure that they will maintain eligibility for
voucher receipt. 8 Taken together, these three aspects of the Section 8 program design—the increase in the
marginal tax rate, the increase in household income, and the income eligibility threshold—provide
8
This consideration will be more salient for cases whose income is near the program eligibility limit.
7
incentives for households to decrease their employment and earnings, relative to a counterfactual of
receipt of no housing assistance. 9
Given the competing conceptual considerations, a clear prediction regarding the labor market
effects of voucher receipt is not possible.
Although it is difficult to make any absolute predictions about the relationship between voucher
receipt and labor market outcomes, the preceding discussion does support a number of relative
predictions. First, all else equal, cases that are near the income eligibility threshold are likely to exhibit
more negative labor market responses relative to cases that lie further from the eligibility threshold.
Hence, we expect that groups with traditionally high relative earnings capacity (e.g., whites, those with
more education, married parents) will exhibit more negative labor market responses than groups that
traditionally have lower earnings capacity. Second, while the income effect stemming from voucher
receipt will affect all recipients, we expect that this income effect would be larger for those population
groups with greater impediments to work, such as those with greater family responsibilities (e.g., families
with children, especially single parents; older workers with lower labor market attachments). Our
empirical analysis will address these conjectures, but we first review existing evidence on this issue.
3.2
Neighborhood quality
Decisions regarding neighborhood quality may also be affected by the fact that voucher receipt
alters the budget constraint for housing. 10 As we described above, as a result of the program design
9
Two of these theoretical considerations—the increased marginal tax rate and the increased income
provided by voucher receipt—apply only when the counterfactual is specified as receipt of no housing assistance, as
it is in this analysis. If the counterfactual is specified as receipt of public housing, as in the Moving to Opportunity
(MTO) experiment, then voucher receipt does not result in increased marginal tax rates or increased household
income because residents of public housing, like voucher recipients, contribute 30 percent of their income toward
rent, with the government subsidizing remaining rental costs. As we explore in greater detail later, varying
counterfactuals may provide a partial explanation for the discrepant evidence on the relationship between voucher
receipt and labor market outcomes in previous studies.
10
It is important to note that Section 8 voucher receipt alters recipients’ budget constraint for housing only
when the counterfactual is specified as receipt of no housing assistance, as it is in this analysis. If the counterfactual
is specified as receipt of public housing, as in the Moving to Opportunity (MTO) experiment, then voucher receipt
Comment [MSOffice1]: Do
we do this?
8
voucher recipients face a zero marginal price of housing up to the locally defined fair market rent. 11
Figure 1, which presents recipients’ budget constraints in the absence and presence of the Section 8
program, illustrates that the program will often result in an increase in recipient housing consumption. If
neighborhood quality represents an important site factor that households consider in choosing housing
option, standard economic theory predicts that voucher receipt would result in improved neighborhood
quality. However, if households place the great majority of weight on unit characteristics or nonneighborhood site factors in selecting housing, voucher receipt may not result in improved neighborhood
quality. Given these incentives, voucher receipt is unlikely to lead to a decrease in neighborhood quality.
[Insert figure 1 about here]
3.3
Household/family composition
Regarding household/family composition, three aspects of the Section 8 program design—the
income gain, the increased marginal tax rate, and the income ceiling for benefits—are likely to affect
recipients’ choices. These include whether the household is a single or multigenerational family for
example.
Income Gain
First, as Figure 1 indicates, the Section 8 program can result in an income boost for recipient
families, which may enable a variety of changes in household/family composition that had previously
been constrained by financial considerations. Recipients may be able to terminate multigenerational
housing arrangements or cohabitation relationships with other adults that were necessary to meet all
financial obligations. Changes in these arrangements would imply a reduction in the number of adult
does not alter recipients’ budget constraint because residents of public housing, like voucher recipients, contribute
30 percent of their income toward rent, with the government subsidizing remaining rental costs.
11
Technically, recipients face a zero marginal price of housing for units above the minimum acceptable
standards of health and safety and fair market rent. Beyond the level of fair market rent, voucher recipients are able
to purchase additional housing (up to 40 percent of their income) by paying the incremental market rent, without
loss of the voucher.
9
members in the household. On the other hand, the increase in income associated with voucher receipt
may provide individuals with the ability to support adult children or other individuals experiencing
hardship. This consideration implies that voucher receipt could result in the addition of adult members to
the household. The additional income provided by a voucher also seems likely to provide recipients with
the resources necessary to support additional children, especially because the value of a rental voucher
can increase with larger housing units, which can be justified to accommodate additional children.
Increase in Marginal Tax Rate
Because the Section 8 program requires recipient households to contribute 30 percent of income
toward rent, the program subsidy is effectively income-conditioned. 12 At the margin, this may discourage
the addition of earning adults to the household, and encourage departure of earning adults who had
previously been in the household primarily to meet financial obligations.
Income Ceiling for Benefits
Third, if a household has income above the income eligibility ceiling, continued voucher receipt
is jeopardized. Hence, after receipt, households—especially those whose income is near the program
eligibility limit—may reduce the number of earning adults in the household in order to ensure that the
household will retain eligibility for voucher receipt.
Taken as a whole, the Section 8 program design seems likely to create an incentive structure that
would lead to a reduction in the number of adult members of a household and encourage the addition of
children. Moreover, the relocation that often accompanies Section 8 voucher receipt is likely to affect the
theoretical considerations identified above. Relocation provides a natural opportunity for households to
alter their composition and represents an important mechanism for families/households to change their
neighborhood surroundings. Because relocation often takes place soon after voucher receipt, significant
change in household/family composition and neighborhood quality may occur in the first year of voucher
12
Again, it is important to note that Section 8 voucher receipt increases household income and the marginal
tax rate only given our counterfactual of no receipt of housing assistance; see note 10.
10
receipt, with subsequent stability in household/family composition and locational characteristics after this
re-optimization. 13 Our empirical analysis will assess the accuracy of these predictions and conjectures.
3.4
Take-up of child care benefits
Given the incentives implicit in housing voucher programs, standard economic theory is not
able to provide unambiguous predictions regarding expected program impacts on the take up of child
care benefits (Shroder, 2002a). On one hand, voucher recipients could use the opportunity provided by
their voucher to find housing closer to areas with available jobs and child care programs. On the other
hand, voucher receipt may result in recipients relocating to areas where they are unfamiliar with the
local child care options. And finally, if voucher recipients move from a multi-generational family living
together to a single generational family, there may be greater need for more formal childcare to replace
“grandparent or other family member care”.
Even if it were clear that receipt of a Section 8 voucher was related to changes in child care
usage, identifying the causal mechanisms leading to these relationships is difficult. For example,
evidence of increased child care take-up over time for Section 8 voucher recipients relative to
nonrecipients could reflect the fact that families obtaining a subsidy are more skillful in navigating
complex bureaucracies, or that they are more persistent, or have other attributes that explain both their
securing a Section 8 voucher and their program take-up choices. In short, securing a housing voucher
and demanding more child care services likely requires many of the same personal characteristics,
making any causal effect of vouchers difficult to identify.
13
Many of the households that we classify as voucher recipients do not use their voucher during the entire
observation period. Indeed, the typical recipient household tends to use their Section 8 voucher for less than four
years. Our analysis describes how the households which received and used a voucher in the treatment year revealed
geographic area and household composition outcomes in subsequent years, whether or not they continued to use
their voucher for the entire observation period.
11
4.
Empirical Research on the Behavioral Effects of Housing Assistance
In this section of the paper we discuss existing analysis of the influence of housing vouchers on
recipient labor market, neighborhood quality, household/family composition, and childcare use responses
attempting to provide some insight on the sources of differences.
4.1
Research on labor market effects
An extensive body of research, both experimental and nonexperimental, examines the labor
market effects of Section 8 voucher receipt. Less research has been conducted on the labor market effects
of project-based housing assistance, such as public housing or housing developed with assistance from the
Low Income Housing Tax Credit Program. Shroder (2002a) comprehensively reviews early work on the
labor market effects of vouchers, describing mixed evidence on the subject; he concludes that housing
assistance has little overall effect on individuals’ labor market supply. Since Shroder’s review, several
additional studies have analyzed these relationships. The experimental studies are described in Table 1a,
and the nonexperimental studies in Table 1b.
The experimental studies all focus on the labor market effects of housing voucher receipt; some
of the studies specify project-based assistance as the counterfactual, while others compare voucher
recipients to those receiving no housing assistance. The nonexperimental studies tend to analyze the labor
market effects of any housing assistance, either voucher-based or project-based. Specifically, five of the
six nonexperimental studies include both voucher recipients and recipients of project-based assistance in
their sample of housing assistance recipients; the sole study that does not include recipients of both types
of assistance—Newman, Holupka, and Harkness (2009)—analyzes only recipients of project-based
assistance. All nonexperimental studies in Table 1b specify the counterfactual as no housing assistance.
[Insert tables 1a and 1b here]
Results from the experimental studies in Table 1a are somewhat inconsistent. Two of the
studies—the Welfare to Work (WtW) and Chicago Housing Authority (CHA) experiments—report a
significant negative effect of voucher receipt on recipient earnings. The other two studies—the Gautreaux
12
study and the Moving to Opportunity (MTO) experiment—find no evidence of a negative relationship
between voucher receipt and recipient earnings. We briefly review each of these studies before discussing
factors that could contribute to the discrepant results.
The Gautreaux study represents the earliest experimental evaluation of voucher-based housing
assistance. This study, which took place in Chicago from the 1970s to the 1990s, compared the earnings
of public housing project residents who were given vouchers and randomly assigned to move to the
suburbs to the earnings of public housing residents who were given vouchers and randomly assigned to
remain in the city. The analysis found voucher recipients assigned to move to the suburbs exhibited
significantly higher earnings than voucher recipients assigned to remain in the city.
MTO is the most prominent experimental evaluation of voucher-based assistance. This study
randomly assigned participating public housing residents in five large urban areas to one of three groups:
a control group that remained in public housing, a comparison group that received a voucher that could be
used anywhere, and an experimental group that received a voucher that could only be used in
neighborhoods with low poverty rates. The study then evaluated the earnings outcomes of these three
groups, and generally failed to find any statistically significant differences between the three groups in
any of the first four years after random assignment (U.S. Department of Housing and Urban
Development, 2003).
A third experimental evaluation—the Chicago Housing Authority (CHA) experiment—exploits
the fact that 82,600 Chicago families were randomly assigned a position on the waiting list for a Section 8
voucher in 1997. The study compares the earnings and employment of families that received a voucher to
the earnings and employment of families that remained on the waiting list and secured housing in the
private market. Results of the study indicate that voucher receipt results in significant decreases in
earnings and employment. These decreases persisted over the full period of observation.
The final experimental evidence on the relationship between voucher receipt and labor market
outcomes is from the Welfare to Work (WtW) evaluation. This study was conducted across six cities and
randomly assigned approximately 8,700 families that were receiving or eligible to receive welfare
13
benefits to either a treatment group that received a Section 8 voucher or a control group that did not
receive a voucher. Analysis of the relationship between voucher receipt and labor market outcomes
revealed that the treatment group initially worked and earned less than the control group, but these
differences dissipated to statistical insignificance by three and a half years after initial receipt.
The divergent findings of the experimental evaluations of tenant-based housing subsidies—
vouchers—may appear somewhat puzzling at first glance, but the basic difference in the counterfactual
adopted by the studies can help reconcile the findings. Consider first the MTO and Gautreaux evaluations.
Both of these studies provided vouchers only to samples that were public housing recipients at the outset
of the program, so that voucher receipt did not alter the marginal work incentives of voucher recipients
from what they experienced in public housing; it did not increase household income or raise the marginal
tax rate that they faced. Hence, these studies evaluated the effects of residential relocation and not the
receipt of an added public benefit. Given the counterfactual in these studies, it is perhaps not surprising
that neither study found a negative labor market effect from voucher receipt.
In contrast, both the WtW and CHA experiments compare voucher recipients to households
receiving no housing assistance, thereby specifying a counterfactual that results in voucher receipt
improving households’ financial situations and increasing the marginal tax rate they face. Both of these
evaluations find voucher receipt to have a negative effect on earnings and employment, particularly in the
years immediately following voucher receipt.
In addition to the varying counterfactuals underlying the studies, the characteristics of the
samples used in these analyses also differ in important ways. Three of the studies—MTO, Gautreaux, and
the CHA—employ samples consisting entirely of households in large urban areas. WtW includes middlesize cities (e.g., Augusta, Fresno, and Spokane), in addition to two large cities (Houston and Los
Angeles). The MTO and WtW samples were limited to families with minor children; the WtW sample
was further limited to families receiving, or eligible to receive, welfare benefits (TANF), and for whom
(in the judgment of housing agency officials) the subsidy “was critical to the families’ ability to obtain or
retain employment” (Mills et al., 2006, 7). The MTO and the CHA samples were also racially selective,
14
with non-Hispanic whites composing only about 3 percent of participants in each study. In sum, previous
experimental work has employed samples that differ in composition—none of which are broadly
representative of the population of voucher recipients nationally—which could restrict comparability of
results.
The nonexperimental research examining the effect of housing assistance on labor market
outcomes is summarized in Table 1b; results in these studies also diverge. As is the case with the
experimental research, these divergent findings can be partially reconciled by examining the
counterfactuals and samples underlying the studies. The two studies that found housing assistance to
result in decreased earnings relied on nationally representative samples and a counterfactual of no receipt
of housing assistance (Olsen et al., 2005; Susin, 2005). The three studies that find housing assistance to
have no effect on earnings are either based on nonrepresentative localized samples (Bania, Coulton, and
Leete, 2003), specify a counterfactual that may not result in housing assistance truly altering work
incentives (Van Ryzin, Kaestner, and Main, 2003), or employ methodological approaches that do not
warrant causal inferences (Harkness and Newman, 2006). While Newman, Holupka, and Harkness (2009)
compare housing assistance recipients to people with no assistance—thus specifying a counterfactual in
which housing assistance receipt alters work incentives—they study only the effect of project-based
assistance.
4.2
Research on neighborhood quality effects
A relatively limited body of empirical research has examined the relationship between voucher
receipt and neighborhood quality; most of it is based on evidence from three experimental studies,
especially the Moving to Opportunity (MTO) experiment. This study randomly assigned participating
public housing residents in five large urban areas to one of three groups: a control group that remained in
public housing (Control group), a comparison group that received a voucher that could be used anywhere
(Section 8 group), and an experimental group that received a voucher that could be used only to secure
housing in areas with low poverty rates (MTO group). The final evaluation of the program finds that,
relative to the control group, families assigned to the MTO group were lived in neighborhoods with lower
15
poverty rates (Sanbonmatsu et al. 2011). When averaged over the 10 to 15-year study period, families
assigned to the MTO and Section 8 groups lived in census tracts with average poverty rates of 31 percent
and 33 percent, respectively. This compares favorably to the average poverty rate of 40 percent in the
census tracts in which families assigned to the control group lived over the study period.
A second source of experimental evidence on the effect of voucher receipt on neighborhood
quality is Jacob’s (2004) analysis of a natural experiment in Chicago. Jacob exploited the fact that the
Chicago Housing Authority offered Section 8 vouchers to public housing residents whose units were
scheduled for demolition. In his analysis, Jacob classified voucher households whose units were
scheduled for demolition as treatment cases while a group of voucher recipients whose units were not
scheduled for demolition served as the control group. Jacob finds that, three years after initially receiving
notification that their public housing unit would close, treatment group members’ average census tract
poverty rate was 53 percent, whereas the average poverty rate for members of the control group was 67
percent.
The final source of experimental evidence on the relationship between voucher receipt and
neighborhood quality is the Welfare to Work (WtW) evaluation. This study was initiated in 2000 and
conducted across six cities—Atlanta, Augusta, Fresno, Houston, Los Angeles, and Spokane. In this
study, approximately 8,700 families that were receiving or eligible to receive welfare benefits were
randomly assigned to either a treatment group that received a Section 8 voucher or a control group that
did not receive a voucher. Results of the study indicate that four years after random assignment, voucher
recipients lived in neighborhoods with lower poverty rates, higher adult employment, and lower welfare
participation rates than did the control group (Mills et al. 2006). 14 Although these differences are
statistically significant, they are quite small in magnitude relative to the two experimental studies
described above. For example, the average poverty rate in census tracts where members of the treatment
14
Notably, voucher receipt did not have a significant impact on census tract poverty rate until the fifth
quarter after random assignment.
16
group resided was only 0.6 percentage points lower than the average poverty rate in census tracts where
members of the control group resided. 15
Taken together, the existing literature suggests that Section 8 voucher receipt can result in
households residing in higher-quality neighborhoods, at least as measured by the poverty rate of the
census tract. However, estimates of the extent to which Section 8 recipients reside in higher-quality
neighborhoods are quite inconsistent across the three experimental studies; estimates of the difference in
the average neighborhood poverty rates for the treatment and control groups appear to be larger for the
studies using public housing recipients (Jacob and MTO) than for those that test the effect of voucher
receipt on a more general population of recipients (WtW). 16
4.3
Research on household composition effects
Research on the effects of housing voucher receipt on household composition is far more limited.
To our knowledge, the Welfare to Work experiment is the only study that has reported on these effects to
a meaningful extent. About four years after random assignment, voucher receipt reduced the proportion of
15
The limited nonexperimental research on the effects of low-income housing vouchers also shows that
Section 8 voucher recipients are less likely than public housing residents to live in high-poverty neighborhoods. For
example, Newman and Schnare (1997) found that 54 percent of public housing residents lived in neighborhoods in
which more than 30 percent of residents were poor, whereas only 15 percent of Section 8 voucher recipients lived in
such neighborhoods.
16
The discrepant magnitudes are likely due in part to contrasting counterfactuals. The counterfactual in the
Jacob and MTO studies is residence in traditional public housing. The findings in these studies that voucher receipt
results in households living in neighborhoods with significantly lower poverty rates may not be surprising, given
that public housing complexes are often located in neighborhoods that are quite poor, particularly in the cities where
Jacob’s study and MTO took place. The counterfactual in the WtW study, in contrast, was specified as no voucher
receipt, and just 7 percent of the households in the experiment lived in public housing at the time of random
assignment. Again perhaps not surprisingly, this study found voucher receipt to result in households living in areas
with only slightly lower poverty rates. (The study finds that those who lived in public and assisted housing were
more likely to move to a different census tract [p=<.01] but not more likely to move to a tract with a lower poverty
rate than were control group members who resided in public or assisted housing at baseline [exhibits D.1 and D.3].)
Because most housing voucher recipients do not transition from traditional public housing to Section 8 voucher
receipt, but rather from no housing assistance to Section 8 receipt, residence in public housing is not always an ideal
counterfactual in the study of housing voucher effects. Residence in public housing in large central cities may be an
especially misleading counterfactual for some studies of housing vouchers, since the majority of voucher recipients
do not reside in central cities. The most recent national data show that less than half (47 percent) of voucher
recipients lived in central cities, 14 percent resided outside of metropolitan statistical areas and 39 percent resided in
suburbs (HUD 2008b).
17
multigenerational households and average household size, but did not have an effect on the likelihood of
residing with a spouse or partner or on the number of children living in the household (Mills et al. 2006).
Given the limited existing empirical research on household composition, our analysis has the potential to
yield insights into the effects of policy on this important social outcome.
4.4
Research on child care take-up
Non-experimental empirical studies have addressed the effects of housing assistance on public
program utilization generally, but none of these studies study the take up of public child care services.
5.
Our Research Approach
Below we analyze the effect of housing voucher receipt on employment and earnings, neighborhood
quality, household composition and childcare use among a broad sample of recipient households in
Wisconsin. The breadth and diversity of our sample indicates that it is broadly representative of voucher
recipients nationwide. 17 The large sample size also enables us to separately analyze the effects of voucher
receipt for a wide variety of policy-relevant socioeconomic subgroups; here we only summarize these
results.
We specify the counterfactual as receipt of no housing assistance, which results in voucher
receipt increasing both the marginal tax rate and household income of recipients; this counterfactual is
similar to those used in the CHA and WtW experiments. Under this counterfactual, our analysis
effectively estimates the effect of a marginal expansion of the Section 8 program on the behavior of new
voucher recipients, relative to a comparison group that receives no housing assistance. 18
17
According to data on the HUD Web site “A Picture of Subsidized Housing: 2008,” the demographic
profile of Section 8 subsidized households in Wisconsin is similar to that for the United States as a whole with the
only major difference being a lower proportion of Hispanics among recipients in Wisconsin (5 percent vs. 17
percent). In terms of other characteristics comparing Wisconsin to the United States: 12 percent vs. 15 percent are 2
adults; 40 percent vs. 36 percent are single adults; 49 percent vs. 48 percent are female heads with dependent
children; 20 percent vs. 18 percent are disabled; 59 percent vs. 56 percent are between ages 25 and 50; 18 percent
are age 62 plus (both); 38 percent vs. 42 percent are black; and 1 percent are Native American (both).
18
For two distinct, yet related, reasons we believe this to be the most appropriate counterfactual for understanding
the behavioral effects of voucher receipt, and hence for evaluating the impacts of this public program. First, given
the lengthy waiting lists for Section 8 vouchers—households that received vouchers in Wisconsin in 2008 spent an
18
5.1.
Data and estimation sample—the CARES database
In our analysis, we use a dataset consisting of detailed administrative records from two large-scale
databases maintained by the State of Wisconsin, supplemented with data from the U.S. Census Bureau.
The CARES Database
The administrative records contained in the large-scale Client Assistance for Re-employment and
Economic Support (CARES) database serve as the basis of the dataset we created for this analysis.
CARES contains extensive information on all individuals residing in a household that receives benefits
from any public program administered by the state. The demographic information contained in CARES
includes age, race, and disability status of all members of the living unit, as well as the years of education
for all adults in the case. In addition, the household’s history of participation in a wide variety of meanstested programs and the address history of the household are all included in the database. Finally, and
most importantly for this analysis, the CARES database indicates whether households receive voucherbased housing subsidies, reside in project-based public housing, or receive no housing assistance. The
CARES database is very large; some 470,000 cases were open in CARES over the course of 2003.
As the first step in creating our dataset, we extracted from CARES all cases that applied for or
received food stamps or TANF in 2001, 2002, or 2003. 19 This yielded three separate calendar-year
cohorts. Within each of the three calendar-year cohorts, we formed two unique groups, one composed of
families that first received a public rental subsidy in that year, and the other made up of families that did
average of over two years on the waiting list—it is reasonable to analyze an expansion of the Section 8 program that
would increase the number of available vouchers. Second, any new voucher recipients resulting from a marginal
expansion of the Section 8 program are likely to have received no housing assistance prior to voucher receipt; data
in our sample indicate that nearly 95 percent of new voucher recipients received no housing assistance prior to
voucher receipt; less than 10 percent moved from receipt of public housing assistance directly to voucher receipt.
19
We extract cases that applied for food stamps or TANF in 2001–2003 because it is during the application
(and renewal) process for these programs that households are asked whether they receive voucher-based subsidies,
reside in project-based public housing, or receive no housing assistance. The cases that receive no housing
assistance serve as potential comparison cases in this analysis.
19
not. 20 To obtain our desired counterfactual of no housing assistance, we excluded from our sample all
cases that resided in public housing. 21 Table 2 summarizes the demographic characteristics of the two
groups used in our analysis extracted from the CARES database.
[Insert table 2 here]
After creating the voucher and non-voucher groups for each of the three cohorts, we pooled the
cohorts to streamline our analysis. Then, for each case in our dataset, we extracted a wide variety of
variables from CARES for use in our analysis. These variables include demographic characteristics,
household composition including in particular the presence of children by age group, address, and
participation in means-tested programs, including in particular childcare subsidies.. We extracted this
information annually, beginning at baseline and extending through 2006, or up to six years after voucher
receipt.
The UI Database
After using the CARES database to identify voucher and potential comparison cases—and to glean
important baseline and up to six years of subsequent information on these cases—we then added
employment and earnings information to each case record. Using administrative records from the
Unemployment Insurance (UI) System—another large-scale database maintained by Wisconsin state
government—we merged individual-level earnings data from 1996 until 2006 into our dataset. 22 This
merge provided us with annual earnings data for all individuals and cases in our dataset from five years
20
A family unit is defined as being in the voucher group if the CARES case file indicates that it first
received a rental subsidy in a particular calendar year or if the case file indicates that the case received a rental
subsidy after a minimum of two consecutive months of nonreceipt. Nonrecipient units are those cases that did not
join the voucher group according to this criterion and were not receiving a rental subsidy. As noted above, those that
resided in public housing at the time of their initial application or receipt of TANF or food stamps are excluded from
our analysis.
21
22
We also excluded a few cases that were recorded as earning over $50,000 in a calendar year.
Although the employer-reported earnings data in the UI database are reported on a quarterly basis, we
analyze annual earnings in this study.
20
prior to voucher receipt up to six years after receipt. In our primary analysis, we estimate the effect of
voucher receipt on casehead earnings.
From the earnings data we derived our employment variable, which measures the number of
quarters an individual worked in each year. An individual was recorded as working in a given quarter if
the UI database contained a positive earnings report for the individual in that quarter. We then summed
the number of quarters worked in each calendar year to generate employment information for all
individuals and cases in our data again from five years prior to voucher receipt up to six years after
receipt.
Census block group data—neighborhood characteristics
As the final step in creating our dataset, we commissioned the University of Wisconsin–Madison’s
Applied Population Lab to match each address in the address history of all cases in our database to a
census block group, and then record a variety of characteristics associated with each block group. 23 We
then merged approximately 20 block group characteristics from the 2000 Census into the dataset. 24 From
these 20 block group characteristics we selected four—the unemployment rate, the percentage of persons
in poverty, the percentage of 16- to 19-year-olds in school, and the median gross rent—to characterize
multiple aspects of neighborhood quality, specifically dimensions related to the labor market, education,
income, and housing quality. 25
23
The block group information that we attach to each case record is based on the dimensions identified in
Feins (2003) and include: percentage of persons in poverty, percentage of households receiving public assistance
income, percentage of female-headed families with children, percentage of high school dropouts, unemployment
rate, labor force participation rate, percentage of families with no workers, percentage of people with incomes twice
the poverty level, percentage of people with education beyond high school, percentage of 16- to 19-year-olds in
school, percentage of housing stock that is owner-occupied, median family income, racial composition, median
house value, and median gross rent.
24
Because the 2000 Census served as our data source on neighborhood quality at the block group level, our
measures of neighborhood quality are not dynamic in nature. Consequently, any observed changes in neighborhood
quality are attributable to household relocation, rather than a change in neighborhood quality over time.
25
Examining the bivariate correlations between these measures suggests that we were largely successful in
identifying different dimensions of neighborhood quality. Although the neighborhood unemployment rate and the
percentage of persons in poverty correlate at approximately 0.7, all remaining correlations fall below 0.3, with most
below 0.2.The precise correlations are available from the authors upon request. Taken together, this suggests the
21
After merging the census block group data into the existing CARES/UI dataset, we have
complete data for 12,170 cases in the voucher group and over 342,000 nonvoucher, or potential
comparison, cases for up to six years after voucher receipt. 26 As we have indicated, this sample is broad
and includes both urban and rural residents, households with a wide variety of racial and ethnic
backgrounds, and cases with a variety of familial compositions, from single individuals to married
couples with multiple children.
5.2.
Identification strategy
Identifying the effect of voucher receipt on labor market outcomes requires leveraging the more than
342,000 potential comparison cases in a manner that results in a comparison group indistinguishable from
the voucher group on all characteristics except voucher receipt, or where any differences between the two
groups can be accounted for in a manner that renders treatment assignment ignorable. Put another way,
treatment status—voucher receipt—must be exogenous. Failure to achieve treatment exogeneity leaves
open the possibility that any observed relationships between voucher receipt and labor market outcomes
are spurious.
We have several options for creating a comparison group meeting these conditions. We ultimately
determined that a propensity score matching procedure represented the best approach for creating a
comparison group that permitted valid causal inferences regarding the effect of voucher receipt on
casehead employment and earnings. 27 In its essence, propensity score matching involves estimating the
existence of multiple dimensions of neighborhood quality and opens the possibility of heterogeneous effects of
voucher receipt across these dimensions.
26
Readers may note that the number of cases for which we have complete data (12,170 voucher cases and
over 342,000 potential comparison cases) is not equal to the sum of rental subsidy and non-rental subsidy cases in
Table 2. This is because Table 2 presents demographic characteristics for all cases extracted from the CARES
database. Not all of these cases were able to be linked to UI earnings records or geocoded to a census block group.
27
The primary papers describing propensity score matching approaches include Rosenbaum and Rubin
(1983); Heckman, Ichimura, Smith, and Todd (1996, 1998); Heckman, Ichimura, and Todd (1997); and Smith and
Todd (2005). Applications of the method include Dehejia and Wahba (1999, 2002); Lechner (2002); Hotz, Imbens,
and Klerman (2002); and Dyke, Heinrich, Mueser, Troske, and Jeon (2006).
22
probability that a specific case will receive the treatment (voucher receipt), given that case’s observed
characteristics, and then matching each treatment case to one or more potential comparison cases with
similar, or ideally identical, propensity scores. The goal of this procedure is to generate a comparison
group that is balanced with the treatment group on all characteristics that may induce a spurious
correlation between treatment status and the outcome being studied; that is, to generate a sample of cases
where treatment status is exogenous. 28 After using a propensity score procedure to generate the matched
comparison group, we isolated the effect of voucher receipt on casehead labor market outcomes—
earnings and employment—in a multivariate framework.
Specifically, we use a difference-in-differences regression model containing a wide variety of
covariates to estimate the effect of voucher receipt on labor market outcomes. By combining propensity
score matching with regression analysis we exploit the advantages of each method while mitigating the
limitations associated with each procedure. Indeed, prior research has found that combining propensity
score matching with regression adjustment is preferable to applying either method by itself (Imbens and
Wooldridge, 2008; Gelman and Hill, 2007). Given the nature of our sample and the estimation approach,
our results reflect the impact of the receipt of a Section 8 voucher on labor market outcomes for a sample
of recipient caseheads, relative to a group of matched comparison caseheads that received no housing
assistance.
In Appendix A, we describe our procedures in more detail.
6.
Mobility Effects: No Housing Assistance Counterfactual
Prior to analyzing the effect of voucher receipt on other behaviors, we estimate the effect of housing
voucher receipt on the probability of the household changing residence within one year and within four
28
The theoretical properties underlying propensity score matching guarantee only that the procedure has
the ability to eliminate differences on all observed characteristics used in the propensity score estimation; this
procedure does not have the ability to correct for bias stemming from unobserved variables. As we describe in
greater detail below, propensity score matching procedures have been shown to be particularly effective in reducing
bias stemming from nonrandom treatment assignment under a number of conditions that we are confident our
procedure meets. Our belief that our approach produces unbiased estimates of the effect of voucher receipt on labor
market outcomes is supported by the results of several sensitivity analyses noted below.
23
years after the end of the month in which the case first received the housing voucher. Because changing
residential location is a potential mechanism through which the other outcomes of voucher recipients may
be altered, assessing the extent to which voucher receipt causes residential location change is relevant to
our analysis of the effect of voucher receipt on these other effects. As noted above, the CARES database
includes information on the address of each family, enabling us to identify changes in residence.
As expected, the receipt of a housing voucher substantially increases the probability of changing
residential location. For the new voucher recipients, 58 percent of voucher recipients had moved within
one year after voucher receipt, compared to 44 percent of matched nonrecipient families. By four years
after voucher receipt, 77 percent and 69 percent of the two groups, respectively, had moved. When
estimated in a regression framework, the marginal effect of voucher receipt on the probability of moving
within one year of receipt is 0.131; the marginal effect on moving within four years of receipt is 0.107. 29
Of course, voucher recipients’ behavioral effects may not only be affected by whether they move,
but also where they move; the neighborhoods into which voucher recipients relocate may influence their
subsequent behaviors. In other work, we have estimated the effect of voucher receipt on the quality of
neighborhoods in which voucher holders reside, as measured by census block group characteristics
(Carlson et al., 2011). Using four indicators of neighborhood quality—the unemployment rate, the
percentage of persons in poverty, the percentage of 16- to 19-year-olds in school, and the median gross
rent—we find voucher receipt to have little effect on neighborhood quality one year after receipt; relative
to a matched comparison group, voucher recipients live in neighborhoods with slightly lower
unemployment rates and slightly lower percentages of 16- to 19-year-olds in school, but the effects are
substantively small. There are no statistically significant differences between voucher recipients and the
matched comparison group on the other two indicators of neighborhood quality. Four years after receipt,
voucher recipients lived in census block groups with a significantly lower percentage of persons in
29
Similar results are observed for each of the demographic subgroups. The results on geographic movement
by demographic subgroup are available from the authors.
24
poverty and continue to reside in neighborhoods with lower unemployment rates. The magnitude of these
effects, however, is rather small—for example, a 0.28 percentage point lower poverty rate and a 0.18
percentage point lower unemployment rate.
7.
Labor Market Effects: No Housing Assistance Counterfactual
We estimate the effect of receiving a Section 8 voucher on both the annual quarters worked and earnings
of recipient caseheads using the entire sample of observations. The estimated model is equation (1), a
generalized least squares difference-in-differences estimation with a case-level random effect.
7.2.1.
Full-sample employment and earnings effects
In results presented in Table 3 we see that, across all voucher recipients, there is no statistically
significant change in quarters worked over the first five years after voucher receipt. However, in the sixth
year after receipt, quarters worked increased by a statistically significant but substantively small .04
quarters per year; the average casehead receiving a Section 8 voucher had 2.4 percent more quarters
worked per year than the mean comparison group casehead in the sixth year after voucher receipt.
Table 4 presents our results for the earnings effect of voucher receipt. For the full sample, we
estimate that receipt of a housing voucher results in an average decline of about $650 in casehead
earnings in the initial year after voucher receipt. This corresponds to about 12 percent of the average
earnings for the matched comparison cases. 30 Over subsequent years, however, the negative earnings
effect is reduced, and by sixth year after voucher receipt, was less than $100. All of the annual negative
differences are statistically significant except that for the sixth year after voucher receipt. 31
The differential pattern of employment and earnings results is interesting. While voucher receipt
does not affect the number of quarters worked by recipient caseheads, it does cause a significant decline
30
Average earnings of the voucher group were not statistically different from the average earnings of the
matched comparison group in the year prior to treatment. Hence, we assess the voucher group’s average earnings
decline due to receipt of a Section 8 voucher as a percentage of the matched comparison group’s average earnings.
31
These results are generally consistent with the findings of the Welfare to Work evaluation. Jacob and
Ludwig (2006, 2008) find that in the Chicago Public Housing experiment, voucher receipt caused a drop in earnings
of approximately 10 percent over the post-receipt evaluation period. See Table 1.
25
in casehead earnings in the short-term, although earnings recover over the longer run. This is suggestive
of a scenario where, in the short-term, caseheads reduce the number of hours they work, but do not cease
employment altogether; perhaps caseheads give up a second job that had been required in order to meet
the rent. However, as time passes, the negative earnings effect dissipates while employment remains
largely unchanged. There are several potential explanations for this pattern. Perhaps caseheads transition
into better-paying jobs over time. Alternatively, it could be that some cases went off voucher receipt and
reverted to pre-voucher labor market behavior, which is likely attributable to the return of the nonvoucher incentive structure (i.e. no income threshold, lower implicit marginal tax rate, no income effect).
[Tables 3 and 4 here]
Family-type subgroup earnings effects of voucher receipt
We estimated the mean effect of voucher receipt on casehead earnings by family type. We hypothesized
that caseheads with larger obstacles to work (e.g., one-parent families) would have larger negative work
and earnings effects.
This hypothesis is largely supported by our estimates. For each family subgroup, the effect of
voucher receipt on casehead earnings is large and negative in the year of receipt. The largest absolute
declines are for single parents (-$833) or minus 21 percent. This is followed by those of couples with
children whose average decline was -$784, far larger than those of households with no child (-$333 on
average). For all of the family subgroups, the negative earnings effects are reduced over time. The pattern
of reductions is not straightforward: single parents fade out quickly, and are only -$309 in year 4. For
couples with children, the reduction is greatest among the three subgroups (-$544) after 4 years. By year
six earnings effects are negative but statistically insignificant for both groups. 32
32
Although we believe a counterfactual of no housing receipt is most relevant, analysis of the effects of
voucher receipt under an alternative counterfactual may be interesting and important. For example, it might be
useful to estimate the effect that “vouchering out” public housing projects would have on the labor market behavior
of recipients. Consequently, we present estimates of the labor market effects of voucher receipt under a second
counterfactual—the receipt of public housing. These estimates are presented in Appendix B.
Comment [MSOffice2]: Che
ck out.
26
Interestingly, while overall there are almost no employment effects for these family type
subgroups, the single exception is the increase in employment for single parents after 5 years (+.06). Thus
for the family type with the greatest obstacle to employment and earnings (single parents), receiving a
voucher appears to eventually lead to greater employment.
8.
Household Composition Effects: No Housing Assistance Counterfactual
In our estimation of the effect of voucher receipt on household composition, we first present results on the
number of adults and number of children living in households. Then, we study the basis for these
observed ‘count effects’ for adults by estimating the effect of voucher receipt on the addition and loss of
adults.
8.1
Effects on number of adults and number of children
Table 5 summarizes our findings concerning the effect of voucher receipt on the number of adults and the
number of children in the household. Full sample results are presented in the first bank of the table; the
remaining banks show effects on age subgroups of the population. We analyze these outcomes for up to
five years after initial voucher receipt.
[Insert table 5 about here]
Our estimates reveal that one year after receipt, voucher receipt led to sizable and statistically
significant reductions in the number of adults. Voucher receipt is also estimated to result in a decrease in
the number of children, but the estimate is substantively small. The reduced number of adults is observed
for the subsequent five years, though the magnitude of the effect diminishes across time. Beginning two
years after voucher receipt, effect of voucher receipt on the number of children became positive, and the
magnitude of the estimated effect increased in each succeeding year. Five years after voucher receipt, the
effect of voucher receipt on the reduction in the number of adults was nearly equal to its effect on the
increase in the number of children.
For households headed by younger people (ages 18-31), the large, statistically significant, and
negative treatment year impact on the number of adults is reduced in subsequent years, though it remains
negative and statistically significant through the observation period. Five years after voucher receipt,
27
older household groups recorded decreases in the number of adults, but no significant effect on the
number of children. Additional subgroup results are available from the authors upon request.
8.2
Effects on addition/losses of adults
In this analysis, shown in Table 6, we separately estimate the effect of voucher receipt on adding and
losing an adult member of the case. This analysis allows us to determine whether the reduction in the
number of adults is primarily attributable to voucher recipients, relatively to the comparison group,
actively shedding adult members from the case or simply being less likely adding additional adult
members. The table presents the estimated marginal effects for the full sample, and age subgroups.
[Insert table 6 about here]
For the full sample, receipt of a housing voucher stimulates significant changes in household
composition in the initial year of receipt. Consistent with the findings in Table 5, voucher receipt is
estimated to increase the predicted probability of losing an adult by more than 5 percentage points in the
treatment year. This effect is both statistically and substantively significant, given that the comparison
group mean in the year of receipt is only 15.4 percent. This effect turns negative and statistically
significant, albeit substantively small, for three of the next five years. Similarly, voucher recipients were
also significantly less likely than cases in the matched comparison group to add an adult in the year of
voucher receipt. Taken together, it appears that initial receipt of a voucher spurs recipients to reconsider,
and generally decrease, the number of adults living in the household. Over the period of observation, the
number of adults in the household appears to stabilize, but at a lower level.
In the treatment year, households with heads in the youngest age category (ages 18-30) have
sizeable and significant positive marginal effect (6.9 percentage points) of voucher receipt on the
probability of losing an adult, and a sizable and significant negative effect (4.6 percent) on the probability
of adding an adult. Both effects work to decreasing the number of adults in these households in the year
of voucher receipt, relative to the matched comparison group. The initial negative effect on adding an
adult is followed by an additional negative effect in the first post-treatment year, and then a positive effect
in the fifth year after voucher receipt. These patterns are consistent with the overall effect of voucher
28
receipt in significantly reducing the number of adults in the households of those in the youngest age group
over the observation period. For the older age groups, the large and significant positive treatment year
effects on the probability of losing an adult and the large and significant negative effect on the probability
of adding an adult are followed by generally small and statistically insignificant marginal effects on these
outcomes in subsequent years.
Taken as a whole, the results indicate that, relative to the matched comparison group, the lower
number of adults in voucher cases is primarily attributable to these cases actively dropping adults in the
initial year of receipt. However, a reduced probability of adding adults among voucher cases is
undoubtedly responsible for a smaller portion of the observed lower number of adults, at least in the
initial year of receipt.
9.
Effects on Child Care Utilization: No Housing Assistance Counterfactual
Table 7 summarizes the effects of voucher receipt on the use of state-subsidized child care services in the
treatment year and over the following three years. The sample for this analysis only includes cases that
contained at least one child under the age of 12 in each of the four years we analyze. The effects are
estimated using the regression model described above, and summarized in the footnote to the table.
Voucher receipt appears to have a statistically significant effect on the use of state-subsidized
child care services, both in the treatment year and in subsequent years. Over the four-year period, the
proportion of voucher families receiving child care subsidies consistently exceeds that of the matched
comparison group families by about 4-6 percentage points. This difference in usage is statistically
significant. Among the users of state-subsidized child care services, voucher recipients receive $200$400 more subsidy per year than matched comparison group members. Voucher receipt does not appear
to have a significant effect on the choice between licensed (generally center-based) and certified
(generally less expensive) care among families who receive child care subsidies.
[Insert table 7 here]
29
Table 8 shows the effect of voucher receipt on state-subsidized child care use and subsidy values
for families that have a child less than five years old in each of the years studied. The overall effects on
child care use are again positive and statistically significant, but substantively larger for families with
younger children. The usage rate of public vouchers exceeds that of matched comparisons by 5-7
percentage points over the years studied, or by about 10-15 percent. While families with young children
using state- subsidized child care services are estimated to receive somewhat greater subsidies than
matched controls, the differences are not statistically significant. Again, among child care users, housing
voucher receipt does not appear to influence the choice of child care services between licensed care and
certified care.
[Insert table 8 here]
10.
Supplemental Analyses and Robustness Tests
Like all propensity score-based analyses, ours is subject to concerns that the results may be biased due to
unobserved heterogeneity between the voucher and matched comparison groups. Hence, we have
undertaken several sensitivity analyses and robustness checks; these diagnostic procedures provide
confidence that our propensity score matching approach has provided unbiased and reliable estimates of
the effect of voucher receipt on labor market outcomes. These additional tests, which also provide
insights into the mechanisms of voucher receipt and behavioral effects, include:

A pre-/post-voucher receipt comparison involving two cohorts of voucher recipients,

A case-level fixed effects estimate,

A test of potential bias due to selection on unobservables,

Estimates based on alternative propensity score matching techniques,

Estimates of behavioral effects using OLS estimation,

Effects by duration of voucher holding, and

Effects by the nature of program-induced neighborhood change.
Taken as a whole, the results of all these sensitivity analyses and robustnesss tests are similar to our
propensity score-based results and provide confidence in our primary results. These robustness tests and
supplementary analyses are available from the authors, upon request.
30
11.
Discussion and Conclusion
We have used a longitudinal dataset containing information on more than 350,000 low-income
households in a medium-sized, diverse state to study the initial and long-term mobility, and a variety of
other behavioral effects of the receipt of a Section 8 housing voucher. Our data have comprehensive
household—and in some cases, individual—demographic, income, and benefit receipt information
extending over six years, and have been merged with data on employment, earnings, and geographic
location. In our primary analysis, we identified a large sample of families that received a housing voucher
during calendar years 2001–2003 and, using propensity score matching techniques, a comparison group
that received no housing assistance. However, we also performed a secondary analysis in which we
estimated the labor market effects of voucher receipt against a counterfactual of public housing receipt.
In both analyses, using a difference-in-differences regression framework coupled with propensity
score-based techniques, we studied the employment and earnings effects of voucher receipt over a sixyear period. By analyzing impacts for a diverse and large group of low-income families, we extend
findings from prior studies typically based on households that have lived in public housing or in medium
to large urban areas. We distinguish these effects by family type subgroups. We tested the robustness of
our results with multiple alternative specifications and other tests, and distinguished impact by the
duration for which voucher receipt is retained and by the nature of program-induced geographic moves.
11.1
Effects on mobility and neighborhood quality
Consistent with prior research, we find that voucher receipt leads to a significantly higher initial
and long-term probability of changing residence, relative to the matched comparison group.
Voucher receipt leads to some improvements in neighborhood quality in the long term, but
appears to have little effect in the short term. Evaluated one year after initial receipt, voucher recipients
exhibit no improvement on three of our four neighborhood quality indicators. However, observations
made four years after initial receipt indicate gains on all four indicators of neighborhood quality for
voucher recipients. These results are consistent with a scenario in which voucher recipients require some
31
time to learn about the new housing options available to them, but once recipients have evaluated the new
options, they make decisions to reside in relatively better neighborhoods.
11.2
Labor market effects
With respect to labor market effects, we found a negligible impact of voucher receipt on work effort
(quarters worked per year) in the years immediately after voucher receipt, although after six years,
voucher recipients recorded a statistically significant, but substantively small, gain in work effort relative
to the matched comparison group. We also find that the early-year impact of voucher receipt on casehead
earnings is negative (about 12 percent of mean casehead comparison group earnings). However, this
negative earnings effect decreased over the five years after voucher receipt; by the end of the six-year
observation period the earnings effects were not significantly different from zero.
Our analysis reveals that single parent families experienced a greater initial loss of earnings than
those without children or couples with children and that couples with children experienced a greater loss
of earnings than households without children. Over time, however, single parents were able to take more
advantage of employment opportunities and experienced a greater fade-out of the negative earnings
effects. In our alternative analysis using public housing receipt as the counterfactual, we find that the
overall labor market effects of Section 8 voucher receipt do not differ significantly from those of public
housing in Wisconsin.
Our primary results generally confirm earlier research that uses a similar counterfactual of no
public housing assistance. Like previous experimental analyses that have a counterfactual of no receipt of
housing assistance—such as WtW and the CHA experiment—our analysis finds voucher receipt to have a
negative effect on casehead earnings. A careful nonexperimental study of the labor market effects of
project-based assistance using a similar counterfactual (Newman, Holupka, and Harkness, 2009) found a
similar pattern: statistically significant declines in earnings in the second and third year after movement
into public housing relative to their matched comparisons who received no housing assistance, but no
statistically significant differences in subsequent years. Similarly, the results of our secondary analysis—
where public housing is specified as the counterfactual—are generally consistent with experimental
32
studies that have employed this counterfactual, specifically Gautreaux and MTO. The fact that the results
of our two analyses, which draw on the same data source and employ similar analytical techniques but
specify different counterfactuals, are consistent with the results of comparable experimental studies is
illuminating. The alignment makes clear that the labor market impacts of a marginal expansion of the
Section 8 voucher program depend on whether the counterfactual is assumed to be receipt of public
housing assistance or receipt of no housing assistance. This finding goes some way toward reconciling the
discrepant experimental results regarding the effect of voucher receipt on employment and earnings. It
also provides some indication of the mechanisms that may be at work. Specifically, the results suggest
that the location change spurred by voucher receipt has little effect on earnings or employment; both
MTO and our analysis that specifies public housing as the counterfactual find no effect while the
Gautreaux experiment finds a small positive effect. The results of our primary analysis, along with the
WtW and CHA experiments and the study by Newman, Holupka, and Harkness (2009), indicate that the
increase in household income and the marginal tax rate that accompany voucher receipt exert a substantial
negative impact on recipient casehead earnings, and this effect initially overwhelms any positive labor
market outcomes stemming from location change. 33
11.3
Effects on household composition
Voucher receipt appears to generate household composition change in the year of receipt, but greater
stability in subsequent years. The change that occurs in the initial year is mainly characterized by a
reduction in the number of adult members of the household. For some, the voucher may allow recipients
to leave unproductive relationships and to establish an independent household. For others, the voucher
33
This analysis may also provide insight into why the negative earnings effect of voucher receipt persists
over the entire period of observation in the CHA experiment, but fades to insignificance in the WtW experiment.
Our duration analysis suggests that cases remaining on voucher assistance exhibit a negative earnings pattern
through the full period of voucher receipt while cases that go off assistance return to their pre-voucher earnings
behavior. Data from HUD’s A Picture of Subsidized Households-2008 dataset indicate that the average duration of
receipt for cases receiving their voucher from the Chicago Housing Authority is nearly nine years, which is much
longer than the average duration of receipt in the six WtW sites. As a result, the differential results between the
CHA and WtW studies may be attributable to recipients remaining on assistance for a much longer time period in
Chicago, relative to cases in the WtW sample.
33
may entail leaving a parental residence and establishing their own living arrangement. Both of these
changes suggest that voucher receipt improves overall well-being, reflecting the additional resources that
come with voucher receipt.
11.5
Effects on child care usage
We find that voucher recipient families with children less than 12 years of age exhibit a greater
participation rate in state-subsidized child care programs than do matched comparison families who
have not received a voucher, and a greater subsidy amount. Somewhat surprisingly, the voucher
program does not seem to have resulted in an increased use of licensed (center–based) care. While this
overall pattern also holds for families with a child less than age 5, it is only statistically significant for
the overall use of childcare benefits.
34
Appendix A
Detailed Matching and Estimation Procedures
Obtaining a Comparison Group: A Propensity Score Matching Approach
Propensity score matching procedures have been shown to be effective in minimizing potential bias due
to nonrandom selection into treatment under several conditions that our procedure meets. 34 First,
propensity score matching procedures work best when a large number of potential comparison cases are
available, relative to the number of treatment cases. As described above, our estimation sample contains
12,170 voucher cases and over 342,000 potential comparison cases for a nearly 30:1 ratio of potential
comparison to treatment cases. Few propensity score applications have had a more desirable distribution
of treatment and potential comparison cases. Second, propensity score matching has been shown to
achieve reliable results when treatment cases are matched to comparison cases that are demographically,
geographically, and contextually similar. Our matching procedures contain variables measuring all these
dimensions. Demographically, we include variables measuring race/ethnicity, age, sex, education, and
several other characteristics. Geographically, all cases reside in Wisconsin and, in addition, we match on
county of residence. Contextually, we match on approximately 20 census block group variables that
measure important neighborhood characteristics. As a result, the treatment and matched comparison
groups are statistically indistinguishable on these important dimensions. Third, propensity score matching
is most effective when extensive baseline (pre-treatment) information on the outcomes to be studied is
available for all of the cases. Our matching model contains casehead earnings and employment,
neighborhood quality and household composition information for up to five years prior to voucher
receipt. We include multiple variables measuring these dimensions in the model used to estimate each
case’s propensity score. In addition, we include dozens of other variables that are likely to be predictive
of Section 8 voucher receipt. As a result, there are no pretreatment differences between the treatment and
matched comparison groups on the outcome measures. In sum, we believe that these characteristics
provide an ideal setting for application of a propensity score matching procedure to estimate the
behavioral effects of voucher receipt.
Upon estimation of the model used to estimate the propensity score for each case, we turned our
attention to identifying the specific matching strategy that we would use in our analysis. After considering
the advantages and drawbacks of several matching strategies, we chose to employ a nearest neighbor
matching strategy in our primary analysis. 35 This strategy matches each treatment case to one or more
nonrecipient cases with the closest (or ideally identical) propensity scores. In the context of this analysis,
each treatment case is matched to the nearest five nonrecipient cases. 36 We match with replacement,
which means that each nonrecipient case can be matched to more than one treatment case.
34
See, for example, Heckman, Ichimura, and Todd (1997); Smith and Todd (2005); and Pirog, Bufardi,
Chrisinger, Singh, and Briney (2009).
35
There are several matching strategies available in the propensity score literature—including “nearest
neighbor,” “kernel,” and “local linear regression” methods. Discussions of various matching metrics and methods
can be found in Mueser, Troske, and Gorislavsky (2007) and Smith and Todd (2005). We also estimated a kernel
matching model and the results were almost identical to those shown here. Results of the estimation are available
from the authors.
36
Nearest neighbor matches were also performed with each treatment case matched to the 1, 3, and 10
nearest neighbors. The results, which are available from the authors, did not differ substantively.
35
As the final step in the matching procedure, we used several metrics to evaluate the quality of the
match. First, we conducted a balance test, which indicated that the matching procedure succeeded in
eliminating pretreatment differences between the voucher and matched comparison groups on every
covariate included in the model used to estimate the propensity score. Second, we compare the
pretreatment behavioral effect trajectories for the treatment and matched comparison groups. Figure A1
compares the voucher and matched comparison group employment and earnings patterns during the fiveyear period immediately before potential receipt of a voucher. As seen there, employment and earnings
patterns during the period prior to potential receipt (nonreceipt) of a voucher are virtually identical for the
voucher and matched control group observations. Taken as a whole, these diagnostic procedures indicate
that based on observed characteristics, the matching procedure succeeded in generating a comparison
group that allows for unbiased and reliable estimates of the behavioral effects of Section 8 voucher receipt
on casehead earnings and employment. 37
Difference-In-Differences Estimates Of The Effect Of Voucher Receipt On Casehead Earnings,
Employment, Neighborhood Quality And Household Composition
Using the balanced sample created through the propensity score matching procedure, we isolate the effect
of voucher receipt on our behavioral effect variables using a difference-in-differences regression
framework. Specifically, we estimate the following model:
(1)
where i and t index cases and the year of effect we are measuring with respect to voucher receipt,
respectively; Y represents the outcome; α represents the intercept; V is a dummy variable indicating
voucher receipt in the treatment year; R represents the calendar year; X is a vector of observed
characteristics of the case and casehead; 38 A represents the year of effect relative to the treatment year;
and C represents a case-level random effect. The model is estimated via generalized least squares (GLS),
with random effects controls. For each year, the total estimated effect of voucher receipt on the relevant
behavioral effect is equal to the sum of 1 and 5 .
37
We also performed balance tests for each of our subgroup matches. The results, which are available from
the authors, show no statistically significant pre-receipt differences on any variable for any subgroup.
38
Specifically, the vector of observed characteristics includes the sex, race/ethnicity, age, marital status,
and baseline earnings of the casehead as well as the presence of a disabled or elderly individual in the case and a
measure of change in the composition of the case.
36
Figure A1. Mean Casehead Earnings and Quarters Worked: 1 to 5 Years Prior to Voucher Receipt
Mean Casehead Earnings Prior to Voucher Receipt
4000
4500
Dollars
5000
5500
6000
by Receipt Status
-5
-4
-3
Year Relative To Receipt
Matched Comparisons
-2
-1
Voucher Group
Mean Casehead Quarters Worked Prior to Voucher Receipt
1
1.5
Quarters
2
2.5
3
by Receipt Status
-5
-4
-3
Year Relative To Receipt
Matched Comparisons
-2
Voucher Group
-1
37
Appendix B
Effect of Voucher Receipt on Casehead Earnings and Employment by year, Full
sample (Dollars): Public Housing Assistance Counterfactual
Earnings
Employment
Year
(Dollars)
(Quarters worked)
Year of receipt
-137.50
(121.71)
-0.006
(0.030)
One year post receipt
-89.84
(155.36)
0.015
(0.035)
Two years post receipt
27.40
(177.78)
0.031
(0.036)
Three years post receipt
62.91
(177.53)
-0.013
(0.037)
Four years post receipt
-27.14
(233.38)
-0.029
(0.045)
Five years post receipt
-395.64
-0.037
(362.59)
(0.069)
Note: Robust standard errors clustered by case in parentheses below point estimates. Point
estimates in bold are statistically significant at alpha = .10 level.
38
References
Altonji, J. G., Elder, T. E., Taber, C. R. 2005. Selection on observed and unobserved variables: assessing
the effectiveness of catholic schools. Journal of Political Economy 113, 151–184.
Bania, N., Coulton, C., Leete, L. 2003. Public housing assistance, public transportation, and the welfareto-work transition. Cityscape: A Journal of Policy Development and Research 6, 7–44.
Carlson, D., Haveman, R., Kaplan, T., Wolfe, B. 2011. Long-term effects of public low-income housing
vouchers on neighborhood quality and household composition. Working Paper, Institute for
Research on Poverty, University of Wisconsin-Madison.
Dehejia, R., Wahba, S. 1999. Causal effects in nonexperimental studies: reevaluating the evaluation of
training programs. Journal of the American Statistical Association 94, 1053–1062.
Dehejia, R., Wahba, S. 2002. Propensity score matching methods for nonexperimental causal studies.
Review of Economics and Statistics 84, 151–161.
Diaz, J. J., Handa, S. 2006. An assessment of propensity score matching as a nonexperimental impact
estimator: Evidence from Mexico’s PROGRESA program (Programa de Educacion, Salud y
Alimentacion). Journal of Human Resources 41, 319–345.
Dyke, A., Heinrich, C. J., Mueser, P. R., Troske, K. R., Jeon, K. S. 2006. The effects of welfare-to-work
program activities on labor market outcomes. Journal of Labor Economics 24, 567–607.
Elder, T., Jepsen, C. 2011. Are Catholic primary schools more effective than public primary schools?
Working Paper. Retrieved May 16, 2011 from http://www.njea.org/njeamedia/pdf/CatholicvsPublicSchools.pdf
Eriksen, M. D. 2009. The market price of low-income housing tax credits. Journal of Urban Economics
66, 141-149.
Eriksen, M. D., Rosenthal, S. S. 2010. Crowd-out effects of place-based subsidized rental housing: new
evidence from the LIHTC Program. Journal of Public Economics 94, 953-966.
Feins, J. D. 2003. A cross-site analysis of MTO’s locational impacts. In Goering, J., Feins, J. D. (Eds.),
Choosing a Better Life? Evaluating the Moving to Opportunity Social Experiment. Washington,
D.C., Urban Institute Press.
Gelman, A., Hill, J. 2007. Data Analysis Using Regression and Multilevel/Hierarchical Models. New
York, NY, Cambridge University Press.
Goering, J. 2003. Comments on future research and housing policy. In Goering, J., Feins, J. D. (Eds.),
Choosing a Better Life? Evaluating the Moving to Opportunity Social Experiment. Washington,
D.C., Urban Institute Press.
Goering, J., Feins, J. D., Richardson, T. 2002. A cross-site analysis of initial moving to opportunity
demonstration results. Journal of Housing Research 13, 1–30.
39
Harkness, J. M., Newman, S. J. 2006. Recipients of housing assistance under welfare reform: trends in
employment and welfare participation. Housing Policy Debate 17, 81–108.
Heckman, J., Ichimura, H., Todd, P. E. 1997. Matching as an econometric evaluation estimator: evidence
from evaluating a job training programme. Review of Economic Studies 64, 605–654.
Heckman, J., Ichimura, H., Smith, J., Todd, P. E. 1996. Sources of selection bias in evaluating social
programs: an interpretation of conventional measures and evidence on the effectiveness of
matching as a program evaluation method. Proceedings of the National Academy of Sciences 93,
13416–13420.
Heckman, J., Ichimura, H., Smith, J., Todd, P. E. 1998. Characterizing selection bias using experimental
data. Econometrica 66, 1017–1098.
Hirano, K., Imbens, G. 2001. Estimation of causal effects using propensity score weighting: an
application of data on right ear catheterization. Health Services and Outcomes Research
Methodology 2, 259–278.
Hotz, V. J., Imbens, G., Klerman, J. A. 2002. The long-term gains from GAIN: a re-analysis of the impact
of the California GAIN program. Working paper, Department of Economics, UCLA.
Imbens, G., 2004. Nonparametric estimation of average treatment effects under exogeneity: a review.
Review of Economics and Statistics 86, 4–30.
Imbens, G., Wooldridge, J. 2008. Lecture 1: estimation of average treatment effects under
unconfoundedness, part I. Lecture presented at the Applied Microeconometrics Workshop,
August 4–6, Institute for Research on Poverty, Madison, WI. Retrieved March 10, 2009, from
http://www.irp.wisc.edu/newsevents/workshops/appliedmicroeconometrics/participants/notes/rev
_lect_1-G.pdf
Jacob, B. A., Ludwig, J. 2006. The effects of means-tested housing assistance on labor supply: new
evidence from a housing voucher lottery. Presentation prepared for a Joint Center for Housing
Studies symposium, Revising Rental Housing, A National Policy Summit, Harvard University,
November.
Jacob, B. A., Ludwig, J. 2008. The effects of housing assistance on labor supply: evidence from a voucher
lottery. Working Paper 14570. Cambridge, MA, National Bureau of Economic Research.
Retrieved March 11, 2009, from http://www.nber.org/papers/w14570.pdf
Katz, B., Turner, M. A. 2000. Who should run the housing voucher program? A reform proposal.
Washington, DC: Brookings Institution, Center on Urban and Metropolitan Policy.
Kling, J. R., Liebman, J. B. Katz, L. F. 2007. Experimental analysis of neighborhood effects.
Econometrica 75, 83–119.
Lechner, M. 2002. Some practical issues in the evaluation of heterogeneous labour market programmes
by matching methods. Journal of the Royal Statistical Society, series A 165 (part 1), 59–82.
Mendenhall, R., DeLuca, S., Duncan, G. 2006. Neighborhood resources, racial segregation, and economic
mobility: results from the Gautreaux Program.” Social Science Research 35, 892–923.
40
Michalopoulos, C., Bloom H. S., Hill, C. J. 2004. Can propensity-score methods match the findings from
a random assignment evaluation of mandatory welfare-to-work programs? Review of Economics
and Statistics 86, 156–179.
Mills, G., Gubits, D., Orr, L., Long, D., Feins, J., Kaul, B., Wood, M. 2006. Effects of Housing Vouchers
on Welfare Families. Washington, D.C.: U.S. Department of Housing and Urban Development,
Office of Policy Development and Research. Retrieved October 8, 2010 from
http://www.huduser.org/publications/commdevl/hsgvouchers.html
Mueser, P. R., Troske, K. R., Gorislavsky, A. 2007. Using state administrative data to measure program
performance. Review of Economics and Statistics 89, 761–783.
Newman, S., Holupka, C. S., Harkness, J. 2009. The long-term effects of housing assistance on work and
welfare. Journal of Policy Analysis and Management 28, 81-101.
Olsen, E. O., Tyler, C. A., King, J. W., Carrillo, P. E. 2005. The effects of different types of housing
assistance on earnings and employment. Cityscape: A Journal of Policy Development and
Research 8, 163–187.
Orlebeke, C. R. 2000. The evolution of low-income housing policy, 1949 to 1999. Housing Policy Debate
11, 489-520.
Pirog, M. A., Bufardi, A. L., Chrisinger, C. K., Singh, P., Briney, J. 2009. Are the alternatives to
randomized assignment nearly as good? Statistical corrections to nonrandomized evaluations.
Journal of Policy Analysis and Management 28, 169–172.
Popkin, S., Buron, L., Levy, D., Cunningham, M., 2000. The Gautreaux legacy: what might mixedincome and dispersal strategies mean for the poorest public housing tenants? Housing Policy
Debate 11, 911–942.
Rosenbaum, J. E. 1995. Housing mobility strategies for changing the geography of opportunity. Housing
Policy Debate 6, 231–270.
Rosenbaum, J. E., DeLuca, S. 2000. Is housing mobility the key to welfare reform? Lessons from
Chicago’s Gautreaux Project. Washington, D.C., Brookings Institution Center on Urban and
Metropolitan Policy.
Rosenbaum, P., Rubin, D. 1983. The central role of the propensity score in observational studies for
causal effects. Biometrika 70, 41–55.
Ross, C., Reynolds, J., Geis, K. 2000. The contingent meaning of neighborhood stability for residents’
psychological well-being. American Sociological Review 65, 581–597.
Shroder, M. 2002a. Does housing assistance perversely affect self-sufficiency? A review essay. Journal of
Housing Economics 11, 381–417.
Shroder, M. 2002b. Locational constraint, housing counseling, and successful lease-up in a randomized
housing voucher experiment. Journal of Urban Economics 51, 315–338.
Sinai, T., Waldfogel, J. 2005. Do low-income housing subsidies increase the occupied housing stock?
Journal of Public Economics 89, 2137-2164.
41
Smith, J., Todd, P.E. 2005. Does matching overcome Lalonde’s critique of nonexperimental estimators?
Journal of Econometrics 125, 305–353.
Susin, S. 2005. Longitudinal outcomes of subsidized housing recipients in matched survey and
administrative data. Cityscape: A Journal of Policy Development and Research 8, 189–218.
Swartz, R., Miller, B. 2002. Welfare reform and housing. Welfare Reform and Beyond Policy Brief 16.
Washington, D.C., Brookings Institution.
Turney, K., Clampet-Lundquist, S., Edin, K., Kling, J. R., Duncan, G. 2006. Neighborhood effects on
barriers to employment: results from a randomized housing mobility experiment in Baltimore. In
Burtless G., Rothenberg Pack, J. (Eds.), Brookings-Wharton Papers on Urban Affairs 2006.
Washington, D.C., Brookings Institution.
U.S. Department of Housing and Urban Development. 2003. Moving to Opportunity interim impacts
evaluation. Washington, D.C., U.S. Department of Housing and Urban Development.
U.S. Department of Housing and Urban Development. 2008. A Picture of Subsidized Households- 2008.
Retrieved August 20, 2010, from http://www.huduser.org/portal/picture2008/index.html.
Van Ryzin, G. G., Kaestner, R., Main, T. J. 2003. The effects of federal and local housing programs on
the transition from welfare to work: evidence from New York City. Cityscape: A Journal of
Policy Development and Research 6, 45–72.
42
Tables and Figures
Table 1a. Summary of Experimental Literature on Relationship between Housing Assistance Receipt and Labor Market Outcomes
Study
Comparison Groups
Location and Timing
Data and Methodology
Effects on Employment or
Earnings
Gautreauxa
Housing project residents who
moved to suburbs; project
residents who left the projects
but remained in the city
Chicago; 1976–1998
Experiment; possible internal
validity concerns
Movers to suburbs had
higher earnings
Moving to
Opportunityb
Housing project residents who
moved to low-poverty
neighborhoods; project residents
who moved to any neighborhood;
project residents who remained
in project housing
Housing assistance
distributed to 4,600 families
in 5 cities (Baltimore,
Boston, Chicago, Los
Angeles, and New York);
1994–1998
Participants were randomly
assigned to a voucher that
could only be used in areas
with poverty rates below 10
percent, a voucher that could
be used in any rental unit that
accepted a Section 8 voucher,
or a control group that
remained in housing project
No significant differences;
study is not yet completed.
However, Web appendices 7
and 8 of Kling, Liebman,
and Katz (2007) show
positive earnings impacts for
younger adults who were
required to move to lowpoverty neighborhoods.
Welfare-to-Workc
Welfare recipients or eligible
recipients who received a
housing voucher; welfare
recipients or eligible recipients
who did not receive a voucher
8,700 families in 6 cities
(Atlanta, Augusta, Fresno,
Houston, Los Angeles, and
Spokane) assigned to
treatment and control
groups between spring 2000
and spring 2001
Random assignment to a
housing voucher or to a
control group that received no
housing voucher through the
program
Voucher recipients initially
worked and earned less, but
the differences disappeared
after 3.5 years; Study is not
yet completed
Chicago Housing
Authority
Experimentd
Members of a housing assistance
waiting list who were randomly
selected to receive a housing
voucher; wait-list members not
selected to receive a housing
voucher
82,600 households in
Chicago in July 1997 were
placed on a waiting list for
Section 8 vouchers and
assigned a position on the
waiting list from 1997–2003
by random assignment
Random assignment of
households that applied for
Section 8 to a housing
voucher or to no housing
voucher
Voucher recipients worked
and earned significantly less
than those who did not
receive a voucher over the
full follow-up period
a
See Rosenbaum (1995); Popkin, Buron, Levy, and Cunningham (2000); Rosenbaum and DeLuca (2000); and Mendenhall, DeLuca, and Duncan (2006).
See Goering, Feins, and Richardson (2002); Shroder (2002b); Goering (2003); Turney, Clampet-Lundquist, Edin, Kling, and Duncan (2006); and Kling,
Liebman, and Katz (2007).
c
See Mills et al. (2006).
d
See Jacob and Ludwig (2006, 2008).
b
43
Table 1b. Summary of Nonexperimental Literature on Relationship between Housing Assistance Receipt and Labor Market Outcomes
Effects on Employment or
Earnings
Study
Comparison Groups
Location and Timing
Data and Methodology
Bania, Coulton, and
Leete (2003)
Welfare leavers receiving Section
8 vouchers and project-based
housing; other welfare leavers
Cleveland/Cuyahoga County;
1996 followed through 1997
Administrative data; Regression
model
No significant effect; no
difference between voucher
and project-based assistance
recipients
Susin (2005)
Housing assistance recipients,
both voucher and project-based;
low-income nonrecipients
Nationally representative
sample; 1996 followed
through 1999
SIPP combined with administrative
data on housing assistance receipt;
Propensity score matching
Earnings of housing subsidy
recipients reduced by about
15 percent; no difference
between voucher and
project-based assistance
recipients
Van Ryzin, Kaestner,
and Main (2003)
Welfare recipients receiving
Section 8 vouchers; welfare
recipients receiving project
housing assistance; welfare
recipients receiving no housing
assistance
New York City; 1995–1996
Local survey data; Logistic
regression controlling for a variety of
observed differences
No significant effect on
earnings of voucher
recipients relative to those
with no assistance
Olsen, Tyler, King, and
Carrillo (2005)
Housing assistance recipients;
housing project residents; lowincome households receiving no
housing aid
Nationally representative
sample; 1995–2002
PSID combined with HUD data on
recipients; Regression model with
income ceiling imposed on
comparison group
Harkness and Newman
(2006)
Single mothers receiving projectbased housing assistance or
vouchers; single mothers
receiving no assistance
Female-headed households who
moved to project-based housing;
matched (through propensity
scoring) female-headed
households who received no
housing assistance
National HUD data; 1996 and
2001
HUD data combined with CPS for
comparison group; Simple
comparison of univariate results with
some focus on regression
PSID combined with HUD data on
project-based recipients; propensity
score matching
Earnings of housing subsidy
recipients decline by $3,600
in first two years of
program participation
relative to those with no
housing assistance
No significant effect
Newman, Holupka, and
Harkness (2009)
Nationally representative
sample, 1970-1995. Housing
assistance recipients were
followed for 2 years before
beginning assistance to 6
years after; the comparisons
were followed in same years
Earnings of households that
moved into public housing
were significantly lower
than matched comparisons
in the 2nd and 3rd years after
public housing entry; no
significant effect after that
44
Table 2. Demographic Characteristics for Those Who Receive Rental Subsidies and Those Who Do Not
Receive Rental Subsidies: 2001–2003 Cohorts
2001
2002
2003
Do Not
Do Not
Do Not
Receive
Receive
Receive
Receive
Receive
Receive
Rent
Rent
Rent
Rent
Rent
Rent
Characteristic
Subsidy
Subsidy
Subsidy
Subsidy
Subsidy
Subsidy
Total Number of Cases
6,159
163,391
6,080
187,276
5,383
216,064
Sex
Male
Female
15.6
84.4
24.8
75.2
15.2
84.8
26.7
73.3
15.3
84.7
28.0
72.1
Age
18–30
31–45
46–59
60+
42.9
29.9
12.6
14.2
37.9
36.0
13.5
12.1
47.0
30.2
12.3
10.3
38.4
36.1
14.2
10.8
48.0
28.3
14.0
9.4
39.1
35.8
14.9
9.7
Race
White
Black
Hispanic
Other race
59.3
29.0
3.3
8.5
48.2
36.2
6.2
9.4
58.7
28.5
3.7
9.2
48.9
35.6
6.3
9.2
60.4
26.9
3.1
9.6
50.9
34.5
6.3
8.4
Education Level
No high school diploma
High school diploma
35.4
64.6
39.7
60.3
34.1
65.9
37.9
62.1
33.2
66.8
36.3
63.7
Marital Status
Single, never married
Divorced or annulled
Separated
Married
Widowed
50.5
20.4
11.3
10.7
7.0
50.2
17.9
10.7
15.1
6.0
52.0
21.4
11.2
10.3
5.1
51.4
17.7
10.4
15.3
5.2
52.0
21.7
11.0
10.7
4.6
52.7
17.5
9.8
15.5
4.5
County Urbanicity
Rural
Urban
Milwaukee
28.0
45.9
26.1
21.6
31.0
47.4
26.9
51.8
21.3
21.6
32.0
46.4
30.6
51.3
18.1
22.2
33.7
44.1
Number of Children
0
1
2
3+
39.8
25.6
19.1
15.5
44.9
21.2
16.5
17.4
36.1
27.6
19.6
16.8
45.9
21.3
16.3
16.4
35.9
27.2
20.1
16.9
48.1
21.1
15.6
15.2
45
Table 3. Effect of Voucher Receipt on Quarters Worked, by Demographic Subgroup: No Housing Assistance
Counterfactual
Subgroup
Full Sample
Family Composition
Couple with child
Single parent
No child
Year of
Receipt
One Year Post
Two Years
Post
Three Years
Post
Four Years
Post
Five Years
Post
-0.02
(0.01)
0.01
(0.01)
0.01
(0.01)
0.02
(0.01)
0.02
(0.02)
0.04
(0.02)
-0.07
(0.05)
-0.02
(0.02)
-0.05
(0.02)
-0.04
(0.05)
0.02
(0.02)
-0.07
(0.02)
-0.02
(0.05)
0.01
(0.02)
-0.07
(0.02)
-0.05
(0.05)
0.01
(0.02)
-0.05
(0.02)
-0.02
(0.06)
0.02
(0.02)
-0.03
(0.03)
-0.04
(0.07)
0.06
(0.03)
-0.03
(0.03)
46
Table 4. Effect of Voucher Receipt on Earnings, by Family status: No Housing Assistance Counterfactual
Subgroup
Full Sample
Family Composition
Couple with child
Single parent
No child
Year Of
Receipt
One Year
Post
Two Years
Post
Three Years
Post
Four Years
Post
Five Years
Post
-647.02
(70.62)
-558.33
(70.62)
-412.54
(70.61)
-309.42
(70.60)
-221.99
(78.87)
-97.49
(100.59)
-783.98
(228.62)
-833.12
(103.06)
-333.34
(83.57)
-723.78
(228.63)
-672.05
(103.04)
-367.85
(83.58)
-355.62
(228.63)
-583.85
(103.02)
-387.91
(83.57)
-321.97
(228.64)
-519.66
(103.00)
-300.14
(83.58)
-544.31
(259.97)
-309.06
(114.87)
-247.56
(93.49)
-422.93
(336.57)
-147.04
(147.02)
-179.89
(117.75)
Note: Standard errors in parentheses below point estimates. Point estimates in bold are statistically significant at alpha
= .10 level.
47
Table 5. Effect of voucher receipt on number of adults and children in the case-Full sample &
subgroups
Year of
One year Two years
Three
Four
Five years
Case composition
Receipt
post
post
years post years post
post
Full Sample
Number of adults
-0.102***
-0.104***
-0.085***
-0.077***
-0.064***
-0.049***
(0.007)
(0.007)
(0.007)
(0.007)
(0.007)
(0.009)
Number of children
-0.014**
(0.007)
-0.006
(0.008)
Number of adults
-0.140***
(0.010)
-0.125***
(0.010)
Number of children
-0.021**
(0.009)
-0.019*
(0.011)
Number of adults
-0.079***
(0.010)
-0.083***
(0.011)
Number of children
0.036***
(0.011)
0.040***
(0.013)
Number of adults
-0.052***
(0.012)
-0.054***
(0.012)
-0.016*
(0.009)
-0.008
(0.009)
Number of children
0.014*
0.024***
(0.008)
(0.009)
Age 18-30
-0.087***
-0.077***
(0.010)
(0.010)
0.004
(0.012)
0.014
(0.014)
Age 31-54
-0.089***
-0.086***
(0.011)
(0.012)
0.033**
(0.014)
0.021
(0.014)
Age 55+
-0.040***
-0.039***
(0.013)
(0.013)
-0.011
(0.010)
-0.012
(0.008)
0.039***
(0.010)
0.053***
(0.013)
-0.057***
(0.010)
-0.045***
(0.012)
0.028*
(0.014)
0.062 ***
(0.019)
-0.077***
(0.012)
-0.064***
(0.016)
0.025*
(0.015)
0.021
(0.019)
-0.037***
(0.012)
-0.050***
(0.014)
-0.009
(0.009)
-0.010
(0.011)
Note: Standard errors in parentheses below point estimates. *p<.10, **p<.05, ***p<.01.
48
Table 6. Marginal effect of voucher receipt on case composition change-Full sample and subgroups
Year of
One year Two years
Three
Four
Five years
Case composition
Receipt
post
post
years post years post
post
Full Sample
Add Adult
-0.027***
-0.012***
0.000
-0.001
0.005**
0.008***
(0.003)
(0.003)
(0.003)
(0.003)
(0.002)
(0.003)
Lose Adult
0.053**
(0.004)
-0.007**
(0.003)
Add Adult
-0.046***
(0.005)
-0.016***
(0.005)
Lose Adult
0.069***
(0.006)
-0.021***
(0.005)
Add Adult
-0.015***
(0.005)
-0.007
(0.004)
Lose Adult
0.038***
(0.006)
Add Adult
Lose Adult
-0.014***
(0.003)
-0.008**
(0.003)
-0.006**
(0.003)
0.000
(0.004)
0.004
(0.004)
0.014***
(0.005)
-0.010**
(0.004)
0.002
(0.005)
Age 31-54
-0.007
-0.001
(0.004)
(0.004)
0.001
(0.004)
0.007
(0.004)
0.005
(0.006)
0.000
(0.005)
-0.002
(0.005)
0.000
(0.005)
-0.002
(0.006)
-0.009*
(0.005)
-0.003
(0.004)
Age 55+
0.002
-0.001
(0.004)
(0.003)
0.000
(0.003)
-0.005***
(0.002)
0.027***
(0.007)
0.003
(0.006)
-0.007
(0.005)
-0.004
(0.004)
-0.002
(0.005)
Age 18-30
0.001
-0.005
(0.005)
(0.004)
-0.025***
(0.005)
-0.012**
(0.005)
-0.001
(0.005)
Note: Standard errors in parentheses below point estimates. *p<.10, **p<.05, ***p<.01.
49
Table 7.
Regression results for effect of voucher receipt on child care use, subsidy amount, and licensed
child care use for all cases with at least one child under age 12 in the year of receipt and each of
the three following years
Proportion/
Difference
Standard
p- value
Mean of
from
error
voucher
matched
recipients
controls
Year
Child care use
Year of voucher receipt
0.5029
0.0464
0.0063
0.000
One year post-receipt
0.4794
0.0603
0.0063
0.000
Two years post-receipt
0.4328
0.0615
0.0062
0.000
Three years post-receipt
0.3768
0.049
0.0061
0.000
Among users-Child care subsidy amount (dollars)
Year of voucher receipt
7461.68
228.56
138.78
0.100
One year post-receipt
8127.90
435.37
132.71
0.001
Two years post-receipt
7253.96
313.84
126.16
0.013
Three years post-receipt
5965.73
341.76
122.87
0.005
Among users- Licensed child care use
Year of voucher receipt
0.8234
0.0082
0.0102
0.420
One year post-receipt
0.8372
0.0080
0.0099
0.421
Two years post-receipt
0.8427
0.0026
0.0096
0.791
Three years post-receipt
0.8234
0.0056
0.0102
0.585
Note: Regressions used to estimate difference from matched controls for child care use and licensed
care use included controls for the casehead's age, sex, years of education, marital status, race/ethnicity,
number of children, number of case members, county of residence, and case composition changes.
These results are not presented here, but are available from the authors. Regressions used to estimate
difference from matched controls for subsidy amount included a control for child care use in addition to
all of the variables listed above.
50
Table 8.
Regression results for effect of voucher receipt on child care use, subsidy amount, and licensed child care use for all cases with at least one
child under age 5 in the year of receipt and each of the three following years
Year
Proportion/
Mean of
voucher
recipients
Difference
From
Matched
controls
Standard
error
P-value
Child care use
0.5277
0.0495
0.0081
0.000
0.5402
0.0672
0.0082
0.000
0.5159
0.0716
0.0082
0.000
0.4681
0.0595
0.0082
0.000
Among users-Child care subsidy amount (dollars)
7387.55
58.96
161.18
0.715
Year of voucher receipt
8388.67
217.25
156.64
0.166
One year post-receipt Two
7560.85
47.13
148.15
0.750
years post-receipt
6407.96
212.57
143.85
0.140
Three years post-receipt
Among users- Licensed child care use
0.8282
0.0157
0.0119
0.188
Year of voucher receipt
0.8402
0.0032
0.0116
0.784
One year post-receipt
0.8491
0.0060
0.0112
0.593
Two years post-receipt
0.8394
-0.0055
0.0115
0.634
Three years post-receipt
Note: Regressions used to estimate difference from matched controls for child care use and licensed care use included controls for the casehead's
age, sex, years of education, marital status, race/ethnicity, number of children, number of case members, county of residence, and case composition
changes. These results are not presented here, but are available from the authors. Regressions used to estimate difference from matched controls
for subsidy amount included a control for child care use in addition to all of the variables listed above.
Year of voucher receipt
One year post-receipt
Two years post-receipt
Three years post-receipt