Organization of the Report

문서에서 Family Options Study (페이지 42-0)

Chapter 1. Introduction

1.4 Organization of the Report

This first chapter of the report has provided an overview of family homelessness, the homeless services system, the evaluation design, and the baseline characteristics of the research sample. The balance of this report is organized as follows. Chapter 2 describes the interventions studied and the implementation of the Family Options Study. Chapter 2 also explains site selection and examines the characteristics of the study sites and the process used to conduct random assignment. Chapter 3 discusses the conceptual framework of the interventions and hypotheses about their potential effects. Chapter 4 presents the methodology and data sources.

Chapter 5 describes the experiences of the usual care group.

It also defines the outcomes derived from participant surveys and administrative data that are used to estimate intervention

effects. Chapters 6 through 9 then present findings about the impacts of the four interventions, organized by the six pairwise comparisons. In particular, Chapter 6 provides im - pact measures for SUB compared to UC, for the five domains of housing stability, family preservation, adult well-being, child well-being, and self-sufficiency. Chapter 7 presents findings from the comparison of CBRR to UC in the five domains and Chapter 8 does so for the comparison of PBTH to UC. Chapter 9 turns to the other pairwise comparisons, reporting impacts of SUB compared to CBRR, SUB compared to PBTH, and CBRR compared to PBTH. Chapter 10 discusses results about the relative impacts of groups of interventions based on pooled comparisons to illuminate other policy ques - tions. Chapter 11 explores the variability of impacts across types of families, using indices related to psychosocial challenges and housing barriers constructed for each family.

Chapter 12 describes the relative costs of the interventions.

Chapter 13 discusses study conclusions. Several technical appendixes support the report. Appendix A provides details about the data sources and dataset construction. Appendix B discusses the construction of adult and child well-being out-comes. Appendix C presents technical details regarding the samples and analysis methods. Survey nonresponse analysis is documented in Appendix D. Appendix E contains sup-plemental tables showing use of transitional housing during the followup period. Appendix F presents exhibits showing the results of the pooled comparisons. Appendix G presents technical details about the cost data collection and analysis.

21 The statistic for PTSD is the national 12-month prevalence rate as measured in the National Comorbidity Survey Replication (NCS-R), which was fielded in 2001 and 2002. The NCS-R used a different instrument to measure PTSD than what was used in the Family Options Study.

22 The statistic for the national rate of serious psychological distress is from the 2011 National Health Interview Survey. This survey used the same measure of psychological distress that was used in the Family Options Study.

23 Percentages are of respondents who reported that a past eviction or no rent history at all presented a “big” or “small” problem for them in finding a place to live. The survey items did not capture whether a respondent had a past eviction or had no rent history at all if the respondent did not think these factors were problems in finding a place to live. Therefore, these percentages are lower bounds on the proportions of the respondent sample who had a history of eviction and who had never been a lease-holder.




his chapter discusses the implementation of the Family Options Study. It begins with an overview of the study interventions and the contrasting fea - tures that were envisioned in the study design. The next section addresses site recruitment and describes key charac-teristics of the 12 study sites. The remainder of the chapter then describes how the study team implemented random assignment.

2.1 Interventions Studied

The Family Options Study examines four interventions. The study team collaborated closely with HUD during the design phase of the study to determine what types of interventions should be studied and to define the distinguishing features.

The study team defined the interventions to include con-trasts in the type and duration of housing assistance and the presence of supportive services. The four interventions were defined as follows.

1. Subsidy (SUB) was defined as a permanent housing subsidy, usually a housing choice voucher (HCV). SUB could have included the assistance to find a unit that qualified for the voucher program that might be available to anyone who receives voucher assistance, but it did not include other supportive services.

2. Community-based rapid re-housing (CBRR) was intended to provide temporary rental assistance for 2 to 6 months (potentially renewable for up to 18 months) paired with limited, housing-focused services to help families find and rent conventional, private-market housing.

3. Project-based transitional housing (PBTH) was intended to provide temporary housing (for up to 24 months with aver - age expected stays of 6 to 12 months) in agency-controlled buildings or apartment units paired with intensive supportive services.

4. Usual care (UC) is defined as any housing or services that a family accesses in the absence of immediate referral to the other interventions. UC typically includes at least some addi - tional stay in the emergency shelter from which families were enrolled.

The intended contrasts across interventions in types of hous - ing subsidies and of services, are shown in Exhibit 2-1.

Detailed findings from the pairwise comparisons of each active intervention (SUB, CBRR, and PBTH) to UC are presented in Chapters 6, 7, and 8. To assess the nature of the housing and services offered to families assigned to the interventions, the introductions to Chapters 6, 7, and 8 describe how the interventions were implemented in the study sites. Those descriptions use information about housing and services col - lected from the programs selected to operate the interventions.

The intervention assessments also make use of data about the extent to which families received services from the assigned intervention and the duration of that assistance.

2.2 Site Selection

After defining the distinguishing features of the study interventions, the study team recruited sites. To select and recruit the sites, the study team canvassed a large group of

Exhibit 2-1. Intended Contrasts in Subsidy and Services for the Family Options Interventions and Usual Care Group

Housing Subsidy Services Provided

Somea None

Some Heavy Light

Permanent SUB

Temporary PBTH CBRR

None UCb

CBRR = community-based rapid re-housing. PBTH = project-based transitional housing. SUB = permanent housing subsidy. UC = usual care.

a Outside of the heavy and light distinction, the nature of services also may differ between PBTH and CBRR. CBRR focuses on services to help with locating housing, leasing up, and settling in. By contrast, PBTH provides more comprehensive social services, such as assessments, provision of and referral to job-related services, counseling, substance abuse treatment, and family- and child-oriented services. See Sections 5-1, 6-1, 7-1, and 8-1 and Chapter 12 for details about the types of services offered.

b UC was intended to involve other assistance that families accessed on their own after emergency shelter. In many emergency shelters extensive services are provided. Information about services offered in emergency shelters is provided in Section 5-1 and in Chapter 12.

Short-Term Impacts of Housing and Services Interventions for Homeless Families 12 Chapter 2. Implementing the Study

Continuum of Care (CoC) programs across the country to locate cities, counties, and metropolitan areasin which the number of families entering emergency shelter was consid-ered sufficient to achieve enrollment goals and where the intervention models defined for the study were present.

Providers of SUB, CBRR, and PBTH in selected communities had to be willing to implement a random assignment evalu-ation. Participating in the study meant that programs had to agree to commit program slots to families in the study and to comply with random assignment as the method of deter-mining which families would be referred to their programs from participating emergency shelters. The team worked closely with the CoC and local homeless system leaders to collect information about the homeless assistance system and then negotiated with program providers and public housing agencies (PHAs) to determine whether a sufficient number of program slots in each intervention were available in the site to make the study viable in the community.

The study team recruited 12 sites to conduct the study. The sites are displayed in Exhibit 2-2 with information about the number of CoCs and the geography covered by each site.

By definition, all sites had the UC option available in their communities, because the study sample was recruited from emergency shelters, and UC was defined as assistance that families accessed after a 7-day stay in emergency shelter with - out priority access to the active interventions (SUB, CBRR, and PBTH). The study team initially sought to select sites that had all three active interventions available. The study team subsequently determined that it would be acceptable to include some sites in which only two of the interventions were available. With assistance from HUD, the study team was able to obtain agreements with PHAs to make housing subsidies available for randomization to the SUB interven-tion in 10 of the 12 sites. HUD and local CoC stakeholders assisted the study team to secure CBRR slots for the study that might not otherwise have been available to families in shelter. Thus, the study increased the resources available to families who were experiencing homelessness in emergency shelters in participating communities.

Part of the site recruitment process involved confirming that all programs included in the study were good representa-tives of their defined intervention. The study team started by defining the distinguishing features of the interventions,

Exhibit 2-2. Family Options Study Sites

Site Name Municipal Areas

(cities/counties/geographic area)

Included in the Study CoCs Included in the Site

Alameda County, California Berkeley CA-502 Oakland/Alameda County CoC

Fremont Hayward Oakland Alameda County

Atlanta, Georgia Atlanta GA-500 Atlanta Tri-County CoC

Baltimore, Maryland Baltimore MD-501 Baltimore City CoC

Boston, Massachusetts Boston MA-500 Boston CoC

Connecticuta Bridgeport CT-503 Bridgeport/Stratford/Fairfield CoC

New Haven CT-501 New Haven CoC

Norwalk CT-506 Norwalk/Fairfield County CoC

Stamford CT-508 Stamford/Greenwich CoC

Fairfield County

Denver, Colorado Denver CO-503 Metropolitan Denver Homeless Initiative

Honolulu, Hawaii Island of Oahu HI-501 Honolulu CoC

Kansas City, Missouri Kansas City MO-604 Kansas City/Independence/Lee’s Summit/Jackson County CoC Jackson County

Louisville, Kentucky Louisville KY-501 Louisville/Jefferson County

Jefferson County

Minneapolis, Minnesota Minneapolis MN-500 Minneapolis/Hennepin County

Hennepin County

Phoenix, Arizona Phoenix AZ-502 Phoenix/Mesa/Maricopa County Regional CoC

Maricopa County

Salt Lake City, Utah Salt Lake City UT-500 Salt Lake City & County

Salt Lake County CoC = Continuum of Care.

a The Connecticut site includes multiple metropolitan areas in the state.

Sources: Site recruitment data and HUD; CoC designations reported reflect designations in effect in September 2010 when study enrollment began; since that time, some CoCs have been reorganized and renamed

such as the housing assistance subsidy duration and level and presence of dedicated services linked to the housing assistance. The challenge in this endeavor was that short-hand terms used by practitioners and researchers, such as “transitional housing” or “supportive housing,” do not reflect uniform approaches. In reality, as Rog and Randolph (2002) noted, even when programs of a particular “type”

are specifically chosen for study, their characteristics can overlap considerably with other programs that nominally use an approach labeled in a different way. To address this challenge, during initial site selection, the team visited potential study programs (and interviewed some by phone), collected data on their operations, and completed an assess-ment for each candidate program. The study team selected programs that fit the study’s definitions of the interventions based on these assessments, rather than based on programs’


Before selecting programs to participate in the study, the study team identified minimum requirements for a program to be considered an example of each intervention. Selecting programs that met these requirements assured that families enrolled in the study would receive comparable levels of housing assistance and service support within an interven-tion regardless of site differences. Such comparability in turn allowed for the evaluation to test the outcomes associ-ated with being randomly assigned to distinct interventions across multiple sites. Overall, the data collected from the participating programs confirm that the interventions were distinct from each other in the ways intended by the study’s design (see Sections 5-1, 6-1, 7-1, and 8-1). Exhibit 2-3 tabulates the number of providers for each intervention that agreed to participate in the study at each site.

As indicated in the exhibit, all four interventions were offered in 9 sites. Two sites (Atlanta and Baltimore) did not offer SUB and one site (Boston) did not offer PBTH.

Exhibit 2-3. Number of Programs, by Site and Intervention


Alameda County, California 3 1 7 9

Atlanta, Georgia 4 7 4

Baltimore, Maryland 2 5 3

Boston, Massachusetts 1 2 8

Connecticuta 3 2 3 9

Denver, Colorado 2 1 3 5

Honolulu, Hawaii 2 6 7 6

Kansas City, Missouri 1 5 3 3

Louisville, Kentucky 1 1 4 3

Minneapolis, Minnesota 1 1 2 1

Phoenix, Arizona 2 1 4 5

Salt Lake City, Utah 2 1 1 1

Total 18 27 46 57

CBRR = community-based rapid re-housing. PBTH = project-based transitional housing. SUB = permanent housing subsidy. UC = usual care.

a The Connecticut site includes multiple metropolitan areas in the state.

Source: Study program data

2.3 Characteristics of Participating Sites

The 12 study sites represent a diverse range of geographic locations, size, population, and housing and labor market characteristics. Gubits et al. (2013) provides a detailed review of population, income, and labor market conditions in the participating sites. To provide context for the impact analysis, this section discusses select housing market char-acteristics and rates of homelessness across the 12 sites at the time study enrollment began in 2010.24 Although not a

randomly selected sample of communities, the sites are varied in geography and conditions that are related to homeless-ness. The sites are located in all four of the Census Bureau- designated regions in the country. Exhibit 2-4 displays the geographic coverage of the sites.

Housing market characteristics offer insight into the con - ditions for obtaining housing in each of the 12 study sites (see Exhibit 2-5). The rental vacancy rate serves as an indi-cator of how difficult it may be for a family to obtain rental housing.

24 Ten of the study sites began enrollment in the fall of 2010; two sites (Baltimore and Louisville) began enrolling families into the study during the spring of 2011.

Short-Term Impacts of Housing and Services Interventions for Homeless Families 14 Chapter 2. Implementing the Study

Exhibit 2-4. Location of Study Sites

Exhibit 2-5. Housing Market Characteristics of Study Sites

Site Rental Vacancy Rate (% 2010) Median Monthly Gross Rent ($ 2010)

Alameda County, Californiaa 5.6 1,198

Atlanta, Georgia 16.4 892

Baltimore, Maryland 7.5 874

Boston, Massachusetts 5.4 1,233

Connecticutb 12.3 1,047

Denver, Colorado 5.5 811

Honolulu, Hawaii 6.1 1,171

Kansas City, Missouria 13.8 738

Louisville, Kentuckya 9.2 670

Minneapolis, Minnesotaa 6.1 861

Phoenix, Arizonaa 11.7 884

Salt Lake City, Utaha 7.5 832

United States 8.2 855

a Because these sites operated at the county level, the data presented are for the county where the study site is located.

b The site includes the cities of Bridgeport and New Haven, Connecticut. The figures shown are averages for the two cities.

Source: 2010 American Community Service 1-Year Estimates, U.S. Census Bureau

Areas with lower rental vacancy rates are considered less likely to have affordable rental housing. In 2010, the nation - wide rental vacancy rate was 8.2 percent. Among the 12 study sites, Boston had the lowest rental vacancy rate (5.4 percent) and Atlanta had the highest rental vacancy rate (16.4 percent).

Another 6 sites—Alameda County, Baltimore, Denver, Honolulu, Minneapolis, and Salt Lake City—had vacancy rates between 5 and 8 percent, and 3 had vacancy rates between 11 and 14 percent.

In 2010, the national median monthly gross rent was $855.

Of the 12 study sites, 6—Atlanta, Baltimore, Denver, Min - neapolis, Phoenix, and Salt Lake City—had median rents between $800 and $900, similar to the national rate. Another 2 sites—Kansas City and Louisville—had rates lower than the national average, and 4 sites—Alameda County, Boston, Connecticut, and Honolulu—had median rents above $1,000, well above the national average. Boston had the highest median rent of all 12 sites, at $1,233.

Family Options Study: Short-Term Impacts of Housing and Service Interventions for Homeless Families

Short-Term Impacts ▌pg. 20 Exhibit 2-4. Location of Study Sites

Housing market characteristics offer insight into the conditions for obtaining housing in each of the 12 study sites (see Exhibit 2-5). The rental vacancy rate serves as an indicator of how difficult it may be for a family to obtain rental housing.

Areas with lower rental vacancy rates are considered less likely to have affordable rental housing. In 2010, the nationwide rental vacancy rate was 8.2 percent. Among the 12 study sites, Boston had the lowest rental vacancy rate (5.4 percent) and Atlanta had the highest rental vacancy rate (16.4 percent).

Another 6 sites—Alameda County, Baltimore, Denver, Honolulu, Minneapolis, and Salt Lake City—had vacancy rates between 5 and 8 percent, and 3 had vacancy rates between 11 and 14 percent.


The prevalence of family homelessness also differed among the 12 study sites. To give a sense of the extent of homeless-ness from one study site to another, Exhibit 2-6 shows the proportion of the population that was homeless as reported in each study site for the point-in-time counts conducted in January 2011 as a percent of the study site’s total population.

Among the sites, Honolulu had the highest rate of homeless-ness (1.26 percent of the population), while Phoenix had the lowest (0.15 percent of the population).25 Of the 12 study sites, 9 had a higher incidence of homelessness than the national rate of 0.20 percent. The exhibit also shows the number of homeless families (overall, not merely in the study) and the number of people in these households in the point-in-time count. Boston had the highest number of homeless families (987 families) and Louisville had the lowest (134 families) reported. The high incidence in Boston may reflect a Massachusetts “right to shelter” policy for homeless fam-ilies, meaning that all families who apply for shelter, lack alternative housing options, and whose income does not exceed 115 percent of the federal poverty line are entitled

to shelter (Institute for Children and Poverty, 2010). The right-to-shelter policy might also increase lengths of stay in shelter in Boston.

Each site offers assistance to homeless families through emer - gency shelter and transitional housing programs. Exhibit 2-6 also shows the level of assistance available as measured by the number of emergency shelter and transitional housing beds that are dedicated to assisting people in homeless families.26 These figures provide an indication of the local homeless service system’s size and the relative prevalence of emergency shelter and transitional housing in each commu-nity’s system for families. One-third of the sites had excess emergency shelter and transitional housing capacity on the night of the point-in-time count in January 2011, whereas the other two-thirds were using overflow capacity or had families who were unsheltered (that is, families that had to stay in cars, on the streets, or in other private or public places not designed for or ordinarily used as a regular sleep-ing accommodation for human besleep-ings).27 Whereas roughly

Exhibit 2-6. Homeless Population in Study Sites

Site Total

Alameda County, Californiaa 1,510,271 0.28 376 1,136 447 852

Atlanta, Georgiab 1,612,474 0.42 365 1,073 484 1,489

Baltimore, Maryland 620,961 0.66 323 934 164 448

Boston, Massachusetts 617,594 0.89 987 2,926 2,648 435

Connecticutc 274,008 0.16 165 498 410 248

Denver, Colorado 600,158 0.80 924 2,609 727 1,635

Honolulu, Hawaii 337,256 1.26 558 2,235 675 1,733

Kansas City, Missouria 674,158 0.41 407 1,548 494 663

Louisville, Kentuckya 741,096 0.22 134 386 178 275

Minneapolis, Minnesotaa 1,152,425 0.27 467 1,572 1,279 823

Phoenix, Arizonaa 3,817,117 0.15 683 2,238 1,130 1,381

Salt Lake City, Utaha 1,029,655 0.20 241 827 322 479

United States 308,745,538 0.20 76,653 234,079 110,679 110,364

a County-level data are presented for these sites because the study was implemented in the county in which the metropolitan area is located.

b Represents the population of DeKalb and Fulton Counties, because CoC GA-500 includes this larger geography, including Atlanta.

c Represents the population of New Haven and Fairfield County, whereas the CoC data represents the four CoCs that participated in the study: CT 501 New Haven; CT 503 Bridge-port; CT-506 Norwalk-Fairfield; and CT-508 Stamford/Greenwich.

Sources: 2010 American Community Service 1-year estimates; U.S. Census Bureau and 2010 Decennial Census (total population figures); 2011 CoC Housing Inventory Chart and Homeless Populations and Subpopulations Data, HUD (HUD, 2011b)

25 The homeless data from point-in-time counts are reported by the CoC and thus do not always align precisely with the geography of the study site.

26 The housing inventory count of transitional housing beds includes all project-based and scattered-site transitional housing beds, because separate counts for the two

26 The housing inventory count of transitional housing beds includes all project-based and scattered-site transitional housing beds, because separate counts for the two

문서에서 Family Options Study (페이지 42-0)