Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
The algorithm of Black homelessness: a classification tree analysis
(USC Thesis Other)
The algorithm of Black homelessness: a classification tree analysis
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Hill
i
THE ALGORITHM OF BLACK HOMELESSNESS:
A CLASSIFICATION TREE ANALYSIS
by
Chyna Yvonne Vandon Hill
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degrees
MASTER OF SOCIAL WORK / DOCTOR OF PHILOSOPHY
(SOCIAL WORK)
August 2020
Copyright 2020 Chyna Yvonne Vandon Hill
ii
Dedication
Many of the people who prayed for me and prayed over me are no longer here. Even so, I
am happy that I was able to pull it together for the people who are. Black women and a couple of
Black men literally saved my life. To them, I am deeply and forever indebted. This is dedicated
to Chelsea Victoria Davis, Zachary John Heffner, and Tessa Yvonne Davis - who isn’t here yet
but is the little girl I always saw coming. This is also dedicated to my late Aunty Anna who
refused to let me fold my wings.
iii
Acknowledgements
This work would not be possible without my committee. Thank you to my chair, Dr.
Rice, whom I would not be here without. While I have a lot to thank you for, I think the most
monumental is being a consistent and genuine force in my life. Thank for believing in my ideas
and providing the mentorship I needed to execute them. Dr. Richardson, thank you for
demanding excellence. I appreciate you urging me to remain persistent and insisting that I had
something important to say. Dr. Kim, thank you for nurturing me, feeding me, and supporting me
along this journey.
I am grateful for the family members who were central to my personal and intellectual
growth. Granny, everything good in me…everything whole in me is because of you. In your
century on earth, you’ve loved me in ways that will transcend lifetimes. While everyone else
holds me to standards of perfection, you hold me to standards of grace. Thank you for
everything—including sticking around to see this. Woowoo, it is quite the adventure having you
as my grandmother. Thank you for being unapologetically you, unapologetically Black, and
unapologetically woman. There would not be a Dr. Chyna Yvonne Davis without Nicki Yvonne
Davis. Thank you for your lessons in strength and resilience. Undoubtedly, you are the strongest
and brightest of us all.
Mom, I don’t know who you were before you were my mother, but I do know that you
traded in whoever you were to be my mother. There is not enough gratitude in the world for that.
Thank you for every single sacrifice you made on my behalf. You did the best you could and
even if your “best” sucked sometimes, you ensured I’d blossomed no matter where you planted
me.
iv
Aunty, you’re the reason I still want to live in places I’ve never seen. You were my first
best friend and you’re still one of my favorite humans. One of the greatest things about our
relationship is it transcends illness. Although I couldn’t save you, I intend to save so many lives
because of you. I love you always.
Tish Johnson and Jen Johnson, thank you for loving me, mothering me, and being there
for me. Both of you have been the best parents I could ask for.
I am thankful for my community of women who supported me through this process.
Cassandra, Shanice, Ski, Ebony, Nneka, Victoria, Trese, and Becky, I could not ask for a better
group of friends. Thank you for dealing with the tantrums (and I literally mean tantrums – Ski, in
particular, because you were in the closest in proximity), picking out my outfits, doing my
makeup, feeding me, sending money, keeping me out of trouble, and being there for me
whenever I called. I’ve been a sucky friend over the last four years but thank you for continuing
to be my best friends anyway.
To my partner, Reggie, thank you for bringing me back to myself and shattering the
world I settled for. Thank you for loving me and being my partner is all things. I adore you.
Lastly, but of significant importance, I'm eternally indebted to my great grandfather, the
late Johnnie Albert Davis. You gave me everything, and while I couldn’t give you the
immortality I was aiming for, I hope this will suffice. I love you and I will miss you for the rest
of my life.
v
TABLE OF CONTENTS
Dedication ....................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iii
Tables ............................................................................................................................................ vii
Figures.......................................................................................................................................... viii
Abstract .......................................................................................................................................... ix
Chapter 1 ......................................................................................................................................... 1
The Origins of Homeownership Inequity.................................................................................... 3
Contemporary Disenfranchisement ............................................................................................. 8
The War on Drugs and Its Collateral Consequences................................................................... 9
Conclusion ................................................................................................................................. 10
Chapter 2 ....................................................................................................................................... 13
Method ...................................................................................................................................... 16
Measures ................................................................................................................................ 16
Analytic Plan ......................................................................................................................... 17
Classification Tree Analyses ................................................................................................. 18
Results ....................................................................................................................................... 19
Black Youth ........................................................................................................................... 22
Latino Youth .......................................................................................................................... 23
White Youth .......................................................................................................................... 24
All Youth ............................................................................................................................... 25
Discussion ................................................................................................................................. 35
Chapter 3 ....................................................................................................................................... 39
Methods ..................................................................................................................................... 42
Measures.................................................................................................................................... 43
Analytic Plan ......................................................................................................................... 44
Classification Tree Analyses ................................................................................................. 44
Results ....................................................................................................................................... 45
All Youth ............................................................................................................................... 48
Black Youth ........................................................................................................................... 49
Latino Youth .......................................................................................................................... 50
White Youth .......................................................................................................................... 51
vi
Discussion ................................................................................................................................. 61
Limitations ................................................................................................................................ 63
Conclusion ................................................................................................................................. 64
Chapter 4 ....................................................................................................................................... 66
Tool Disparities ......................................................................................................................... 69
Theoretical Foundations ............................................................................................................ 72
Recommendations ..................................................................................................................... 74
Family reunification as a false data point. ............................................................................. 75
The call for deeper data. ........................................................................................................ 76
Better assessment and evaluation of mitigating factors (e.g., family disruption). ................ 77
Cross-validation. .................................................................................................................... 79
Conclusion and Implications ..................................................................................................... 80
Bibliography ................................................................................................................................. 81
Appendices .................................................................................................................................... 95
Appendix A: NST Tool ............................................................................................................. 95
vii
Tables
2.1. Overall Descriptive Characteristics ....................................................................................26
2.2. Bi-Variate Analyses ............................................................................................................29
2.3. Feature Importance by Race & Ethnicity.............................................................................30
2.4 Fit Statistics by Race & Ethnicity ........................................................................................30
3.1. Overall Descriptive Characteristics ....................................................................................53
3.2. Bi-Variate Analyses ............................................................................................................55
3.3. Feature Importance by Race & Ethnicity.............................................................................56
3.4 Fit Statistics by Race & Ethnicity ........................................................................................56
viii
Figures
2.1. Predictors of Black Youth who Return to Family ............................................................. 31
2.2. Predictors of Latino Youth who Return to Family ............................................................ 32
2.3. Predictors of White Youth who Return to Family ............................................................. 33
2.4. Predictors of All Youth who Return to Family .................................................................. 34
3.1. Predictors of Remaining Stable for 180 Days or More for All Youth ............................... 57
3.2. Predictors of Remaining Stable for 180 Days or More for Black Youth........................... 58
3.3. Predictors of Remaining Stable for 180 Days or More for Latino Youth ......................... 59
3.4. Predictors of Remaining Stable for 180 Days or More for White Youth .......................... 60
ix
Abstract
To address rising levels of youth homelessness, communities around the country have
implemented continuums of care (COCs) to link youth to housing resources (Rice et al., 2018).
Recent work has found that formal exits allocated through COCs account for more than 30% of
youth exits from homelessness, with 25% of those exits attributed to rapid re-housing (RRH)
programs and another 5.8% attributed to permanent supportive housing (PSH) interventions
(Rice et al., 2018). Independent of the continuum of care models, youth are also exiting
homelessness by returning to family. Studies of youth homelessness suggest that familial
relationships are positively associated with increased likelihood of exiting homelessness as well
as maintaining stable housing over time among homeless youth (Milburn et al., 2009). Although
youth may exit homelessness in various ways, exits via formal housing programs or family are
the most visible. Therefore, the predictors of utilizing these exits and the subsequent stability of
these exits need intentional focus.
Furthermore, while homelessness is not exclusively Black, the factors that result in initial
housing loss and subsequent housing stability may be unique to those who identify as Black. A
central argument of this dissertation is that although several racial and ethnic communities
experience homelessness, the systemic inequities that drive homelessness impact Black people
differently and, in some cases, to a greater degree. This exploration leads to what I call the
“algorithm of Black homelessness,” which refers to a finite set of policies that, when attached to
identity markers such as race, systemically hinder Black people from remaining stably housed.
The findings from this dissertation will elucidate important policy implications and encourage
the development of assessment tools that capture the unique needs of Black youth experiencing
homelessness.
x
Keywords: Black youth homelessness, classification tree analyses, housing outcomes,
marginalization, algorithm of Black homelessness, algorithmic redlining
1
Chapter 1
On any given day, as many as 500,000 people experience homelessness in the United States
(Henry, 2019). While Black people comprise 13% of America’s general population, 40% of
America’s homeless population is Black (Jones, 2016; Carter, 2011; HUD 2018). Although all
racial and ethnic communities experience homelessness, it disproportionately impacts the Black
community. Studies of homelessness illustrate that Black populations have higher odds and
longer durations of homelessness (Henry, 2019; Morton et al., 2018; Choi et al., 2015). Among
Black adults, findings from a recent study suggest that relative to their White counterparts,
Blacks have higher rates of homelessness over their life course as well (Furaso et al., 2018).
Even when adjusting for a range of covariates, such as education, veteran status, and region, the
disparity between Black people who experienced homelessness and their White peers remained
significant (Furaso et al., 2018). The relationship between race and homelessness, therefore,
requires additional exploration.
In studies of homelessness, housing loss is consistently linked to individual-level factors such
as substance abuse, poor mental health, and abuse (Morton et al., 2018; Morton et al., 2017;
Barman-Adhikari et al., 2016; Barman-Adhikari & Rice 2014; Tyler, Akinyemi, & Butler, 2012;
Berdahl et al., 2005; Cauce et al., 2000; De Rosa et al., 1999; Gerber 1997). However, for people
who identify as Black and/or African American, there are a series of events that precede
homelessness. At the individual and family level, research suggests that these events are
reflective of generational loss (Freeman, 2018). That loss includes chronic unemployment, a lack
of familiarity with one’s surroundings due to displacement, and diminished social status and self-
esteem. For Black Americans who have been subject to systemic disadvantages and structural
inequalities, chronic and cumulative loss exacerbates the risk of homelessness (Olivet et al.,
2
2019). At the generational level, chronic and cumulative loss complicates an already wounded
family structure because of the necessity of other family members to fill in gaps and overextend
themselves mentally, emotionally, and financially. Historical disenfranchisement through
legislative processes has also created generational losses, which now appear normative and
endemic.
A premise of this dissertation is that Black homelessness is reflective of historic “algorithms”
that have been embedded in legislative processes. An algorithm can be defined as “a set of
instructions, rules, and calculations designed to solve problems” (Race after Technology, 2018,
pg. 11). In this dissertation, the use of the term “algorithm” refers to a set of preconceived
notions that have been used to dictate the legislative processes that have historically targeted and
disproportionately penalized the Black community. Since my argument centers around those who
identify or who have been identified as Black, I intend to introduce what I call, the “algorithm of
Black homelessness,” which I define as “an additive formula that assigns systematic biases to
Black or any identity attached to Black and produces homelessness as an output.” Further
explained, inherent biases are assigned to Black identities systematically and, once compounded,
yield homelessness. The collective sum of identities and the relationship of those identities to
each other plus the hierarchical placement of those identities in systems that are historically and
inherently unequal generate unique vulnerabilities among Black Americans experiencing
homelessness.
The first and central argument of this paper is that Black homelessness (and, subsequently,
Black youth homelessness) is the direct result of American oppressive legal processes that have
severely reduced the number of housing options available to Black Americans. This has resulted
in intergenerational consequences. While the individual-level factors that have been repeatedly
3
associated with homelessness are critical, the systemic factors that precede them are equally
important. Before I begin to examine the racial dimensions of youth homelessness, it is necessary
to first examine housing disenfranchisement through a historical lens. To reduce the number of
Black youth experiencing homelessness, it is imperative to understand the many pathways that
Black youth follow into homelessness. Then those pathways must be examined from within the
communities from which they originate. This chapter describes how codified systems have
introduced what I call the “algorithm of Black homelessness” and concludes with a discussion on
the implication of these restrictive laws on Black youth homelessness.
The Origins of Homeownership Inequity
For Black Americans, homelessness is the direct result of plans, policies, and practices that
have systematically and generationally excluded Black people from the housing market. When
Africans were brought to the United States and sold into slavery, they were classified as property
that could not, therefore, own property (Olivet, 2019). One of the first laws to enforce this policy
was the Three-Fifths Compromise of 1787, which was written into the United States Constitution
(Lynd, 1966). This compromise declared enslaved Africans as 3/5ths of a White person and
rendered any concept of “human rights” invalid. As a legal extension of physical branding, the
Three-Fifths Compromise ensured that land and property ownership were logically and legally
unobtainable. Although there are numerous cases of formerly enslaved Africans who inherited
land or were able to purchase land, these transactions were not legally recognized or protected.
As a result, even before the United States had a recognizable housing market, logically and
legally, enslaved Africans were disenfranchised.
In 1865, 78 years after the ratification of the Three-Fifths Compromise, the United States
ended one of its earliest and most long-standing definitions of enslavement. Following the Civil
4
War, General William Sherman issued Special Field Order #15. Under Special Field Order #15,
those who were formerly enslaved would be awarded 40 acres and a mule (Cimbala, 1989).
Unfortunately, Special Field Order #15 was rescinded as quickly as it was enacted (Cimbala,
1989). Even so, in support of Special Field Order #15 and those who were formerly enslaved, the
Southern Homestead Act of 1866 was passed. This act opened up public land to settlement and
farming to African Americans and White persons who had serviced the Union (Canaday et al.,
2015). Unfortunately, these lands were inaccessible to African Americans because 262 years of
enslavement had disrupted the accumulation of financial resources needed to occupy these
settlements. In addition, a year later, Union soldiers who were charged with enforcing Special
Order #15 and the Southern Homestead Act of 1866 were removed from the US South (Canaday
et al., 2015; Cimbala, 1989). Consequently, multiple generations of Black families were denied
land ownership.
Sharecropping in the US South, Restrictive Covenants in the North
Although Africans were no longer bound to plantations or slave owners, slavery had rendered
Africans destitute. Without financial resources, strong networks of support, or government
intervention, sharecropping became a new means of survival. In the South, sharecropping
allowed individuals to farm the land in exchange for housing and a share of crop production.
Under this system, former slave masters often owned these acreages of land and had complete
control over sharecropping wages. In many cases, these wages were not enough to survive or to
establish savings. To escape poverty and escalating levels of violence, such as lynching and
whitecapping, Black Americans headed North (Wilkerson, 2011).
At the same time, America entered World War I. During this time, millions of people
(predominately White males) were stripped from the American economy (Gotham, 2000). This
5
forced companies to recruit and utilize Black labor (Jones-Correa, 2000). The steady inclusion of
Black Americans into the labor market decreased already limited resources and led to increased
racial tensions (Jones-Correa, 2000). In an effort to maintain status and power, racially restrictive
covenants were introduced. With racially restrictive covenants, property owners could choose
not to sell, rent, or lease property on the basis of race (Rose, 2016). Largely enforced in cities,
these covenants forbade Black Americans and other minoritized groups from occupying or
owning property (Rose, 2016; Jones-Correa, 2000). This made it harder for Blacks to pursue new
employment opportunities since they were unable to occupy and/or secure adequate housing near
employment. Although racially restrictive covenants were not federal policies, they were
implemented and enforced at the state level.
Whitecapping and Redlining
As a response to a changing population, states around the country (most notably the in the
South) turned a blind eye to increased levels of racialized violence. Absent of legal protections,
Black people were physically and violently stripped of land and home ownership (Pietila, 2012;
Rothstein, 2017). A notable example is the Tulsa Race Riot of 1921. Greenwood, Oklahoma,
then known as America’s “Black Wall Street,” was an affluent Black town (Brophy, 2003).
Reports have suggested that money re-circulated within the community from 36 to 100 times
before exiting Greenwood’s economy (Bell, 2002). In Greenwood, Black residents owned
airplanes, extravagant homes, and an array of businesses (Bell, 2002; Montford 2014). In May of
1921, the Ku Klux Klan led a riot that resulted in the slaughter of 300 African Americans, the
destruction of 600 businesses, and propelled countless residents into abject homelessness
(Messer et al., 2018). Police participated in the riot, rather than quelling it. The Tulsa Race Riot
not only rendered a major economic setback for African Americans but also encouraged a culture
6
of violence that stunted land ownership, wealth accumulation, and upward mobility for untold
generations of African Americans.
Ten years after the Greenwood massacre, the United States entered the Great Depression. To
promote and sustain homeownership in a deteriorating economy, the Homeowners’ Loan
Corporation (HOLC) was established. The establishment of the HOLC allowed for the
refinancing of loans over a fifteen-year period, with an interest rate of 5% per year (Woods, 2012
pg. 1037). Between 1933 and 1936, the HOLC approved the administration of $3,093,451,321 in
mortgages to over one million Americans (Woods, 2012). Recognizing the need for a long-term
finance program for homeownership, Congress introduced the Federal Savings and Loan
Insurance Corporation (FSLIC) in 1934. The FSLIC developed and implemented lending and
appraisal practices that continued to exclude Black people from the housing market (Woods,
2012 pg. 1039).
When the HOLC was signed into law, the purpose was to prevent foreclosures by making
rental housing and home ownership affordable. To achieve this, the HOLC created maps for
underwriting that illustrated the risk of mortgage refinancing. The problem with this was that
these maps used a neighborhood’s racial composition to define mortgage risks, ultimately
deeming non-White neighborhoods as hazardous. The HOLC outlined these hazardous locations
in red and that process became known as redlining. As a consequence of redlining, Black
communities were deemed financially insecure or a threat to property values. This meant that
homes in predominantly Black neighborhoods appreciated slower and at lower rates than homes
in White neighborhoods. Black Americans, therefore, could not build generational wealth that
was equivalent to that of Whites. This gave each successive generation of Whites a considerable
7
advantage over their Black peers (Blackmon, 2009). Ultimately, 300 years of discrimination
created a wealth gap that continues to hinder housing stability for Black Americans.
As evidenced above, policies born out of racial discrimination restricted Black people from
accessing mortgage refinancing and federal underwriting opportunities (Pietila, 2012; Rothstein,
2017; Taylor, 2019). To put this in perspective, the National Community Reinvestment Coalition
(NCRC) released a report in 2018 that noted: “Most of the neighborhoods (74%) that the HOLC
graded as high-risk or ‘Hazardous’ eight decades ago are low-to-moderate income (LMI) today.
Additionally, most of the HOLC graded ‘Hazardous’ areas (nearly 64%) are minority
neighborhoods now” (NCRC, 2018).
The Exclusion of Black Veterans Adds to Housing Precarity
Following the Great Depression, the United States immediately entered World War II
between 1939 and 1945. To reward returning veterans for their service, the Servicemen’s
Readjustment Act of 1944, commonly known as the G.I. Bill, was established to guarantee an
array of benefits. Written in partnership with the American Legion, an organization which
historically denied memberships to Black Americans (McKenna, 2008), and sponsored by
Senator John Rankin, who introduced legislation to deport 12 million African Americans’ back
to Africa (Mettler, 2005; McKenna, 2008), the new legislation guaranteed mortgage loans to US
veterans (Woods, 2012). However, despite their service, Black veterans remained excluded from
the housing market, since Jim Crow laws prevented the obtainment of home loans from local
banks. Banks were free, therefore, to deny loans on the basis of race.
Sheer racism prevented Black veterans from integrating White neighborhoods. For example,
the G.I. Bill prescribed that states administer mortgage loans as they saw fit, ultimately
8
subjecting Black Americans to an additional bevy of discriminatory housing practices, on top of
whitecapping (threatening violence), redlining, denial of mortgages, and restrictive covenants
(Logan & Parman, 2017; Woods, 2018). As conceived, the G.I. Bill was supposed to allow
veterans to advance to middle-class status. Instead, this advancement occurred largely among
White Americans. Black Americans, many of whom still had living grandparents who were
former slaves, were left behind yet again.
Contemporary Disenfranchisement
In 1955, President Dwight Eisenhower announced plans to develop an Interstate Highway
System that would reduce traffic jams, expedite travel, and strengthen the nation’s security
(Karas, 2015). The following year, the Federal-Aid-Highway Act of 1956 was ratified. The
construction of interstate highways ultimately uprooted minoritized communities and required
the demolition of numerous African American historical sites, such as parks, churches, schools,
waterfronts, and cemeteries (Mohl, 2004; Karas 2015). Although interstate highways ended
isolation for some rural cities, many Black Americans were forced from their homes (Mohl,
2004). Others were unable to continue operating their businesses. For example, prior to the
construction of Interstate 40 through Nashville, there was significant legal opposition.
Neighborhood representatives argued that the construction of I-40 would displace a prominent
Black community by effectively “isolating Black owned businesses from their client base”
(Karas, 2015, pg. 13). While the legal opposition failed and construction was authorized to
continue, the ramifications appeared within a year. Following I-40’s completion, property values
declined 33%. Most Black-owned businesses were forced to close (Mohl, 2014; Karas 2015).
The consequences of the Interstate Highway System’s implementation were felt around the
country. Hamtramck, Michigan serves as another example. Over a six-year period, 1,800 Black
9
Americans were displaced when Interstate 75 was constructed (Sanchez et al., 2003). Around the
same time, a neighborhood in Birmingham, Alabama, which contained 60 blocks of Black
Americans, was cleared to accommodate interstate construction (Karas, 2015; Connerly, 2002).
The racial boundaries that the highway created eerily reflected Birmingham’s early racial zoning
policies for housing (Karas, 2015; Connerly, 2002). Unsurprisingly, scholars have found that
Birmingham’s construction of its Interstate Highway System largely contributed to significant
declines in the city’s black community (Connerly, 2002). Between the physical destruction of
neighborhoods and the predictive destruction of neighborhoods, property values and investments
in Black neighborhoods dropped drastically.
The War on Drugs and Its Collateral Consequences
In the 1970s, President Richard Nixon declared a “War on Drugs,” which ushered in the
modern era of mass incarceration. One of the major collateral consequences of being swept into
the judicial system for drug-related charges was eventual homelessness (Metraux et al., 2007;
Galea & Vlahov, 2002). The cascading effects of the War on Drugs commonly occurred in this
way: States were authorized to increase their enforcement of anti-drug regulations and to
implement harsher prison sentences. Low-level drug offenders typically received more than 20
years for small possession charges. Since 1980, arrests for drug-related sentences have tripled.
Black Americans were six times more likely to be incarcerated for drug-related offenses than
their White counterparts (Wagner & Sawyer, 2018; NAACP, 2018). Although Black Americans
are 13% of the population—and comprise only 12% of overall drug users—34% of those
arrested for drug offenses are African American (Mauer, 2009).
Black Americans constitute 46.2% of all incarcerations (Alexander, 2020; Chin, 2002; Nunn
2002; Miller, 2005). Because as many as 1.4 million of America’s inmates are Black men (Chin,
10
2002; Miller, 2005), the collateral consequences of the War on Drugs cannot be understated.
When African Americans are released from prison, their status as former felons relegates them to
a “New Jim Crow” class (Alexander, 2020). Around the country, subsidized and unsubsidized
housing properties have policies in place that prevent people with felonies from applying for
and/or receiving housing (Olivet et al., 2019). Convicted felons also do not have the right to
receive student loans to further their education. Their business and professional licenses can be
revoked. They can lose or be denied public benefits. They can be denied the right to vote.
Additionally, in some states, former felons are forbidden from fostering or adopting children
(Chin, 2002). Convicted felons cannot serve on juries. They are ineligible for security clearances
for federal jobs. They are ineligible for various employment options. They cannot enlist in the
military. They are subject to imprisonment if they are found to possess a firearm, thereby
denying them their Second Amendment rights (Chin, 2002). While the list of collateral
consequences provided above is expansive, the impact on housing is perhaps most germane to
this study. Convicted felons are excluded from public housing programs. An individual living in
public housing who houses a convicted felon may be subject to eviction (Carey, 2004; Bushway,
2007). For the convicted felons who have no place to live and work, homelessness and
recidivism may be the unfortunate outcomes.
Conclusion
The Fair Housing Act of 1968 aimed to end housing discrimination on the basis of race,
religion, nationality, or sex (Pietila, 2012; Rothstein, 2017;). However, more than 40 years after
its inception, studies show that there is still a lack of enforcement and insufficient funding for
necessary investigations (Pietila, 2012; Rothstein, 2017). The legacy of depriving African
Americans of one of the core tenets of the American Dream—home ownership—has become
11
deeply entrenched in the US economy. First, African slaves who were deemed property upon
their arrival to the US in 1619 could not, therefore, own any property of their own. Slaves that
fought eventually for the Union in the Civil War were promised parcels of land and livestock.
However, once the war ended in 1865 and these slaves were freed, the promises were rescinded.
Instead, Black Americans were met with racialized violence for nearly 50 years as they
attempted to buy their own land and set up their own cities in places like Greenwood, Oklahoma
in the 1920s. Two decades later, in the 1940s, Black American veterans who fought for their
country were excluded again, this time from the G.I. Bill, which awarded mortgages to purchase
homes in desirable places. In the places that Black Americans could settle, highway projects cut
through communities in the 1950s, a mere decade later. What was left of many predominantly
Black communities received a final blow when the War on Drugs, which lasted from the 1970s
until the late 2000s, incarcerated many Black male would-be heads of households and put them
into a permanent underclass (Alexander, 2020; Pietila, 2012; Rothstein, 2017; Olivet et al.,
2019).
This study, therefore, breaks new ground for homelessness scholars. While homelessness is
not exclusively Black, many of the historical factors that precede and drive initial housing loss
and future housing instability are unique to those who identify as Black. Current studies of
homelessness have done an exceptional job of elucidating the individual-level factors that
increase susceptibility to homelessness. However, less attention has been paid to the policies,
practices, and plans that have historically and systematically placed Black people at elevated
risks for homelessness. Recognizing that discriminatory legislation and policies have largely
contributed to the exclusion and inaccessibility of Black people from the housing market is
12
essential, as is understanding the mechanics of policy diffusion and its relevance to
understanding the algorithm of black homelessness.
13
Chapter 2
Nationally, communities around the country are observing rising levels of youth
homelessness. Although an array of identities comprise youth experiencing homelessness, race is
becoming increasingly salient. Youth who identify as Black or African American are 83% more
likely to be homeless than their White peers (Morton, 2017). Contrary to their numbers in the
general population, Black youth comprise 33% of youth experiencing homelessness (Henry et
al., 2017). In addition to being disproportionately represented in the homeless population, Black
youth’s disproportionate involvement in juvenile justice and foster care systems place them at
elevated risk of homelessness (Morton et al., 2017). Studies of youth homelessness suggest that,
in the present system, there are four main exits from homelessness. These include (1) permanent
supportive housing (PSH), (2) rapid re-housing (RRH), (3) locating housing independent of
service providers, and (4) returning to family (Rice et al., 2018).
While Black youth’s trajectories out of homelessness via PSH and RRH seemingly mirror
those of White youth and other youth of Color, research suggests that Black youth’s trajectories
out of homelessness may be disrupted by early systemic inequities that disrupt their ability to
return home and remain home. Specifically, a recent study found that Black youth who return
home have lower odds of remaining stably housed for 180 days or more (Hill et al., 2020 under
review). Given the salience of race to Black youth’s trajectories in and out of homelessness, this
study seeks to examine whether the predictors of Black youth who return to family mirror the
predictors of their racial counterparts while simultaneously identifying the potential unique
predictors that hinder Black youth from returning to family.
In recent years, providers have implemented coordinated systems of care (COC) to
connect homeless youth with housing services and resources. Within a COC, youth may be
14
eligible for a variety of support services. Although the primary goal of COCs is to end youth
homelessness, there are more youth experiencing homelessness than there are resources
available. Consequently, communities have adopted various assessment tools to assist providers
in allocating housing resources. For youth, the Next Step Tool (NST), informed by the
Transitional Age Youth Triage Tool and the Vulnerability Index-Service Prioritization Decision
Assistance Tool, is used most often. The NST is a closed-ended survey that assesses risk for
long-term homelessness and subsequent instability. Based on their NST score, youth may be
eligible for PSH, a time-unlimited housing program that includes wraparound social services, or
RRH, a time-limited rental subsidy program accompanied by various social services. A recent
study using data from the Homeless Management Information System (HMIS) found that youth
with lower prioritization scores are exiting homelessness by self-resolve (finding their own
accommodations) or returning to family (Rice et. al., 2018). While self-driven exits are common
among homeless youth, studies repeatedly suggest that familial relationships are critical to
exiting homelessness (Rice et. al., 2018; Barman-Adhikari et al., 2014; Wenzel et al., 2012; De
La Haye et al., 2012; Milburn et al., 2009).
Despite complicated pasts and complex familial relationships, research suggests that
homeless youth maintain familial connections and are more likely to count on relatives than
other network members for support (De la Haye et al. 2012; Wenzel et al., 2012). Positive family
relationships are associated with reduced risk behaviors and shorter durations of homelessness
(Milburn et al., 2009; Rice, Milburn, and Monro, 2011; Milburn et al., 2012; Barman-Adhikari et
al., 2014). As research indicates that family relationships are an important source of both
tangible and emotional support for youth experiencing homelessness, it is important to examine
the role of family in exits from homelessness (Johnson, 2005; De La Haye et al., 2012). In
15
studies of youth homelessness, familial relationships are positively associated with an increased
likelihood of exiting homelessness as well as maintaining stable housing over time (Milburn et
al., 2009). Furthermore, youth who have increased contact with parents have an increased
likelihood of exiting homelessness and remaining housed (Milburn et al., 2009).
Contrary to the positive role that familial relationships can have in the lives of homeless
youth, an array of studies suggest that fractured familial relationships are a central contributor to
youth homelessness (Edidin, 2012; Shelton, 2009; Ferguson, 2009; Mallet, 2009; Mallet, 2005;
Reeg, 2003). Housing loss attributed to unhealthy family relationships deter youth from returning
to family or seeking support from extended family (Edidin, 2012; Shelton, 2009; Ferguson,
2009; Mallet, 2009; Mallet, 2005; Reeg, 2003). Family conflict within the immediate family,
historically or contemporarily, may prevent youth from being able to identify or geographically
locate extended family networks. Youth who do have knowledge of extended family may avoid
seeking support because of suspicions that extended family networks have an allegiance to their
immediate families (Barker, 2012). Alternatively, youth who do not attribute housing loss to
family disruption and are able to return home have higher odds of membership in short-term
inconsistency trajectories (Tevendale et al., 2011). As evidenced by the studies highlighted
above, family relationships play a critical role in exits from homelessness.
Across racial and ethnic groups, previous research suggests that family reunification
accounts for 12% of exits from homelessness. As funding limitations prevent COCs from
allocating RRH or PSH to every youth experiencing homelessness, identifying predictors that
encourage or hinder returns to family may assist providers in developing programs or
interventions that repair and strengthen family relationships, ultimately making family
reunification a more viable, accessible, and long-term exit from homelessness. With those aims
16
in mind, this study will explore predictors of exiting homelessness by returning to family and
identify variations in predictors across racial and ethnic groups.
Method
Sampling Procedures. This study utilized de-identified administrative archival data
retrieved from the HMIS database generated on May 1, 2017. This data set included a total of
10,922 homeless youth from 16 communities across the United States. The data were collected
by local community providers and input into the HMIS at the time of housing eligibility
assessment. All youth input into the HMIS between January 1, 2015 and May 1, 2017 were
included in the data set. Information captured by the data set included housing eligibility
assessment dates, demographic information (e.g., age, race/ethnicity, and sexual orientation),
community type (e.g., rural, suburban, and urban), NST score (i.e., a vulnerability score
produced by the NST, a measurement widely adopted by homeless service providers to prioritize
housing resources), homelessness exits (e.g., PSH, RRH, and family reunification), homelessness
exit dates, and return to homelessness dates if stable housing was lost.
Measures
Demographic Characteristics. Age, gender, and sexual orientation were included as
control variables in this analysis. Age was a dichotomous variable in which one represents youth
aged 17 or younger. Sexual orientation was a dichotomous variable in which zero represents
heterosexual and one represents lesbian, gay, bisexual, queer or questioning, or intersex. Gender
was a dichotomous variable with female coded as one and male coded as zero. Community type
was a nominal variable with three categories: rural, suburban, and urban.
17
Homeless Experiences. For this analysis, several categories of homeless experiences
were deemed significant by bivariate analysis. These included physical violence; reasons for
housing loss (family violence, friends or family, unhealthy relationships, gender or sexual
orientation, and difference in religious or cultural beliefs); system interactions (ER care, legal
issue, incarcerated as a minor, use of crisis services, ambulance use, incarceration, police
interaction, and involvement in the correctional system); health (presence of mental health issue,
physical disability, chronic illness, leaving housing because of health, head injury, pregnancy or
indication of self-harm); substance use (use of marijuana before age 12, prescription misuse,
leaving housing because of substance use, hard to be housed because of substance use); need (no
subsistence need); behaviors (risk behavior, forced behavior, not taking medicine, avoiding
taking medicine, planned self-care, owing money, receiving money, and interest in HIV
housing); and living situation (couch, car, outdoors, shelter).
Dependent Variable. The study sought to examine predictors of youth who exit
homelessness by returning to family. Family reunification was a binary variable and youth who
exited homelessness via reunification were coded as one. Youth who exited homelessness via
PSH, RRH, self-resolve, or did not exit at all were coded as 0.
Analytic Plan
First, a series of bivariate logistic regression analyses were conducted to explore the
relationship between each assessment item and family reunification. The risk assessment item
served as the predictor variable and family reunification served as the outcome variable. Next, all
significant predictor variables were placed in a classification tree analysis to test linear and non-
linear relationships among significant risk assessment items (as determined by the bivariate
analysis).
18
Classification Tree Analyses
In this study, classification trees were used to classify characteristics that predicted
exiting homelessness by returning to family between and within racial groups. Within
classification tree analyses, recursive partitioning is used to generate dichotomous splits among
the selected variables. For each CART analysis, only significant risk assessment items are
included. To discover the relationship between variables, the Chi-square Automatic Interaction
Detector (CHAID) technique was used. CHAID does not require the data to be normally
distributed and because the dependent variable is categorical, chi-square tests are used to
partition. This analysis used cost-complexity pruning to limit over-fitting and determine the best
size of the tree (Breiman et al., 1984). The tree with the lowest misclassification rate was
selected as the optimal tree. The fit statistics used in this study included sensitivity, specificity,
area under the curve (AUC), and misclassification rate. The sensitivity revealed what percentage
of homeless youth who exited to family were correctly identified and the specificity showed
what percentage of homeless youth who did not return to family were correctly identified. The
importance or relevance of sensitivity and specificity depends on the interest in a particular
outcome. In this case, as this study was interested in knowing the predictors that prohibited youth
from utilizing family exits from homelessness, the specificity statistic was of more importance.
A specificity of 100% would mean that 100% of the homeless youth who did not return to family
were correctly identified while a sensitivity of 100% would show that 100% of homeless youth
who exited to family were correctly identified. The AUC is a measurement of how well
predictions are classified and ranked. The AUC ranges in value from 0 to 1. The AUC value is
assessed in terms of its proximity to 1. For example, a model with an AUC of 100% would
indicate that predictions are ranked 100% correctly. Alternatively, an AUC of .5 is the same as
19
random assignment to outcomes so anything greater than .5 moving is an improvement over
random assignment.
Results
Table 1 provides sample characteristics included in this analysis. The data set was
comprised of 10,250 youth. Youth (N = 654) who did not identify as White, Black, or Latino
were excluded from this analysis because these youth comprised a number of other racial and
ethnic categories and could not be disaggregated without compromising the sample size.
Furthermore, as an argument of this paper is that homeless youth are not monolithic, this study
chose not to lump the remaining youth (N=654) into a single category (such as Other or Youth of
Color) because while corresponding results would be specific to the racial groups comprised in
“Other” or “Youth of Color” as a whole, the results might not be specific to each racial group
that comprises the whole (i.e., “Other” or “Youth of Color”).
In the data set used for this analysis, most youth identified as White (51%), 32%
identified as Black, and Latino youth accounted for the remaining 16%. Of the community types
(rural, urban, suburban) recognized in this data set, 67% of the homeless youth belonged to urban
communities. This study predominantly consisted of youth who were male (78%), heterosexual
(70%), and 18 years or older (70%). LGBTQ youth accounted for 30% of this data set with
White youth comprising 49% LGBTQ youth, Black youth accounting for 30% LGBTQ youth,
and Latino youth comprising 15% of LGBTQ youth. Running away from home was the most
frequently cited reason for housing loss (75%). Roughly 42% of all homeless youth reported
physical violence and 19% reported self-harm. Relative to their Black (34%) and Latino (15%)
youth, White youth reported the highest rates of self-harm behaviors (44%). Among the youth
who reported physical harm, 44% were white, 34% were Black, and 15% were Latino. Relative
20
to their Black (29%) and Latino peers (14%), 50% of White youth reported owing money.
Across racial and ethnic groups, 10% of homeless youth reported avoiding medical attention.
This was most evident among White youth who comprised 48% of the youth who indicated an
avoidance of medical services. Of youth who reported a mental health issue (12%), 47% were
White, 31% were Black, and 15% were Latino. Among the various variables exploring substance
use, 13% of homeless youth attributed housing loss to substance abuse, 7% did not take
medications as prescribed, 4% reported that cannabis use started at age 12 or younger, and 4%
reported misusing prescription drugs. For housing exits, 5% of homeless youth exited
homelessness via PSH. Among youth who exited homelessness via PSH, 45% were White, 37%
were Black, and 11% were Latino. RRH accounted for 26% of exits from homelessness. For
youth who exited via RRH, 47% were White, 32% were Black, and 16% were Latino. Returns to
family accounted for 11.5% of exits from homelessness. White youth comprised 59% of family
exits, Black youth comprised 22%, and Latino youth comprised 12%. As it relates to returns to
family, this finding was particularly interesting because Black youth returned to family at lower
rates than their White peers and an odds ratio calculation suggests that, relative to Black youth,
White youth are 75% more likely to report returning home.
Table 2 provides the results of the bivariate analysis. The variables listed in this table
were significantly associated with exiting homelessness by returning to family (dependent
variable) and, therefore, included in subsequent classification tree analyses. Specifically, the
results listed in this table suggested that LGBTQ status, being female, reasons for housing loss,
living condition (car, shelter, outdoors, couch), police interaction, correctional system
involvement (stayed one night or more in jail), substance abuse, legal issues, risk behaviors,
forced behaviors, incarceration as a minor, prescription misuse, mental health issues, head
21
injuries, hard living independently, self-harm, owing money, received money, no subsistence
use, cannabis use (use occurred before age 12), chronic illness, community type (rural, urban),
not taking medication, avoiding medication, ambulance use, ER care, being incarcerated as a
minor, leaving housing because of health issues, physical disability, interest in HIV housing,
hard to be housed because of substance abuse, leaving housing because of substance use, planned
act of self-care, victim of physical violence, chronic illness, crisis service use, and pregnancy
were significantly associated with attributing housing loss to family disruption.
Table 3 reflects the most important predictors of exiting homelessness by returning to
family by race. The most important predictor was denoted by 1. When all youth were aggregated,
the most important predictor of exiting homelessness by returning to family by was community
type (rural). The subsequent predictor for all youth was identifying as a victim of physical
violence. Being a victim of physical violence was 60% as important as residing in a rural
community. Receiving ER care was a third predictor for youth experiencing homelessness. ER
care was 30% as important as residing in a rural community. When conducting these analyses for
each racial/ethnic group in separate analyses, for Black youth, couch surfing was the most
important predictor of exiting homelessness by returning to family. This was followed by being a
victim of physical violence which was 60% as important as the most important predictor (couch
surfing) for Black youth. Lastly, family violence, use of crisis services, and residing in an urban
community acted as important predictors of returning to family. These predictors were 50% as
important as the most important predictor for Black youth (couch surfing). Among Latino youth,
residing in a rural community was the most important predictor of exiting homelessness by
returning to family. Being a victim of physical violence, utilizing ER care, and attributing
housing loss to differing cultural beliefs were subsequent predictors of exiting homelessness by
22
returning to family and were 50% as important as the most important predictor (residing in a
rural community).
Black Youth
Figure 1 illustrates the characteristics that are most important to the prediction of exiting
homelessness via family or not. For this sample, the sensitivity was 11% and the specificity was
99%. The AUC was 69% and the misclassification rate was 7%. In this analysis, low sensitivity
was attributed to the sample size. Explained further, family exits only accounted for 11.5% of
total exits from homelessness. Youth who exited homelessness by returning to family were
compared to youth who utilized other exits from homelessness. As a result, although there were
3,382 Black youth in this study, only 273 utilized family exits. Of those 273, only 252 were
included in the final classification tree because Black youth with missing data points for the
response variable or missing data points for the significant predictor variables were excluded
from the analysis. These youth were then placed in contrast with the total number of Black youth
who utilized another exit from homelessness. The optimal tree structure for Black youth
suggested that couch surfing, physical violence, community type, family violence, and the use of
crisis services were the most important variables for predicting exits to family. Of the previously
mentioned predictors, couch surfing (parent node) appeared to be the most important. Among
Black youth who were not couch surfing, 93% did not exit to family. For Black youth who were
not couch surfing, experiencing physical violence was a subsequent predictor. Among Black
youth who were not couch surfing and experienced physical violence, 4% exited homelessness
by returning to family. For Black youth who were not couch surfing and did not report physical
violence, 10% exited homelessness by returning to family. For Black youth who were not couch
surfing and did not report physical violence, family violence was a subsequent predictor. 12% of
23
Black youth who were not couch surfing and did not report physical or family violence returned
home, relative to 5% of Black youth who were not couch surfing and did not report physical but
reported family violence. Among Black youth who did report couch surfing, 23% exited to
family. For Black youth who were couch surfing, the second important predictor was community
type (urban). 18% of Black youth who were couch surfing and reported living in an urban
community exited homelessness by returning home. In contrast, among Black youth who were
couch surfing and did not reside in a rural community, 33% returned to family. The third most
important predictor for exiting homelessness by returning to family for Black youth who were
couch surfing and did not reside in an urban community was the use of crisis services. 44% of
Black youth who were not couch surfing, did not live in an urban community, and used crisis
services less than 2.5 times exited homelessness by returning to family. Among Black youth who
were not couch surfing, did not live in an urban community, and used crisis services more than
2.5 times, only 18% exited homelessness by returning to family.
Latino Youth
Figure 2 illustrates the characteristics that were most important to exiting homelessness
via family or not. In this sample, the sensitivity was 10% and the specificity was 99%. The AUC
was 56% and the misclassification rate was 8%. For Latino youth, the most important predictor
(parent node) was residing or not residing in a rural community. Among Latino youth who
resided in a rural community, 36% exited to family relative to 8% of those who did not reside in
a rural community. Of Latino youth who did reside in a rural community, the second predictor
was ER care. For Latino youth who resided in a rural community and had less than four visits to
the ER, 43% exited homelessness by returning to family, relative to 9% or Latino youth who
resided in a rural community and had four or more visits to the ER. For Latino youth who lived
24
in a rural community and had less than four ER visits, housing loss attributed to different
religious or cultural beliefs was the third most important predictor. Among Latino youth who
lived in a rural community, had less than four ER visits, and cited different cultural or religious
beliefs as the reason for housing loss, 18% exited homelessness by returning to family, relative
to 52% of Latino youth who lived in a rural community, had less than four ER visits, and who
did not cite different cultural or religious beliefs as the reason for housing loss.
White Youth
Figure 3 shows the characteristics that were most important to exiting homelessness via
family among White youth. The sensitivity for this model was 8% and the specificity was 99%.
The AUC was 70% and the misclassification rate was 14%. The most important characteristic for
predicting exiting to family was being a victim of physical violence since becoming homeless.
8% of White youth who have been a victim of physical violence since becoming homeless exited
to family. For White youth who have not been victims of physical violence since entering
homelessness, 18% exited to family. The second most important predictor for White youth who
have been a victim of physical violence was community type (rural). For White youth who have
been victims of physical violence and resided in a rural community, 31% exited homelessness by
returning to family. For White youth who have been victims of physical violence and did not
reside in a rural community, 15% exited homelessness by returning to family. Among White
youth who were victims of physical violence and did not reside in a rural community, family
violence was a subsequent predictor. 19% White youth who were victims of physical violence,
did not reside in a rural community, and did not report family violence exited to family whereas
only 6% of White youth who were victims of physical violence, did not reside in a rural
community, and reported family violence exited to family. For White youth who were victims of
25
physical violence and resided in a rural community, family violence was also a subsequent
predictor. 35% of White youth who were victims of physical violence, resided in a rural
community, and did not report family violence exited to family whereas only 22% of White
youth who were victims of physical violence, resided in a rural community, and reported family
violence exited to family.
All Youth
Figure 4 shows the characteristics that were most important to exiting homelessness via
family among all youth experiencing homelessness. The sensitivity for this model was 6% and
the specificity was 99%. The AUC was 61% and the misclassification rate was 11%. In this
model, among youth experiencing homelessness, the most important characteristic for predicting
exits to family was community type. 24% of youth who lived in rural communities exited
homelessness by returning to family relative to 9% who did not live in rural communities. For
youth who did reside in rural communities, the second predictor of importance was being a
victim of physical violence since becoming homeless. 32% of youth who lived in rural
communities and did not report physical violence exited homelessness by returning to family
relative to 14% of youth who lived in rural communities and did report physical violence. Third,
for youth who lived in a rural community and who did not report physical violence, living
outdoors was the third predictor of importance. 33% of youth who lived in a rural community,
did not report physical violence, and were not living outdoors reported exiting homelessness by
returning to family. In contrast, 3% of youth who lived in a rural community, did not report
physical violence, and lived outdoors reported exiting homelessness by returning to family. For
youth who lived in rural communities and reported physical violence, ER care was a predictor of
importance. 10% of youth who lived in rural communities, reported physical violence, and
26
reported one or more ER visits exited homelessness by returning home, compared to 25% of
youth who lived in rural communities, reported physical violence, and reported no ER visits.
Table 2.1
Overall Descriptive Characteristics
Descriptive Characteristics N White Black. Latino
10, 922 5212 3382 1656
Gender
Male 8487 (78%)
4075
(78%)
2586
(77%)
1300
(79%)
Female 2429 (22%)
1133
(22%)
786 (23%) 355 (21%)
Sexual Orientation
LGBTQ 3319 (30%)
1619
(31%)
984 (29%) 511 (31%)
Heterosexual 8487 (78%)
3593
(69%)
2398
(71%)
1145
(69%)
Age
< 17 3303 (30%)
1653
(32%)
953 (28%) 508 (31%)
> 17 7619 (70%)
3559
(68%)
2429
(72%)
1148
(69%)
Living Situation
Car 766 (7%) 527 (10%) 114(3%) 74 (4%)
Couch 665 (6%) 289 (6%) 252 (7%) 73 (4%)
Outdoors 798 (7%) 258 (5%) 344 (10%) 126 (8%)
Shelter 7188 (66%)
3464
(66%)
2208
(65%)
1112
(67%)
Transitional Housing 1505 (14%) 674 (13%) 463 (14%) 271 (16%)
Service Engagement
ER 1+ 6194 (57%)
2787
(53%)
2055(61%) 936 (57%)
Ambulance Use 1+ 2333(21%)
1042
(20%)
792 (13%) 351 (21%)
Crisis Services 1+ 3419 (31%)
1615
(21%)
1058(31%) 508 (33%)
Police 1+ 4603 (42%)
2121
(39%)
1470(43%) 699 (42%)
Incarceration (Minor) 2784 (25%)
1296
(25%)
898 (27%) 406 (25%)
Legal Issue 2816 (26%) 1313 925 (27%) 403 (24%)
27
(25%)
Correctional System 1+ 1985 (18%) 940 (18%) 602 (9%) 294 (18%)
Violence
Physical Harm 4635 (42%)
2053
(39%)
1565(46%) 716(43%)
Self-Harm 2077 (19%) 914 (18%) 711 (21%) 299 (18%)
Behavior
Forced Behavior 1545 (14%) 764 (15%) 446 (13%) 241 (15%)
Risk Behavior 2614 (24%)
1293
(25%)
810 (24%) 356 (22%)
Finances
Owe Money 1861 (17%) 938 (18%) 563 (17%) 252 (16%)
Received Money 3222 (30%)
1602
(31%)
948 (28%) 467 (28%)
Fulfillment
Planned Activities 2891 (26%)
1424
(27%)
822 (24%) 454 (27%)
Need
No Subsistence 9506 (87%)
4509
(87%)
2962
(88%)
1451
(88%)
Reason Homeless
Ran Away 8147 (75%)
3904
(75%)
2482
(73%)
1264
(76%)
Religion 2750 (25%)
1323
(25%)
839 (25%) 393 (24%)
Family or Friends 2549 (23%)
1182
(23%)
813 (24%) 390 (24%)
Gender/Sexual Orientation 1790 (16%) 873 (17%) 537(16%) 268 (16%)
Family Violence 3288 (30%)
1570
(30%)
1013
(30%)
492 (30%)
Unhealthy Relationship 2314 (21%)
1170
(22%)
647 (19%) 336 (20%)
Health
Physical Health 455 (4%) 202 (4%) 157 (5%) 62 (4%)
Chronic Illness 562 (5%) 267 (5%) 180 (5%) 82 (5%)
HIV/Aids 453 (4%) 226 (4%) 140 (4%) 57 (3%)
Physical Disability 988 (9%) 414 (8%) 338 (10%) 168 (10%)
Avoidance of Medical
Attention
1082 (10%) 516 (10%) 340 (10%) 162 (10%)
Pregnant 952 (9%) 415 (8%) 365 (11%) 122 (7%)
Substance Use
Substance Use Disrupted
Housing
1449 (13%) 646 (12%) 488 (14%) 224 (14%)
Substance Use Disrupts 1350 (12%) 615 (12%) 439 (13%) 200 (12%)
28
Housing
Not Taking Medication 780 (7%) 361 (7%) 257(8%) 124 (7%)
Prescription Misuse 447 (4%) 184 (4%) 166 (5%) 60 (4%)
Pot Use < 12 421 (4%) 180 (3%) 135 (4%) 85 (5%)
Mental Health
Mental Health Issue 1328 (12%) 622 (12%) 418(12%) 196 (12%)
Head Injury 334 (3%) 161 (3%) 113 (3%) 38 (2%)
Developmental Disability 204 (2%) 98 (2%) 64 (2%) 28 (2%)
Disability = no Independence 905 (8%) 399 (8%) 281 (8%) 165 (18%)
Exit Type
Permanent Support Housing 579 (5%) 260 (5%) 217 (6%) 66 (4%)
Rapid Re-Housing 2885 (26%)
1362
(26%)
910 (27%) 466 (28%)
Formal Exit (PSH & RRH) 3464 (32%)
1622
(31%)
1127(33%) 532 (32%)
Family 1259 (12%) 745 (14%) 273 (8%) 146 (9%)
Community Type
Urban 7285 (67%)
3093
(59%)
2584
(76%)
1251
(76%)
Suburban 2046 (19%) 979 (19%) 612 (18%) 350 (21%)
Rural 1591 (15%)
1140
(22%)
186 (6%) 55 (3%)
Stability
Stable 180 Days + 4393 (74%)
2240
(74%)
1270(74%) 625(74%)
29
Table 2.2.
Bivariate analyses of risk factors and exiting homelessness via family reunification
Variable p value
Rural <.0001
Victim of Physical Violence <.0001
Urban <.0001
Legal Issue <.0001
Violence Between Family <.0001
Couch <.0001
Police Interaction <.0001
Outdoors <.0001
Incarcerated Minor <.0001
Car <.0001
Unhealthy Relationship <.0001
HDTBH Substance Abuse <.0001
Planned Self Care Act <.0001
Friends or Family Caused Homelessness <.0001
Gender or Sexual Orientation Caused Homelessness <.0001
Correctional System <.0001
Self-Harm or Others <.0001
No Subsistence Need <.0001
Mental Health Issue <.0001
Risk Behavior <.0001
Forced Behavior <.0001
Difference in Religious Beliefs Caused Homelessness <.0001
Housing loss attributed to Substance Use <.0001
ER Care <.0001
Not Taking Medicine <.0001
Prescription Misuse <.0001
Owe Money <.0001
Incarceration <.0001
Pregnant <.0001
Female <.0001
Physical Disability <.0001
Interest in HIV Housing <.0001
Shelter <.0001
Housing loss attributed to Health <.0001
Crisis Service Use <.0001
Ambulance Use <.0001
Chronic Illness <.0001
Received Money 0.0001
Avoid Medication 0.0002
Head Injury 0.0004
30
Hard Living Independently 0.0004
Pot Use < 12 0.0005
LGBTQ 0.0123
Table 2.3
Feature Importance by Race & Ethnicity
Note. The term n/a indicates that a particular risk item does not appear in the tree of that specific
group.
Table 2.4
Fit Statistics by Race & Ethnicity
Note. AUC = area under the curve, MR = misclassification rate
Variable All Youth White Black Latino
Rural 1.0 0.9 n/a 1.0
Victim of Physical Violence 0.6 1.0 0.6 0.5
Family Violence n/a 0.8 0.5 n/a
Outdoors n/a 0.3 n/a n/a
Couch n/a n/a 1.0 n/a
ER Care 0.3 n/a 0.5
Use of Crisis Services n/a n/a 0.5 n/a
Different Cultural Beliefs n/a n/a n/a 0.5
Urban n/a n/a 0.5 n/a
Sensitivity Specificity AUC MR
All Youth 0.06 0.99 0.61 0.11
White 0.08 0.99 0.7 0.14
Black 0.11 0.99 0.69 0.07
Latino 0.10 0.99 0.56 0.08
31
Figure 2.1.
Predictors of Black Youth Who Return to Family
32
Figure 2.2.
Predictors of Latino Youth who Return to Family
33
Figure 2.3.
Predictors of White Youth who Return to Family
34
Figure 2.4.
Predictors of All Youth Who Return to Family
35
Discussion
Recent work explores four viable exits from homelessness (Rice et al., 2018). These
include PSH, RRH, self-resolve (locating housing independent of service providers), and family
reunification (Rice et al., 2018). Previous research found that RRH and PSH account for 32% of
exits from homelessness. Even so, COCs cannot allocate RRH or PSH to every youth
experiencing homelessness. For youth who cannot exit homelessness via RRH or PSH, family
reunification may be a viable solution. As highlighted in previous findings, family reunification
accounts for 12% of exits from homelessness (Hsu et al., 2019; Rice et al., 2018). In studies of
youth homelessness, the presence of familial relationships has played a critical role in how youth
transition from homelessness and when youth transition from homelessness (Edidin, 2012;
Shelton, 2009; Ferguson, 2009; Mallet, 2009; Mallet, 2005; Reeg, 2003). For instance, within six
months of initial homelessness, one study found that more than 66% of homeless youth return
home (Milburn et al., 2006b). As evidenced above, despite initial reasons for housing loss, youth
are exiting homelessness by returning to family (Rosenthal et al., 2007; Milburn et al., 2006).
The findings in this study highlight potential intervention points for providers. Most notably,
these findings may help distinguish which subgroups of homeless youth may utilize or be more
inclined to exit homelessness by returning to family.
For White and Black youth experiencing homelessness, physical violence was a bi-
directional predictor of exiting homelessness by returning to family. Research repeatedly
attributes youth homelessness to familial factors such as violence, abuse, and poverty (Whitbeck
& Hoyt, 2000; Alvi et al., 2010). However, there is room for additional exploration of how
familial relationships become strained in the first place. For instance, a Canadian study found
that the inability to tolerate children who act outside of household norms and rules as well as
36
family violence (whether it is parental violence or not being protected from violence happening
within the family) may be reflective of structural or systemic factors that act as precursors to
strained family relationships (Alvi et al., 2010). These findings suggest that familial relationships
and the origin of familial strains require further exploration. It is also important to note that
familial strains and the underlying factors that cause them may differ across subgroups of
homeless youth because of systematic and structural disadvantage. This may be particularly
relevant to Black youth whose family structures are more likely to be susceptible to system
interferences such as generational foster care, incarceration, wage inequality, neighborhood
disadvantage, and external violence.
Among youth experiencing homelessness, Black youth did not return to family at the
same rates as their White peers. For instance, as reflected in table 1, returns to family comprised
12% of exits from homelessness. Of youth who exited homelessness by returning to family, 14%
of White youth, 8% of Black youth, and 9% of Latino youth returned home. While the data set
does not provide details to further explore this finding, the classification tree analysis for Black
youth established couch surfing as the most important predictor of exiting homelessness by
returning to family. The presence of couch surfing in the classification tree may be indicative of
informal networks of support (family, extended relatives, or friends) who can house couch
surfing Black youth for short periods of time. This finding does not imply a difference in
distribution but rather a difference in predictors that generates additional thought and room for
further exploration.
Couch surfing as a predictor of returning to family may also imply that, for Black youth,
returns to family are an extension of couch surfing (i.e., temporary housing) rather than an exit
from homelessness. For instance, a recent study found that Black youth who returned home had
37
lower odds of remaining stably housed for 180 days or more (Hill et al., 2020 under review).
This finding further suggests that, for Black youth, returns to family may be short-term
arrangements rather than long-term placements. As a result, Black youth may resurface in
subsequent pulls of administrative data and/or in another COC because the original housing
placement (couch surfing) was coded as an exit from homelessness rather than a short-term
placement.
Limitations
The findings from this study should be evaluated in the context of its limitations. First,
this study can only account for youth who report returning to family. Given the current time
limitation of the data, this study cannot conclude if this attempt was successful. Consequently,
some youth may be returning to homelessness but this may not be accounted for in the current
version of the data used for this examination. Secondly, a larger sample size and more data are
needed to better understand the factors that precede or contribute to predictors of returning to
family. Lastly, several missing data points limit the generalizability of the data. Nonetheless, the
results of this study are promising and emphasize further and more rigorous investigation of the
homeless management system for Black youth experiencing homelessness.
Conclusion
Black youth’s use of informal housing (couch surfing) may impact the ways in which
providers assess the needs of Black youth and, subsequently, how providers allocate housing
resources. The findings in this study reinforce the need for COCs to focus on strengthening
existing familial bonds or assisting homeless youth with forging new support systems. While
temporary support and structures (couch surfing) may be predictive of Black youth returning to
38
family, returns to family may only be an extension of couch surfing rather than a viable exit from
homelessness.
The findings in this study also raise important questions around visibility. Specifically,
providers may be asking about exits incorrectly, categorizing exits incorrectly, tracking exits
incorrectly, or not asking enough questions about exits. For instance, while the previous assertion
may be more obvious among Black youth who are coded as exiting to incarceration or
categorized as lost, pending, and unknown, it is probably less obvious that Black youth and other
Youth of Color (YOC) are coded as returning to family or self-resolving because they are not
visible in the homeless system. Despite this classification, it is possible that Black youth and
other YOC who have been coded as returning to family have not exited homelessness. Instead,
these youth may be utilizing temporary housing arrangements in conjunction with low barrier
services (drop-in centers and/or community resources outside of the HMIS that do not track
service use) and possibly resurface elsewhere. By coding youth as returning to family without
further exploration or follow-up, when youth do resurface in COCs, these youth are often
classified as re-entering homelessness when, in fact, these youth have never left a state of
housing instability and characterizing them as “returning to homelessness” may be inaccurate.
Future studies should investigate the visibility (or lack thereof) of Black youth in the HMIS, and
how that visibility impacts the housing assessments of Black youth.
39
Chapter 3
Annually, as many as 4.2 million teens and young adults experience homelessness (Morton et
al., 2018). In the United States, youth homelessness has reached crisis levels (Rice et al., 2018).
To combat youth homelessness, communities around the country have implemented continuums
of care (COC). A COC is defined as a “regional or local planning body that coordinates housing
and service funding from the U.S. Department of Housing and Urban Development” (Rice et al.,
2018, pg. 70). The primary goal of COCs is to assess and link youth who are in the greatest need
of care to housing services and resources. Youth who enter COCs are assessed using the
Transition Age Youth-Vulnerability Index-Service Prioritization Decision Assistance Tool
(TAY-VI-SPDAT): Next Step Tool (NST) (Orgcode Consulting, 2015). Based on responses to
28 multiple-choice questions, youth are assigned a score between 0 and 17. Youth who score 8 or
higher are recommended for long-term housing with high service intensity. Youth who score
between 4 and 7 are recommended for time-limited housing with moderate service intensity.
Youth whose scores are between 0 and 3 are not recommended to a housing program or for
moderate-to-high intensity services (Rice et al., 2018, pg.71).
Data from the HMIS suggests that there are four main exits from homelessness: (1) RRH, (2)
PSH, (3) self-resolve (exiting homelessness independent of COCs), and (4) returning to family.
Rapid re-housing interventions are time-limited rental subsidies programs that offer various
social services. Alternatively, PSH interventions provide wraparound social services that are
offered for an extended period of time. While initial data suggest that RRH and PSH are
allocated fairly, one study elucidates a potential racial disparity. A recent finding revealed that
relative to their non-White peers, YOC who exit homelessness by returning to family have a 43%
reduction in odds of remaining stably housed 180 days or more (Rice et al., 2018). In addition,
40
across gender and sexual orientation, Black youth have lower odds of returning to family than
being lost or pending (Hill et al., 2020 under review). Reductions in youth homelessness are
largely dependent on the ability to remain stably housed following an episode of homelessness.
Based on the previously mentioned studies, the predictors of housing stability for Black youth
may not mirror their peers. This chapter aims to explore predictors of remaining stably housed
for 180 days or more among Black youth experiencing homelessness. This chapter also compares
these data across racial groups.
Research shows that service engagement is critical to successful exits from homelessness
(Morton et al., 2018; Chan et al., 2017; Barman-Adhikari et al., 2016; Barman-Adhikari & Rice
2014; Tyler, Akinyemi, & Butler, 2012; Berdahl et al., 2005; Cauce et al., 2000; De Rosa et al.,
1999; Gerber 1997). Even so, among homeless youth, studies consistently show low rates of
service utilization and varying levels of engagement across service types (De Rosa et al., 1999;
Kipke, Montgomery, & MacKenzie, 1993). In fact, as it relates to mental health, less than 50%
of youth experiencing homelessness (YEH) use mental health services (Kort-Butler & Tyler,
2012; Milburn et al., 2006; Sweat et al., 2008; Tyler et al., 2012). Additionally, YEH are more
likely to pursue services that meet immediate needs such as food, clothing, and showers (De
Rosa et al., 1999; Pergamit, Ernst, & Hall, 2010; Sweat et al., 2008) rather than services like
employment or education that can assist with long-term stability.
While there is a plethora of research exploring correlates of service use amongst YEH—and
even though service use is used in assigning prioritization scores—findings around correlates of
service use among Black YEH are limited. For instance, relative to their non-Black peers, youth
who identified as Black had higher rates of service use among most services (e.g., employment
assistance, shelter, and healthcare) (Hill et al., 2019 – under review). In addition, the same study
41
found that 64% of Black youth reported no use of therapy services in the last 30 days. (Hill et al.,
2019 – under review). Similarly, an earlier study suggested that racial and ethnic minorities were
less likely to receive services than their White peers (Garland et al., 2005). As findings around
Black youth and service use are mixed, factoring service use as an item of prioritization in the
NST score may prevent Black youth from being correctly allocated to the appropriate housing
resource and, consequently, Black youth who may be in the greatest need of formal housing
resource may not receive it.
To successfully exit homelessness, some groups of homeless youth may use specific sets of
services (RRH, PSH, therapy, healthcare, employment assistance, and/or shelter) to exit
homelessness or use specific services at varying rates to exit homelessness. The use of these
services may enable subsequent stability. For instance, while service usage patterns are
extremely limited among LGBT youth, a 2010 study suggests that LGBT homeless youth use
basic (food), therapy, and health (STI/HIV testing) services more often their heterosexual
counterparts (Choi et al., 2015). Additionally, relative to White youth, an early study found that
homeless YOC use shelters more frequently (De-Rosa et al., 1999). In addition to identity,
exposure and access to services within housing programs also appear to impact the stability of
exits from homelessness. As it relates to RRH, a program that offers time-limited services and
short-term rental assistance, lower engagement in other services (shelter and transitional
housing) has been associated with a lower probability of exiting homeless (Rice et al., 2018).
Among youth who exited homelessness via PSH (a long-term intervention that includes access to
extensive services), there were lower rates of returns to homelessness (Rice et al., 2018). While
the studies cited above offer insight on patterns of service utilization, these studies do not expand
upon the relevance of service use to subsequent housing stability. To our knowledge, studies of
42
youth homelessness have not examined predictors of remaining stably housed for 180 days or
more across racial groups or the relevance of service use to remaining stably housed for 180 days
or more. Therefore, this study aims to address the following research questions:
RQ1 – What are the predictors of remaining stably housed for 180 days or more among
Black youth experiencing homelessness?
RQ2 – What predictors promote or discourage housing stability for 180 days or more?
RQ3 – Do the predictors of remaining stably housed for 180 days or more differ among
Black, White, and Latino youth experiencing homelessness?
This analysis is a data-driven exercise in machine learning that is guided by CART, which will
determine what factors “predict” housing stability at 180 days or more.
Methods
Sampling Procedures. This study utilized de-identified administrative archival data
retrieved from the HMIS database generated on May 1, 2017. This data set included a total of
10,922 homeless youth from 16 communities across the United States. Data from youth were
entered into the HMIS by local community providers when assessing a youth’s housing
eligibility. All youth input into the HMIS between January 1, 2015 and May 1, 2017 were
included in this dataset. Information captured by the dataset includes housing eligibility
assessment dates, demographic information (e.g., age, race/ethnicity, and sexual orientation),
community type (e.g., rural, suburban, and urban), NST score (i.e., a vulnerability score
produced by the NST, a measurement widely adopted by homeless service providers to prioritize
housing resources), homelessness exits (e.g., PSH, RRH, and family reunification), homelessness
exit dates, and return to homelessness dates if stable housing was lost.
43
Measures
Demographic Characteristics. Age, race, gender, and sexual orientation were included
in this analysis. Age was a dichotomous variable; it represented youth aged 17 or younger. Race
was a three-category nominal variable that included White, Black, and Latino. Sexual orientation
was a dichotomous variable in which zero represents heterosexual and one represents lesbian,
gay, bisexual, queer or questioning, or intersex. Gender was a dichotomous variable with female
coded as one and male coded as zero. Community type was a nominal variable that included
three categories: rural, suburban, and urban.
Independent Variables. The NST is a linear scoring system that incorporates 28
multiple-choice, dichotomous, and frequency-type questions to assess homeless youth’s level of
vulnerability (Rice et al. 2018). These questions cover the history of housing and homelessness,
risks, socialization, daily functions, and wellness (Sample question: Is your current lack of stable
housing because of violence at home between family members?). The analysis for this study
began with the inclusion of every risk assessment item used to determine the NST score. The
relevance of risk assessment items was determined by bivariate analyses.
Dependent Variable. The study sought to examine youth who remained stably housed
across any exit type for 180 days or more. Youth could be categorized into one of seven exit
types. Exit type was a nominal variable that included seven categories: PSH, RRH, family
reunification, self-resolve, pending, lost (no contact with providers), and incarceration. PSH and
RRH represent youth who exit homelessness by receiving PSH or RRH. Family reunification
depicts youth who exit homelessness by returning to family. Youth who exit homelessness via
self-resolve have located housing independent of formal programs or family reunification. Youth
are categorized as pending if they are awaiting housing. Youth who are categorized as lost have
44
lost contact with homeless service providers. Incarceration encompasses youth who have exited
homelessness via incarceration. Youth who remained stably housed for 180 days across any exit
type were coded as one.
Analytic Plan
First, bivariate analyses were generated to explore correlations between the DV (Stable
180 Days+) and 52 predictor variables. The results of the bivariate analyses excluded ten risk
assessment items as potential predictors of housing stability for 180 days or more.
Next, CART analyses were used to split all possible combinations of values for the predictor
variables into non-overlapping regions (Gordon, 2013; Sledjeski et al., 2008). Within CART
analyses, data division was achieved through recursive partitioning (Gordon, 2013; Sledjeski et
al., 2008). Individual splits were determined by finding the best predictor variable and a cut point
that assigned the observations in the parent node to the child nodes. For the purpose of this study,
CHAID was used to assess model fit. Only significant variables as determined by the bivariate
analyses were included in each classification tree analysis.
Classification Tree Analyses
In this study, classification trees were used to identify characteristics predictors of
remaining stably housed 180 days or more between and within racial groups. This analysis used
cost-complexity pruning to limit over-fitting and determine the best size of the tree (Breiman et
al., 1984). The tree with the lowest misclassification rate was selected as the optimal tree. The fit
statistics used in this study included sensitivity, specificity, AUC, and misclassification rate. The
sensitivity revealed what percentage of homeless youth who remain stably housed for 180 days
or more were correctly identified and the specificity showed what percentage of homeless youth
who did not remain stably housed for 180 days or more were correctly identified. The
45
importance or relevance of sensitivity and specificity depends on the outcome of interest. As this
study is interested in the predictors that promote and hinder housing stability for 180 days or
more, sensitivity and specificity are important. In terms of what makes a good specificity or not,
a 100% specificity would suggest that the model correctly identified 100% of homeless youth
who did not remain stably housed for 180 days or more, while a 100% sensitivity would suggest
that 100% of the homeless youth who remained stably housed for 180 days or more were
correctly identified. The AUC is a measurement of how well predictions are classified and
ranked. The AUC ranges in value from 0 to 1. The AUC value is assessed in terms of its
proximity to 1. For example, a model with an AUC of 100% would indicate that predictions are
ranked 100% correctly.
Results
Table 1 reflects the descriptive characteristics included in this analysis. The data set was
comprised of 10,250 youth. Youth (N = 654) who did not identify as White, Black, or Latino
were excluded from this analysis because these youth could not be disaggregated into another
racial or ethnic background without compromising the sample size. Furthermore, as an argument
of this paper is that homeless youth are not monolithic, I chose not to lump the remaining youth
(N=654) into a single category (such as Other or Youth of Color) because corresponding results
would be specific to the racial groups comprised in “Other” or “Youth of Color” but would not
necessarily be specific to each racial group that comprises “Other” or “Youth of Color.” In the
data set used for this analysis, most youth identified as White (51%), one-third identified as
Black, and Latino youth accounted for the remaining 16%. Of the community types (rural, urban,
suburban) recognized in this data set, 67% of homeless youth belonged to urban communities.
46
Aligning with national reports, homeless youth in this data set were predominantly male
(78%), heterosexual (70%), and 18 years or older (70%). LGBTQ youth accounted for 30% of
this dataset, with White youth comprising 49% of LGBTQ youth, Black youth accounting for
30% of LGBTQ youth, and Latino youth comprising 15% of LGBTQ youth. Running away from
home was the most frequently cited reason for housing loss (75%). Roughly 42% of all homeless
youth reported physical violence and 19% reported self-harm. Relative to their Black (34%) and
Latino (15%) counterparts, White youth reported the highest rates of self-harm behaviors (44%).
Among youth who reported physical harm, 44% were white, 34% were Black, and 15% were
Latino. Relative to their Black (29%) and Latino peers (14%), 50% of White youth reported
owing money.
Across racial and ethnic groups, 10% of homeless youth reported avoiding medical
attention. This was most evident among White youth who comprised 48% of youth who
indicated an avoidance of medical services. Of youth who reported a mental health issue (12%),
47% were White, 31% were Black, and 15% were Latino. Among the various variables
exploring substance use, 13% of homeless youth attributed housing loss to substance abuse, 7%
did not take medications as prescribed, 4% reported that cannabis use began at age 12 or
younger, and 4% reported misusing prescription drugs. For housing exits, 5% of homeless youth
exited homelessness via PSH. Among youth who exited homelessness via PSH, 45% were
White, 37% were Black, and 11% were Latino. RRH accounted for 26% of exits from
homelessness. For youth who exited via RRH, 47% were White, 32% were Black, and 16% were
Latino. Returning to family accounted for 11.5% of exits from homelessness. White youth
comprised 59% of family exits, Black youth comprised 22%, and Latino youth comprised 12%.
For youth who exited homelessness, 40% remained stable for 180 days or more. Of those who
47
did exit homelessness and maintained stability for 180 days or more, 51% were White, 29% were
Black, and 14% were Latino.
Table 2 provides the results of the bivariate analyses. The variables listed in this table
were significantly associated with remaining stably housed for 180 days or more (dependent
variable) and, therefore, included in subsequent CART analyses. Specifically, the results listed in
this table suggest that exit type (PSH, RHH, family, or self-resolve), age, LGBTQ status, reasons
for housing loss, living condition (car, shelter, outdoors, couch), police interaction, correctional
system involvement, substance abuse, legal issues, risk behaviors, forced behaviors, prescription
misuse, head injuries, self-harm, owing money, received money, no substance use, cannabis use,
chronic illness, community type (rural, urban), not taking medication, ambulance use, ER care,
being incarcerated as a minor, leaving housing because of health issues, victim of physical
violence, crisis service use, and pregnancy were significantly associated with attributing housing
loss to family disruption.
Table 3 reflects the most important predictors of housing stability for 180 days or more
by race. The most important predictor is denoted by 1. For all youth, the most important
predictor of remaining stably housed for 180 days or more was exit type. Among Black and
Latino youth, exiting homelessness using an RRH intervention was most predictive of housing
stability for 180 days or more. For White youth, the most important predictor of remaining stably
housed was exiting homelessness via self-resolve (independent of COCs). For Black youth, the
second most important predictors were exiting homelessness via self-resolve (0.69) or PSH
(0.69). While exiting homelessness by returning to family appeared in the trees of Latino and
White youth as an important predictor of housing stability at 180 days or more, exiting
homelessness by returning to family was not among the most important predictors for Black
48
youth. After exit type, shelter use appeared to be a subsequent predictor of housing stability for
Black (0.16) and White (0.11) youth. For Latino youth, after exit type, community type (rural –
0.15) followed by ambulance use (0.1) were important predictors of housing stability.
All Youth
Figure 1 shows the characteristics that were most important to remaining stable for 180
days or more. The sensitivity for this model was 99% and the specificity was 77%. The AUC
was 90% and the misclassification rate was 14% (table 4). In this model, among youth
experiencing homelessness, the most important characteristic for predicting remaining stably
housed for 180 days or more was exiting homelessness by entering RRH. From RRH, several
distinctive pathways emerged. For youth who exited homelessness via RRH, 76% remained
stably housed for 180 days or more. Of youth who exited via RRH, the second most important
predictor for housing stability was the use of shelter services. 79% of youth who exited via RRH
and used shelter services remained stable 180 days or more. For youth who exited via RRH but
did not use shelter services, 67% remained housed 180 days or more. Conditional upon youth
who exited homelessness via RRH and did not use shelter services, use or non-use of crisis
services was a subsequent predictor. 73% of youth who used RRH, did not use shelter services,
and used crisis services less than <0.500 remained stably housed 180 days or more. Only 63% of
youth who exited homelessness via RRH, did not use shelter services, and used crisis services
more than 0.500 of the time remained stably housed 180 days or more.
For youth who did not exit homelessness via RRH, the second most important predictor
was exiting homelessness independent of COC models (self-resolve). 76% of youth who did not
exit RRH but exited homelessness via self-resolve remained stable 180 days or more. Only 19%
of youth who did not exit homelessness via RRH or self-resolve remained stable 180 days or
49
more. For youth who did not exit homelessness via RRH or self-resolve, exiting homelessness by
returning to family was the next most important predictor. 68% of youth who exited
homelessness by returning to family remained stable 180 days or more whereas only 8% of
youth who did not return to family remained stable 180 days or more.
Black Youth
Figure 2 illustrates the characteristics that were most important to remaining stable for
180 days or more among Black youth. For this sample, the sensitivity was 98% and the
specificity was 81%. The AUC was 92% and the misclassification rate was 13% (table 4). The
optimal tree structure for Black youth suggests that RRH, exiting homelessness independent of
COCs, shelter, PSH, and the use of crisis services were the most important variables for
predicting stability for 180 days or more. Of the previously mentioned predictors, RRH (parent
node) appeared to be the most important.
Among Black youth who did not exit homelessness by entering RRH, only 23% remained
stable for 180 days or more. If Black youth did not exit via RRH, the second important predictor
was exiting homelessness independent of COCs (self-resolve exit). 75% of youth who exited
homelessness independently remained stable 180 days or more. For youth who did not exit
homelessness via RRH or self-resolve, PSH was the third most important predictor. 78% of
youth who exited via PSH remained stable 180 days or more relative to 9% who did not exit via
PSH, RRH, or self-resolve.
Among Black youth who exited homelessness by entering RRH, 77% remained stable
for180 days or more. For Black youth who exited homelessness by entering RRH, the use of
shelter services was the second most important predictor. Of Black youth who used shelter
services and exited homelessness via RRH, 82% remained stable for 180 days or more,
50
compared to 65% of Black youth who exited homelessness via RRH and did not use shelter
service. For Black youth who exited homelessness via RRH but who did not use shelter services,
crisis services were a subsequent predictor. 66% of Black youth who exited via RRH, did not use
shelter services, and used crisis services less than 13 times remained stable 180 days or more,
relative to 29% of Black youth who exited RRH, did not use shelter services, and used crisis
services 13 times or more.
Latino Youth
Figure 3 illustrates the characteristics that were most important to remaining stable 180
days or more among Latino youth. In this sample, the sensitivity was 97% and the specificity
was 82%. The AUC was 92% and the misclassification rate was 12% (table 4). For Latino youth,
the most important predictor (parent node) of exiting homelessness was via RRH. 75% of Latino
youth who exited via RRH remained stable, relative to 23% who did not exit via RRH. For those
who did not exit via RRH, self-driven exits were the second predictor. 80% of Latino youth who
exited homelessness independently remained stably housed for 180 days or more, relative to 14%
who did not exit homelessness via RRH or self-resolve. For Latino youth who did not utilize
RRH or a self-driven exit, family exits were the third predictor. 63% of youth who exited to
family remained stably housed, relative to 6% of youth who did not exit to family.
For Latino youth who did exit homelessness by entering RRH, ambulance use was a
subsequent predictor. 74% of youth who exited homelessness via RRH and used ambulance
services five times or less remained stably housed 180 days or more. For youth who exited via
RRH and used ambulance services five times or less, living in a rural community was the last
predictor. 75% of Latino youth who exited homelessness via RRH, used an ambulance less than
five times, and were not rural remained stably housed for 180 days or more relative to 33% of
51
Latino youth who exited homelessness via RRH, used an ambulance more than five times, and
were rural.
White Youth
Figure 4 shows the characteristics that were most important to remaining stably housed
for 180 days or more among White youth. The sensitivity for this model was 98% and the
specificity was 76%. The AUC was 89% and the misclassification rate was 15% (table 4). The
most important predictor of remaining stably housed for 180 days or more was RRH. 70% of
White youth who exited homelessness by entering RRH remained stably housed relative to 32%
of White youth who did not exit homelessness via RRH.
For White youth who did exit homelessness by utilizing RRH, the second most important
predictor was shelter. 78% of White youth who exited homelessness through RRH and used
shelter services remained stably housed for 180 days or more, compared to 68% of White youth
who exited homelessness through RRH and did not use shelter services The third most important
predictor for White youth who exited homelessness through RRH and used shelter services was
police interaction. 93% of White youth who exited homelessness through RRH, used shelter
services, and had more than 5.5 interactions with police remained stably housed, relative to 77%
of White youth who exited homelessness through RRH, did not use shelter services, and had less
than 5.5 interactions with police remained stably housed.
For White youth who did not exit homelessness by entering RRH, the second most
important predictor was exiting to family. 72% of White youth who did not exit homelessness by
RRH exited to family and remained stably housed relative to 22% of White youth who did not
exit homelessness by RRH or by returning to family. Among White youth who did not exit to
family, the third most important predictor was utilizing a self-driven exit. 76% of White youth
52
who did not exit homelessness by utilizing RRH or returning to family, exited via self-resolve
and remained stably housed relative to 8% who did not utilize a self-driven exit, family exit, or
RRH.
53
Table 3.1.
Overall Descriptive Characteristics
Descriptive Characteristics N White Black. Latino
10, 922 5212 3382 1656
Gender
Male 8487 (78%) 4075 (78%) 2586 (77%) 1300 (79%)
Female 2429 (22%) 1133 (22%) 786 (23%) 355 (21%)
Sexual Orientation
LGBTQ 3319 (30%) 1619 (31%) 984 (29%) 511 (31%)
Heterosexual 8487 (78%) 3593 (69%) 2398 (71%) 1145 (69%)
Age
< 17 3303 (30%) 1653 (32%) 953 (28%) 508 (31%)
> 17 7619 (70%) 3559 (68%) 2429 (72%) 1148 (69%)
Living Situation
Car 766 (7%) 527 (10%) 114(3%) 74 (4%)
Couch 665 (6%) 289 (6%) 252 (7%) 73 (4%)
Outdoors 798 (7%) 258 (5%) 344 (10%) 126 (8%)
Shelter 7188 (66%) 3464 (66%) 2208 (65%) 1112 (67%)
Transitional Housing 1505 (14%) 674 (13%) 463 (14%) 271 (16%)
Service Engagement
ER 1+ 6194 (57%) 2787 (53%) 2055(61%) 936 (57%)
Ambulance Use 1+ 2333(21%) 1042 (20%) 792 (13%) 351 (21%)
Crisis Services 1+ 3419 (31%) 1615 (21%) 1058(31%) 508 (33%)
Police 1+ 4603 (42%) 2121 (39%) 1470(43%) 699 (42%)
Incarceration (Minor) 2784 (25%) 1296 (25%) 898 (27%) 406 (25%)
Legal Issue 2816 (26%) 1313 (25%) 925 (27%) 403 (24%)
Correctional System 1+ 1985 (18%) 940 (18%) 602 (9%) 294 (18%)
Violence
Physical Harm 4635 (42%) 2053 (39%) 1565(46%) 716(43%)
Self Harm 2077 (19%) 914 (18%) 711 (21%) 299 (18%)
Behavior
Forced Behavior 1545 (14%) 764 (15%) 446 (13%) 241 (15%)
Risk Behavior 2614 (24%) 1293 (25%) 810 (24%) 356 (22%)
Finances
Owe Money 1861 (17%) 938 (18%) 563 (17%) 252 (16%)
Received Money 3222 (30%) 1602 (31%) 948 (28%) 467 (28%)
Fulfillment
Planned Activities 2891 (26%) 1424 (27%) 822 (24%) 454 (27%)
Need
54
No Subsistence 9506 (87%) 4509 (87%) 2962 (88%) 1451 (88%)
Reason Homeless
Ran Away 8147 (75%) 3904 (75%) 2482 (73%) 1264 (76%)
Religion 2750 (25%) 1323 (25%) 839 (25%) 393 (24%)
Family or Friends 2549 (23%) 1182 (23%) 813 (24%) 390 (24%)
Gender/Sexual Orientation 1790 (16%) 873 (17%) 537(16%) 268 (16%)
Family Violence 3288 (30%) 1570 (30%) 1013 (30%) 492 (30%)
Unhealthy Relationship 2314 (21%) 1170 (22%) 647 (19%) 336 (20%)
Health
Physical Health 455 (4%) 202 (4%) 157 (5%) 62 (4%)
Chronic Illness 562 (5%) 267 (5%) 180 (5%) 82 (5%)
HIV/Aids 453 (4%) 226 (4%) 140 (4%) 57 (3%)
Physical Disability 988 (9%) 414 (8%) 338 (10%) 168 (10%)
Avoidance of Medical Attention 1082 (10%) 516 (10%) 340 (10%) 162 (10%)
Pregnant 952 (9%) 415 (8%) 365 (11%) 122 (7%)
Substance Use
Substance Use Disrupted Housing 1449 (13%) 646 (12%) 488 (14%) 224 (14%)
Substance Use Disrupts Housing 1350 (12%) 615 (12%) 439 (13%) 200 (12%)
Not Taking Medication 780 (7%) 361 (7%) 257(8%) 124 (7%)
Prescription Misuse 447 (4%) 184 (4%) 166 (5%) 60 (4%)
Pot Use < 12 421 (4%) 180 (3%) 135 (4%) 85 (5%)
Mental Health
Mental Health Issue 1328 (12%) 622 (12%) 418(12%) 196 (12%)
Head Injury 334 (3%) 161 (3%) 113 (3%) 38 (2%)
Developmental Disability 204 (2%) 98 (2%) 64 (2%) 28 (2%)
Disability = no Independence 905 (8%) 399 (8%) 281 (8%) 165 (18%)
Exit Type
Permanent Support Housing 579 (5%) 260 (5%) 217 (6%) 66 (4%)
Rapid Re-Housing 2885 (26%) 1362 (26%) 910 (27%) 466 (28%)
Formal Exit (PSH & RRH) 3464 (32%) 1622 (31%) 1127(33%) 532 (32%)
Family 1259 (12%) 745 (14%) 273 (8%) 146 (9%)
Community Type
Urban 7285 (67%) 3093 (59%) 2584 (76%) 1251 (76%)
Suburban 2046 (19%) 979 (19%) 612 (18%) 350 (21%)
Rural 1591 (15%) 1140 (22%) 186 (6%) 55 (3%)
Stability
Stable 180 Days + 4393 (74%) 2240 (74%) 1270(74%) 625(74%)
55
Table 3.2.
Bivariate Analyses
Variable
p
RRH
<.001
PSH
<.001
Self
<.001
Family
<.001
Age
<.001
Shelter
<.001
Unhealthy or Toxic Relationship
<.001
Household Violence
<.001
Outdoors
<.001
Ran Away from Home
<.001
Gender Identity & Sexual Orientation
<.001
Risk Behaviors
<.001
Victim of Physical Violence
<.001
Car
<.001
Forced Behaviors
<.001
Hard to be housed – Substance Abuse
<.01
No Subsistence Need
<.0001
Couch
<.001
Crisis Service Use
<.001
Prescription Misuse
<.001
Police Interaction
<.001
Pot Use < 12
<.001
Rural
.0005
Leave BC Substance Use
.001
Religious Beliefs
.002
Correctional System
.002
Harm Self or Others
.002
Leave BC of Health
.003
Mental Health Issue
.003
Pregnant
.004
Avoid Medication
.004
Family & Friends caused Homelessness
.004
LGBTQ
.005
Received Money
.009
Owe Money
.014
Not Taking Medication
.016
Urban
.018
Incarceration of a Minor
.0181
Legal Issue
.027
56
Ambulance Use
.030
ER Care
.049
Table 3.3.
Feature Importance by Race & Ethnicity
Variable All Youth White Black Latino
RRH 1.00 0.93 1.00 1.00
Family 0.86 0.83 n/a 0.67
Self 0.80 1.00 0.69 0.82
Shelter 0.12 0.11 0.16 n/a
Crisis Service Use 0.09 n/a 0.07 n/a
PSH n/a n/a 0.69 n/a
Rural n/a n/a n/a 0.15
Ambulance Use n/a n/a n/a 0.10
Police Interaction n/a 0.07 n/a n/a
Note. The red boxes indicate that the risk assessment item was not associated with the race or
ethnicity group in question.
Table 3.4.
Fit Statistics by Race and Ethnicity
Sensitivity Specificity AUC MR
All Youth 0.99 0.77 0.9 0.14
White 0.98 0.76 0.89 0.15
Black 0.98 0.81 0.92 0.13
Latino 0.97 0.82 0.92 0.12
Note. AUC = Area under the curve, MR = Misclassification rate
57
Figure 3.1.
Predictors of Remaining Stable for 180 Days or More for All Youth
Stably Housed 180 Days + (All)
N = 10,922
DN Remain Stably Housed: 60%
Remained Stably Housed: 40%
Self Driven Exit
N = 6892
DN Remain Stably Housed: 81%
Remained Stably Housed: 19%
RRH
N = 8037
DN Remain Stably Housed: 73%
Remained Stably Housed: 27%
RRH
N = 2885
DN Remain Stably Housed: 24%
Remained Stably Housed: 76%
Self Driven Exit
N = 1145
DN Remain Stably Housed: 24%
Remained Stably Housed: 76%
Shelter
N = 792
DN Remain Stably Housed: 33%
Remained Stably Housed: 67%
Shelter
N = 2093
DN Remain Stably Housed: 21%
Remained Stably Housed: 79%
Family
N = 5633
DN Remain Stably Housed: 92%
Remained Stably Housed: 8%
Family
N = 1259
DN Remain Stably Housed: 32%
Remained Stably Housed: 68%
No
No
No
No
Yes
Yes
Yes
Yes
Crisis Service Use > 0.500
N = 428
DN Remain Stably Housed: 37%
Remained Stably Housed: 63%
Crisis Service Use < 0.500
N = 364
DN Remain Stably Housed: 27%
Remained Stably Housed: 73%
No Yes
58
Figure 3.2.
Predictors of Remaining Stable for 180 Days or More for Black Youth
Stably Housed 180+ (Black Youth)
$ N = 3382
DN Remain Stably Housed: 62%
Remained Stably Housed: 38%
RRH
N =2472
DN Remain Stably Housed: 77%
Remained Stably Housed: 23%
RRH
N = 910
DN Remain Stably Housed: 23%
Remained Stably Housed: 77%
Self
N =2181
DN Remain Stably Housed: 84%
Remained Stably Housed: 16%
Self
N =291
DN Remain Stably Housed: 24%
Remained Stably Housed: 76%
Shelter
N =248
DN Remain Stably Housed: 35%
Remained Stably Housed: 65%
Shelter
N =662
DN Remain Stably Housed: 18%
Remained Stably Housed: 82%
PSH
N = 217
DN Remain Stably Housed: 22%
Remained Stably Housed: 78%
PSH
N = 1964
DN Remain Stably Housed: 91%
Remained Stably Housed: 9%
No
No
No
No
Yes
Yes
Yes
Yes
Crisis Service Use < 13.2
N =241
DN Remain Stably Housed: 34%
Remained Stably Housed: 66%
Crisis Service Use >= 13.2
N =7
DN Remain Stably Housed: 71%
Remained Stably Housed: 29%
59
Figure 3.3.
Predictors of Remaining Stable for 180 Days or More for Latino Youth
60
Figure 3.4.
Predictors of Remaining Stable for 180 Days or More for White Youth
61
Discussion
The findings in this study illustrate that formal networks of support are critical to youth
maintaining housing stability for 180 days or more. In COCs, formal networks of support are
comprised of RRH and PSH interventions. PSH is a long-term housing intervention that is
allocated to chronically homeless individuals who possess greater vulnerabilities and/or risks
than their peers (Hsu, et al., 2019; Rice et al, 2018; Padgett, Henwood, & Tsemberis, 2016).
RRH is a temporary housing intervention that provides short-term rental and/or relocation
assistance (HUD, 2013). Among youth experiencing homelessness, RRH is the most allocated
exit from homelessness and more than 25% of youth experiencing homelessness utilize RRH
(Hsu, et al., 2019; Rice et al., 2018). In this study, 2,885 youth (28%) exited homelessness by
entering RRH. Of youth who utilized RRH, 76% remained stable for 180 days or more. While
other exits did appear in the CART analyses, these exits were only predictive of housing stability
for youth who did not receive an RRH intervention. Although COCs cannot allocate RRH
resources to every youth who is experiencing homelessness, the findings in this study reinforce
the importance of RRH interventions for youth (Hsu, et al., 2019).
Overall findings surrounding the effectiveness of RRH programs are mixed. For instance,
HUD’s Family Option Study concluded that, after a three-year period, RRH played a minimal
role in improving housing stability (Gubits et al, 2015). Other research highlights the potential
promise of RRH to individuals experiencing homelessness (Hsu, et al., 2019; Rice et al., 2018;
Rodriguez & Eidelman, 2017; Byrne et al., 2016; Finkel et al., 2016; Cunningham et al., 2015;
Taylor, 2014). Studies that explore RRH among youth experiencing homelessness show that
RRH is primarily allocated for youth with mid-level vulnerability scores and that age,
community type, and low service engagement decrease the probability of youth exiting
62
homelessness via RRH (Hsu et al., 2019). Aligning with previous studies, the findings from this
research suggest that RRH programs may be critical in decreasing the number of youth who
return to homelessness (Hsu Ta et al., 2019; Rice et al., 2018). As RRH is a short-term rental
program (and has proven less costly than other interventions), it is possible the youth with mid-
level vulnerability scores, as identified in other studies, are more readily able to adapt to living
independently and to the subsequent responsibilities attached to RRH interventions. Furthermore,
previous studies have found that youth who are 18 years or older are more likely to receive RRH
(Hsu Ta et al., 2019; Rice et al., 2018). While age did not surface as one of the most important
predictors in the CART analyses, age may still be a contributing factor to the success of RRH
programs for youth experiencing homelessness.
In addition to exit type, among youth experiencing homelessness, service engagement
(shelter use, crisis service use, ambulance use, and police interaction) was an important predictor
of remaining housed for 180 days or more. In studies of youth homelessness, findings
consistently suggest that youth need a holistic set of services to successfully exit homelessness
(Aviles & Helfrich, 2004; de Winter & Noom, 2003). Even so, research has consistently found
low rates of service utilization among youth experiencing homelessness (Hsu et al., 2019; Hill et
al., 2020 – under review; Rice et al., 2018; De Rosa et al., 1999; Kipke, Montgomery, &
MacKenzie, 1993). An array of behavioral and individual factors contributes to different rates of
service utilization among youth experiencing homelessness (Berdahl et al., Hoyt, & Whitbeck,
2005; Morton et al., 2018; Morton et al., 2017; Tyler et al., 2012). These may include race,
sexual orientation, community, NST score, foster care experience, and mental health status (Hill
et al., 2020 – under review; Hsu et al., 2019; Rice et al., 2018). Similar to findings reported in
previous studies, youth who were engaged in services had a higher probability of remaining
63
stably housed for 180 days or more (Hsu, et al., 2019; Rice et al., 2018). For homeless youth,
consistent engagement with shelters and RRH or PSH interventions may yield more exposure to
providers and longer-term housing placements (Hsu, et al., 2019). Service engagement may also
heighten the visibility of homeless youth, which may reduce the likelihood of youth becoming
lost in the system. In particular, for Black and White youth who exited homelessness through
RRH interventions, shelter use was a secondary predictor. An earlier study found that relative to
60% of their non-Black peers, 65% of Black youth used shelter services (Hill et al., 2020 – under
review). As Black youth are using shelter services, the shelter may be an appropriate location to
further engage Black youth in supportive services and assess their immediate needs. Further, as
Black youth are visible in the shelter system, the shelter system may have one of the best chances
at tracking and monitoring Black youth as they navigate the housing system. Future studies
should investigate the salience of race to service engagement and subsequent housing stability.
Limitations
Although these data provide visibility to the factors that predict housing stability at 180
days or more, there are notable limitations. First, as this is an administrative data set, this
analysis is constrained to two years of observations. For instance, using 180 days as the time
point of interest may not reveal long-term racial differences that may emerge at 365 days or 730
days. Additionally, this data does not capture youth who may have been stable for 180 days or
more within a local COC but returned to homelessness within 180 days in a different community.
Explicitly stated, if youth leave an RRH intervention and do not return to the COC for shelter
services, providers have little visibility into a youth’s subsequent trajectory. Consequently, a
youth may be coded into the “remained housed for 180 days or more” category because the
youth did not resurface at a corresponding COC or the youth has been lost. Furthermore, while
64
this data set includes data on service use before the exit and provides information on exit type,
the services offered in RRH interventions are unknown. Also unknown is the threshold at which
services become effective and/or the amount of monetary support that is sufficient in maintaining
stability. The findings in this study suggest that exit type matters, as does the service use prior to
and within RRH interventions. However, the current dataset does not shed any light on this
phenomenon. Lastly, this data does not capture the factors or characteristics that may have aided
youth in remaining stable for 180 days or more. These may have included enduring relationships
with family, employment, government assistance (e.g., EBT or Medi-Cal), or access to other
stabilization programs. Despite these limitations, the findings in this study offer important
considerations for housing providers.
Conclusion
Echoing previous research, the findings in this study elucidate RRH as a viable path
forward. From a policy perspective, expanding RRH provisions may seemingly be a necessary
and appropriate solution. This can be attributed to the cost-effectiveness associated with RRH.
However, COCs cannot allocate RRH interventions to every youth experiencing homelessness
nor is RRH an appropriate resource for every youth experiencing homelessness. In light of this,
additional research is needed to evaluate the impact and effectiveness of RRH over a sustained
period. Specifically, we do not know how much exposure to RRH interventions is sufficient to
maintain long-term housing stability. Additionally, the findings in this study raise the following
question: what kind of services are provided through RRH interventions and how do these
services impact success in RRH? While these data cannot answer this question, this topic should
be explored in future work. Before policymakers can consider an expansion of RRH provisions,
further research is needed to examine the characteristics of RRH programs that contribute to their
65
success as well as an exploration of youth who have the greatest success with RRH interventions.
The future of homeless youth rests on service engagement and successful exits from
homelessness.
66
Chapter 4
Every night, as many as 568,000 people experience homelessness (Henry et al., 2019).
Aware of the rising homeless population, communities around the country have implemented
COCs to provide and allocate housing resources (U.S. Department of Housing & Urban
Development). However, COCs cannot house or serve every person experiencing homelessness
(Hsu et al., 2019; Rice et al., 2018). To prioritize those in greatest need of these services, COCs
use assessment tools with embedded algorithms to prioritize and allocate housing resources. For
example, several studies have found that mental health issues, substance use, and physical
violence are frequently associated with housing loss (Mojtabai, 2005; Padgett et al., 2008).
Consequently, algorithms embedded in housing assessment tools are often skewed toward these
vulnerabilities.
Individuals who disclose mental health issues, substance use, or physical violence, or
seek treatment receive a higher score than individuals who do not and are deemed to be in
greater need of services or specific services. The problem with this approach is that the
algorithms incorporated in housing assessment tools are skewed toward a finite set of
vulnerabilities—vulnerabilities that may vary in marginalized communities. For example,
housing assessments routinely capture mental health vulnerabilities. Even so, in the Black
community, there is a historical distrust of medical providers. Consequently, Black people are
less likely to use mental health services or report mental health issues (Breland-Noble, 2004;
Coker et al., 2009; Lindsey et al., 2010; Caldwell et al., 2016; Motley & Banks, 2018). As Black
people are less likely to use mental health services and less likely to report mental health issues,
housing assessment tools that place greater weight on mental health vulnerabilities may place
Black people at an algorithmic disadvantage. Since Black people comprise 40% of the homeless
67
population, it is rather concerning that housing assessment tools may not be capturing their
vulnerabilities adequately. Black people, therefore, may not be prioritized correctly for housing
or assigned appropriate services.
While homelessness is not exclusively Black, there is a specific set of systematic
interactions (input) that may act as catalysts in creating and maintaining Black homelessness
(output) in the United States. I surmise that there is an “algorithm of Black homelessness” at
play. I define this concept as an addictive formula that relies on a set of underlying identity rules
set forth by a history of systemic oppression. Identities are compounded (not necessarily equally
but potentially hierarchically) into discriminatory systems that yield homelessness as an output.
The collective sum of identities and the relationship of those identities to each other—plus the
hierarchical placement of those identities in systems that are historically and inherently
unequal—generate unique vulnerabilities among Black Americans experiencing homelessness.
The problem, as I have explored in this study, is that current housing assessment tools do not
factor in algorithms of Black homelessness. Thus, Black people are being systematically deemed
ineligible for several services, resulting in contemporary redlining. In some cases, the data that
are used to compute who is at the highest risk of remaining unsafely housed do not
mathematically account for race at all. In other cases, race is not scored within housing
assessment algorithms. Finally, when race is a factor that is inputted into one of these
coordinated assessment tools, those data seem to serve a purely demographic purpose, and not as
possible breadcrumbs that will lead to a longer historical story of displacement and economic
precarity.
Among homeless populations, the VI-SPDAT and the NST are frequently used to assess
medical, social, and mental health vulnerabilities (Brown et al., 2018; Community Solutions &
68
OrgCode Consulting, Inc, 2014). Although the VI-SPDAT is primarily administered to adults
and the NST is administered to youth, both assessment tools generate recommendations for PSH,
which is recommended for individuals with the greatest vulnerabilities and includes permanent
housing subsidies as well as intensive support services or RRH (Hsu et al., 2019; Brown et al.,
2018; Rice et al., 2018). The VI-SPDAT may also make a recommendation for mainstream
housing, which often takes the form of affordable housing options in the private market for
individuals with minimal vulnerabilities (Brown et al., 2018).
Despite widespread utilization of the NST and VI-SPDAT, there is limited research that
establishes the NST and VI-SPDAT as evidence-based assessments (Hsu et al., 2019; Rice et al.,
2018; Brown et al., 2018). General critiques of the previously mentioned tools highlight the
absence of items that align with predictors of housing stability for individuals triaged into PSH,
RRH, or, in the case of the VI-SPDAT, mainstream affordable housing (Brown et al., 2018).
Additionally, research has not yet identified the specific items included in current versions of the
NST and VI-SPDAT that predict the success of one housing intervention over another as
determined or suggested by the NST or VI-SPDAT score (Balagot et al., 2019; Brown et al.,
2018). Lastly, NST and VI-SPDAT assessment scores rely on self-reporting. This implies that
the scores are only as accurate as the information that is provided. With self-reporting, the
accuracy of responses is dependent on an individual’s rapport with the interviewer, trust of the
system, and provider advocacy (Fritsch et al., 2017). Additionally, racial discordance (between
provider and client) may further skew self-reporting. These problems may result in scores that
are inaccurate and unreliable. Consequently, individuals experiencing homelessness will not
always be allocated into an optimal housing resource or triaged into an appropriate level of
services. While this is undoubtedly a problem for the homeless population as a whole, for Black
69
people—who are disproportionately represented in the homeless population and who historically
underreport medical, social, and mental health vulnerabilities—the impact of using standardized
tools that have not been rigorously and culturally tested for validity and reliability may further
reduce the likelihood of ever exiting homelessness.
Tool Disparities
Research is beginning to explore potential racial disparities in standardized assessment tools.
For youth, the TAY-VI-SPDAT is the most widely used assessment tool. (Rice et al., 2018; Rice,
2017). The TAY-VI-SPDAT assesses vulnerability, and youth with the highest score (greatest
need) are linked to appropriate services (Rice et al., 2018). A recent study found that the VI-
SPDAT, an assessment tool used to allocate housing resources to homeless adults and inform the
TAY-VI-SPDAT, perpetuates racial inequities by prioritizing and capturing vulnerabilities that
White individuals are more likely to report (CES Racial Equity Analysis of Assessment Data,
2019). This appears evidenced by the inclusion of mental health status and the utilization of
formal services as items that are associated with higher prioritization scores on the TAY-VI-
SPDAT and VI-SPDAT (Hsu et al.,2019; Rice et al., 2018; CES Racial Equity Analysis of
Assessment Data, 2019).
The inclusion of mental illness and mental health treatment as eligibility items and
components of care are prominent in housing interventions because they have been associated
with initial housing loss and subsequent housing instability (Mojtabai, 2005; Padgett et al.,
2008). However, the aforementioned findings are concerning because housing assessments that
include the presence or treatment of mental illnesses as items of eligibility may not be
appropriate because Black people are, in general, less likely to use mental health services or
report mental health issues (Breland-Noble, 2004; Coker et al., 2009; Lindsey et al., 2010;
70
Caldwell et al., 2016; Motley & Banks, 2018). In adult homeless literature, for example, racial
disparities in mental health reporting and utilization of mental health services have been
repeatedly documented (Motley & Banks, 2018; Caldwell et al., 2016; Lindsey et al., 2010;
Coker et al., 2009). A 2018 study found that Black males with trauma histories are less likely to
utilize mental health services than other ethnic groups (Motley & Banks, 2018). An earlier study
attributed the underutilization of therapy services by Black men to mental health stigmas in the
Black community (Lindsey et al., 2010).
While there is not as much empirical evidence for Black youth experiencing homelessness,
several studies suggest that there are disparities in mental health utilization for Black youth
overall (Coker et al., 2009). Black youth are more likely to underutilize mental health services
than their White counterparts (Breland-Noble, 2004; Caldwell et al., 2016). One of the few
studies that did explore mental health service use among Black youth identified ethnicity,
perceived socioeconomic status, self-reported health, and psychiatric disorders as predictors
(Caldwell et al., 2016). The authors of that study attributed the lower utilization of mental health
services by Black youth to religious beliefs, financial barriers, and the fear of community stigma
(Caldwell et al., 2016). Moreover, Black youth may not be achieving long-term housing stability
because they are not being referred to the appropriate services in the first place. When their
unique vulnerabilities are not identified or assessed because the allocation criteria are skewed to
vulnerabilities that the dominant majority are more likely to report and experience, they are not
likely to be triaged as high-risk.
Alongside housing allocation tools and housing interventions, providers have also prioritized
service engagement (CES Racial Equity Analysis of Assessment Data, 2019). Even so, among
homeless youth, studies consistently show low rates of service utilization and varying levels of
71
engagement across service types (De Rosa et al., 1999; Kipke, Montgomery, & McKenzie,
1993). In fact, as it relates to mental health, less than 50% of YEH use mental health services
(Milburn et al., 2006; Sweat et al, 2008; Tyler et al., 2012). When presented with services such
as employment or education, YEH are more likely to instead pursue basic needs such as food,
clothing, or showers (De Rosa et al., 1999; Pergamit et al., 2010; Sweat et al., 2008). Echoing the
findings above, using service engagement as an item of prioritization for the allocation of
housing resources—and as a component of housing interventions—may not assist Black youth in
achieving long-term stability either, since service engagement is not the most pressing
vulnerability.
While there is a plethora of research exploring correlates of service use among YEH, and
although service use is used in assigning prioritization scores, findings around correlates of
service use among Black YEH are mixed. For instance, relative to their non-Black peers, youth
who identified as Black had higher rates of service use (e.g., employment assistance, shelter, and
healthcare) (Hill et al., 2019 – under review). However, the same study found that 64% of Black
youth reported no use of therapy services over the last 30 days (Hill et al., 2019 – under review).
Similarly, another study found that YOC (inclusive of Black youth) used shelters more
frequently than their White peers (De Rosa et al., 1999). An earlier study suggested that racial
and ethnic minorities are less likely to receive services than their White peers (Garland et al.,
2005), which aligns with even earlier findings attributing the lack of service use by Black youth
to provider discrimination (Gerber, 1997). As findings around Black youth and service use are
mixed, factoring service use into prioritization scores and allowing it to assist youth in obtaining
a higher prioritization score may not be capturing Black youth who are in the greatest need of
housing resources.
72
The inclusion of mental health and service use in prioritization scores highlights a potential
gap in housing assessment and eligibility tools for minoritized communities. Most notably,
people from minoritized communities who either (1) under-report mental illness, (2) do not
report mental illness, (3) do not experience mental illness, or (4) experience mental health
symptoms different from their counterparts, may be placed in interventions that do not address
their pertinent needs. For example, people from minoritized communities who are being placed
in interventions like Housing First that are effective only for people with dominant mental health
symptoms or illnesses, may not be being placed appropriately. Therefore, they may not achieve
long-term stability since the root causes of their initial stability have not yet been addressed. If
current housing interventions are focused on mental health and service use but there are
subgroups of homeless people who historically under-report or simply do not engage with
service providers due to historical trauma or community stigmas, the tools developed to
adequately assess and allocate housing resources may be perpetuating the very gaps they were
intended to fill. Consequently, it is necessary to know what those needs are and how
interventions can best address them.
Theoretical Foundations
Motivated by studies that have highlighted growing inequities in housing prioritization
tools, COCs are in need of theoretical frameworks that will allow for the inclusion of race and
other salient identities in assessment instruments (CES Racial Equity Analysis of Assessment
Data, 2019; Hsu et al., 2019; Brown et al., 2018; Rice et al., 2018; Milburn et al; 2006). Critical
Race Theory (CRT), initially developed by Derrick Bell, is one such concept that could
undergird the shift in assessment data that I am suggesting. Bell has written that race interacts
with systems of power to perpetuate and sustain oppression (Delgado & Stefancic, 2017). The
73
use of CRT as an analytical framework allows for the consideration of race in housing
recommendations and, ultimately, encourages a critical examination of the relevance of parental
incarceration and/or foster care involvement to homeless experiences. Furthermore, housing
providers and homeless scholars should consider minoritized identities and the interactions of
these identities in various legal systems as experiences that are attached to long-term
consequences. For instance, in this dissertation, couch surfing was the most important predictor
of exiting homelessness for Black youth. However, couch surfing was not a predictor of
remaining stably housed for 180 days or more; exiting homelessness via a formal network of
support (RRH or PSH) was. For Black youth, therefore, these findings suggest that family
reunification may not be an exit from homelessness at all. Instead, couch surfing may be a form
of unstable housing in between longer stints of homelessness. The HMIS data set used in this
analysis provides no insight into why this might be. To better understand these findings, more
information is needed about a youth’s initial housing loss. Critical race theory challenges social
work scholars, therefore, to examine the role that involvement in judicial systems has played
both generationally and historically. For example, one possible explanation for the frequent
couch-surfing phenomenon could be that the person with whom the youth is staying may not be
housed very stably either. Or, conversely, youth who have been involved in the judicial system
may have outstanding probation requirements (mandated and random visits from probation
officers) and additional surveillance that forbid them from remaining housed with their host for
long periods of time.
These kinds of interlocking forces of oppression are best described by the term
“intersectionality,” which was coined by the legal scholar Kimberlé Crenshaw. Crenshaw has
written that rather than one discriminatory factor superseding another factor, there are instead
74
many forms of bias working in tandem to create differential experiences. For example, a case
worker’s racism, sexism, and ageism might coalesce to deem a White, female, 14-year-old as
more worthy of immediate care than a Black, male, 18-year-old. Current assessment tools do not
account for pre-existing provider biases or agency biases in tool administration practices. While
Crenshaw’s initial conception of intersectionality posits that Black women experience oppression
at the intersection of multiple identities (i.e., race, gender, and class) (Crenshaw, 1989), the
theory of intersectionality allows and encourages the examination of the interactions that occur
among multiple identities to create differential experiences with exclusion and subordination
(Howard & Navarro, 2016).
As it relates to housing prioritization, the theory of intersectionality suggests that a deeper
assessment of identity is warranted. Current research indicates that homeless youth are
disproportionately Black and/or LGBTQ, and predominantly male (HUD, 2017; Morton et. al.,
2017). Studies also indicate that youth who identify as racial, sexual, or gender minorities are
marginalized at rates higher than those of their White cisgender peers (Berdahl, Hoyt, &
Whitbeck, 2005). Assessments tools, therefore, should account for these intersectional
identities—and the biases that each identity may elicit—in housing recommendations.
Recommendations
The NST and VI-SPDAT capture perceived and observed vulnerabilities of individuals
experiencing homelessness. Unfortunately, these tools provide limited visibility into the
algorithms of persistent Black homelessness. This study has found, for example, that exiting
homelessness via a formal network of support is the most important predictor of remaining stably
housed for 180 days or more. Although enlightening, this data point does not identify the specific
characteristics of those formal networks of support that enable subsequent stability. These
75
findings also do not highlight the individual characteristics that encourage or contribute to
remaining housed for 180 days or more. Similarly, across race and ethnicity, the CART analyses
consistently ranked RRH as the best predictor of remaining stably housed. While RRH programs
cannot house or serve every person experiencing homelessness, 25% of YEH are triaged into an
RRH intervention (Hsu et al., 2019; Rice et al., 2018). Relative to PSH, RRH interventions are
allocated to youth with mid-level vulnerabilities (Hsu et al., 2019; Rice et al., 2018). The irony in
this finding is that RRH is a fairly new housing intervention and because of this, the role that
RRH plays in housing stability is unclear. It is also unknown if there are individual
characteristics that predict who will be most successful in RRH interventions. As RRH is more
cost-efficient than other housing programs, increasing the accessibility and availability of RRH
interventions could play a critical role in housing homeless youth. However, to truly reduce the
number of YEH, researchers need a better understanding of what works, why it works, and for
whom it works.
Family reunification as a false data point. Family reunification may be a viable exit from
homelessness (Hsu et al; 2019; Rice et al., 2018). However, the NST does not operationalize
returning to family in a way that is insightful or interpretable. For example, while returns to
family have been coded as an exit type, relatively little is known about how providers define
returns to family and what returns to family look like. Additionally, how youth describe returns
to family, to whom youth are returning, or where youth are returning are unknown. It is also
plausible that although youth are returning to family, youth are not ceasing to be homeless.
Returns to family, therefore, may preclude youth from fitting HUD’s literal definition of
homelessness even though they have not escaped true homelessness. I return to the issue of
couch-surfing. Although one cannot infer causality from the data utilized in this study, one can
76
suspect that “exiting to family” may be an extension of couch surfing rather than a legitimate exit
from homelessness. US studies of youth homelessness have not explored this subject in detail.
International findings suggest, however, that White youth are more likely to sleep on the streets,
while minoritized youth have higher odds of seeking temporary shelter with friends or relatives
(Somerville & Steele, 2002). The “invisibility” of Black youth on the streets suggests that they
may also be absent from need assessments, street counts, or other inventories used to determine
services, allocate resources, and identify accurate housing statuses. It is the position of this paper
that Black youth are not ceasing to be homeless when they return to family. Instead, Black youth
are just ceasing to be visible, possibly due to safety concerns, such as police violence, gang
violence, or fear of brushes with the judicial system. This may partially explain why we see high
numbers of Black youth accessing short-term services such as those offered at drop-in centers
but far lower numbers in formal systems of care. As Black youth may not be experiencing literal
homelessness on a consistent basis due to their ability to temporarily find housing off the streets,
better assessment tools that incorporate these factors would perhaps make the current system of
allocating assistance more equitable. It would also make Black youth who fall into this nebulous
kind of homelessness more visible to caseworkers.
The call for deeper data. Persons experiencing homelessness disproportionately belong to
marginalized identities. Therefore, it seems imperative to include marginalized identities as
components of assessment scores. Most notably, race has become particularly salient in studies
of homelessness. For instance, although Black people comprise 13% of the general population,
they comprise 40% of the homeless population (Henry et al., 2019). Similarly, Black youth
account for 34% of youth experiencing homelessness and, compared to their White counterparts,
77
youth who identify as Black or African American are 83% more likely to be homeless (Henry et
al., 2017; Morton, 2017).
In addition to race, sexual orientation requires considerable attention. A previous study found
that 68% of LGBT youth experiencing homelessness attribute housing loss to family rejection
(Durso & Gates, 2012). In this study, sexual orientation was the most important predictor of
attributing housing loss to family disruption. Previous reports have found that LGBT youth are
120% more likely than other youth to be homeless (Morton, 2017). A report has also suggested
that as many as 40% of homeless youth identify as LGBT (Shelton & Winkelstein, 2014). LGBT
youth face elevated risks of homelessness and have higher rates of victimization, substance use,
mental health problems, and suicidal acts than their heterosexual peers (Keuroghlian et al.,2014).
Considering current statistics, an intersectional approach to how assessment scores are calculated
may be critical to correctly triaging youth into housing and appropriate services. This effort can
potentially be supplemented with questionnaires that further assess the impact of marginalized
identities on acquiring and maintaining housing.
Better assessment and evaluation of mitigating factors (e.g., family disruption).
Among YEH, an array of studies suggest that disruptions in the family structure mediate housing
loss (Edidin, 2012; Shelton, 2009; Ferguson, 2009; Mallet, 2009; Mallet, 2005; Reeg, 2003).
Family disruption is often characterized by a variety of factors, but neglect, abuse, and family
violence are cited most often (Barker, 2012; Toro et al., 2011). A recent study exploring family
separation among homeless youth found that in addition to abuse or neglect, a lack of resources
and the inability to provide necessary care often result in youth being kicked out or voluntarily
leaving the family home (Barker, 2012). Housing loss attributed to family relationships often
deter youth from returning to family or seeking future support (Edidin, 2012; Shelton et al.,
78
2009; Ferguson, 2009; Mallet, 2009; Mallet, 2005; Reeg, 2003). Additionally, family conflict
within the immediate family, historically or contemporarily, may prevent youth from being able
to identify or geographically locate extended family networks (Barker, 2012). Undoubtedly,
familial relationships shape youth’s trajectories into homelessness.
External factors that mediate family disruption may highlight the relevance of race to this
discussion. For instance, among Black youth, disproportionate representation in systems such as
juvenile justice and foster care may disrupt traditional family networks. For example, relative to
their white counterparts, Black youth are more likely to report foster care experience (Auerswald
& Puddefoot, 2012) and less likely to be reunited with their families (Dettlaff et al., 2011).
Furthermore, while in foster care, Black youth have fewer visits with family members, and spend
more time in out of home placements (Dettlaff et al., 2011). Removals that result in the
incarceration of a parent, the separation of siblings, and/or permanent custody loss may further
cripple existing family relationships (Roberts, Shattered Bonds, pg. 228 – 229). As returns to
family have the potential to be a cost-effective and viable exit from homelessness, housing
providers need to assess the factors that mediate family disruption and ultimately cause housing
disruption in the first place. For Black youth who remain on the receiving end of systematic
inequities, the onset of family disruption may not only be mediated differently but also
manifested differently.
Exploring the “why” of housing loss, especially for Black youth, may have critical policy
implications. Specifically, systematic inequities that alter family structures or physically separate
family members may instigate the onset of breakdowns in the family structure (Freeman, 2018).
For Black youth who are assigned scores that render them ineligible for RRH or PSH, returns to
family may be one of few remaining options. If this is the case, housing providers must
79
understand the factors that created network disruption and craft ways to address network
disruption before Black youth can successfully return to family and remain with family.
Cross-validation. The vulnerabilities captured by housing assessments are self-reported.
As individuals experiencing homelessness may have had contact with other systems, data
integrated from systems such as child welfare, juvenile justice, or hospital systems may enable
providers to better assess the needs of those experiencing homelessness. For marginalized
populations, there may be sensitivity around subsets of questions. For example, undocumented
youth and single adults may be hesitant to disclose legal system involvement or any issues that
may place them or their families at risk of discovery or deportation. Among Black people
experiencing homelessness, fear of discrimination or a distrust of providers may prevent the
reporting of substance abuse issues, mental health issues, or involvement in legal systems. While
the previously mentioned vulnerabilities should not prevent one from obtaining housing or
receiving support services, among marginalized individuals, history suggests otherwise. The
integration of data from other service systems may also highlight additional vulnerabilities or
illustrate the ways in which characteristics such as race and sexual orientation interact in various
systems to create experiences among marginalized groups that do not mirror their White
counterparts. For instance, incarceration data suggests that Black Americans, especially Black
males, are disproportionately arrested, prosecuted, convicted, and sentenced (Chin, 2002).
Expanded further, Black males are not treated the same as their White counterparts in the
criminal justice system. As a result of these experiences (longer sentences and harsher
punishments for identical crimes), via collateral consequences, Black males may have longer
durations of homelessness (attributed to the inability to find employment or housing and, in some
states, the inability to apply for government aid). This recommendation is not a call for increased
80
surveillance. Instead, I am recommending that providers integrate data from other service
systems to better assess housing needs and recommend appropriate services.
Conclusion and Implications
The findings highlighted in this study situate the allocation and prioritization of housing
resources into an argument of equity versus equality. To better triage people into appropriate
services, the NST and VI-SPDAT should consider an assessment model that recognizes how
various identities have interacted in historically unjust systems, which would imply, ultimately,
that some populations experience homelessness differently. The goal of future assessment tools,
then, should not be to allocate and prioritize housing resources equally, but to allocate and
prioritize housing resources equitably. This would mean supplementing existing assessment tools
with items that encourage increased consideration of individuals from minoritized identities and
communities, rather than adhering to the current colorblind, one-size-fits-all approach. It is not
my position that the NST and VI-SPDAT are asking the wrong questions. Instead, I am
suggesting that the NST and VI-SPDAT are not asking enough of the right questions. This means
that provider assessments should capture minoritized risks and vulnerabilities—especially when
the youth in question are the likely heirs of centuries-long housing disparities. Providers who
administer housing assessments like the NST or VI-SPDAT could benefit from the approach
outlined in this paper because it would better equip them to adequately identify housing needs,
allocate appropriate resources, and correctly prioritize those in greatest need.
81
Bibliography
Aguiar F, Almeida L, Ruffino-Netto A, Mello F, Werneck G, Aguiar F. Classification and
regression tree (CART) model to predict pulmonary tuberculosis in hospitalized
patients. BMC Pulmonary Medicine. 2012;12(1):40-40. doi:10.1186/1471-2466-12-40
Alexander, M. (2020). The new Jim Crow: Mass incarceration in the age of colorblindness. The
New Press.
Alvi, S., Scott, H., & Stanyon, W. (2010). “We’re locking the door”: family histories in a sample
of homeless youth.(Report). The Qualitative Report, 15(5).
Auerswald, C., and Puddefoot, G. (2012). “Comparing white and African American homeless
youth in San Francisco: Research findings and policy implications.” California Homeless
Youth Project Voices from the Street.
Aviles, A., & Helfrich, C. (2004). Life skill service needs: Perspectives of homeless
youth. Journal of youth and adolescence, 33(4), 331-338.
Balagot, C., Lemus, H., Hartrick, M., Kohler, T., & Lindsay, S. P. (2019). The homeless
Coordinated Entry System: the VI-SPDAT and other predictors of establishing eligibility
for services for single homeless adults. Journal of Social Distress and the Homeless,
28(2), 149-157.
Barker, J. (2012). Social capital, homeless young people and the family. Journal of Youth
Studies, 15(6), 730–743. https://doi.org/10.1080/13676261.2012.677812
Barman-Adhikari, A., & Rice, E. (2014). Social networks as the context for understanding
employment services utilization among homeless youth. Evaluation and program
planning, 45, 90-101.
Barman-Adhikari, A., Bowen, E., Bender, K., Brown, S., & Rice, E. (2016, October). A social
capital approach to identifying correlates of perceived social support among homeless
youth. In Child & Youth Care Forum (Vol. 45, No. 5, pp. 691-708). Springer US.
Bell, G. S. (2002). In the black: A history of African Americans on Wall Street. John Wiley &
Sons.
82
Benjamin, R. (2019). Race after technology: Abolitionist tools for the new jim code. Social
Forces.
Berdahl, T. A., Hoyt, D. R., & Whitbeck, L. B. (2005). Predictors of first mental health service
utilization among homeless and runaway adolescents. Journal of Adolescent
Health, 37(2), 145-154.
Blackmon, D. A. (2009). Slavery by another name: The re-enslavement of black Americans from
the Civil War to World War II. Anchor.
Breiman, L., Friedman, J., Olshen, R. A., and Stone, C. J. (1984). Classification and Regression
Trees . Belmont, CA: Wadsworth.
Breland-Noble, A. M. (2004). Mental healthcare disparities affect treatment of Black
adolescents. Psychiatric Annals, 34(7), 534-538.
Brophy, A. L. (2003). Reconstructing the dreamland: The Tulsa riot of 1921: Race, reparations,
and reconciliation. Oxford University Press.
Brown, M., Cummings, C., Lyons, J., Carrión, A., & Watson, D. P. (2018). Reliability and
validity of the Vulnerability Index-Service Prioritization Decision Assistance Tool (VI-
SPDAT) in real-world implementation. Journal of Social Distress and the
Homeless, 27(2), 110-117.
Bushway, S. D., & Sweeten, G. (2007). Abolish lifetime bans for ex-felons. Criminology & Pub.
Pol'y, 6, 697.
Byrne, T., Treglia, D., Culhane, D. P., Kuhn, J., & Kane, V. (2016). Predictors of homelessness
among families and single adults after exit from homelessness prevention and Rapid Re-
Housing Programs: Evidence from the Department of Veterans Affairs Supportive
Services for Veteran Families program. Housing Policy Debate, 26(1), 252-275.
Caldwell C.H., Assari S., Breland-Noble A.M. Handbook of Mental Health in African American
Youth 2016. Springer International Publishing; Zurich, Switzerland: 2016. The
Epidemiology of Mental Disorders in African American Children and Adolescents; pp.
3–20.
Canaday, N., Reback, C., & Stowe, K. (2015). Race and Local Knowledge: New Evidence from
83
the Southern Homestead Act. The Review of Black Political Economy, 42(4), 399-413.
Carey, C. A. (2004). No second chance: People with criminal records denied access to public
housing. U. Tol. L. Rev., 36, 545.
Carter III, G. R. (2011). From exclusion to destitution: Race, affordable housing, and
homelessness. Cityscape, 33-70.
Cauce, A. M., Paradise, M., Ginzler, J. A., Embry, L., Morgan, C. J., Lohr, Y., & Theofelis, J.
(2000). The characteristics and mental health of homeless adolescents: Age and gender
differences. Journal of Emotional and Behavioral Disorders, 8(4), 230-239.
Chan, H., Rice, E., Vayanos, P., Tambe, M., & Morton, M. (2017, October). Evidence from the
past: Ai decision aids to improve housing systems for homeless youth. In 2017 AAAI Fall
Symposium Series.
Chan, H., Rice, E., Vayanos, P., Tambe, M., & Morton, M. (2018, September). From Empirical
Analysis to Public Policy: Evaluating Housing Systems for Homeless Youth. In Joint
European Conference on Machine Learning and Knowledge Discovery in Databases (pp.
69-85). Springer, Cham.
Chin, G. (2002). Race, the war on drugs, and the collateral consequences of criminal
conviction. Journal of Gender, Race and Justice, 6(2).
Choi, S. K., Wilson, B. D., Shelton, J., & Gates, G. J. (2015). Serving our youth 2015: The needs
and experiences of lesbian, gay, bisexual, transgender, and questioning youth
experiencing homelessness.
Cimbala, P. A. (1989). The Freedmen's Bureau, the Freedmen, and Sherman's Grant in
Reconstruction Georgia, 1865-1867. The Journal of Southern History, 55(4), 597-632.
Coker, T. R., Elliott, M. N., Kataoka, S., Schwebel, D. C., Mrug, S., Grunbaum, J. A., ... &
Schuster, M. A. (2009). Racial/ethnic disparities in the mental health care utilization of
fifth grade children. Academic Pediatrics, 9(2), 89-96.
Community Solutions, & OrgCode Consulting, Inc. (2014). The Vulnerability Index-Service
Prioritization Decision Assistance Tool (VI-SPDAT) manual for single person
households. Retrieved from http://www.orgcode.com/ wordpress/wp-
84
content/uploads/2014/08/VI-SPDATManual-2014-v1.pdf
Community Solutions, & OrgCode Consulting, Inc. (n.d.). VI-SPDAT Version 2 and Family VI-
SPDAT Version 2. Retrieved from http://www.community.solutions/sites/
default/files/about_the_vi-spdat_2.0.pdf
Connerly, C. E. (2002). From racial zoning to community empowerment: The interstate highway
system and the African American community in Birmingham, Alabama. Journal of
Planning Education and Research, 22(2), 99-114.
Copeland RW. In the Beginning: Origins of African American Real Property Ownership in the
United States. Journal of Black Studies. 2013;44(6):646-664.
doi:10.1177/0021934713506010.
Crenshaw, K. (1989). Demarginalizing the intersection of race and sex: A black feminist critique
of antidiscrimination doctrine, feminist theory and antiracist politics. u. Chi. Legal f.,
139.
Crenshaw, K. W. (2010). Twenty years of critical race theory: Looking back to move
forward. Conn. L. Rev., 43, 1253.
CSH. (2019, March 22). OrgCode, Community Solutions & Launch Next Step Tool for Youth.
https://www.csh.org/2015/06/orgcode-community-solutions-csh-launch-next-step-tool-
for-youth/
Cunningham, M., Gillespie, S., & Anderson, J. (2015). Rapid Re-housing.
Delgado, R., & Stefancic, J. (2017). Critical race theory: An introduction. NYU Press.
De la Haye, K., Green Jr, H. D., Kennedy, D. P., Zhou, A., Golinelli, D., Wenzel, S. L., &
Tucker, J. S. (2012). Who is supporting homeless youth? Predictors of support in
personal networks. Journal of Research on Adolescence, 22(4), 604-616.
De Rosa, C. J., Montgomery, S. B., Kipke, M. D., Iverson, E., Ma, J. L., & Unger, J. B. (1999).
Service utilization among homeless and runaway youth in Los Angeles, California: Rates
and reasons. Journal of Adolescent Health, 24(3), 190-200.
Dettlaff, A. J., Rivaux, S. L., Baumann, D. J., Fluke, J. D., Rycraft, J. R., & James, J. (2011).
85
Disentangling substantiation: The influence of race, income, and risk on the
substantiation decision in child welfare. Children and Youth Services Review, 33(9),
1630-1637.
de Winter, M., & Noom, M. (2003). Someone who treats you as an ordinary human being…
Homeless youth examine the quality of professional care. British Journal of Social
Work, 33(3), 325-338.
Durso, L.E., & Gates, G.J. (2012). Serving Our Youth: Findings from a National Survey of
Service Providers Working with Lesbian, Gay, Bisexual, and Transgender Youth who
are Homeless or At Risk of Becoming Homeless. Los Angeles: The Williams Institute.
Edidin JP, Ganim Z, Hunter SJ, Karnik NS. The mental and physical health of homeless youth: a
literature review. Child Psychiatry Hum Dev. 2012;43(3):354–375.
Ferguson, K. M. (2009). Exploring family environment characteristics and multiple abuse
experiences among homeless youth. Journal of interpersonal violence, 24(11), 1875-
1891.
Finkel, M., Henry, M., Matthews, N., Spellman, B., & Culhane, D. (2016). Rapid Re-Housing
for Homeless Families Demonstration Programs Evaluation Report Part II:
Demonstration Findings—Outcomes Evaluation. Available at SSRN 3055243.
Free, L. (2009). Gendering the Constitution: Manhood, Race, Woman Suffrage and the
Fourteenth Amendment, 1865-1866.
Freeman, E. M. (2018). Homelessness Among African American Families. In The Black
Family (pp. 67-82). Routledge.
Fritsch, A., Hiler, H., Mueller, B., Wu, M., & Wustmann, J. (2017). The Vulnerability of
Assessments, A Qualitative Analysis of Housing Professionals' Experiences with the VI-
SPDAT in Minnesota and a Comparative Review of Alternative Housing Triage
Assessments.
Fusaro, V. A., Levy, H. G., & Shaefer, H. L. (2018). Racial and Ethnic Disparities in the
Lifetime Prevalence of Homelessness in the United States. Demography, 55(6), 2119-
86
2128.
Galea, S., & Vlahov, D. (2002). Social determinants and the health of drug users: socioeconomic
status, homelessness, and incarceration. Public health reports, 117(Suppl 1), S135.
Garland, A. F., Lau, A. S., Yeh, M., McCabe, K. M., Hough, R. L., & Landsverk, J. A. (2005).
Racial and ethnic differences in utilization of mental health services among high-risk
youths. American Journal of Psychiatry, 162(7), 1336-1343.
Geber, G. M. (1997). Barriers to health care for street youth. Journal of Adolescent Health,
21(5), 287-290. https://doi.org/10.1016/S1054-139X(97)00111-0.
Greene, S., Turner, M. A., & Gourevitch, R. (2017). Racial residential segregation and
neighborhood disparities. Washington, DC: US Partnership on Mobility from Poverty.
Gotham, K. F. (2000). Racialization and the state: The Housing Act of 1934 and the creation of
the Federal Housing Administration. Sociological Perspectives, 43(2), 291-317.
Gordon, L. (2013, April). Using classification and regression trees (CART) in SAS® enterprise
miner TM for applications in public health. In SAS Global Forum (Vol. 2013, p. 2013).
Gubits, D., Shinn, M., Bell, S., Wood, M., Dastrup, S. R., Solari, C., ... & McInnis, D. (2015).
Family options study: Short-term impacts of housing and services interventions for
homeless families. US Department of Housing and Urban Development, Office of Policy
Development and Research.
Henry M, Cortes A, Shivji A, Buck K; Office of Community Planning and Development, U.S.
Department of Housing and Urban Development. The 2014 Annual Homeless
Assessment Report (AHAR) to Congress: Part 1: point-in-time estimates of
homelessness. Published October 2014. Available at: www. hudexchange. info/
resources/ documents/ 2014- AHARPart1. pdf. Accessed March 18, 2016.
Henry M, Watt R, Rosenthal L, et al. The 2017 Annual Homeless Assessment Report
87
(AHAR) to Congress: Part 1: Point-in-Time Estimates of Homelessness.
https://files.hudexchange.info/resources/documents/2017-AHAR-Part-1.pdf: United
States Department of Housing and Urban Development: Office of Community
Planning and Development; 2017.
Henry, M., Mahathey, A., Morrill, T., Robinson, A., Shivji, A., & Watt, R. (2019). Abt
Associates; US Department of Housing and Urban Development Office of Community
Planning and Development. In The 2019 annual homeless assessment report (AHAR) to
Congress. Available at: https://files.hudexchange.info/resources/documents/2019-
AHAR-Part-1.pdf. Accessed February (Vol. 6).
Hill, C. (2020). Exploring Predictors of Service Use among Youth Experiencing
Homelessness. Manuscript submitted for publication.
Howard, T. C., & Navarro, O. (2016). Critical race theory 20 years later: Where do we go from
here?. Urban Education, 51(3), 253-273.
Hsu, H. T., Rice, E., Wilson, J., Semborski, S., Vayanos, P., & Morton, M. (2019).
Understanding Wait Times in Rapid Re-Housing Among Homeless Youth: A Competing
Risk Survival Analysis. The journal of primary prevention, 40(5), 529-544.
Johnson, R. A. (2010). African Americans and homelessness: moving through history. Journal of
Black Studies, 40(4), 583-605.
Johnson KD, Whitbeck LB, Hoyt DR. Predictors of social network composition among homeless
and runaway adolescents. Journal of Adolescence. 2005;28(2):231–248. doi:
10.1016/j.adolescence.2005.02.005.
Jones‐Correa, Michael. The Origins and Diffusion of Racial Restrictive Covenants. Political
Science Quarterly. 2000;115(4):541-568. doi:10.2307/2657609
Jones, M. M. (2016). Does race matter in addressing homelessness? A review of the
literature. World medical & health policy, 8(2), 139-156.
Karas, D. (2015). Highway to inequity: The disparate impact of the interstate highway system on
88
poor and minority communities in American cities. New Visions for Public Affairs, 7, 9-
21.
Keuroghlian, A. S., Shtasel, D., & Bassuk, E. L. (2014). Out on the street: a public health and
policy agenda for lesbian, gay, bisexual, and transgender youth who are
homeless. American Journal of Orthopsychiatry, 84(1), 66.
Kimble, J. (2007). Insuring inequality: The role of the Federal Housing Administration in the
urban ghettoization of African Americans. Law & Social Inquiry, 32(2), 399-434.
King, B. T. (2018). Assessment and Findings of the Vulnerability Index (VI-SPDAT): Survey of
Individuals Experiencing Homelessness in Travis County, Tx.
Kipke, M. D., Montgomery, S., & MacKenzie, R. G. (1993). Substance use among youth seen at
a community-based health clinic. Journal of Adolescent Health, 14(4), 289-294.
Kort-Butler, L. A., & Tyler, K. A. (2012). A cluster analysis of service utilization and
incarceration among homeless youth. Social science research, 41(3), 612-623.
Lindsey, M. A., Joe, S., & Nebbitt, V. (2010). Family Matters: The Role of Mental Health
Stigma and Social Support on Depressive Symptoms and Subsequent Help Seeking
Among African American Boys. Journal of Black Psychology, 36(4), 458–482.
https://doi.org/10.1177/0095798409355796.
Logan, T. D., & Parman, J. M. (2017). Segregation and Homeownership in the Early Twentieth
Century. American Economic Review, 107(5), 410-14.
Lynd, S. (1966). The Compromise of 1787. Political Science Quarterly, 81(2), 225-250.
Johnson, H. B. (1998). Black Wall Street: From riot to renaissance in Tulsa's historic
Greenwood district. Marion Koogler McNay Art Museum.
Mallett, S., Rosenthal, D., & Keys, D. (2005). Young people, drug use and family conflict:
Pathways into homelessness. Journal of adolescence, 28(2), 185-199.
Mallett, S., Rosenthal, D. A., Keys, D., & Averill, R. (2009). Moving out, moving on: young
people's pathways in and through homelessness. Routledge.
Mauer, M. (2009). The changing racial dynamics of the war on drugs. Washington, DC:
Sentencing Project.
89
McKenna, C. (2008). The homeownership gap: How the post-World War II GI bill shaped
modern day homeownership patterns for black and white Americans (Doctoral
dissertation, Massachusetts Institute of Technology).
Mitchell, B. S. R. A. A. J. F. (2018, December 18). HOLC “redlining” maps: The persistent
structure of segregation and economic inequality » NCRC. https://ncrc.org/holc/
Messer, C. M., Shriver, T. E., & Adams, A. E. (2018). The Destruction of Black Wall Street:
Tulsa's 1921 Riot and the Eradication of Accumulated Wealth. American Journal of
Economics and Sociology, 77(3-4), 789-819.
Mettler, S. (2005). “The only good thing was the GI Bill”: Effects of the education and training
provisions on African-American veterans' political participation. Studies in American
Political Development, 19(1), 31-52.
Metraux, S., Roman, C. G., & Cho, R. S. (2007, September). Incarceration and homelessness.
In National symposium on homelessness research. US Department of Housing and Urban
Development Washington, DC.
Milburn, N. G., Ayala, G., Rice, E., Batterham, P., & Rotheram-Borus, M. J. (2006).
Discrimination and exiting homelessness among homeless adolescents. Cultural
Diversity and Ethnic Minority Psychology, 12(4), 658.
Milburn, N. G., Rice, E., Rotheram‐Borus, M. J., Mallett, S., Rosenthal, D., Batterham, P., ... &
Duan, N. (2009). Adolescents exiting homelessness over two years: The risk
amplification and abatement model. Journal of research on adolescence, 19(4), 762-785.
Milburn, N. G., Iribarren, F. J., Rice, E., Lightfoot, M., Solorio, R., Rotheram-Borus, M. J., ... &
Eastmen, K. (2012). A family intervention to reduce sexual risk behavior, substance use,
and delinquency among newly homeless youth. Journal of Adolescent Health, 50(4), 358-
364.
Miller, E. J. (2005). Foundering Democracy: Felony Disenfranchisement in the American
Tradition of Voter Exclusion. Nat'l Black LJ, 19, 32.
90
Mohl, R. A. (2004). Stop the road: Freeway revolts in American cities. Journal of Urban History,
30(5), pp. 674- 706.
Mohl, R. A. (2014). Citizen activism and freeway revolts in Memphis and Nashville: The road to
litigation. Journal of Urban History, 40(5), pp. 870-893.
Mojtabai, Ramin. "Perceived reasons for loss of housing and continued homelessness among
homeless persons with mental illness." Psychiatric Services 56.2 (2005): 172-178.
Michela Zonta, “Racial Disparities in Home Appreciation: Implications of the Racially
Segmented Housing Market for African Americans’ Equity Building and the
Enforcement of Fair Housing Policies” (Washington: Center for American Progress,
2019.
Morton, M. H., Dworsky, A., & Samuels, G. M. (2017). Missed opportunities: Youth
homelessness in America. National estimates. Chicago, IL: Chapin Hall at the
University of Chicago.http://voicesofyouthcountorg/wpcontent/uploads/2017/11/VoYC-
National-Estimates-Brief-Chapin-Hall-2017. pdf.
Morton, M. H., Dworsky, A., Matjasko, J. L., Curry, S. R., Schlueter, D., Chávez, R., & Farrell,
A. F. (2018). Prevalence and correlates of youth homelessness in the United States. Journal
of Adolescent Health, 62(1), 14-21. https://doi.org/10.1016/j.jadohealth.2017.10.006
Motley, R., & Banks, A. (2018). Black males, trauma, and mental health service use: A
systematic review. Perspectives on social work: the journal of the doctoral students of
the University of Houston Graduate School of Social Work, 14(1), 4.
NAACP, “Criminal Justice Fact Sheet,” available at http://www.naacp.org/criminal-justice-fact-
sheet.
Nunn, K. B. (2002). Race, crime and the pool of surplus criminality: or why the war on drugs
was a war on blacks. J. Gender Race & Just., 6, 381.
Olivet, J., Dones, M., & Richard, M. (2019). The intersection of homelessness, racism, and
mental illness. In Racism and psychiatry (pp. 55-69). Humana Press, Cham.
91
Padgett, D. K., Henwood, B., Abrams, C., & Drake, R. E. (2008). Social relationships among
persons who have experienced serious mental illness, substance abuse, and
homelessness: Implications for recovery. American journal of orthopsychiatry, 78(3),
333-339.
Padgett, D., Henwood, B. F., & Tsemberis, S. J. (2016). Housing First: Ending homelessness,
transforming systems, and changing lives. Oxford University Press, USA.
Pergamit, M. R., Ernst, M., & Hall, C. (2010). Runaway youth’s knowledge and access of
services. National Runaway Switchboard, Chicago IL. Available at http://www.
1800runaway. org/media/documents/NORC_Final_Report_4_22_10. pdf.
Pietila, A. (2012). Not in my neighborhood: How bigotry shaped a great American city. Rowman
& Littlefield.
Provine, D. (n.d.). Race and Inequality in the War on Drugs. Annual Review of Law and Social
Science, 7(1), 41–60. https://doi.org/10.1146/annurev-lawsocsci-102510-105445
Reeg, B. (2003). The Runaway and Homeless Youth Act and disconnected youth. In Leave no
youth behind: opportunities for congress to reach disconnected youth. Center for Law
and Social Policy, Washington, DC.
Rice, E., Holguin, M., Hsu, H. T., Morton, M., Vayanos, P., Tambe, M., & Chan, H. (2018).
Linking Homelessness Vulnerability Assessments to Housing Placements and Outcomes
for Youth. Cityscape, 20(3), 69-86.
Rheingold, Ira, Fitzpatrick, Michael, Hofeld, Al, Jr. From redlining to reverse redlining: a history
of obstacles for minority homeownership in America. (New Legal Problems Low-Income
Clients Face When They Work). Clearinghouse Review. 2001;34(9 10).
Roberts, D. E. (2011). Prison, foster care, and the systemic punishment of black
mothers. Ucla L. Rev., 59, 1474.
Roberts, D. (2009). Shattered bonds: The color of child welfare. Civitas Books.
Rodriguez, J. M., & Eidelman, T. A. (2017). Homelessness interventions in Georgia: Rapid re-
housing, transitional housing, and the likelihood of returning to shelter. Housing Policy
92
Debate, 27(6), 825-842.
Rose, C. M. (2016). Racially Restrictive Covenants—Were They Dignity Takings?. Law &
Social Inquiry, 41(4), 939-955.
Rosenthal, D., Rotheram-Borus, M. J., Batterham, P., Mallett, S., Rice, E., & Milburn, N. G.
(2007). Housing stability over two years and HIV risk among newly homeless
youth. AIDS and Behavior, 11(6), 831-841.
Rothstein, R. (2017). The color of law: A forgotten history of how our government segregated
America. Liveright Publishing.
Sanchez, Thomas W., Stolz, Rich, Ma, Jacinta S. Moving to Equity: Addressing Inequitable
Effects of Transportation Policies on Minorities. June 2003.
http://www.escholarship.org/uc/item/5qc7w8qp.
Shelton, K. H., Taylor, P. J., Bonner, A., & van den Bree, M. (2009). Risk factors for
homelessness: Evidence from a population-based study. Psychiatric Services, 60(4), 465-
472.
Shelton, J., & Winkelstein, J. (2014). Librarians and social workers: Working together for
homeless LGBTQ youth. Young Adult Library Services, 13(1), 20.
Sledjeski, E. M., Dierker, L. C., Brigham, R., & Breslin, E. (2008). The use of risk assessment to
predict recurrent maltreatment: A classification and regression tree analysis
(CART). Prevention science, 9(1), 28-37.
Somerville, P., & Steele, A. (2002). 'Race', Housing and Social Exclusion. Jessica Kingsley
Publishers.
Sweat, J., Nyamathi, A., Christiani, A., & Mutere, M. (2008). Risk behaviors and health care
utilization among homeless youth: Contextual and racial comparisons. Journal of
HIV/AIDS prevention in children & youth, 9(2), 158-174.
Taylor, J. (2014). Housing assistance for households experiencing homelessness. In National
Alliance to End Homelessness National Conference on Ending Family and Youth
Homelessness. New Orleans, LA.
93
Tenner, A. D., Trevithick, L. A., Wagner, V., & Burch, R. (1998). Seattle YouthCare’s
prevention, intervention, and education program: A model of care for HIV-positive,
homeless, and at-risk youth. Journal of Adolescent Health, 23(2), 96-106.
Tevendale, H. D., Comulada, W. S., & Lightfoot, M. A. (2011). Finding shelter: Two-year
housing trajectories among homeless youth. Journal of Adolescent Health, 49(6), 615-
620.
Toro, P. A., Lesperance, T. M., & Braciszewski, J. M. (2011). The heterogeneity of homeless
youth in America: Examining typologies. National Alliance to End Homelessness:
Washington, DC.
Tyler, K. A., Akinyemi, S. L., & Kort-Butler, L. A. (2012). Correlates of service utilization
among homeless youth. Children and Youth Services Review, 34(7), 1344-1350.
Villagran, G. J. The American Empire: Grotesque Inequity is Corrosive to Democracy.
Wagner, P., & Sawyer, W. (2018). Mass incarceration: The whole pie. The Prison Policy
Initiative.
Wenzel, S., Holloway, I., Golinelli, D., Ewing, B., Bowman, R., & Tucker, J. (2012). Social
networks of homeless youth in emerging adulthood. Journal of Youth and
Adolescence, 41(5), 561-571.
Whitbeck, L. B., Hoyt, D. R., & Bao, W. N. (2000). Depressive symptoms and co‐occurring
depressive symptoms, substance abuse, and conduct problems among runaway and
homeless adolescents. Child development, 71(3), 721-732.
Whitbeck, L. B., Chen, X., Hoyt, D. R., Tyler, K. A., & Johnson, K. D. (2004). Mental
disorder, subsistence strategies, and victimization among gay, lesbian, and bisexual
homeless and runaway adolescents. Journal of sex research, 41(4), 329-342.
Wilkerson, I. (2011). The warmth of other suns: The epic story of America's great migration.
Vintage.
Wilkey, C., Donegan, R., Yampolskaya, S., & Cannon, R. (2019). Coordinated Entry Systems:
Racial Equity Analysis of Assessment Data | The Homeless Hub. Retrieved July 9, 2020,
94
from https://www.homelesshub.ca/resource/coordinated-entry-systems-racial-equity-
analysis-assessment-data
Woods, L. L. (2012). The Federal Home Loan Bank Board, redlining, and the national
proliferation of racial lending discrimination, 1921–1950. Journal of Urban
History, 38(6), 1036-1059.
Woods, L. L. (2018). “The Inevitable Products of Racial Segregation”: Multigenerational
Consequences of Exclusionary Housing Policies on African Americans, 1910–
1960. American Journal of Economics and Sociology, 77(3-4), 967-1012.
95
Appendices
Appendix A: NST Tool
96
97
98
99
100
101
102
103
Abstract (if available)
Abstract
To address rising levels of youth homelessness, communities around the country have implemented continuums of care (COCs) to link youth to housing resources (Rice et al., 2018). Recent work has found that formal exits allocated through COCs account for more than 30% of youth exits from homelessness, with 25% of those exits attributed to rapid re-housing (RRH) programs and another 5.8% attributed to permanent supportive housing (PSH) interventions (Rice et al., 2018). Independent of the continuum of care models, youth are also exiting homelessness by returning to family. Studies of youth homelessness suggest that familial relationships are positively associated with increased likelihood of exiting homelessness as well as maintaining stable housing over time among homeless youth (Milburn et al., 2009). Although youth may exit homelessness in various ways, exits via formal housing programs or family are the most visible. Therefore, the predictors of utilizing these exits and the subsequent stability of these exits need intentional focus. ❧ Furthermore, while homelessness is not exclusively Black, the factors that result in initial housing loss and subsequent housing stability may be unique to those who identify as Black. A central argument of this dissertation is that although several racial and ethnic communities experience homelessness, the systemic inequities that drive homelessness impact Black people differently and, in some cases, to a greater degree. This exploration leads to what I call the “algorithm of Black homelessness,” which refers to a finite set of policies that, when attached to identity markers such as race, systemically hinder Black people from remaining stably housed. The findings from this dissertation will elucidate important policy implications and encourage the development of assessment tools that capture the unique needs of Black youth experiencing homelessness.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Social network engagement and HIV risk among homeless former foster youth
PDF
Homeless youth: Reaching the Hard-To-Reach
PDF
Youth homelessness policy change
PDF
Transitional housing and wellness center: a holistic approach to decreasing homelessness and mental illness in the Black community
PDF
Couch-surfing among youth experiencing homelessness: an examination of HIV risk
PDF
Ending homelessness: evolution of the Qad Prep Academy
PDF
We are our neighbors' keeper: an innovative field kit of outreach and assessment tools to help end homelessness
PDF
Discrimination at the margins: perceived discrimination and the role of social support in mental health service use for youth experiencing homelessness
PDF
Social network norms and HIV risk behaviors among homeless youth in Los Angeles, California
PDF
WISER women’s program: well-being innovation with support and education for resilience—a homelessness prevention intervention
PDF
Increasing access to mental health counseling for homeless youth: Peer2Peer Counseling Supports
PDF
The 500 Day Initiative: an innovative standard operating procedure assisting Detroit Housing Commission voucher recipients with 24 hour case management services
PDF
Mitigate microaggressions against Black women in the workplace for improved health outcomes
PDF
Addison’s Neighbor: permanent supportive housing for parenting youth transitioning out of foster care
PDF
Understanding emotional regulation and mood of young adults in the context of homelessness using geographic ecological momentary assessment
PDF
Homeless female veterans—silent epidemic
PDF
Ensuring healthy development for all youth: homelessness is trauma connecting with the invisible families: fostering the parent-child bond
PDF
Conjoint homeless prevention services for older adults
PDF
Rethink Homelessness project
PDF
Unto the least of these homeless ministry: ending homelessness within the co-occurring population
Asset Metadata
Creator
Hill, Chyna Yvonne Vandon
(author)
Core Title
The algorithm of Black homelessness: a classification tree analysis
School
Suzanne Dworak-Peck School of Social Work
Degree
Master of Social Work / Doctor of Philosophy
Degree Program
Social Work
Publication Date
07/28/2020
Defense Date
06/23/2020
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
algorithm of Black homelessness,algorithmic redlining,Black homeless youth,Black youth homelessness,classification tree analysis,housing outcomes,housing stability,OAI-PMH Harvest,youth experiencing homelessness
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Rice, Eric (
committee chair
), Kim, Elizabeth (
committee member
), Richardson, Allissa (
committee member
)
Creator Email
chynahil@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-348391
Unique identifier
UC11663516
Identifier
etd-HillChynaY-8804.pdf (filename),usctheses-c89-348391 (legacy record id)
Legacy Identifier
etd-HillChynaY-8804.pdf
Dmrecord
348391
Document Type
Dissertation
Rights
Hill, Chyna Yvonne Vandon
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
algorithm of Black homelessness
algorithmic redlining
Black homeless youth
Black youth homelessness
classification tree analysis
housing outcomes
housing stability
youth experiencing homelessness