NEW PERSPECTIVES ON COMMUNITY-LEVEL

DETERMINANTS OF HOMELESSNESS
THOMAS BYRNE and ELLEN A. MUNLEY
U.S. Department of Veterans Affairs, National Center on
Homelessness Among Veterans; University of Pennsylvania

JAMISON D. FARGO
U.S. Department of Veterans Affairs, National Center on
Homelessness Among Veterans; Utah State University

ANN E. MONTGOMERY
and DENNIS P. CULHANE
U.S. Department of Veterans Affairs, National Center on
Homelessness Among Veterans; University of Pennsylvania

ABSTRACT: Understanding the root causes of homelessness is important for developing effec-
tive solutions to the problem. This fact has not gone unnoticed by researchers, who have made
numerous attempts to identify the underlying structural determinants of homelessness by modeling
inter-community variation in the rate of homelessness as a function of community-level variables.
Yet, prior studies in this area have a number of serious limitations, principally their reliance on
methodologically flawed estimates of the size of the homeless population. The present study ad-
dresses this and other limitations by using newly available and more reliable estimates from the
U.S. Department of Housing and Urban Development to model variation in the rate of homelessness
across a large and diverse sample of communities throughout the United States. In doing so, this
study builds on the analysis conducted by Lee, Price-Spratlen, and Kanan (2003), and its findings
have implications for policy and future research.

Since its emergence as a public problem (Stern, 1984) in the 1980s, homelessness has
become an entrenched phenomenon, posing a consistent challenge to policymakers, advocates,
and service providers alike. According to the most recent estimates, on any given night in the
United States, there are roughly 645,000 persons residing in homeless shelters or in unsheltered
street locations. Over the course of a year, approximately 1.6 million persons, or about 1 in
every 195 Americansand 1 in every 25 persons living below the poverty thresholdexperience
homelessness (U.S. Department of Housing and Urban Development, 2011). The sheer scope
of the problem as well as the myriad negative health, economic, and social outcomes linked to

Direct correspondence to: T. Byrne, National Center on Homelessness Among Veterans, 4100 Chester Avenue, Suite 201,
Philadelphia, PA 19104. E-mail: [email protected]

JOURNAL OF URBAN AFFAIRS, Volume 35, Number 5, pages 607625.

Copyright  C 2012 Urban Affairs Association

All rights of reproduction in any form reserved.

ISSN: 0735-2166. DOI: 10.1111/j.1467-9906.2012.00643.x
608 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

homelessness (Burt, 2001; Hawkins & Abrams, 2007; Hwang, 2001; Lee & Farrell, 2003; Lee &
Greif, 2008; Wolitski, Kidder, & Fenton, 2007; Zerger, 2002) underscore the need to develop and
implement effective policies to prevent and end homelessness.
The emergence and persistence of contemporary homelessness has been accompanied by a
proliferation of studies that seek to identify its causal mechanisms. One approach has been to
identify explanations for the geographic variation in the prevalence of homelessness. Researchers
have identified conditions that co-vary with levels of homelessness, and described these conditions
as possible causal factors of homelessness at the neighborhood, city, metropolitan, and state
levels.
The present study adds to the existing body of literature consisting of cross-sectional studies of
geographic variation in homelessness rates by (a) using more recent and reliable estimates of the
homeless population; (b) conducting the first analysis to date of community-level determinants
of homelessness that includes both metropolitan and non-metropolitan communities; and (c)
increasing the comparability of findings across studies.
Our specific objective is to replicate and extend the community-level model of homelessness
examined by Lee, Price-Spratlen, and Kanan (2003). Our goal is to assess whether using more
recent data and including a broader set of communities yields different, and potentially more
robust, findings than Lee et al. (2003). Our findings have potential implications for the use of
these data and sample of communities in future research and might shed new light on policy
changes that could reduce rates of homelessness.

Review of Previous Studies

Research on geographic variation in homelessness emerged in the 1990s in response to the
predominance of individual-level studies in the existing homeless literature. By their nature,
individual-level studies focused on characteristics and conditions of individuals and households,
and were based on theoretical models that conceptualized homelessness as a result of individual-
level factors as varied as adverse childhood experiences, disability, mental illness, substance abuse
disorders, lack of social or human capital, a history of institutional involvement, and exogenous
health and income shocks (Bassuk, Rubin, & Lauriat, 1984; Calsyn & Roades, 1994; Curtis,
Corman, Noonan, & Reichman, 2011; Hopper, Jost, Hay, Welber, & Haugland, 1997; Jones,
1983; Koegel, Melamid, & Burnam, 1995; Metraux, Roman, & Cho, 2008; OFlaherty, 2009;
Shinn et al., 1998; Susser, Lin, & Conover, 1991).
In contrast, structural models, which provide the basis for many community-level studies,
portray macro-level trends such as decreases in the availability of affordable housing, labor
market conditions, cutbacks in safety net spending, prevalence of disabilities, and demographic
factors as the primary drivers of homelessness (Burt, 1991; Koegel, Burnam, & Baumohl, 1996;
Main, 1996; McChesney, 1990; OFlaherty, 1995; Rossi, 1989; Wright & Lam, 1987). Over
time, research on determinants of homelessness has moved toward a general consensus that
individual and structural explanations are not mutually exclusive, and theoretical models have
been developed that integrate the two types of factors (Culhane Lee, & Wachter, 1996; OFlaherty,
2004). However, as a study of homelessness rates at the community level, this paper focuses on
structural community-level determinants of homelessness.
Much of the previous research on geographic variation in homelessness shares a similar
methodological approach: most studies model the rate of homelessness in a city (or metropolitan
area, county, or state) as a function of a set of city-level (or metropolitan area, county, or state)
factors. However, studies have tested a range of community-level factors as potentially important
determinants of homelessness, and likewise, the specific measure of the rate of homelessness that
 I New Perspectives on Community-Level Homelessness I 609

has been used as the dependent variable has varied across studies. Table 1 provides a summary
of prior studies, including their measures of homelessness and the community-level factors that
they included as predictors of homelessness.

Community-Level Determinants of Homelessness

The set of community-level factors hypothesized to be associated with homelessness varies
across studies, making it a challenge to synthesize this body of research. Some consistencies
have emerged, but for many factors, their inclusion in studies has been too inconsistent to allow
conclusions to be drawn. Lee et al. (2003) categorized the community-level factors that predict
rates of homelessness into the following broad domains: housing market, economic conditions,
demographic composition, safety net, climate, and transience. Each domain is described in more
detail in the following sections.

Housing Market Factors

At least one housing market factor, including rent levels and rental vacancy rates, regulation on
housing construction, or presence of rent control, was associated with homelessness rates in each
of the reviewed studies. Many researchers have placed primary focus on the shortfall of available
affordable housing, resulting from a mismatch among housing cost, housing availability, and
household income (Burt, 1991; Koegel, Burnam, & Baumohl, 1996; Main, 1996; McChesney,
1990; OFlaherty, 1995; Rossi, 1989; Wright & Lam, 1987).
A number of prior studies have consistently identified significant relationships between in-
creased rent levels, decreased vacancy rates, and increased homelessness. The proportion of
renter households in a community has been positively associated with homelessness, while va-
cancy rate has been negatively associated with homelessness. The most consistent housing market
finding has been a significant positive relationship between increased rent level and homelessness
(Bohanon, 1991; Early & Olsen, 2002; Honig & Filer, 1993; Lee et al., 2003; Quigley, 1990;
Quigley & Raphael, 2002; Quigley, Raphael, & Smolensky, 2001; Troutman, Jackson, & Ekelund,
1999).

Economic Conditions
Local economic conditions also play an important role in determining the level of housing
affordability in an area. Individual-level studies have established that homelessness generally
occurs among people who are very poor (Burt & Cohen, 1989), suggesting that poverty rates
as well as additional factors that impact a households income should be included in predictive
models of homelessness rates. Prior studies on the community-level determinants of homelessness
have most frequently used poverty and unemployment rates as proxies for economic conditions,
and both have been found in numerous studies to be positively associated with the rate of
homelessness (Appelbaum, Dolny, Dreier, & Gilderbloom, 1991; Bohanon, 1991; Burt, 1993;
Early & Olsen, 2002; Quigley, 1990; Quigley & Raphael, 2002; Quigley et al., 2001; Troutman
et al., 1999).

Demographic Composition
While there is compelling evidence from other areas of literature on homelessness that de-
mographic characteristics affect risk of homelessness, studies of geographic variation in rates
TABLE 1

Summary Matrix of Previous Studies of Community-Level Determinants of Homelessness

Significant independent variables

Dependent Housing Economic Demographic Transience

Study variable market conditions composition Safety net Climate /other
Appelbaum et al. Homelessness per 1,000 % Renter households Unemployment Temperature
(1991) people1 rate
Bohanon (1991) % Homeless in poverty Rent level Unemployment Household sizeSize of institutionalized
population1 rate mental health
population
Burt (1993) Shelter beds per capita2 Ratio: Low-income Employment % Single-person General Assistance
renters to units sector households benefit level
Unemployment General Assistance
rate eligibility
Early & Olsen Sheltered, unsheltered, all Rent level Extreme poverty Shelter quality Temperature
(2002) homeless per 10,000 rate
people in poverty3
% Sheltered homeless in Subsidized housing
homeless population3 targeted toward
extremely poor
Elliott & Krivo % Homeless in total Low-rent units Unskilled jobs % Black Mental health
610 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

(1991) population1 expenditures

% Female-
headed
households
% Hispanics
Grimes & % Sheltered homeless in Rent control % Group Medicaid expenditures Temperature Region of
Chressanthis population3 quarters USA
(1997) residents
% Unsheltered homeless Rent gap between Population size Violent crime
in population3 low-cost/median-cost rate
units
% Total homeless in
population3
Honig & Filer Homelessness per Rent level for low-cost Employment % Black AFDC benefit level
(1993) 100,000 people1 units growth
SSI benefit level

(Continued)
TABLE 1

Continued
Significant independent variables

Dependent Housing Economic Demographic Safety Transience

Study Variable market conditions composition net Climate /other
Hudson (1998) Homelessness per 10,000 Service sector Population McKinney expenditures 1990 Census
people3 employment density S-Night
search
effort
Lee et al. (2003) Homelessness per Rent level % Single-person Precipitation Residential
10,0003 households mobility
Quigley (1990) Homelessness per 1,000 Rent level Poverty rate Population Temperature
people1 growth
Rental vacancy rate
Quigley & Multiple rates of Rent level Income level Size of institutionalized Temperature
Raphael homelessness among mental health
(2002), individuals and population
Quigley et al. families2,3
(2001)
Rental vacancy rate Poverty rate SSI benefit receipt rate
Raphael (2010) % Homeless in poverty Housing market Temperature
population4 regulation
Rent-to-income ratio
Troutman % Homeless in poverty Rent level for low-cost Poverty rate Alcohol, drug, mental
et al.(1999) population1,3 units health expenditures
Rent control Unemployment Federal housing
rate assistance
expenditures
McKinney expenditures
Rental vacancy rate for
low-cost units
Tucker (1987, Homelessness per 1,000 Rent control Population Temperature
1989) people1 growth

Notes: Sources for dependent variables are indicated as follows:

1
1984 HUD estimates.
2
Burt 1989 shelter estimate.
3
1990 Census S-Night Enumeration.
4
AHAR PIT estimate.
I New Perspectives on Community-Level Homelessness I 611
612 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

of homelessness have found inconsistent results for the relationship between variation in demo-
graphic composition of the population and variation in the rate of homelessness.
Cross-sectional studies have shown demographic characteristics, including race, to affect risk
of homelessness among individuals. For example, African Americans have been consistently
shown to be overrepresented in the homeless population (Burt, 2001; Culhane & Metraux, 1999;
U.S. Department of Housing and Urban Development [HUD] 2011); therefore, areas with higher
proportions of African Americans may have higher rates of homelessness. In addition, evidence
from a study of intra-city homelessness variation suggests that the proportion of female-headed
households with young children is positively associated with homelessness (Culhane et al., 1996).
Similarly, the concentration of single-person households in a community may be an important
determinant of homelessness, both because single-person households cannot rely on the support
of a second wage earner, thereby placing them at an increased risk of homelessness, and because
communities with more single-person households may also face increased competition for low-
cost rental units. Finally, the age distribution of the population in a particular community may
affect the size of the homeless population, especially in light of evidence that members of the
latter half of the baby-boomer age cohortborn between 1946 and 1964make up a highly
disproportionate share of the single adult homeless population (Culhane, Metraux, Byrne, Stino,
& Bainbridge, in press).
Studies at the community level, in terms of the geographic variation in homelessness rates,
have found also significant relationships between increased rates of homelessness and increased
proportions of single-person and female-headed households (Burt, 1993; Elliott & Krivo, 1991;
Lee et al., 2003), African American households (Elliott & Krivo, 1991; Honig & Filer, 1993),
and Hispanic households (Elliott & Krivo, 1991). However, specific demographic measures have
not been consistently shown to be significant across studies, and no study has found evidence of
a relationship between a communitys age distribution and its rate of homelessness.

Safety Net
Studies of homeless individuals have provided evidence that the size of the social safety net and
the extent to which social safety net programs provide an adequate level of assistance can impact
the chances that households will experience homelessness (Burt, 1991; Koegel et al., 1996; Rossi,
1989). Under this line of reasoning, more generous and more widely available safety net programs
help protect low-income, disabled, and other vulnerable households from becoming homeless.
For example, the degree to which income support programs like Temporary Assistance for Needy
Families (TANF), Supplemental Security Income (SSI), or General Assistance (GA), are able to
provide larger numbers of vulnerable households with resources adequate enough to help them
obtain or maintain housing may impact the rate of homelessness. Similarly, increased spending
on treatment programs and other services for persons with physical or mental disabilities, as
well as for those with substance abuse disorders, who are all overrepresented in the homeless
population, can provide households with supports that may keep them from becoming homeless.
However, because there are a variety of programs and ways to measure assistance, few studies
have consistently included a single measure for a single program. Despite this lack of consistent
results, studies have generally found that more extensive and more generous social safety net
programs have a negative relationship with homelessness (Burt, 1993; Elliott & Krivo, 1991;
Honig & Filer, 1993; Quigley & Raphael, 2002; Quigley et al., 2001; Troutman et al., 1999).

Climate
Some studies attempting to explain geographic variation in homelessness have also added
climate measures to models, hypothesizing that less precipitation or higher temperatures could
 I New Perspectives on Community-Level Homelessness I 613

contribute to higher measured rates of homelessness by making homeless people more visible,
less vulnerable to mortality, or more likely to gravitate to a region. Among these studies, most
have found climate to have a significant relationship with rates of homelessness, and in the
expected direction, with higher temperatures and less precipitation associated with higher rates
of homelessness, and higher proportions of persons experiencing homelessness in unsheltered
locations (Appelbaum et al., 1991; Grimes & Chressanthis, 1997; Lee et al., 2003; Quigley, 1990;
Quigley & Raphael, 2002; Quigley et al., 2001; Raphael, 2010; Tucker, 1987, 1989).

Transience
Lee et al. (2003) also examined transience as a possible determinant of homelessness, including
the proportion of persons who have recently moved and the number of highways and railroads
serving an area. The authors argue that areas with high residential mobility have more competitive
housing and labor markets, which may increase the vulnerability to homelessness of those less
well-suited to compete in these arenas. In addition, the extent to which an area is a frequent
destination for tourists, migrant laborers, students, and others may have an impact on its rate of
homelessness as some of these persons may wind up stranded and become homeless. Lee et al.
find that the proportion of recently moved persons was a significant positive predictor, suggesting
the importance of the relationship between the degree of residential mobility in an area and its
rate of homelessness.

Homelessness Dependent Variables

The main challenge in executing studies that investigate determinants of the variation in
the rate of homelessness across jurisdictions is obtaining accurate estimates of the number of
persons experiencing homelessness in each location. Prior to 2005, few large-scale estimates
of homelessness were available, and as such, the studies summarized in Table 1with the
exception of Raphaels 2010 study, which used state-level rates of homelessness from the 2007
Annual Homeless Assessment Report to Congresshave relied on one of three estimates of
homelessness:

1. 1984 U.S. Department of Housing and Urban Development (HUD) estimate, based on a
survey of local experts in 60 metropolitan areas, who reported the size of the homeless
population in their area (HUD, 1984);
2. 1989 Burt survey of shelter capacity, covering 182 cities with populations greater than
100,000 (Burt, 1993); and
3. 1990 Census S-Night Enumeration, which was an effort by the Census Bureau to include
more homeless people in the decennial census by counting homeless persons at emergency
shelters and outdoor locations within municipalities with populations greater than 50,000
(U.S. Bureau of the Census, 1990).

Each of these efforts represented a methodological improvement over previously available

data, but each also has flaws that may affect the reliability of findings from prior studies that
have made use of these data. The primary methodological flaw with the 1984 HUD estimates
was its sole reliance on informant interviews, which may have resulted in an undercount of
the number of persons experiencing homelessness by applying estimates of homelessness in
downtown areas to entire metropolitan areas (Appelbaum et al., 1991). Alternatively, the 1989
Burt survey used shelter bed capacity as a proxy for homelessness. This estimate did not account
for unsheltered homelessness, turnover in shelter beds, and unused beds (Burt, 1993). The 1990
614 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

Census S-Night Enumeration improved on these two studies by relying on actual enumeration
of people experiencing homelessness. However, methodological flaws and inconsistency in how
enumeration protocols were implemented across communities may have resulted in an undercount
of the sheltered and unsheltered homeless population (Martin, 1992; U.S. General Accounting
Office, 1991).

Goals of the Current Study

The sources of data used in this study further improve on past enumeration efforts. The bi-
ennial point-in-time count that HUD requires of communities nationwide enumerates sheltered
homeless households using computerized systems that must meet a set of standard criteria to
ensure accuracy in reporting (HUD, 2011). Enumeration of unsheltered households also must
meet HUDs methodology standards (HUD, 2008). HUD has begun producing annual estimates
based on enumerations that occur in a large number of communities of diverse size and urban-
ization. HUDs ongoing and extensive technical support to communities has resulted in marked
improvement in the reliability of homeless counts over the past several years (HUD, 2011).
The current study will also expand the scope of geographic areas included in this type of
research. By necessity, many prior studies restricted their sample sizes to a limited number of cities
or metropolitan areas, due either to availability of estimates of the size of the homeless population,
or due to a mismatch between geographies at which the independent variables (predictive factors)
and dependent variables (rates of homelessness) were measured. This focus on larger urban
areas is problematic given evidence that homelessness is shifting from principal cities to rural
and suburban areas (HUD, 2011). The current study addresses this gap by conducting parallel
analyses for separate samples of metropolitan and non-metropolitan communities. In doing so, it
not only provides a more complete picture of structural determinants of homelessness, but also
contributes to the body of knowledge on rural homelessness, a topic that has been excluded from
most prior research.
The current study also aims to build on prior research by testing a set of potential structural
determinants of homelessness that closely mirror those used by Lee et al. (2003), a rigorously
designed study using homelessness data from the 1990 Census. The results from this study will
be directly comparable to those reported by Lee and colleagues, allowing progress to be made
towards more definitive identification of community-level factors that are the most important
determinants of homelessness.

METHODS

Data
Our dependent variables were estimates of the number of persons experiencing homelessness
in communities throughout the United States and our independent variables were characteristics
of these communities along the six domains described previously.

Dependent Variables (Homelessness)

This study uses the HUD point-in-time (PIT) counts of unsheltered and sheltered homeless
persons collected on a single night in January 2009 in 447 Continuums of Care (CoCs) throughout
the United States. CoCs are geographic units at which providers of homelessness assistance share
federal resources and work collaboratively to develop a strategic plan to address homelessness
 I New Perspectives on Community-Level Homelessness I 615

within their jurisdiction. CoCs vary in size and composition and can be comprised of single cities,
individual counties, several counties, or entire states. Regardless of their size and composition,
CoCs are geographically meaningful contexts for understanding the scope and determinants
of homelessness since they are themselves spatial manifestations of how efforts to address
homelessness are organized and administered. While 54 mainly rural counties are not part of a
CoC, more than 99% of the United States population lived within the boundaries of a CoC in
2009.
CoCs constitute the unit of analysis used in this study for assessing rates of homelessness. We
use the HUD PIT estimates to construct two measures of the rate of homelessness, which parallel
those used by Lee and colleagues: (1) the number of homeless adults per 10,000 adults in the
general population, and (2) the number of homeless adults per 10,000 adults in poverty. We use
homeless adults, which includes adults who are homeless as part of a family with children, rather
than the overall number of persons experiencing homelessness as our numerator in constructing
these rates to control for any potential variation in family size across CoCs. In addition, as persons
experiencing homelessness are nearly universally poor, the second measure approximates the rate
of homelessness among those who face the highest risk.

Independent Variables (CoC and State Characteristics)

To replicate the analysis conducted by Lee et al. (2003) as closely as possible, we collected
independent variables mirroring those from a number of sources. These variables, their sources,
and how they compare with the original variables used by Lee and colleagues are summarized in
Table 2. The housing market measures (rent level, homeownership rate, vacancy rate), indicators of
demographic composition (% Black,% Hispanic,% baby boomers,% single-person households),
and safety net measures (public assistance recipients, SSI recipients, TANF payment level, mental
health expenditures) were almost identical to those used by Lee and colleagues. While several
of the economic indicatorsextreme poverty and unemploymentmatch Lee and colleagues,
we were unable to include an indicator of low-wage jobs. Similarly, our transience measure was
limited to mobility rate and excludes a variable for transport access. Finally, it was not feasible
to include measures of climate given that CoCs, which form our unit of analysis, can be large
enough that there was significant within-CoC climate variation.

Constructing the Research Dataset

While the CoC was the unit of analysis for the homelessness dependent variables, CoCs
constitute geographies with irregular boundaries. Consequently, measures of CoC-level charac-
teristics (e.g., housing market and economic conditions) are virtually non-existent. As a result,
we constructed CoC-level independent variables from county-level measures using a two-step
process. First, geospatial matching procedures linked all counties with their appropriate CoC.
Second, county-level measures were statistically adjusted, where necessary, to transform them
into CoC-level variables. Given the irregular geographic composition of CoCs, there were three
relationship types possible between county and CoC boundaries:

1. Boundary for a single CoC and a single county was identical;

2. A single CoC may be comprised of an aggregation of two or more counties; and
3. Multiple CoCs may fall within a single county.

We used Geographic Information System (GIS) software to identify the appropriate CoC-
county relationship type for each CoC and, in turn, to match each CoC with its corresponding
TABLE 2

616 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

Summary of Study Variables and Comparison with Lee et al. (2003)
Variable Lee et al. (2003) Current Study Source
Homelessness Rate (1) Homeless persons per 10,000 residents (1) Homeless adults per 10,000 residents 2009 HUD PIT
Estimates1
(2) Homeless persons per 10,000 residents with income (2) Homeless adults per 10,000 adults in poverty
<50% of poverty
Housing Market
Rent level Median contract rent ($) Median rent (in $100) for efficiency unit 2009 HUD FMR2
Homeownership % Housing units occupied by owner % Housing units occupied by owner ACS3
Vacancy rate % All housing units vacant % Housing units vacant ACS
Economic Conditions
Unemployment rate % Civilian labor force unemployed % Civilian labor force unemployed ACS
Low-wage jobs % Employed persons 16+ in service/unskilled N/A
Extreme poverty % Persons with incomes <50% of poverty level % Persons with incomes <50% of poverty level ACS
Demographics
Blacks % Black in total population % Black in total population ACS
Hispanics % Hispanic in total population % Hispanic in total population ACS
Baby boomers % Persons 2544 in total population % Persons 4564 in total population ACS
1-person households % Occupied housing units with single occupant % Occupied housing units with single occupant ACS
Safety Net
PA recipients % Households receiving public assistance % Households in poverty receiving public assistance ACS
SS recipients % Persons receiving Social Security benefits % Households in poverty receiving SSI ACS
AFDC/TANF payment Mean monthly AFDC payment ($) per recipient, for state Maximum monthly TANF payment (in $100) for Urban Institute
family of 3, for state
MH expenditures Per capita mental health expenditures ($), for state Per capita expenditures (in $100) by state mental Kaiser Family
health agency, for state Foundation
Climate
Temperature July average maximum minus January average minimum N/A
temperature
Precipitation Average annual precipitation (inches) N/A
Transience
Mobility rate % Persons 5+ living in different houses than 5 years prior % Total population residing in U.S for 1+ year, who ACS
moved in past year
Transport access N of interstates, US routes, and railroads serving area N/A

Notes:
1
HUD PIT Estimate = U.S. Department of Housing and Urban Development (HUD) (PIT) estimate of persons experiencing homelessness.
2
HUD FMR = HUD Fair Market Rents database.
3
ACS = U.S. Census Bureau, 20052009 American Community Survey 5-year estimates.
 I New Perspectives on Community-Level Homelessness I 617

county or counties. To complete the matches, we superimposed county centroids (i.e., points
representing the geographic center of counties) on a map of CoC boundaries. Approximately
51% (N = 227) of CoCs matched directly to one county while 38% (N = 171) of CoCs were
comprised of multiple counties. The remaining 11% (N = 49) of CoCs fit the third type of
CoC-county relationship described earlier. In these cases, the multiple CoCs that were fully
encompassed by a single county were merged into a single new CoC, with its boundaries being
coterminous with the county.
After appropriately matching CoCs and counties, we statistically adjusted the 171 CoCs that
fit the second type of relationship described above and the 49 CoCs that fit the third type to
complete the construction of CoC-level variables from county measures (no adjustments were
necessary for the 227 CoCs that met the criteria for the first type of relationship). In the case
of the second type of relationship, we constructed CoC-level variables from county measures
by taking either the sum or a population-weighted average of the county measures from all of
the counties within a given CoC. In the third type of relationship, where multiple CoCs within
a single county were merged, we summed the HUD PIT estimates of persons experiencing
homelessness from all of the CoCs located within a single county. In turn, county-level measures
then became CoC-level variables for these combined CoCs, with no statistical adjustment required.
The merging of several CoCs in this fashion resulted in a reduction in the number of CoCs
from 447 to 414, which we then stratified into metropolitan (N = 338) or non-metropolitan area
(N = 76) groups using the U.S. Department of Agricultures (USDA) Economic Research Service
definitions of rurality (USDA, 2003).
Although the CoC was the basic unit of analysis for this study, several variables of interest
(TANF payments, mental health expenditures) were only available at the state level. Variables
measured at the state level were not adjusted and therefore all CoCs within a state assumed the
same value.

Statistical Analysis
Because CoCs were nested within states, data from CoCs located within the same state were not
considered to be independent from one another and are likely to have more similar characteristics
than they would share with CoCs in different states. This clustering violates the basic assumption
of independence in ordinary least squares (OLS) regression and such clustering must be accounted
for in the statistical analysis to correctly model the variation in relationships between independent
and dependent variables. Using a multilevel modeling approach rather than OLS regressions
allowed us address this problem: we used the CoC as the first level of analysis and the state as
the second level of analysis.
We conducted a series of linear mixed-effects models to understand which CoC variables were
significantly associated with each homelessness outcome (homeless adults per 10,000 adults in the
[a] general population and [b] population in poverty). OLS models were also estimated for each
outcome, and had similar results to, but lower explanatory power than the linear mixed-effects
models, and therefore are not reported here. In addition, we stratified the sample into metropolitan
and non-metropolitan CoCs and conducted analyses separately for each subgroup. We applied a
natural logarithmic transformation to each outcome variable due to their highly skewed natures.
We included the following predictor variables in models with homeless adults per 10,000 adults in
the general population: rent level; homeownership rate; housing vacancy rate; unemployment rate;
extreme poverty rate; size of the Black, Hispanic, and baby boomer populations; public assistance
recipients; SSI recipients; TANF benefit level; mental health expenditures; and mobility rate. With
the exception of the percent of the population in extreme poverty, the same predictor variables
618 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

were included in models where the population in poverty served as the denominator of the
outcome.
Random intercepts were specified in all models. Analyses were conducted using the R envi-
ronment for statistical computing (R Development Core Team, 2011). Due to difficulties inherent
in calculating accurate confidence intervals and p-values for mixed-effects models, we calculated
highest posterior density intervals and Monte Carlo derived p-values (Chen & Shao, 1999), which
are Bayesian analogues to these estimates.

RESULTS
Data were available for 414 CoCs (338 metropolitan, 76 non-metropolitan) and descrip-
tive statistics for all study variables are presented in Table 3, stratified by metropolitan/non-
metropolitan status.

Modeling Adult Homelessness per 10,000 Adults

Several variables for models based on metropolitan and non-metropolitan samples were signif-
icantly associated with the rate of adult homelessness per 10,000 adults in the general population
(see Table 4.) Although predictors in both metropolitan and non-metropolitan models accounted
for a large degree of the variation in CoC homelessness rates (R2 for metropolitan = 58%, non-
metropolitan = 67%), notable differences were observed between models in terms of the pattern
of significant predictors. For example, in the model for metropolitan CoCs, rent level, homeown-
ership rate, the size of the Hispanic and baby-boomer populations, the proportion of single-person
households, and the proportion of recently moved households were all positively associated with
the outcome. In the model for non-metropolitan CoCs, however, rent level and unemployment
rate were positively associated, and the size of the Black population was negatively associated
with the outcome. The effects for rent level and the unemployment rate were particularly strong
in the non-metropolitan CoC model, with a 32% increase in homelessness per $100 increase in
median rent and a 27% increase in homelessness per 1% increase in unemployment.

Modeling Adult Homelessness per 10,000 Adults in Poverty

Similarly, several variables for models based on metropolitan and non-metropolitan samples
were found to be significantly associated with the rate of adult homelessness per 10,000 adults in
the population in poverty (see Table 5). The pattern of results was somewhat different between
models based on metropolitan and non-metropolitan CoCs, although the degree of variation
in CoC homelessness rates accounted for by the independent variables was similar (R2 for
metropolitan = 52%, non-metropolitan = 56%). In the metropolitan model, rent level, the size
of the Hispanic and baby-boomer populations, the proportion of single-person households, rate
of public assistance receipt, and mobility rate were positively associated and state mental health
expenditures were negatively associated with the outcome. In contrast, in the non-metropolitan
model, rent level was positively associated and the size of the Black population negatively
associated with the rate of homelessness. The effect for rent level was particularly strong in both
metropolitan and non-metropolitan CoC models, with a 15% and 39% increase in homelessness
per $100 increase in median rent, respectively.
TABLE 3

Descriptive Statistics

Metro CoCs (N = 338) Non-Metro CoCs (N = 76)

Variable M Mdn SD Min Max M Mdn SD Min Max

Homelessness Rate
Homeless adults per 10,000 residents 20.60 15.92 18.77 0.91 137.78 19.55 9.00 30.86 0.00 157.50
Homeless adults per 10,000 adults in poverty 200.01 150.58 178.17 9.62 1198.74 159.13 72.33 244.51 0.00 1450.37
Housing Market
Rent level $ 699.17 640.21 193.51 384.83 1230.00 533.80 479.38 125.76 383.42 1006.00
Homeownership% 0.68 0.69 0.08 0.34 0.87 0.72 0.72 0.05 0.57 0.85
Vacancy% 0.11 0.10 0.05 0.03 0.54 0.17 0.15 0.08 0.07 0.43
Economic Conditions
Unemployment% 0.08 0.07 0.02 0.03 0.15 0.07 0.07 0.02 0.04 0.12
Extreme poverty% 0.06 0.05 0.02 0.01 0.21 0.06 0.06 0.02 0.03 0.13
Demographics
Blacks% 0.12 0.09 0.12 0.00 0.64 0.07 0.03 0.10 0.00 0.42
Hispanics% 0.11 0.07 0.13 0.01 0.81 0.06 0.03 0.07 0.01 0.37
Baby boomers% 0.26 0.26 0.02 0.16 0.33 0.27 0.27 0.02 0.22 0.35
1-person households% 0.27 0.27 0.05 0.13 0.47 0.27 0.28 0.02 0.22 0.32
Safety Net
PA recipients% 0.21 0.20 0.09 0.03 0.88 0.19 0.18 0.09 0.06 0.68
SS recipients% 0.31 0.30 0.08 0.08 0.58 0.31 0.31 0.08 0.15 0.61
AFDC/TANF payment $ 449.34 426.00 171.31 170.00 923.00 452.78 432.00 189.98 170.00 923.00
MH expenditures $ 130.03 122.03 68.18 36.09 381.90 122.71 103.76 67.08 36.09 277.03
Transience
Mobility% 0.16 0.16 0.04 0.07 0.29 0.16 0.15 0.03 0.09 0.22
I New Perspectives on Community-Level Homelessness I 619
620 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

TABLE 4

Summary of Mixed-Effects Model for Variables Predicting Adult Homelessness per 10,000 Adults in
General Population

Metro CoCs (N = 338) Non-Metro CoCs (N = 76)

Variable B 95% CI p B 95% CI P

Intercept 0.60 2.801.50 0.493 3.26 5.7011.38 0.509

Housing Market
Rent level 6.34 1.6611.19 0.012 31.78 10.3854.13 0.005
Homeownership 2.05 3.750.68 0.006 5.99 12.920.38 0.070
Vacancy rate 0.18 1.011.66 0.677 0.96 2.384.38 0.575
Economic Conditions
Unemployment rate 1.91 2.658.02 0.332 26.93 8.6640.84 0.004
Extreme poverty 1.88 3.236.53 0.540 10.90 28.6310.22 0.357
Demographics
Blacks 0.07 0.750.79 0.944 4.15 6.701.49 0.002
Hispanics 1.05 0.291.83 0.006 1.68 1.235.37 0.210
Baby boomers 7.43 3.6312.20 <0.001 4.89 8.2621.08 0.381
1-person households 3.58 1.015.37 0.003 2.84 8.3815.20 0.600
Safety Net
PA recipients 1.05 0.122.08 0.080 3.41 5.620.88 0.153
SS recipients 0.76 0.421.96 0.212 0.61 3.662.73 0.773
AFDC/TANF payment 1.17 8.416.63 0.786 11.62 36.206.51 0.161
MH expenditures 12.58 28.822.13 0.096 8.10 52.2451.87 0.925
Transience
Mobility rate 5.39 3.268.66 <0.001 1.77 11.588.01 0.739
SD Random Intercepts 0.25 0.35
Model R 2 0.58 0.67

DISCUSSION
The overarching objective of this study was to contribute to research on community-level
structural determinants of homelessness by introducing new perspectives and directions to this
body of research. On the whole, the use of HUD PIT data measuring homelessness rates has
allowed us to confirm some findings of Lee and colleagues, while increasing the explanatory
power of the model from 35% to 58%. Our analysis also confirms their findings that rent
level, single-person households, and recently moved households are positively associated with
the rate of homelessness in the general population. While the significance of the public assis-
tance receipt rate was not observed in our analysis, in both models a positive association was
found for this variable. As a new contribution to this body of research, we found the home-
ownership rate to be negatively associated, and the size of the Hispanic population and baby-
boomer age cohort to be positively associated with the rate of homelessness among the general
population.
Our results also confirmed findings of Lee and colleagues on the rate of homelessness among
the population in poverty. Once again, rent level and the proportion of single-person households
were confirmed to be positive and highly significant predictors of homelessness in metropolitan
areas. In addition, we found the size of the Hispanic population and the baby-boomer cohort,
as well as the proportion of recently moved households to be positively associated, and mental
health expenditures to be negatively associated, with homelessness. One result contradicted
the findings of the earlier study: both studies found the public assistance receipt rate to be a
significant predictor, however, Lee and colleagues found it was negatively associated with the
 I New Perspectives on Community-Level Homelessness I 621

TABLE 5

Summary of Mixed-Effects Model for Variables Predicting Adult Homelessness per 10,000 Adults in
Poverty

Metro CoCs (N = 338) Non-Metro CoCs (N = 76)

Variable B 95% CI p B 95% CI p

Intercept 0.08 2.571.90 0.811 0.23 9.879.06 0.971

Housing market
Rent level 14.53 9.9620.17 <0.001 39.19 12.2665.71 0.004
Homeownership 0.46 2.181.14 0.570 1.32 8.815.41 0.622
Vacancy rate 0.31 1.801.12 0.707 0.58 3.644.46 0.794
Economic conditions
Unemployment rate 4.95 9.651.31 0.142 15.14 1.8431.58 0.086
Demographics
Blacks 0.17 0.671.03 0.768 3.99 7.240.80 0.014
Hispanics 1.03 0.201.90 0.014 2.64 1.146.94 0.171
Baby boomers 8.58 4.3513.70 <0.001 10.92 5.5029.77 0.173
1-person households 4.07 1.356.20 0.001 2.14 12.7816.36 0.752
Safety net
PA recipients 2.13 0.943.21 <0.001 2.05 5.012.47 0.471
SS recipients 1.12 0.182.40 0.104 0.64 4.603.25 0.757
AFDC/TANF payment 1.58 9.896.69 0.718 6.61 35.9716.49 0.487
MH expenditures 19.57 36.522.75 0.024 8.12 79.1845.74 0.592
Transience
Mobility rate 5.61 3.189.18 <0.001 0.72 12.3411.94 0.878
SD random intercepts 0.27 0.29
Model R 2 0.52 0.56

rate of homelessness among the poverty population while we found a positive association between
these rates. Using the HUD PIT data increased the models ability to explain overall variance in
homelessness from 49% to 52%.
Three primary points can be drawn from the results of our study. First, our findings provide
additional evidence that homelessness has its roots in housing market dynamics, and particularly
in the difficulty in obtaining affordable housing. Second, both of our metropolitan area models
find the size of the baby-boomer cohort, the size of the Hispanic population, and the number
of recently moved households to be positively associated with homelessness. Third, and finally,
our study points to the great potential in using the newly available HUD PIT estimates of the
homeless population to build on prior research and arrive at a better understanding of the structural
determinants of homelessness.
Each of these key findings suggests the need for further study or policy changes. Our findings
on the importance of affordable housing stock for decreasing homelessness underscore the need
for policies that either increase the supply of affordable housing or provide additional safety
net supports to households to help them afford housing and decrease competition for a finite
number of low-rent units. Given the finding that rates of homelessness were higher in areas
with relatively more single-person households, the need for additional safety net supports may
be particularly acute for persons in this group who are especially vulnerable to homelessness
in the event of job loss, illness, or other income shocks. While an expansion of the Section
8 Housing Choice Voucher Program, which is the primary federal housing assistance program
for low-income families, would be the most straightforward remedy to the affordable housing
problems that appear to be the most important drivers of homelessness, recent history and the
622 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

current political and fiscal environment provide little reason to believe that such an expansion is
likely to be forthcoming.
However, there are a number of pragmatic alternative options to a pure expansion of housing
subsidies that might help address the prevailing lack of affordable housing. For example, Khadduri
(2010) describes a plan for overhauling the Section 8 program such that subsidies are targeted
more directly to those individuals and jurisdictions at highest risk of or with the highest rates of
homelessness, thereby making it a more effective tool for preventing homelessness. Providing a tax
credit to all low-income renters similar to the existing Earned Income Tax Credit (EITC), an idea
proposed by Landis and McClure (2010), would also go far in helping those at risk of homelessness
maintain housing. It would also serve as an important counterbalance to the mortgage interest
deduction, which almost exclusively benefits middle- and upper-income Americans. Others have
suggested that relaxing zoning requirements or other local regulations on housing construction
might encourage the creation of more affordable housing units (Joint Center for Housing Studies
at Harvard University, 2008). While these are just a sampling of the ideas that have been put forth
as potential solutions to problems related to housing affordability, they underscore the potential
effectiveness of creative housing policy reform.
Our findings on demographic groups and residential mobility merit closer attention, both in
future research and from a policy standpoint. The finding regarding the baby-boomer age group is
consistent with evidence that baby boomers are highly overrepresented in the homeless population,
and underscores the need to develop targeted interventions to address homelessness among
members of this population. Indeed, substantial reductions in the overall rate of homelessness
could be achieved by targeting this demographic sub group. Here, housing subsidies that offer a
more limited form of support than a full Section 8 voucher might be an especially viable approach,
but additional research is needed to test the effectiveness of such shallow subsidies. The findings
regarding Hispanic ethnicity and residential mobility suggest that migration patterns may have a
more important relationship with the rate of homelessness than has been previously considered.
Future research should delve more deeply into these issues, particularly the finding that rates of
homelessness were higher in areas with relatively more Hispanics. This finding seems inconsistent
with prior research showing Hispanics to be underrepresented in the homeless population, which
has been credited largely to their greater propensity for relying on informal housing arrangements
to avoid homelessness (Rosenheck, Bassuk, & Salomon, 1999). One plausible interpretation of
our finding that could be investigated by future studies is that the documented migration of
Hispanics out of large cities and into smaller communities and rural areas (Kandel & Cromartie,
2004), has strained their social and kinship networks, limiting their opportunities to rely on
informal housing arrangements when needed.
Although this study made progress towards its objective, it also had a few limitations that
bear mentioning. Similar to Lee and colleagues study, the present study does not control for
the capacity of homelessness assistance programs; therefore, our dependent variables may be
conflated with the magnitude of a communitys response to the problem. Even though this study
applies rigorous methods to match the geography at which homelessness is measured to the
geography at which structural factors are measured, the mismatch between the two continues to
present problems in identifying important community-level influences on these rates; alternative
methods to classify CoCs as urban or rural could yield different results.
While this is the first study to examine the structural determinants of homelessness using
a set of non-urban jurisdictions, we have merely taken the first step in expanding this body
of research. The HUD PIT estimates of homelessness used in this study are also available for
a range of sub-populations of persons experiencing homelessness, including families, persons
with serious mental illness, veterans, and persons experiencing chronic homelessness. It is likely
that community-level determinants of homelessness may operate differently for each of these
 I New Perspectives on Community-Level Homelessness I 623

sub-populations, and this should be investigated in future research. Similarly, this study, by
design, considered a fairly small set of factors as potential community-level determinants of
homelessness; there is certainly room to expand research to include a wide range of additional
measures, including those that have received less attention in the literature.
The results of this study demonstrate that the dynamics operating at the macro level are impor-
tant for understanding homelessness and that, correspondingly, macro-level policy interventions
are ultimately necessary to prevent and end homelessness. Future research that expands on this
study and provides additional insight regarding structural factors that are important determinants
of homelessness would be crucial for informing policy-level interventions.

REFERENCES
Appelbaum, R., Dolny, M., Dreier, P., & Gilderbloom, J. (1991). Scapegoating rent control: Masking the causes
of homelessness. Journal of the American Planning Association, 57, 153164.
Bassuk, E. L., Rubin, L., & Lauriat, A. (1984). Is homelessness a mental health problem? American Journal of
Psychiatry, 141, 15461550.
Bohanon, C. (1991). The economic correlates of homelessness in sixty cities. Social Science Quarterly, 72,
817825.
Burt, M. R. (1991). Causes of the growth of homelessness during the 1980s. Housing Policy Debate, 2, 903936.
Burt, M. R. (1993). Over the edge: The growth of homelessness in the 1980s. New York: Russell Sage Foundation.
Burt, M. R. (2001). Helping Americas homeless: Emergency housing or affordable housing? Washington, DC:
The Urban Institute.
Burt, M. R., & Cohen, B. E. (1989). Differences among homeless single women, women with children, and single
men. Social Problems, 36, 508524.
Calsyn, R. J., & Roades, L. A. (1994). Predictors of past and current homelessness. Journal of Community
Psychology, 22, 272278.
Chen, M. H., & Shao, Q. M. (1999). Monte Carlo estimation of Bayesian credible and HPD intervals. Journal of
Computational and Graphical Statistics, 8, 6992.
Culhane, D. P., & Metraux, S. (1999). One-year rates of public shelter utilization by race/ethnicity, age, sex and
poverty status for New York City (1990 and 1995) and Philadelphia (1995). Population Research and Policy
Review, 18, 219236.
Culhane, D. P., Lee, C. M., & Wachter, S. M. (1996). Where the homeless come from: A study of the prior address
distribution of families admitted to public shelters in New York City and Philadelphia. Housing Policy
Debate, 7, 327365.
Culhane, D. P., Metraux, S., Byrne, T., Stino, M., & Bainbridge, J. (in press). The aging of contemporary
homelessness. Contexts.
Curtis, M. A., Corman, H., Noonan, K., & Reichman, N. (2011, February). Life shocks and homelessness (Working
paper No. 16826). Cambridge, MA: National Bureau of Economic Research.
Early, D. W., & Olsen, E. O. (2002). Subsidized housing, emergency shelters, and homelessness: An empirical
investigation using data from the 1990 Census. Advances in Economic Analysis and Policy, 2(1), 2, 234.
Elliott, M., & Krivo, L. J. (1991). Structural determinants of homelessness in the United States. Social Problems,
38, 113131.
Grimes, P. W., & Chressanthis, G. A. (1997). Assessing the effect of rent control on homelessness. Journal of
Urban Economics, 41(1), 2337.
Hawkins, R. L., & Abrams, C. (2007). Disappearing acts: The social networks of formerly homeless individuals
with co-occurring disorders. Social Science and Medicine, 65, 20312042.
Honig, M., & Filer, R. K. (1993). Causes of intercity variation in homelessness. The American Economic Review,
83, 248255.
Hopper, K., Jost, J., Hay, T., Welber, S., & Haugland, G. (1997). Homelessness, severe mental illness, and the
institutional circuit. Psychiatric Services, 48, 659665.
Hudson, C. G. (1998). Estimating homeless populations through structural equation modeling. Journal of Sociology
and Social Welfare, 25, 136154.
624 I JOURNAL OF URBAN AFFAIRS I Vol. 35/No. 5/2013

Hwang, S. W. (2001). Homelessness and health. Canadian Medical Association Journal, 164, 229233.
Joint Center for Housing Studies at Harvard University. (2008). Americas rental housing policy: The key to a
balanced national policy. Cambridge, MA: Author.
Jones, R. E. (1983). Street people and psychiatry: An introduction. Hospital and Community Psychiatry, 34,
807811.
Kandel, W., & Cromartie, J. (2004). New patterns of Hispanic settlement in rural America. Washington, DC: U.S.
Department of Agriculture, Economic Research Service.
Khadduri, J. (2010). Rental subsidies: Reducing homelessness. In I. G. Ellen & B. OFlaherty (Eds.), How to
house the homeless (pp. 5988). New York: Russell Sage Foundation.
Koegel, P., Burnam, M. A., & Baumohl, J. (1996). The causes of homelessness. In J. Baumohl (Ed.), Homelessness
in America (pp. 2434). Westport, CT: Oryx.
Koegel, P., Melamid, E., & Burnam, A. (1995). Childhood risk factors for homelessness among homeless adults.
American Journal of Public Health, 85, 16421649.
Landis, J. D., & McClure, K. (2010). Rethinking federal housing policy. Journal of the American Planning
Association, 76, 319348.
Lee, B. A., & Farrell, C. R. (2003). Buddy, can you spare a dime? Urban Affairs Review, 38, 299324.
Lee, B. A., & Greif, M. J. (2008). Homelessness and hunger. Journal of Health and Social Behavior, 49(1), 319.
Lee, B. A., Price-Spratlen, T., & Kanan, J. W. (2003). Determinants of homelessness in metropolitan areas. Journal
of Urban Affairs, 25, 335356.
Main, T. J. (1996). Analyzing evidence for the structural theory of homelessness. Journal of Urban Affairs, 18,
449457.
Martin, E. (1992). Assessment of S-night street enumeration in the 1990 Census. Evaluation Review, 16, 418438.
McChesney, K. Y. (1990). Family homelessness: A systemic problem. Journal of Social Issues, 46(4), 191
205.
Metraux, S., Roman, C. G., & Cho, R. (2008). Incarceration and homelessness. In D. Dennis, G. Locke, &
J. Khadduri (Eds.), Toward understanding homelessness: The 2007 National Symposium on Homelessness
Research. Washington, DC: U.S. Department of Housing and Urban Development.
OFlaherty, B. (1995). An economic theory of homelessness and housing. Journal of Housing Economics, 4(1),
1349.
OFlaherty, B. (2004). Wrong person and wrong place: For homelessness, the conjunction is what matters. Journal
of Housing Economics, 13(1), 115.
OFlaherty, B. (2009). What shocks precipitate homelessness? (Discussion paper No. 080914). Retrieved from
http://jagiellonia.econ.columbia.edu/bo2/research/whatrisks.pdf
Quigley, J. M. (1990). Does rent control cause homelessness? Taking the claim seriously. Journal of Policy
Analysis and Management, 9(1), 8993.
Quigley, J. M., & Raphael, S. (2002). The economics of homelessness: The evidence from North America.
European Journal of Housing Policy, 1, 323336.
Quigley, J. M., Raphael, S., & Smolensky, E. (2001). Homeless in America, homeless in California. Review of
Economics and Statistics, 83(1), 3751.
R Development Core Team. (2011). R: A language and environment for statistical computing, version 2.13. Vienna,
Austria: R Foundation for Statistical Computing. http://www.R-project.org
Raphael, S. (2010). Housing market regulation and homelessness. In I. G. Ellen, & B. OFlaherty (Eds.), How to
house the homeless (pp. 110140). New York: Russell Sage Foundation.
Rosenheck, R., Bassuk, E., & Salomon, A. (1999). Special populations of homeless Americans. In L. B. Fosburg
& D. L. Dennis (Eds.), Practical lessons: The 1998 National Symposium on Homelessness Research (pp.
2.12.31). Washington, DC: U.S. Department of Housing and Urban Development, U.S. Department of
Health and Human Services.
Rossi, P. H. (1989). Down and out in America: The origins of homelessness. Chicago: University of Chicago Press.
Shinn, M., Weitzman, B. C., Stojanovic, D., Knickman, J. R., Jimenez, L., Duchon, L., James, S., & Krantz, D.
H. (1998). Predictors of homelessness among families in New York City: From shelter request to housing
stability. American Journal of Public Health, 88, 16511657.
Stern, M. J. (1984). The emergence of the homeless as a public problem. The Social Service Review, 58,
291301.
 I New Perspectives on Community-Level Homelessness I 625

Susser, E. S., Lin, S. P., & Conover, S. A. (1991). Risk factors for homelessness among patients admitted to a state
mental hospital. American Journal of Psychiatry, 148, 16591664.
Troutman, W. H., Jackson, J. D., & Ekelund, R. B. (1999). Public policy, perverse incentives, and the homeless
problem. Public Choice, 98, 195212.
Tucker, W. (1987). Where do the homeless come from? National Review, 25, 3243.
Tucker, W. (1989). Americas homeless: Victims of rent control. Washington, DC: The Heritage Foundation.
U.S. Bureau of the Census. (1990). Standard Tape File 2C (STF-2C).
U.S. Department of Agriculture (USDA). (2003). Measuring rurality. Retrieved from http://www.
ers.usda.gov/Briefing/Rurality/
U.S. Department of Housing and Urban Development (HUD). (1984). A report to the Secretary on the homeless
and emergency shelters. Washington, DC: U.S. Government Printing Office.
U.S. Department of Housing and Urban Development (HUD). (2008). A guide to counting unsheltered homeless
people. Washington, DC: Author.
U.S. Department of Housing and Urban Development (HUD). (2011). The 2010 Annual Homeless Assessment
Report to Congress. Washington, DC: Author.
U.S. General Accounting Office (GAO). (1991). 1990 Census: Limitations in methods and procedures to include
the homeless. Washington, DC: Author.
Wolitski, R. J., Kidder, D. P., & Fenton, K. A. (2007). HIV, homelessness, and public health: Critical issues and a
call for increased action. AIDS and Behavior, 11, 167171.
Wright, J. D., & Lam, J. A. (1987). Homelessness and the low-income housing supply. Social Policy, 17(4), 4853.
Zerger, S. (2002). A preliminary review of literature: Chronic medical illness and homelessness. Nashville, TN:
National Health Care for the Homeless Council.