Estimating the Number of People Who Are Homeless: Homeless Management Information Systems

Contents

• Introduction
• What Are Homeless Management Information Systems?
• HUD's Continuing Role in Collecting Information About Homeless Persons
• Development of the HMIS Network
• Congressional Direction
• HUD Actions
• HMIS Data and Technical Standards
• Confidentiality of Domestic Violence Victims
• Status of the HMIS Network
• HMIS Implementation
• Participation of Service Providers in HMIS
• Estimates of the Number of People Who Are Homeless
• CoC Point-in-Time Estimates of Homeless Individuals
• Most Recent CoC Estimates
• The Annual Homeless Assessment Report (AHAR)
• Development of a Nationally Representative Sample of Communities
• Trends Among Sheltered Homeless Persons between 2007 and 2008
• Previous Attempts to Estimate the Number of People Who Are Homeless
• The Urban Institute (1987)
• The National Survey of Homeless Assistance Providers and Clients (1996)
• Sources of Demographic Information About Homeless Persons

Summary

In 1998, Congress directed the Department of Housing and Urban Development (HUD) to develop a process for collecting data about homeless persons. Together with local communities, HUD began in 2001 to implement a series of Homeless Management Information Systems (HMIS). Two categories of federal fund recipients are required to participate in HMIS: organizations that receive grants through the Housing Opportunities for Persons with AIDS (HOPWA) program and organizations that receive HUD Homeless Assistance Grants. The HOPWA program provides housing and supportive services for persons living with HIV/AIDS, while the Homeless Assistance Grants fund transitional and permanent housing, as well as services, for homeless individuals.

Local jurisdictions called "Continuums of Care" (CoCs)—typically cities, counties, or combinations of both—are the entities that implement HMIS. Homeless service providers in these CoCs collect and store information about homeless individuals they serve, and the information is aggregated in computer systems at the CoC level. The data in these systems are being used by some CoCs to assess client needs and to better direct their services.

Congress initially allocated funds for data collection regarding homeless persons in the FY2001 HUD Appropriations Act (P.L. 106-377), and has continued to allocate funds in all HUD spending bills from FY2002 through FY2009. Local communities can then apply to HUD for available funds that they may use to implement HMIS. The last time that HUD reported to Congress about the progress of HMIS (in 2007), community implementation of the data systems had increased. According to HUD, 91% of local CoCs were implementing HMIS in 2006 (compared to 72% in 2005), meaning that they had established systems into which data are entered. Because more CoCs were implementing HMIS in 2006, the percentage that were only in the planning stage had decreased from 20% in 2005 to approximately 9%. As of 2006, only 1% of CoCs were not yet planning an HMIS (compared to 7% in 2005).

HUD released its fourth analysis of data from a sample of participating HMIS jurisdictions—the fourth Annual Homeless Assessment Report (AHAR)—in July 2009. The fourth AHAR used HMIS data from a sample of 222 communities (compared to 98 in the previous AHAR) to derive a national-level estimate of the number of homeless persons from October 2007 through September 2008. The fourth AHAR is the second report in which HUD used data from an entire 12-month period to estimate the number of people who were homeless (the first and second AHARs used three months and six months, respectively). For the first time, then, an entire year's worth of data can be compared to that from the previous year. In addition to the AHAR estimates using HMIS data, local CoCs conduct point-in-time counts of homeless individuals on one day in January at least every two years. HUD has published these results as part of each AHAR.

This report describes the development of HMIS, reports on the continuing progress of HMIS, summarizes information released in the four AHARs, and describes previous attempts to estimate the number of people who are homeless. It will be updated as events warrant.

---

Estimating the Number of People Who Are Homeless: Homeless Management Information Systems

Introduction

It is difficult to ascertain the number and characteristics of persons experiencing homelessness due to the transient nature of the population, although attempts to collect information about homeless individuals have been made in recent decades.¹ Beginning in the mid-1990s, for example, the Department of Housing and Urban Development (HUD) required its grant recipients to provide information about the homeless clients they served. In addition, comprehensive attempts to count homeless individuals were made in both the 1980s and 1990s, first via Census data and then through a national collaborative survey called the National Survey of Homeless Assistance Providers and Clients. However, no systematic method for collecting information about homeless persons has existed until recent years. In response to a directive from Congress in 1998, HUD began in 2001 to develop a system to collect information about homeless individuals; the processes of data collection, organization, and storage systems, which take place at the local level, have been termed Homeless Management Information Systems (HMIS). In July 2009, HUD released results of its fourth analysis of HMIS data—the fourth Annual Homeless Assessment Report (AHAR). This CRS report describes the development of HMIS, the results of the four AHARs, and previous attempts to count homeless individuals.

What Are Homeless Management Information Systems?

Homeless Management Information Systems (HMIS) are databases established at the local level through which homeless service providers collect, organize, and store information about homeless clients who receive services. HUD is implementing the HMIS initiative through local "Continuums of Care" (CoCs), which acquire and process data from all participating local service providers. CoCs are local boards formed by communities—typically cities, counties, or combinations of both—made up of representatives from nonprofit service providers, advocacy groups, local government, and other interested organizations.² Local boards identify the needs of homeless persons in their communities and try to ensure that they receive the appropriate mix of preventative assistance, emergency services, transitional housing, supportive services, and permanent housing. Local homeless services providers submit requests for funding to their local CoCs, which each in turn submit single consolidated applications to HUD. As of 2009, there were approximately 450 CoCs, including those in the territories.³

Two types of organizations are required to participate in HMIS: those that receive funding through the Housing Opportunities for Persons with AIDS (HOPWA) program and those that receive Homeless Assistance Grants. The HOPWA program, enacted in 1990 (P.L. 101-625) provides housing and supportive services for persons living with HIV/AIDS.⁴ The Homeless Assistance Grants, enacted as part of the McKinney-Vento Homeless Assistance Act (P.L. 100-77),⁵ consist of both formula grants, which are distributed through the Emergency Shelter Grants program, and competitive grants, which are available through the Shelter Plus Care program, Supportive Housing Program, and Section 8 Moderate Rehabilitation Assistance for Single Room Occupancy Dwellings program.⁶ Other service providers that serve homeless individuals and families but do not receive federal funds from these sources are also encouraged to participate in HMIS.

HUD's Continuing Role in Collecting Information About Homeless Persons

Even prior to the congressional directive to implement HMIS (described in the next section of this report, "Development of the HMIS Network"), HUD began efforts to collect information about homeless clients served in the communities that receive HUD Homeless Assistance Grants. Beginning in the mid-1990s, about the time that the Continuum of Care system developed, HUD required applicants for Homeless Assistance Grants to include in their applications information about the number of persons receiving assistance and the type of assistance they received. Initially this was done in narrative form. However, by 2003, the grant application required CoC applicants to complete a housing activity chart that included a point-in-time count of homeless individuals and families receiving services, though HUD did not specify when this count should take place.⁷ The 2003 application also asked applicants to categorize subpopulations served, including the number of chronically homeless individuals, veterans, those with severe mental illnesses, those with HIV/AIDS, and victims of domestic violence. Some CoCs used database systems similar to HMIS to keep track of homeless individuals who were served; these predecessor systems are sometimes referred to as "legacy systems."⁸

The 2005 HUD point-in-time count of homeless persons marked the first time that HUD required all CoCs to conduct a count of both sheltered and unsheltered homeless individuals, and to do it at a particular time of year. HUD directed CoCs to conduct a one-night count during the last week of January of both clients who used homeless services and those who were on the street.⁹ HUD continues to require CoCs to conduct point-in-time counts every two years, though some CoCs choose to conduct counts every year. For example, in 2008, approximately 67% of CoCs voluntarily conducted counts.¹⁰ The most recent point-in-time count in which all CoCs participated occurred in January 2009. The results of CoC point-in-time counts are described later in this report, in the section "CoC Point-in-Time Estimates of Homeless Individuals."

Development of the HMIS Network

Congressional Direction

HUD's ongoing attempts to collect information about homeless individuals were given greater direction beginning in 1998, when Congress instructed HUD to count homeless persons and gather data about both their characteristics and use of homeless assistance services. The FY1999 HUD spending bill (P.L. 105-276) set aside up to 1% of the total appropriation for Homeless Assistance Grants for systems to collect information about those persons experiencing homelessness. Specifically, Congress directed HUD to produce an unduplicated count of homeless persons and to collect information about homeless individuals surveyed such as age, race, sex, disability status, health status, and income; the types of services that homeless clients received; and client outcomes such as length of stay in transitional housing, success in acquiring permanent housing, and employment status.¹¹ Congress concluded that this information would allow HUD to better assess the quality of service programs supported with federal funds.¹²

Congress provided further direction to HUD in the HUD Appropriations Act for FY2001 (P.L. 106-377). The law made Supportive Housing Program funds available for local CoCs to implement management information systems.¹³ Congress directed HUD to work with local jurisdictions to develop a system to collect data, and to be ready to analyze the data within three years of passage of the appropriations bill.¹⁴ Congress also requested that HUD provide Congress with a report on its findings containing an unduplicated count of homeless persons and a descriptive profile of the population.¹⁵ The FY2001 Appropriations Act once again allocated funds to pay for data collection, this time setting aside 1.5% of the $1.02 billion appropriation for Homeless Assistance Grants. Congress has continued to allocate funds for homeless data collection in spending bills from FY2002 to FY2009.

HUD Actions

In the time since Congress directed HUD to implement a system to count homeless persons and collect information about their characteristics, HUD has issued six annual reports to Congress updating its progress. In an initial report, dated August 2001, HUD stated that it would help CoCs collect homelessness data through four means:¹⁶

• flexibly implementing the new Homeless Management Information System (HMIS) eligible activity under the Supportive Housing Program in the 2001 McKinney-Vento competition;
• initiating a comprehensive technical assistance program to help local jurisdictions collect unduplicated client-level data by 2004;
• developing an approach to obtaining meaningful data for an Annual Homeless Assessment Report from a nationally representative sample of jurisdictions; and
• analyzing the most viable approaches to obtaining homeless client-level reporting.

Since issuing this 2001 report, HUD has initiated a number of activities to follow through on its pledge to assist CoCs. It specified that CoCs may use Supportive Housing Program funds for computer hardware, software, and personnel to manage and operate information systems, analyze HMIS data, and produce reports. HUD technical assistance teams hold training sessions for CoCs across the country, and HUD sponsors national conferences in which it provides sessions on a wide range of topics, including data entry, strategies for including data on domestic violence clients and chronically homeless individuals, and how to use HMIS to evaluate program performance and improve services to persons experiencing homelessness. HUD established a website—HMIS.Info—where information about HMIS implementation across the country can be disseminated.¹⁷

HMIS Data and Technical Standards

On July 30, 2004, HUD released its final notice on HMIS data and technical standards that local CoCs are expected to follow when they collect information about their homeless clients.¹⁸ The standards describe two levels of data collection—universal data, which homeless service providers must collect from all clients, and program-specific data, which programs that receive certain types of funding must collect but that other programs are encouraged to collect as well.

All participants must report on universal data elements, which include name, date of birth, race, ethnicity, gender, veteran status, Social Security Number, prior residence, and disabling conditions.¹⁹ In general, all programs that receive funds under the McKinney-Vento Homeless Assistance Grants and HOPWA are required to provide program-specific data; this requirement is not new, as HUD already requires grantee organizations to provide this information in their Annual Progress Reports.²⁰ Included in program-specific data elements are amount and sources of income, receipt of non-cash benefits, physical and developmental disabilities, HIV status, mental illness, substance abuse status, domestic violence status, services provided, and client outcomes.²¹

The HUD data standards were updated in 2009. Among the changes are a new universal data element regarding homeless status that is meant to distinguish those at risk of becoming homeless from those who meet HUD's current definition of homelessness.²² This category will capture those individuals and families being served through the new Homelessness Prevention and Rapid Re-Housing Program (HPRP).²³ The new data standards also allow for HMIS to capture information regarding financial assistance and housing relocation and stabilization services provided to clients; recipients of HPRP funds are required to enter this information into HMIS while it is optional for other providers.

Confidentiality of Domestic Violence Victims

Due to the sensitive nature of much of the information that homeless service providers must collect, some groups that provide services to domestic violence victims raised privacy concerns to HUD after its release of proposed data and technical standards but prior to release of final standards in 2004. These organizations requested that information about domestic violence victims not be included in HMIS.²⁴ At the time, HUD acknowledged the sensitivity of certain information, but concluded that CoCs could collect the information in such a way that would protect the identity of those in the system. To this end, HUD included privacy and security standards in the data and technical standards that all organizations must follow.

However, on January 5, 2006, President Bush signed the Violence Against Women and Department of Justice Reauthorization Act (P.L. 109-162), which included provisions to (1) amend the McKinney-Vento Homeless Assistance Act to prevent victim service providers from disclosing personally identifying information through HMIS, and (2) permit disclosure of non-personally identifying information only after a public notice and comment period. On March 16, 2007, HUD released a notice regarding HMIS and the amendments to McKinney-Vento made by P.L. 109-162.²⁵ In the notice, HUD confirmed that it would require disclosure of non-personally identifying information only after going through a notice and comment period. Since then, Congress enacted the Homeless Emergency Assistance and Rapid Transition to Housing (HEARTH) Act as part of P.L. 111-22, which contained the same language as P.L. 109-162. In its data standards and other guidance, HUD has instructed domestic violence service providers not to disclose personally identifying information in HMIS. In the case of the new Homelessness Prevention and Rapid Re-Housing Program (HPRP), HUD has directed that service providers not enter information into HMIS, and instead use a "comparable database."²⁶

Status of the HMIS Network

Two aspects of HMIS implementation contribute to a CoC's ability to capture data regarding homeless persons. The first aspect is whether a data collection system has been established at the CoC level, and the second is the degree to which homeless service providers within a CoC are participating in the system. Although almost all CoCs have established an HMIS system into which data may be entered, the extent to which data are actually entered into these systems remains incomplete, on average. Once established, a comprehensive HMIS network is meant to improve the ability of communities to provide services to homeless persons as well as to help HUD determine how best to allocate resources.²⁷

HMIS Implementation

HUD's initial goal was that every CoC implement an HMIS by October 2004—meaning establish a system into which service providers are entering data. Although this goal was not accomplished by 2004, the number of CoCs participating in HMIS has increased in every year since 2001.²⁸ According to the most recent HUD report to Congress (from 2007), between 2005 and 2006 the percentage of CoCs that had implemented an HMIS (meaning they were actually inputting data) increased from 72% to 91%.²⁹ From 2005 to 2006, the percentage of CoCs that had decided to implement an HMIS but were still in the planning stages decreased from 20% to 9%, and the percentage that were not yet planning an HMIS dropped from 7% to 1%.³⁰

At the local level, CoCs have several options for implementing and maintaining their HMIS databases. Not all CoCs are implementing their own HMIS. Some are collaborating to create a multi-jurisdictional HMIS with two or more CoCs. Others are planning to make individual CoC data accessible at the state level, while 19 states have decided to implement a state-level HMIS.³¹ Local initiatives also differ in their methods of incorporating service providers into HMIS. Local CoCs may use one central HMIS, into which all service providers input client information. Another option is to allow service providers to use different database systems, but to have technical specialists available at the CoC level to merge all data into one unified system. A third option is to use side-by-side systems where individual service providers enter data into their own systems, and also enter data into a CoC-wide HMIS.

Participation of Service Providers in HMIS

Even where CoCs have successfully implemented HMIS, coverage of homeless service providers may be incomplete. HUD uses the term "bed coverage" to describe the rate at which local service providers within a CoC participate in HMIS. The term refers to the percentage of available beds in a CoC that are actually accounted for in HMIS. If not all service providers within a CoC participate in HMIS, then bed coverage may be low. Issues with bed coverage may arise in cases of domestic violence shelters that are reluctant to report data due to confidentiality concerns, or where service providers do not receive HUD funds and are not required to participate in HMIS. In addition, even when service providers report data to HMIS, they might not include all clients served, which could result in another limitation on the usefulness of the data.³²

HUD keeps track of bed coverage rates both by the type of shelter provided, such as emergency shelter, transitional housing, and permanent housing, and by household type, such as homeless individuals and homeless families. The most recent HUD report to Congress that discussed bed coverage rates was issued in 2007. Therefore, current bed coverage rates may be greater than those described here. From 2005 to 2006, the average number of beds across CoCs that were included in HMIS increased in all categories.³³ HUD reports bed coverage as an average rate—the average of all CoCs' bed coverage rates.

• Emergency Shelter: The average bed coverage rate for shelters serving individuals went from 43% in 2005 to 55% in 2006. For shelters serving homeless families, the average bed coverage rate went from 45% in 2005 to 51% in 2006.
• Transitional Housing: The average bed coverage rate for transitional housing serving homeless individuals increased from 41% in 2005 to 50% in 2006. Average bed coverage rates for homeless families increased from 51% to 62%.
• Permanent Housing: Average bed coverage rates for permanent supportive housing for individuals went from 46% in 2005 to 58% in 2006. Average bed coverage rates for homeless families went from 54% in 2005 to 58% in 2006.

Estimates of the Number of People Who Are Homeless

Since the 1980s, a number of attempts have been made to estimate the total number of homeless persons in the country as well as to describe their characteristics. Although the specific methods used in the studies have varied, in most, researchers surveyed a sample of the homeless population and used the sample to estimate the total number of homeless persons in the country. The time periods covered by these counts vary. Some are "point-in-time" counts that estimate the number of homeless people on a single night during the year. Others estimate the number of persons who are homeless during longer periods—a week or span of months. Researchers have also used samples to estimate the total number of persons who are homeless at some point during the year.³⁴

The HMIS initiative differs from these previous efforts to count homeless people and gather information. Instead of sampling only certain communities or counting homeless individuals on only a single night, CoCs gather information from all homeless assistance providers regarding all homeless individuals who use their services each day of the year. Eventually, once communities have fully implemented HMIS, the network of systems is expected to provide an annual unduplicated count of homeless persons from each jurisdiction. Counting homeless populations on the street might continue to be important, however, as their use of services is unknown.³⁵ HUD released its first report to Congress using HMIS data, the Annual Homeless Assessment Report (AHAR), in February 2007. Since then, HUD has released three more AHARs; the most recent was released in July 2009.³⁶ Because HMIS is not fully implemented in all jurisdictions around the country, the four AHARs, like previous efforts to count homeless persons, rely on a sample of jurisdictions.

This section describes several efforts to estimate the number of homeless individuals over the years. These include CoC point-in-time counts that take place at least every two years, estimates in the four AHARs using HMIS data, and previous estimates from the 1980s and 1990s. This section also includes a description of demographic information regarding homeless persons from various sources.

CoC Point-in-Time Estimates of Homeless Individuals

As mentioned earlier in this report, in 2005, 2007, and 2009, HUD required all CoCs to conduct point-in-time counts of both the sheltered and unsheltered homeless individuals in their jurisdictions. HUD has directed that the counts take place on one day during the last week of January, when shelter use is expected to be high. Some CoCs also conduct counts in off years. In 2008, 67% of CoCs voluntarily conducted point-in-time counts.³⁷ Although CoCs do not necessarily use HMIS in their point-in-time counts,³⁸ eventually HUD expects the HMIS initiative to be part of this point-in-time collection of information about homeless individuals. As HMIS develops, CoCs will be able to use the systems as part of the data collection process in estimating the number of sheltered homeless people.³⁹

The reliability of CoC point-in-time data vary by Continuum, particularly in the case of estimates of unsheltered homeless individuals. Unsheltered individuals are those living in places not meant for human habitation, such as cars, abandoned buildings, highway underpasses, and public parks. Although HUD has published guidance on how to conduct street counts⁴⁰ and provides technical assistance to CoCs, the task is complicated, and not all CoCs are able to conduct statistically reliable surveys of those individuals who are not sheltered.⁴¹

During the point-in-time counts, HUD also asks participating CoCs to collect information about homeless individuals, which is referred to as "subpopulation information." CoCs are to ask homeless individuals whether they are chronically homeless; have severe mental illnesses, substance abuse disorders, or HIV/AIDS; are veterans; have experienced domestic violence; or are unaccompanied youth. CoCs are not always able to gather this information, and even when they do, according to HUD, the subpopulation information is less reliable than the estimates of the number of homeless individuals.⁴² Information about homeless subpopulations from each point-in-time count is available on HUD's website.⁴³

Most Recent CoC Estimates

In January 2009, all CoCs were required by HUD to conduct a point-in-time count. Because those data are not yet available, 2007 marks the most recent year in which data are available from all CoCs. The 2007 results for the 50 states, the District of Columbia, Puerto Rico, and the territories are as follows:

• the sheltered homeless population consisted of 391,401 persons on a single day during the last week of January 2007;
• the unsheltered homeless population numbered 280,487;
• the total number of homeless individuals counted on one day during the last week of January 2007 was 671,888.

In 2008, approximately 67% of CoCs voluntarily conducted point-in-time counts on one night during the last week of January.⁴⁴ The approximately 33% of CoCs that did not conduct counts in 2008 reported their counts from the previous year (when every CoC was required to conduct a point-in-time count).⁴⁵ These CoCs reporting 2007 estimates represented about 43% of the sheltered and unsheltered homeless persons in the 2008 count.⁴⁶ CoCs with large homeless populations, such as the City of Los Angeles, Chicago, and Detroit, did not conduct 2008 counts. As such, the 2008 count may not be representative of communities nationwide.⁴⁷ The 2008 results for the 50 states, the District of Columbia, Puerto Rico, and the territories are as follows:

• the sheltered homeless population consisted of 386,361 persons on a single day during the last week of January 2008; ⁴⁸
• the unsheltered homeless population numbered 278,053;
• the total number of homeless individuals counted on one day during the last week of January 2008 was 664,414.

The Annual Homeless Assessment Report (AHAR)

As mentioned earlier in this report, when Congress directed HUD to create homeless management information systems, it also required HUD to issue a report to Congress containing an unduplicated count and descriptive profile of homeless individuals. HUD released the first Annual Homeless Assessment Report in February 2007.⁴⁹ The report used HMIS data to estimate the number of people who experienced homelessness during a three-month period in 2005—from February through April. Since then, HUD has released three more AHARs; in the last two, HUD estimated the number of people who were homeless during a 12-month period. (The second AHAR estimated the number of people who were homeless during a six-month period.⁵⁰) For each of the four AHARs, researchers relied on HMIS data collected from a sample of communities during a period of time and used these data to derive national-level estimates of the number of homeless persons. See Table 2 for estimates from each of the four AHARs.

The HMIS data in the four AHARs provide estimates only of the sheltered homeless population—individuals living in emergency shelters and transitional housing—and do not include estimates of individuals living on the street or other places not meant for human habitation. Only HMIS data from sample sites in which there is at least 50% bed coverage in at least one of four categories (emergency shelter for individuals, emergency shelter for families, transitional housing for individuals, and transitional housing for families) is used in each AHAR.⁵¹ Only the first AHAR includes data from domestic violence shelters. ⁵² The four AHARs also reported data collected from CoCs during their one-night counts of homeless persons in January of 2005, 2006, 2007, and 2008, each of which included individuals and families who were on the street or similar locations, as well as those in emergency shelters or transitional shelters.

Each AHAR has also used HMIS data to provide national estimates of the demographic characteristics of sheltered homeless persons. (See Table 3 for demographic information.) In each AHAR, individual adults made up the majority of the homeless population (at least 65%) compared to people in families. Single adult men represented the largest percentage of the homeless population in each of the four AHARS, ranging between 47% and 53%. Among racial groups, the highest percentage of homeless people were African American in each AHAR. The four AHARs also published reports regarding disability and veteran status. However, depending on the AHAR, large percentages of records submitted to HUD were missing information on veteran and disability status, though these levels have been decreasing. For example, in the first AHAR, 35% of records were missing information on veteran status; in the fourth AHAR, 7.5% of records were missing this information.

Development of a Nationally Representative Sample of Communities

The national estimates of the number of homeless persons that used emergency shelters and/or transitional housing at some point during a given time period, presented in each AHAR, were produced using methodology developed by Abt Associates Inc. and the University of Pennsylvania's Center for Mental Health Policy and Research.⁵³ The estimates depend on a nationally representative sample of communities (the AHAR sample). The original AHAR sample contained 80 Community Development Block Grant (CDBG) jurisdictions located within 71 CoCs.⁵⁴ However, minimum HMIS requirements meant that some sample communities were excluded from the analysis. In order to participate, each jurisdiction was required to have a minimum level of bed coverage—only CoCs in which at least 50% of beds in at least one of four categories (emergency shelter for individuals, emergency shelter for families, transitional housing for individuals, and transitional housing for families) could participate in the AHAR.⁵⁵

The primary sampling units in the AHAR are CDBG jurisdictions.⁵⁶ A total of 3,142 CDBG jurisdictions, comprising the 50 states, was used to select the sample. There were essentially two parts to setting up the first AHAR sample. First, since research had shown that different geographic areas experience different rates of homelessness, and a few of the CDBG jurisdictions accounted for a disproportionately large share of the U.S. population, the researchers decided to select some sampling sites with certainty (i.e. not randomly).⁵⁷ This meant that these jurisdictions would only represent themselves in the national estimates rather than representing jurisdictions of similar size in the nation as a whole. This method reduces sampling variability; otherwise, the estimates produced from a sample drawn completely at random could fluctuate wildly depending on whether such large population CDBG jurisdictions had been chosen for inclusion in the sample. A total of 18 of these so-called certainty sites were selected for inclusion in the AHAR sample.⁵⁸

Second, the researchers divided the remaining CDBG jurisdictions into 16 strata based on four types of geography (central cities, cities of 50,000 or more that are not central cities, urban counties, and rural areas) and Census regions—for example, one stratum would be Northeast Central City.⁵⁹ Each stratum was further divided into groups based on size. Then the number of non-certainty sites were allocated across the strata based on adjusted population.⁶⁰ The 62 sample sites were then drawn at random, one site from each group in each stratum. The 80 sample sites (the 18 certainty sites and 62 randomly chosen sites) contained approximately 14% of the U.S. population and over 40 million persons.⁶¹

At the beginning of data collection for the first AHAR, 55 of the 80 sample communities had implemented HMIS and were included in the report's analysis.⁶² In addition, nine communities outside of the AHAR sample, labeled "contributing communities," met the minimum standards and were included to arrive at estimates in the first AHAR. ⁶³ The contributing communities, similar to the sample sites selected with certainty, only represent themselves in the national estimates.⁶⁴

For the second AHAR, 58 sample communities were able to contribute HMIS data.⁶⁵ This included 49 communities that participated in the first AHAR, plus an additional nine sample communities that were able to meet the minimum standards for the first time. Five sample communities that participated in the first AHAR did not meet the minimum standards to contribute to the second AHAR.⁶⁶ An additional 16 contributing communities not in the sample were able to provide data for the second AHAR,⁶⁷ bringing the total number of sites used to derive the national estimates to 74 communities. The third AHAR included data from 98 communities—61 sample sites and 37 contributing communities.⁶⁸

For the fourth AHAR, modifications were made to the AHAR sample to address a number of issues that emerged in the previous reports. Some communities in the original sample, especially in rural areas, had not been able to contribute data because of HMIS implementation issues or other concerns regarding data quality. In addition, many rural communities did not have any homeless service providers from which to report data.⁶⁹ To address these issues, 22 rural communities were added to the AHAR non-certainty sample, bringing the total number of AHAR sample communities to 102 (18 certainty sites and 84 non-certainty sites).⁷⁰ For the fourth AHAR, 87 of these 102 sample sites had implemented HMIS and were able to provide data,⁷¹ supplemented by 135 contributing communities, bringing the total number of communities providing data for the fourth AHAR to 222 sites.⁷²

Trends Among Sheltered Homeless Persons between 2007 and 2008

The fourth AHAR marks the first time that year-to-year comparisons can be made in sheltered homeless estimates derived from HMIS data. ⁷³ The number of people that used an emergency shelter, transitional housing, or both, in 2008 compared to 2007 increased by about 5,200 people to 1,593,794. (See Table 4.) The number of sheltered homeless individuals (versus homeless people in families) decreased from 2007 to 2008, by 22,422 people, while the number of sheltered homeless persons in families increased by 43,183 people over the same period. The number of sheltered homeless individuals as a percentage of all sheltered homeless persons decreased from 70.2% in 2007 to 68.6% in 2008. Conversely, the percentage of sheltered homeless persons in families as a percentage of all sheltered homeless persons increased from 29.8% in 2007 to 32.4% in 2008.⁷⁴

The majority of sheltered homeless persons (68%) continue to be concentrated in urban areas. (See Table 5.) However, there was a noticeable uptick of homeless persons residing in suburban or rural areas.⁷⁵ While in 2007 about 23% of sheltered homeless persons were located in suburban or rural areas, in 2008 they made up 32% of the sheltered homeless population. This shift is even more prominent when looking only at persons in families (versus individuals). (See Table 6.) The share of sheltered homeless persons in families located in suburban or rural areas increased from almost 27% in 2007 to over 38% in 2008, an increase of more than 11 percentage points. In 2008, 29% of sheltered homeless individuals resided in suburban or rural areas, compared with a little over 21% of homeless individuals that resided in suburban or rural areas in 2007.

The fourth AHAR also showed changes in the racial and ethnic makeup of sheltered homeless individuals and families. (See Table 7.) The number of homeless persons considered White, non-Hispanic increased from 42.6% to 44.6% in the case of individuals and from 21.3% to 24.4% in the case of persons in families. The number of homeless individuals who were African American increased (from 33.2% to 37.0%) while those in families decreased (from 55.2% to 50.9%). Those homeless persons who were White, Hispanic followed the opposite pattern, with a decreased percentage of individuals who were homeless (from 14.1% to 11.0%) and an increase among homeless people in families (from 9.8% to 13.1%).

The previous living situation of sheltered homeless persons (as measured by the living arrangement the night before entering a shelter) showed a change in pattern from 2007 to 2008. (See Table 8.) In 2007, about 30% of sheltered homeless adults in families had already been homeless before entering an emergency shelter or transitional housing, while a little more than 54% of adults in families had come from some type of housing—their own or staying with friends or family. In 2008, there was an increase in the proportion of adults in families coming from some type of housing of 7.1 percentage points. The proportion of adults in families who had already been homeless before entering an emergency shelter or transitional housing decreased by over 4 percentage points, to a little less than 26% of the population of sheltered homeless adults in families.

In 2008, the median length of stay in shelters also showed some changes. (See Table 9.) In 2008, the median length of stay for individuals in emergency shelters increased by four nights to a total of 18 nights. Persons in families continued to stay in emergency shelters a median length of 30 nights. In the case of transitional housing, both individuals and persons in families increased their median length of stay from the previous year. In 2008, individuals increased their length of stay by 16 nights from 2007, to a total of 107 nights, while persons in families increased their length of stay by 10 nights, to a total of 161 nights.

While African Americans continued to be overrepresented among those in families who stayed in emergency shelters for more than 180 days, their proportion in 2008 decreased by more than 17 percentage points, from almost 88% of long-term stayers in 2007 to approximately 71% in 2008. (See Table 10.) Conversely, those identifying as White, non-Hispanic/Latino; White, Hispanic/Latino; or those identifying as other racial groups (including those who identify as multiple racial groups) made up larger proportions of persons in families who stayed in emergency shelters more than 180 days in 2008. These groups showed increases of 1.2, 6.9, and 9.3 percentage points, respectively.

Previous Attempts to Estimate the Number of People Who Are Homeless

Previous attempts have been made both to arrive at an accurate count of the number of homeless persons in the United States and to describe their characteristics. The first national estimate occurred in 1984, when HUD reported an estimate of homeless individuals using various means to arrive at a result. These included surveying persons knowledgeable about homelessness in randomly selected metropolitan areas, asking service providers to estimate the number of homeless individuals in their area, and assessing the results of various homelessness counts that had been conducted in local areas.⁷⁶ Through this process, HUD estimated that between 250,000 and 350,000 individuals were homeless at a given point in time. Two more recent, comprehensive estimates are described below.

The Urban Institute (1987)

In March 1987, the Urban Institute conducted interviews of a sample of homeless individuals living in 34 different cities with a population of 100,000 or more and who used soup kitchens and shelters.⁷⁷ The researchers estimated that the number of homeless persons during an average seven-day period in March 1987 ranged from 496,000 to 600,000.⁷⁸ They used this seven-day estimate to project that approximately 1 million individuals were homeless at some time during 1987.⁷⁹

The National Survey of Homeless Assistance Providers and Clients (1996)

The Urban Institute released a second estimate in 2000 using data collected in 1996 by the Census Bureau as part of the National Survey of Homeless Assistance Providers and Clients (NSHAPC). The NSHAPC surveyed both homeless individuals and service providers. Surveys were conducted in 76 communities of varying size and included clients and staff of numerous organizations such as emergency shelters, transitional and permanent housing facilities, soup kitchens, food pantries, and drop-in centers.⁸⁰ Although the purpose of the NSHAPC was not to arrive at a count of homeless individuals,⁸¹ researchers used the data to arrive at an estimate of the number of homeless individuals who relied on homeless services during two different seven-day periods in 1996.⁸² During a seven-day period in the fall of 1996, an estimated 444,000 clients used homeless assistance services,⁸³ and during a seven-day period in the winter of that year, the number was estimated to be 842,000.⁸⁴ The researchers used these numbers to estimate that during all of 1996, between 2.3 million and 3.5 million individuals were homeless at some time.⁸⁵

Sources of Demographic Information About Homeless Persons

In addition to efforts to count the number of homeless individuals, attempts have been made to describe the characteristics of the national homeless population. Each of the four Annual Homeless Assessment Reports have collected information regarding age, familial status, race, disability and veteran status. The demographic findings of these reports are described earlier in this report. In addition, earlier efforts as part of the National Survey of Homeless Assistance Providers and Clients and through the Census Bureau provide demographic information regarding homeless persons from the 1990s. Further, ongoing surveys conducted by the U.S. Conference of Mayors annually captures information about homeless individuals, and the Department of Veterans Affairs attempts to estimate the number of homeless veterans on an annual basis. These efforts are described below.

National Survey of Homeless Assistance Providers and Clients

The NSHAPC data resulted in demographic, income, and other information about homeless individuals in 1996.⁸⁶ Among the findings were that homeless clients were predominantly male (68%) and nonwhite (53%); 23% of homeless clients were veterans.⁸⁷ Large proportions of homeless adults had never married (48%) and had not received a high school diploma (38%).⁸⁸ The NSHAPC also found that although 48% of homeless adults had minor children, only 31% of those with children lived with them.⁸⁹ Thirty-eight percent of homeless clients reported alcohol problems during the past month, and 39% reported mental health problems during that period.⁹⁰ Over one-quarter (27%) of homeless clients had lived in foster care, a group home, or other institutional setting for part of their childhood.⁹¹ Twenty-five percent reported childhood physical or sexual abuse.⁹²

Census Bureau

The Census Bureau released a report using data collected during the 2000 Census of individuals living in emergency and transitional housing. The information was collected on one day in March 2000 and captured information from nearly 171,000 respondents. The report described some basic demographic characteristics of those who were included in the survey.⁹³ Of those persons who were interviewed, 74% were adults (age 18 and older), and of the entire population (adults and children), 61% were male and 39% were female.⁹⁴ The most respondents were white (41%), slightly fewer were African American (40%), and 20% reported that they were Hispanic.⁹⁵

U.S. Conference of Mayors Hunger and Homelessness Survey

The U.S. Conference of Mayors has issued an annual report since 1984, in which between 20 and 30 large cities survey their social service providers' efforts to combat hunger and homelessness and provide housing.⁹⁶ In 2008, the U.S. Conference of Mayors appointed 27 mayors to serve on its Task Force on Hunger and Homelessness. The cities where those 27 mayors serve were asked to respond to a survey for the organization's annual report on hunger and homelessness for the time period between October 1, 2007, and September 30, 2008; 25 cities responded.⁹⁷ Regarding the demographics of the homeless population, the 2008 report focused on the living situation of single adults compared to families. The surveyed cities reported that the population living on the streets consisted overwhelmingly of single adults (94%).⁹⁸ Persons living in families made up approximately 4% of the total population living on the streets, and unaccompanied youth about 2%. Single adults also made up the greatest share of the population living in emergency shelters (nearly 67%), with members of families making up 29% of those in shelters and unaccompanied youth 1%. Those individuals living in families made up the majority of the transitional housing population – approximately 56% versus 43% who were single adults. In permanent supportive housing, however, single adults represented 60% of the population, compared to slightly less than 40% who were people living in families. Among cities that were able to provide more detailed information about their homeless residents, an estimated 26% of the homeless population had serious mental illnesses, 13% had physical disabilities, 2% were HIV positive, 13% were veterans, and 15% percent were victims of domestic violence.⁹⁹

Department of Veterans Affairs

In the area of veterans who experience homelessness, the Department of Veterans Affairs (VA) annually estimates the number of veterans who are homeless through the "Community Homelessness Assessment, Local Education and Networking Groups" (CHALENG) process. The estimates are based on a variety of sources, although the VA is attempting to make its process consistent with HUD's CoC point-in-time counts of homeless individuals. In its most recent report, the VA estimated that in 2008 approximately 131,000 veterans were homeless on one day during the last week of January.¹⁰⁰ For more information about the CHALENG process and estimates, see CRS Report RL34024, Veterans and Homelessness, by [author name scrubbed].

Acknowledgments

The sections of this report entitled "Most Recent CoC Estimates" and "The Annual Homeless Assessment Report (AHAR)" were written by CRS intern Mark Strayer.

Footnotes

1.	Currently, the definition of "homeless individual" under the McKinney-Vento Homeless Assistance Act (P.L. 100-77) is "(1) an individual who lacks a fixed, regular, and adequate nighttime residence; and (2) an individual who has a primary nighttime residence that is—(A) a supervised publicly or privately operated shelter designed to provide temporary living accommodations (including welfare hotels, congregate shelters, and transitional housing for the mentally ill); (B) an institution that provides a temporary residence for individuals intended to be institutionalized; or (C) a public or private place not designed for, or ordinarily used as, a regular sleeping accommodation for human beings." Note, however, that the definition of "homeless individual" will change somewhat as the result of enactment of the HEARTH Act (see P.L. 111-22). For more information about the change in definition, see the section entitled "Defining Homelessness" in CRS Report RL30442, Homelessness: Targeted Federal Programs and Recent Legislation, coordinated by [author name scrubbed].
2.	States may also constitute CoCs to coordinate funding in sparsely populated areas.
3.	"HUD-Defined CoC Names and Numbers Listed by State," revised July 2009, http://www.hudhre.info/documents/2009CoCNamesNumbers.pdf.
4.	For more information on the HOPWA program, see CRS Report RL34318, Housing for Persons Living with HIV/AIDS, by [author name scrubbed].
5.	P.L. 100-77 is codified at 42 U.S.C. §§11301-11435.
6.	For more information about the Homeless Assistance Grants, see CRS Report RL33764, The HUD Homeless Assistance Grants: Distribution of Funds, by [author name scrubbed].
7.	The FY2003 application is available on HUD's website, http://www.hud.gov/library/bookshelf12/supernofa/nofa03/cocapp.doc.
8.	See U.S. Department of Housing and Urban Development, The Annual Homeless Assessment Report to Congress, February 2007, p. 2, http://www.huduser.org/Publications/pdf/ahar.pdf (hereinafter, First AHAR).
9.	Ibid., p. 16.
10.	U.S. Department of Housing and Urban Development, Office of Community Planning and Development, The 2008 Annual Homeless Assessment Report to Congress, July 2009, p. 9, http://www.hudhre.info/documents/4thHomelessAssessmentReport.pdf (hereinafter, Fourth AHAR).
11.	See House Committee on Appropriations, Department of Veterans Affairs and Housing and Urban Development and Independent Agencies Appropriations Act 1999, report to accompany H.R. 4194, H.Rept. 105-610, 105th Cong., 2nd sess., July 8, 1998. The FY1999 HUD Appropriations Act referred to the House Committee Report language for specific requirements.
12.	Ibid.
13.	The provision allowing HMIS funding from the Supportive Housing Program (SHP) is codified at 42 U.S.C. §11383(a)(7). HUD enumerated the ways in which CoCs may use SHP funds for management information systems in Federal Register, volume 69, no. 146, July 30, 2005, p. 45890.
14.	See Conference Committee, Department of Veterans Affairs and Housing and Urban Development and Independent Agencies Appropriations Act 2001, conference report to accompany H.R. 4635, H.Rept. 106-988, 106th Cong., 2nd sess., October 18, 2000.
15.	See Senate Committee on Appropriations, Department of Veterans Affairs and Housing and Urban Development, and Independent Agencies Appropriations Act 2001, report to accompany H.R. 4635, S.Rept. 106-410, 106th Cong., 2nd sess., September 13, 2000.
16.	U.S. Department of Housing and Urban Development, Report to Congress: HUD's Strategy for Homeless Data Collection, Analysis and Reporting, August 2001, p. 1, http://www.hud.gov/offices/cpd/homeless/hmis/strategy/congressreport.pdf.
17.	The website is http://www.hmis.info.
18.	Federal Register vol. 69, no. 146, July 30, 2004, pp. 45888-45934 (hereinafter, "HMIS Data and Technical Standards").
19.	Ibid., p. 45905.
20.	Ibid., pp. 45913-45914.
21.	Ibid., p. 45914.
22.	U.S. Department of Housing and Urban Development, Office of Community Planning and Development, Homeless Management Information System (HMIS) Data Standards: Revised Draft Notice, July 2009, p. 4, http://www.hudhre.info/documents/HMISDataStandards_July2009.pdf.
23.	HPRP was created as part of the American Recovery and Reinvestment Act (P.L. 111-5). For more information about HPRP, see CRS Report RL33764, The HUD Homeless Assistance Grants: Distribution of Funds, by [author name scrubbed].
24.	"HMIS Data and Technical Standards," pp. 45891-45892.
25.	U.S. Department of Housing and Urban Development, "The Violence Against Women and Department of Justice Reauthorization Act of 2005: Applicability to HUD Programs," Federal Register, vol. 72, no. 51, March 16, 2007, pp. 12695-12700.
26.	U.S. Department of Housing and Urban Development, Guidance on HPRP Subgrantee Data Collection and Reporting for Victim Service Providers, July 7, 2009, http://www.hudhre.info/documents/HPRPVictimServiceReportingGuidance.pdf.
27.	First AHAR, p. 1.
28.	U.S. Department of Housing and Urban Development, Report to Congress: Sixth Progress Report on HUD's Strategy for Homeless Data Collection, Reporting and Analysis, May 2007, p. 4, http://www.hud.gov/offices/cpd/homeless/library/improvingDataCollection.pdf (hereinafter, Sixth Progress Report to Congress).
29.	Ibid.
30.	Ibid.
31.	Ibid., p. 5.
32.	First AHAR, p. 13.
33.	Sixth Progress Report to Congress, p. 5.
34.	For an explanation of how annual counts are estimated using data from point-in-time counts, see Martha R. Burt and Carol Wilkens, Estimating the Need: Projecting from Point-in-Time to Annual Estimates of the Number of Homeless People in a Community and Using this Information to Plan for Permanent Supportive Housing, Corporation for Supportive Housing, March 2005, http://documents.csh.org/documents/pubs/csh_estimatingneed.pdf.
35.	U.S. Department of Housing and Urban Development, HUD's Homeless Assistance Programs: A Guide to Counting Unsheltered Homeless People, Second Revision, January 15, 2008, p. 14, http://www.hudhre.info/documents/counting_unsheltered.pdf (hereinafter, A Guide to Counting Unsheltered Homeless People).
36.	The Fourth AHAR is available at http://www.hudhre.info/documents/4thHomelessAssessmentReport.pdf.
37.	Fourth AHAR, p. 9 (see footnote 10).
38.	First AHAR, p. 17.
39.	Ibid.
40.	A Guide to Counting Unsheltered Homeless People (see footnote 35).
41.	First AHAR, p. 18.
42.	Ibid.
43.	See http://www.hudhre.info/index.cfm?do=viewHomelessRpts.
44.	Fourth AHAR, p. 9.
45.	Ibid., p. 16.
46.	CRS calculations based on Fourth AHAR, Appendix C, pp. C-6 to C-17. Each CoC count in which sheltered and/or unsheltered homeless persons was the same for 2007 and 2008 was added together and expressed as a percentage of the 2008 CoC point-in-time count taken from the Third Annual Homeless Assessment Report. See U.S. Department of Housing and Urban Development, Office of Community Planning and Development, Third Annual Homeless Assessment Report to Congress, July 2008, p. 10, http://www.hudhre.info/documents/3rdHomelessAssessmentReport.pdf (hereinafter, Third AHAR).
47.	Fourth AHAR, p. 9.
48.	Ibid., p. 8.
49.	The first AHAR is available at http://www.huduser.org/Publications/pdf/ahar.pdf.
50.	U.S. Department of Housing and Urban Development, Office of Community Planning and Development, The Second Annual Homeless Assessment Report to Congress, March 2008, http://www.hudhre.info/documents/2ndHomelessAssessmentReport.pdf (hereinafter, Second AHAR).
51.	Fourth AHAR, pp. B-14 to B-16. For each participating community, only the category data in which the bed coverage rate was equal to or greater than 50% was included.
52.	Second AHAR, pp. 4-5. As explained earlier in this report, the Violence Against Women and Department of Justice Reauthorization Act (P.L. 109-162) prevented domestic violence service providers from participating in HMIS. The first AHAR data collection period occurred prior to enactment of P.L. 109-162, and some of these providers were still participating in HMIS at that time.
53.	First AHAR., p. 59.
54.	Ibid., p. 11.
55.	Ibid., p. 13. For each participating community, only the category data in which the bed coverage rate was equal to or greater than 50% was included. It was not possible to create a sample based on HMIS implementation requirements that was also nationally representative. Ibid., p. 12.
56.	Ibid., p. 65.
57.	Ibid., p. 66.
58.	Ibid., pp. 67-68.
59.	Ibid., p. 69.
60.	Specifically, researchers used the square root of an area's population to determine that 27 sites would be central cities, eight would be from cities with populations greater than 50,000 that did not qualify as central cities, 13 would be from urban counties, and 14 from rural areas. Ibid.
61.	Ibid., p. 70.
62.	Ibid, p. 13. It should be noted that only 40 sample communities are reported as actually providing data for the AHAR, however. Of the 55 communities that had implemented HMIS, fifteen did not have any emergency shelters or transitional housing as of early 2005, and therefore no sheltered homeless population. See First AHAR, Appendix A, pp. 55-57. These sites were still included in the analysis "because they represent other communities with zero providers," First AHAR, Appendix B, pp. 73-74.
63.	Ibid., p. 13.
64.	Ibid., p. 70.
65.	Second AHAR, p. 4. This includes 14 sample communities that reported having no emergency shelters or transitional housing in their jurisdictions in early 2006. See Appendix A, pp. 47-49.
66.	Ibid., p. 4.
67.	Ibid., p. 5.
68.	Third AHAR, Appendix A, pp. A-1 to A-4. Thirteen participating sample communities did not have any emergency shelters or transitional housing in their jurisdictions in early 2007.
69.	Fourth AHAR, Appendix B, p. B-11.
70.	Ibid., p. B-13.
71.	Ibid., p. B-16.
72.	Ibid., p. B-15.
73.	The time period of the data collected between October 1, 2006, and September 30, 2007, is referred to as 2007, while the time period of the data collected between October 1, 2007, and September 30, 2008, is referred to as 2008.
74.	Fourth AHAR, p. 42. Percentage and numerical differences discussed in this section have not been tested for statistical significance, and were not included in the Fourth AHAR. Therefore, it remains to be shown which differences were least likely to have been produced by chance alone (due to sampling variability). Generally speaking, the larger the difference, the less likely it is for the difference to have been produced by chance alone.
75.	The data for 2007 and 2008 do not allow for separate estimates for suburbs and rural areas since many rural areas did not submit useable HMIS data to the AHAR.
76.	U.S. Department of Housing and Urban Development, A Report to the Secretary on the Homeless and Emergency Shelters, May 1984, pp. 8-19.
77.	Martha R. Burt and Barbara E. Cohen, America's Homeless: Numbers, Characteristics, and Programs that Serve Them (Washington, DC: The Urban Institute Press, July 1989).
78.	Ibid., p. 29. The range varies based on estimates of homeless individuals who did not use homeless services, and therefore were not counted.
79.	Ibid., p. 32.
80.	Martha R. Burt, Laudan Y. Aron, et al., Homelessness: Programs and the People They Serve: Findings of the National Survey of Homeless Assistance Providers and Clients, Technical Report, December 1999, Chapter 2, p. 2-1, http://www.huduser.org/publications/homeless/homeless_tech.html (hereinafter, "Homelessness: Programs and the People They Serve").
81.	Ibid., p. 1-7.
82.	Martha Burt and Laudan Y. Aron, America's Homeless II: Population and Services, The Urban Institute, February 1, 2000, http://www.urban.org/UploadedPDF/900344_AmericasHomelessII.pdf.
83.	The estimate for one week during the fall of 1996 was based on service usage by homeless individuals.
84.	The estimate for one week during winter of 1996 was based on service provider estimates.
85.	America's Homeless II: Population and Services.
86.	Homelessness: Programs and the People They Serve.
87.	Ibid., p. 3-4.
88.	Ibid., pp. 3-5 to 3-7.
89.	Ibid., p. 3-3.
90.	Ibid., pp. 8-3 to 8-8.
91.	Ibid., p. 10-2.
92.	Ibid., p. 10-10.
93.	Annetta C. Smith and Denise I. Smith, Emergency and Transitional Shelter Population: 2000, U.S. Census Bureau, October 2001, http://www.census.gov/prod/2001pubs/censr01-2.pdf.
94.	Ibid., p. 6.
95.	Ibid., p. 8.
96.	For the most recent U.S. Conference of Mayors report, see U.S. Conference of Mayors, Hunger and Homelessness Survey: A Status Report on Hunger and Homelessness in America's Cities, December 2008, http://www.usmayors.org/pressreleases/documents/hungerhomelessnessreport_121208.pdf (hereafter Hunger and Homelessness Survey).
97.	The cities surveyed were Boston, Charleston, Charlotte, Chicago, Cleveland, Dallas, Denver, Des Moines, Gastonia (NC), Kansas City (MO), Los Angeles, Louisville, Miami, Minneapolis, Nashville, Philadelphia, Phoenix, Portland (OR), Providence, Salt Lake City, San Francisco, Santa Monica, Seattle, St. Paul, and Trenton.
98.	Hunger and Homelessness Survey, p. 16.
99.	Ibid., p. 18.
100.	John H. Kuhn and John Nakashima, The Fifteenth Annual Progress Report on P.L. 105-114: Services for Homeless Veterans Assessment and Coordination, U.S. Department of Veterans Affairs, March 11, 2009, http://www1.va.gov/homeless/docs/CHALENG_15th_Annual_CHALENG_Report_FY2008.pdf.