Potential Analyses with Homelessness Data: Ideas for Policymakers and Researchers
Report to the Office of the Assistant Secretary for Planning and Evaluation (ASPE) andHealth Resources and Services Administration (HRSA)U.S. Department of Health and Human Services (HHS)
Prepared by: Michelle Wood and Jill Khadduri Abt Associates Inc.
This report is available on the Internet at:http://aspe.hhs.gov/hsp/09/HomelessnessData/PotentialAnalyses
- Participation in TANF and Medicaid by People Experiencing Homelessness and People At Risk of Homelessness
- Comparing the rate of homelessness among program applicants to the rate among all low income families and individuals
- Comparing the rate of homelessness among program participants to the rate among all low-income families and individuals
- Comparing risk factors for homelessness among program applicants and participants to risk factors among all low income families and individuals
- Measuring the rate at which sheltered homeless families and individuals participate in the TANF and Medicaid programs
- Do TANF and Medicaid Protect People from Becoming Homeless?
- Do People Experiencing Homelessness Use TANF and Medicaid in Different Ways from Other Low-income People?
In 2006, the U.S. Department of Health and Human Services (HHS) undertook a study to explore the extent to which states collect data on housing status and homelessness from applicants for Medicaid and/or Temporary Assistance for Needy Families (TANF), the two largest HHS mainstream programs that may serve individuals or families experiencing homelessness. The project complements Departmental efforts to increase access to HHS mainstream resources for persons experiencing homelessness.
The study found that thirty states currently collect information on homelessness or risk factors for homelessness from applicants for TANF or Medicaid. Abt Associates has developed some ideas for potential uses of this information for policymakers (especially at the state level) and researchers. This document summarizes ideas for data analysis that would help answer three sets of questions that should be of interest to policymakers and researchers:
Do families and individuals experiencing homelessness or at risk of homelessness participate in TANF and Medicaid to the same extent as other low-income families and individuals? Do the programs need to conduct special outreach to people with highly unstable housing situations?
Do participation in TANF and Medicaid help protect people from becoming homeless?
Do families and individuals experiencing homelessness use TANF and Medicaid in different ways from other program participants? Do homeless families have more difficulty complying with TANF requirements and using the program to transition to employment? Do homeless families and individuals use Medicaid-reimbursed services in inefficient ways?
Each of these issues can be analyzed to some extent through data on homelessness collected only at the time of application for TANF or Medicaid benefits, but such analysis is limited by the nature of homelessness. Many families and individuals experience only single, brief episodes of homelessness, and others are homeless repeatedly but with short episodes. Only a few individuals or families remain homeless for months or years. The time of application for TANF or Medicaid may coincide with a period of homelessness or housing instability or may not.
Altogether, 28 states collect homelessness indicators or risk factors from TANF applicants and 27 states do so for Medicaid applicants. (See Table 1 at the end of this document and also Table 3 from the final report). For 11 TANF programs and 12 Medicaid programs the application is the only time this information is collected and recorded. In these cases the usefulness of the data is somewhat limited since it is not possible to use such data to track whether housing status changes over time, or to assess the length of the episode of homelessness or the timing of the beginning of the homelessness episode relative to the application for TANF or Medicaid.
The ability to answer the questions is greater if states update the information on homelessness or risk factors at annual recertification or at other points of contact between participants and program staff, increasing the possibility that the data will capture a period of homelessness and making it possible to analyze patterns of homelessness over time in relation to patterns of use of the TANF and Medicaid programs.
As shown in Table 1, homelessness data are updated at annual recertifications for 17 TANF programs and 15 Medicaid programs, providing the opportunity to track homelessness among TANF and Medicaid participants after they enroll in the programs.
The best platform for analysis exists when states are willing to match data from TANF and Medicaid administrative data systems to data from Homeless Management Information Systems (HMIS) maintained either at the state level or for Continuums of Care covering particular cities or metropolitan areas within the state. HMIS data make it possible to determine whether a TANF or Medicaid participant was ever in the residential services system for people experiencing homelessnessbefore, during, or following a period of homelessness.
An HMIS is an electronic data collection system that stores person-level information about homeless persons who access assistance from the homeless assistance system. An HMIS allows service providers to collect information on homeless persons over time, providing detailed information on demographic characteristics, timing of episodes of homelessness, and service use patterns. HMIS can be used to analyze lengths of spells of homelessness, and types of homeless assistance received. HMIS typically do not cover entire states, so the data matching and analysis would need to be focused on specific communities where an HMIS is well developed. Based on HMIS data collection for the 2007 Annual Homeless Assessment Report (AHAR), it appears that most of the states that collect homelessness data from TANF or Medicaid applicants and update it at recertification, also had at least one community that provided HMIS data to the AHAR.
The first section of this document focuses on the question of participation in TANF and Medicaid by people experiencing homelessness. The next section looks at whether TANF and Medicaid are protecting people from becoming homeless. The third section turns to the issue of whether people experiencing homelessness are using TANF and Medicaid in ways that are different from other program participants.
To implement any of these ideas, states would need to assess the quality of available data. For example, most of the states that collect information on homelessness from applicants or participants in TANF or Medicaid reported to the studys survey that they do not routinely verify the information, nor have they attempted to use the information for analysis. HMIS systems are at various stages of development across states and communities within states. The data are populated for some Continuums of Care, while others remain at early stages of system development and implementation. This document does not provide guidance on steps the states would need to take to assess and improve data and to conduct these analyses. Any ideas offered here would need to be further developed by states before implementation. This document is simply intended to provide starting point suggestions for how these data could be used and their potential value to policymakers and researchers.
[ Go to Contents ]
Participation in TANF and Medicaid by People Experiencing Homelessness and People at Risk of Homelessness
Comparing the rate of homelessness among program applicants to the rate among all low income families and individuals
The estimates of homelessness available from application data could be used to compare the rate of homelessness among Medicaid or TANF applicants with the rate of homelessness at a single point in time among the states low income population in general. This comparison would use the data HUD requires Continuums of Care to collect every two years to estimate the number of people experiencing homelessness who are sheltered (from administrative data) and the number of people experiencing homelessness who are unsheltered (from street counts). Those point in time counts make it possible to estimate what percentage of the low-income population (or of members of low-income families) is homeless on a particular day. This rate could then be compared to the percentage of Medicaid or TANF applicants who are homeless at the time of application, which also is a single point in time.
If the rate reporting they are homeless on their TANF or Medicaid application is substantially lower than the rate at which low-income people in the Continuum of Care are homeless (for Medicaid) or the rate at which low-income families are homeless (for TANF), this might imply that mainstream programs are not reaching the homeless population and that better coordination between the homeless services and the mainstream system is needed. The state could repeat the analysis to measure results of efforts to increase access to mainstream benefits for people who are homeless.
Comparing the rate of homelessness among program participants to the rate among all low-income families and individuals
This would be a similar analysis, but instead of focusing on the rate of homelessness among TANF and Medicaid applicants it would focus on the rate among participants in the programs. This may be a more meaningful comparison, because a low rate of homelessness at the time of the application may simply mask the fact that many families or individuals were already participants in TANF or Medicaid before they became homeless, so they would not appear in data on new program applications. Data on the status of families or individuals at the time they are recertified would also be single day, point in time data, and therefore legitimate to compare with point in time estimates from the Continuum of Care.
Comparing risk factors for homelessness among program applicants and participants to risk factors among all low income families and individuals
Only one state (Idaho) collects data, not on the homelessness of program applicants, but on whether the applicant currently lives with friends and relatives, a potential risk factor for homelessness. Other states collect data on this item as well as presence of an eviction notice (another potential risk factor), but these states also collect data from applicants regarding homelessness. Data on risk factors alone are less useful for measuring the extent to which families or individuals at risk of homelessness are served by TANF and Medicaid, since a large percentage of at-risk individuals and families never actually become homeless.
Measuring the rate at which sheltered homeless families and individuals participate in the TANF and Medicaid programs
For states or communities within states that have functioning and well-populated HMIS systems, data matches between the HMIS and TANF and Medicaid administrative records can determine the extent to which people who were homeless at any time during the course of a year were served by the mainstream programs during their period of homelessness. HMIS data have entry and exit dates for the use of particular homeless programs (emergency shelters and transitional housing), and those dates can be used to establish a period of sheltered homelessness (a period during which the family or individual was either in a residential program or a repeat user of one or more programs). That period of homelessness can be compared to the same time period for TANF and Medicaid records. Periods of homelessness and of mainstream program participation could be measured using actual start and end dates or as whether homelessness and program participation occurred in the same month.
For the most part, HMIS data do not include information on people experiencing homelessness who are unsheltered. However, there is evidence that most homeless individuals are sheltered at some point, and homeless families and individuals are unlikely to spend much time on the street, so this data match would provide an excellent estimate of the extent to which TANF and Medicaid are serving people experiencing homelessness.
[ Go to Contents ]
States that collect information about homelessness both at application and at recertification could do longitudinal analysis of participant administrative records to determine 1) the extent to which participants homeless at application are still homeless in later years and 2) the extent to which participants not homeless at application become homeless later on. This analysis might suggest whether the mainstream programs are helping to protect people from becoming homeless or from protracted or repeat episodes of homelessness. Those states that collect data on risk factors could extend this analysis to determine the patterns of such risks relative to TANF or Medicaid participation.
For TANF, data permitting the analysis could also include whether post-application episodes of homelessness or housing instability occur after families have been sanctioned, are approaching, or reach their TANF time limit.
Data on type of Medicaid service used might be used to assess whether Medicaid protects even those families and individuals with severe medical crises from becoming or remaining homeless.
Using matches to HMIS data to determine how patterns of use of TANF and Medicaid relate to the timing of sheltered homelessness
States that can match TANF and Medicaid data to HMIS data could use the data match to analyze whether people experiencing homelessness already participated in the mainstream programs before they became homeless or whether starting to participate was associated with leaving homelessness. A recently-completed study of the costs of homelessness found that first-time homeless families tend to already be enrolled in Medicaid, whereas the rate of use of Medicaid for first-time homeless individuals was very low.
The data matches could also determine whether mainstream program participants are likely to remain homeless for shorter periods of time than those who do not participate.
The data matches could also be used to determine whether families participating in TANF are less likely to become homeless than non-participants. For those participating in TANF, the data match could examine how the pattern of homelessness relates to the pattern of TANF participation. For example:
- Are TANF recipients who reach the TANF time limits more likely than other recipients to be homeless at some point during their participation in the program?
Are TANF recipients who receive sanctions more or less likely than other participants to be homeless at some point?
Are TANF recipients who request exemption from work requirements more or less likely to be homeless at some point?
For Medicaid, such data matches could also be used to determine how the onset of homelessness is related to particular types of medical crises. This could help state policymakers plan health-based strategies for reducing homelessness.
[ Go to Contents ]
Do People Experiencing Homelessness Use TANF and Medicaid in Different Ways from Other Low-income People?
States that collect data on homelessness or risk factors only from application data could link that data to information on later use of TANF and Medicaid benefits to determine whether applicants who were homeless at the time of application have different program experiences (e.g., Medicaid utilization; TANF program participation) than other applicants. The information could be used to identify applicants who may face greater barriers to achieving program requirements than other program applicants. If families and individuals who are homeless or at risk of homelessness make different demands on the mainstream programs than their housed counterparts, this information could be used to identify families and individuals who may need special, targeted services or assistance.
TANF. For TANF, applicants who are homeless or at risk of homelessness may have more difficulty than others in meeting the TANF work requirements. These families might be considered harder to serve and in need of more intensive case management in order to make best use of TANF assistance and to achieve self-sufficiency. Program managers could use applicant data to determine which families should receive more targeted employment services or case management to address the barriers resulting from homelessness. Identifying families with more serious risks would therefore allow state programs to focus their services more effectively and to ensure that families with the greatest barriers are identified early.
Medicaid. In the case of Medicaid it is possible that individuals and families who are homeless may use Medicaid assistance less efficiently than other beneficiaries. For example, beneficiaries who are homeless may be more likely than others to seek routine health care services from hospital emergency rooms rather than primary care physicians, or to use in-patient rather than out-patient care, resulting in higher costs to the program and less efficient use of health care resources.
States that collect information on homelessness status at the time of Medicaid application could use this information to track utilization of Medicaid over time and compare utilization among beneficiaries who were homeless when they applied to those who were stably housed. This analysis would provide additional evidence to inform the debate on this topic. If such a pattern of Medicaid usage is observed in a state, the data on applicants could be used to target education and assistance to improve the utilization of Medicaid benefits. States could use data collected from applicants about housing status, homelessness, or risk factors, to identify individuals or families who might benefit from education about how to use Medicaid benefits effectively and assistance in seeking health care via the Medicaid program.
States with well-populated Homeless Management Information Systems could conduct more detailed analysis on how homelessness relates to patterns of Medicaid and TANF use.
Medicaid. For Medicaid, a match to HMIS data could be used to learn whether people who follow particular patterns of homelessnessfor example, single brief episodes or repeat episodes, or sustained periods of homelessnessare particularly likely to be heavy users of Medicaid or to use medical services in inefficient ways.
TANF. For TANF, a match to HMIS data could permit a state to determine further the way in which patterns of homelessness can help identify families at high risk of being sanctioned by TANF for failing to meet work requirements or exhausting their time limits.
[ Go to Contents ]
Data being collected by states regarding homelessness among TANF and Medicaid applicants and recipients may serve several purposes for policymakers and researchers, especially at the state level. The survey conducted for the study did not indicate that a great deal of analysis has been done with the data being collected by states. However, we believe that there are several options that could be considered that would provide researchers and policymakers with important information about how to target program resources and better understand how mainstream programs are serving individuals and families who experience homelessness.
[ Go to Contents ]
|State||Homelessness Indicators Included on Application||Homelessness Risk Factors Included on Application||Updated at Recertification?|
|Are you homeless?||Do you reside in a shelter?||Are you staying in a domestic violence shelter?||Do you have a permanent home?||Do you live with friends or relatives?||Do you have an eviction notice?|
|Louisiana (TANF only)||X||X||X|
|Massachusetts (Medicaid only)||X||X|
|Mississippi (Medicaid only)||X|
|Missouri (Combined only)||X||X||X|
|New Mexico (Combined only)||X||X||X|
|New York (Combined only)||X||X||X||X||X|
|Oregon (Combined only)||X||X|
|Rhode Island (Combined only)||X||X||X||X||X|
|South Carolina (TANF only)||X||X|
|Utah (Combined only)||X||X|
|Virginia (TANF only)||X||X||X||X|
|West Virginia (Combined only)||X||X|
|Source: 2008 interviews with TANF and Medicaid program officials.|
[ Go to Contents ]
 U.S. Department of Health and Human Services. 2009. Homelessness Data in Health and Human Services Mainstream Programs (available at http://aspe.hhs.gov/hsp/09/HomelessnessData/index.shtml).
 U.S. Department of Housing and Urban Development. 2008. The 2007 Annual Homeless Assessment Report (AHAR). Office of Community Planning and Development. Kuhn, R., & Culhane, D. P. (1998.) Applying cluster analysis to test a typology of homelessness by pattern of shelter utilization: Results from the analysis of administrative data. American Journal of Community Psychology, 26(2), 207-232. Culhane, Dennis P., Stephen Metraux, Jung Min Park, Maryanne Schretzman, and Jesse Valente. Testing a Typology of Family Homelessness Based on Patterns of Public Shelter Utilization in Four U.S. Jurisdictions: Implications for Policy and Program Planning, Housing Policy Debate 18(1), 2007, 128. Spellman, Brooke, Jill Khadduri, Brian Sokol, and Josh Leopold, Costs of First Time Homelessness for Families and Individuals. Cambridge MA: Abt Associates, Inc., forthcoming 2009.
 Homelessness information might be updated at other points of contact between TANF or Medicaid program staff and program participants but the survey conducted for the study asked specifically about recertification. If updates are recorded at other points, the same type of use could be made of the data.
 Since 1994, HUD has encouraged communities to address the problems of homelessness and housing in a coordinated, comprehensive, and strategic fashion through a Continuum of Care (CoC). The CoC is a community plan to organize and deliver housing and services to meet the specific needs of people who are homeless as they move to stable housing and maximum self-sufficiency (HUD, 2001, Guide to Continuum of Care Planning and Implementation). At a minimum, the CoC plan encompasses providers who receive McKinney-Vento homeless assistance funding. In many communities other key providersfor example, faith-based providers of emergency shelter and transitional housing that do not seek government fundingparticipate fully in the CoC.
 It is possible that states have developed informal methods of using the homelessness information to analyze program participation or inform policymaking but these were not reported to us in the survey.
 Spellman, Brooke, Jill Khadduri, Brian Sokol, and Josh Leopold, Costs of First Time Homelessness for Families and Individuals. Cambridge MA: Abt Associates, Inc., forthcoming 2009.