Social Security Administration Disability Programs and Individuals Facing Homelessness

by
Social Security Bulletin, Vol. 81 No. 2, 2021

This article examines the geographic, demographic, socioeconomic, and program-participation characteristics of initial Supplemental Security Income (SSI) and Social Security Disability Insurance (DI) applicants who faced homelessness during 2007–2017. Using Social Security Administration data, we chart the distribution of homeless SSI/DI applicants and beneficiaries across county-equivalent areas in the contiguous United States. We also use a text-mining method to identify 162,536 potentially homeless disability-program applicants, in addition to the 647,790 applicants identified using the standard homeless-status indicators in the administrative data. We find that homelessness among disability-program applicants was largely an urban phenomenon, with almost half (42.1 percent) of applicants living in one of 25 urban areas. Relative to their domiciled counterparts, homeless disability-program applicants were far more likely to be male, aged 18–64, and without a high school or general equivalency diploma. Allowance rates varied among studied applicants differentiated by program, mortality status, and primary impairment.


Joyce Nicholas is a social science research analyst in the Office of Retirement and Disability Policy (ORDP), Social Security Administration (SSA). When this article was written, Thomas Hale was an economist with ORDP, SSA.

Acknowledgments: The authors thank Susan Wilschke, Jeffrey Hemmeter, Joyanne Cobb, Chris Tamborini, Angela Hood, Michael Compson, Hilary Waldron, Bert Kestenbaum, and Sherria Green for their support and thoughtful comments on drafts of this article.

The findings and conclusions presented in the Bulletin are those of the authors and do not necessarily represent the views of the Social Security Administration.

Introduction

Selected Abbreviations
DDS Disability Determination Service
DI Disability Insurance
EDCS Electronic Disability Collect System
HUD Department of Housing and Urban Development
ISM in-kind support and maintenance
MSSICS Modernized Supplemental Security Income Claims System
SSA Social Security Administration
SSI Supplemental Security Income
USICH U.S. Interagency Council on Homelessness

This study provides new quantitative information about individuals who applied for Supplemental Security Income (SSI) and Social Security Disability Insurance (DI) disability benefits when they were experiencing or at risk of homelessness. The Social Security Administration (SSA) places great importance on identifying homeless disability-program applicants because their unmet housing needs, along with their health challenges, make it harder for them to navigate the application process. By understanding the geographic distribution of homeless disability-program applicants across SSA's service areas, and their demographic, socioeconomic, and program-participation characteristics, SSA can improve its efforts to ensure that homeless applicants receive needed supports.

For this article, we supplement structured data from SSA disability-benefit applicant intake forms with text mined from the “residential address” and “note” fields of those forms to identify individuals who were experiencing or at risk of homelessness.1 Our primary purpose is to provide an overview of the prevalence of homelessness among SSA's service population. We identify 810,326 individuals experiencing homelessness who submitted an initial SSI/DI disability-benefit application during the years 2007 through 2017 and had a medical decision made by a state Disability Determination Service (DDS) after September 2007.2 This study is the first to examine the distribution of homeless SSI/DI applicants and beneficiaries across county-equivalent areas in the contiguous United States.3

To compile our count of homeless disability-program applicants, we began by identifying the individuals who were recorded as experiencing homelessness in one of two ways in the administrative records. The first is the “homeless flag,” which an SSA field officer activates in the DI or SSI applicant's file to alert other SSA and DDS staff to follow the special case-processing procedures required in cases involving homelessness. The second is the “transient indicator,” which is attached to an SSI applicant's file for the same purpose as the homeless flag but is also used in postentitlement in-kind support and maintenance (ISM) evaluations.4 To the count of individuals identified by the homeless flag and the transient indicator, we added applicants that we identified as experiencing homelessness by mining the text in “residential address” and “administrative note” fields in those application files. With the text-mining experiment, this article explores whether SSA's processes and mechanisms for recording homeless and transient status potentially miss any disability-program applicants who face housing instability.

Background

In administering the SSI and DI programs, SSA provides income stability for individuals with disabilities who meet the program requirements and are experiencing homelessness. SSI and DI, in concert with other programs, can help individuals transition from homelessness toward stable and permanent housing. SSA promotes and seeks to improve collaboration with government and nonprofit stakeholders who serve individuals experiencing homelessness and can assist that population during both the initial SSI/DI application and the medical determination process.5 SSA is one of 19 agencies participating in the U.S. Interagency Council on Homelessness (USICH), which oversees and coordinates the federal response to homelessness.6 In addition to this study, SSA has conducted various data analyses to inform USICH efforts. For example, in 2014, SSA evaluated the outcomes of Social Security disability applications submitted through the Benefits Entitlement Services Team (BEST) demonstration project to determine if the project successfully increased access to SSI/DI benefits for individuals experiencing homelessness (Kennedy and King 2014).7 SSA also conducted and evaluated the Homeless with Schizophrenia Presumptive Disability pilot. The evaluation found that providing support during the application process for homeless individuals with a serious mental illness led to higher allowance rates at the initial adjudication level, fewer requests for consultative examinations, and reduced time to allowance (Bailey, Engler, and Hemmeter 2016).

SSA Disability Programs

The SSI program makes payments to individuals with a qualifying disability and limited income and resources; the DI program provides benefits to disabled workers who are insured (based on their earnings records) and, in some cases, to their eligible family members. Section 223 of the Social Security Act defines disability as “the inability to engage in any substantial gainful activity by reason of any medically determinable physical or mental impairment which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than 12 months.” For both programs, individuals must meet that definition of disability. The SSI program is means-tested; qualifying applicants must have income and assets below certain levels. To qualify for DI benefits, individuals must have accrued sufficient work credits based on their earnings histories.

The disability determination process begins when the individual applies for SSI, DI, or both and submits the application(s) to an SSA field office, where a staff member first verifies nonmedical eligibility by determining whether the applicant is engaged in substantial gainful activity, as indicated by an annually adjusted earnings threshold.8 If so, the field office denies the application; otherwise, the field office sends the case to a state DDS office.

In both programs, the DDS determines disability based on vocational and medical evidence from the applicant's medical or behavioral care providers or from a consultative examination—that is, a physical or mental examination or test purchased by SSA. If the DDS determines that the applicant is not disabled, the applicant may request reconsideration, in which the DDS thoroughly reexamines all evidence used in the initial determination and any additional evidence or information submitted with the reconsideration appeal. If the DDS denies the application at the reconsideration level, the claimant may request an appeal hearing before an administrative law judge (ALJ). If the claim is denied at the ALJ level, the applicant can then bring the case to the SSA Appeals Council; if the Council denies the claim or decides not to review the case, the applicant can appeal to federal district court.9

SSA Definitions of Housing Instability

SSA uses two definitions of housing instability in its disability programs. The first definition is the one that must be met to activate the homeless flag. It therefore applies to both the SSI and DI programs, and it has two components, one reflecting current status and the other reflecting prospective risk. SSA defines a disability-program applicant as “homeless” if he or she (1) does not have a fixed, regular, and adequate nighttime residence; or (2) is at risk of losing or is expected to lose his or her current accommodations within 14 days and will not have a fixed, regular, and adequate nighttime residence (SSA 2014a). SSA uses this definition to flag disability-program applications for special expedited processing so that individuals who face homelessness and meet the eligibility criteria can begin to receive stable income sooner. If an applicant meets this definition, SSA policy requires field office staff to activate the homeless flag manually in the agency's Electronic Disability Collect System (EDCS). Thus, we use the EDCS homeless flag to identify applicants meeting this first definition.

The second definition applies only to the SSI program. SSA defines an applicant as “transient” if he or she has no permanent living arrangement or fixed place of residence. A member of a household or a resident of an institution is not considered transient; an individual who is homeless, or who stays with a succession of friends or relatives with no permanent arrangement, is considered transient (SSA 2005). SSA operational policy instructs field office staff to apply a “transient” indicator in the Modernized SSI Claims System (MSSICS) to record SSI applicants and recipients experiencing current housing instability.10 SSA uses this information primarily to determine the applicant's living-arrangement category (which may affect SSI payment amounts) and to help account for ISM (such as food or shelter received from family or friends) at the time of application or, if SSI payments have begun, at the time of an ISM evaluation (Nicholas 2014). SSA operational policy also instructs field office staff to activate the EDCS homeless flag on any pending disability-program application for a claimant whose SSI living arrangement is flagged as transient. We use the MSSICS transient indicator to identify applicants meeting this second definition of homelessness.

For this study, we also apply a third definition of homelessness, which more closely aligns with the Department of Housing and Urban Development (HUD) definition adopted by USICH. That definition identifies an individual as chronically homeless if he or she can be diagnosed with a physical or mental disability, is (or was) without a home, and experienced housing instability for at least 12 months either consecutively or during at least four separate occasions within the last 3 years (HUD 2015). For this study, we use the HUD/USICH homeless definition, which we identify in SSA records via text mining, to detect members of the SSI/DI population who may be experiencing homelessness but do not have a homeless flag or transient indicator on their record.11 Specifically, for applicants either filing an SSI/DI claim or undergoing an SSI ISM evaluation, we search the content of the residential-address and administrative-note fields in their records for terms and phrases that reflect similarities with the USICH definition of homelessness.12 This approach is broad, but it represents a first step toward understanding whether the homeless flag and transient indicator alone might undercount the homeless population.

Data and Methods

We used administrative data available from four SSA sources as of August 16, 2017. First, we used the Disability Analysis Support Hub (DASH) for programmatic information and ZIP Codes for all initial SSI/DI applications transferred from an SSA field office to a state DDS where a medical decision occurred after September 2007. We also used the DASH to detect the use of the homeless flag and transient indicator, and to provide the address and note field contents needed to identify homeless applicants via text mining. Second, we used the 2017 release of the Disability Research File (DRF) to obtain a 10-year view of SSI/DI application and payment information.13 Third, we used the 2015 version of the Disability Analysis File to obtain 2015 earnings data and any additional or more current SSI and DI payment data. We concluded our analysis using death records available as of December 31, 2018 from the restricted-access Death Master File.

Several data limitations influenced the parameters of our study. First, the reference periods of available data sources permitted us to study only homeless individuals who submitted an initial application during calendar years 2007 through 2017 and had a medical decision rendered by a DDS after September 2007. Second, the limited availability of recent and accurate annual income data at the time of writing prevented us from examining earnings data for years since 2015.

Identification of Study Group

We applied the three methods of detecting homeless status to identify the subset of 2007–2017 disability-program applicants we sought to include in our study. We selected DI applicants who had a homeless flag or text in the residential-address or administrative-note field indicating that they were homeless when they filed their application.14 We chose SSI applicants who had a homeless flag, transient indicator, or text in the residential-address or administrative-note field specifying that they were homeless at the time of either an SSI application or a subsequent ISM evaluation.

Many applicants experiencing homelessness may not complete the SSI/DI application process or may have their applications denied because they lack supporting documentation of medical impairments (Bailey, Engler, and Hemmeter 2016). As such, many individuals apply for benefits multiple times. To support a person-level analysis and to avoid double counting, we limited our study to the administrative records for only the most recent application of each homeless disability-program applicant whose initial application was received by SSA during 2007–2017. We examined data from the last application filed before an allowance or denial decision in which an SSA staff member identified the applicant as homeless.15,16 Likewise, to avoid double counting members of our comparison group of domiciled disability-program applicants, we applied the same selection criteria and methodology.

The study group is composed of 810,326 individuals, hereafter referred to as “homeless disability applicants.” Of these, we identified 64,264 cases (7.9 percent) with an EDCS homeless flag but no transient indicator or text-mining results indicating homelessness; 339,697 cases (41.9 percent) with an MSSICS transient indicator but no homeless flag or text-mining results indicating homelessness; and 162,536 cases (20.1 percent) of homelessness indicated by only the text-mining method (Table 1 and Chart 1). We also identified 181,496 applicants (22.4 percent) by any combination of two of the three methods, as well as 62,333 applicants (7.7 percent) whose file met all three definitions. The 162,536 applicants who were identified by text mining alone—20.1 percent of the total group—would have been excluded from the study if we had used only the SSI transient indicator and SSI/DI homeless flag to identify homeless disability applicants. This outcome confirms that our text-mining method, using the HUD definition of homelessness, greatly increases the number of disability applicants identified as experiencing homelessness.

Table 1. Disability-program applicants experiencing homelessness, by method of identifying homeless status, 2007–2017
Measure Number Percent
Total 810,326 100.0
One method only 566,497 69.9
EDCS homeless flag 64,264 7.9
MSSICS transient indicator 339,697 41.9
Text mining 162,536 20.1
Two methods 181,496 22.4
EDCS homeless flag and MSSICS transient indicator 73,422 9.1
EDCS homeless flag and text mining 38,815 4.8
MSSICS transient indicator and text mining 69,259 8.5
All three methods 62,333 7.7
SOURCE: Authors' calculations using administrative data from SSA.
Chart 1. Disability-program applicants experiencing homelessness, by method of identifying homeless status, 2007 through 2017. Venn diagram depicting set relationships and values listed in Table 1.

Before conducting our geospatial analysis, we assessed how frequently field office staff applied the EDCS homeless flag and the MSSICS transient indicator. The activation of the MSSICS transient indicator requires the activation of the EDCS homeless flag but only at the time an active SSI application is available for expedited processing (SSA 2005, 2014b, 2014c). As a result, for SSI allowances, we were unable to determine whether SSA staff had applied the MSSICS transient indicator at the time of application or during a postentitlement ISM evaluation. Therefore, we assessed the use of the EDCS homeless flag and the MSSICS transient indicator by focusing on SSI denials because it is certain that field office staff applied the MSSICS transient indicator for this subgroup only at the time of application, and not for a postentitlement ISM evaluation.

About one-quarter (25.6 percent) of denied SSI applications had neither an EDCS homeless flag nor an MSSICS transient indicator; we identified the applicants as homeless using text mining (Table 2). Another 13.7 percent of denied SSI applications were identified with only an EDCS homeless flag. The remaining 60.7 percent of SSI denials had an MSSICS transient indicator; and although this entire subgroup should have had an EDCS homeless flag activated as well, only about one out of four had one.

Table 2. Denied disability-benefit applications from individuals experiencing homelessness, by program and method of identifying homeless status, 2007–2017
Method Total DI only SSI a
Number Percent Number Percent Number Percent
All 439,422 100.0 27,507 100.0 411,915 100.0
EDCS homeless flag only 62,003 14.1 5,375 19.5 56,628 13.7
MSSICS transient indicator only 194,326 44.2 6,022 21.9 188,304 45.7
Both homeless flag and transient indicator 62,361 14.2 586 2.1 61,775 15.0
Text mining only 120,732 27.5 15,524 56.5 105,208 25.6
SOURCE: Authors' calculations using administrative data from SSA.
a. Includes individuals who applied for concurrent SSI and DI benefits.

Among all 439,422 denied SSI/DI applications, we found that only 28.3 percent had an EDCS homeless flag activated for them and received expedited processing of their disability claims because of homelessness; however, there are many other reasons for which SSA may flag claims for expedited processing.17 Despite the limits of our study data, the analysis of denied applications begins to illuminate how frequently field office staff use the homeless flag and transient indicator for individuals facing disability and homelessness.

Geospatial Analysis

Although our study group consists of 810,326 homeless disability applicants in SSA's entire domestic service area, we focused our geospatial analysis on applicants in the 48 contiguous states. We anchored our geospatial analysis on the ZIP Codes of homeless disability applicants with a mailing address. We used Public Use Microdata Areas, developed for the Census Bureau's American Community Survey, to provide the conversion factors needed to generate county-based statistics from ZIP Code–level data. Our study covers 2,274 county-equivalent areas across the lower 48 states.18,19 Of the full study group, about 82.6 percent (669,298) had ZIP Code data indicating residence in the lower 48 states.20 Another 7.6 percent had ZIP Code data indicating residence in Alaska, Hawaii, or U.S. territories such as Guam and Puerto Rico. The remaining 9.8 percent of studied homeless disability applicants had no recorded ZIP Code.

Findings

We present our findings from three perspectives. First, we examine the geographic distribution of the homeless disability applicants. Second, we consider their demographic and socioeconomic characteristics. Third, we look at the differences (or similarities) between DI and SSI homeless disability applicants.

Geospatial Distribution of Homeless Disability Applicants

Charts 2–4 are maps of the contiguous United States respectively showing homeless disability applicants per 50,000 residents, homeless disability beneficiaries per 50,000 residents, and the 25 metropolitan areas21 with the highest numbers of homeless disability applicants, all for the period 2007–2017. These maps provide four main takeaways. First, across the lower 48 states, the most prominent clusters of homeless disability applicants appeared along the West Coast and the northeastern Interstate 95 corridor, in the Great Lakes region, and in Florida (Chart 2). Second, the geographic distributions of homeless disability applicants and beneficiaries were similar, based on a visual comparison of Charts 2 and 3. Third, most clusters of homeless disability applicants occurred in urban counties with at least 50,000 residents; about 9.8 percent of homeless disability applicants lived in either the Los Angeles or the New York City metropolitan area and an additional 32.3 percent lived in 23 other urban areas (Chart 4 and Table 3). Fourth, less than 1 percent of homeless disability applicants resided in a band of counties in the central states running continuously from North Dakota through western Texas (Chart 2). Our geospatial analysis revealed that 98 percent of our study group in the lower 48 states resided in county-equivalent areas with at least 50,000 inhabitants and that homelessness among disability applicants is largely an urban phenomenon. This finding is consistent with HUD's point-in-time estimates of the population experiencing homelessness, which indicate that California and New York have the largest numbers of homeless individuals (driven by Los Angeles and New York City), followed by Florida (HUD 2017). By contrast, the share of the entire U.S. population that lived in urban areas at the end of our study period was 80 percent (Census Bureau 2017).

Chart 2. Homeless disability applicants per 50,000 residents, by county-equivalent area, 2007 through 2017. Population density map summarized in the narrative. SOURCE: Authors' calculations using administrative data from S S A.
Chart 3. Homeless disability beneficiaries per 50,000 residents, by county-equivalent area, 2007 through 2017. Population density map summarized in the narrative. SOURCE: Authors' calculations using administrative data from S S A.
Chart 4. Twenty-five core-based statistical areas with the most homeless disability applicants, 2007 through 2017. Bubble map depicting relative sizes of populations listed in Table 3.
Table 3. Twenty-five core-based statistical areas ranked by largest homeless disability applicant population in the period 2007–2017
Rank and core-based statistical area Airport code identifier Homeless disability applicants
Number As a percentage of study group a
 1. Los Angeles-Long Beach-Anaheim, CA LAX 46,135 5.7
 2. New York-Newark-Jersey City, NY-NJ-PA NYC 33,525 4.1
 3. Boston-Cambridge-Newton, MA-NH BOS 22,446 2.8
 4. Miami-Fort Lauderdale-West Palm Beach, FL MIA 18,420 2.3
 5. Chicago-Naperville-Elgin, IL-IN-WI ORD 15,769 1.9
 6. San Francisco-Oakland-Hayward, CA SFO 15,677 1.9
 7. Seattle-Tacoma-Bellevue, WA SEA 15,228 1.9
 8. Baltimore-Columbia-Towson, MD BWI 14,905 1.8
 9. Washington-Arlington-Alexandria, DC-VA-MD-WV WAS 14,489 1.8
10. Atlanta-Sandy Springs-Roswell, GA ATL 12,611 1.6
11. Dallas-Fort Worth-Arlington, TX DFW 12,454 1.5
12. Philadelphia-Camden-Wilmington, PA-NJ-DE-MD PHL 11,906 1.5
13. Detroit-Warren-Dearborn, MI DTW 10,821 1.3
14. Sacramento-Roseville-Arden-Arcade, CA SMF 9,765 1.2
15. Denver-Aurora-Lakewood, CO DEN 9,714 1.2
16. San Diego-Carlsbad, CA SAN 9,083 1.1
17. Riverside-San Bernardino-Ontario, CA RAL 8,772 1.1
18. Cincinnati, OH-KY-IN LUK 8,352 1.0
19. Phoenix-Mesa-Scottsdale, AZ PHX 8,251 1.0
20. Providence-Warwick, RI-MA PVD 8,231 1.0
21. Tampa-St. Petersburg-Clearwater, FL TPA 8,036 1.0
22. Houston-The Woodlands-Sugar Land, TX IAH 7,458 0.9
23. Portland-Vancouver-Hillsboro, OR-WA PDX 7,145 0.9
24. Columbus, OH CMH 6,285 0.8
25. Minneapolis-St. Paul-Bloomington, MN-WI MSP 6,001 0.7
Top 25 combined . . . 341,479 42.1
Total United States . . . 810,326 100.0
SOURCE: Authors' calculations using administrative data from SSA and Office of Management and Budget.
NOTE: . . . = not applicable.
a. "Study group" comprises the total SSA domestic service area rather than only the contiguous United States.

Demographic and Socioeconomic Characteristics

Relative to domiciled disability applicants (that is, those not identified as homeless), homeless disability applicants were more likely to be men, of working age (18–64), and without a high school diploma or equivalent (Table 4). They were also more likely to have died as of December 31, 2018.

Among the homeless disability applicants, 47,178 (5.8 percent) worked during 2015.22 Some earnings-related statistics, not shown in Table 4, provide interesting perspectives on the applicants we identify as homeless. For example, those who worked had median annual earnings of $3,261. Furthermore, those whose applications were denied had median earnings that nearly doubled those of applicants who were allowed benefits ($5,273 versus $2,724). Surprisingly, earners in our study sample had an allowance rate of 72.6 percent, while nonearners had an allowance rate of 44.1 percent. The reasons we see higher allowance rates for earners than for nonearners are unclear.

Table 4. Selected characteristics of homeless and domiciled individuals who applied for disability-program benefits during the period 2007–2017
Characteristic Homeless Domiciled
Number Percent Number Percent
Total 810,326 100.0 21,648,926 100.0
Sex
Male 550,335 67.9 11,505,359 53.1
Female 259,991 32.1 10,143,568 46.9
Age
0–17 13,775 1.7 2,879,307 13.3
18–64 750,362 92.6 16,215,046 74.9
65 or older 46,189 5.7 2,554,573 11.8
Educational attainment
No high school diploma or equivalent 280,065 34.6 4,816,913 22.3
High school diploma or equivalent 356,614 44.0 8,093,621 37.4
Some college 112,010 13.8 3,225,254 14.9
College graduate 34,570 4.3 1,527,840 7.1
Missing data 27,067 3.3 3,985,299 18.4
Earnings status in 2015
Yes 47,178 5.8 a a
No 763,148 94.2 a a
Vital status on December 31, 2018
Living 705,908 87.1 19,195,496 88.7
Deceased 104,418 12.9 2,453,430 11.3
SOURCE: Authors' calculations using administrative data from SSA.
NOTES: Rounded components of percentage distributions do not necessarily sum to 100.0.
a. We did not obtain earnings data for domiciled disability applicants.

As of December 31, 2018, the respective death rates of homeless and domiciled individuals who had applied for disability benefits in the period 2007–2017 were 12.9 percent and 11.3 percent; this difference is statistically significant, with a p-value of less than 0.01. To account for age differences between the groups, we also analyzed death rates by age group. We found that the age-normalized death rates likewise were higher for homeless disability applicants than for their domiciled counterparts. These findings are consistent with those in social science and medical literature (O'Connell 2005).

Males and individuals with physical primary impairments were overrepresented among the homeless disability applicants who had died by year-end 2018 (not shown). Males constituted 77.3 percent and 66.5 percent of deceased and living applicants, respectively. Yet the characteristic with the largest difference between the percentages of deceased and living disability applicants is the physical primary impairment (74.1 percent versus 55.6 percent). No statistically significant differences in educational attainment existed among homeless disability applicants, living or dead.

Program Type

In this section, we examine the SSA disability programs from which homeless applicants sought benefits (SSI, DI, or SSI and DI concurrently). Table 5 shows that more applicants sought only SSI payments (31.2 percent of all homeless disability applicants) than only DI benefits (5.2 percent). The remaining 63.6 percent of the study subjects claimed concurrent SSI and DI benefits on their application records.

Table 5. Selected characteristics of individuals experiencing homelessness who applied for disability-program benefits during the period 2007–2017, with distributions by program
Characteristic All Program
DI only SSI only Concurrent DI and SSI
Number Percent Number Percent Number Percent Number Percent
Total 810,326 100.0 41,698 5.2 252,855 31.2 515,773 63.6
Sex
Male 550,335 100.0 25,764 4.7 174,862 31.8 349,709 63.5
Female 259,991 100.0 15,934 6.1 77,993 30.0 166,064 63.9
Educational attainment
No high school diploma or equivalent 280,065 100.0 7,565 2.7 108,789 38.8 163,711 58.5
High school diploma or equivalent 356,614 100.0 18,666 5.2 95,746 26.8 242,202 67.9
Some college 112,010 100.0 9,974 8.9 21,917 19.6 80,119 71.5
College graduate 34,570 100.0 4,819 13.9 5,563 16.1 24,188 70.0
Missing data 27,067 100.0 674 2.5 20,840 77.0 5,553 20.5
Earnings status in 2015
Yes 47,178 100.0 2,638 5.6 12,424 26.3 32,116 68.1
No 763,148 100.0 39,060 5.1 240,431 31.5 483,657 63.4
Primary impairment type
Mental or cognitive 340,266 100.0 14,058 4.1 116,503 34.2 209,705 61.6
Physical 470,060 100.0 27,640 5.9 136,352 29.0 306,068 65.1
SOURCE: Authors' calculations using administrative data from SSA.
NOTES: Rounded components of percentage distributions do not necessarily sum to 100.0.

Table 6 shows that homeless disability applicants were more likely to have a physical condition than a mental or cognitive one recorded as their primary impairment (58.0 percent versus 42.0 percent).23 For applicants with a physical impairment, the death rate was slightly more than double that of applicants with a mental/cognitive impairment (16.4 percent versus 8.0 percent).

Table 6. Selected characteristics of individuals experiencing homelessness who applied for disability-program benefits during the period 2007–2017, with distributions by type of primary impairment
Characteristic All Primary impairment type
Mental or cognitive Physical
Number Percent Number Percent Number Percent
Total 810,326 100.0 340,266 42.0 470,060 58.0
Sex
Male 550,335 100.0 218,378 39.7 331,957 60.3
Female 259,991 100.0 121,888 46.9 138,103 53.1
Educational attainment
No high school diploma or equivalent 280,065 100.0 124,702 44.5 155,363 55.5
High school diploma or equivalent 356,614 100.0 144,906 40.6 211,708 59.4
Some college 112,010 100.0 42,411 37.9 69,599 62.1
College graduate 34,570 100.0 14,450 41.8 20,120 58.2
Missing data 27,067 100.0 13,797 51.0 13,270 49.0
Program
DI only 41,698 100.0 14,058 33.7 27,640 66.3
SSI only 252,855 100.0 116,503 46.1 136,352 53.9
Concurrent DI and SSI 515,773 100.0 209,705 40.7 306,068 59.3
Died as of December 31, 2018
Number 104,418 27,094 77,324
Death rate 12.9 8.0 16.4
SOURCE: Authors' calculations using administrative data from SSA.

The allowance rate for homeless disability applicants overall was 45.8 percent (Table 7). Of the applicant subgroups, DI-only applicants had the lowest allowance rate of 34.0 percent, while those filing only an SSI claim had an allowance rate of 46.8 percent. Applicants with a physical primary impairment had an allowance rate of 41.4 percent while those with a mental or cognitive primary impairment had an allowance rate of 51.8 percent.

Table 7. Selected characteristics of individuals experiencing homelessness who applied for disability-program benefits during the period 2007–2017, with distributions by application outcome
Characteristic All Application decision
Not allowed a Allowed
Number Percent Number Percent Number Percent
Total 810,326 100.0 439,422 54.2 370,904 45.8
Sex
Male 550,335 100.0 290,684 52.8 259,651 47.2
Female 259,991 100.0 148,738 57.2 111,253 42.8
Educational attainment
No high school diploma or equivalent 280,065 100.0 149,989 53.6 130,076 46.4
High school diploma or equivalent 356,614 100.0 198,025 55.5 158,589 44.5
Some college 112,010 100.0 61,784 55.2 50,226 44.8
College graduate 34,570 100.0 17,389 50.3 17,181 49.7
Missing data 27,067 100.0 12,235 45.2 14,832 54.8
Program
DI only 41,698 100.0 27,511 66.0 14,187 34.0
SSI only 252,855 100.0 134,643 53.2 118,212 46.8
Concurrent DI and SSI 515,773 100.0 277,268 53.8 238,505 46.2
Earnings status in 2015
Yes 47,178 100.0 12,906 27.4 34,272 72.6
No 763,148 100.0 426,516 55.9 336,632 44.1
Vital status on December 31, 2018
Living 705,908 100.0 402,036 55.8 303,872 42.2
Deceased 104,418 100.0 37,386 35.8 67,032 64.2
Primary impairment type
Mental or cognitive 340,266 100.0 164,128 48.2 176,138 51.8
Physical 470,060 100.0 275,294 58.6 194,766 41.4
SOURCE: Authors' calculations using administrative data from SSA.
a. Denied or decision pending.

Among the homeless disability applicant subgroups, one of the highest allowance rates was for those who died after their DDS decision (64.2 percent). This outcome might be attributed to an SSA initiative to expedite processing for certain applications by flagging them as terminal illness (or TERI) cases. SSA and DDS staff expedite the SSI/DI claims of homeless disability applicants who have a terminal illness at the initial step of the disability determination process (Rajnes 2012). In our study, the highest observed allowance rates were for those who died after they began receiving benefits and who belonged to diagnostic groups involving many of the descriptors used by SSA and DDS staff to identify a potential TERI case. Examples of diagnostic groups common in TERI cases include various types of malignant neoplasms such as cancers of the esophagus or liver. Subsequently deceased homeless disability applicants with primary impairments involving neoplasms or diseases of the digestive system had allowance rates of 95.4 percent and 76.6 percent, respectively (not shown). However, given the high number of homeless disability applicants with a mental/cognitive or musculoskeletal impairment, not all who died were TERI cases.

Many homeless disability beneficiaries may be unable to manage their SSI and DI payments. SSA appointed a representative payee to manage the SSI/DI payments received by 24.4 percent of homeless disability beneficiaries in our study (not shown).24 In comparison, during December 2019, an estimated 18.6 percent of all working-age DI/SSI disability beneficiaries had a representative payee who helped them manage their program payments (SSA 2020a, 2020b). Of the homeless disability beneficiaries we identified as having a representative payee, half had their SSI and DI benefits managed by a natural or adoptive parent or an authorized social service agency or custodial institution. Finally, the majority (69.1 percent) of homeless disability beneficiaries with a payee at any point during program participation had a mental or cognitive condition rather than a physical one recorded as their primary impairment, consistent with needing assistance in managing one's benefits.

Summary and Conclusions

This study provides new insights about the SSI and DI programs in the context of homelessness as well as new statistics about the geographic, demographic, socioeconomic, and program-participation characteristics of homeless disability-program applicants. Some highlights of our findings follow.

Homelessness among disability-program applicants was largely an urban phenomenon, involving individuals living within concentrated areas in the contiguous United States. At least 98 percent of homeless SSI/DI applicants in the lower 48 states resided in urban counties, in contrast with 80 percent of the general population (Census Bureau 2017). About 42.1 percent of homeless disability applicants lived in one of 25 urban areas (Chart 4 and Table 3).

Several demographic subgroups were overrepresented among the study group. Relative to their domiciled counterparts, homeless disability applicants were far more likely to be male, aged 18 to 64, and without a high school diploma or equivalent (Table 4).

Allowance rates varied by program, postdecision mortality rate, and primary impairment. The overall allowance rate of homeless disability applicants was 45.8 percent (Table 7). Those who applied for only DI had one of the lowest allowance rates (34.0 percent) of any applicant subgroup while those applying for only SSI had an allowance rate of 46.8 percent. Applicants who subsequently died had one of the highest allowance rates, at 64.2 percent. SSA was more likely to allow SSI/DI benefits for applicants with a mental or cognitive primary impairment than for those with a physical condition (51.8 percent versus 41.4 percent).

Not all homeless disability applicants had an EDCS homeless flag on their files to prompt expedited processing of their applications. We examined the activation of the EDCS homeless flag (along with the MSSICS transient indicator) by focusing on SSI disability applicants facing housing instability whose claims were denied. Under SSA operational policy, field office staff can activate the MSSICS transient indicator only at the time of application and are required to activate the EDCS homeless flag for every applicant with an activated MSSICS transient indicator. Only 28.3 percent of files for denied SSI/DI applications had an EDCS homeless flag activated (with or without an MSSICS transient indicator) and thereby received expedited processing of their disability claim (Table 2). Future studies should explore the specific situations of these cases to determine whether certain circumstances that we did not observe precluded the need for the homeless flag.

Finally, a significant share of our study sample would not have been identified as homeless if we had relied on only the EDCS homeless flag and the MSSICS transient indicator. About 20 percent of our study group (162,536 claimants) would not have been included in this research if we had used only the homeless flag and transient indicator to identify those experiencing or at risk of homelessness (Chart 1). The application of a text-mining approach, informed by the HUD definition of homelessness, provides additional insight about the subset of disability-program applicants who may be experiencing or at risk of homelessness. Although additional research is needed to validate the current analysis or improve the methods used here, text mining could be a way to identify individuals facing disability and housing instability to ensure that they receive appropriate supports and assistance during the application process.

Appendix A: Text-Mining Search Terms and Phrases

Listed below are the text-mining search terms and phrases we used to identify SSI/DI applicants experiencing homelessness. We searched the residential-address and administrative-note fields of the claimants' files (including those for SSI ISM evaluations) to detect any of the listed terms, which we selected because they align with the HUD/USICH definition of homelessness. We began building the list with a set of search terms and phrases generated by SSA researchers who attempted to identify 2009–2011 disability-program claimants who were experiencing homelessness. Then, we checked and augmented the initial list of search terms and phrases by comparing them with those appearing in the residential-address and administrative-note fields of the files for 6,941 individuals belonging to the treatment group of the Homeless Outreach Projects and Evaluation (HOPE) demonstration from January 2005 through April 2007 (McCoy and others 2007). The HOPE demonstration had targeted chronically homeless individuals who applied for DI and SSI benefits to participate in the project. Validating our terms against the HOPE list further assured the appropriateness of the terms we included; however, we acknowledge that further validation—including checking for false positives—would be necessary prior to any operationalization of this method to inform new policy or service delivery practices.

Search terms and phrases (70 items)

abandoned

airport

angels

angels watch

angel's watch

bench

bus

cardboard box

camping

car

catholic charities

clinic

coalition

corr fac

corr facility

correc

correction

correctional

couch

double(d) (up)

empty

field office

forest

garage

general delivery

homeless

hotel

inn

live(s) with (friend or parent or relative or neighbor…)

metro

mission

motel

motor lodge

no address

no permanent

no place to live

park

park bench

pathfinder

rescue

residing with

salvation army

shelter

skid row

sofa

SSA FO

station

stay with

staying with

staying with friends

street

temp(orary) housing

tent

tent off

train

transcient

transient

transition housing

truck

under the bridge

undomicile

undomiciled

undomociled

vacant

vacant home

van

vehicle

woods

YMCA

YWCA

Appendix B: Mapping Methods

We executed seven steps before creating the density maps (Charts 2 and 3) of the contiguous United States. First, we assumed that homeless disability applicants who had a recorded ZIP Code were dispersed among the component counties by the same proportions with which the ZIP Code's land area fell within those counties. (The majority of homeless disability applicants in the lower 48 states had a recorded ZIP Code that was contained within a single county.) Second, we summed the number of homeless disability applicants living within each county-equivalent area of the lower 48 states. Third, we extracted Census county-resident counts and divided them by 50,000, the minimum number of residents living in an urban county (Missouri Census Data Center 2016). Fourth, we divided county homeless disability-applicant counts by the factor resulting from our third step to compute homeless disability applicants per 50,000 county residents. Fifth, we sorted county-level records in ascending order of homeless disability applicants per 50,000 residents. Sixth, we divided the records into quintiles and identified the minimum and maximum values for each quintile. Finally, we used those values to assign each county-equivalent into a quintile or density category, shown in Chart 2. We then replicated this procedure for homeless disability beneficiaries, shown in Chart 3.

Notes

1 Hereafter, our use of the term “experiencing homelessness” should be taken to include individuals at risk of, but not necessarily currently experiencing, homelessness.

2 We use the acronym “SSI/DI” to refer collectively to three types of disability-program participation: (1) SSI only, (2) DI only, and (3) concurrent SSI and DI.

3 The contiguous United States includes the lower 48 continental states, and excludes Alaska, Hawaii, and U.S. territories (Census Bureau 2013).

4 For information on how SSA uses the homeless flag and the transient indicator, see SSA (2014a) and SSA (2005), respectively.

5 Homeless-service stakeholders include providers of health care, behavioral health, and social services, as well as faith- and community-based organizations and partners. One example of collaboration is SSA's participation in the Substance Abuse and Mental Health Services Administration's SSI/DI Outreach, Access, and Recovery (SOAR) program. SOAR aims to increase access to SSA disability-program benefits for eligible children and adults who are experiencing or at risk of homelessness and have a serious mental illness, medical impairment, and/or substance-use disorder (see https://www.samhsa.gov/homelessness-programs-resources/grant-programs-services/soar).

6 The McKinney-Vento Homeless Assistance Act of 1987 established USICH.

7 Kennedy and King found that BEST contributed to increased access to disability benefits for applicants. Relative to other disability cases, the BEST cases had high allowance rates and short processing times.

8 In 2015, for example, SSA considered substantial gainful activity to be indicated by monthly earnings of at least $1,090 for a nonblind individual and at least $1,820 for a blind individual.

9 For detailed information on SSA's sequential disability determination process, see Wixon and Strand (2013).

10 When using non-MSSICS paper records, SSA field office staff note transience in the remarks field.

11 SSA field office staff record homeless status only at the time of submission of a disability-program application or, in the case of SSI, a recipient's most recent ISM evaluation. Because the SSA definition of “homeless” focuses on housing status at the time of application, disability-program staff are not required to follow up with applicants recorded as homeless or transient to determine the severity or duration of their housing instability (or to check whether domiciled applicants later become homeless).

12 Appendix A lists all search terms and phrases used to inform our text-mining method for selecting study members. The residential-address and administrative-note fields are associated with application forms SSA-3368 (for DI) and SSA-8000BK (for SSI).

13 SSA restricts DRF adjudicative data to the first three levels of the SSI/DI disability determination process (initial DDS decision, DDS reconsideration, and administrative law judge hearing) because of data-reporting issues associated with the higher adjudicative levels.

14 Following guidelines in SSA (2006), we included individuals who faced housing instability and met the requirements for Old-Age and Survivor's Insurance (OASI) benefits under the assumption that they had converted from DI to OASI on reaching their full retirement age (or age 55, if they were blind).

15 Examining the last application indicating homelessness may bias allowance rates upward because the likelihood of being allowed benefits increases with the number of applications submitted. However, we used the most recent application because it is more likely to reflect current information for homeless SSI/DI disability applicants.

16 Nearly 28 percent of studied homeless disability applicants submitted multiple disability-program applications and had homelessness indicated on at least one.

17 Quick Disability Determination, Compassionate Allowance, Terminal Illness, Wounded Warrior, and other flags may likewise expedite handling.

18 Appendix B details the methodology of our geospatial analysis.

19 The federal government describes noncounty administrative or statistical areas that are comparable to counties as “county equivalents” (Census Bureau 2013). Louisiana parishes; the organized boroughs of Alaska and New York City; the District of Columbia; and the independent cities of the states of Virginia, Maryland, Missouri, and Nevada are equivalent to counties for administrative purposes.

20 Among the 2,274 county equivalents in the lower 48 states with homeless disability applicants, about 34.7 percent had no more than one applicant per 50,000 residents and 28.3 percent had at least 50 applicants per 50,000 residents. About 40.9 percent of county equivalents with homeless disability applicants had no more than one beneficiary per 50,000 residents and 13.5 percent had at least 50 beneficiaries per 50,000 residents.

21 We used the Office of Management and Budget core-based statistical areas to define the metropolitan areas.

22 For this study, we did not access the earnings data of the 21,648,926 individuals who were domiciled and who submitted at least one disability application from calendar year 2007 through 2017.

23 SSA statistical publications provide statistics by diagnostic group for beneficiaries but not for applicants. The rate of mental/cognitive primary impairments we found among our sample of homeless disability-program beneficiaries (47 percent; not shown) was greater than that of all DI beneficiaries (29 percent) but less than that of all SSI recipients (57 percent; SSA 2017a, 2017c).

24 SSA appoints a representative payee for an adult beneficiary who is physically or mentally incapable of managing his or her own funds. In addition, SSA usually appoints a payee to receive benefits on behalf of a child younger than 18 (SSA 2017b).

References

Bailey, Michelle Stegman, Debra Goetz Engler, and Jeffrey Hemmeter. 2016. “Homeless with Schizophrenia Presumptive Disability Pilot Evaluation.” Social Security Bulletin 76(1): 1–25.

Census Bureau. 2013. “County Population Totals: 2010–2012; Datasets. Population, Population Change, and Estimated Components of Population Change: April 1, 2010 to July 1, 2012.” https://www.census.gov/data/tables/time-series/demo/popest/2010s-counties-total.html.

———. 2017. “One in Five Americans Live in Rural Areas.” https://www.census.gov/library/stories/2017/08/rural-america.html.

[HUD] Department of Housing and Urban Development. 2015. “Homeless Emergency Assistance and Rapid Transition to Housing: Defining 'Chronically Homeless.'” Federal Register 80(233): 75791–75806 (December 4). https://www.gpo.gov/fdsys/pkg/FR-2015-12-04/pdf/2015-30473.pdf.

———. 2017. The 2016 Annual Homeless Assessment Report (AHAR) to Congress. Part 1: Point-in-Time Estimates of Homelessness. https://www.hudexchange.info/resources/documents/2016-AHAR-Part-1.pdf.

Kennedy, Elizabeth, and Laura King. 2014. “Improving Access to Benefits for Persons with Disabilities Who Were Experiencing Homelessness: An Evaluation of the Benefits Entitlement Services Team Demonstration Project.” Social Security Bulletin 74(4): 45–55.

McCoy, Marion L., Cynthia S. Robins, James Bethel, Carina Tornow, and William D. Frey. 2007. Evaluation of Homeless Outreach Projects and Evaluation (HOPE). Task 6: Final Evaluation Report. Rockville, MD: Westat.

Missouri Census Data Center. 2016. “Geocorr 2014: Geographic Correspondence Engine.” https://mcdc.missouri.edu/applications/geocorr2014.html.

Nicholas, Joyce. 2014. “Source, Form, and Amount of In-kind Support and Maintenance Received by Supplemental Security Income Applicants and Recipients.” Social Security Bulletin 74(3): 39–54.

O'Connell, J. J. 2005. “Premature Mortality in Homeless Populations: A Review of the Literature.” Nashville, TN: National Health Care for the Homeless Council.

Rajnes, David. 2012. “'Fast-Track' Strategies in Long-Term Public Disability Programs Around the World.” Social Security Bulletin 72(1): 79–108.

[SSA] Social Security Administration. 2005. “Program Operations Manual System (POMS) SI 00835.060: Transients, Homeless Individuals, and LA/ISM Determinations.” https://secure.ssa.gov/apps10/poms.nsf/lnx/0500835060.

———. 2006. “Program Operations Manual System (POMS) DI 10105.080: Age Requirement.” https://secure.ssa.gov/apps10/poms.nsf/lnx/0410105080.

———. 2014a. “Program Operations Manual System (POMS) DI 11005.004: Identifying and Flagging Homeless Cases.” https://secure.ssa.gov/apps10/poms.nsf/lnx/0411005004.

———. 2014b. “Program Operations Manual System (POMS) DI 81010.080: Alerts, Flags, and Messages in the Electronic Disability Collect System (EDCS).” https://secure.ssa.gov/apps10/poms.nsf/lnx/0481010080.

———. 2014c. “Program Operations Manual System (POMS) DI 81020.085: Certified Electronic Folder (CEF) Flags.” https://secure.ssa.gov/apps10/poms.nsf/lnx/0481020085.

———. 2017a. Annual Statistical Report on the Social Security Disability Insurance Program, 2016. Publication No. 13-11826. Washington, DC: SSA. https://www.ssa.gov/policy/docs/statcomps/di_asr/2016/index.html.

———. 2017b. Annual Statistical Supplement to the Social Security Bulletin, 2016. Publication No. 13-11700. Washington, DC: SSA. https://www.ssa.gov/policy/docs/statcomps/supplement/2016/index.html.

———. 2017c. SSI Annual Statistical Report, 2016. Publication No. 13-11827. Washington, DC: SSA. https://www.ssa.gov/policy/docs/statcomps/ssi_asr/2016/index.html.

———. 2020a. Annual Statistical Report on the Social Security Disability Insurance Program, 2019. Publication No. 13-11826. Washington, DC: SSA. https://www.ssa.gov/policy/docs/statcomps/di_asr/2019/index.html.

———. 2020b. SSI Annual Statistical Report, 2019. Publication No. 13-11827. Washington, DC: SSA. https://www.ssa.gov/policy/docs/statcomps/ssi_asr/2019/index.html.

Wixon, Bernard, and Alexander Strand. 2013. “Identifying SSA's Sequential Disability Determination Steps Using Administrative Data.” Research and Statistics Note No. 2013-01. Washington, DC: SSA.