Out of Sight, Out of Mind: Including Group Quarters Residents With Household Residents Can Change What We Know About Working-Age People With Disabilities

Abstract

Information about residents of institutional and noninstitutional group quarters (GQ), particularly those with disabilities, has been limited by gaps in survey data, and statistics based on data that exclude some or all GQ residents are biased as estimates of total population statistics. We used the 2006 and 2007 American Community Survey (ACS) to identify the distribution of working-age populations with and without disabilities by major residence type and to assess the sensitivity of disability statistics to GQ residence. Our findings show that (1) of those with disabilities, about 1 in 13 males and 1 in 33 females live in GQ; (2) GQ rates are higher for individuals reporting mental, self-care, or go-outside-the-home disabilities than for those reporting sensory, physical, or employment disabilities; (3) younger males with disabilities are more likely to reside there, particularly at institutional GQ, reflecting their relatively high incarceration rate; (4) individuals with and without disabilities who are black, American Indian, were never married, or have less than a high school education have higher GQ residence rates; (5) 40% of male and 62% of female GQ residents have a disability; (6) adding GQ residents to household residents increases estimated disability prevalence for males by 6%, and the estimated difference between disability prevalence rates by gender nearly disappears; and (7) inclusion of the GQ population substantially lowers employment rate estimates for young males, blacks, and American Indians.

Introduction

A large proportion of working-age people with disabilities live in institutional group quarters (GQ), especially correctional and nursing facilities, and in some noninstitutional GQ, such as group homes. Statistics based on data that exclude some or all of these residents—which has been the standard practice—are biased as estimates of total population statistics. Although such biases are small for many population groups, they can be disproportionately large for people with disabilities, resulting in inferences that are misleading with regard to the size, characteristics, and well-being of the disability population and how they vary over time and geographic areas.

However, information about GQ residents, particularly those with disabilities, has been limited by gaps in survey coverage, infrequent data collection, and variation in disability definitions among surveys (She and Stapleton 2009). Gaps in survey coverage occur because institutional populations are omitted and also vary with respect to the survey frame for noninstitutional GQ. Federal surveys for specific institutions, especially local jails, state prisons, federal prisons (Harlow 1998; James 2004; Maruschak and Beck 2001), and nursing facilities (Decker 2005), have provided many insights into these populations. Such surveys, however, are conducted irregularly and use disability definitions that make comparisons of disability statistics from other surveys problematic. The Census Bureau collected disability information from the long-form census samples in 1990 and 2000, but the data are not comparable because of changes in the disability questions and a technical issue in administering the 2000 disability questions.

The purpose of this article is to address these issues by exploring two topics regarding GQ residence of working-age people with disabilities:

1. What is the distribution of the working-age population with disabilities by major residence type (household, institutional, and noninstitutional GQ), and how does it compare with the corresponding distribution for those without disabilities? How does the distribution vary by demographic characteristics, disability type, and state?

2. How sensitive are disability statistics to the inclusion or exclusion of GQ residents from survey samples? What are the effects on prevalence estimates and economic status?

We drew on the 2006 and 2007 American Community Survey (ACS) to answer these questions. The ACS is unique because since 2006, it has included GQ populations, which allows for estimates of institutional and noninstitutional GQ residents. Although more detail about residence status is contained in the survey, the Census Bureau does not make those data available for analysis in its public-use files because of small sample sizes.

We compared GQ residence rates of working-age people with disabilities (that is, the percentage residing in each type of GQ) with those without disabilities. We also considered how the GQ residence rates of those with disabilities vary with other personal characteristics and assessed the sensitivity of statistics on disability prevalence and employment rates to the inclusion or exclusion of GQ residents. This information is useful for policy makers, disability researchers, and advocates who need accurate information on the population with disabilities, including those who reside in GQ.

A full understanding of the intersection between disability and GQ requires inclusion of the entire age range. We have chosen, however, to focus on those aged 25 to 64—often termed the “working-age population”—because the concept of disability is linked to major social activities, and the major social activities of this age group are, on average, distinctly different from those of younger and older individuals.¹ Further, public policies and social institutions that support people with disabilities vary by age of the recipient in a manner that reflects variation in major social activities.² Parallel analyses for younger and older age groups are of sufficient importance, and contextually sufficiently different, to warrant separate treatment.

This is not to deny, of course, that many people under age 25 or over age 64 work. We use 25 and 64 as the bounds for inclusion in the analysis for practical reasons. Many individuals under age 25 live in college dormitories, and the available data do not distinguish between those living in dormitories and those living in other noninstitutional group quarters. Use of age 64 as the upper bound enhances comparability of our statistics to statistics in other publications, including U.S. Census Bureau publications based on the ACS, even though the historical basis for this limit—the full retirement age for Social Security—recently increased from 65 to 66.

Our findings show that (1) about 1 in 13 working-age males and 1 in 33 females with disabilities live in GQ; (2) GQ rates are higher for individuals who reported mental, self-care, or go-outside-the-home disabilities than for those who reported sensory, physical, or employment disabilities; (3) younger males with disabilities are more likely to reside in GQ, particularly in institutional GQ—a reflection of the greater incarceration rate for this demographic group; (4) similar to individuals without disabilities, individuals with disabilities who are black, American Indian, or never married or who have less than a high school education have higher GQ residence rates; (5) 40% of male GQ residents and 62% of female GQ residents have a disability; (6) adding GQ residents to household residents increases estimated disability prevalence for males by 6%, and the estimated difference between the disability prevalence rates for males and females nearly disappears; and (7) employment rates for males with disabilities—especially young blacks, and American Indians—are substantially lower when the GQ population is included than when it is excluded.

What Do We Know About GQ and Their Residents?

GQ residence units differ from housing units in terms of ownership and the provision of services, and GQ types differ by the level of supervision provided. The Census Bureau (2007b) defines a housing unit as “a house, an apartment, a mobile home, a group of homes, or a single room that is occupied (or, if vacant, intended for occupancy) as separate living quarters.” The defining characteristic is that individuals in households live apart from others who might reside in the building. In contrast, a GQ is a place “where people live or stay that is normally owned or managed by an entity or organization providing housing and/or services for the residents.”

GQ are divided into two categories based on the level of supervision. Institutional GQ house residents under legal care, custody, or supervision, including adult correctional facilities such as federal or state prisons and local jails, juvenile correctional facilities and group homes, nursing facilities, psychiatric hospitals, and residential schools for people with disabilities; residents are not free to come and go as they please. Noninstitutional GQ include college student housing, military quarters, emergency or transitional shelters, group homes, residential treatment centers for adults, workers’ dormitories, Job Corps centers, religious GQ, YMCA/YWCA, and hostels; residents are largely free to come and go as they please.

There have been dramatic changes in GQ populations in the past three decades. Although the percentage of the total population residing in GQ has increased modestly, from 2.6% in 1980 to 2.8% in 2000 (U.S. Census Bureau 1980a, b, 2007c), there has been a large shift away from nursing homes and toward correctional facilities. In 1980, 18.7% of institutional residents lived in correctional facilities, compared with 57.2% living in nursing facilities (U.S. Census Bureau 1980a). By 2000, the proportion of people in correctional facilities (49%) exceeded the proportion in nursing facilities (42%) (U.S. Census Bureau 2000).

Because a large majority of adult inmates of correctional facilities are of working age, whereas most nursing home residents are elderly, the change in the residential composition has been accompanied by a large increase in the GQ residence rate for the working-age population. By 2006, an estimated 73.6% of all GQ residents were between the ages of 18 and 64, including 36% between ages 25 and 64 (U.S. Census Bureau 2009). We focus on the latter age range for this article partly because many 18- to 24-year-olds are still preparing for employment and partly because we could not differentiate the large share who reside in college housing from other noninstitutional GQ residents in the public-use data. Among GQ residents 25 to 64 years old in 2006, an estimated 56.5% were incarcerated, 8.5% were in nursing homes, and 33.8% were in noninstitutional GQ, including just 1.6% in college housing (U.S. Census Bureau 2007c).

These dramatic changes also reflect two societal trends: increasing rates of incarceration (Western 2007) and efforts to help people with disabilities live in less restrictive environments. These trends may be interrelated to some extent: as the number of residents in psychiatric hospitals and state mental institutions has declined, more individuals with mental illnesses have entered jails or prisons, a process that has been referred to as “criminalization of the mentally ill” (Scott et al. 2006). A 2005 survey of prison and jail inmates found that about half either reported a recent history of mental illness (such as having a prior diagnosis or having received mental health treatment) or had symptoms of mental illness, although it is unclear how either measure might be influenced by being incarcerated (James and Glaze 2006). Individuals who are incarcerated have disability rates that are much greater than those found in the general population. A 1997 survey of federal and state prison inmates found that 21% of state inmates and 18% of federal inmates reported having a condition that limited their ability to work (Maruschak and Beck 2001). Between 23% and 31% of inmates reported having a disability, either because of difficulty seeing or hearing; a speech, learning, or physical disability; or a mental or emotional condition. Further, with increasing emphasis on mandatory sentencing, particularly for repeat offenders, the incarcerated population is expected to grow older, with a corresponding increase in the proportion with disabilities (Schmertmann et al. 1998).

Given the relatively large number of working-age people with disabilities residing in GQ, statistics on their disability prevalence and characteristics based on household data are likely to be biased as estimates for all working-age people with disabilities. The effect of the higher share of individuals with disabilities living in GQ means that the prevalence of disability in the full U.S. population is higher than in the household population. To illustrate, Brault (2008) reported the following disability prevalence rates for individuals aged 5 and older based on the 2006 ACS: 14.9% in the household population, 15.1% in the noninstitutional population (household plus noninstitutional GQ), and 15.7% in the entire population.

The inclusion of institutional and noninstitutional GQ populations in the ACS since 2006 represents a very significant step in addressing the data and survey issues mentioned earlier. The ACS sample now includes residents of essentially all institutional and noninstitutional GQ as well as all housing units. Unfortunately, detailed information about the GQ type is not available in public-use files; instead, only two broad GQ categories are identified: institutional and noninstitutional. This is regrettable because each category is far from homogeneous. For example, Brault (2008), who had access to the detailed ACS data, found that disability prevalence is low (4.1%) among residents of college housing (the largest noninstitutional GQ residence group), while much higher among residents of all other noninstitutional GQ (62.7%).

Methods

Data

We used the 2006 and 2007 ACS Public Use Microdata Samples (PUMS) for our analyses (U.S. Census Bureau 2008). The ACS, a component of the Census Bureau’s Decennial Census program, is an annual survey of approximately 3 million U.S. households. The combination of multiple approaches to collect data (mail, telephone, and in-person interviews) and the fact that housing units selected for the survey are required to participate by law leads to a response rate of between 95% and 97%. The ACS PUMS files used in this analysis are a subset of the full 2006 and 2007 ACS and include approximately 1.3 million U.S. households.

The ACS has three advantages over other data sources for answering our research questions. First, the ACS is the only data source with information on GQ residents. Depending on the topic, we present results according to the following categories of residence: institutional GQ, noninstitutional GQ, all GQ (institutional and noninstitutional), household, and total population (household plus all GQ). Second, the large sample size allows estimates at the national level, for smaller geographic locations, and for specific demographic groups. In addition to national statistics, we created selected statistics for each state, the District of Columbia, and Puerto Rico (hereafter referred to as “states”). We pooled the 2006 and 2007 data in order to increase the sample sizes for small groups and improve the accuracy of our estimates. Third, the ACS includes a series of broad disability questions that include impairment, activity limitations, and participation restrictions, with respondents able to answer affirmatively to more than one disability question. We categorized those responses into the following disability types: (1) sensory impairment (blindness, deafness, or a severe vision or hearing impairment), (2) physical impairment (substantial limitations of one or more basic physical activities such as walking, climbing stairs, reaching, lifting, or carrying), (3) mental impairment (difficulty learning, remembering, or concentrating), (4) self-care limitation (difficulty dressing, bathing, or getting around inside the home), (5) go-outside-the-home restriction (difficulty going outside the home alone to shop or visit a doctor’s office), and (6) employment limitation (difficulty working at a job or business). Each question specified either that the condition is “long-lasting” or has lasted six months or more. We classified any individual who responded affirmatively to one of these questions as having a disability.

The ACS question on employment limitation is problematic because it is highly subjective and very sensitive to the individual’s environment and other personal characteristics. An employment-focused measure by itself underestimates disability prevalence, or rather, identifies only a specific disability subpopulation (Burkhauser et al. 2002). At the same time, however, employment limitations are central to the concept that disability is a consequence of a mismatch between personal characteristics and environmental demands (Verbrugge and Jette 1994; World Health Organization 2002). In fact, the ACS dropped this question after 2007. Hence, we also produced some “any disability except employment disability only” statistics from which we excluded respondents who reported only an employment limitation. The ACS disability questions did not allow us to identify individuals with specific impairments (e.g., those with intellectual limitations/developmental disabilities (I/DD), or mental health conditions).

We produced many statistics for specific demographic groups because of our interest in how these characteristics are related to GQ residence. These include many commonly used demographic variables: age (classified either in narrowly defined age groups—25 to 29, 30 to 39, 40 to 49, 50 to 59, and 60 to 64—or more broadly defined age groups—25 to 34, 35 to 49, 50 to 64), sex, race (white, black, American Indian, Asian or other Pacific Islander, and other), ethnicity (Hispanic or non-Hispanic), marital status (married, divorced or separated, widowed, or never married), and education status (no high school diploma nor equivalent degree, high school diploma or equivalent, some college but not a four-year college degree, and four-year college degree or higher). Because the ACS allows individuals the option of identifying themselves as belonging to more than one racial category, the race categories overlap.³ For example, if a respondent identified herself as both white and black, we included the respondent in the statistics for both groups. We also produced statistics for the following characteristics that might be related in various ways to both disability and residence status: military (active duty, veteran, or nonveteran); immigrant/citizenship (native-born; naturalized immigrant; or noncitizen by entry period of 2000 or later, 1990 to 1999, or before 1990); and primary language (English, Spanish, or other).

To illustrate the effect of population definition on economic outcomes for working-age people with disabilities, we produced statistics on one such outcome: employment. We used the broadest measure of employment available in the ACS: working in a paid job, for oneself, or without pay in a family business or farm. The reference period is the week before the ACS survey.

Methodology

All the statistics presented are estimated numbers or percentages of people within various groups: the percentage of a group living in a residence type, the percentage having a disability (i.e., disability prevalence for the group), and the percentage having a demographic or socioeconomic characteristic.

For each statistic, we calculated margins of error to identify the likely range of sampling error. Margins of error are especially important for state statistics, in which samples for some groups can be relatively small. The interval defined by a point estimate plus or minus its margin of error is the 90% confidence interval for the population value. The margins of error were calculated using a formula provided by the U.S. Census Bureau (2007a).

For all statistics, we followed the U.S. Census Bureau’s convention of suppressing point estimates for which the coefficient of variation (the standard error relative to the point estimate) is greater than 30% or the sample size for the denominator is less than 50 cases.

Residence Type for the Disability and Nondisability Populations

Residence Type by Disability Status

The two most striking findings of the analysis are that (a) the percentage of working-age individuals with disabilities residing in GQ is much higher than the percentage of those without disabilities, and (b) the percentage for males with disabilities is much higher than that for females (Fig. 1). During this period, 8.4% of males with disabilities lived in GQ, compared with 2.0% of males without disabilities. For women, 3.4% of those with disabilities lived in GQ compared with just 0.3% for those without disabilities.

Differences between the institutional GQ statistics for males and females in Fig. 1 reflect much higher incarceration rates for males than for females. These differences might also reflect larger numbers of males in other institutions. For instance, Lakin et al. (2009) reported that, in 2008, 64.8% of residents in large state facilities for people with I/DD were male.⁴ We also find that males with disabilities are more likely than females to live in noninstitutional GQ. One reason for these gender differences might be that families are less willing to have females with disabilities reside in noninstitutional GQ because of safety concerns in general and sexual abuse in particular. In addition, active-duty military personnel, veterans with disabilities, and former inmates with disabilities sometimes live in noninstitutional GQ, and all are more likely to be male than female. We consider statistics for active-duty military personnel and veterans in the next section.

Males and females who reported mental, self-care, or go-outside-the-home disabilities are more likely to reside in GQ than those who reported sensory, physical, or employment disabilities. Individuals with the former types of disabilities may have more need for assistance, increasing their likelihood of residing in GQ. For men, this pattern is most distinctive for noninstitutional GQ and much less distinctive for institutional GQ. Excluding those who report only an employment limitation increases the percentage residing in both GQ types for both sexes. This is because those who report having only an employment disability are more likely than others with disabilities to live in the household population. One reason for this might be that those living in GQ, especially institutional GQ, think employment limitations are irrelevant because of their residential circumstances. Excluding those with only an employment limitation increases the GQ residence rate for males from 8.4% to 8.8%, and for females from 3.4% to 3.6%.

Variation in Residence Type by Characteristic

In this section, we examine variation in GQ residence across groups defined by the various characteristics described above.

Variation in residential status by age (Fig. 2) reflects incarceration rates by age as well as age-related changes in the composition of those with disabilities. An estimated 9.1% of 25- to 29-year-old males with disabilities live in institutional GQ, and another 3.3% live in noninstitutional GQ. As age increases, the percentage of individuals with disabilities living in GQ declines. This is attributable, in part, to increases in the percentage of people who experienced disability onset late in their working years. Such individuals perhaps have different family and community support systems in place, such as support from spouses and/or adult children, compared with those who experience disability earlier in life and may receive more support from their parents (see, e.g., Krauss et al. 2005). Older individuals are also less likely to have engaged in illegal activities, thereby reducing the number of individuals with disabilities residing in correctional facilities. In addition, the age-related changes in GQ may be influenced by differential mortality rates (which may be higher for individuals with disabilities who reside in GQ earlier in life) and disability composition (because older individuals with disabilities are less likely to have I/DD and more likely to have age-related chronic conditions).

Male-female differences in GQ residence rates diminish with age for those in both institutional and noninstitutional GQ. The statistics for 60- to 64-year-old females with disabilities appear anomalous; the percentage living in GQ for this age group is higher than for those 50 to 59 years old, in contrast to the results for males.

GQ residence also varies by race and ethnicity: blacks and American Indians with disabilities are much more likely to reside in GQ, especially institutional GQ, than those in other racial groups (Fig. 3). Variation is especially high among males, but the pattern for females is the same. An estimated 16.8% of black males with disabilities reside in GQ, including 12.0% in institutional GQ. The corresponding figures for American Indians are 12.8% and 9.3%. GQ residence—especially institutional—is much less common for Asian and Pacific Islander males with disabilities.

Hispanic men with disabilities are somewhat less likely than non-Hispanic men to reside in noninstitutional GQ but somewhat more likely to reside in institutional GQ, while Hispanic women are somewhat less likely than non-Hispanic women to live in GQ of either type. Statistics by primary language (English versus Spanish) are qualitatively similar.

Native-born individuals of either sex have higher GQ residence rates than foreign-born individuals (Fig. 3). These males with disabilities have a GQ residence rate of 8.7% compared with 5.7% for foreign-born males; the corresponding rates for females are 3.6% and 1.4%. For male immigrants, there is a sharp difference between the rate for those who are naturalized citizens (3.6%) and those who are not (7.8%), and among the latter, the GQ residence rate is especially high for those who entered the country before 1990 (9.5%) (data not shown). Patterns are similar for immigrant males without disabilities, but at much lower levels (data not shown). Most of the variation in the GQ residence rate by immigrant status reflects variation in the institutional GQ rate. These patterns suggest relatively high rates of incarceration among immigrants who have not become citizens even after living in the country for at least 16 years. For females with disabilities, however, GQ residence rates are lower for immigrants than for the native-born, and slightly lower for noncitizen immigrants than for those who are naturalized. Differences between the patterns for males and females by immigration status likely reflect much higher incarceration rates for males.

Relatively few married working-age people with disabilities reside in GQ—just 2.3% for men and 0.8% for women. Those who have never been married are especially likely to live in GQ, a finding that reflects disability severity for some and the consequences of incarceration for others, especially men. Similar patterns apply to those without disabilities, but at a much lower level.

The relationship between educational attainment and GQ residence is also quite strong. Very high percentages of people both with disabilities and with less than a high school education reside in GQ—14.1% for men and 6.4% for women. This finding might reflect the severity of disability for some and, especially for men, educational status at the time of incarceration for others. Similar patterns apply to those without disabilities, but at a much lower level (data not shown).

Male veterans with disabilities are less likely than nonveterans to reside in either institutional or noninstitutional GQ (Fig. 3), and 17.2% of active-duty military men with disabilities reside in noninstitutional GQ, compared with 11.6% for those without disabilities. Presumably, these estimates reflect residence in military barracks as well as military hospitals or rehabilitation facilities. Active-duty military males with disabilities are a tiny fraction of all working-age men with disabilities—an estimated 21,000 out of more than 11 million, or less than 0.2%. Hence, these statistics have little influence on those for all males with disabilities. There were too few female active-duty military personnel in the ACS sample to support reliable estimates.

High GQ residence rates for young working-age men with disabilities, coupled with high rates for working-age men with disabilities in certain demographic and educational categories, suggest that GQ residence rates for young men with disabilities in these same categories are extraordinarily high. This is confirmed by the statistics shown in Fig. 4, which is identical to Fig. 3 except that the sample is males and females aged 25–39 rather than 25–64. Just under one-quarter (23.2%) of black males with disabilities in the age group reside in GQ, including 19.2% in institutional GQ, compared with 9.3% for whites. The corresponding figures for males without disabilities are, respectively, 9.8% and 2.0%. American Indian males with disabilities in this age group also have extremely high GQ residence rates: 17.9% for all GQ and 14.6% for institutional GQ, compared with 7.2% and 5.8%, respectively, for those without disabilities. Widowers with disabilities have the highest GQ residence rate of any of the demographic groups presented (25.8%); similarly, widowers without disabilities also have a high GQ rate (7.4%). We also find that 20.1% of males with disabilities aged 25–39 who had less than a high school education reside in GQ, including 13.7% in institutional GQ; the corresponding figures for those without disabilities are 7.4% and 6.6%. GQ residence rates for young females with disabilities in the same groups are also high relative to those without disabilities but are much lower than the corresponding rates for males.

State Variation in Residence Type

The percentage of working-age people with disabilities who reside in GQ varies widely across states (Fig. 5). The state median for the proportion of males with disabilities residing in GQ is 7.6%. At one extreme, 3.4% ± 0.4% of males with disabilities in Puerto Rico and 3.6% ± 1.3% of those in Vermont reside in GQ. At the other extreme, 17.4% ± 2.7% of males with disabilities in the District of Columbia reside in GQ, and four states have more than 10%, led by Virginia (11.2% ± 0.7%). Kentucky and Nevada have the lowest rates for noninstitutional GQ residence (1.0% ± 0.3% and 1.0% ± 0.4%, respectively), while Washington, D.C., has the highest (11.6% ± 2.4%)—well beyond the rate of Minnesota, which ranks second (6.9% ± 0.8%). For institutional residence, the range among states is narrower than for all GQ, from a low of 1.5% ± 0.5% (New Hampshire) to a high of 9.3% ± 0.6% (Virginia).

For females with disabilities, Puerto Rico (0.9% ± 0.2%) and Alabama (1.6% ± 0.3%) have the lowest GQ residence rates, while Washington, D.C., (8.0% ± 2.0%) and New York (5.0% ± 0.3%) have the highest (Fig. 5). Puerto Rico has the lowest rates for noninstitutional (0.3% ± 0.1%) and institutional (0.6% ± 0.2%) GQ residence rates among all states. When Puerto Rico is excluded, there is an almost fivefold difference between the states with the lowest (Alabama, 0.7% ± 0.2%; Nevada, 0.7% ± 0.3%) and highest (Minnesota, 3.3% ± 0.6%) noninstitutional GQ residence rates, and more than a threefold difference between states with the lowest (South Carolina, 0.8% ± 0.2%) and highest (Illinois, 2.5% ± 0.3%) institutional rates.

For both sexes, rates of GQ residence for those without disabilities are much lower in all states, and the range is narrower (data not shown). In the median state, 1.8% of its male residents are in GQ, and the range is from 0.5% ± 0.2% (Vermont) to 4.6% ± 0.6% (Alaska). For females without disabilities, several states have GQ residence rates around 0.2%, and the sample size in some states is too low to produce an accurate statistic. Only one jurisdiction has a GQ residence rate above 1.0% for females: Washington, D.C. (1.4% ± 0.3%). States that have high GQ rates for individuals with disabilities tend to also have high GQ rates for individuals without disabilities (and vice versa), although the correlation is not perfect and the tendency is stronger for male rates than for female rates.⁵

Sensitivity of Disability Statistics for Working-Age People to Population Definition

Disability Prevalence by Population Definition

Given the differences in GQ residence rates for the populations with and without disabilities, it is no surprise that disability prevalence statistics for the entire population differ from those for the household population (Table 1). Differences are not very large, however, because even though GQ residence rates are much higher for those with disabilities than for those without, only a small minority of those with and without disabilities reside in GQ. Estimated disability prevalence for all working-age males is 13.9%, which is 6.0% higher than the 13.1% estimate for the household population alone; the total population figure for females is 14.1%, or 2.7% higher than the 13.7% estimate for the household population. The total population estimates do, however, present a somewhat different picture of sex differences in disability prevalence for the working-age population than the household population estimates do. The latter indicate that disability prevalence is 0.6 percentage points lower for males than for females, but when the GQ population is included, the estimated difference is reduced to 0.2 percentage points.

Disability prevalence in the total population is higher than in the household population for each disability type (Table 1). In percentage terms, the differences for both sexes are greatest for mental, self-care, and go-outside-the-home disabilities. With one exception, prevalence rates are higher for females than for males, and sex differences are smaller for the total population than for the household population. Most notably, the sex difference in prevalence of mental disabilities observed in the household population (an estimated 0.3 percentage points higher for females) essentially disappears when the GQ population is included. The exception is sensory disabilities: males are considerably more likely to report sensory disabilities than females (3.8% in the total population versus 2.7% for females), and the difference is slightly larger in the total population than in the household population (1.1 percentage points versus 1.0).

When those who report only an employment limitation are excluded, the prevalence of any disability in the total population drops by about 1 percentage point for both males and females. Excluding those with only employment limitations increases the effect of including the GQ population in prevalence statistics because these individuals account for a smaller share of those with disabilities living in GQ, compared with those with disabilities living in the household population. For men, prevalence in the total population is 6.5% higher than in the household population when those with only employment limitations are excluded, versus 6.0% higher when they are included. The corresponding values for women are 2.9% and 2.7%.

Relative differences in prevalence rates for the total and household populations are larger for young adults than for older adults in the working-age population, reflecting the decline in GQ residence rates with age. This difference is particularly notable for young males, 25 to 34 years old, whose prevalence rate for the total population is 9.3% higher than for the household population alone; the corresponding figures for males 35 to 44 years old and 50 to 64 years old are, respectively, 8.5% and 4.1%.

Table 1 also shows differences in disability type by residence. Males in households most commonly reported having physical (7.9%) or employment (7.6%) disabilities, while those in GQ residences most commonly reported a mental disability (24.1%) or an employment disability (22.7%). Females in GQ reported higher disability rates than their male counterparts. Presumably this is because the share of females in GQ who reside in facilities designed to serve people with disabilities is larger than for males. Females residing in households most commonly reported having physical (9.1%) and employment (8.3%) disabilities, while females residing in GQ most commonly reported employment (48.0%) and mental (44.3%) disabilities.

Disability prevalence increases with age at approximately the same rate for both household and GQ residents (Table 1). Although disability prevalence is much higher in the GQ than in the household population, people with disabilities represent a minority of the GQ population in the youngest age group. For the middle and oldest age groups, however, people with disabilities are in the majority. This partly reflects lower incarceration rates at older ages as well as the absence of significant numbers of college students and military personnel. Note, however, that disability prevalence is less than 50% for males in the middle age group, most likely because a large share of this group is incarcerated. Disability prevalence is exceptionally high for the oldest female group residing in GQ (84.8%).

Employment Rates by Population Definition

As is well known, the employment rate for working-age individuals with disabilities (i.e., the number employed divided by the total number) is substantially lower than for those without disabilities. Most cited employment statistics are for the household or noninstitutional populations, depending on which data source is used. Here, we examine how sensitive employment rates are to the inclusion of the noninstitutional and total GQ populations in the ACS samples (Table 2). We do not include a separate column for institutional GQ because the ACS does not count any institutional residents as employed. Employment rates for the total population of people with disabilities include institutional residents in the denominator, but none are counted in the numerator.

The employment rate statistics for males are sensitive to the inclusion of both institutional and noninstitutional GQ populations, but the effects are small because these populations are not very large. While the employment rates are lower for residents of noninstitutional GQ, no individuals in institutional GQ are counted as employed. Hence, holding disability status constant, the employment rate for all working-age males is necessarily lower than the rate for those in the household population. In the household population, the estimated employment rate for working-age males with disabilities is 40.7%, and the employment rate for working-age males who live in noninstitutional GQ is 24.0%. When males in all residence types are included, the employment rate for those with disabilities is 38.1% (2.6 percentage points lower than for the household population, or a relative change of 6.4%). Adding the noninstitutional and institutional GQ populations reduces the employment rate of males with disabilities by more than the rate for males without disabilities because a smaller proportion of those without disabilities reside in institutional GQ. The employment rate of males without disabilities is 87.9% in the household population, 58.8% in the noninstitutional GQ population, and 86.4% in the total population.

Differences are larger for groups of males with relatively high institutional GQ residence rates. For 25- to 34-year-olds, the employment rate in the household population is 51.0%, while in the total population it is 45.8% (a 10.2% relative difference). The corresponding numbers are 29.2% and 25.1% for blacks (a 14.0% relative difference) and 38.2% and 34.1% for American Indians (a 10.7% relative difference).

For females, the addition of the GQ population to the household sample has a much smaller effect on employment statistics. The employment rate for working-age females with disabilities in the household population is 33.3%; the corresponding statistic when the GQ population is included is 32.6%—less than 1 percentage point lower. As a result, employment rates for females are much closer to those for males when the GQ population is included than when it is excluded.

Excluding those who report only an employment limitation from the disability population increases the household and total population employment rates for that population (second rows in the top and bottom halves of Table 2). This is because those reporting only an employment limitation have a lower employment rate than those reporting at least one additional disability, holding other characteristics constant. This change in definition also increases the difference between the household and total population employment rates because a disproportionately large share of those who report only an employment limitation are in the household population.

Because GQ residence rates vary substantially across states, comparisons of employment rates for the household population across states can lead to misleading conclusions about relative employment outcomes, especially for young men with disabilities. To illustrate, we produced state statistics on employment rates for 25- to 39-year-old men with disabilities for both the total and household populations. We plotted both series against the state’s estimated GQ residence rate for each population, along with lines for the least squares regression of each series against the GQ residence rate (Fig. 6). As is evident from the figure, total population employment rates for states near the high end of the GQ residence range are substantially lower than for states near the low end; based on the regression line, the predicted employment rate for a state with a 20.0% GQ residence rate is 9.9 percentage points lower than the corresponding rate for a state with a 5.0% residence rate. Differences in household employment rates across states with high and low GQ residence rates are much smaller; based on the household employment rate regression line, the predicted employment rate for a state with a 20.0% GQ residence rate is 3.8 percentage points lower than the corresponding rate for a state with a 5.0% GQ residence rate.

Conclusion

In summary, among the working-age population (25 to 64 years old), we find that about 1 in 13 males with disabilities and 1 in 33 females with disabilities live in GQ. Younger males with disabilities in this group are the most likely to reside in GQ, particularly institutional GQ—a reflection of a high incarceration rate. Similar to men without disabilities, men with disabilities who are black, American Indian, never married, or have less than a high school education are substantially more likely than others to reside in GQ. There is also substantial variation at the state level in the proportion of people with disabilities who live in GQ.

The prevalence of disability is very high among working-age GQ residents: 40% for males and 62% for females. Adding GQ residents to household residents increases estimated disability prevalence for males by 6.0%, compared with just 2.7% for females. The estimated 0.6 percentage point difference between the disability prevalence rates for males and females found in the household population is reduced to 0.2 percentage points when the total population is considered.

Inclusion of the GQ population in the calculation of employment rates slightly reduces the rates for people with disabilities, more for males than for females and especially for subgroups with high institutional GQ rates. Employment rates for young males, blacks, and American Indians with disabilities are noticeably lower when the GQ population is included. Cross-state comparisons of employment rates for young men with disabilities can be misleading if the rates are based on the household population alone and the states have substantially different GQ residence rates.

In considering the implications of these findings, there are three important caveats to keep in mind. First, we have not been able to produce information on residents in detailed types of institutional and noninstitutional GQ because such information is not available in the public-use file. This access issue necessarily limits inferences drawn from these data. Differentiation between residents of correctional facilities and residents of institutions designed to serve people with disabilities is especially important, as is differentiation between college students and other noninstitutional GQ residents. It would be highly desirable for the Census Bureau to allow researchers greater access to the detailed GQ information, perhaps through a special file that pools data across several years, so that sample sizes are sufficiently large to protect confidentiality, or by making it easier for screened researchers to gain access to restricted data.

Second, we do not know the extent to which ACS respondents residing in GQ would answer the disability questions differently were they not residing in GQ. This issue seems particularly problematic for the incarcerated population, many of whom might be quick to blame a mental or physical impairment for their incarceration. The Census Bureau has not studied this issue.⁶ The fact that disability prevalence estimates for males remain lower than for females when the GQ population is included suggests that any such effect might not be large. Note, though, that the prevalence rate for males 25 to 34 years old, the age group with the highest GQ residence rate, is larger than for females in the same age group by a statistically significant amount—7.7% versus 7.0% (Table 1). That difference is not entirely caused by the male-female difference in GQ residence rates for this age group, however, because the prevalence rate for males in the household population is also higher than for females, by 0.3 percentage points. In older age groups, the prevalence rate for males in the total population is lower than the prevalence rate for females, a fact that could be related to lower GQ residence rates for older males.

The third important caveat concerns the interpretation of state institutional GQ residence rate statistics. Some of the cross-state variation in GQ residence rates reflects where institutions happen to be located and how state of residence is determined. Some institutional residents reside in institutions that are not in the same state in which they lived prior to entry. This is often true for federal prisoners, and possibly for persons convicted of a crime committed in a state other than their own or those incarcerated by their state of origin in an institution in another state. The Census Bureau counts such individuals as residents of the state in which the institution is located. An earlier effort by the Census Bureau to develop an alternative classification system determined that there was too much ambiguity about the former state of residence to implement any reasonable alternative.

Despite these caveats, the statistics are instructive. They leave the distinct impression that young males with disabilities, especially those who are black, American Indian, or poorly educated, are more likely to reside in GQ, particularly institutional GQ, than their counterparts without disabilities or other groups with disabilities. Although the data do not indicate the type of institutional facility, it appears that the majority reside in correctional facilities. If this is true, the link between incarceration and disability would seem to be a critical area for future research. One possible interpretation of these findings is that a mental or physical impairment contributes to an individual’s decision to participate in criminal activity, but we think this explanation is too simplistic. Disability is now understood to be a product of both an individual’s physical and mental characteristics and the individual’s environment. Hence, a more plausible interpretation is that it is very difficult for individuals with certain physical or mental conditions to succeed in their social environment without resorting to criminal activity. Lack of opportunities to succeed in school or to obtain a reasonable job, reliance on government or others for financial support, and an absence of health and social supports to promote independence might all increase the chance that individuals with disabilities participate in criminal activity. It is also possible that their impairments disadvantage them in the criminal justice system; that is, holding criminal activity constant, they might be more likely than others to be arrested, convicted, and sentenced to prison. A final possibility is that the incidence of disability after incarceration is high, potentially because of the nature of the prison environment.

The fact that far fewer women with disabilities are incarcerated presumably reflects the same personal and social factors that lead to much lower incarceration rates among all women. While it seems likely that disability prevalence is high among women who are incarcerated, we cannot confirm that from the public ACS data because the share of incarcerated women is much smaller than it is for men.

For some purposes, the GQ population—or at least the institutional GQ population—should be excluded in calculations of employment rates, even if data for the full population are available. For many policy purposes, the institutional GQ population might be irrelevant (e.g., enforcement of the employment provisions of the Americans with Disabilities Act). Yet, if we want to understand the consequences of public policies and other factors (such as the economy) on the extent to which working-age people with disabilities are self-sufficient, it would be a mistake to ignore those who reside in GQ because the very same factors might affect the number who reside in GQ. It seems likely, for instance, that a prolonged and substantial downturn in the economy both reduces the employment of people with disabilities and increases the number residing in institutions. If statistics for the non-GQ population alone are considered, as is usual, the magnitude of the decline will be understated.

While institutional GQ is the predominant form of GQ residence, the proportion of individuals with disabilities in noninstitutional residences—particularly women—is not inconsequential, either from a resident or policy perspective. This issue may be particularly important for people with I/DD, for whom noninstitutional GQ (in the form of smaller residential units that are increasingly being operated by nonstate agencies) has replaced larger state-run institutions over the past 30 years (Lakin et al. 2009). Many of those with I/DD spend their entire lives in noninstitutional GQ. While this shift away from institutional GQ represents a clear leap forward, there still may be issues with autonomy and integration for noninstitutional GQ residents. Furthermore, the cost of providing noninstitutional GQ is considerable, and, as evidenced by waiting lists of those seeking services, state budgets for such programs are not sufficient to meet demand. More research is needed to understand the various types and costs of noninstitutional residences available for people with disabilities, the quality of care, and the level of community integration they provide.

Another possible avenue for research is the exploration of the reasons for state differences in GQ rates for individuals with disabilities. Do some states have policies that promote certain GQ types (such as group or congregate care homes), while others have more home- and community-based supports that provide those with disabilities more opportunities to reside in households? Do the legal systems of states with low institutional GQ rates differ from those with high institutional GQ rates, or do the former send their prisoners to other states? How do laws addressing nursing care facilities affect GQ residence rates?

It would be interesting to conduct similar analyses for younger and older populations. Incarceration rates are much lower for both groups than for the working-age group. Many in the younger group reside in GQ associated with educational institutions, while most GQ residents in the older group live in assisted-living facilities or nursing homes. Including or excluding GQ residents in disability prevalence statistics for older adults may have a much greater effect on disability prevalence than was observed for working-age adults.

People with disabilities are disproportionately represented in the GQ population, a fact often overlooked by policy makers because few statistical tabulations include them. GQ residence is not necessarily negative or harmful and often provides necessary supports for residents with disabilities. However, given the movement toward more independence and autonomy for this population, GQ, particularly institutions, may impede these goals. Most notably, incarceration for young adults with disabilities may have long-term consequences for their later ability to live in the community. Although there is already an emphasis on home and community-based care for individuals with disabilities in order to avoid nursing facilities, no such policies are in place for those at risk of being involved in the criminal justice system. The cost of changing the environment in ways that will reduce the number of men and women with disabilities who become incarcerated is likely quite high. So far, efforts to improve adult outcomes for youths with disabilities have had only limited success.⁷ But perhaps we have not invested enough in this effort, given the high cost of the alternative—the cost of the criminal activities that many engage in now, the high cost of caring for them as inmates, and the tragic waste of lives lived behind bars.

Acknowledgments

This research was sponsored by the Department of Education’s National Institute for Disability and Rehabilitation Research (NIDRR) under its Rehabilitation, Research and Training Center on Disability Demographic and Statistics grant to Cornell University (No. H133B031111). The contents do not necessarily represent the policy of the Department of Education, and readers should not assume endorsement by the Federal Government (Edgar, 75.620 (b)). The authors gratefully acknowledge John Czajka, Dale Garrett, Andrew Houtenville, and David Wittenburg, who offered instructive technical assistance, and Bonnie Hart, who assisted with programming. In addition, the comments of three anonymous reviewers helped to improve the clarity and focus of this article.

Notes

References