A Guide to Disability Statistics from the American Community Survey (2008 Forward)

 

Author: William Erickson Employment and Disability Institute, Cornell University

 

The user guide has been developed through the Center for Rehabilitation Research using Large Datasets (CRRLD). The CRRLD is a collaboration between the University of Texas Medical Branch (UTMB) and Cornell University's Employment and Disability Institute (EDI) and is funded through a grant from the National Institutes of Health (NIH grant # R24HD065702-01). The goal of the Center is to build rehabilitation research capacity by increasing the quantity and quality of rehabilitation outcomes research using large administrative and research datasets. The contents of this paper do not necessarily represent the policy of the U.S. Department of Health and Human Services, and you should not assume endorsement by the Federal Government (Edgar, 75.620 (b)). 

 

Acknowledgement 

Many thanks to the following individuals who provided valuable assistance with the development of this paper: Melissa Burress, Zafar Nazarov, Margaret Waelder, Sara VanLooy, and Sarah von Schrader.

 

University at Texas-Medical Branch Project Team  Ken Ottenbacher, Ph.D., OTR (Principal Investigator)   Amol Karmarkar, Ph.D 

 

Employment and Disability Institute (EDI) Project Team Susanne Bruyère, Ph.D., CRC, Director of EDI William Erickson, Research Specialist Arun Karpur, Extension Faculty

Table of Contents

Introduction. 2

ACS Background, Methodology and Definitions. 5

Purpose of the ACS. 5

Development of the ACS. 7

Universe and Sample Design. 7

Data Collection Methodology. 8

Household Sample Data Collection. 8

Group Quarters Sample Data Collection. 8

Definitions. 9

Employment Definitions. 18

Dissemination. 21

Changes to the ACS and Implications. 23

Future Changes. 24

ACS Description of Disability Population. 25

Prevalence. 26

ACS Employment and Economic Well-Being Estimates. 29

ACS Group Quarter (GQ) Population Estimates. 31

ACS State Level Estimates. 32

Summary and Conclusions. 35

References. 37

Appendix A: Quick Start to Analysis of the ACS PUMS Data. 40

Appendix B.  ACS Sample Design and Computation of Standard Errors. 43

Appendix C. ACS Disability Question Set History. 50

Appendix D. International Classification of Functioning, Disability and Health (ICF) concepts. 56

Appendix E. Disability Prevalence rate by state. 58

Appendix F. Employment rate by state and disability status. 60

 


Introduction

This User Guide focuses on the information relevant to disability contained in the American Community Survey (ACS), a nationally representative survey of households and “group quarters” (GQ) populations conducted by the U.S. Census Bureau. The ACS is a valuable source of disability information, making it possible for researchers and policymakers to track changes in prevalence rates, employment and economic indicators across states and over time.  It provides vital information on how the labor market and the social and policy environments may influence the status and economic well-being of the population with disabilities. The ACS offers researchers and others a tremendous amount of population-based information that can be used in a myriad of ways, from identifying potential populations and localities in need of services to providing baseline measures for comparison to other studies. The ACS is invaluable for the purposes of monitoring the progress of the population with disabilities and is an important component of the nation’s efforts to reach the goals of full participation, independent living, and economic self-sufficiency for the population with disabilities.

Cornell University produced the original series of User Guides to disability data sources as a part of its Rehabilitation Research and Training Center on Disability Demographics and Statistics designed to bridge the divide between the sources of disability data and the users of disability statistics.  This ACS User Guide is an update and extension of the original guide written by Robert Weathers in 2005.    It has been produced in collaboration with the University of Texas Medical Branch under funding from the National Institutes of Health, and contains valuable information on the disability population that can be used to inform rehabilitation researchers and other disability data users. The purpose of this User Guide is to provide:

1.   A guide to the disability information available in the American Community Survey (ACS);

2.   A discussion of the strengths, limitations and unique features of the survey;

3.   Examples of estimates based on analysis of the ACS Public Use Microdata Sample (PUMS) dataset, the anonymized ACS data the Census Bureau makes publically available for researchers. Estimates include the size of the population, the prevalence rate, the employment rate, and measures of economic well-being that may be of interest to rehabilitation researchers

4.   A brief “quick start” section that includes basic information and informative resources to facilitate researcher’s ability to perform analysis of the ACS PUMS data. (See Appendix A)

 

The ACS is an annual survey and provides national, state, and local level data on demographic, social, economic, and housing characteristics as well as detailed information on the population living in institutionalized and non-institutionalized “group quarters” (GQ).  In past decades, these were collected in the Decennial Census long form survey. As of 2010, the Decennial Census will only field the basic short form survey, focusing on the collection of the basic population count information, and the ACS will replace the Decennial long form survey for the purpose of collecting in-depth and detailed population and housing information. The ACS includes six questions that are used to identify the population with disabilities as well as two questions focused on identifying veterans with service-connected disabilities.

Many features of the ACS are useful to rehabilitation researchers, policymakers, disability service providers, and the disability advocacy community.  First, the ACS contains a unique combination of data on disability, demographic characteristics, economic well-being, and employment.  Second, the sample size and the design of the ACS allows users to examine a variety of annual disability statistics at the national, state, metropolitan statistical area (MSA), and county level. Third, although the ACS has undergone several alterations to its survey design and the content and phrasing of its disability questions that currently limit the feasibility of looking across time (see limitations below), it is possible to estimate trends in various disability statistics at a level of geographic detail (i.e., the county level) that is not possible in any other national survey. In 2005, the ACS was implemented in Puerto Rico as well (referred to as the Puerto Rico Community Survey –PRCS).   Also, as of 2006 the ACS includes a sample of the population living in group quarters (GQ).  Group quarters include such places as college residence halls, residential treatment centers, skilled nursing facilities, group homes, military barracks, correctional facilities, and workers’ dormitories. Several types of GQ house significant populations with disabilities. These strengths of the ACS allow users to track changes to the disability population so that services can be better targeted to the population, publicly and privately funded disability programs can be more effectively administered, and new programs can be evaluated. In recent years, the ACS matured and stabilized, allowing for estimates across time from 2008 forward.

While the ACS can provide knowledge about a wide variety of topics, it has some limitations.  First, the ACS is limited to six basic questions that are used to identify the disability population. These questions provide insight into several relatively specific impairments but fail to identify other common impairments and disability types, including mental illness and upper body impairments such as disabling back injuries. These questions also do not allow for the identification of specific health conditions (e.g., cancer, paralysis, or HIV/AIDS) or the cause of a disability.  Second, the ACS definitions do not explicitly include the societal and environmental factors that may contribute to a disability, such as discrimination and lack of reasonable accommodations. Third, although the ACS sample does include group quarters, the sample sizes are small and the publically available ACS data does not provide information regarding the type of group quarters. For example, it is not possible to distinguish the population in correctional facilities from nursing homes, limiting the usefulness of this information.  Finally, due to changes in the disability question design (in 2003 and 2008) and sampling (in 2005), examining disability over time is problematic except within specific short time periods. 

ACS Background, Methodology and Definitions

The survey methodology can have an important impact on the information that a survey collects on the population with disabilities.  Mathiowitz (2000) provides a good review of the general methodological issues as well as those specific to the population with disabilities.  The purpose of this section is to describe the development of the ACS, the methods used by the ACS to collect information on the population, and the precise definitions used to describe the population with disabilities. This guide will focus on the most recent ACS survey and methodology as of 2009. Please refer to the earlier ACS User Guide (Weathers, 2005) for details regarding the earlier versions of the ACS.

Purpose of the ACS

The ACS is a continuous data collection effort by the U.S. Census Bureau that is used to produce annual estimates at the national, state, and local level on the characteristics of the United States population. The ACS collects information on an annual basis from approximately 3 million household addresses in the United States and an additional 36,000 addresses in Puerto Rico.  From 2006 forward, it includes 2.5 percent of the population living in Group Quarters.

The U.S. Census Bureau has three main objectives for the ACS (U.S. Census Bureau, 2003).  The first objective is to provide federal, state and local governments with an information base for the administration and evaluation of government programs.  The second objective is to use the ACS as a replacement for the Decennial Census long form so that the Decennial Census efforts can focus solely on counting the population.  The third objective is to provide data users with timely information each year on demographic, housing, social and economic statistics that can be compared across states, communities, and population groups.

Table 1 is a list of the topics covered in the ACS. Later in this document, we focus on disability, employment and economic measures; however, any of the topics listed below can potentially be examined with respect to disability using the ACS. 

 

Table 1

 ACS Survey Topics

 

Basic (Population)

Financial (Housing)

Economic (Population and Housing)

·      Age

·      Business on Property

·      Class of Worker

·      Sex

·      Cost of Utilities

·      Food Stamps Benefit

·      Hispanic Origin

·      Condominium Fee

·      Health Insurance

·      Race

·      Insurance

·      Income

·      Relationship

·      Mobile Home Costs

·      Vehicles Available

 

·      Mortgage

·      Work Status Last Year

Social (Population)

·      Real Estate Taxes

·      Industry

·      Ancestry

·      Rent

·      Journey to Work

·      Citizenship Status

·      Tenure

·      Occupation

·      Disability

·      Value of Property

·      Place of Work

·      Educational Attainment

 

·      Labor Force Status

·      Fertility

Physical (Housing)

 

·      Grandparent caregiving

·      Number of people

 

·      Language Spoken

·      Acreage

 

·      Marital History

·      Agricultural Sales

 

·      Marital Status

·      Bedrooms

 

·      Military Service Period

·      House Heating Fuel

 

·      Place of Birth

·      Kitchen Facilities

 

·      School Enrollment

·      Plumbing Facilities

 

·      Residence 1 Year Ago

·      Telephone Service

 

·      BA/BS Field of Degree

·      Rooms

 

·      Veteran Status

·      Units in Structure

 

·      VA Service-Connected Disability Rating

·      Vehicles Available

 

·      Year of Entry to U.S.

·      Year Moved Into Unit

 

 

·      Year Structure Built

 

Note. Reprinted from U.S. Census Bureau (2012). ( “Questions on the form and why we ask”  http://www.census.gov/acs/www/about_the_survey/questions_and_why_we_ask/)

Development of the ACS

The development of the ACS began in the 1990s, and the survey went through several testing phases prior to full implementation, this Guide will focus on its history since 2003. For more background on the ACS prior to 2003, please see the previous ACS User Guide (Weathers, 2005).  The ACS is designed to produce reliable estimates at the national level as well as for small geographic areas (e.g., counties and congressional districts).  It differs from the Decennial Census in that it collects data on a continuous basis and produces reliable estimates by pooling the data over one-year, three-year, or five-year periods, depending upon the size of the area and other considerations.  Full implementation of the ACS occurred in 2005 and included three million household addresses per year, 2.5 percent of those living in Group Quarters per year, and 36,000 addresses in Puerto Rico per year. As of June of 2011 the ACS sample has been further increased to 3.54 million households per year.

Universe and Sample Design

The ACS collects data each year from a nationally representative sample drawn from the universe of U.S. households and the population living in a sample of “Group Quarters” as described below.

The ACS utilizes a two-stage stratified sample design.  Population estimates based upon the sample have some degree of sampling error and non-sampling error.  Standard errors and confidence intervals that account for the sample design describe the degree of uncertainty in the estimates due to sampling error and some forms of non-sampling error.  Appendix B provides 1) additional information on the sample design for the ACS and the ACS PUMS; 2) describes the efforts that the Census Bureau uses to minimize non-sampling error, and 3) references the documents that describe the techniques used to compute standard errors. 

Data Collection Methodology

Household Sample Data Collection

The survey uses three different methods to collect data from households:

1.       A survey delivered by mail to a household member who is responsible for completing it and mailing it back to the Census;

2.      A telephone survey conducted by a Census Bureau employee using Computer Assisted Telephone Interview (CATI) technology; and

3.      In-person interviews using Computer Assisted Personal Interview (CAPI) technology.  A person referred to as the “householder,” usually the person who either owns the housing unit or who pays the rent for the housing unit, is responsible for completing the ACS questionnaire for the household. 

The Census Bureau first attempts to administer all of the questionnaires by mail.  Approximately six weeks after the questionnaires are mailed, the Census Bureau begins conducting telephone interviews for all households who have not responded by mail and have a telephone number.  The Census Bureau then identifies a sub-sample of households that do not respond by mail or telephone and a trained Census Bureau field representative is sent to these households to conduct in-person interviews.  This multi-method approach, combined with the legal obligation for recipients to respond to the survey, results in high response rates, generally between 97 to 98 percent.

Group Quarters Sample Data Collection

The GQ data collection is performed by a U.S. Census Bureau Field Representative  and is divided into two phases. The first phase is the Group Quarters Facility Questionnaire (GQFQ), a facility-level interview with the facility contact person or administrator. The GQFQ determines the facility type, population size and obtains a list of all residents, which is used as the basis of the sample of persons to be interviewed.

The second GQ data collection phase is the person-level interview. All residents of small GQs (15 or fewer residents) are interviewed. A sample of ten residents from each large GQ is systematically selected. The ACS GQ survey is a bilingual (English/Spanish) booklet for collecting person level information. The survey does not include housing questions, but does ask about food stamp benefits. The Field Representative assembles the appropriate number of survey packages and records location information on the sampled individuals (i.e., room number). The preferred method of data collection is face-to-face with the sampled resident. Other approaches can be used, such as telephone interviews or leaving the survey with the resident or the GQ facility contact to distribute. When surveys are left, the Field Representative returns to collect completed surveys within two days. The Field Representative also ensures that the sample residents are mentally and physically able to understand and complete the survey on their own, and some cases may perform a proxy interview with a relative, guardian or GQ contact for the sampled resident.

Definitions

A description of the survey questions and the population used to produce data on disability, demographics, employment, and economic well-being are shown in Table 2.

  Table 2

Disability Definitions from the 2009 American Community Survey

Description

PUMS Variable

Name

Question

Population

Hearing difficulty

DEAR

Q17a. Is this person deaf or does he/she have serious difficulty hearing?

 

All ages

Visual difficulty

DEYE

Q17b. Is this person blind or does he/she have serious difficulty seeing even when wearing glasses?

All ages

 Cognitive difficulty

DREM

Q18a. Because of a physical, mental, or emotional condition, does this person have serious difficulty concentrating, remembering, or making decisions?

Ages 5 and older

Ambulatory difficulty

DPHY

Q18b. Does this person have serious difficulty walking or climbing stairs?

Ages 5 and older

Self-care difficulty

DDRS

Q18c. Does this person have difficulty dressing or bathing?

Ages 5 and older

Independent Living difficulty

DOUT

Q19. Because of a physical, mental, or emotional condition, does this person have difficulty doing errands alone such as visiting a doctor's office or shopping?

Ages 15 and older

Disability recode

DIS

If a person responds “yes” to one or more of the six questions above (Q17a to Q19), then the Census classifies the person as having a disability.

All Ages

Veteran Service-Connected Disability Rating

DRATX

Q28a. Does this person have a VA service-connected disability rating? (checkbox-yes/no)

All Veterans

Veteran Service-Connected Disability rating

DRAT

Q28b. What is this person’s service-connected

disability rating? : 0 percent; 10 or 20 percent; 30 or 40 percent; 50 or 60 percent; 70 to 100 percent;  Not reported

Veterans with a rating

 

Disability.  The ACS has undergone a number of minor and major changes to its disability questions. For a history of the changes to the disability questions since 2003, please see Appendix C.  Significant changes were also made in 2008 to other ACS questions and response categories, including those regarding employment status and the number of weeks worked in the past 12 months (used to calculate full-time/full-year employment status), that impact estimates related to those topics. Because of these changes, comparisons cannot be made with previous year estimates.[1] This section will focus on the development of the 2008 survey questions currently in use.

A number of disability data users voiced dissatisfaction with the disability questions contained within the Census 2000 and the ACS. By 2003, a number of issues with the questions had been discovered including evidence of apparent misinterpretation of two of the disability questions on the part of respondents (Stern, 2003; Stern & Brault, 2005). As a result, the Office of Management and Budget Interagency Committee created an ACS Subcommittee on Disability Measurement, with the National Center on Health Statistics taking the lead in determining the adequacy of the Census disability questions.  The subcommittee “found that surveys could identify certain aspects of disability and estimate a population who would be likely to experience restrictions in participation due to physical, social and other environmental barriers.”  The subcommittee used the International Classification of Functioning, disability and health (ICF) as a basis for developing a new set of questions. Appendix D contains a description of ICF concepts.

The work group identified four basic areas of functioning that identified the largest proportion of the population with disabilities: vision, hearing, mobility, and cognitive function (Brault, Stern & Raglin, 2007). They also determined that there were elements that could be used to identify service needs and independent living, specifically the ability to move around the community without assistance (independent living) and the ability to bathe and dress oneself (self-care). Versions of these questions were then cognitively tested to determine both the validity and the impact of the use of different terms and qualifiers (see Miller & DeMaio (2006) for more details on the cognitive testing). As a result of these efforts, the subcommittee developed an alternate group of six disability questions. These new questions were designed to better identify specific portions of the population of persons with disabilities, and to clearly define disability as a functional limitation that may increase a person's risk of participation limitation.

The six new disability questions then underwent a content testing and were compared against the 2003 disability questions as a control. A content re-interview test was used to evaluate the variance in question response to investigate longer term measures of disability and functioning.  The proposed questions were found to have a “simple response variance” equal to or lower than the control questions in addition to lower item non-response rates. In conclusion, “the proposed changes result in better questions, in terms of reliability and response and their ability to better identify the population of persons with disabilities” (Brault, Stern, & Raglin, 2006, p. iii).

The six new disability questions were integrated into the 2008 ACS questionnaire and are described in the first section of Table 2. Note that the Census Bureau refers to each of the individual types as “difficulty” while in this report the term “disability” is used.[2] A person is defined as having a hearing disability if they responded “yes” to Q17a and a visual disability if they responded “yes“ to Q17b. Both of these questions (Q17a, Q17b) are asked of all household members regardless of age and are consistent with the impairment concept from the ICF (See Appendix D). People ages 5 and older are asked about physical, mental, or emotional conditions that resulted in serious cognitive impairments (Q18a-cognitive disability) such as concentrating, remembering, or making decisions; ambulatory difficulties (Q18b) - walking and climbing stairs.    

 

 

The ACS also includes a question (Q18c) consistent with the ICF activity limitations concept regarding difficulties in the performance of activities of daily living: difficulty dressing or bathing.  The remaining disability question (Q19) is consistent with the ICF participation restriction concept.  It asks if a physical, mental or emotional condition affects participation in usual life activities such as going outside the home alone to visit a doctor’s office or going shopping (ages 15 and older).

The ACS uses these six questions to identify seven disability categories that are described in Table 1a. The Census Bureau created a seventh category, referred to as a disability, as a “yes” response to one or more of the six disability questions.  This definition is similar to the ICF use of the term disability in that it includes impairments, activity limitations, or participation restrictions.  As noted previously, these six questions are different from those asked in the previous iteration of the ACS in the period 2003-2007.  For more information on the evolution of the disability questionnaire items on the ACS see Appendix C.

In addition to these six defined disability categories in 2008, in that year the Census Bureau also introduced veteran’s service connected disability questions.  These items are for all surveyed persons who have ever served on active duty in the U.S. Armed Forces, military reserves, or National Guard. The first question (Q28a) asks if the person has a VA service–connected disability rating. If the response is “yes,” a follow-up question (Q28b) is asked regarding what the person’s actual service-connected disability rating is (0 percent, 10 or 20 percent, 30 or 40 percent, 50 or 60 percent, 70 percent or higher).

Demographics.  Data on demographics are drawn from the “list of residents” section of the ACS and include age, gender, race, and ethnic origin.  Each household member is listed and their name is associated with the numerical order in that list.  Person 1 is the “householder” - the person living or staying at the address in whose name the house or apartment is owned, being bought or rented. The survey repeats each question for every person in the household who lived or stayed at the address for more than 2 months in the past 12 month period. Table 3a below provides the question text, response categories and the 2009 variable names for each demographic characteristic used in the estimates later in this report. Note that it does not include all the demographic characteristics available in the ACS data (See Table 1 above). 

Table 3a

Demographic Definitions from the 2009 American Community Survey

Census Term

PUMS Variable Name

Question

Population

Relationship

REL

(RELP in 2010)

Q2. How is this person related to Person 1? Mark (X) ONE box.

Husband or wife, Adopted son or daughter, Brother or sister, Biological son or daughter, Stepson or stepdaughter, Father or mother, Grandchild, Parent-in-law, Son-in-law or daughter-in-law, Roomer or boarder, Unmarried partner, Other relative, Housemate or roommate, Foster child, Other nonrelative

 ** REL variable also identifies Institutionalized Group Quarters population and Non-institutionalized Group Quarters population.

*** Variable name changed to RELP in 2010 ACS PUMS data***

All

Gender

SEX

Q3. What is person 1’s sex?

All

Age

AGEP

Q4. What is person 1's age and what is this person's date of birth?

All

Race

(detailed)

RAC2P, RAC3P

Q6. What is person 1's race? Mark (X) one or more boxes.  Responses include the following: White; Black or African-American or Negro; American Indian or Alaska Native (print name of enrolled or principal tribe); Asian Indian; Chinese; Filipino; Japanese; Korean; Vietnamese; Other Asian (Print Race); Native Hawaiian; Guamanian or Chamarro; Samoan; Other Pacific Islander (Print Race Below); Some other race (print race).

All

Census Race Recode

RAC1P

The Census Bureau recoded to the following: White Alone; Black or African American Alone; American Indian Alone; Alaska Native Alone; American Indian and Alaska Native Alone; Asian Alone; Native Hawaiian or Other Pacific Islander Alone; Some other race alone; or two or more races.  Alone means that this category was the only race category selected.  The householder is allowed to select one or more races for a household member.  See Census website for details of race recode.

All

Additional Race Recode

RAC1P recode

American Indian Alone, Alaska Native Alone, and American Indian and Alaska Native Alone are grouped into one category in this paper and called American Indian or Alaska Native.

All

Hispanic Origin

HISP

Q5.  Is person of Hispanic, Latino or Spanish origin?  Responses include the following: No, not of Hispanic, Latino or Spanish origin; Yes, Mexican, Mexican Am., Chicano; Yes, Puerto Rican; Yes, Cuban, Yes, other Spanish/Hispanic/Latino -Hispanic, Latino or Spanish origin Print origin, for example, Argentinean, Colombian, Dominican, Nicaraguan, Salvadoran, Spaniard, and so on.

All

Hispanic Recode

HISP recode

Recoded to 1 if Yes to question Q5 above, 2 if No.

 

 

The ACS includes several questions on education (See Table 3b). Two of the questions are related to recent participation in an educational program (Q10a, b) and a third queries the person’s highest level of educational attainment at the time of the survey. For persons who have a bachelor’s degree there is an additional question regarding the specific major(s) studied (Q12.).


Table 3b

Education Definitions from the 2009 American Community Survey

Census Term

PUMS Variable Name

Question

Population

School enrollment

SCH

Q10a. At any time IN THE LAST 3 MONTHS, has this person attended school or college? Include only nursery or preschool, kindergarten, elementary school, home school, and schooling which lead to a high school diploma or a college degree. Responses include the following::

 No, has not attended in the last 3 months ; Yes, public school, public college; Yes, private school, private college, home school

All

Grade level attending

SCHG

Q10b. What grade or level was this person attending? Responses include the following: Nursery school, preschool; Kindergarten;  Grade 1 through 12 – Specify grade 1 – 12;

College undergraduate years (freshman to senior)

Graduate or professional school beyond a bachelor’s degree (for example: MA or PhD program, or medical or law school)

All

Educational Attainment

SCHL

Q11. What is the highest level of schooling this person has completed? 
If currently enrolled, mark the previous grade or highest degree received.

All

Educational  Attainment     Recode:      

 

SCHL recode

 

All

Less than High   School

 

Responses: No schooling completed; Nursery school; Kindergarten;

                   Grade 1 through 11 –Specify; 12th grade no diploma

 

High School

 

Responses: Regular high school diploma; GED or alternative credential

All

Greater than High School

 

Responses: indicates at least some college or higher education

All

Bachelor’s Degree Major

FOD1P,

FOD2P

Q12. This question focuses on this person’s BACHELOR’S DEGREE. Please print below the specific major(s) of any BACHELOR’S DEGREES this person has received. (For example: chemical engineering, elementary teacher education, organizational psychology) [codes up to 2 majors]

holder of a BA degree

Source: Author's adaptation from ACS website: http://www.census.gov/acs/www/UseData/Def.htm

Employment Measures.  The Census Bureau’s definition of employment status is drawn from two questions.  The employment reference period is defined as the week before the date that the householder completed the questionnaire.   Table 3c describes the ACS information on the employment status of each household member age 15 and older.  A household member is considered employed if he met one of the two following criteria: (1) is “at work” during the reference period—that is, worked as a paid employee, worked in his or her own business or profession, worked on his or her own farm, or worked 15 or more hours as an unpaid worker on a family farm or business; or (2) was “with a job but not at work” during the reference period—that is, he or she had a job but temporarily did not work at that job during the reference period due to illness, bad weather, industrial dispute, vacation or other personal reasons.  Institutionalized people as well as those whose only activity was work around the house or unpaid volunteer work are not classified as employed.

Another measure of employment is referred to as “full-time full year” employment.  It is defined by the Census Bureau as working 50 to 52 weeks in the previous 12 months and at least 35 hours per week during that period. Note that as the ACS is continuously collected the 12 month period usually will span across two years – if completed in July 2009 the period would span the period July 2008 through July 2009.

Table 3c

 

Employment Definitions and PUMS variables (2009 American Community Survey)

 

Census Term

PUMS Variable Name

Question

Ages

Employment status

 

Q29a. LAST WEEK, did this person work for pay

at a job (or business)?

Ages 15 and older

 

 

Q29b. LAST WEEK, did this person do ANY work for pay, even for as little as one hour?

Ages 15 and older

 

 

Q35b. LAST WEEK, was this person TEMPORARILY

absent from a job or business?: Yes, on vacation, temporary illness, maternity leave, other family/personal reasons, bad weather, etc.)

Ages 15 and older

Employment Status recode

ESR

Employed if ESR=1 .Civilian employed, at work

2 .Civilian employed, with a job but not at work

4 .Armed forces, at work 5 .Armed forces, with a job but not at work

Not employed if ESR=3 .Unemployed or 6 .Not in labor force

 

Number of weeks worked

WKW

Q39a.  During the PAST 12 MONTHS (52 weeks), did this person work 50 or more weeks? Count paid time off as work.

Q39b. How many weeks DID this person work, even for a few hours, including paid vacation, paid sick leave, and military service?: 50 to 52 weeks; 48 to 49 weeks; 40 to 47 weeks; 27 to 39 weeks; 14 to 26 weeks;13 weeks or less

 

Ages 15 and older

Number of hours worked

WKHP

Q40. During the PAST 12 MONTHS, in the WEEKS WORKED, how many hours did this person usually work each WEEK? Usual hours worked each WEEK:

Ages 15 and older

Employment Definitions

 

 

 

Employed: Reference Period

ESR recode

The person is classified as employed if they respond “yes” to Q29a, Q29b, or Q35b. [ESR=1,2,4,5]

Ages 15 and older

Employed:                                     Full-time year round

ESR, WKW, WKHP recode

At least 50 weeks during the previous year and at least 35 hours per week.  Determined by condition that weeks worked is greater than or equal to 50 (from Q39a) and usual hours per week is greater than or equal to 35 hours (from Q40).

Ages 15 and older

Source:  Author's adaptation from ACS website:  http://www.census.gov/acs/www/UseData/Def.htm

     

Income and Poverty Data.  The economic well-being measures use information from the ACS on annual income, family size, family composition, household size, and household composition.  Table 3d describes the income measures and summarizes the method used by the Census Bureau to construct a poverty measure.


Table 3d

ACS Economic Well-Being Measures and Related Variables (2009 American Community Survey PUMS data)

Census Term

PUMS Variable Name

Question

Ages

Income

PINCP
(total person’s income)

(Person Section) Q47a-h. Asks the person to list the amount of income received from the following sources: wages, salary, commissions, bonuses, or tips from all jobs (before deductions for taxes, bonds, dues or other items); self-employment income from own non-farm businesses or farm businesses, including proprietorships and partnerships (net income after business expenses); interest, dividends, net rental income, royalty income, or income from estates and trusts; Social Security or Railroad Retirement; Supplemental Security Income (SSI); any public assistance or welfare payments from the State or local welfare office; retirement, survivor or disability pensions (not including social security); and any other sources of income received regularly such as Veterans’ (VA) payments, unemployment compensation, child support or alimony (not including lump sum payments such as money from an inheritance or the sale of a home).

Ages 15 and older

Income and earnings adjustment factor

ADJINC

Adjustment factor for income and earnings dollar amounts  (6 implied decimal places)

Note: The value of ADJINC inflation-adjusts reported income to dollars of the current survey year. ADJINC applies to variables FINCP and HINCP in the housing record, and variables INTP, OIP, PAP, PERNP, PINCP, RETP, SEMP, SSIP, SSP, and WAGP in the person record.

NA

Poverty

POVPIP
(person level)

The Census Bureau used information on the family income and household composition, along with standard poverty thresholds, to construct a poverty measure.  See the Census Bureau website http://www.census.gov/hhes/www/poverty/poverty-cal-in-acs.pdf   for details. 

All ages except unrelated HH members below the age of 15.

Household Size

NP

The sum of all people who the householder reports living in the housing unit.

All ages

Household Income

HINCP

The sum of income for each household member age 15 and older in the household unit.

All ages

Source:  Author's adaptation from ACS website: http://www.census.gov/acs/www/UseData/Def.htm

 

The income measure uses income received in the past 12 months (i.e., income received in the year preceding the completion of the survey) from each individual household member.  For a household that completes the survey in July 2009, the year is July 2008 to June 2009.  The questions are located in the “person” section of the survey.  Questions 47a through 47h are used to collect information on the sources of income.  Table 3d comprehensively lists the types of income included in calculating person’s total income.  The annual total income is the sum of all of the income sources for the household member.

The poverty measure is computed using poverty threshold standards indexed to current dollars for that data year using poverty factors based upon the Consumer Price Index (CPI-U). The family is used as the income sharing unit, with family income being the sum of total income from each family member living in the household.  The poverty threshold depends upon the size and composition of the family, taking into account the age of the householder (i.e., the person who owns or pays rent for the housing unit and who fills out the ACS questionnaire for the household) for one-member families and two-member families, as well as the number of related children under the age of 18.  Family income is compared to the relevant poverty threshold to determine the family’s poverty status.[3] For household members who are unrelated to the head of household, the poverty threshold is based upon the person’s own total income and uses a different threshold.  A poverty measure is not created for unrelated household members who are under the age of 15 because the ACS does not collect income information from persons under the age of 15. 

Note that poverty statistics do not adjust for expenses that are the result of a health condition or a disability (e.g., personal assistance, equipment, and medications).    For these reasons, household income relative to the poverty line is substantially limited as an indicator of a household’s poverty if the household contains a person with a disability.  Additionally poverty measures  do not adjust for in-kind benefits, such as health insurance, food stamps, housing, transportation, and child-care.

Dissemination

The U.S. Census Bureau disseminates thousands of ACS summary data tables to the public for a wide variety of geographic levels including national, state, metropolitan statistical area (MSA), and county.   It also provides tables for larger towns and cities (with total populations of 65,000 or more for single-year estimates and 20,000 for three-year “pooled” estimates), and down to the Census Block level in the five-year pooled estimates. Pooled data utilizes multiple years of data (3 or 5)  to provide a larger sample size that will produce more reliable estimates for smaller geographic areas or smaller populations. Due to the changes made to the disability questions, three-year pooled estimates  (based on ACS 2008-2010 data) are available as of Fall 2011, and five-year estimates will be available in 2014 (based on ACS 2008-2012 data). The ACS summary statistics and detailed data tables can be accessed through the Census Bureau’s American Factfinder website.[4] The Factfinder summary tables provide users with data aggregated to various geographic levels; however, they cover only a few topics and demographic breakdowns, which can limit their usefulness.  

The ACS Public Use Microdata Sample (PUMS) is also available from the U.S. Census Bureau.  The PUMS contains data at the household level and person level, allowing users to produce customized statistics not available from the American Factfinder tables. The ACS PUMS data are a subsample of the full ACS data and include approximately one percent of the housing units and one percent of the GQ persons in the United States and Puerto Rico (the ACS samples approximately 2.5 percent of the population annually). All people residing in PUMS Housing Units (HUs) are included in the PUMS data.  According to the Census Bureau, estimates based on the ACS PUMS files, such as those included in this document, may differ slightly from the ACS summary tables produced by the Census Bureau because they are subject to additional sampling error and further data processing operations[5].

The Census Bureau uses procedures to assure the confidentiality of respondents included in the PUMS data.  The smallest identifiable geographical area available is a Public Use Microdata Area (PUMA) that contains a minimum total population of 100,000 people. In most cases, rural PUMAs are based on counties (or county equivalents), while urban PUMAs are usually based on city boundaries, census tracts or metro areas. Typically a large county of 400,000 would be divided into four PUMAs, each containing approximately 100,000 people. If a county contains fewer than 100,000 people it is generally combined with other adjoining small-population county areas until it meets the area minimum 100,000 person PUMA criteria. All PUMAs are nested within state boundaries and generally each PUMA is geographically contiguous.[6]

PUMS data are available for 2000 through 2010.  Single-year, as well as three- and five-year, PUMS data files are available. The multi-year PUMS provide larger sample sizes in situations where a single year sample is inadequate for analysis; however, the geographic specificity of these multi-year files is still limited to the PUMA level.  Due to changes in the disability questions in 2008, the five-year PUMS files do not include disability at this time. Disability variables will be included in the 2008–2012 five-year PUMS that will be available in in early 2014. For individuals without access to statistical packages or the needed computational resources the Census Bureau’s DataFerrett application (http://dataferrett.census.gov) also allows users to perform basic ACS PUMS data analysis on the web.

Note that other major and minor changes to the ACS survey, data collection and sampling methodologies have occurred over time and these alterations, and their implications, are described below.

Changes to the ACS and Implications

The ACS has evolved over time and will likely continue to do so in the future. Many of these changes resulted in data “seams” that have significant impacts on the measurement of disability. These changes include alterations to sampling, sample size, and  the disability questions, and the Census Bureau recommended not comparing disability estimates across these seams. These changes included:

·       2003: alterations to the structure of the disability questions (see Weathers, 2005)

·       2005: the ACS housing sample size was increased and sampling methodology altered. The previous sample only included addresses in approximately half of the county/county equivalents, while the new sample included addresses in all 3,143 counties/equivalents.

·       2006: institutionalized and non-institutionalized Group Quarters were sampled..

·       2008: major changes to the set of six disability questions and those regarding employment status during the reference week and the number of weeks worked in the past 12 months (used to calculate full-time/full year employment status)

For a more detailed history of the ACS disability question sets, see Appendix C. For a summary of all changes specific to the ACS 2008 survey see the following Census Bureau 2006 content test document: http://www.census.gov/acs/www/Downloads/methodology/content_test/SummaryResultsACS2006ContentTest.pdf   For a synopsis of all ACS question changes across its history see: http://www.census.gov/acs/www/methodology/questionnaire_changes/

The 2008 changes result in lower overall disability estimates that are different from previous years. See Tables C1 and C2 in Appendix C for working age prevalence rate estimates in 2007 and 2008. The revised questions are more focused than the previous versions, allowing more accurate identification of specific issues or problems. The ability to distinguish between visual and hearing impairments was a major improvement, and the ambulatory disability question provides more specific information on persons with mobility issues. Thee new question set does not identify persons with upper body disabilities or those with back problems, however, and people with mental illness or learning disabilities are also unlikely to be identified by any of the current questions. Unless these issues result in self-care or independent living disability, persons with such issues would not be identified as having a disability.

Future Changes

As with all survey instruments, further changes to questionnaire items are likely to occur, methodologies change, and estimates are likely to be affected, just as they were in 2005 when the sample was increased and new sampling methodology was implemented, or when the disability questions were revised in 2003 and 2008. When performing analysis across years with any data source, it is important to carefully examine the specific questions, variable names, and values included in the analysis to confirm that changes have not occurred during that period.  The ACS has seen changes in a number of questions and variables over the years, and it is never safe to assume that specific questions have remained unchanged. The current version of the six ACS disability questions may stabilize now that they (or minor variants) have been integrated into a number of other major surveys including the 2009 American Housing Survey (AHS) and are going to be used in the upcoming fielding of the Survey of Income and Program Participation (SIPP).  A slightly modified version of the six questions was adopted in June 2008 in the monthly Current Population Survey (CPS). [7] This standardization across multiple surveys may make changes to the disability questions less likely. Despite the apparent advantages to utilizing identical questions across different surveys, prevalence rates can vary significantly (Rietschlin & MacKenzie1, 2004). This variation may potentially be due to differences in sampling, data collection methodologies as well as the topical focus of the survey (i.e., health vs. economic focus) among other factors.

ACS Description of Disability Population

In this section we present tables with estimates derived from the ACS PUMS as examples of what is available. Tables include national prevalence rates across all ages, as well as outcomes including employment, poverty, and health insurance calculated specifically for the working-age population (ages 21 to 64). State level estimates illustrate differences across states in prevalence and employment rates as well as group quarter’s estimates. All estimates provided below are based on the author’s analysis of the 2009 ACS PUMS data and most are excerpted from the 2009 Disability Status Reports (Erickson, Lee, & von Schrader, 2011).  See www.DisabilityStatistics.org  for additional ACS PUMS-based estimates including access to the most recent annual Disability Status Reports.

Each table below includes a percentage estimate and a population estimate (“Number”) along with the 90% margin of error (MOE) provided to illustrate sampling variability. The “Base population” column contains the estimated number of individuals upon which the calculation is based. For percentages, this is the denominator. The “Sample Size” is the number of survey participants used to calculate the statistic.

Note that when developing PUMS based estimates it is important to realize that due to the ACS sampling and weighting procedure each person in the ACS PUMS data actually represents a number of people in the population. To adjust for this each person and housing unit has a weight variable associated with it. In order to develop valid PUMS based estimates the data must be “weighted” using the appropriate weighting variable that “inflates” the sample estimate to the full population. The person level weight variable (PWGTP) is applied to person level estimates while the housing level weight (WGTP) is used for household level estimates. 

Prevalence

As shown in Table 4a, according to the ACS, 36,230,100 (+/- 157,290) million persons with a disability lived in the U.S. in 2009 out of a total population of 302,783,200 non-institutionalized persons, a 12.0 +/-0.05% prevalence rate. This estimate is based on a sample size of 2,979,835. Ambulatory disabilities have the highest prevalence rate at 6.9% followed by independent living (5.4%), cognitive disability (4.8%), hearing disability (3.4%), self-care disability (2.6%), and visual disability (2.1%).Note that the disability types will not sum to the total population number or percentage with a disability as individuals may report more than one disability type (i.e., the types are not mutually exclusive). 

 


Table 4a

 

Disability Prevalence among Non-Institutionalized People of All Ages in the United States in 2009

Disability Type % MOE Number MOE Base Pop. Sample

Disability Type

Percent

MOE

Number

MOE

Base Population

Sample Size

Any Disability

12.0

0.05

36,230,100

157,290

302,783,200

2,979,835

Visual

2.1

0.02

6,453,300

69,940

302,783,200

2,979,835

Hearing

3.4

0.03

10,221,000

87,460

302,783,200

2,979,835

Ambulatory

6.9

0.04

19,425,100

118,690

281,613,500

2,799,892

Cognitive

4.8

0.04

13,581,200

100,250

281,613,500

2,799,892

Self-Care

2.6

0.03

7,189,100

73,730

281,613,500

2,799,892

Independent Living

5.4

0.04

13,041,100

98,330

240,963,700

2,414,589

 

Table 4b below provides prevalence rates by age group. Note that the two younger groups are not asked all six disability questions. There is a steady increase in disability prevalence rates as the population ages from 10.4% of the working age population, to a quarter (26.0%) of those ages 65-74 and half of those 75 or older (50.8%).

 

Table 4b

Disability Prevalence across Age Groups (U.S. Non-Institutionalized Population, 2009)

Age

Percent

MOE

Number

MOE

Base Population

Sample Size

0-4*

0.7

3.29

157,000

11,020

21,169,700

179,963

5-15*

5.1

0.09

2,299,900

42,040

44,761,000

426,632

16-20

5.5

0.14

1,215,700

30,620

21,930,700

201,911

21-64

10.4

0.06

18,382,600

115,670

177,004,700

1,709,245

65-74

26.0

0.27

5,346,000

63,770

20,550,800

253,484

75 and older

50.8

0.33

8,828,900

81,480

17,366,300

208,620

* Note: Children under the age of five were only asked about vision and hearing disabilities. The independent living disability question was only asked of persons aged 16 years old and older.

 

Table 4c illustrates the prevalence rates for the working-age population by gender, race, and Hispanic/Latino origin. In this age group, disability prevalence is the same for both males and females. Native Americans or Alaskan natives as a category have the highest prevalence rate, with nearly one in five (18%) having a disability as compared to only one in twenty Asians (4.5%). Persons of Hispanic/Latino origin have slightly lower disability prevalence rates than non-Hispanic/Latinos.

Table 4c

 

U.S Disability prevalence by demographic characteristics (U.S. Non-Institutionalized Population ages 21-64, 2009)

 

Characteristic

Percent

MOE

Number

MOE

Base Population

Sample Size

Gender-Male

10.4

0.09

9,107,500

82,720

87,396,200

822,262

Gender-Female

10.4

0.09

9,275,100

83,450

89,608,500

886,983

Race-White

10.1

0.07

13,570,800

100,210

133,865,000

1,349,321

Race-Black/African American

14.1

0.21

2,976,800

47,770

21,165,500

164,366

Race-Native American or
Alaska Native

18.0

0.91

248,400

13,860

1,383,800

14,774

Race-Asian

4.5

0.19

391,500

17,400

8,756,800

83,646

Race-Some other race(s)

10.1

0.24

1,195,100

30,360

11,833,600

97,138

Race-Hispanic/Latino

8.3

0.15

2,162,900

40,780

26,052,200

210,562

Race-Not Hispanic/Latino

10.7

0.07

16,219,600

109,060

150,952,500

1,498,683

 

A “service-connected disability” is one that has been determined by the Department of Veterans Affairs (VA) to be being a result of disease or injury incurred or aggravated during military service. Table 4d shows that about one in six (17.5%) working-age veterans have a service-connected disability. Nearly a half million (447,800) working-age civilian veterans in the U.S. had the most severe service-connected disability rating (70% or above). Note that a veteran can receive disability compensation for a wide range of conditions, and a veteran with a service-connected disability may not report having one of the six ACS functional or activity limitation disabilities.


Table 4

 

Civilian Veterans with a Service-Connected Disability (U.S. Non-Institutionalized Population, ages 21-64, 2009)

 

Service–Connected Disability

Percent

MOE

Number

MOE

Base Population

Sample Size

Has any service-connected disability rating (0-100%)

17.5

0.21

2,201,900

29,040

12,609,500

132,558

0 percent Disability Rating

6.9

0.34

153,000

7,680

2,201,900

23,718

10 or 20 percent Disability Rating

35.6

0.63

783,700

17,370

2,201,900

23,718

30 or 40 percent Disability Rating

17.6

0.50

386,400

12,200

2,201,900

23,718

50 or 60 percent Disability Rating

10.7

0.41

235,900

9,540

2,201,900

23,718

70 percent or higher Disability Rating

20.3

0.53

447,800

13,130

2,201,900

23,718

 

Figure 1 shows the educational attainment of the working-age population by disability status. Nearly one-quarter of persons with disabilities have less than a high school education as compared to only one in ten persons without disabilities. Nearly one-third of persons without disabilities have a Bachelor’s degree or higher, while slightly more than one in ten persons with disabilities have a BA or higher degree.

Figure 1.  Educational Attainment by Disability Status (U.S. Non-Institutionalized Population ages 21-64, 2009)

 

Educational attainment by disability status

 

With Disability:          

23.9%  Less than High School

 34%   High School Diploma/Equivalent

29.9%  Some College/Associates degree

12.2%  BA degree or higher

 

Without Disability     

10.8%  Less than High School

26.1%  High School Diploma/Equivalent

32.3%  Some College/Associates degree

30.8%      BA degree or higher

 

End Figure 1

ACS Employment and Economic Well-Being Estimates

The 2009 ACS shows that the employment rates for persons with a disability are lower than the employment rates for persons without a disability.  Table 5a shows this result for each of the three employment measures for the working-age population.  The first section shows that while 76.8% of the population without a disability was employed during the reference week, only 36.0% of the population with a disability was employed during that period.  The same pattern holds for full-time/full-year employment rates with 57.0% of persons without disabilities working FT/FY as compared to 22.5% of persons with disabilities.

Table 5a

 

Employment by disability status (U.S. Non-Institutionalized Population ages 21-64, 2009)

 

Disability Status

Percent

MOE

Number

MOE

Base Population

Sample Size

Reference period employment Without Disability

76.8

0.09

121,778,200

237,410

158,622,200

1,528,114

Reference period employment With Disability

36.0

0.31

6,612,900

70,770

18,382,600

181,131

Full time/Full Year Employment Without Disability

57.0

0.08

90,418,700

169,940

158,622,200

1,528,114

Full time/Full year Employment With Disability

22.5

0.21

4,130,100

42,950

18,382,600

181,131

 

The economic well-being of the population with disabilities is substantially worse than that of the population without disabilities based upon the two measures presented in Table 5b.  The annual median household income for households that include any working-age person with a disability is $37,200, as compared $60,000 for households that do not contain any working-age persons with a disability. As would be expected, poverty rates are much higher for persons with disabilities, with 26.4% of the population with a disability living below the poverty line as compared to 10.8% of the population without a disability. 

 


Table 5b.1

 

Economic well-being by disability status (U.S. Non-Institutionalized Population ages 21-64, 2009)

 

Annual Household Income

Median

MOE

 

 

Base Population

Sample Size

Without Disability

$60,000

$190

 

 

79,747,000

798,887

With Disability

$37,200

$350

 

 

14,718,000

155,378

 

Table 5b.2

 

Economic well-being by disability status (U.S. Non-Institutionalized Population ages 21-64, 2009)

 

Poverty

Percent

MOE

Number

MOE

Base Population

Sample Size

Without Disability

10.8

0.07

16,987,200

111,460

158,005,900

1,522,895

With Disability

26.4

0.29

4,846,900

60,770

18,349,300

180,847

 

ACS Group Quarter (GQ) Population Estimates

 As noted previously another advantage of the ACS is that it includes within its sampling frame both institutionalized and non-institutionalized populations. All estimates prior to this only have included the non-institutionalized population: those residing in housing units or in non-institutionalized GQs. This distinction is made as the institutionalized population is generally not able to participate in activities such as employment, or eligible for the benefits or programs that serve the non-institutionalized population. Table 6 breaks this sample into the three types of housing: housing units, non-institutionalized GQs and institutionalized GQs provides estimates and sample sizes for each housing type. Approximately 298.7 million people live in households, or 97.3% of the total population, while an additional 4 million live in non-institutionalized GQs and 4.2 million live in institutions. The prevalence rate varies greatly across these different housing types. About one in ten persons living in households have a disability (11.8%), while one in five (21.1%) of non-institutionalized GQ residents have a disability. Over half (57.2%) of institutionalized persons have a disability, accounting for 6.3% of the total population with disabilities living in the U.S.

 

Table 6

U.S. Total population and population with disabilities by residence type (all ages, 2009)

 

Housing Unit Population

Group Quarters Population Non-Institutionalized

Group Quarters Population Institutionalized

Total Population

Sample Size

2,947,932

31,903

50,893

3,030,728

Population

298,700,000

4,053,800

4,223,300

307,000,000

Percent of population

97.3

1.3

1.4

100.0%

Population with disabilities

35,370,000

856,400

2,414,700

38,641,100

Prevalence rate

11.8

21.1

57.2

12.6

Percent of population with disabilities

91.5

2.2

6.3

100.0%

 

 

ACS State Level Estimates

One of the strengths of the ACS is the ability to develop estimates at the state level due to the large sample size the data provides. This section provides some of state estimates of prevalence and employment rates based on the 2009 PUMS data. As can be seen in Figure 2 below there is wide variability in the prevalence rate between states for the non-institutionalized population ages 21-64. The four states with the lowest prevalence rates are Hawaii (7.7%), New Jersey (7.8%), Colorado (8.2%) and Illinois (8.2%). The states with the highest prevalence rates are: Alabama (15.5%), Kentucky (16.4%), Arkansas (17.0%), and West Virginia (18.4%). Puerto Rico has an even higher prevalence rate of 19.1%.  For comparison the overall U.S. prevalence rate is 10.1%. See Appendix E for prevalence by state data table.

 

Figure 2. Disability Prevalence rate by state (Non-Institutionalized Population ages 21-64, 2009)

 

US 10.4

Alabama 15.5

Alaska 12.0

Arizona 10.3

Arkansas 17.0

California 8.4

Colorado 8.2

Connecticut 8.6

Delaware 11.2

District of Columbia 10.0

Florida 9.9

Georgia 10.5

Hawaii 7.7

Idaho   11.2

Illinois 8.2

Indiana            11.3

Iowa    9.4

Kansas 10.6

Kentucky 16.4

Louisiana 13.0

Maine 14.4

Maryland 8.4

Massachusetts 9.2

Michigan 11.9

Minnesota 8.4

Mississippi 15.1

Missouri 12.6

Montana 11.8

Nebraska 9.2

Nevada 9.0

New Hampshire 9.2

New Jersey 7.8

New Mexico 12.3

New York 9.1

North Carolina 11.7

North Dakota 9.2

Ohio 12.0

Oklahoma 15.2

Oregon            11.2

Pennsylvania 11.0

Puerto Rico 19.1

Rhode Island 10.2

South Carolina 12.2

South Dakota 9.7

Tennessee 13.9

Texas 10.3

Utah 8.3

Vermont 11.4

Virginia 9.3

Washington 10.7

West Virginia 18.4

Wisconsin 8.9

Wyoming 11.1

 

End Figure 2

 

Employment rates also vary greatly by state, as Figure 3 demonstrates. The states with the lowest employment rates for working age (21-64) persons with disabilities are West Virginia (28.0%), Kentucky (28.1%), the District of Columbia (28.3%) and Mississippi (28.7%). The states with the highest employment rates are Alaska            (50.2%), Wyoming (50.3%) and North Dakota (56.1). See Appendix F for employment rates for all states. As overall employment rates vary by state due to differences in their economic environment and opportunities, one of the most useful measures of the differences is the employment ratio. The employment ratio is calculated by dividing the employment rate of those with disabilities by the rate of those without disabilities. The higher the ratio the better the relative employment rate is, with a value of 1.0 meaning that persons with disabilities are employed at the same rate as those without. The overall employment ratio in the U.S. is .47, which means that people with disabilities are employed at slightly less than half the rate of those without disabilities.  As can be seen in Figure 3, the employment ratio varies significantly between states. The employment ratio ranges from a low of .36 in the District of Columbia to a high of .66 in North Dakota.

 

Figure 3. Employment ratio by state (Non-Institutionalized Population ages 21-64, 2009)

United States 0.47

Alabama 0.40

Alaska 0.65

Arizona 0.46

Arkansas 0.45

California 0.46

Colorado 0.59

Connecticut 0.52

Delaware 0.45

District of Columbia 0.36

Florida 0.44

Georgia 0.46

Hawaii 0.51

Idaho 0.51

Illinois 0.47

Indiana 0.47

Iowa 0.56

Kansas 0.55

Kentucky 0.38

Louisiana 0.46

Maine 0.42

Maryland 0.54

Massachusetts 0.44

Michigan 0.42

Minnesota 0.54

Mississippi 0.38

Missouri 0.45

Montana 0.58

Nebraska 0.54

Nevada 0.54

New Hampshire 0.48

New Jersey 0.52

New Mexico 0.50

New York 0.44

North Carolina 0.46

North Dakota 0.66

Ohio 0.44

Oklahoma 0.51

Oregon 0.51

Pennsylvania 0.46

Rhode Island 0.45

South Carolina 0.40

South Dakota 0.55

Tennessee 0.40

Texas 0.52

Utah 0.56

Vermont 0.53

Virginia 0.50

Washington 0.51

West Virginia 0.38

Wisconsin 0.49

Wyoming 0.61

Puerto Rico 0.40

 

End Figure 2

 

Summary and Conclusions

The design of the ACS provides several advantages over other data collection efforts.  First, it has gone through a rigorous testing phase.  Second, the survey methodology and design result in a high response rate, with estimates supported at the national, state, Metropolitan Statistical Area, and county level. Third, the redesigned disability questions were developed by a federal inter-agency workgroup and subjected to cognitive testing.  Fourth, the sample includes both institutional and non-institutional group quarters, which few other surveys include. Finally, the ACS provides a wealth of other potentially interesting data including housing information (i.e., rent, utility costs, etc.) and household composition, as well as employment and economic well-being indicators beyond those presented within this User Guide.

The ACS Public Use Microdata Sample estimates allow users to better understand the population with disabilities, including the size of the population, the prevalence rate, the demographic composition, the employment rate, and economic well-being measures.   At the national level, the ACS estimates that there are approximately 303 million non-institutionalized people of all ages in the U.S., of whom 36.2 million are people with a disability, for a disability prevalence rate of approximately 12.0%. The majority of institutionalized persons have a disability, accounting for an additional 4.2 million persons. Compared to the population without disabilities, the population with disabilities is older, more likely to be of African American and Native American decent, and more likely to have an education below the high school level.  The employment and economic well-being measures indicate major and persistent disparities between the population with disabilities and the population without disabilities. At the state level, the ACS estimates show significant differences in the prevalence of disability, as well as employment levels. The differences in employment rates exist both in absolute terms and relative to the population within the state without a disability.

The ACS allows users to examine trends over time.  However, due to major changes in the questions used to identify disability as well as changes to the sampling design currently limit the usefulness of the time span available. Time trends from 2008 onwards will be valuable to researchers in the future.

The ACS is a valuable source of disability information.  As the ACS data collection effort continues, researchers and policymakers will be able to track changes in prevalence rates, as well as employment and economic indicators across states and over time.  These differences may provide vital information on how the labor market environment, the social environment and the policy environment influence the employment and economic well-being of the population with disabilities. The use of the ACS to monitor the progress of the population with disabilities will be an important component of the nation’s efforts to reach the goals of full participation, independent living, and economic self-sufficiency for the population with disabilities.


References

Beaghen, M. & Stern, S., (2009). Usability of the American Community Survey Estimates of the Group Quarters Population for Substate Geographies. 2009 Joint Statistical Meetings: Proceedings of the Survey Research Methods Section. American Statistical Association.  Retrieved from  http://www.amstat.org/sections/srms/proceedings/y2009/Files/303932.pdf.  

Brault, M., Stern, S., & Raglin, D. (2007). Evaluation report covering disability: 2006 American Community Survey content test report. Washington, DC: U.S. Census Bureau. Retrieved from http://www.census.gov/acs/www/Downloads/methodology/content_test/P4_Disability.pdf.

Brault, M. (2009) Review of Changes to the Measurement of Disability in the 2008 American Community Survey. Washington, DC: U.S. Census Bureau. Retrieved from http://www.census.gov/hhes/www/disability/2008ACS_disability.pdf

Erickson, W., Lee, C., & von Schrader, S. (2011). 2009 Disability status report: United States. Ithaca, NY: Cornell University Employment and Disability Institute (EDI).

Jordan, J., & Beaghen, M. (2011). Analysis of the variances of American Community Survey estimates of the group quarters population. Retrieved from http://www.census.gov/acs/www/Downloads/library/2011/2011_Jordan_01.pdf.

Mathiowitz, N. (2000).  Methodological issues in the measurement of work disability. In N. Mathiowetz & G. Wunderlich (Eds.), Survey of measurement of work disability: Summary of a workshop (pp. 28-52). Washington DC: National Academy Press.

Miller, K. & DeMaio, T. (2006). Report of cognitive Research on proposed American Community Survey disability questions. Retrieved from http://www.census.gov/srd/papers/pdf/ssm2006-06.pdf.

Rietschlin, J. & MacKenzie1, A. (2004). Statistics Canada International Symposium Series proceedings: 2004 innovative methods for surveying difficult-to-reach populations. Retrieved from http://www5.statcan.gc.ca/bsolc/olc-cel/olc-cel?lang=eng&catno=11-522-XIE2004001.   

Social Security Advisory Board. (2001). Charting the future of Social Security’s disability programs: The need for fundamental change. Washington DC: Social Security Advisory Board. Retrieved from http://www.ssab.gov/publications/disability/disabilitywhitepap.pdf.

Social Security Advisory Board. (2003). The Social Security definition of disability. Washington DC:  Social Security Advisory Board. Retrieved from http://www.ssab.gov/documents/socialsecuritydefinitionofdisability.pdf.

Stern, S. (2003). Counting people with disabilities: How survey methodology influences estimates in the Census 2000 and the Census 2000 Supplementary Survey: Census Bureau Staff Research Report. Washington DC: U.S. Census Bureau, Poverty and Health Statistics Branch. Retrieved from http://www.census.gov/hhes/www/disability/finalstern.pdf.

Stern, S. & Brault, M.  (2005). Disability data from the American Community Survey: A brief examination of the effects of a question redesign in 2003. Washington DC: U.S. Census Bureau, Housing and Household Economic Statistics Division Working Paper. Retrieved from http://www.census.gov/hhes/www/disability/ACS_disability.pdf.

Weathers, R., R. (2005, May). A Guide to Disability Statistics from the American Community Survey. Rehabilitation Research and Training Center on Disability Demographics and Statistics, Cornell University, Ithaca, NY.
http://digitalcommons.ilr.cornell.edu/edicollect/123

World Health Organization. (2001). International Classification of Functioning, Disability, and Health..  Geneva, Switzerland: World Health Organization. Retrieved from http://www.who.int/classifications/icf/en/.

U.S. Census Bureau (2007). New and Modified Content on the 2008 ACS Questionnaire: Results of Testing Prior to Implementation: Retrieved from http://www.census.gov/acs/www/Downloads/methodology/content_test/SummaryResultsACS2006ContentTest.pdf

 

U.S. Census Bureau. (2009). American Community Survey Puerto Rico Community Survey 2009 subject definitions. Retrieved from http://www.census.gov/acs/www/Downloads/data_documentation/SubjectDefinitions/2009_ACSSubjectDefinitions.pdf .

U.S. Census Bureau. (2009). American Community Survey accuracy of the data. Retrieved from http://www.census.gov/acs/www/Downloads/data_documentation/Accuracy/ACS_Accuracy_of_Data_2009.pdf.

U.S. Census Bureau (2012).  American Community Survey: Questions on the form and why we ask.   Retrieved from http://www.census.gov/acs/www/about_the_survey/questions_and_why_we_ask/

 

 

 


Appendix A: Quick Start to Analysis of the ACS PUMS Data

General ACS information:

 

Main Census Bureau ACS website:

http://www.census.gov/acs/www/

 

Questionnaire archive:

Contains English and Spanish pdfs of the household and GQ questionnaires as well as the associated instruction guides for all years of the ACS.

               http://www.census.gov/acs/www/methodology/questionnaire_archive/

 

Downloading and working with PUMS data

 

A Compass for Understanding and Using American Community Survey Data: What PUMS Data Users Need to Know (Feb. 2009):

An excellent resource describing the basics of the ACS and how to work with the ACS PUMS files.

 http://www.census.gov/acs/www/Downloads/handbooks/ACSPUMS.pdf

 

PUMS documentation: http://www.census.gov/acs/www/data_documentation/pums_documentation/

·               Data dictionary: Provides all variable names, response categories and values for both housing records and person records available in the PUMS

·               Accuracy of the PUMS: Describes ACS and PUMS sample design, sampling and non-sampling error, data confidentiality and disclosure avoidance measures taken. Also includes: weighting and estimation methodology, standard error calculation including both replicate weight and design factor methods, how to work with dollar amounts, and national and state level design factors (used for calculating generalized standard errors).

 

Data downloading: http://www.census.gov/acs/www/data_documentation/pums_data/

ACS PUMS data can be downloaded for the entire U.S., or by individual state, including D.C. and Puerto Rico in two formats: SAS data files or comma separated value (.CSV) files. There are two separate files for each geography:

·         Population Records: 1 record/person

·         Housing Unit Records: 1 record/ Housing unit or GQ person

National level PUMS files: Due to its size both the person and housing data files at the national level are split into two sub-files: a and b.

 PUMS data can also be downloaded using the Census Bureau’s DataFerrett application. (FERRETT stands for Federated Electronic Research, Review, Extraction, and Tabulation Tool.) DataFerrett provides basic data analysis functionality including the ability to recode data, develop customized tables and create data graphs and maps. It is particularly useful for those without access to statistical packages. http://dataferrett.census.gov/

 

Creating Accurate PUMS Estimates: New users of the ACS should fully explore the PUMS documentation to understand how to create estimates from the PUMS data and Standard Error calculations. In particular users should familiarize themselves with the following features of the data.

Population and housing weights:

The appropriate weight variable should always be used when performing analyses:

·         Person level weight (for person level analysis i.e. age, disability): PWGTP

·         Housing level weight (for housing level analysis i.e. household income): WGTP

To ensure proper use of the weights, compare estimates generated to the “PUMS estimates for user verification” available on the PUMS documentation page (see link above).

 

Standard Errors: See Accuracy of the PUMS documentation for calculation methodology

       http://www.census.gov/acs/www/data_documentation/pums_documentation/

 

Identifying housing unit and Group Quarters[8] (GQ) samples:

·         Population record file: Use relationship variable: REL (renamed RELP in 2010 PUMS)

·         Housing unit record file: Use type of unit variable (TYPE)

 

Dollar amounts:

Census Bureau provides an inflation adjustment factor variable to multiply and standardize dollar amount variables. There are two different adjustment factors provided which one is used depends upon the specific dollar amount variable to be adjusted. See data dictionary for more information: http://www.census.gov/acs/www/data_documentation/pums_documentation/

 

Merging person and housing level files:

The variable SERIALNO is used as a key to merge the two person and housing level files together (after sorting by SERIALNO). SERIALNO is a unique Housing unit/GQ person serial number. Note that some housing units may be vacant. There is no need to merge unless the analysis requires both housing and person data – the PUMS data files are large.

 

Public Use Microdata Area (PUMA) Resources:

State level PUMA maps can be found here:

·         PDF versions: http://www.census.gov/geo/www/maps/puma5pct.htm

·         ArcView Shape files: http://www.census.gov/geo/www/cob/pu5_2000.html#shp

 

     Identifying where specific locations fall within PUMAs:

The Missouri Census Data Center (MCDC) MABLE/Geocorr2010 is a Geographic Correspondence Engine that provides a crosswalk between a wide variety of geographical areas and Census PUMAs. Very useful in determining in which PUMA(s) a specific city, county or other area is located and understanding how the total population is distributed across PUMAs or areas within a PUMA(s): http://mcdc1.missouri.edu/MableGeocorr/geocorr2010.html


Appendix B.  ACS Sample Design and Computation of Standard Errors

 

The population estimates reported in the paper are drawn from a sample and, as with any sample, are subject to sampling and non-sampling error.  Standard errors and Margin of Error (MOE) are used to describe the magnitude of sampling error and some forms of non-sampling error.  The formulas used to compute standard errors and MOE must take into account the sample design. 

The purpose of this technical appendix is to provide a brief description of the ACS sample design as well as the ACS PUMS sample design.  It also provides links to Census Bureau documentation regarding how to compute standard errors that account for the ACS and ACS PUMS sample design.  Standard errors may be used to construct the Margin of Error.  The Census Bureau uses 90% MOE in their tables.      

Sample Design

 

The 2009 ACS sample is designed to identify approximately 3 million housing unit addresses (HUs) in the U.S., 36,000 HUs in Puerto Rico (for the PRCS) and approximately a 2.5 percent sample of the expected number of persons residing in GQ facilities.[9]  The sampling frame used for these is the Census Bureau’s Master Address File (MAF) - the Bureau’s inventory of known living quarters. HU address samples were selected independently for each of the 3,142 counties/county equivalents. There are two phases involved in the HU address sampling. The first phase identifies the sample of HUs to be surveyed. In the second phase, a sub-sample of HUs who have not responded to the mailed survey or the telephone follow-up, are selected for Computer Assisted Personal Interviewing (CAPI).

First Phase Household sampling:

 

·      First Stage Sampling: Each address in the MAF sampling frame is assigned to one of the five representative sampling subframes by block. Each subframe is associated with a specific year. This subframe undergoes further sampling as described below. The purpose of these subframes is to meet the requirement that no address can be in the sample more than once in a five-year period. Also incorporated is a 20 percent sample of “new” HUs (not previously included in the MAF).

·      Assignment of blocks to a second-stage sampling stratum: The sampling rates are determined for each stratum for the current sample year. Sampling rates are determined for most blocks of addresses based on the Measure Of Size (MOS) for the smallest sampling entity to which the any part of the block belongs. The seven sampling strata/entities include counties, school districts, places with active and functioning governments, American Indian Areas/Alaska Native Areas/Hawaiian Home Lands (AIANHH), American Indian Tribal Subdivisions with active and functioning governments, minor civil divisions (MCDs) with active and functioning governments in 12 states, and Census designated places in Hawaii only. The overall sampling rate are calculated to yield a target sample size of approximately 3 million HU addresses in the U.S., and 36,000 in Puerto Rico.

·      Second Stage sampling rate calculation: A systematic sample of addresses is selected from the second-stage universe within each county/county equivalent.

·       Sample month assignment: Each of the selected addresses identified in the second-stage sampling is randomly allocated to one of the 12 months of the year for data collection.

Second-Phase Household Sample Selection:

·      This involves the subsampling of un-mailable and/or non-responding addresses. All un-mailable addresses are subsampled at the rate of 2 out of 3. Non-respondent addresses are sampled at rates ranging from 1-in-2 and 1-in-3, with higher rates applied to areas with lower completion levels.

 

Group Quarters (GQ) sample selection:

All GQs are divided into three strata: Small GQs (with 15 or fewer residents), GQs that were closed on Census day 2000, and large GQ (greater than 15 residents). There are approximately 105,000 small GQ facilities and 73,000 large GQs and an additional 2,400 of unknown size. The small GQ and those closed on Census day 2000 were combined into a single stratum. The small and large GQs had different sample selection methods as described below.

Small GQs:

Small GQs are sampled similarly to the HU’s. Address sample and data are collected for all people in the facility. There are two phases: the first identifying the small GQs to be included in the survey and the second phase determining the residents within the GQ to be surveyed in cases where the GQ has a resident population greater than 15.

The first phase sample involves two stages:

·         First-stage sample: As with the HU sample small GQs are only eligible for the ACS once every five years. To ensure this all small GQs are systematically assigned to one of five partitions and each of the partitions is rotated over the five year period, becoming the universe for the second stage sample.

·         Second-stage sample: GQs are selected from the first stage sample based on the state’s target sampling rate. In 2009, most states had a 2.5 percent target sampling rate, while 16 states had higher rates (up to 7.11% in Wyoming) to meet Census Bureau publication thresholds.

·         Second phase sample: All residents are eligible to receive a survey if the actual count is 15 or fewer residents. If the actual count exceeds 15, a field subsample of 10 is randomly selected.

 

Large GQs:

 

Unlike small GQs and HUs, all large GQs are eligible for sampling every year so there is no partitioning stage. The ultimate sampling units for large GQs are not the facility itself, but rather the people selected in groups of ten to be interviewed within the facility.

·         First phase sample: For each large GQ, a GQ Measure of Size (GQMOS) is calculated by dividing the expected population by 10. Groups of 10 people are identified for interview and the number of these groups in a GQ is determined by its GQMOS. For example a GQ with 500 individuals would have approximately 50 groups. The sampling rate across these GQs is based on the state GQ targeted sampling rates. Very large GQs are likely to have multiple “hits” (groups of ten interviews).

·         Second phase sample: selection of persons within large GQs. A random selection of a subsample of ten people from the roster is made for each “hit” on a large GQ and all individuals in this subsample receive a survey.

       

More details regarding the 2009 ACS sample may be found in the document “Accuracy of the Data (2009)” (http://www.census.gov/acs/www/Downloads/data_documentation/Accuracy/ACS_Accuracy_of_Data_2009.pdf ).  Also see chapters 3 (Frame development) and 4 (sample design and selection) in the ACS Design and Methodology document. (http://www.census.gov/acs/www/Downloads/survey_methodology/acs_design_methodology.pdf )

As noted previously the ACS has a very large sample size, second to only the Decennial Census.  Initially, 2,897,256 HU addresses in the U.S. were selected to be potential sample members in 2009, and that resulted in 1,917,748 final completed household interviews. Note that some of the initial addresses were commercial or non-existent and were not interviewed, while others were non-respondents.  After excluding addresses determined to be commercial or non-existent the response rate in the ACS is very high for both HUs and GQ persons, ranging from 97%-98% over the period between 2005-2009.

ACS Public Use Microdata Sample (PUMS).  The ACS PUMS file consists of a sample drawn from the ACS sample.  The 2009 PUMS data contains 1,314,006 housing unit records and 2,979,656 person records from households and 83,248 person records from GQs. The basic unit for the GQ sample is the individual while the basic unit for the housing unit is the individual household. The population sample includes all persons living in the selected households plus persons selected from the GQ sample.

After stratification and sorting, the Census designed a systematic method of selecting household units and GQ persons.  Household level weights and person level weights were then constructed in the PUMS to allow a user to create population estimates.

To further assure the privacy of individual and household information, the U.S. Census Bureau limits the geographic detail provided in the files. The smallest identified unit is the Public Use Microdata Area (PUMA). The Census Bureau also applies “confidentiality edits”  The confidentiality edit involves introducing a small degree of uncertainty into the estimates of ACS characteristics.  This includes matching person records based upon a set of key characteristics and swapping their data.   The methods used maintain the quality and usefulness of the data.

Sampling and Non-Sampling Error

 

Both sampling error and non-sampling errors introduce some degree of uncertainty into estimates.  Sampling error occurs when population characteristics are estimated based upon a sample and are not based upon the entire population.  Because many samples may be drawn from a population, and each sample can produce a different estimate, there is always some degree of uncertainty when samples are used to estimate characteristics of a population.  The variability of estimates drawn from samples, sometimes referred to as uncertainty, is described by standard errors.  Standard errors are used to construct confidence intervals, which describe the likelihood that a particular estimate falls within a certain range of estimates.

Non-sampling error results from other forms of error and includes errors keying in data, errors editing the data, misinterpretation of questions by respondents, non-random non-response to the survey or survey questions, and other factors.  To the degree that the error occurs at random, additional variability will arise in the estimates and the standard errors will describe the variability due to this non-sampling error.  However, non-sampling errors may occur in a systematic manner (i.e., non-random errors).  Systematic errors that arise in the data collection process are not described by standard errors.  Thus, it is important to assess the role of systematic non-sampling errors that may arise in an estimate. 

The Census Bureau attempts to minimize systematic errors by researching and analyzing new sampling techniques, questionnaire designs, and data collection and processing procedures. The ACS also uses other methods to minimize systematic error, such as following up on mail non-respondents during the CATI and CAPI phases.  Information on potential ACS non-sampling errors that are identified by the Census Bureau are posted on the ACS website under “errata,” which may be found at the following address: http://www.census.gov/acs/www/UseData/Errata.htm.

Calculation of Standard Error

 

Due to the complex sample design of the ACS special techniques are required to properly calculate standard errors (SE). There are two basic methods that the Census Bureau provides for calculating SEs: the direct standard error and the more basic generalized standard error. The generalized standard error calculation utilizes the design factors that the Census Bureau provides and is the technique used for developing the SE and Margin of Errors (MOE) provided in this guide. A more accurate but more computationally intensive method is referred to as direct standard errors. This method utilizes the 80 replicate weights associated with each person and household level observation in the PUMS data. For full details of the ACS PUMS design and information on the methods for calculating standard errors, design factor tables, and examples of SE calculations see “PUMS Accuracy of the Data (2009).”  (http://www.census.gov/acs/www/Downloads/data_documentation/pums/Accuracy/2009AccuracyPUMS.pdf .)


Appendix C. ACS Disability Question Set History

            Since the ACS was first fielded in 2000, there have been three variants of the disability questions. In 2003 the question structure was altered to address some issues discovered with the original questions. In 2008 the entire set of six questions was completely replaced with a new set. This new set of questions has been used up through 2012 at the time of this writing. 

ACS 2000-2002

 

            Between 2000 and 2002, the disability questions used the Census 2000 wording and structure:

Q15.  Does this person have any of the following long lasting conditions:

a.       Blindness, deafness, or a severe vision or hearing impairment?

b.      A condition that substantially limits one or more basic physical activities such as walking, climbing stairs, reaching, lifting, or carrying?

Q16.  Because of a physical, mental, or emotional condition lasting 6 months or more, does this person have any difficulty in doing any of the following activities:

a.       Learning, remembering, or concentrating?

b.      Dressing, bathing, or getting around inside the home?

{Page break on survey}

c.       (Answer this if the person is 16 YEARS OLD OR OVER.) Going outside the home alone to shop or visit a doctor's office?

d.      (Answer this if the person is 16 YEARS OLD OR OVER.) Working at a job or business?

 

An analysis of the data by Stern and Brault (2005) suggests that some of the people responding to Questions 16c and 16d may not have understood that it was linked to the introductory sentence on the previous page, “Because of a physical, mental, or emotional condition lasting 6 months or more, does this person have any difficulty in doing any of the following activities…?”  Some sample members may have responded as if the introductory sentence did not exist, which can lead to a different interpretation of the question.  For example, without the introductory sentence, a person may misinterpret Question 16d as instead asking, “Are you working at a job or business?” A “yes” to this interpretation of the question would indicate that they are currently working, not that they have a health condition that makes it difficult for them to work at a job or business.  Therefore, it is possible that these last two questions may have incorrectly identified some people without a disability as being people with a disability.  

ACS 2003 changes:

            In an attempt to address these issues described above the Census Bureau restructured the disability questions in 2003 as follows:

Q15.  Does this person have any of the following long lasting conditions:

c.       Blindness, deafness, or a severe vision or hearing impairment?

d.      A condition that substantially limits one or more basic physical activities such as walking, climbing stairs, reaching, lifting, or carrying?

Q16.  Because of a physical, mental, or emotional condition lasting 6 months or more, does this person have any difficulty in doing any of the following activities:

a.       Learning, remembering, or concentrating?

b.      Dressing, bathing, or getting around inside the home?

Answer Question 17 only if this person is age 15 or older.  Otherwise skip to question for Person 2 on page 10.

Q17.  Because of a physical, mental, or emotional condition lasting 6 months or more, does this person have any difficulty in doing any of the following activities:

a.       Going outside the home alone to shop or visit a doctor's office?

b.      Working at a job or business?

While this change may appear to be minor, there was a major change in the prevalence rates for the employment disability and the go-outside-the-home disability. Significantly lower employment rates and economic well-being estimates were also found for the population with disabilities between the 2002 ACS and the 2003 ACS especially for the population with disabilities identified by the two altered questions (Weathers, 2005, Appendix C.). The differences are large, and it is possible that the difference is due to a change in the structure of the survey. 

Users should be aware that differences between estimates from the ACS disability data before the 2003 survey may differ from the 2003 ACS estimates because of this difference in the questionnaire.  The difference is likely to affect the ACS overall disability definition, the go-outside-the-home disability definition and the employment disability definition. It is likely that the other four disability questions were not affected by the change in the questionnaire and may be used to estimate changes over time.

ACS 2008 Changes

Major changes were made to the ACS disability questions in 2008. The changes were both a conceptual and empirical break from the previous ACS disability questions. The U.S. Census Bureau made these changes to better identify specific portions of the population of persons with disabilities, and to more clearly define disability as a functional limitation that may increase a person’s risk of participation limitation. These changes made to the disability questions mean that the population identified in 2008 is different from that identified in previous years and the results should not be compared. The Census Bureau’s American Community 2006 content test found that the new set of six disability questions had equal or better reliability and higher response rates than the previous ACS question set.

Below are the current disability questions used in the 2008-2011 ACS. The question numbers are those used for the 2008 version of the survey.

Q16a. Is this person deaf or does he/she have serious difficulty hearing?

Q16b. Is this person blind or does he/she have serious difficulty seeing even when wearing glasses?

Answer question 17a – c if this person is 5 years old or over.

Q17a. Because of a physical, mental, or emotional condition, does this person have serious difficulty concentrating, remembering, or making decisions?

Q17b. Does this person have serious difficulty walking or climbing stairs?

Q17c. Does this person have difficulty dressing or bathing?

Answer question 18 if this person is15 years old or over.     

Q18. Because of a physical, mental, or emotional condition, does this person have difficulty doing errands alone such as visiting a doctor’s office or shopping?

The major changes from the previous ACS disability questions used between 2003-2007 and the 2008 questions are:

·         Duration of limitation was removed and the term “serious” is used to focus on longer term/more severe impairments.

·         Employment Disability question is no longer asked.

·         Vision and Hearing Disabilities (was Sensory Disability) were separated into two independent questions and the population includes children under the age of 5.

·         Cognitive Disability (was Mental Disability): dropped “learning” as an activity, added the activity “making decisions.”

·         Ambulatory Disability (was Physical disability) is now limited to mobility related activities; “reaching, lifting, or carrying” activities dropped.

·         Self-Care Disability (was Self-Care disability): no longer includes phrase “Because of a physical, mental, or emotional condition.” The new question is limited to dressing and bathing activities, and the “getting around inside the home” activity is dropped.

·         Independent Living Disability (was Go-Outside-Home disability): dropped both the time restriction “lasting 6 months or more” and the phrase “Outside the home.”

The 2008 questions resulted in a lower overall disability prevalence rate than the previous set of questions (10.4% as compared to 12.8%), although the separate hearing and vision disability questions resulted in a higher prevalence rate than the pre-2008 sensory disability question.

Table C1

 

US Non-institutionalized population ages 21-64 prevalence rate: ACS 2007

 

Disability Type

Percent

MOE

Number

MOE

Base Population

Sample Size

Any Disability

12.8

0.05

22,295,000

89,200

174,206,000

1,692,615

Sensory

2.9

0.02

5,033,000

43,700

174,206,000

1,692,615

Physical

7.9

0.04

13,789,000

71,200

174,206,000

1,692,615

Mental

4.7

0.03

8,273,000

55,700

174,206,000

1,692,615

Self-Care

2.3

0.02

4,056,000

39,300

174,206,000

1,692,615

Go-Outside-Home

3.4

0.03

5,931,000

47,400

174,206,000

1,692,615

Employment

7.6

0.04

13,297,000

70,000

174,206,000

1,692,615

 

Table C2

 

US Non-institutionalized population ages 21-64 prevalence Rate: ACS 2008

 

Disability Type

Percent

MOE

Number

MOE

Base Population

Sample Size

Any Disability

10.4

0.06

18,312,900

115,430

175,368,200

1,693,675

Visual

1.9

3.29

3,314,200

50,380

175,368,200

1,693,675

Hearing

2.3

0.03

3,990,400

55,220

175,368,200

1,693,675

Ambulatory

5.4

0.05

9,498,200

84,400

175,368,200

1,693,675

Cognitive

4.1

0.04

7,213,700

73,840

175,368,200

1,693,675

Self-Care

1.8

3.29

3,240,900

49,820

175,368,200

1,693,675

Independent  Living

7.6

0.04

13,297,000

70,000

174,206,000

1,692,615

 

See Brault, 2009 (pages 12-16) for a detailed comparison between the 2007 and 2008 disability questions and Census. For a summary of all changes to the ACS 2008 survey see the U.S.Census Bureau documentation regarding the 2006 ACS content testing.


Appendix D. International Classification of Functioning, Disability and Health (ICF) concepts

The 2008 ACS disability questions used the International Classification of Functioning, disability and health (ICF) as the basis for their development. In this section we describe the ICF related concepts of impairment, activity limitation, participation restriction, and disability (see WHO, 2001). A prerequisite to each of these concepts is the presence of a health condition. Examples of health conditions are listed in the International Classification of Diseases, Tenth Edition (ICD-10) and they encompass diseases, injuries, health disorders, and other health related conditions.

                                                                                                                  

An impairment is defined as a significant deviation or loss in body function or structure. For example, the loss of a limb or vision loss may be classified as impairments. In some surveys, impairments are defined as long lasting health conditions that limit a person’s ability to see or hear, limit a person’s physical activity, or limit a person’s mental capabilities.

 

An activity limitation is defined as a difficulty an individual may have in executing activities. For example, a person who experiences difficulty dressing, bathing or performing other activities of daily living due to a health condition may be classified as having an activity limitation. In some surveys, activity limitations are identified based upon a standard set of activities of daily living questions (ADL's).

 

A participation restriction is defined as a problem that an individual may experience in involvement in life situations. For example, a working-age person with a severe health condition may have difficulty participating in employment as a result of the physical environment (e.g., lack of reasonable employer accommodations) and/or the social environment (e.g., discrimination). In some surveys, participation restrictions are identified by questions that ask whether the person has a long lasting health condition that limits his or her ability to work, or whether a health conditions affects his or her ability to go outside his or her home to go shopping, to church or to the doctor’s office.

 

The final ICF concept that used is a disability. The term disability is used to describe the presence of an impairment, an activity limitation and/or a participation restrictions. This concept is similar to the definition used in the Americans with Disabilities Act of 1990 (ADA). The ADA defines a disability as “a physical or mental impairment that substantially limits one or more of the major life activities, a record of such an impairment, or being regarded as having such an impairment.”

 

While these concepts may seem to follow a progression—that is, an impairment leading to an activity limitation leading to a participation restriction—it is not necessarily the case. It is possible that a person may have a participation restriction without an activity limitation or impairment. For example, a person diagnosed as HIV positive may not have an evident impairment or activity limitation but may not be able to find employment due to discrimination resulting from his health condition. Similarly, a person with a history of mental illness, but who no longer has a loss in capacity or activity limitation, may also be unable to finding employment due to discrimination resulting from his health condition.

 

Figure 1 provides a useful summary of the ICF concepts. It illustrates that while there is an overlap across these concepts, it is possible that one of them can occur without a relation to the others. The universe of the ICF is the health of the population as a whole. The shaded area of Figure 1 illustrates the ICF concept of a disability.

 

 

Simplified Conceptual Model of Disability Using ICF Concepts

Venn diagram showing that health conditions are made up a combination of health conditions (diseases, disorders, injuries, traumas, etc.), Impairment, Activity Limitation, and Participation Restriction. 

 

End Venn diagram image.

 

 


 

Appendix E. Disability Prevalence rate by state

(non-institutionalized population ages 21-64, 2009)

Location

Percent

MOE

Number

MOE

Base Population

Sample Size

United States

10.4

0.06

18,382,600

115,670

177,004,700

1,709,245

Alabama

15.5

0.51

417,500

14,130

2,688,000

27,080

Alaska

12.0

1.07

50,700

4,620

421,700

3,648

Arizona

10.3

0.36

377,400

13,670

3,673,600

33,723

Arkansas

17.0

0.68

274,900

11,430

1,621,000

15,814

California

8.4

0.14

1,791,600

29,920

21,451,300

201,116

Colorado

8.2

0.36

245,800

11,080

3,006,600

29,080

Connecticut

8.6

0.45

174,400

9,330

2,037,200

19,843

Delaware

11.2

1.01

56,700

5,280

508,100

4,728

District of Columbia

10.0

1.12

37,600

4,300

376,200

3,536

Florida

9.9

0.21

1,029,900

22,600

10,365,300

101,880

Georgia

10.5

0.29

593,300

17,110

5,661,400

54,412

Hawaii

7.7

0.7

58,800

5,430

762,900

7,616

Idaho

11.2

0.78

96,100

6,880

859,900

8,528

Illinois

8.2

0.23

613,000

17,510

7,484,700

71,810

Indiana

11.3

0.38

414,900

14,270

3,668,100

36,918

Iowa

9.4

0.48

158,800

8,250

1,692,400

16,802

Kansas

10.6

0.55

170,100

9,160

1,611,000

15,527

Kentucky

16.4

0.54

409,800

13,950

2,500,400

24,817

Louisiana

13.0

0.48

332,900

12,720

2,557,500

24,733

Maine

14.4

0.85

111,800

6,810

777,900

7,184

Maryland

8.4

0.35

284,100

11,900

3,367,300

32,773

Massachusetts

9.2

0.33

361,300

13,390

3,919,600

37,541

Michigan

11.9

0.31

684,200

18,290

5,755,700

55,854

Minnesota

8.4

0.36

259,900

11,390

3,087,600

29,896

Mississippi

15.1

0.64

245,400

10,870

1,624,800

15,848

Missouri

12.6

0.41

430,600

14,490

3,426,500

33,534

Montana

11.8

0.99

66,400

5,700

562,100

4,999

Nebraska

9.2

0.66

93,900

6,840

1,017,000

9,896

Nevada

9.0

0.53

138,900

8,310

1,541,500

15,186

New Hampshire

9.2

0.74

72,500

6,000

792,100

7,593

New Jersey

7.8

0.27

396,400

14,100

5,099,900

49,612

New Mexico

12.3

0.71

138,600

8,230

1,126,400

10,408

New York

9.1

0.19

1,042,400

22,760

11,487,400

107,488

North Carolina

11.7

0.32

635,600

17,640

5,419,700

53,578

North Dakota

9.2

1.08

34,200

4,120

373,400

3,755

Ohio

12.0

0.29

798,100

19,750

6,651,000

66,036

Oklahoma

15.2

0.57

313,200

12,270

2,060,400

20,078

Oregon

11.2

0.48

252,300

11,120

2,255,100

21,821

Pennsylvania

11.0

0.27

795,900

19,790

7,216,700

69,949

Rhode Island

10.2

0.89

62,800

5,570

613,600

5,914

South Carolina

12.2

0.46

317,200

12,450

2,601,400

25,926

South Dakota

9.7

1.01

43,900

4,670

451,600

4,424

Tennessee

13.9

0.41

510,000

15,690

3,665,800

36,130

Texas

10.3

0.19

1,439,200

26,680

13,995,400

132,863

Utah

8.3

0.52

125,600

7,940

1,508,600

14,474

Vermont

11.4

1.11

41,900

4,210

368,400

3,512

Virginia

9.3

0.31

429,800

14,610

4,636,300

45,613

Washington

10.7

0.36

426,700

14,480

3,980,000

38,745

West Virginia

18.4

0.8

195,800

8,890

1,063,300

10,385

Wisconsin

8.9

0.33

294,400

11,240

3,291,700

33,445

Wyoming

11.1

1.18

35,500

3,870

319,500

3,174

Puerto Rico

19.1

0.56

424,500

13,100

2,228,000

17,275


Appendix F. Employment rate by state and disability status

(non-institutionalized population ages 21-64, 2009))

Location

Employment rate: With Disabilities

MOE: With Disabilities

Sample Size: With Disabilities

Employment rate: WithOUT Disabilities

MOE: WithOUT Disabilities

Sample Size: WithOUT Disabilities

Employment Ratio: WithOUT Disabilities

United States

36.0

0.31

181,131

76.8

0.09

1,528,114

0.47

Alabama

30.0

1.63

4,327

74.8

0.66

22,753

0.40

Alaska

50.2

4.36

404

77.7

1.34

3,244

0.65

Arizona

33.9

1.77

3,544

73.4

0.56

30,179

0.46

Arkansas

34.6

2.08

2,827

76.8

0.83

12,987

0.45

California

34.2

0.81

16,991

74.3

0.23

184,125

0.46

Colorado

46.7

2.31

2,438

78.7

0.56

26,642

0.59

Connecticut

41.7

2.71

1,662

79.8

0.67

18,181

0.52

Delaware

35.8

4.61

512

80.1

1.36

4,216

0.45

District of Columbia

28.3

5.33

341

78.0

1.63

3,195

0.36

Florida

32.9

1.06

10,536

74.6

0.33

91,344

0.44

Georgia

34.6

1.42

5,957

75.4

0.44

48,455

0.46

Hawaii

41.1

4.65

593

80.3

1.09

7,023

0.51

Idaho

38.5

3.60

969

75.7

1.12

7,559

0.51

Illinois

36.3

1.41

6,282

76.5

0.37

65,528

0.47

Indiana

36.0

1.71

4,275

77.0

0.53

32,643

0.47

Iowa

47.0

2.46

1,632

84.0

0.58

15,170

0.56

Kansas

44.7

2.76

1,651

81.7

0.74

13,876

0.55

Kentucky

28.1

1.61

3,965

74.7

0.69

20,852

0.38

Louisiana

35.0

1.89

3,451

76.3

0.65

21,282

0.46

Maine

33.8

3.01

1,037

81.0

1.02

6,147

0.42

Maryland

44.0

2.13

2,797

81.0

0.51

29,976

0.54

Massachusetts

34.8

1.82

3,305

79.8

0.49

34,236

0.44

Michigan

30.4

1.27

6,564

71.7

0.46

49,290

0.42

Minnesota

44.3

2.23

2,543

82.0

0.52

27,353

0.54

Mississippi

28.7

2.09

2,660

74.9

0.85

13,188

0.38

Missouri

35.6

1.67

4,320

78.5

0.54

29,214

0.45

Montana

45.5

4.43

577

78.6

1.33

4,422

0.58

Nebraska

45.8

3.73

937

84.7

0.86

8,959

0.54

Nevada

40.9

3.02

1,474

75.4

0.83

13,712

0.54

New Hampshire

39.3

4.16

706

82.0

1.04

6,887

0.48

New Jersey

40.2

1.78

3,795

77.8

0.44

45,817

0.52

New Mexico

37.0

2.97

1,331

74.7

1.00

9,077

0.50

New York

33.9

1.06

10,154

76.4

0.30

97,334

0.44

North Carolina

34.9

1.37

6,442

76.2

0.45

47,136

0.46

North Dakota

56.1

6.15

367

85.4

1.39

3,388

0.66

Ohio

34.0

1.21

7,683

76.5

0.40

58,353

0.44

Oklahoma

40.2

2.01

3,249

78.5

0.71

16,829

0.51

Oregon

38.1

2.22

2,416

74.7

0.70

19,405

0.51

Pennsylvania

36.3

1.23

7,444

78.1

0.37

62,505

0.46

Rhode Island

36.6

4.40

584

80.6

1.22

5,330

0.45

South Carolina

30.1

1.87

3,235

74.7

0.66

22,691

0.40

South Dakota

45.8

5.45

411

83.3

1.34

4,013

0.55

Tennessee

30.2

1.47

5,205

75.7

0.55

30,925

0.40

Texas

39.8

0.94

14,272

77.1

0.27

118,591

0.52

Utah

43.3

3.20

1,236

77.9

0.81

13,238

0.56

Vermont

44.2

4.77

384

83.6

1.27

3,128

0.53

Virginia

39.6

1.71

4,292

80.0

0.45

41,321

0.50

Washington

39.1

1.71

4,076

76.7

0.51

34,669

0.51

West Virginia

28.0

2.16

1,919

74.2

1.00

8,466

0.38

Wisconsin

40.3

1.92

3,007

81.6

0.48

30,438

0.49

Wyoming

50.3

5.21

352

82.5

1.40

2,822

0.61

Puerto Rico

22.8

1.37

3,436

57.6

0.78

13,839

0.40

 

 

 



[1] For a summary of all changes to the ACS 2008 survey see the following Census Bureau 2006 Content test summary: http://www.census.gov/acs/www/Downloads/methodology/content_test/SummaryResultsACS2006ContentTest.pdf

[2] The Census Bureau frequently refers to these questions as “difficulties” while the overall category is referred to as “disability.” In this document we refer to all types as “disabilities.”

[3] For more information regarding how the Census Bureau calculates poverty status see page 100 in the American Community Survey Puerto Rico Community Survey 2009 Subject Definitions: :http://www.census.gov/acs/www/Downloads/data_documentation/SubjectDefinitions/2009_ACSSubjectDefinitions.pdf

[4] U.S. Census Bureau American Factfinder: http://factfinder2.census.gov/faces/nav/jsf/pages/index.xhtml.

[6] For more information regarding the PUMA creation guidelines, see 2010 Public Use Microdata Areas (PUMAs) http://www.census.gov/geo/puma/puma2010.html .

 

[7]The 2008/2009 National Health Interview Survey includes a split-ballot format to test the ACS and CPS disability question variants. For more information on this testing see the following CDC document Disability Questions Tests 2008/2009 File (7/15/2010): ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHIS/2009/srvydesc_disbtest.pdf

[8] Note that for many states the GQ PUMS population sample is small. Estimates may vary significantly from year to year at the state level and for smaller geographies due to the GQ sampling variability (Beaghen & Stern, 2009). Users interested in GQ populations may want to utilize multiyear PUMS files and use caution with regards to sample sizes. See PUMS GQ sample sizes in Table 2 of the appropriate year “PUMS Accuracy of the Data.” available under the “Accuracy of the PUMS” heading: http://www.census.gov/acs/www/data_documentation/pums_documentation/

[9] In 2008 the GQ sampling rate was increased to meet Census Bureau publication thresholds for 16 states with small GQ populations. Beginning in 2011sample size will increase to 3.54 million HU addresses annually.