Disability Prevalence Measurement

Definition

The set of survey instruments and conceptual frameworks used to identify the working-age population with disabilities in nationally representative U.S. data. Two primary instruments coexist with partially non-overlapping coverage: (1) the traditional Current Population Survey Annual Social and Economic Supplement (CPS-ASEC) work-activity limitation question (in use since 1981), which asks whether a health condition limits the amount or kind of work a person can do; and (2) the American Community Survey (ACS)/CPS Basic Monthly Survey (CPS-BMS) six-question sequence (introduced 2008), which asks about serious difficulties in six functional domains (vision, hearing, cognitive, ambulatory, self-care, independent living) without any work-activity component. The conceptual target for both is the World Health Organization (WHO) ICF (International Classification of Functioning, Disability and Health) framework, which defines disability as a health condition-based impairment, activity limitation, and/or participation restriction.

Key Ideas

ICF concentric-circle model: In theory, the disability population should nest as three concentric circles — ICF-based disability (outer) $\supset$ work-activity limitation (middle) $\supset$ Social Security Disability Insurance (SSDI)/Supplemental Security Income (SSI) beneficiaries (inner). In practice, the work-activity limitation and six-question populations only partially overlap; neither is a subset of the other.
2010 CPS Venn diagram (ages 25–61, civilian noninstitutionalized): $A$ = six-question only (3.3%, 5.6 million); $B$ = both (4.6%); $C$ = work-limit only (3.7%, 5.0 million); $A+B$ = 7.9% (12.0 million); $B+C$ = 8.3% (12.5 million); $A+B+C$ = 11.6% (17.5 million). Using either instrument alone understates the combined population by $\approx 35\%$ .
Composition bias: The three sub-populations have dramatically different labor market and program participation profiles. Work-limit-only ( $C$ ): employment 22.8%, SSDI/SSI receipt 37.6%. Six-q-only ( $A$ ): employment 57.3%, SSDI/SSI receipt 11.6%. Using $A+B$ (six-question sample) systematically overstates employment (30.8% vs. 16.6% for $B+C$ ) and understates SSDI/SSI receipt (38.9% vs. 49.2% for $B+C$ ).
Face validity test: SSDI and SSI-Disabled Adults beneficiaries have medically-verified work-preventing impairments; any disability measure claiming ICF coverage should capture them. Six-question sequence captures only 63.3% of SSDI/SSI recipients; work-activity limitation captures 84.1%; seven-question union captures 92.0%.
Trend validity vs. level bias: Despite large level differences, employment and program participation trends are similar across all definitions (1981–2010). The work-activity limitation question is valid for trend analysis but not for cross-sectional benchmarking.
Historical ACS revision: In 2005–2007, the ACS used a six-question sequence that included a work-activity limitation question. In 2008, the vision/hearing questions were separated and the work-activity limitation was dropped, despite objections from the American Statistical Association president (Keller-McNulty 2006).

How It Works

Work-activity limitation question (CPS-ASEC, 1981–present): A single question — does a health condition limit the amount or kind of work you can do? Identifies a population with severe functional limitations specifically affecting work capacity. Captures the middle circle ( $B+C$ ) of the ICF model. Yields $\approx 8.3\%$ disability prevalence among working-age adults (stable 1981–2010).

ACS/CPS-BMS six-question sequence (2008–present): Six domain-specific questions covering: vision difficulties, hearing difficulties, cognitive difficulties (remembering/concentrating/making decisions), ambulatory difficulties (walking/climbing stairs), self-care difficulties (dressing/bathing), and independent living difficulties (running errands). Captures $A+B$ . Yields $\approx 7.9\%$ prevalence but misses $C$ — people whose impairments specifically affect work capacity but don't manifest as difficulty in the six enumerated functional domains.

Seven-question combined sequence: The union of both ( $A+B+C$ ). Best current operationalization of ICF-based disability for policy purposes; prevalence $\approx 11.6\%$ . Only available in March CPS (when both ASEC and BMS are administered simultaneously).

Why It Matters

Bureau of Labor Statistics monthly disability statistics (published since June 2008) are based on the CPS-BMS six-question sequence and therefore systematically overstate the employment rate and understate the SSDI/SSI participation rate of the true working-age population with disabilities. Researchers using ACS or CPS-BMS data after 2008 to study disability employment trends or program take-up face potential composition bias if they do not supplement with a work-activity limitation question. The 2008 revision also introduced a structural break in disability prevalence series, making pre- and post-2008 ACS data not directly comparable for policy analysis.

Open Questions

What alternative to the work-activity limitation question (if any) could be added to the ACS/CPS-BMS to recover coverage of the $B+C$ population without reintroducing the work-endogeneity critique (Anderson-Burkhauser 1985: non-workers more likely to report work limitations)?
Does the composition bias in six-question-only samples affect measured trends in disability employment gaps, or only levels?
How do the ACS and CPS-BMS disability series behave post-2010, particularly during and after the COVID-19 period when both employment and program participation shifted dramatically?