Assessing the Need for a National Disability Survey: Final Report. F. Augment Samples in Existing Surveys


Federal and state agency staff indicated that one way they attempt to address issues related to small survey sample sizes is to pool data across survey years. This is an effective and efficient means for increasing sample sizes when data have been collected consistently over time and it is reasonable to presume that population distributions for relevant variables have changed little. To facilitate these types of analyses, NCHS provides weights and survey design adjustment information to use when multiple years of the NHIS are being pooled. Other surveys, including the MEPS, NBS, CPS, and SIPP are good candidates for pooling multiples years to increase sample sizes. For example, studies have pooled multiple years of the MEPS to study the health care expenditures of people with disabilities (Olin and Dougherty 2006), and multiple years of the NHIS to study SSDI beneficiaries in the 24-month waiting period for Medicare (Riley 2006; Livermore et al. 2010).

Modifications to survey sample frames and methods might also identify more people with disabilities in surveys. One respondent suggested, “Enhancements to survey sampling frames should be considered as a way to enhance sample size of people with disabilities in existing surveys.” The exclusion of people residing in group quarters or institutions may inadvertently omit a large number of people with disabilities (She and Stapleton 2009). Improvements in the manner with which surveys interview respondents may also help to identify more people with disabilities. People with disabilities may require additional follow-ups to successfully complete a survey compared to people without disabilities. Alternative measures for conducting the survey may be required, such as providing surveys in Braille or verbally for people with vision difficulties.

Another way to increase survey sample sizes so that particular subpopulations of people with disabilities might be studied is to augment the samples of existing surveys with such individuals. An example of this is SSA’s collaboration with the Census Bureau to include an additional sample of SSDI and SSI beneficiaries in the 2001 SIPP. To increase the number of SSDI and SSI program participants available for analysis, SSA contracted with the Census Bureau to interview a sample of known SSDI beneficiaries and SSI recipients identified from SSA administrative records using the SIPP core questionnaire. This resulted in 2,447 additional interviews of known program participants. By adding this oversample of interviews to the 2,575 program participants already in the sample, the sample size of SSI and SSDI beneficiaries in the 2001 SIPP increased to a total of 5,022 (DeCesaro and Hemmeter 2008).

The addition of the standard six-question disability series to federal surveys might also offer future opportunities to augment sample sizes in surveys that obtain their samples from larger national surveys. For example, the MEPS obtains its sample from the NHIS. Sample sizes for working-age people with disabilities in the MEPS are fairly small, necessitating the pooling of data across years to study this population. With the addition of standard disability questions to both the NHIS and MEPS, sample sizes of people with disabilities in the MEPS could be augmented by using the disability questions as a screener in the NHIS, and oversampling NHIS respondents with disabilities in the MEPS. Several federal agencies noted unanswered disability questions related to health insurance, health care utilization, expenditures, and effects of the Affordable Care Act that could potentially be addressed with augmented samples and data from the MEPS.

Another option is for surveys to oversample from identifiable groups with relative high disability prevalence. In the past, the SIPP has used this approach to oversample from low-income households, and the HRS has used it to oversample African Americans. For instance, using ACS data pooled across years, it might be feasible to identify fairly small areas where disability prevalence is relatively high. Those areas could be oversampled for the SIPP.

View full report


"NatlDS.pdf" (pdf, 1.5Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®