Representativeness of the Low-Income Population in the Health and Retirement Study

Abstract

We study to what extent the Health and Retirement Study (HRS) is representative of all income groups, but with a particular emphasis on low-income groups. To focus on the HRS sample composition and abstract from potential measurement issues associated with measures of income and program participation, we exploit the SSA administrative data matched to the HRS sample and compare their distribution against the distribution of the same variables for the same population in the SSA databases. We find that overall, for cohorts and years that can be most reliably compared, the distributions are very similar and conclude that the HRS is representative for the population it covers. However, for some subgroups in the low-income population (e.g., recipients of Supplemental Security Income, Medicaid beneficiaries), there are some differences and thus we caution against estimating population totals for such small subpopulations. The HRS samples for which restricted matched administrative data are available are often not representative of a broad population of interest, because not all HRS respondents were asked permission to match in any given year. Therefore, the restricted HRS datafiles are generally not suitable for estimating population distributions, although they are still very useful for modeling purposes.