OBJECTIVES: Measurement of health inequalities based on self-reports may be biased if individuals use response scales in systematically different ways. We use anchoring vignettes to test and adjust for reporting differences by education, race/ethnicity, and gender in self-reported health in 6 domains (pain, sleep, mobility, memory, shortness of breath, and depression). METHOD: Using data from the 2006 U.S. Health and Retirement Study (HRS) and the 2007 Disability Vignette Survey, we estimated generalized ordered probit models of the respondent's rating of each vignette character's health problem, allowing cut-points to vary by age, gender, education, and race/ethnicity. We then used one-step hierarchical ordered probit (HOPIT) models to jointly estimate the respondent's cut-points from the vignettes and the severity of the respondent's own health problems based on these vignette cut-points. RESULTS: We found strong evidence of reporting differences by age, gender, education, and race/ethnicity, with the magnitude depending on the specific health domain. Overall, traditional models not accounting for reporting differences underestimated the magnitude of health inequalities by education and race/ethnicity. DISCUSSION: These results suggest caution in relying on self-reported health measures to quantify and explain health disparities by socioeconomic status and race/ethnicity/ethnicity in the United States. The findings support expansion of the use of anchoring vignettes to properly account for reporting differences in self-reports of health.
Stored only when both papers are in the corpus — 174 citations exist upstream.
Imported from OpenAlex (CC0) — nothing here is generated.