Reporting error in weight and height among older adults: Implications for estimating healthcare costs

We find substantial error in older adults' reports of weight and height, and this error is neither classical nor independent of common regressors in econometric models. Use of self-reports leads to bias in estimates of mean BMI, obesity prevalence, coefficients in healthcare utilization regressions, and obesity-attributable healthcare costs. Correction algorithms can reduce but not eliminate bias, and in certain cases they can actually worsen bias. These findings have implications for accurate estimates of the impact of obesity on healthcare utilization and costs among older adults.