Abstract
Most research on health insurance in the United States uses the Current Population Survey Annual Social and Economic Supplement. However, a recent redesign of the health insurance questions disrupted the historical time trend in 2013. Using data from the American Community Survey, which has a parallel trend in the uninsured rate, we model a bridge estimate of the uninsured rate using the traditional questions. Also, we estimate the effect of changing the questionnaire. We show that the impact of redesigning the survey varies substantially by subgroup. This approach can be used to produce bridge estimates when other questionnaires are redesigned.