One of the most well-documented problems with using survey data to study the dynamics of health insurance in the United States is the undercount: the difference between estimates of Medicaid enrollment derived from surveys and counts from state administrative records. Research points to measurement error in the reporting of Medicaid enrollment as the main explanation for this undercount. In order to improve survey data on insurance, it is important to learn why respondents misreport and why rates of misreporting vary across surveys. In this study, we evaluate the extent and causes of false-negative reporting about Medicaid in the Medical Expenditure Panel Survey – Household Component (MEPS-HC). 2 We attempt to replicate methods from previous studies of different surveys so that the findings are more comparable and therefore more helpful in discussions about how specific survey features may affect the propensity to false-negatively report Medicaid. Results show that MEPS-HC respondents correctly classify an estimated 82.5% of the people known to have Medicaid during 2003 and incorrectly classify 9.2% as having some other type of coverage and 8.3% as having no insurance. This corroborates the findings that respondents have trouble accurately reporting Medicaid enrollment but most are classified as having some form of coverage (Call et al., 2009; U.S. Census Bureau, 2008). Results also corroborate findings that there is consistency across surveys in terms of the enrollee characteristics predictive of misreport. We conclude that poverty level and timing (of service, enrollment and reporting periods) are robust predictors of misreport. Closer analysis of predictive factors and how their interaction with misreporting varies across surveys may yield clues about how specific survey features may account for differences in reporting accuracy across surveys.