Survey counts of those enrolled in Medicaid are significantly lower than administrative data counts. Because surveys are the only source of information on key policy populations, such as the uninsured and those eligible for Medicaid but not enrolled, this disparity can have significant implications, depending on why the two counts are systematically different. We are examining matched survey data and administrative data from the Current Population Survey, the National Health Interview Survey, and the Medicaid Statistical Information System to better understand how people who are enrolled in Medicaid answer the two surveys. This will allow us to understand the roll that survey measurement error plays in explaining the difference. We also examine the impact of survey item non-response (imputation), survey editing, unit non-response and sample coverage have on explaining the difference.