The Medicaid Undercount: Real or Perceived Bias in Estimates of Coverage in General Population Surveys

Given that the general population surveys of health insurance coverage may undercount the number of individuals enrolled in Medicaid by 15 to 50%, depending on the estimate, does such an undercount lead to inflated estimates of the uninsured? A Minnesota study suggests that, for the most part, Medicaid enrollees know whether or not they have insurance, whether they have public or private coverage, but are less sure which public program they are enrolled in. This has implications for how the Medicaid undercount is used, especially for adjusting the estimates of uninsured. However, it is unclear how well the Minnesota findings generalize to other states. This study replicated the Minnesota study in three other states with different population characteristics, different state health insurance programs, and different survey designs to determine whether the currently used estimates for undercounts are appropriate or whether other methodological adjustments are necessary. The purpose of this study was to provide a better methodology for adjusting counts of Medicaid recipients and the uninsured in national surveys. Since survey estimates of the uninsured play a central role in setting policy and allocating resources, the results of this study would improve the confidence with which these survey results are used.