Using Logistic Regression to make County‐level Estimates of the Medically Uninsured in North Carolina

Often, policy analysts are asked to produce data for which there are no universally accepted methods. Policymakers and legislators are continually searching for accurate estimates of the magnitude of the problem with which to inform their debate, but often need the estimates within a short period of time—too short to allow large, population‐based sample surveys. This means that such estimates must be produced from data that may lack the specificity sought by policymakers and legislators, using techniques not perfectly suited for the analysis. The recent focus on health care financing policies has created a situation where estimates of the number of medically uninsured persons are key to decision‐making about coverage policies. This article describes the use of the 1992 Current Population Survey and logistic regression analysis to explain the determinants of women of childbearing age in North Carolina without medical insurance, and to develop county‐level estimates of their population. This is an example of using logistic regression as a fool for prediction and projection of data crucial to the policymaking process as well as adapting a method normally used in the academic environment to the policy world.