Predictions of county uninsured rates: accuracy and stability

Uninsured rates of all counties in an area can be predicted by combining survey data from a subset of area counties and secondary data on economic indicators that are available for all counties. The authors present a set of indicators that are related to the uninsured rate and develop prediction models from surveys taken two years apart in the same counties. The accuracy and stability of the prediction models are assessed. Accuracy levels are highest when contemporaneous rates are predicted, but accuracy deteriorates when rates in a later time period are predicted using models from a prior period.