The geographical distribution of nursing homes can significantly distort small-area life expectancy estimations. Consequently, uncorrected life expectancies should not be used for small-area life expectancy comparisons. Instead, several nursing home corrections have been proposed. The practical use of these corrections, however, is severely limited by data availability. This paper introduces a new, model-based nursing home correction that requires considerably less detailed nursing home data. A formal comparison shows that the proposed model-based approach is the recommended correction for all small-area life expectancy estimations where detailed previous residential address information of the nursing home population is not available. This makes the approach highly relevant for a wide range of empirical applications.
Copyright © 2012 Elsevier Ltd. All rights reserved.