Estimating transition probabilities of age misstatement

Abstract

Abstract This article outlines a method of estimating probabilities of gross transfers from one age to another due to misreporting of age. An essential ingredient in the computation is the information on the true age structure of the population, which may be estimated by using generalized stable population relationships. The method consists essentially of iteratively adjusting rows and columns of an initial guess matrix so that the application of the resultant transition matrix to the true age distribution produces the recorded age distribution. The initial guess matrix can be either empirically based or theoretically derived. The method is illustrated by using data on India, 1971–1981. The application reveals that in India, although the tendency to exaggerate age is strong at adult ages, the bias does not increase appreciably at older ages, as is commonly believed.