A model of age heaping with applications to population graduation that retains informative demographic variation

Dalkhat M. Ediev, North-Caucasian Academy; Lomonosov Moscow State University; International Institute for Applied Systems Analysis;

Age heaping remains an issue in both historical and contemporary demographic studies. Traditional smoothing techniques are problematic in dealing with age heaping in cases where the population age structure shows both the digit preference and informative variation by age that should not be lost during the graduation. Same applies to more recent modelling-based smooth reconstructions of the latent population distributions. We generalize and modify an earlier model where age rounding's propensity depends on the distance to the round age and the strength of age heaping at that age. Efficient and robust estimation method is proposed for parameterizing the model that allows reconstructing the latent population distribution by age. We test our model in comparison to the traditional alternatives on an ample set of empirical data. Our method is capable of removing age heaping without substantial distortions of the actual population variation by age. In comprehensive empirical testing, our method appears the best in both the quality of heaping removal and retaining the informative population variation. The method has good potential for a wide practical application.

Keywords: Age structure, Methodology, Applied demography, Mathematical demography

See paper.

  Presented in Session 113. Measurement of Age and Age Structures