Abstract for: Birth Cohorts Approach to Modeling Aging Populations
Traditional approaches to modeling aging populations require disaggregating populations into multiple age cohorts and defining transitions using continuous-time first-order delays or discrete-time delays. A recognized limitation of continuous-time methods is a phenomenon known as cohort blending which results in distortions of age distribution. Discrete-time delays and cohort age shifts impose other model limitations making it difficult to integrate with other dynamics of the system. We present an alternative approach to modeling aging populations that disaggregates population into multiple birth year cohorts where birth year is the time-invariant attribute of the stock and age is a time-varying parameter of the system. This approach is well-suited for problems where it is important to understand the impact of exogenous parameters that change significantly over the time-horizon of interest that are both age- and time-dependent, such as mortality. We apply this method to create a model of the US national population from 1975 to 2010 that compares well with census data. The birth-cohort national model is a generic aging population model that can be adapted for study-specific purposes where both age and time are important factors.