A Latent Class Analysis Approach to Multiple Systems Estimation with Longitudinal Register Data

Lucy Brown , University of Kent
Eleni Matechou, Queen Mary University of London
Bruno Santos, Universidade de Lisboa
Eleonora Mussino, Umeå University

Overcoverage occurs when individuals are registered as living in a country but in fact live elsewhere (imperfect emigration registration) or have passed away and their death was not recorded (imperfect death registration). Overcoverage can lead to serious bias in population estimates, negatively influencing policymaking and research. Demographers around the world have been working towards estimating overcoverage, particularly with the use of official population registers. As more countries move towards a register-based system as a means to either replace or complement traditional censuses, an efficient, standardised method to estimate population size and in turn overcoverage is essential. Many previous approaches rely on multiple systems estimation (MSE) but are only able to consider annual snapshots of register data. We propose a method based in MSE but extended to allow knowledge exchange over time in a longitudinal approach, as well as incorporating individual heterogeneity in the observation process. Instead of using log-linear models for contingency tables which is the traditional approach for MSE, we consider latent class models (LCMs) which by design assume individuals belong to two (or more) latent classes, each of which behave differently in the observation process. This model will be employed on population register data for Sweden and Norway, providing insights into population dynamics, both latent and observable.

See extended abstract

 Presented in Session P8. Demographic Trends, History, Data and Methods