Using population register data and capture-recapture models to estimate over-coverage in Sweden.

Over-coverage occurs when individuals who reside in a country leave or pass away, and this demographic event is not recorded in population registers, leading to population size overestimation. This problem can have important policy and decision-making consequences. With the increased reliance on incomplete but overlapping official registers for documenting whole populations or subgroups of populations, there is a need for more sophisticated modelling techniques that reliably estimate population size, and hence over-coverage, from such registers. Previous approaches have considered multiple systems estimation (MSE) for monitoring over-coverage, but MSE does not naturally extend to cases where individuals are followed over time. In this paper, motivated by the case study of Sweden, we develop a capture-recapture (CR) modelling framework for population registers that allows us to estimate the population size each year, the probability of presence for each individual in the population, conditional on their records, each year and to quantify the effect of demographic characteristics on the probability of emigration and re-immigration, amongst other parameters. Our results suggest that the CR approach, which accounts for the whole time series for each individual, gives a more realistic estimate of the population size compared to existing, deterministic approaches, especially when considering the subgroup of newly arrived individuals, and that it provides new insights on individual behaviour in terms of migration patterns than existing MSE approaches.