The reliability of intercensal population estimates and nowcasts

László G. Radnóti, Hungarian Central Statistical Office

A stochastic version of the cohort-component method of population forecasts is studied in the paper. Assuming that the evolution of the population can be modelled with a Markov process, where the number of deaths in each age group and births to females of fertile ages follow Poisson distributions it is possible to derive interval estimates for the composition of the population at a given moment of time in the future, if the initial population composition and the parameters of the process – e.g. age specific mortality and birth rates – are given. It is shown that for large enough populations the resulting confidence intervals for forecasted population will be very tight, therefore in the case of short term population forecast of larger countries it is enough to forecast the expected population counts. Results are illustrated using recent Hungarian census, mortality, birth and migration data.