Modern approaches to forecasting mortality

In recent years, Russia has seen an increase in the death rate, which entails a natural decline in the population, which, in turn, has negative consequences, such as restrictions on economic growth. In order to effectively formulate a development strategy for individual regions and for the whole country, it is necessary to have a reliable mortality forecast. There are many approaches to predicting mortality; the choice of a specific approach depends on the type, quantity and quality of available data, the setting of the problem, and the forecast horizon. Usually, demographic studies use extrapolation, interpolation and multicusal approaches to predict mortality. The article discusses their main features, advantages and disadvantages. Key differences in approaches to predicting mortality are determined by the nature of the information used. Extrapolation approaches assume that the trends in the baseline period in the development of mortality rates will continue. Interpolation, multi-causal and Bayesian approaches are also imply the use of qualitative (or subjective) information. For interpolation, a target mortality distribution must be determined. The development trajectory from the initial to the target distribution is determined based on mathematical interpolation. In multi-causal approaches, theoretical positions from related scientific fields can be integrated into mortality projections in order to correct trends observed in the past. In connection with Bayesian approaches, the corresponding probability distribution of the values of the factors that explain the dynamics of mortality must be taken into account. It is also worth to pay attention to the fact that the basis for high-quality forecasts of mortality and its age distributions is, firstly, reliable initial information about the levels of its indicators in the past (baseline) periods of time and (if possible) about the trends and causes of variability of mortality in the forecast period.