Predicting Demographic Indicators by Splines

Abstract

The current demographic pattern in Ukraine features the decreasing birth rate and the increasing mortality, resulting in ageing and decline of the population, which breaks the favourable demographic balance. At the Ukrainian territories affected by radioactive contamination because of the accident at the Chornobyl nuclear power plant, these processes differ from those on other territories. Given the considerable impact of emergency situations on the course of demographic processes, developing and testing prediction techniques specifically designed for those territories has essential importance. The purpose of this work is to forecast changes in demographic indicators (population number, birth rate, mortality,infant mortality and stillbirth) by spline functions, forthe areas with the heaviest radioactive contamination after the Chornobyl disaster, which are located in Zhytomyr region (Korosten, Luhyno, Narodychy, Ovruch and Olevsk), for 2020–2023, and the town of Korosten for 2021–2023.The data sources for the research were State Statistic Service of Ukraine for 1979–2020, and the Ministry of Health of Ukraine for the period of the COVID-19 pandemic. At the first phase, spline functions were used to forecast the abovementioned indicators in 2017–2020 by data for 1979–2016 for the town of Korosten. A comparison of the resulting forecast with the actual data led to the conclusion that that spline models of observations could be an effective tool for short-term forecasting of population number, birth rate and mortality. The most adequate prediction of population numbers could be achieved with cubic splines, whereas the best prediction of birth rate and mortality – with linear continuous splines. The forecasts of birth and mortality rates for 2017–2020 proved to be quite optimistic. However, in the pandemic conditions, a deviation of the predicted population numbers, birth rate and mortality was revealed: actual numbers for all the three indicators in 2020 were beyond the confidence region, which had not been the case in 2017–2019. The actual figures in 2020, found to be far worse than the predicted ones, may be caused by the impact of the COVID-19 pandemic in 2020, which is an unpredictable factor. The extra mortality caused by COVID-19 in 2020 was estimated for the town of Korosten. At the second phase, probable changes in the population number, birth rate and mortality were predicted for 2021–2023 in all the areas. The 95% confidence region and confidence intervals were built for the predictions. It was found that the last years’ trends in demographic indicators in radioactively contaminated territories would continue in a short-term perspective. It was shown that because annual numbers of infant mortality and stillbirthcould not be predicted due to their significant variations, averaging for 5-year periods should be used. A forecast of the average numbers was made for the 5 five-year periods where actual data were unknown. Also, it was emphasized that for the indicators predicted with linear continuous splines, actual numbers might turn to be far from the forecasted ones, because of the existence of extreme points, with growth suddenly changing for decline, and vice versa. Because such points cannot be predicted by extrapolation of observed trends, it is necessary to find other methods for their prediction. Further research will focus on other prediction methods, to achieve higher prediction accuracy,and on inclusion of economic indicators in the prediction models.