• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

Article

Статистические подходы к анализу и моделированию сезонности в демографических данных

Демографическое обозрение. 2019. Т. 6. № 2. С. 104-141.

According to the May Presidential Decree (2018), one of the national goals and strategic objectives of the development of the Russian Federation for the period up to 2024 is “ensuring sustainable natural growth in the population of the Russian Federation and increasing life expectancy to 78 years”. Thus, the increased need to monitor the current demographic situation, the study of the structure of demographic indicators, and the close attention of the community to the realization of national goals led to the choice of the topic of this study. The paper studies the problems of modeling the seasonality of demographic indicators in the Russian Federation (the number of births, the number of deaths, infant mortality, the number of marriages) according to monthly data of Rosstat for the period 2007-2018. Foreign studies have shown that, along with traditional demographic methods, ARIMA models give good results in forecasting of demographic indicators (population size, birth and death rates, life expectancy). Using the approach based on SARIMA models in this work allowed us to obtain adequate models with good statistical and prognostic properties. The stationarity of processes was analyzed on the basis of the HEGY test. The indicators studied in the work had a number of features that must be taken into account when modeling. The series of the number of births and the number of deaths had second and first integration orders respectively and contained deterministic seasonality, the series of the number of marriages had the first integration order and seasonal integration, and the infant mortality series did not contain seasonality, which was confirmed based on the analysis of the autocorrelation function and periodogram. Point and interval estimates of the forecast for 2019 were built for all indicators here studied. To compare the quality of forecasting SARIMA models, seasonal Holt -Winters models were also evaluated.