预测误差的相关性对人口预测结果的影响

Effects of Correlated Forecast Errors on Population Forecasts

文章应用时间序列模型,以中国和瑞典的历史人口数据为例,深入探讨了预测误差的相关性对人口预测结果的影响。预测误差的相关性是指预测误差与时间、年龄、地区和性别,以及生育、死亡和迁移预测误差之间的相关关系。研究发现,只有充分考虑了误差相关性的预测区间才能覆盖实际人口的发展变化。预测误差的相关性对总人口数量预测结果的影响相对较小,但是对分年龄段人口数量和人口年龄结构指标预测结果的影响较大。在进行人口预测时,需要将预测误差相关性纳入模型,才能得出准确和科学的预测结果。

Based on the historical data from China and Sweden,this paper examines the effects of the correlation of forecast errors on population forecasts by using time series model.The correlation of forecast errors means forecast errors in the demographic rates,fertility,mortality and migration are associated with age,region and sex,and associated with themselves over time(auto-correlation)as well as the demographic rates correlated with one another.The study finds that only the predictive intervals with consideration of the correlations are able to cover the true values of the population indicators.The effects of correlations of forecast errors on total population forecasts are relatively low.However,its effects on forecasts of population by age and age structure indicator are huge.When forecasting population,we should put the correlation of forecast errors into the model in order to get accurate and scientific results.