Demographic analysis and forecasting using of inter-municipal population growth and distribution matrices

Abstract Challenges in the field of demographic projections include, among others, the volatility of the migration component - critical for the projection of small areas; the compatibility between projections of small and large areas; and the measurement and inclusion of uncertainty in future scenarios of population growth. This article presents a new probabilistic method to conduct interregional population forecasting dealing with these three challenges. The proposed method has the following advantages: 1) it only requires information about the last place of residence and the population distributions of the last two Censuses; 2) it generates confidence intervals for the projected populations; 3) it makes the role of migration flows in the growth dynamics explicit and; 4) it facilitates the elaboration of counterfactual scenarios and sensitivity analysis using matrices of interregional population growth and distribution. We describe the patterns and trends in migration flows in the state of São Paulo applying spatial visualization tools and identifying areas in which migration is responsible for considerable shares of the demographic dynamics. About 95% of the 572 municipal projected populations of São Paulo had good precision and were within expected confidence intervals. We used data from the 1980, 1991 and 2000 Brazilian Censuses.摘要 人口预测（demographic projections）领域面临的诸多挑战包括：迁移成分（migration component）的波动性——这对小区域人口预测至关重要；大小区域预测结果间的兼容性问题；以及人口增长未来情景中不确定性的测度与纳入。本文提出一种全新的概率方法开展区域间人口预测（interregional population forecasting），以应对上述三项挑战。所提出的方法具备以下优势：1）仅需获取最新居住地信息与最近两次人口普查的人口分布数据；2）可生成预测人口的置信区间（confidence intervals）；3）清晰阐明人口迁移流在增长动态中的作用；4）借助区域间人口增长与分布矩阵，便于构建反事实情景（counterfactual scenarios）并开展敏感性分析（sensitivity analysis）。我们通过空间可视化工具（spatial visualization tools）分析圣保罗州（São Paulo）的人口迁移流模式与趋势，并识别出迁移对人口动态贡献占比可观的区域。圣保罗州572个市政区域的预测人口中，约95%具备良好精度，且落在预设置信区间内。本研究使用了巴西1980年、1991年及2000年的人口普查（Brazilian Censuses）数据。