广西人口结构非线性演进与生育率影响因素

Nonlinear Evolution of Population Structure and Factors Influencing Fertility in Guangxi

摘要: 目的：刻画广西人口结构演进的非线性特征并识别生育率的关键影响因素，基于省级人口数据开展实证研究。方法：利用2005~2024年《广西统计年鉴》数据，运用光滑样条、Chow断点检验、Bootstrap与随机森林等非参数方法，对老龄化趋势、教育结构与初婚人数关系及出生率驱动因素进行建模分析。结果：广西老龄化率呈分段加速上升，年均增速由2005~2014年的0.01个百分点跃升至2015~2024年的0.44个百分点，2016年经Chow检验确认为显著结构性转折点；光滑样条拟合优度(R² = 0.958)显著优于线性模型(R² = 0.700)；大专及以上占比与初婚人数呈倒U型关系，拐点为5.82% (95%置信区间：5.33%~6.49%)；老龄化率、初婚人数与高等教育占比是影响出生率的前三大因素，相对重要性为100%、98.9%、96.0%，随机森林袋外R² = 0.889。结论：广西人口结构演进具有显著非线性特征，非参数方法能更准确地刻画分段变化，低生育率是人口老化、婚姻市场萎缩与教育水平提升共同作用的结果，可为人口监测预警与婚育政策优化提供量化依据。

Abstract: Objective: To depict the nonlinear characteristics of population structure evolution in Guangxi and identify the key factors influencing fertility, and to conduct an empirical study based on provincial population data. Methods: Using data covering 2005~2024 from the Guangxi Statistical Yearbook, this paper applies nonparametric methods including smoothing splines, the Chow breakpoint test, the Bootstrap method, and random forests to model and analyze the aging trend, the relationship between education structure and the number of first marriages, and the drivers of the fertility. Results: The aging rate in Guangxi showed segmented acceleration; its average annual growth rate jumped from 0.01 percentage points during 2005~2014 to 0.44 percentage points during 2015~2024, and 2016 was confirmed by the Chow test as a significant structural turning point. The smoothing spline achieved a goodness of fit (R² = 0.958) significantly better than the linear model (R² = 0.700). The proportion of the population with an associate degree or above and the number of first marriages exhibited an inverted U-shaped relationship, with an inflection point of 5.82% (95% confidence interval: 5.33%~6.49%). The aging rate, the number of first marriages, and the proportion of higher education were the top three factors affecting the fertility, with relative importance of 100%, 98.9%, and 96.0%, respectively; the out-of-bag R² of the random forest was 0.889. Conclusions: The evolution of Guangxi’s population structure has significant nonlinear characteristics, and nonparametric methods can more accurately capture segmented changes. The low fertility rate is the joint result of population aging, the shrinking marriage market, and rising educational attainment, which can provide quantitative support for population monitoring and early warning and for optimizing marriage and fertility policies.