中国人口流动重力模型回归系数的定量估算及其空间异质性研究

Quantitative Estimation and Spatial Heterogeneity of the Regression Coefficient of the Population Flow Gravity Model in China

基于中国337个地级及以上行政区划间人口省际流动数据,对人口流动重力模型代理变量的回归系数进行实证推算。结论包括：（1）在地级行政单元研究精度下,中国人口流动重力模型代理变量的回归系数呈现出明显的空间异质性,及难以预期的作用效应,因此研究精度的提升将加剧空间交互模型构建的复杂程度,同时假定重力模型回归系数的空间均质性将导致模拟结果空间差异趋势的收敛;（2）中国人口省际流动距离衰减系数为1.975,337个地级及以上行政区划距离衰减系数值域为0.434-4.438。研究所展示的各省该系数值域及直辖市、省会城市结果为今后模型应用提供了基础参考数据;（3）地级行政单元下的距离衰减系数存在空间异质性与梯度特征,高值区主要分布在东北沿疆、新疆和珠三角地区,低值区主要位于中国中部和西南地区。结合以往研究成果比较,中国人口省际流动的空间衰减特征与铁路客运网络的可达性格局存在一定程度的一致性。

The spatial distribution of natural factors and social economic factors is not balanced, and the spatial interaction is the basic theory of geography, which has the characteristics of mutual strength and attenuation.Gravity model is widely used in the simulation of the relationship between geographic factors. The parameters of the gravity model are the important representation of the general characteristics of the specific geographic system. The parameter calibration is the precondition of the reasonable application of the model and the key step of the suitability of the model. Different modeling and estimation methods will lead to different parameter estimation and goodness of fit. However, a large number of studies at home and abroad have been based on the assumption that the model is too idealistic and simple, or it can be done in the opaque process. The calibration of the regression coefficients of the model requires a large amount of data to be used for empirical estimation,but the constraints of the relational data are led to the fact that most of the studies currently use the existing coefficient calibration results and assume their spatial homogeneity. Based on the relational data of the inter-provincial flow of 337 prefectures and above administrative divisions in China, the composite quality parameters of the 4 economic indicators of regional GDP, public finance income, total investment in fixed assets and total social consumption. The regression coefficients of the gravity model of the population in China are estimated,and the spatial heterogeneity of the coefficients is analyzed. The conclusion includes： 1） In the prefecture level administrative unit of precision, the regression coefficient Chinese population gravity model agent variables showing a spatial obvious heterogeneity, effects and unpredictable, so the study of accuracy improvement will increase the complexity of constructing the spatial interaction model, spatial heterogeneity also assumes that the gravity model of regression coefficient will lead to the convergence of simulation results spatial differences in trends; 2） The distance attenuation coefficient China interprovincial flow attenuation coefficient range 1.975337 level and above administrative distance is 0.434-4.438. The value range of the provinces and the municipalities directly under the central government and the provincial capital cities have provided the basic reference data for the future model application; 3）There are spatial heterogeneity and gradient characteristics in the distance attenuation coefficient under the administrative unit of the prefecture level. The high value area is mainly distributed in the northeast along the Yangtze River, Xinjiang and the Pearl River Delta. The low value area is mainly located in the central and southwest China. Compared with previous research results, the spatial attenuation characteristics of inter provincial mobility of Chinese population are in a certain degree consistent with the achievable character of railway passenger network.