Spatial analysis of inter-provincial migration flows in China based on spatial OD models

• Published in: 2015 23rd International Conference on Geoinformatics pp. 1 - 5
• Main Authors: Pu, Yingxia, Wang, Yaping, Han, Honglin, Liu, Dawei, Song, Xuetao, Feng, Yuanyuan
• Format: Conference Proceeding Journal Article
• Language: English
• Published: IEEE 01.06.2015
• Subjects: Analytical models; Autocorrelation; China; Computational modeling; Conferences; Construction; Gravitation; Gravity; Migration; network autocorrelation; Networks; Origins; population migration flows; spatial mechanism analysis; spatial OD model
• ISSN: 2161-024X
• DOI: 10.1109/GEOINFORMATICS.2015.7378568

Based on the network autocorrelation concept and related modeling methods, this paper analyzes the existence of network autocorrelation effects and builds several spatial OD models to explore the spatial mechanisms among Chinese inter-provincial migration flows using the sixth national census data. Several results are as follows: (1) there is a significant network dependence relationship among inter-provincial migration flows in China during the period of 2005-2010. The estimates of Moran's I with three different network structures (W o , W d and W w ) are 0.5025, 0.2867 and 0.1078. It shows that the origins of migration flows arrived at the same destination (D) are spatially agglomerated, and the destinations of migration flows from the same origin (O) are spatially clustered. In addition, the flows from spatially neighbored origins to spatially neighbored destinations are mutually dependent; (2) the goodness-of-fit of all spatial OD models (model 2 to 5) is much better than traditional gravity model. There are significant network autocorrelation effects among migration flows. In comparison with several forms of spatial OD models, it is confirmed that the network structure effect of the origins is much stronger than the other two situations. In fact, the origins in the central part of China are more concentrated across the space; (3) when simultaneously considering three different network structures, the variables' influences in model 5 are in line with the expectations, but decrease to some different degrees compared with traditional gravity model. Therefore, taking full account of network autocorrelation effects will comprehensively reveal the spatial mechanism of migration flows.