A Research On Spatio-temporal Model Of Interprovincial Migration Flows In China

The paper relies on the NSFC "Spatial Origin-Destination flow data mining and patterns discovery"(project number:41271388) to research spatial and temporal autocorrelation effects and spatio-temporal model of interprovincial migration flows in China.The migration research is of great practical significance, and so far a lot of theoretical, methodological and empirical achievements have been achieved. However, most of them use areas as study units, and the direct driving forces of society, economy, culture and politics are considered and the abstract factors of space and time are often ignored. The real migration processes take on some clustering patterns in space and unfold in time. This paper considers migration flows as the research subject, focuses on the effect of spatial autocorrelation and temporal autocorrelation, and builds migration flows models to capture this mechanism in the interprovincial migration flows in China.In terms of spatial autocorrelation, the paper analyzes this effect from global and local perspective respectively. Moran’s I index is applied to evaluate the global spatial autocorrelation, while G statistics is used to evaluate the local spatial autocorrelation. Specifically, the global Moran’s I indices of five periods of Chinese interprovincial migration flows in1985-1990,1990-1995,1995-2000,2000-2005and2005-2010are calculated, and we find the effect of global spatial autocorrelation in interprovincial migration has become much more stronger, outmigration and in-migration all have high global Moran’s I value in2005-2010. Then the paper use G statistics to analyze the local spatial autocorrelation pattern of in-migration flows in Guangdong, Zhejiang and Shanghai, and outmigration flows in Sichuan, Anhui and Hunan, which are representative destinations and origins. In the temporal autocorrelation aspect, the paper uses the auto-covariance and autocorrelation coefficient to analyze migration flows data from1985to2010, and we find significant temporal autocorrelation effect exists in the Chinese interprovincial migration flows. Then the paper constructs the different temporal lag migration model from first order to fifth order, and finds the second order temporal lag is the best by comparison. In order to express these two effects, the pure spatial OD(Origin-Destination) model and the spatio-temproal OD model of the Chinese interprovincial migration flows are constructed and estimated respectively. By comparison, we find the spatio-temporal OD model is more appropriate according to AIC criterion. Consequently, the spatial and temporal effects can’t be ignored in analyzing the Chinese interprovincial migration flows. If only the spatial effect is considered, it will lead to overestimation of the other explanatory variables. Finally, the paper uses the spatio-temporal OD model to predict the Chinese interprovincial migration flows in2010-2015, and some policy suggestions are presented based on the forecast result.