A spatial econometric modeling of online social interactions using microblogs

Abstract

With the advent of Information and Communication technology (ICT) in modern age, the statement of “death of distance” has received numerous discussions. This article contributes a new empirical study to the debate of “death of distance” by considering the effect of spatial autocorrelation in the estimation of distance decay effect with the incorporation of network autocorrelation in spatial econometric modeling. This work is based on a city-level dataset from China's largest social networking site called Weibo. The findings are shown as following. First, the coefficient value of network autocorrelation term (0.007, significant at 0.01 level) suggests that the city-level online social links are spatially dependent. In other words, these social connections are not randomly distributed across space but tend to form spatial clusters where neighboring links are more similar. Second, controlling spatial autocorrelation in the data, a distance decay effect on the formation of online social links is unveiled with a much smaller scaling exponent of the distances (i.e., 0.276) as compared to those (e.g., 2.0, 1.8, 1.45, 1.06, 1.03, 0.4, and 0.5) in existing studies. This research provides a useful modeling framework to analyze the real-world driving forces that characterize the patterns of social interactions in virtual space and thus advance our understanding in the connection of virtual and real spaces.