Modeling and Evaluating Hierarchical Network: An Application to Personnel Flow Network

Abstract

This study evaluates (1) the properties of a hierarchical network of personnel flow in a large and multilayered Chinese bureaucracy, in light of selected classical network models, and (2) the robustness of the hierarchical network with regard to the edge weights as strength of “weak ties” that hold different offices together. We compare the observed hierarchical network with the random graph model, the scale-free model, the small-world model, and the hierarchical random graph model. The empirical hierarchical network shows a higher level of local clustering (in both LCC and GCC) and a lower level of fluidity of flow (i.e., high APL) across offices in the network, as compared with the small-world model and the hierarchical random graph model. We also find that the personnel flow network is vulnerable to the removal of “weak ties” that hold together a large number of offices on an occasional rather than regular basis. The personnel flow network tends to dissolve into locally insulated components rather than to maintain an integrated hierarchy.