Rank-size distribution of teletraffic and customers over a wide area network

Zipf's law is used to model the uneveness of customers and telecommunications traffic distribution among the geographical areas covered by a nationwide network. Three datasets, relative to the traffic intensity in the telephone network and to the number of Internet customers as recorded on an Italian network, are used to assess the validity of the law. The estimated Zipf parameter is lower than 1 (typical of town population) in all three cases but larger than those observed for population over the same administrative groupings (code areas and provinces). Therefore the largest code areas and provinces account for a larger share of the overall traffic than what their population share would suggest, but far lower than that expected from the popularised 80/20 Pareto rule. Copyright (C) 2006 AEIT.